The bizarre pharmacology of the ATP release channel pannexin1

Abstract

Pannexins were originally thought to represent a second and redundant family of gap junction proteins in addition to the well characterized coonexins. However, it is now evident that pannexins function as unapposed membrane channels and the major role of Panx1 is that of an ATP release channel. Despite the contrasting functional roles, connexins, innexins and pannexins share pharmacological properties. Most gap junction blockers also attenuate the function of Panx1, including carbenoxolone, mefloquine and flufenamic acid. However, in contrast to connexin based gap junction channels, Panx1 channel activity can be attenuated by several groups of drugs hitherto considered very specific for other proteins. The drugs affecting Panx1 channels include several transport inhibitors, chloride channel blockers, mitochondrial inhibitors, P2X7 receptor ligands, inflammasome inhibitors and malaria drugs. These observations indicate that Panx1 may play an extended role in a wider spectrum of physiological functions. Alternatively, Panx1 may share structural domains with other proteins, not readily revealed by sequence alignments.

Introduction

Pannexins were discovered in a data base search on the basis of their limited sequence homology to the invertebrate gap junction proteins, the innexins (Panchin et al., 2000). While there are three pannexins (Panx1, Panx2 and Panx3) in most vertebrate genomes, thorough functional studies presently are only available for Panx1. The structural similarity with the gap junction channel-forming innexins led to the initial misconception that pannexins have a similar gap junction function. However, no in vivo evidence for such a function exists. Instead, the expression of Panx1 in single cells such as erythrocytes (Locovei et al., 2006a) and exclusive expression in the apical membrane of polarized cells such as airway epithelial cells (Ransford et al., 2009) rules out a gap junction function of Panx1 at least in these cells. The list of evidence against an in vivo gap junction function of pannexins is growing. For example, recently it has been shown that in kidney epithelial cells the Panx1 expression occurs exclusively on the apical membrane while the gap junction bearing basolateral membrane of these cells is devoid of the Panx1 protein (Hanner et al., 2012). Similarly, Panx1 in the heart is localized in the nonjunctional membrane and does not give rise to punctate staining (Dolmatova et al., 2012).

Overexpression of Panx1 yielded conflicting results. Despite an abundance of Panx1 protein in the plasma membrane, no gap junction function was found in some studies (Boassa et al., 2007; Huang et al., 2007; Penuela et al., 2007), while in others dye flux from cell to cell or intercellular current spread as a consequence of Panx1 overexpression was reported (Lai et al., 2007; Vanden Abeele et al., 2006). There are several possible explanations for the conflicting results without invoking physiological pannexin-based gap junctions. First, the measurements used in these studies were crude and, moreover, could not distinguish between connexin and pannexin channels. It is possible that the overexpression of Panx1 resulted in an upregulation of a connexin in these cells and consequently the formation of connexin gap junction channels. Therefore, to show that pannexins form gap junction channels, one needs to demonstrate both the absence of any connexin and the presence of a gap junction channel with properties distinct from any connexin based gap junction channel. Second, the pannexins are glycoproteins and as such are prevented from gap junction formation (Boassa et al., 2007; Dahl et al., 1994; Penuela et al., 2007). It is possible that the notoriously impaired glycosylation in the oocyte expression system is the basis of artifactual gap junction formation by Panx1 in paired oocytes (Bruzzone et al., 2003). Indeed, enzymatic removal of the carbohydrate moieties facilitated that process (Boassa et al., 2008). Nevertheless, the rate of Panx1 gap junction channel formation lagged that observed with connexins by orders of magnitude. These data, together with the failure of a glycosylation deficient mutant to effectively induce gap junction channels in paired oocytes, indicate that the large majority of Panx1 in the plasma membrane of the oocytes is in the glycosylated form (Boassa et al., 2008).

Western blots reveal that a variable portion of Panx1 protein is present in the unglycosylated form in different tissues. It therefore could be argued that in certain tissues the unglycosylated Panx1 could be used for gap junction channel formation. However, such gap junction channel formation is unlikely, because it has been demonstrated in independent studies that the unglycosylated Panx1 is impaired in trafficking and consequently is in undetectable (Boassa et al., 2007; Penuela et al., 2007), or minute (Penuela et al., 2009) quantities found in the plasma membrane.

The clarification of the question of gap junction channel formation by pannexins unfortunately is obfuscated by approaches with inadequate techniques. Such attempts (Tang et al., 2008) are particularly problematic when the data indicate the absence of gap junction formation by pannexins while the title of the paper and its discussion claim the opposite. For example, the authors state in the Results section of their paper:”…indicating that Panx1-specific immunolabeling was uniformly distributed in cell bodies of the neurons.”. Indeed, the low magnification micrographs do not hint at any punctate staining typical for gap junctions. Nevertheless the Conclusion section of the paper states: “We presented Western blot, reverse transcription PCR and immunolabeling data to show the expression of Panx1 and Panx2 in the neuronal cells in the cochlea. Panx1 was also found in the cochlear supporting cells. These results suggested a new family of gap junction proteins suitable for assembling into gap junctions and hemichannels in the intercellular transport pathways in the cochlea.”

Additional arguments against a gap junction channel function of Panx1 can be found in Dahl and Locovei (2006), Sosinsky et al. (2011) and Dahl and Keane (2012). Realizing the absence of unequivocal evidence for the ability of pannexins to form gap junction channels, a large number of pannexin researchers, therefore, advocated “that in the absence of firm evidence that pannexins form gap junctions, the use of the term “hemichannel” be discontinued within the pannexin literature.” (Sosinsky et al., 2011).

In contrast, the role of Panx1 channels residing in unapposed cell membranes is well established (Bao et al., 2004; Bruzzone et al., 2003; Dahl and Locovei, 2006; Scemes, 2011; Sosinsky et al., 2011). Specifically, the role of Panx1 channels as the major ATP release channel in a great number of cell types is supported by several lines of independent evidence. Over the last eight years a dozen arguments have been presented to that effect (Dahl and Keane, 2012; Dahl and Locovei, 2006; Scemes, 2011). The list of evidence includes high permeability of Panx1 channels to ATP, a correlation of Panx1 expression with ATP release sites at the cellular and subcellular level, and matching pharmacologies of channel mediated ATP release and of the Panx1 channel. Considering the poorly selective properties of Panx1 channels it is plausible, that in addition to ATP, Panx1 channels also can mediate the release of other compounds in the size range of ATP. Indeed, it has been reported that Jurkat cells release UTP in addition to ATP (Chekeni et al., 2010). Furthermore, it has been shown that erythrocytes release epoxyeicosatrienoic acids via Panx1 channels (Jiang et al., 2007). The flux of other compounds remains to be determined.

Panx1 has been recognized as a major factor in the immune system, where it is involved in the innate immune response and associated cell death (Locovei et al., 2007b; Pelegrin and Surprenant, 2006a; Silverman et al., 2009). Bacterial products like muramyl dipeptide induce IL-1β release, which is diminished by Panx1 inhibitors (Marina-Garcia et al., 2008), T-cell activation and associated inflammatory responses involve Panx1 activity (Schenk et al., 2008), and the neutrophil activation involves the ability of Panx1 to mediate release of ATP (Chen et al., 2010).

A series of disease states involve Panx1 channel activity, including Crohn’s disease (Gulbransen et al., 2012), AIDS (Seror et al., 2011), epilepsy (Santiago MF, 2011) and neuronal cell death due to oxygen deprivation (Thompson et al., 2006). Panx1 channels also may play a role in secondary cell death associated with CNS injury (Dahl and Keane, 2012; Silverman et al., 2009). Thus, Panx1 deserves much interest as a drug target for therapeutic intervention in several major diseases.

Gap junction blockers acting on Panx1

Pharmacological inhibitors of gap junction channel activity were originally identified in studies on invertebrates (Johnston et al., 1980). Despite the lack of sequence homology between the structural proteins, gap junctions formed by either innexins or connexins are equally affected by the inhibitors of junctional communication. This means that structural motifs not easily recognized in the primary sequence are shared between innexins and connexins. Thus it is not surprising that most gap junction inhibitors also affect channels formed by pannexins, which do share sequence homology with the innexins.

The list of gap junction blockers includes a diverse set of agents including alcohols, flufenamic acid, glycyrrethinic acid, mefloquine and other quinines. The longer chain alcohols, heptanol and octanol, were first found to reduce junctional conductance in the crayfish septate axon (Johnston et al., 1980). Flufenamic acid (FFA) and other fenamates were first described by (Harks et al., 2001) to inhibit Cx43 gap junction channels. Glycyrrhetinic acid (GA) and the related compound carbenoxolone (CBX) were first reported by (Davidson et al., 1986) to act as reversible inhibitors on gap junction channels in cultured human fibroblasts. Subsequently, CBX was shown to be a nearly universal inhibitor of junctional communication. Quinine and mefloquine were found to induce graded responses for gap junction channels formed by different connexins (Cruikshank et al., 2004).

None of these agents acts exclusively on gap junction channels. Instead they target not only other channels, but also transporters and non-membrane proteins. It thus is not very meaningful that these compounds also affect pannexin channels. The inhibition of Panx1-mediated membrane currents by CBX and FFA was first described by (Bruzzone et al., 2005). These authors noted a distinct quantitative differences for the effects of these drugs between connexins and pannexins. FFA was more potent in inhibiting connexin channels while CBX required considerably lower concentrations to inhibit Panx1 channels than were required to close connexin channels. Figure 1 and table 1 show the effects of alpha- and beta- glycyrrhetinic acid (GA) and of flufenamic acid (FFA) on Panx1 channel currents in Xenopus oocytes. As reported earlier (Bruzzone et al., 2005), β GA at 50 μM concentration inhibited Panx1 currents by approximately 50 %. For a similar level of current inhibition 500 μM of FFA was required. Contrary to the conclusion by (Chekeni et al., 2010) based on indirect measurements that α-GA at even higher concentrations did not inhibit Panx1 in Jurkat cells, this stereoisomer was as potent as β-GA in inhibiting currents in Xenopus oocytes expressing full length Panx1. Chekeni et al. argued that Panx1 channel activity could only be observed in their assay system after Panx1 was cleaved by caspase 3, which is in contrast to the reversible Panx1 channel activation in all other cells where Panx1 mediates ATP release (Bao et al., 2004; Locovei et al., 2006a; Ransford et al., 2009; Suadicani et al., 2012). Since the truncation of Panx1 could eliminate an essential component of the drug mediated inactivation mechanism, we tested a Panx1 deletion mutant mimicking the protease cleaved form. As Figure 1 reveals, the truncated Panx1 (Panx1 378stop) still was inhibited by both isomers of GA and by FFA, indicating that the inactivation mechanism was operative in the absence of the carboxyterminal amino acid sequence cleaved by caspase 3. Table 1 shows a summary of the effects of these drugs.

Figure 1.

Effect of α- and β-GA on Panx1 channel currents in Xenopus oocytes. The membrane potential was held at −50 mV and stepped to +50 mV at a rate of 0.1 Hz to elicit Panx1 currents. Both drugs reversibly inhibited currents in oocytes expressing either wtPanx1 or the truncated mutant Panx1, 378stop.

Table 1.

Effects of alpha and beta GA and of FFA on Panx1 currents.

drug	Wt Panx1	Panx1 378stop
18 α-GA (50 μM)	56.3 ± 3.2 (3)	69.7 ± 4.5 (3)
18 β-GA (50 μM)	55.7 ± 3.2 (3)	67.3 ± 8.1 (3)
FFA (500 μM)	52.0 ± 3.1 (4)	52.7 ± 11.2 (3)

% inhibition of Panx1 currents in oocytes are given as means ± SD. The number of oocytes assayed is given in parentheses.

It is clear that α-GA and β-GA cannot be used to discriminate between effects mediated by Panx1 or connexins. Similary, in the absence of a systematic analysis of the effects of CBX and of FFA on all connexin isoforms, even their different quantitative effects on some connexins and Panx1 cannot be generalized. An example for such a situation is encountered with mefloquine. This drug has differential effects on different connexins, with Cx36 (IC50 ~300 nM) being considerably more sensitive to the drug than Cx 50 (IC50~1.1 μM), while connexins 43, 32 and 26 required 10 to 100 times higher concentrations to observe inhibitory effects on junctional currents. Mefloquine has been observed to inhibit Panx1 mediated currents with an IC50 of ~50 nM (Iglesias et al., 2009b).

Moreover, all known gap junction drugs are acting on several additional targets, making their usefulness for functional assignments rather questionable.

“Mimetic” peptides inhibit Panx1 channels regardless of what they mimic

Because of the notorious lack of specificity of gap junction channel blockers many laboratories have searched for more precise agents. Gap junction channels form by docking of preassembled connexin hexamers (aka connexons or hemichannels) localized in apposing cell membranes to each other (Dahl et al., 1992; Musil and Goodenough, 1993). This docking step mediated by the extracellular loop domains of connexins initiates the opening of the formerly closed connexons to form the complete gap junction channel that establishes the permeation pathway between cells (Bukauskas and Weingart, 1993; Dahl et al., 1992). The whimsical idea that peptides with the sequence of the connexin extracellular domains could mimic the docking event and consequently open connexons was not confirmed when experimentally tested (Dahl et al., 1994). However, the extracellular loop peptides appear to bind to connexons since they do inhibit the formation of gap junction channels, probably by interfering with docking but without activating the docking gate (Dahl et al., 1994; Dahl et al., 1992; Warner et al., 1995). The use of the peptides for interference with gap junction function is limited as they inhibit formation of new channels but do not affect existing gap junctions. Thus, the action of the peptides depends on the turnover rate and consequently is slow.

The gap junction mimetic peptides experienced a resurgence when it was found that they affected channel mediated ATP release from cells (Braet et al., 2003). Together with the observation of similar effects by such gap junction blockers as CBX, this led to the conclusion that Cx43 and other connexins can mediate ATP release in the form of open “hemichannels.” This notion has gotten wider acceptance than the data support. In particular, ATP is released from cells that do not express connexins, such as erythrocytes (Locovei et al., 2006a). Similarly, ATP is released in some cells in a polar fashion exclusively at the apical membrane while connexin expression is exclusively at the basolateral membrane, such as in the airway epithelia (Ransford et al., 2009; Wiszniewski et al., 2007).

Since Panx1 expression is more consistent with a general ATP release function, connexin mimetic peptides were tested for interference with Panx1 channel activity. Despite the claim of high specificity of the mimetic peptides for their connexin target, they were observed to significantly inhibit Panx1 channels (Wang et al., 2007). The likely mechanism is a partitioning effect of the peptides in the channel. Peptides had to be at least a certain size and polyethyleneglycols of similar size exhibited the same effect (Dahl, 2007; Wang et al., 2007).

Although Pannexin mimetic peptides do inhibit Panx1 channels, they appear to suffer from the same limitations as the connexin mimetic peptides. Yet because of the large bore size of the Panx1 channel, which is not shared by many channels, the mimetic peptides are specific by way of their cut-off size.

In studies using mimetic peptides, specificity of the peptide effect is often documented by having a peptide with scrambled sequence serve as control. However, single “scrambled” peptides are appropriate controls only under the condition that specific binding of the peptide to the target protein is the operative mechanism. If, as suggested, the inhibitory effect is a steric interference with channel function, a scrambled peptide may or may not be folded to the same dimensions as the “specific” peptide and thus could fail to exert the same effect.

It appears that the mimetic peptides that are “active” for parameters such as ATP release were identified by a selection process rather than through rational design. A prime example is Gap24, which has the sequence of the cytoplasmic loop of the Cx32 protein. Because Gap26 and Gap27 have the sequence of portions of the extracellular loops of the protein, they at least have the potential to mimic the docking process that occurs during gap junction formation. It is likely that Gap24 was designed as a control peptide, because an intracellular sequence was thought to be a suitable control for an extracellular sequence, but when Gap24 was found to be as effective as the rationally designed peptides the rationale was changed (De Vuyst et al., 2006).

Inhibition of Panx1 channels by transport inhibitors

The properties of Panx1 channels suggested a main function of Panx1 as an ATP release channel (Bao et al., 2004). To test this hypothesis, the possibility of a correlation between ATP release and Panx1 channel activity was probed. One of the tests involved determination of whether the pharmacological profile of the Panx1 channel matched that of ATP release. With the exception of the drugs known to inhibit vesicular release most drugs interfering with channel mediated ATP release also attenuated Panx1 channel currents. One glaring exception was dipyridamole, which reduced ATP release from erythrocytes but did not interfere with Panx1 channels suggesting the existence of an additional, non-vesicular ATP release pathway (Qiu et al., 2011).

Among the drugs attenuating ATP release are the transport inhibitors probenecid and NPPB (Ballerini et al., 2002; Darby et al., 2003; Mitchell et al., 1998; Reigada et al., 2006; Sabirov et al., 2001). This has led to the conclusion that ATP release can also be mediated by transport processes in addition to vesicular and channel-mediated release. However, both drugs have been shown to inhibit Panx1 channel currents, ATP release from Panx1 expressing oocytes and dye uptake by erythrocytes (Silverman et al., 2008). Furthermore, additional transport inhibitors, such as sulfinpyrazone, benzbromarone and estrone sulfate inhibited Panx1 channels (Figures 2 and 3).

Figure 2.

Dose-response curves for inhibition of Panx1 channel currents in Xenopus oocytes by various drugs. Oocytes expressing Panx1 were voltage clamped and the membrane potential was held at −50 mV. To open Panx1 channels repetitive brief voltage steps to + 50 mV were applied. The drugs were applied to the bath by perfusion.

Figure 3.

Inhibition of Panx1 channel currents by transport inhibitors. All drugs were applied at a concentration of 500 μM except dipyridamole (200 μM). The experimental procedure was the same as described in legend to figure 2.

This remarkable overlap of drug sensitivity between Panx1 channels and various transporters including those of the URAT family has led us to speculate that Panx1 actually could be a component of a transport protein complex, where Panx1 represents the permeation pathway and the cloned transport proteins impart directionality and specificity to the transport process (Silverman et al., 2008). However, this has not been tested.

Chloride channel blockers

Some of the transport inhibitors such as NPPB affect transport by inhibiting the chloride channel leg of the transport process. It is therefore of interest whether other chloride channel blockers also affect Panx1 channels. Indeed, several of them including DIDS, SITS and IAA94, were observed to inhibit Panx1 currents in the micromolar concentration range (Ma et al., 2009b). This pharmacological property and permeability measurements on a low conductance channel led to the conclusion that Panx1 forms an anion selective channel (Ma et al., 2012). Several observations about Panx1 channels, however, are not consistent with such a conclusion because they unequivocally demonstrate a cation permeability for the channels. 1. When a Panx1 channel in an excised membrane patch is exposed to a K⁺/ATP gradient, the reversal potential lies between the reversal potential for the two ions (actually closer to the K⁺ reversal potential). This indicates a substantial K⁺ permeability together with an ATP permeability of the Panx1 channel. 2. Retinal ganglion cells show CBX and probenecid sensitive cation currents elicited by membrane swelling and thus are likely to be mediated by the mechanosensitive Panx1 present in these neurons (Xia et al., 2012). 3. A large conductance cation channel in cardiac myocytes exhibits key properties of Panx1 channels including single channel conductance and sensitivity to CBX and probenecid. This cation channel activity also correlates with Panx1 expression (Kienitz et al., 2011). 4. ATP release from many cell types is correlated with an uptake of extracellular tracer molecules. This correlation is so good that dye uptake is an accepted surrogate measure for ATP release. The correlation holds true both for negatively charged and positively charged dyes. For example, ATP release from erythrocytes is correlated with the uptake of the negatively charged carboxyfluorescein (Locovei et al., 2006a) and the positively charged YoPro. That the dye uptake is mediated by Panx1 channels is indicated by its absence in Panx1^−/− eythrocytes and by the pharmacology of the dye uptake. YoPro flux in erythrocytes is inhibited by, for example, probenecid, BBG, ATP and BzATP (Qiu and Dahl, 2009; Qiu et al., 2011; Silverman et al., 2008). The effectiveness of the last two compounds rules out P2X7R as permeation pathway for the dye.

Like Panx1 channels, connexin channels are also inhibited by NPPB. However, no effect of DIDS, SITS or IAA-94 on Cx46 and Cx50 hemichannels was observed (Eskandari et al., 2002; Silverman et al., 2008).

Mitochondrial ADP/ATP translocation inhibitors

One of the key functions of mitochondria is to provide ATP to the cytoplasm of cells. Synthesis of ATP takes place in the mitochondrial matrix and ATP gets access to the host cells’ cytoplasm. Since mitochondria are wrapped by two membranes, ATP has to traverse both. The inner mitochondrial membrane has an ADP/ATP translocase facilitating the directional flux of ATP towards the cytoplasm. How ATP permeates the outer mitochondrial membrane is not clear. Because the mitochondrial permeability allows for high rate of ATP permeation one may wonder whether it shares characteristics with Panx1 channels. Two compounds, bongkretic acid and atractyloside, are well known inhibitors of the mitochondrial ATP efflux. As shown in Figure 4, both drugs inhibited Panx1 currents in oocytes. The IC50 was 7 μM for bonkrekic acid and 250 μM for atractyloside. For comparison, the IC50 for inhibition of the mitochondrial pore by bongkrekic acid is 20 nM (Stubbs, 1979). The difference in IC50 between Panx1 and the mitochondrial pore could be interpreted that the pore has to be distinct from Panx1. However, as discussed later, the affinity of Panx1 to drugs can be altered by other proteins. Since the IC50 for the mitochondrial inhibitors on Panx1 were determined in isolation, i.e. without a putative mitochondrial partner protein, the difference in IC50 should not be used to a priory rule out a Panx1 contribution to the mitochondrial pore.

Figure 4.

Effects of bongkrekic acid and atractyloside on Panx1 functions. a) atractyloside induces reversible inhibition of Panx1 channel currents in a dose dependent fashion. b) uptake of Ethidium bromide by (nucleated) frog erythrocytes induced by high extracellular K⁺ (60 mM). Inhibition of dye uptake by 4 μM bongkrekic acid (c) and 50 μM atractyloside (d). Dose response curves for inhibition of Panx1 channel currents in Xenopus oocytes by bongkrekic acid (e) and atractyloside (f) (n=5 for all points). The experimental procedure was the same as described in legend to figure 2.

Bongkrekic acid and atractyloside also inhibited the uptake of dyes by erythrocytes, which is known to be mediated by Panx1 channels. The added importance of this observation is that erythrocytes do not contain mitochondria, so any mitochondrial contribution to this drug effect can be ruled out. Thus both drugs must act directly on the Panx1 channel. It is conceivable that Panx1 is part of the ATP translocation mechanism in mitochondria, perhaps as a continuation of the ATP translocation pathway through the outer membrane. Although Panx1 protein has not been reported in mitochondria, this possibility might need direct testing. Alternatively, the data could be consistent with shared structural domains between Panx1 and either the ADP/ATP translocase or a yet to be identified protein in the outer mitochondrial membrane.

Under conditions of oxidative stress or increased cytoplasmic calcium concentration mitochondria open what is called the “mitochondrial permeability transition pore” (mPTP) eventually resulting in cell death. The mPTP is formed by a protein complex that spans both the inner and the outer mitochondrial membranes and somehow enables the transit of cytochrome c to the cytoplasm. The composition of this protein complex is presently unresolved. Until recently, (see (Abou-Sleiman et al., 2006; Crompton, 1999; Halestrap et al., 2002) for reviews) the main components of the mPTP were considered to be represented by the ADP/ATP translocase and VDAC. However, this view has been challenged because ablation of these proteins does not prevent pore formation (Gouriou et al., 2011; Vaseva et al., 2012; Zorov et al., 2009). The mPTP is inhibited by bonkrekic acid, which in the past has been attributed to the effect on the nucleotide translocase. However, with Panx1 as an alternative target of the drug, a contribution of Panx1 to the mPTP should be considered.

The mPTP has been shown to relate to neuronal cell death and also play a role in reperfusion injury in the heart (Das and Steenbergen, 2012). Bongkrekic acid at 16 microg/kg has been documented to be neuroprotective in ischemic conditions (Muranyi and Li, 2005).

ATP and other P2X7 receptor (P2X7R) ligands

Panx1 channels coexpressed with purinergic receptors can be activated by extracellular ATP (Locovei et al., 2007b; Locovei et al., 2006b). Because of positive feedback when ATP is released from the cell, this allows for amplification of the ATP signal. ATP binding to the purinergic receptor, which can be metabotropic of the P2Y type or ionotropic of the P2X type, opens the Panx1 channel and thereby induces further ATP release into the extracellular space. Without protective measures, activation of this feedback mechanism would quickly exhaust the cellular ATP pool and eventually result in cell death. The countermeasure has been identified in form of self-inhibition of the Panx1 channel (Qiu and Dahl, 2009). Extracellular ATP inhibits Panx1 channel currents and dye uptake and attenuates ATP release.

The inhibition of Panx1 channels is not restricted to ATP but is also observed with BzATP, which is more potent, and a series of other P2X7R ligands. Surprisingly, regardless of whether the ligands act on P2X7R as agonist or antagonist, the effect on Panx1 is inhibitory. The ligands tested included ATP, BzATP, suramin, KN62, BBG and the “specific” P2X7R antagonist A438079. The concentrations of these drugs required for inhibition of Panx1 channel currents all exceed that required for the effects on the P2X7R, suggesting a similar yet distinct binding sites on the two proteins. The binding site on Panx1 has been characterized and indeed exhibits similarities with that on P2X7R (Qiu et al., 2012). The difference in affinities of the binding sites allows for both positive and negative feedback control of the Panx1 channel by ATP (Figure 5). At low ATP concentrations, the purinergic receptors activate Panx1 resulting in amplified ATP release. As the ATP concentration builds up in the vicinity or within the vestibulum of the Panx1 channel, the self inhibition will limit further ATP release.

Figure 5.

ATP feedback loops controlling Panx1 channels. ATP from an external source activates the P2X7 receptor (P2X7R), which in turn activates by a presently unknown mechanism Panx1. The open Panx1 channel allows the efflux of ATP following its concentration gradient. ATP binds to Panx1 at the extracellular surface with lower affinity than to the P2X7R, inhibiting further ATP release through the Panx1 channel.

Inflammasome inhibitors, including glibenclamide, affect Panx1

The inflammasome, a multiprotein complex comprising NLRP, ASC and XIAP (for definitions see footnote), mediates the innate inflammatory response and can be activated in several ways, including by extracellular ATP through P2X7 receptors. It is well established that ATP plays a key role in the activation of the inflammasome (Latz, 2010). However, it also has become evident, that in nervous tissues additional factors aid or even can replace ATP for inflammasome activation. For example, in neurons and astrocytes elevation of extracellular K⁺ activates the inflammasome as indicated by cleavage of caspase 1 and by release of interleukin 1 β (IL-1β) from these cells (Silverman et al., 2009). Thus increased extracellular ATP and K⁺ resulting from leakage from cells damaged by a traumatic event may cooperate to trigger inflammation in the surrounding cells and thus mediate “secondary cell death.” Additional stimuli are glutamate, activating Panx1 through NMDA receptors (Thompson et al., 2008) and oxygen deprivation, also activating Panx1 (Locovei et al., 2006a; Sridharan et al., 2010; Thompson et al., 2006).

Several independent lines of evidence indicate a crucial role for Panx1 in the innate immune response. 1. Inhibitors of Panx1 channels and knockdown of Panx1 with siRNA technology attenuate ATP-induced release of IL-1β in macrophages (Pelegrin and Surprenant, 2006b, 2009). 2. In oocytes co-expressing Panx1 and P2X7R, extracellular application of ATP induces CBX sensitive currents not seen in oocytes expressing only P2X7, suggesting that ATP acting on P2X7R indirectly opens Panx1 channels (Locovei et al., 2007a). Similarly, two component-currents with differential pharmacology are elicited in cultured astrocytes by ATP (Iglesias et al., 2009a). 3. Activation of caspase1 and of IL-1β release induced by bacterial products like muramyl dipeptide are reduced by Panx1 inhibitors (Marina-Garcia et al., 2008). 4. Caspase activation in neurons and astrocytes can be triggered by extracellular K⁺ and the inflammasome activation can be inhibited by probenecid, a Panx1 inhibitor (Silverman et al., 2009). 5. Co-immunoprecipitation studies reveal an association of Panx1 with P2X7R in HEK cells co-expressing P2X7 and Panx1 (Pelegrin and Surprenant, 2006a) and in neurons and astrocytes (Silverman et al., 2009). In neurons and astrocytes, Panx1 and P2X7 are also associated with the components of the inflammasome. 6. In oocytes co-expressing Panx1 and P2X7, extracellular ATP can induce cell death with characteristics of pyroptosis (Silverman et al., 2009). 7. T-cell activation and associated inflammatory responses involve Panx1 activity (Schenk et al., 2008). 8. neutrophil activation involves the ATP-release function of Panx1 (Chen et al., 2010).

Glibenclamide has been widely used to elucidate the mechanism of inflammasome activation (Hamon et al., 1997; Lamkanfi et al., 2009; Marty et al., 2005; Muhl et al., 2003). Glibenclamide, also known as glyburide, binds with high affinity (EC₅₀: 20 nM) to the SUR subunit of ATP-regulated potassium channels (Mikhailov et al., 2001; Sturgess et al., 1985). This effect underlies the therapeutic benefit of glibenclamide in type 2 diabetes. Recently it has been found that glibenclamide also affects innate immunity by inhibiting the inflammasome (Hamon et al., 1997; Lamkanfi et al., 2009; Marty et al., 2005; Muhl et al., 2003). This observation led to the hypothesis that inflammasome activation requires a potassium ion efflux (Gross et al., 2009; Muruve et al., 2008; Schroder et al., 2010). However, the concentrations of glibenclamide required to inhibit activation of caspase and release of cytokines from various cells is 3 orders of magnitude higher than that for inhibition of potassium channels. Thus an alternate target for the drug is likely to mediate the inhibitory effect on the inflammasome. Panx1 has been identified as such a target of glibenclamide (Qiu et al., 2011). In oocytes expressing Panx1, voltage activated currents were attenuated by glibenclamide in a dose dependent fashion. The concentration range for inhibition of Panx1 currents was very similar to that described for inhibition of the inflammasome. The IC50 was 45 μM.

Colchicine has long been used for the treatment of acute gouty arthritis. The mechanism of action of colchicine on the inflammasome is mainly ascribed to the well established effect of the drug on microtubules. For example, urate crystal delivery to the NALP 3 inflammasome is impaired by colchicine in this way (see (Cronstein and Terkeltaub, 2006; Nuki, 2008) for review). Table 2 shows that colchicine at least at high concentrations also impairs Panx1 currents. The time course of channel inactivation by colchicine is similar to that observed for drugs acting directly on the Panx1 protein. Thus for colchicine to impair Panx1 channel currents, a microtubular depolimerization by the drug may not be required.

Table 2.

Pannexin channel inhibitors

	drug	IC50	references
Gap junction blockers	Carbenoxolone	5 μM	(Bruzzone et al., 2005)
	FFA	500 μM > 50% inhib	***
	α GA	50 μM > 50% inhib	***
	β GA	50 μM > 50% inhib	***
“mimetic” peptides	Gap27	200 μM > 20% inhib	(Wang et al., 2007)
	Gap24	200 μM > 10% inhib	“
	PanxE1b	200 μM >18% inhib	“
	Panx10	200 μM > 18% inhib	“
	Gap 24 (dye uptake)	blocked	“
	10Panx1(dye uptake)	blocked	(Pelegrin and Surprenant, 2006a; Wang et al., 2007)
Transport blockers	Probenecid	150 μM	(Silverman et al., 2008)
	Benzbromarone	500 μM > 50% inhib	*
	Sulfinpyrazone	550 μM	*
	Estrone sulfate	115 μM	*
	NPPB	50 μM	(Silverman et al., 2008)
Cl− channel blockers	DIDS	9 μM	(Ma et al., 2009a)
	SITS	11 μM	“
	IAA94	95 μM	“
P2X7 receptor ligands	ATP	680 μM	(Qiu and Dahl, 2009)
	BzATP	100 μM	“
	BBG	3 μM	“
	BBG (ATP release)	0.1 μM	“
	Suramin	100 μM	“
	A438079	75 μM	“
	KN-62	50 μM > 40% inhib	**
Malaria drugs	Mefloquine	50 nM	(Iglesias et al., 2009b)
	Quinine	150 μM > 20% inhib	**
	Artesunate	450 μM	**
	Artemisinin	200 μM > 20% inhib	**
	Artemisinin (dye uptake)	0.14 μM	**
Inflammasome inhibitors	Glyburide	45 μM	(Qiu et al., 2011)
	colchicine	200 μM > 50% inhib	**
Mitochondrial AAC inhibitors	Atractyloside	250 μM	**
	Bongkrekic acid	7 μM	**
Thiol reagents	MBB	100 μM > 20% inhib	(Bunse et al., 2009; Wang and Dahl, 2010)
	MBB (ATP release)	1 mM > 80%	“
	TCEP	10 mM > 20%	(Bunse et al., 2009)

Effect of various drugs on Panx1 channel currents (or on dye uptake or ATP release as indicated). IC50 values are given if available. For some drugs percentage of current inhibition is indicated for specific drug concentration.

^*Locovei and Dahl unpublished

^**Qiu and Dahl unpublished

^***Wang and Dahl unpublished

NLR: NOD like receptor; ASC: apoptosis-associated speck-like protein containing a CARD; XIAP: X-linked inhibitor of apoptosis protein; NOD: nucleotide-binding oligomerization domain; CARD: caspase recruitment domain (of NLR, for example); SPECK: large aggregates of ASC in leukemia cells are called specks because of appearance.

Malaria drugs

As indicated earlier, the malaria drug mefloquine can inhibit Panx1 channels with an IC50 of 50 nM. Mefloquine is a derivative of the older malaria drug quinine. As shown in Table 2 quinine also inhibits Panx1 channels, but much less efficiently. This is not surprising considering the structural similarity between the two drugs. Several drugs unrelated to quinine have proven to be effective for the treatment of malaria (Price et a. 1997). These drugs are of particular interest both medically and conceptually because they are effective on multidrug resistant falciparum malaria, suggesting a different drug target in the pathogen. These drugs include artemisinin and artesunate. As shown in Figure 5 and Table 2, artesunate and artemisinin inhibited Panx1 currents. Artemisinin also inhibited the Panx1 mediated dye uptake in erythrocytes. Artesunate inhibited current with an IC50 of 450 μM and artemisinin at 200 μM inhibited current by ~20% while the IC50 for inhibition of dye uptake by artemisinin was 0.14 μM. A discrepancy between IC50ies for drugs affecting Panx1 currents and Panx1 mediated dye uptake is not uncommon and probably is due to the induction of subconductance states. Pushed into a subconductance state, the flux of dyes as the larger permeants will be more impaired (or even abolished) than the flux of the main current carrying small ions.

Since erythrocytes express Panx1 at high levels and since plasmodia residing in erythrocytes can only receive nutrients through the erythrocyte membrane it is conceivable that in vivo the effect of artesunate on erythrocyte Panx1 is a contributing factor to the therapeutic action of the drug on malaria. However, artesunate and other malaria drugs are capable of acting on plasmodium falciparum in vitro (Wongsrichanalai et al.). Thus, it is likely that plasmodia either express a protein related to the innexin/pannexin family (the long sought ancestor?) or that a plasmodium protein shares a domain with structural similarity to the equivalent domain in Panx1.

Thiol reagents

The Panx1 protein has a carboxyterminal cysteine that can be modified by thiol reagents (Wang and Dahl, 2010). Application of thiol reagents during a pulse protocol that transiently opens Panx1 channels resulted in progressive attenuation of Panx1 channel currents in cells expressing wt Panx1. Replacement of the terminal cysteine but of no other cysteine in Panx1 by serine eliminated this effect. In addition to attenuation of Panx1 channel currents, reaction of the carboxyterminal cysteine also attenuated the release of ATP (Wang and Dahl, 2010). The effect of MBB on ATP release was significantly higher than its effect on Panx1 currents (90 % versus 25% inhibition). Inhibition of ATP release by MBB was only prominent when the thiol reagent was applied to the open channel, indicating that the reagent reached the carboxyterminal cysteine through the channel pore.

These observations demonstrate that the channel can open with the carboxyterminus intact and consequently truncation of the Panx1 protein by caspase 3 cleavage cannot be an absolute prerequisite for channel opening as has been suggested based on the failure of imposed voltage steps to activate uncleaved Panx1 while the same voltage protocol opened the channels after cleavage of the protein (Chekeni et al., 2010). Furthermore, reversible use of Panx1 channels for ATP release in many cell types, including erythrocytes and airway epithelial cells, also is not compatible with a requirement for caspase cleavage of the Panx1 protein because constitutively open Panx1 channels lead to cell death (Bunse et al., 2010; Wang and Dahl, 2010).

Given the modulation of Panx1 channel currents by thiol reaction, it is plausible that channel activity is regulated by the redox state of the cell. Indeed, reducing agents such as TCEP attenuated channel currents (Bunse et al., 2009).

Mechanism of Panx1 channel inhibition

The determination of mechanism of the drug actions requires single channel analysis. The Panx1 channel has been characterized as a large conductance channel with multiple subconductance states (Bao et al., 2004). The various subconductance states differ in amplitude, making analysis cumbersome. Application of inhibitors of macroscopic Panx1 channel currents led to an initial diminution of amplitudes of single channel current events before complete loss of channel activity was observed. Several different drugs including CBX, BBG and BzATP caused these effects (Qiu and Dahl, 2009). Thus it is not clear to what extent channel block, gating to subconductance states or modification of the open probability, p_o occurred. It is quite possible that different mechanisms are mediated by the drugs but that the effects look alike because of the multiple irregularly sized subconductance states, which make it impossible to discriminate between a partial block of the full channel and its activity in a subconductance state. Considering that the binding sites for several drugs most likely differ, identical mechanism of channel closure by the various drugs are unlikely. For example, the binding site for ATP and BzATP on Panx1 channels has been mapped to moieties in the two extracellular loops of the protein close to or within the channel vestibulum (Qiu et al., 2012). Drug effects at this position could easily induce steric channel block. On the other hand, the action of the drugs affecting both gap junction channels made of connexins and membrane channels of Panx1 occurs most likely within the membrane or intracellularly, because at least in gap junction channels the extracellular loops of connexins are shielded in the docked state of connexins. Other hydrophobic drugs can also be expected to act within the transmembrane segments of Panx1. However, the precise binding sites remain to be elucidated.

Typically, the effect of the drugs are more prominent on Panx1 mediated dye uptake or ATP release than on Panx1 channel currents. Irrespective of whether the drug blocks the channel sterically or whether the drug pushes the channel in a subconductance state such a difference is to be expected. A minor constriction of the channel by either mechanism will exclude ATP or dye from passage through the channel, while the remaining channel pore will still be larger than many selective ion channels, such as calcium, sodium or potassium channels.

The monkey wrench in Pannexin pharmacology

A large number of diverse drugs affect Panx1 channels, working reproducibly and predictably in expression systems as well as in normal differentiated cells and tissues. However, an observation made in oocytes expressing Panx1 together with the Kv β3 subunit of potassium channels indicates that there may be pharmacological surprises about Panx1 in store (Bunse et al., 2009). The co-expression with Kv β3 attenuates responses of Panx1 channels to several drugs including CBX, probenecid and TCEP.

In vivo, Panx1 and Kvβ3 are co-expressed in the hippocampus and the cerebellum (Bunse et al., 2009). Whether this co-expression also modulates the pharmacology of Panx1 in these tissues has not been determined yet. Nevertheless, one has to be cautious with the interpretation of pharmacological results; lack of probenecid or CBX effects on a measured process in cells co-expressing Panx1 with Kvβ3 does not rule out a contribution of Panx1 to that function. Whether there are additional proteins with the ability to modify Panx1 pharmacology is not known.

Pannexin as drug cloaca

Table 2 summarizes the drug effects on Panx1 channels and it is clear that pannexin channels are affected by a large number of drugs that are diverse in their chemical composition. These drugs furthermore are known as “specific” inhibitors of various proteins with distinct functional roles. How is it possible that Panx1 is such a sewer for drugs?

For some drugs, such as CBX, the explanation is simple. Panx1 is just one more of the many targets for this rather unspecific agent. Glibenclamide at low concentrations (nM) remains a very specific tool for inhibition of a subtype of potassium channels. At higher, but still micromolar concentrations the effect of glibenclamide is more widespread and includes effects on CFTR, ATP binding cassette transporters and also Panx1.

For other drugs a misappropriation of target may have occurred. It is likely that this interpretation applies to the effect of glybenclamide on the inflammasome. In the past this has been ascribed to an effect on potassium channels. However, the logic of that argument is flawed because much higher glibenclamide concentrations are required for inhibition of the inflammasome than for inhibition of potassium channels. Furthermore, the speculated lowering of the cytoplasmic potassium concentration as inflammasome trigger is unlikely considering that in most tissues the volume of the extracellular space is much lower than the cytoplasmic volume. A more plausible explanation is that glibenclamide directly affects Panx1 channels, which play a key role in inflammasome activation.

For several of the drugs the effect on Panx1 cannot be easily dismissed as a nuisance side effect, so other factors have to be considered. For example, it is plausible that the various target proteins share epitopes even though these epitopes may not be evident by sequence alignments. This argument may apply to the shared effects of drugs between Panx1 and the protein targets P2X7R, connexin and innexin gap junctions and the drug target in plasmodium falcifarum.

Alternatively, Panx1 may be part of a protein complex exerting certain functions, and drug effects may have been inaccurately ascribed to other components of that complex. We have speculated that the remarkable overlap of the pharmacologies of anion transporters and Panx1 could indicate a role of Panx1 in the transport mechanism in concert with the cloned transport proteins. For example Panx1 could provide the permeation pore while the transport proteins determine specificity and directionality of the transporter complex. A similar situation may also apply to ATP efflux from mitochondria, which in addition to the ADP/ATP translocase in the inner mitochondrial membrane also may involve Panx1 for ATP release through the outer membrane. Furthermore, as long as the composition of the mPTP remains unsettled, a Panx1 contribution in this protein complex may be considered.

Conclusion

Although Panx1 channels are affected by a large number of drugs, the search for a compound that interferes only with this protein continues. Panx1 is too important a drug target to accept the existing repertoire of available drugs. Since Panx1 is involved in a series of diverse diseases including HIV infection, stroke, secondary cell death in CNS trauma, Crohn’s disease and epilepsy, it certainly deserves concerted attention as a major target for drug discovery.

Figure 6.

Effect of malaria drugs on Panx1 functions. a) Inhibition of Panx1 channel currents in Xenopus oocytes by artesunate is dose dependent with an IC 50 of 450 μM. b) Ethidium bromide uptake into frog erythrocytes was inhibited by artemisinin with an IC50 of 0.14 μM. Dye uptake was determined as described in Qiu and Dahl (2009) except ethidium bromide replaced YoPro. unstim. = unstimulated. KGlu = 60 mM potassium gluconate replacing Na⁺ in oocyte Ringer solution. N = 5 for all conditions.

Highlights.

• Pannexin1 is inhibited by gap junction blockers but also by several diverse drugs.

• Panx1 is affected by inhibitors of transport, malaria, and mitochondria.

• Panx1 inhibitors are Cl channel blockers, P2X7R ligands and inflammasome inhibitors.

• Thus, Panx1 may play roles in diverse functions in concert with other proteins.

• Or, Panx1 may share epitopes with a wide range of proteins with different functions.

Abbreviations

CBX

carbenoxolone

GA

glycyrrethinic acid

NPPB

5-nitro-2-(3-phenylpropylamino)benzoic acid

DIDS

disodium 4,4′-diisothiocyanatostilbene-2,2′-disulfonate

IAA-94

indanyloxyacetic acid 94

FFA

flufenamic acid

SITS

disodium 4-acetamido-4′-isothiocyanato-stilben-2,2′-disulfonate

BzATP

3′-O-(4-benzoyl)benzoyl adenosine 5′-triphosphate

IL-1b

interleukin-1b

MBB

maleimidobutyrylbiocytin

MTSET

2-(trimethylammonium)ethyl methanethiosulfonate

TCEP

tris(2-carboxyethyl)phosphine