Gap Junction–Mimetic Peptides do Work, but in Unexpected Ways

Abstract

Gap junction mimetic peptides containing sequences of the extracellular loops of connexins inhibit the de-novo formation of gap junction channels but do not impair the function of existing cell-cell channels. Recently, a flurry of publications appeared showing that such “GAP” peptides attenuate ATP release and/or surrogate measures of it. Although no direct effect on putative connexin “hemichannels” has ever been shown, the peptide effect has been used as diagnostic tool for demonstrating the existence of such channels. However, testing of the peptides on genuine unapposed membrane channels formed by connexins failed to reveal any inhibitory action of the peptides on channel activity. Instead, membrane channels formed by the unrelated pannexin1 were inhibited in the same concentration range as described for the release of ATP. Consequently, rather than indicating connexin involvement in ATP release, the GAP peptide effects represent supporting evidence for a role of pannexin1 in this process.

Keywords:

• mimetic peptide
• connexin
• pannexin
• gap junction
• ATP release
• calcium wave

HISTORY OF GAP JUNCTION MIMETIC PEPTIDES

Gap junctions arise from a pool of precursors, i.e., hexameric connexons in the plasma membrane (Dahl et al., 1992; Musil and Goodenough, 1993). The unpaired connexons are typically in a nonconducting state and only after the docking of the extracellular loops of two connexons in adjacent cells will the complete gap junction channel open. To mimic this docking gate (Dahl, 1996), peptides with the amino acid sequence of the extracellular loops of connexin 32 were applied to unpaired oocytes expressing connexin 32 with the expectation that connexons would now provide a non-junctional membrane conductance (Dahl et al., 1992, 1994). While this attempt to open connexons failed miserably, evidence for peptide interactions with the connexons came in the form of subsequent inhibition of gap junction channel formation in oocyte pairs (Dahl, 1992; Dahl et al., 1994). This inhibition of gap junction function by peptides was also observed in tissue culture cells (Warner et al., 1995) and a series of “Gap” peptides with specificity for certain connexins were developed (Chaytor et al., 1997; Chaytor et al., 1998; Evans and Boitano, 2001; Kwak and Jongsma, 1999). Because the peptides only inhibit the formation of new gap junction channels but do not interfere with the function of existing channels the time course of action in cultured cells or in tissues is bound to be slow and to depend on the turnover rate of gap junctions (Evans and Boitano, 2001; Kwak and Jongsma, 1999; Wang et al., 2007).

More recently a faster action of connexin-mimetic peptides has been observed on ATP release and on calcium wave propagation in the absence of effects on dye coupling (Braet et al., 2003b; Gomes et al., 2005; Leybaert et al., 2003). Lacking other logical alternatives at the time, these data were interpreted as evidence for the involvement of membrane channels formed by connexins, a.k.a. connexons or “hemichannels”, in the measured processes. However, to date no peptide effects on genuine connexon channels have been published. Nevertheless, an increasing number of publications have appeared in the last few years where connexin-mimetic peptides were indiscriminately used as proof of connexin “hemichannel” activity although exclusively downstream events had been measured (Table 1).

Table 1 List of publications where connexin-mimetic peptides were used to attempt channel identification

Cell Calcium 33: 37–48, 2003 (Braet et al., 2003b)

Cell Commun Adhes 10: 251–257, 2003 (Leybaert et al., 2003)

J Cell Physiol 197, 205–213, 2003 (Braet et al., 2003a)

J Neurochem 88, 411–421, 2004 (Vandamme et al., 2004)

Neuron 46: 731–744, 2005 (Pearson et al., 2005)

Invest Ophthalmol Vis Sci 46: 1208–1218, 2005 (Gomes et al., 2005)

Embo J 25: 34–44, 2006 (De Vuyst et al., 2006)

Biochem J 397: 1–14, 2006 (Evans et al., 2006)

Nat Med 12, 950–954. 2006 (Wong et al., 2006)

Exp Eye Res 83: 1225–1237, 2006 (Gomes et al., 2006)

J Biol Chem 281, 21362–21368, 2006 (Takeuchi et al., 2006)

Circ Res 99: 1100–1108, 2006 (Eltzschig et al., 2006)

Embo J 21:5071–5082, 2006 (Pelegrin and Surprenant, 2006)

Invest Ophthalmol Vis Sci 48: 120–133, 2007 (D'Hondt et al., 2007a)

Mol Biol Cell 18: 34–46, 2007 (De Vuyst et al., 2007)

Invest Ophthalmol Vis Sci 48, 1518–1527, 2007 (D'Hondt et al., 2007b)

Embo J 26: 657–667, 2007 (Pelegrin and Surprenant, 2007)

Am J Physiol Heart Circ Physiol. 2007 (Shintani-Ishida et al., 2007)

Proc Natl Acad Sci U S A 104, 8322–8327, 2007 (Retamal et al., 2007)

Embo J. 26, 657–667, 2007 (Romanov et al., 2007)

Not a single one of these studies shows a peptide effect on an unequivocally identified channel. Instead, peptide effects were observed on measured parameters like ATP release or calcium wave propagation and the identity of the channel mediating that activity was deduced in form of a circular argument.

No Effect of Connexin-Mimetic Peptides on the Ideal Target Channel Cx32E₁43

A major obstacle to test the effect of mimetic peptides on wild-type connexon channels is their disinclination to be in the open state in the unpaired configuration. Even under extreme experimental conditions such as a membrane potential of +50 mV their open probability is so low that single-channel events can be resolved in whole cell recordings (Contreras et al., 2003).

In contrast to wild-type connexins whose connexon activity is notoriously difficult to document, a chimera between Cx32 and Cx43 happens to provide robust membrane currents in response to mild depolarization (Pfahnl et al., 1997). Probably the swapping of the extracellular loop altered the docking gate in this mutant connexin. The remaining gating and permeability properties of this channel are consistent with that of Cx32 gap junction channels and the unitary conductance (∼ 50 pS) is twice that of a Cx32 gap junction channel (Hu and Dahl 1999; Pfahnl et al., 1997; Purnick et al., 2000). Because the Cx32E₁43 chimera contains at least two targets for the three most commonly used mimetic peptides (Gap24, Gap 26 and Gap 27), it is the ideal study subject to test connexin-mimetic peptide effects.

Wang et al. (2007) determined that the peptides were active by testing their ability to interfere with gap junction formation in paired oocytes (Table 2). As in previous studies, formation of new gap junction channels was inhibited by peptide binding to unpaired connexons, while existing junctions remained unaffected. Application of the peptides to single oocytes with activated Cx32E₁43 connexons, however, had no acute effects on channel currents (Fig. 1, Table 3). Only over a time span of hours a slight attenuation was observed, possibly because connexons with bound peptide were tagged for degradation.

Figure 1 Membrane currents in oocytes expressing (a) Cx32E₁43 or (b) Panx1. Currents were induced by 5-mV voltage steps from a holding potential of −20 mV in (a) or 120 mV steps from a −60 mV holding potential in (b). Peptide ³²Gap24 applied at 200 μ M did not affect Cx32E₁43 channels but reversibly inhibited Panx1 channels. Cytoplasmic acidification by CO₂ reversibly inhibited Cx32E₁43 channels.

Table 2 Inhibition of gap junction channel formation

	^43Gap26	^43Gap27
Cx32E143	+	nd
Cx43	nd	+

Oocytes expressing connexins were exposed to the peptides before and after pairing. Data from Wang et al., 2007.

Table 3 Inhibition of “hemichannel” activity

	^43Gap26	^43Gap27	^32Gap24	^10Panx1	PEG 1500	γ
Cx32E143	−	−	−	−	−	50pS
Cx46	nd	nd	−	+	−	300pS
Pannexin1	nd	+	+	+	+	475pS

Summary of data from Wang et al., 2007. Acute effects of the peptides on membrane currents and dye flux in oocytes expressing connexins or pannexin1 are listed.

It is formally possible that the chimera may have altered gating properties upon docking to the extracellular loops as compared to wild-type Cx32 or Cx43. However, the general properties of the chimera channel, including single-channel conductance, ability to form complete gap junction channels and the ability of connexin-mimetic peptides to inhibit this channel formation, indicate that the gating characteristics are preserved.

Connexin-Mimetic Peptides Inhibit the Unrelated Pannexin1 Channel

Because the data regarding peptide effects on ATP release and calcium wave propagation are compelling, a peptide effect on an alternate and more likely candidate ATP release channel, pannexin1 (Bao et al., 2004; Dahl and Locovei, 2006; Locovei et al., 2006a, 2006b) was also tested by Wang et al. (2007). As summarized in Table 3, membrane currents through pannexin1 channels were inhibited by both connexin- and pannexin-“mimetic” peptides. If a sequence-specific interaction was at the root of the peptide effects, a stronger effect of the pannexin-“mimetic” peptides on pannexon currents should be observed than that by the connexin-based Gap peptides. Yet all peptides inhibited to similar levels of residual current, and, furthermore, the pannexin-“mimetic” peptide ¹⁰Panx1 (Pelegrin and Surprenant 2006) also inhibited Cx46 connexon currents. Moreover, the peptide effects on pannexon currents could be mimicked by application of the inert molecule PEG 1500. The same pattern of inhibition, albeit more pronounced, was also observed for dye uptake. Carboxyfluorescein uptake by Cx32E₁43-expressing oocytes was unaffected by Gap24 and ¹⁰Panx1, while the same peptides as well as PEG1500 inhibited uptake of the dye by pannexin1-expressing oocytes.

Specificity or Pseudo-Specificity?

Typically, scrambled versions of the peptide sequence serve as controls for sequence specificity. As long as sequence-specific interactions between peptide and a target in a protein are the basis for the observed peptide effect, scrambled peptides are adequate controls. However, the rules change with other mechanisms of action. Overall, the peptide data are consistent with a steric effect. The peptides, because of their size, are excluded from the small Cx32E₁43 channels and thus have no effect on the connexon currents and dye uptake. The considerably larger pannnexin1 channel, on the other hand, can accommodate PEG 1500 and similarly sized peptides, As a consequence, current inhibition is observed. Consistent with a steric block mechanism, inhibition of uptake of dyes that are of larger size than the major current carrying ions was more pronounced (Wang et al., 2007).

With a steric blocking mechanism the ability of a scrambled peptide to affect channel function will depend on the peptide fold. With a 10mer the number of possible folds is 10! (ten factorial), or 3,628,800. It is therefore not surprising that non-identical scrambles yield different effects. In the original study on Gap 24 a scrambled version was ineffective on ATP release. In the study by Wang et al., the scrambled peptide had almost the same activity as Gap 24 (Table 4). On the other hand, a “pseudospecificity” was observed for the ¹⁰Panx1 peptide. The scrambled version was only marginally effective apparently fulfilling the standard criteria of specificity. However, the authentic peptide not only inhibited pannexon currents but also the currents through the channels formed by the completely unrelated Cx46 sequence. Thus, the effect of the ¹⁰Panx1 peptide, like that of the connexin-mimetic peptides, cannot be seen as a specific action on the channel.

Table 4 “Pseudospecificity” of peptide effect

^32Gap24	^32Gap24 scr	^10Panx1	^10Panx1 scr
1	0.83	1	0.26

Fractional effect of scrambled peptides on membrane currents in oocytes expressing pannexin1. Data from Wang et al., 2007.

However, because very few channels in the size range required for accommodating the peptides are expressed in vertebrates, a case for their use as a semi-specific agent could be made. However, a cheaper way to obtain the same effects is the use of PEG 1500. Cx32E₁43 is the perfect target to test the sequence-specific interaction of at least two of the commonly used connexin-mimetic peptides. Yet, not the slightest hint of an effect of the peptides on channel activity could be observed (Wang et al., 2007). Cx32E₁43 forms a small channel that even excludes sucrose, for example (Ma and Dahl, 2006). Thus the channel is too small to observe peptide effects based on steric block given the size mismatch of channel pore and peptide.

Pannexin1 as Candidate ATP Release Channel

The study by Wang et al. (2007) shows that the peptide effect on membrane currents is not sequence-specific and the term “mimetic peptide” is therefore misleading. Even if documentation of effects on connexon channels were to be forthcoming, the specificity problem remains. The study, however, validates the data obtained with the peptides on ATP release and calcium wave propagation. The peptides do indeed inhibit a channel which is highly permeant for ATP (Bao et al., 2004). Only in the light of new findings is the identity of the channel likely to be different. The pannexin1 channel is inhibited by the peptides. Moreover, based on its properties including mechanosensitivity, ATP permeability, activation by micromolar cytoplasmic calcium and localization in places of ATP release the pannexin1 channel is a prime contender for the ATP release channel (Bao et al., 2004; Dahl and Locovei, 2006; Locovei et al., 2006a, 2006b).

Recent findings in the taste bud reemphasize the point of pannexin1 being the ATP release channel and being the target of connexin-mimetic peptides. These peptides have been shown to inhibit ATP release and a current in primary taste cells (Romanov et al., 2007). The authors were wise to attribute this effect to either connexon or pannexon channels, since it subsequently was shown that the ATP-releasing cells in the taste bud express pannexin1, but not connexins (Huang et al., 2007).

SUMMARY

It is well established that connexin-mimetic peptides can inhibit the formation of gap junction channels, while they are ineffective on existing gap junction channels. The claim that these peptides also affect non-junctional membrane channels formed by connexins has been made abundantly but is exclusively based on assumptions and circular arguments. Specifically, there is no published record that the peptides alter properties of any unequivocally identified connexin “hemichannel.” On the contrary, the chimera Cx32E₁43, which at least in theory should be the ideal test object for these peptides, forms membrane channels that are not affected by the peptides. Instead, connexin-mimetic peptides inhibit membrane channels formed by Panx1, which is unrelated in sequence to connexins. Inhibition of the Panx1 channels is non-specific and occurs in a concentration range consistent with channel block. Correspondingly, pannexin-mimetic peptides do affect pannexin1 channels but are similarly non-specific as they also inhibit non-junctional membrane channels formed by Cx46.

Unless the documented effects of connexin-mimetic peptides on ATP release, and surrogate measures of it, are attributable to completely unrelated proteins, they invoke Panx1 rather than connexin channels as mediators of non-vesicular ATP release. Additional arguments that Panx1 channels are prime candidates for representing ATP-release channels can be made. For example, non-vesicular ATP release occurs in the absence of connexins in a variety of settings, including erythrocytes, taste cells, and ATP-releasing cells of most invertebrates. In contrast, pannexin1 and the related invertebrate innexins are not only found in these ATP-releasing cells but also are located in the right places, e.g., the luminal membrane of airway epithelia.