I.  INTRODUCTION

Top Previous Next References

ATP is present outside cells. Many cell types release ATP, and the mechanisms and physiological circumstances range from relatively well understood to quite controversial (see Refs. 51, 135, 161, 191, 407, 485). Extracellular ATP acts on cell surface receptors of the P2X and P2Y types (53, 347); it may be involved in phosphorylation reactions through ectokinases (110), and it is rapidly degraded by a series of cell surface enzymes to ADP, AMP, and adenosine (523), the last of which is taken back into cells by a specific transporter (9).

The first cDNAs encoding P2X receptor subunits were isolated in 1994. Their expression in heterologous cells substantiated the view that P2X receptors were ion channels gated by ATP. This review deals first with the molecular properties of the P2X receptors when heterologously expressed and is organized into sections according to the identified subunits. The second part of the review deals with the functional properties of P2X receptors expressed in native cells, reporting studies to establish their molecular identity and physiological role. The emphasis here is on work that most directly addresses the molecular characterization of the receptors; ideally, such studies would use the approaches of 1) gene knock-out, 2) antisense knock-down, 3) biophysical methods such as the kinetics of the responses or the permeation properties of the channel, and 4) quantitative pharmacological studies with a range of agonists and antagonists. Although the era of the molecular physiology of P2X receptors began with the cloning of the cDNAs, there was already a substantial and highly credible body of work that showed the importance of signaling by extracellular nucleotides in many tissue and organ systems. This has been extensively reviewed previously (1, 50, 53, 376).

	II.  THE P2X RECEPTOR GENE FAMILY

Top Previous Next References

There are seven genes for P2X receptor subunits. Their chromosomal locations are summarized in Table 1. P2X₄ and P2X₇ subunit genes are located close to the tip of the long arm of chromosome 12 (12q24.31), where 230 kb of genomic DNA contain also the gene for calmoldulin-dependent kinase type II. On the basis of radiation hybrid mapping, they were judged to be <130 kb apart (46). In fact, the genes are adjacent in the genomes of humans (23,492 bp separating) and mice (26,464 bp separating; chromosome 5). This presumably reflects gene duplication, and P2X₄ and P2X₇ subunits are among the most closely related pairs in amino acid sequences (Figs. 1 and 2). P2X₁ and P2X₅ genes are also very close together (and close to the gene encoding the vanilloid receptor VR1) on the short arm of chromosome 13 (Table 1). The remaining genes are on different chromosomes (Table 1).

View larger version (98K): [in this window] [in a new window]   Fig. 1. P2X receptor subunits. Alignment of rat amino acid sequences is shown. Open boxes indicate conserved amino acids. Shading indicates conserved cysteines. Solid overlines indicate hydrophobic, membrane-spanning regions. Positions corresponding to the beginning of each exon are indicated. Sequences and gene structure are deduced from NCBI accession numbers P47824 (rP2X1), 2020424A (rP2X2), CAA62594 (rP2X3), CAA61037 (rP2X4), CAA63052 (rP2X5), CAA63053 (rP2X6), and CAA65131 (rP2X7).

View larger version (25K): [in this window] [in a new window]   Fig. 2. Dendrogram to show relatedness of 29 P2X receptor subunits. Full-length amino acid sequences were aligned with Clustal W using default parameters. The dendrogram was constructed with TreeView. h, Human (Homo sapiens); r, rat (Rattus norvegicus); m, mouse (Mus musculus); gp, guinea pig (Cavia porcellus); c, chicken (Gallus gallus); zf, zebrafish (Danio rerio); bf, bullfrog (Rana catesbeiana); x, claw-toed frog (Xenopus laevis); f, fugu (Takifugu rubripes). The ellipses indicate the apparent clustering by relatedness into subfamilies.

The genes vary considerably in size (e.g., mP2X₃: 40 kb, Ref. 434; hP2X₆: 12 kb, Ref. 471). The full-length forms have 11-13 exons, and all share a common structure, with well-conserved intron/exon boundaries (Fig. 1). Many spliced forms of the receptor subunits (or fragments thereof) have been described (Table 2); the majority of these represent simple forms in which one or more exons have been spliced out, although some have altered exons through the use of alternative donor/acceptor sites. Several full-length nonmammalian vertebrate sequences are available (Fig. 2). There are no reports of homologous sequences from invertebrate species, although there is considerable functional evidence that extracellular ATP and other nucleotides can directly gate ion channels in invertebrates including protists (8, 71, 241, 372).

	III.  THE P2X RECEPTOR PROTEIN FAMILY

Top Previous Next References

The P2X subunit proteins are 384 (cP2X₄) to 595 (P2X₇) amino acids long. Each has two hydrophobic regions of sufficient length to cross the plasma membrane (37, 346, 472) (Fig. 1); the first of these extends from residue 30 to 50, and the second from residue 330 to 353 (numbers refer to the rat P2X₂ receptor). These hydrophobic regions are separated by the bulk of the polypeptide; considerable evidence presented below indicates that much, perhaps all, of this lies on the extracellular aspect of the membrane. The NH₂ and COOH termini are therefore presumed to be cytoplasmic. The COOH-terminal regions diverge in sequence considerably. Considering the region of the protein which includes the two transmembrane domains and the intervening extracellular domain (i.e., amino acids 30-353 of P2X₂), the proteins are from 40 to 55% pairwise identical (Table 3). The P2X₄ sequence is most closely related to more of the other forms, and the P2X₇ sequence is least like the others; these observations are true whichever species are considered (Table 3).

The amino acid identity between P2X receptor subunits is distributed throughout the extracellular domain, a striking feature of which is the conservation of 10 cysteine residues among all known receptors (Fig. 1). These are not obviously conserved in blocks with respect to exonic structure; the first half of the domain contains six cysteines (exons 2, 4, and 5), and the four further cysteines are in sequence encoded by exons 7 and 8. It is generally thought that such cysteines in an extracellular location would be oxidized and thus contribute to the tertiary structure of the protein by disulfide bond formation; there is no direct evidence for this in the sense that treatment with reducing agents has no effect on channel function (74, 114, 379). The possible pattern of disulfide bond formation has been approached by systematic cysteine to alanine substitutions. Clyne et al. (74) compared the effects of such substitutions (in the rat P2X₂ receptor) on sensitivity to ATP and potentiation by zinc and found that the results could be grouped according to residue. Ennion and Evans (114) carried out similar experiments for the human P2X₁ receptor, but used a clever additional approach. This was to demonstrate that the receptor became accessible to labeling by MTSEA-biotin after a cysteine to alanine mutation, presumably as a result of a free sulfhydryl becoming available. By adding a second cysteine to alanine mutations, they were able in some cases to assign partners, although not all possibilities were tested. The results of these experiments are illustrated schematically in Figure 3. The finding by Ennion and Evans (114) that Cys-124, Cys-130, Cys-147, and Cys-158 (rat P2X₂ numbering) were able to interact promiscuously might indicate that these residues are clustered, as would be expected for a metal ion binding site. However, the ion seems not to be zinc (see Ref. 74).

View larger version (22K): [in this window] [in a new window]   Fig. 3. Glycosylation, phosphorylation, and possible disulfide bonding of P2X2 receptors. Solid circles (N) indicate the three sites that are glycosylated in the native P2X2 receptor (data from Refs. 340, 459, 460). Open circles (T, S) indicate the positions of Thr-18 (threonine phosphorylated by protein kinase C: Ref. 33) and Ser-431 (serine phosphorylated by protein kinase A: Ref. 66). Open circles (C) indicate the 10 conserved cysteines. Alanine substitution and MTSEA-biotin labeling experiments indicate possible disulfide bond formation; data are from P2X1 receptor (114) and from P2X2 receptor (74). Open squares (H) indicate histidine residues involved in zinc (His-120, His-213) and proton (His-319) binding (data from Ref. 74).

There is no reported homology of sequence between P2X receptors and other proteins, although a similarity has been suggested to class II aminoacyl-tRNA synthetases (138). This similarity is mostly between the predicted secondary structure of the second half of the extracellular domain (residues 170-330) and that known from X-ray crystallography of the synthetases, which form their catalytic site from a seven-stranded antiparallel -pleated sheet (92). It was stated that the first half of the extracellular domain (residues 110-170) may provide a metal ion binding site (138), but there is no evidence that the cysteines are involved in this (74).

All the P2X receptor subunits have consensus sequences for N-linked glycosylation (Asn-X-Ser/Thr), and some glycosylation is essential for trafficking to the cell surface. The P2X₁ subunit sequence has five such consensus sites, four of which are conserved among human, rat, and mouse sequences (asparagines 153, 184, 284, 300 in rat P2X₁). These four sites can all be glycosylated (341). The P2X₂ subunit has three such sites (asparagines 182, 239, and 298 in rat P2X₂), and all are glycosylated in oocytes (340) and HEK293 cells (459). The consequences of removal (by tunicamycin) or prevention (by mutagenesis) of glycosylation have been studied. Receptors in which any two of the three sites are glycosylated appear at the cell surface and are fully functional. Receptors in which only one site is glycosylated give barely detectable currents in response to ATP, and channels with no sites glycosylated give no current. These double and triple mutant receptors are retained within the cell, as detected by immunohistochemistry of a COOH-terminal epitope tag (340), or immunoprecipitation of cell surface membrane protein labeled with sulfo-NHS-LC-biotin labeling (459). The other P2X receptor subunits also have consensus sequences for N-linked glycosylation; these are well conserved in their positions among species variants but incompletely conserved among the receptors (P2X₃, four sites; P2X₄, six sites; P2X₅, two sites; P2X₆:, three sites; P2X₇, three sites).

The membrane topology of the protein has also been addressed by determining the location of glycosylation sites; thus the studies on the P2X₂ receptor indicate that asparagines 182, 239, and 298 are all localized to the extracellular domain (Fig. 3). Site-directed mutagenesis has been used to introduce new consensus sites into a background P2X₂ receptor in which the three natural sites have been removed (340, 459). These studies provide direct support for the proposed topology, with a large extracellular domain between the two membrane-spanning regions. Further evidence that the NH₂ and COOH termini reside on the same side of the membrane comes from studies in which two cDNAs have been joined in tandem (340, 442, 460). Such constructs express fully functional channels, and point mutations in one or other of the concatenated domains indicate that both contribute to the channel (340, 442). Finally, confocal immunofluorescence microscopy has been carried out on HEK293 cells transfected with P2X₂ receptors carrying a FLAG epitope at the NH₂ or COOH terminus; in either case, the epitope was accessible to antibody only when the cells had been permeabilized (460).

The P2X₇ subunit has a much longer COOH terminus than the other subunits, and this contains an additional hydrophobic domain (residues 510-530) that is sufficiently long to cross the plasma membrane. There is no published definitive evidence that places the COOH terminus of this receptor inside or outside the cell, but membrane topology algorithms suggest an intracellular location.

Evidence for heteromultimeric receptors has come from functional expression studies, whereas although these show that at least two different subunits can contribute to the ion channel, they are inconclusive with regard to the actual number of subunits. Three kinds of biochemical approaches have also been used. Schmalzing and colleagues (341) cross-linked P2X₁ and P2X₃ receptors, either in intact oocytes or after solubilization with digitonin. The receptors were NH₂-terminally tagged with hexahistidine sequences and cross-linked either with 3,3'-dithiobis(sulfosuccinimidyl-propionate) or with bifunctional analogs of the antagonist pyridoxal-phosphate-6-azophenyl-2',4'-disulfonic acid (PPADS). One of these analogs (CLII) has a flexible spacer between the phenyl group so as to provide up to 3.4 nm between the two pyridoxal aldehyde moieties; it was able to cross-link digitonin-solubilized, purified P2X₁ (or P2X₃) subunits almost quantitatively to homotrimers, and this was reversed to monomers by dithionite, which cleaves the azo bonds of CLII. Cross-linking with CLII of octylglucoside-solubilized P2X₁ receptors led to the appearance of hexamers and trimers, but not intermediate forms (341).

In a second approach, blue native polyacrylamide gel electrophoresis was used to estimate the molecular mass of the P2X₁ receptor isolated under nondenaturing conditions from digitonin extracts of oocytes. These were almost exclusively trimers, whereas parallel experiments on the muscle type nicotinic receptor (coexpression of , , , and  subunits) clearly resolved the expected pentameric structure. Generally consistent results have been reported for rat P2X₇ receptors (239).

The third approach used the hexahistidine-tagged ectodomain of the rat P2X₂ receptor (residues Lys-53 to Lys-308). This was expressed in Escherichia coli, solubilized in urea, and purified by nickel-affinity chromatography (240). After sulfitolysis and refolding, the protein was photoaffinity labeled with [-³²P]ATP; the labeling was prevented by an excess of cold ATP and by suramin (1 µM) and cibacron blue (10 µM). The molecular size of the labeled protein was estimated by equilibrium sedimentation centrifugation as 132 kDa, which is about four times the calculated size of the ectodomain (29 kDa). Obviously, one difficulty of this approach is knowing whether the ectodomain is correctly refolded and whether the ectodomain alone can reconstitute the original ATP binding site. In fact, more recent work by Egan, Voigt, and colleagues (461) indicates that residues critical for multimerization are in or near the second membrane-spanning segment (461), which was not present in the ectodomain experiments.

Voigt, Egan, and colleagues (460, 462) have also determined which pairs of subunits are potentially able to coassemble. The approach was based on coimmunoprecipitation of epitope-tagged subunits after expression in HEK293 cells (460, 462). Table 4 summarizes their results, which are also consistent with the findings of others with respect to the P2X₂/P2X₃ (374), P2X₄/P2X₆ (269), and P2X₁/P2X₅ (270, 447, 463). Thus at one extreme P2X₇ subunits will coassemble with no others (in this biochemical test); they are also the most distinct in sequence (Table 4). P2X₅ receptors will assemble with any others, except P2X₇.

In summary, the biochemical evidence that the protein readily forms stable trimers and hexamers is suggestive that the intact receptor assembles from three or six subunits in heterologous expression systems. However, there are two types of caveat. First, similar approaches resulted in similar conclusions for the large-conductance mechanosensitive channel (mscL) of E. coli (27); this is a channel in which the subunits have a similar transmembrane topology to that proposed for P2X subunits. Electron microscopic images of two-dimensional crystals of reconstituted mscL channels were also interpreted as hexamers (396), but subsequent crystallization of the Mycobacterium tuberculosis mscL shows that this channel actually forms as a pentamer (59). Second, assembly in native cells may be influenced significantly by associated proteins that are not present in heterologous expression systems.

	IV.  HETEROLOGOUS EXPRESSION OF CLONED RECEPTORS

Top Previous Next References

A cDNA encoding the P2X₁ receptor was isolated by direct expression in Xenopus oocytes, beginning with a cDNA library made from rat vas deferens (472). The deduced protein has 399 amino acids. It was noted by Valera et al. (472) that the database already contained a cDNA (RP-2) identical in sequence to part of the P2X₁ receptor cDNA (Table 2); RP-2 cDNA was isolated by subtractive hybridization from thymocytes undergoing apoptosis (353). Human and mouse cDNAs have also been cloned and expressed (473).

ATP-gated channels express well in oocytes and HEK293 cells after injection or transfection with the P2X₁ subunit cDNA (121, 472, 495). Approximately equal currents can be elicited by ATP or -methylene ATP (meATP), each having an EC₅₀ close to 1 µM (121, 472). 2',3'-O-(benzoyl-4-benzoyl)-ATP (BzATP) is also an effective agonist (25, 121); it is particularly potent when calcium flux is measured, with an EC₅₀ in the low nanomolar range (25). The human receptor was cloned from urinary bladder and is basically similar in properties to the rat receptor; both resemble closely the responses of smooth muscle cells of the vas deferens or bladder (122, 473). The most striking property of the P2X₁ receptor is the mimicry of the agonist actions of ATP by meATP, which distinguishes P2X₁ and P2X₃ receptors from the other homomeric forms. MeATP is also useful in this respect; although it does cause maximal currents as large as those evoked by ATP, it activates P2X₁ receptors at concentrations (10 µM) that are ~30-fold less than those needed to activate homomeric P2X₃ receptors (25, 121, 147, 472).

Ennion et al. (116) have mutated the positively charged residues in the human P2X₁ receptor, in an effort to determine which might contribute to the ATP binding site. They found that the lysines most sensitive to substitution by alanine or arginine were Lys-68 and Lys-70 (corresponding to Lys-69 and Lys-71 in the rat P2X₂ sequence); other positively charged residues closer to the COOH-terminal end of the extracellular loop may also be involved (particularly Lys-309) (116). Negatively charged residues have also been mutated to alanine (117). However, even though these (Asp-86, Asp-89, Glu-119, Asp-129, Glu-160, Glu-168, Asp-170, Glu-183, Asp-262, Asp-264, Asp-316 P2X₁ numbering, see Fig. 1) are highly conserved among all P2X receptors, in no case did the substitution by alanine cause any significiant change in the sensitivity to ATP.

The deletion of one leucine residue at the inner end of the second transmembrane domain results in a receptor that does not express and a dominant negative phenotype when the mutated form is coexpressed with wild-type P2X₁ receptors (352); this mutation was made because it was detected in a 6 yr old with a bleeding diathesis that appeared to be due to deficient platelet aggregation, but cause and effect remain obscure. P2X₁ receptors are expressed by platelets (see sect. vI3). Finally, a spliced form of the hP2X₁ receptor that lacks most of exon 6 (including the conserved glycosylation site Asn-184) has been found in platelets and megakaryocyte cell line (156). When expressed in fibroblasts and studied by calcium imaging, this receptor showed a much reduced sensitivity to meATP.

P2X₁ receptors are blocked by suramin and PPADS (121), but there are now newer antagonists that are more P2X₁ selective. MRS2220 (cyclic pyridoxine-4,5-monophosphate-6-azo-phenyl-2',5'-disulfonate) blocks at ~10 µM but has no effect on currents evoked at P2X₂ or P2X₄ receptors (or human P2Y₂, human P2Y₄, or rat P2Y₆) (210). The structures of the main antagonists are shown in Figure 4. Certain suramin analogs also exhibit a relatively high affinity for P2X₁ receptors: 8,8'-carbonylbis(imino-3,1-phenylene carbonylimino)bis(1,3,5-naphthalenetrisulfonic acid) (NF023) blocks P2X₁ receptors more effectively than P2X₂, P2X₃, and P2X₄ receptors (432), and 8,8'-carbonylbis(imino-4,1-phenylene carbonylimino)bis(1,3,5-naphthalenetrisulfonic acid) (NF279) blocks P2X₁ receptors in oocytes with an IC₅₀ of 50 nM (249). The PPADS analog pyridoxal-5'-phosphate-6-(2'-naphthylazo-6'-nitro-4',8'-disulfonate) (PPNDS) blocks P2X₁ receptors with an IC₅₀ of ~10 nM (266). Another useful antagonist at P2X₁ receptors is 2',3'-O-(2,4,6-trinitrophenyl)-ATP (TNP-ATP), which has an IC₅₀ of ~1 nM (483). Among the other receptors, only the P2X₃ homomers and P2X₂/P2X₃ heteromers are similarly sensitive. This action of TNP-ATP is shared by TNP-GTP, TNP-ADP, and TNP-AMP, but not by TNP-adenosine. Finally, di-inosine pentaphosphate (Ip5I) has been described as a selective antagonist at recombinant P2X₁ receptors (242).

View larger version (29K): [in this window] [in a new window]   Fig. 4. Structural formulas for several antagonists used in the study of P2X receptors.

Little information is available with respect to the regions of the receptor involved in antagonist binding. Ennion et al. (116) have determined the effects on suramin antagonism of mutating positively charged amino acids in the extracellular loop. In oocytes expressing human P2X₁ receptors, the block by suramin was slightly increased in receptors with K70R, K215R, and K309R substitutions and decreased in the case of R202A and R292A.

The homomeric P2X₁ receptor is a cation-selective channel that shows little selectivity for sodium over potassium (122). It has a low permeability to larger organic cations such as Tris (P_Tris/P_Na 0.18) or N-methyl-D-glucamine (P_NMDG/P_Na 0.04), at least when tested with brief agonist applications (see below). It has a relatively high permeability to calcium, as estimated from reversal potentials in bi-ionic conditions (P_Ca/P_Na 4 in 112 mM Ca, corrected for ionic activities) (122). Extracellular calcium has little or no inhibitory effect on P2X₁ receptor currents, and this is in marked contrast to the P2X₂ receptor (122). Extracellular acidification inhibits currents at P2X₁ receptors. There are only preliminary reports of the single-channel currents at P2X₁ receptors; the unitary conductance was ~18 pS (122, 472).

Desensitization means the decline in the current elicited by ATP during the continued presence of ATP. The time domain is important; in some P2X receptors this decline occurs in milliseconds (fast desensitization: P2X₁, P2X₃), and in others it occurs 100-1,000 times more slowly (slow desensitization: P2X₂, P2X₄). Figure 5 summarizes the fast and slow desensitization observed for the six P2X receptors that express as homomers in HEK293 cells.

View larger version (12K): [in this window] [in a new window]   Fig. 5. Fast (top) and slow (bottom) desensitization compared for homomeric rat P2X receptors. Note 10-fold difference in time scale. Fast desensitization is observed only with P2X1 and P2X3: brief applications (2-s duration) of ATP (30 µM, except 1 mM for P2X7). Slow desensitization is observed for P2X2 and P2X4: more prolonged applications (60-s duration) of ATP (30 µM, except 1 mM for P2X7). HEK293 cells were transfected with 1 µg/ml cDNA (each in pcDNA3.1) 48 h before these whole cell recordings were made. In all cases except P2X7, the response shown is that seen for the first application of ATP to that cell. For P2X7, one 2-min application of ATP had been made before the application shown. (Figure kindly provided by A. Surprenant.)

P2X₁ receptors undergo fast desensitization when the agonist application is continued for more than several hundred milliseconds (Fig. 5). The desensitization is not marked at lower concentrations (less than or equal to EC₅₀) but becomes prominent at concentrations above 1 µM. Recovery from desensitization is extremely slow; second and subsequent applications of ATP do not elicit as large currents as the first application, and such subsequent applications must be made at long intervals (>15 min) for reproducible responses to be obtained.

The consequences of desensitization can be profound with respect to the detection of functional effects of ATP. The human leukemia cells (HL60) and rat basophilic leukemia cells (RBL) express P2X₁ receptor mRNA and protein, but inward currents in response to extracellular ATP can only be observed after treating the cells with apyrase (45). This surprising observation suggested that ATP was being continuously released from the cells (which was also shown directly by the luciferin-luciferase assay), and responses to exogenous ATP were not observed because the receptor was desensitized. Treatment with apyrase allowed the receptors to recover from desensitization. In view of the increasing number of cell types shown to release ATP (see Refs. 135, 178, 407, 485), this is likely to be a considerable experimental problem in a wide range of tissues.

The marked contrast in the kinetics of desensitization between P2X₁ and P2X₂ receptors prompted a series of experiments with chimeric constructs in an effort to map the domains involved (495). These experiments indicated that desensitization required two regions of the P2X₁ receptor; if either region was replaced by the equivalent segment from the P2X₂ receptor, then desensitization no longer occurred. Each region is 34 amino acids long, comprising the transmembrane segment and the contiguous residues (~14) on its intracellular aspect. These results suggest that closure of the channel during the continued presence of the agonist requires concerted conformational changes involving both transmembrane segments.

Mutations of positively charged residues in the extracellular loop of the human P2X₁ receptor can also have dramatic effects on desensitization. The substitution K68A produces a receptor in which desensitization is greatly slowed (~100-fold), and smaller effects were seen for R292K, K309A, and K309R. Activation of P2X receptors with these mutations also requires much higher concentrations of ATP (see above). Parker (359) found that the rate of desensitization of wild-type P2X₁ receptors stably expressed in HEK293 cells slowed from ~60 ms to several seconds when the cells were passaged in culture; this change was not seen in M332I and T333S mutations, and it was reversed by cytochalasins B and D (5 µM, 2-4 h). The threonine residue at position 18 of the P2X₁ receptor is completely conserved and lies in a protein kinase C consensus sequence. Ennion and Evans (115) showed that replacing this threonine by alanine resulted in a receptor that desensitized 10 times faster than the wild-type receptor, but there is no direct evidence that this change results from its inability to be phosphorylated. The residue lies within the domains identified by Werner et al. (495) as being responsible for the swap in desensitization among P2X₁, P2X₂, and P2X₃ receptors.

Adenoviral expression of a P2X₁ receptor-green fluorescent protein (GFP) construct in vas deferens shows the receptor to be localized in clusters, with larger ones apposing nerve varicosities (105). Heterologous expression in rat dissociated superior cervical ganglia presented a similar picture (284); these cells normally exhibit a nondesensitizing response to 1-s applications of ATP, so the time course of the appearance of the P2X₁ subunit was followed functionally by the presence of a desensitizing current in response to meATP. Exposure to meATP for ~60 s resulted in a loss of GFP from the plasma membrane, with its appearance in acidic endosomes (as judged by monensin sensitivity). Ennion and Evans (113) have made similar conclusions; they found that a 30-min treatment with meATP (100 µM) resulted in a 50% loss of biotinylated P2X₁ receptor on the cell surface. Even a 2-min treatment with meATP (10 µM) was sufficient to cause a long-lasting inhibition of the contractile response. Cell surface receptors recovered within 10 min of terminating the agonist application, and the contractile response recovered more slowly. Therefore, sustained application of agonist to P2X₁ receptors results in 1) rapid (few milliseconds) channel opening, 2) fast desensitization ( ~300 ms), and 3) receptor internalization ( ~1-3 min). If the agonist application is terminated, the receptors reappear at the cell surface ( ~10 min).

The rat P2X₂ receptor cDNA was isolated from a library constructed from NGF-differentiated PC12 cells by testing pools for functional expression in Xenopus oocytes (37). The human receptor cDNA was amplified from pituitary gland (292).

The current elicited by ATP differs prominently from that observed at P2X₁ receptors in that the agonist action of ATP is not mimicked by meATP. There are no agonists currently known that are selective for P2X₂ receptors, but certain effects of ions are useful. Thus P2X₂ receptors are potentiated by protons (97, 244, 441, 500) and by low concentrations of zinc and copper (37, 500, 511). Systematic mutation of cysteine and histidine residues in the rat P2X₂ receptor has indicated that 2 of the 9 histidines (His-120, His-213) but none of the 10 cysteines seem to contribute to the binding of zinc (74). In contrast, the potentiation by protons was much reduced by removing a different histidine residue (His-319) (74).

Homomeric P2X₂ receptors have been thoroughly studied at the single-channel level after expression in oocytes and HEK293 cells (Fig. 6) (97-99). Several models were fitted to the kinetics of the single channels, and the most likely (Fig. 6) had the following features: 1) three molecules of ATP bind to the channel; 2) the binding steps are not independent, but positively cooperative; 3) two open states connect to a common ATP-independent closed state; 4) activation and inactivation proceed along the same pathway; and 4) channels only open when fully liganded.

View larger version (24K): [in this window] [in a new window]   Fig. 6. Single-channel currents elicited by ATP in oocytes expressing P2X2 receptors. A, left: typical unitary currents in response to ATP (2 µM; 1 mM magnesium, 1 mM calcium) at the membrane potentials indicated. Right: current-voltage plot for unitary currents shows strong rectification; this was unchanged by removal of calcium and magnesium. Outside-out recordings from stably transfected HEK293 cell. B: kinetic scheme for gating of the P2X2 receptor which best fits the single-channel records has three sequential ATP binding steps (C1 to C2, C2 to C3, and C3 to C4). [From Ding and Sachs (97). Reproduced from Journal of General Physiology, 1999, by copyright permission of The Rockefeller University Press.]

Efforts have been made to identify amino acid residues that might contribute to the ATP binding site. On the basis that hydrogen bonding with polar or charged side chains were likely to be involved, such amino acids were mutated individually to alanine (217). A region was identified proximal to the first transmembrane domain that contained two lysine residues that were critical for the action of ATP (Lys-69 and Lys-71); these correspond to the residues identified by Ennion et al. (116) in the P2X₁ receptor. Further analysis of this region showed that the attachment of negatively charged methanethiosulfonates to a cysteine introduced at Ile-67 resulted in a parallel rightward shift in the ATP concentration-effect curve, consistent with a reduced affinity for ATP. Positive or uncharged methanethiosulfonates depressed the maximal responses to ATP, consistent with an impairment of the conformational changes leading from binding to channel opening. This inhibition by the methanethiosulfonates was prevented by preexposure to ATP, suggesting occlusion of the binding site (217). Taken together, these results are consistent with Ile-67 being located close to the binding pocket for ATP.

There are no antagonists selective for P2X₂ receptors. The responses to brief applications of ATP are inhibited by calcium ions, with an IC₅₀ of ~5 mM (121), and it may be possible to take advantage of this to differentiate them from other forms. The divalent cations cause a fast (i.e., low affinity) block of single P2X₂ channels (98, 99). The order of potency is Mn > Mg > Ca > Ba, which is the order of ionic radii. This suggests that the divalent ions are binding to a charged site within the channel (98). In the case of calcium, the concentration giving 50% block was 3.8 mM. These observations correlate well with those made by Nakazawa and Hess (326) for PC12 cells.

3.  Permeation properties

A) SINGLE-CHANNEL RECORDING. Single-channel recordings made on outside-out patches from HEK293 cells expressing P2X₂ receptors have been described (97, 98). Openings were associated with an unusually large increase in current noise, suggestive of several open states interchanging more rapidly than could be resolved. The maximal probability of opening observed was 0.61; the EC₅₀ for ATP was ~10 µM, and the Hill coefficient was 2. 3. The unitary currents showed strong inward rectification and had a conductance of 30 pS at 100 mV (Fig. 6). Current flow through the channels was associated with excess current noise, which could not be accounted for by the flickery block of impermeant ions. The permeant ions are ordered in selectivity according to Eisenman's sequence IV (K⁺ > Rb⁺ > Cs⁺ > Na⁺ > Li⁺), and the channels were essentially impermeant to NMDG, Tris, and tetraethylammonium (TEA).

B) RECTIFICATION. At the whole cell level, the currents induced by ATP also show strong inward rectification (37, 122). This is very variable from cell to cell (oocytes or HEK293) cells, with occasional cells showing almost linear current-voltage relations (122). The rectification results in part from rectification in the unitary currents; unitary conductance falls from ~20 pS at 120 mV to ~10 pS at 50 mV. The mechanism of this rectification is not known; its persistence in divalent-free solutions indicates that it does not simply result from block of the permeation pathway by divalent cations (97, 98). Voltage-jump experiments indicate that there is an additional time-dependent component of inward rectification in the voltage range of 100 to 40 mV; when the membrane is stepped to 100 mV, the new conductance is reached with a time constant of ~12 ms (522).

C) CALCIUM PERMEABILITY. P2X₂ receptors are permeable to calcium. P_Ca/P_Na is ~2.5 in 5 mM external calcium; this is less than homomeric P2X₁ (122) and P2X₄ (145) receptors but more than homomeric P2X₃ receptors (482). Unfortunately, it is not straightforward to make an accurate measurement of the calcium permeability of the P2X₂ receptor. The preferred experiment, in which calcium is the only extracellular cation, is difficult because of the block of the current that this causes. The alternative approach is to combine extracellular calcium with another extracellular cation that is impermeant. NMDG is commonly used, but this can be complicated by the time-dependent increase in permeability to NMDG that occurs in some cells transfected with P2X₂ receptors (see below).

D) CYSTEINE SUBSTITUTION. Amino acid residues that might contribute to the permeation path have been identified by the substituted cysteine accessibility method. Rassendren et al. (379) used three methanethiosulfonates to probe the region from Val-316 to Thr-354 in the rat P2X₂ receptor. They found that application of methanethiosulfonates inhibited the currents evoked by ATP in the cases of I328C, N333C, T336C, and D349C and augmented the current for S340C and G342C. In the case of L338C and D349C, only the small positively charged methanethiosulfonate [ethylammonium-methanethiosulfonate (MTSEA)] was effective; for D349C (but not L338C), this block required channel opening. Because MTSEA can permeate the open channel, it was suggested that Asp-349 lies on the internal side of the channel "gate." For the other three positions (I328C, N333C, and T336C), inhibition occurred with methanethiosulfonates that were negatively charged [sulfonatoethyl-methanethiosulfonate (MTSES)] or positively charged [ethyltrimethylammonium-methanethiosulfonate (MTSET)]. It was concluded that these residues lay outside the membrane electric field. On the other hand, the development of block by methanethiosulfonates at T336C introduced new rectification into the channel, which suggests that it might lie in the permeation path. These authors drew attention to the difficulties in using MTSEA, which gave much more variable results that MTSES and MTSET. Substitutions at Ile-328, Asn-333, and Thr-336 (with Ala, Gly, Asn, Asp, Glu, Lys, Ser, and Gln) also increase the dilation of the channel; all cells expressing N333A show a large increase in NMDG permeability and YO-PRO-1 uptake (481). The results of the substituted cysteine accessibility experiments are summarized schematically in Figure 7.

View larger version (37K): [in this window] [in a new window]   Fig. 7. Schematic summary of cysteine substitution studies on rat P2X2 receptor. Two segments of the P2X2 receptor are depicted; Asp-15 to Lys-71 includes the first transmembrane domain, and Val-316 to Thr-354 includes the second transmembrane domain. The representation of the membrane-spanning segments as -helices is hypothetical and suggested only by secondary structure prediction algorithms. Each amino acid shown has been individually mutated to cysteine. Mutated receptors were expressed in HEK293 cells, and currents were elicited by ATP. Outline letters (Tyr-16, Arg-34, Tyr-43, Gln-56, Lys-71) indicate those positions where no functional channels were expressed. Solid gray circles indicate those positions where a methanethiosulfonate inhibited the current (including membrane-permeable methanethiosulfonates). The line between Val-48 and Ile-328 indicates the disulfide bond that forms when both residues are substituted by cysteine; these residues are not necessarily provided by the same subunit. [Data from Rassendren et al. (379), Jiang et al. (216), and Jiang et al. (217).]

E) PERMEABILITY INCREASE WITH TIME. In some cells expressing P2X₂ receptors, the permeation pathway of the P2X₂ receptor appears to dilate during agonist applications lasting for several seconds (HEK293 cells, Refs. 481, 480; oocytes, Ref. 229). This is evidenced by a progressive increase in the permeability to large organic cations, including NMDG, Tris, and TEA (Figs. 8 and 9). Measured under bi-ionic conditions in mammalian cells, the permeability to NMDG is initially very low (<5% that of sodium), but this increases (exponentially with time constant 7 s) until NMDG is ~50% as permeable as sodium (480, 481).

View larger version (14K): [in this window] [in a new window]   Fig. 8. Structures and dimensions of some cations used to estimate the pore size of P2X receptors. These dimensions were measured from space-filling models (van der Waals radii) of energy-minimized conformations drawn in CSC ChemDraw.

View larger version (22K): [in this window] [in a new window]   Fig. 9. Time-dependent permeability increase in cells expressing P2X2 and P2X4 receptors. A: HEK293 cells expressing P2X2 receptors (left) and serotonin (5-HT3) receptors (right). In each case, the only external cation was N-methyl-D-glucamine (NMDG); the internal solution was sodium. At 60 mV, the application of agonist elicits an outward current, reflecting the low permeability to NMDG. In the case of the P2X2 receptor, the current turns inward within a few seconds, and this is accompanied by a positive shift in reversal potential (A, bottom panels). In the case of the 5-HT3 receptor, the current declines (due to desensitization) but never becomes inward; the reversal potential does not change. B: nodose ganglion neurons show responses to ATP (100 µM) and 5-HT (30 µM) that closely resemble those seen in transfected HEK293 cells. C: Xenopus oocytes expressing wild-type P2X4 receptors show a biphasic response to ATP (left). The initial transient current reflects current flow through a sodium-selective channel, and the later, larger, and more prolonged current reflects current flow through an NMDG-permeable channel (see Ref. 229). Mutation of a glycine residue in the second membrane-spanning domain selectively eliminates either the fast, sodium-selective current (G347R) or the sustained NMDG-permeable component (G347Y). [A and B modified from Virginio et al. (481); C modified from Khakh et al. (229).]

1

1

1

1

With whole cell recording, currents at P2X₂ receptors decline little during agonist applications of a few seconds (37, 81) (Fig. 5). For this reason, the P2X₂ receptor is generally described as nondesensitizing, compared with the P2X₁ and P2X₃ receptors. However, there is a progressive decline in the current that occurs during applications of several tens of seconds (slow desensitization; Fig. 5). This has been investigated in two respects: 1) by mutagenesis and 2) by studies on its calcium dependence. Amino acid residues in the NH₂ terminus, the transmembrane domains, and the COOH terminus can influence this slow desensitization.

In the NH₂ terminus, Thr-18 can be phosphorylated by protein kinase C (33). The mutants T18A or T18N show much accelerated slow desensitization; this is complete within 1-2 s, which is still considerably slower than the rate of fast desensitization observed for homomeric P2X₁ and P2X₃ receptors. A similar effect was observed with K20T, which removes the consensus site for protein kinase C phosphorylation while leaving the conserved threonine unchanged. These results suggest that the wild-type channel is constitutively phosphorylated by protein kinase C, and when this does not occur, the channel exhibits more rapid desensitization (33). However, it is not clear whether this explanation can be generalized among P2X receptors. Threonine occupies the position corresponding to Thr-18 in all P2X receptors. P2X₁ receptors exhibit fast desensitization, and this becomes even faster for P2X₁[T18A] (115); however, P2X₁ receptor desensitization is unaltered by phorbol esters (495). P2X₃ receptors with the corresponding mutation do not express functional currents (364).

As for the COOH terminus, it is known that the splice variant of the rat P2X₂ receptor with a shortened COOH terminus (P2X_2b; missing the 69 amino acids from Val-370 to Gln-438 inclusive) shows a rather faster current decay (time constant ~24 s) than the wild-type receptor (time constant ~111 s) (rP2X_2a) (40, 418, 421). This difference, some fourfold, is not seen for the human receptors (292). The additional amino acids found in P2X_2a compared with rP2X_2b begin with Val-370; the last hydrophobic acid of the second membrane-spanning domain is Leu-353. The rat P2X₂ receptor truncated so as to end at Val-370 desensitizes with intermediate time constant when expressed in oocytes (~60 s; Ref. 421). However, the valine is critical because the receptor truncated at Lys-369 desensitizes very much faster (<1 s). Smith et al. (421) identified other residues in the segment of the P2X receptor beginning with Val-370 (Val-Arg-Thr-Pro-Lys-His-Pro in P2X_2a) as being important in desensitization. This is generally consistent with results from Koshimizu and colleagues (255-257) using whole cell calcium measurements as the assay for P2X receptor activation. They studied the changes in intracellular calcium elicited by ATP in GT1 cells expressing P2X₂ receptors and found that positively charged residues in this segment played a role in determining the kinetics of desensitization. Zhou et al. (522) found that certain substitutions at Asp-349, near the inner border of the second transmembrane domain, can also accelerate desensitization. It has been suggested that its negatively charged side chain might interact with the positive charges following Val-370 to stabilize a long-lived channel open state (256, 257, 522). One might equally speculate that an attached phosphate group at Thr-18 interacts with these positive charges.

The role of Ser-431 has also been studied (66); this is situated within the region that is spliced out in the P2X_2b form. The residue is situated at a protein kinase A consensus site, and introduction of the catalytic subunit of protein kinase A into the cytoplasm of HEK293 cells expressing the P2X₂ receptor led to an inhibition of the ATP-evoked currents. The effect was not seen in the S431C receptor. The inhibition was associated with an increased rate of desensitization. In the experiments of Werner et al. (495) (see sect. IVA4), chimeras were made between the P2X₁ and P2X₂ subunits. To make the P2X₂ receptor desensitize as rapidly as the P2X₁ receptor, it was necessary to provide it with both segments 14-47 and 332-365 of the P2X₁ receptor. These sequences include Thr-18 (in P2X₁ and P2X₂), but they do not include Lys-369 (P2X₂, corresponds to Lys-370 in P2X₁).

The calcium dependence of the decline in the current during the application of ATP was studied by Ding and Sachs (99). In whole cell recording mode, currents decline almost linearly with time; they reach half their initial amplitude in ~2 min. This decline was not seen in calcium-free external solution. In outside-out patches, currents at P2X₂ receptors decline much more rapidly than in whole cell configuration; with normal extracellular calcium (1 mM) this decline occurs within tens of milliseconds (99, 521). This basic observation implies that the decline of the current is prevented in the whole cell configuration because of the presence of some intracellular modulator, which is lost slowly in the whole cell recording but lost rapidly in outside-out patches (99). On the other hand, it is extracellular calcium that plays the key role in the decline of the current. Ding and Sachs (99) term this decline inactivation (i.e., inactivation by calcium) rather than desensitization (which may imply involvement of only the receptor protein and the ligand ATP). In the promotion of inactivation, calcium is better than magnesium, barium, and manganese (EC₅₀ values are respectively 1, 2, 3, and 5 mM). The maximum rate of decline of the ATP-induced current, observed with 2.5 mM calcium, is 40 s¹ (corresponding to a time constant of 25 ms). The decline of the current (inactivation) is steeply dependent on the ATP concentration (EC₅₀ 19 µM, Hill coefficient 2.8), the calcium concentration (EC₅₀ 1.3 mM, Hill coefficient 4.0), and membrane potential (inactivation was faster with hyperpolarization, changing e-fold for 26 mV in potential) (99).

In summary, extracellular divalent cations have (at least) two distinct actions on the homomeric P2X₂ receptor. First, they block the open channel; in this case the EC₅₀ for calcium is ~5 mM, the order of effectiveness is Mn > Mg > Ca > Ba, and the results fit well to a single binding site. Second, they reduce the probability of a channel being open; in this case they bind to the liganded channel, the EC₅₀ for calcium is about 1.3 mM, the order of effectiveness is Ca > Mg > Ba > Mn, and the results are best fit by the binding of four Ca ions.

ATP currents increase in size with repeated applications in the case of hippocampal neurons expressing heterologous P2X₂ receptors. Khakh et al. (235) used Sindbis virus to infect neonatal hippocampal neurons in culture with a P2X₂-GFP construct. The cells responded to ATP with currents typical of P2X₂ receptors in other expression systems, but these currents doubled in amplitude when ATP was applied repetitively at 1 Hz. This increase was correlated with a redistribution of the receptor, as visualized by its GFP tag, over distances of several micrometers into varicose "hot spots." The redistribution was not seen with the T18A mutant receptor, suggesting that it might result from activity-dependent phosphorylation by protein kinase C.

When oocytes are injected with RNAs encoding P2X₂ receptors, and also the ₃- and ₄-subunits of nicotinic receptors, they show responses to both ATP and acetylcholine; these can be selectively antagonized with appropriate receptor blockers (237). However, with concomitant application of both agonists, the resultant current is less than the expected sum of the two independent currents. A similar observation had been made previously in several native cells (see sect. VE5). Such occlusion of the currents indicates an interaction between the two receptors. It was more marked when the channels were expressed at high levels and was not seen in oocytes injected with lower amounts of RNAs. This might suggest the need to generate critical amounts of a signaling molecule for the interaction to occur.

P2X₃ receptor subunit cDNAs were isolated from rat dorsal root ganglion cDNA libraries (60, 274), from a human heart cDNA library (147), and from a zebrafish library (32, 108).

The mimicry of ATP by meATP makes these receptors similar to P2X₁ and distinct from the other homomeric forms. 2-Methylthio-ATP (2-MeSATP) is as potent as (274) or more potent than (60, 147) ATP at P2X₃ receptors. Diadenosine pentaphosphate (Ap5A) is a full agonist, as measured by calcium fluxes in transfected 1321N1 human astrocytoma cells (25). The actions of ATP are potentiated by zinc (rat P2X₃: EC₅₀ ~10 µM) (501) and cibacron blue (human P2X₃: EC₅₀ 3 µM). Diadenosine triphosphate (Ap3A) is more potent than at P2X₁ receptors (499), whereas meATP is strikingly less so (60, 147). The zebrafish receptor is notably less sensitive to meATP than the rat and human counterparts (32, 108).

The antagonists suramin, PPADS, and TNP-ATP do not readily distinguish between P2X₁ and P2X₃ receptors, but NF023 is ~20 times less effective at P2X₃ than P2X₁ receptors. Protons inhibit currents at rat P2X₃ receptors, with an EC₅₀ of ~1 µM (pK_a 6). The P2X₃ receptor is remarkably insensitive to block by extracellular calcium (EC₅₀ ~90 mM) (482).

Rat P2X₃ receptors are cation-selective channels (274). The relative permeability of calcium to sodium (P_Ca/P_Na) is ~1.2 (in 5 mM calcium, NMDG solution) (482).

At low concentrations (30-300 nM), ATP elicits currents that are sustained for several seconds, but with higher concentrations the currents show prominent desensitization (Fig. 5). The desensitization occurs with a time constant of <100 ms at concentrations of 30 µM ATP (274). As for P2X₁ receptors, recovery from this desensitization is very slow, and reproducible responses to ATP (or meATP) can only be obtained when applications are separated by at least 15 min.

Cook, McCleskey, and colleagues (82, 83) found that recovery from desensitization can be greatly accelerated by increasing the extracellular calcium concentration. The time constant for recovery was 7 min at 1 mM calcium and 3.5 min at 10 mM; gadolinium had a similar accelerating effect at 10 µM. This effect of calcium was related to the period of time for which the concentration was elevated and occurred whether or not the calcium concentration was increased at the same time that ATP was applied. Indeed, an elevation of calcium concentration was effective to accelerate recovery from desensitization even when it was applied several minutes before the next application of ATP. This suggests that calcium and gadolinium can bind to a desensitized form of the channel and accelerate its recovery into a nondesensitized, closed state.

In certain sensory neurons, sympathetic ganglion cells, and brain neurons, the action of ATP is mimicked by meATP, but there is no desensitization in the millisecond time scale (445). This type of response is mimicked by coexpression of P2X₂ and P2X₃ receptors (274). Direct association between the subunits has been shown by coimmunoprecipitation after expression in insect cells using baculovirus expression (374, 462).