P2X5 Subunit Assembly Requires Scaffolding by the Second Transmembrane Domain and a Conserved Aspartate*

Functional homomeric and heteromeric ATP-gated P2X receptor channels have been shown to display a characteristic trimeric architecture. Of the seven different isoforms (designated P2X1-P2X7), P2X5 occurs in humans primarily as a non-functional variant lacking the C-terminal end of the ectodomain and the outer half of the second transmembrane domain. We show that this truncated variant, which results from the splice-skipping of exon 10, is prone to subunit aggregation because the residual transmembrane domain 2 is too short to insert into the membrane. Alleviation of the negative hydrophobic mismatch by the addition of a stretch of moderately hydrophobic residues enabled formation of a second membrane-spanning domain and strictly parallel homotrimerization. Systematic mutagenesis identified only one transmembrane domain 2 residue, Asp355, which supported homotrimerization in a side chain-specific manner. Our results indicate that transmembrane domain 2 formation contributes 2-fold to hP2X5 homotrimerization by tethering the end of the ectodomain to the membrane, thereby topologically restricting conformational mobility, and by intramembrane positioning of Asp355. While transmembrane domain 2 appears to favor assembly by enabling productive subunit interactions in the ectodomain, Asp355 seems to assist by simultaneously driving intramembrane helix interactions. Overall, these results indicate a complex interplay between topology, helix-helix interactions, and oligomerization to achieve a correctly folded structure.

P2X receptors comprise a family of ligand-gated ion channels that are activated by extracellular ATP and mediate rapid signaling in a large variety of cells including neurons, smooth and cardiac muscles, epithelia, and lymphocytes (1). Seven subunit isoforms, designated P2X 1 -P2X 7, have been identified and share a common topology with cytosolic N-and C-terminal domains and two membrane-spanning regions (transmembrane (TM) 1 and TM2) 3 connected by a large N-glycosylated ectodomain comprising 10 conserved cysteine residues. Like all other known ligand-gated ion channels, P2X receptors are oligomers of identical (homomeric) or homologous (heteromeric) subunits. Biochemical (2)(3)(4), biophysical (5), and functional studies (6) revealed that a trimeric architecture driven by noncovalent intermolecular interactions is a structural hallmark of functional homomeric and heteromeric P2X receptors. The three subunits are thought to arrange circularly, contributing one or both transmembrane domains to form a central cationconducting pore.
The assembly pathway leading to this trimeric structure is not well characterized. The question of which molecular determinants lead to the assembly of subunits into P2X receptors has so far been addressed by examining the ability of deletion mutants and chimeric constructs to associate with full-length P2X subunits in a co-immunoprecipitation assay (7). Neither the N-nor C-terminal cytoplasmic domain was found to be important for assembly. However, a deletion mutant terminating 25 amino acids before the start of the TM2 domain was unable to interact with either of the wild type subunits or with itself. This suggested that either TM2 or a region immediately upstream of TM2 carried a critical determinant of specific subunit-subunit interactions, although the possibility that the lack of co-assembly was the result of an inadequate secondary or tertiary structure was not ruled out (7).
To further examine the role of both TM2 and the pre-TM2 region in P2X receptor assembly, we utilized the P2X 5 isoform, which occurs in humans predominantly as a natural deletion mutant (designated hP2X 5 ⌬328 -349 ) and lacks much of the TM2 and pre-TM2 regions as a result of the splicing-out of exon 10 (8). The hP2X 5 ⌬328 -349 cDNA does not encode functional channels, but a receptor chimera consisting of the N-terminal and C-terminal halves of the human and the rat P2X 5 subunits, respectively, has been expressed in Xenopus laevis oocytes and is functional (8). A full-length hP2X 5 subunit generated by incorporating a sequence corresponding to exon 10 (as identified by a TBLASTN search of human genomic DNA) was expressed well in human embryonic kidney 293 cells and provided robust ATP-dependent currents (9). In this report we sequentially generated a full-length hP2X 5 subunit cDNA by progressively and repeatedly inserting codons for 4 -8 amino acids of the spliced-out exon 10 and then determined the assembly state of the resultant constructs by blue native PAGE. We also challenged homotrimerization of the full-length P2X 5 subunit by block replacements of amino acids throughout the pre-TM2 and TM2 regions to test for the presence of residues important for P2X 5 subunit-subunit recognition. Our results are consistent with the view that TM2 supports homotrimerization in a relatively sequence-unspecific manner by tethering the C-terminal end of the large ectodomain to the membrane; thus, by defining the overall shape of the ectodomain, TM2 spatially constrains folding events in a productive way. We detected a specific role in assembly recognition for only one TM2 residue, Asp 355 , which is fully conserved among P2X family members and may stabilize helix-helix interactions by satisfying hydrogen bonding groups in the membrane.

MATERIALS AND METHODS
P2X cDNA Constructs-To indicate the species origin, P2X subunit names are preceded by "h" or "r" for human or rat, respectively. Constructs from previous work include His-rP2X 1 (2) and His-rP2X 5 (4), which encode the rat P2X 1 and P2X 5 subunit with N-terminal His tags. The original rP2X 5 clone was kindly provided by Dr. Florentina Soto (10). A cDNA encoding the hP2X 5 subunit (GenBank TM accession number AAC51931 (8)) was isolated by PCR from a human brain cDNA library (Invitrogen) using sequence-specific primer pairs (forward, aaagaattcCA-TGGGGCAGGCGGGCTGCAA; reverse, aaagaattcGAGGCAA-TTCACGTGCTCCTGTGGGGCT; EcoRI cloning sites are underlined) subcloned into the oocyte expression vector pNKS2 (11) and entirely sequenced. Codon insertions and replacement mutations were introduced by QuikChange site-directed mutagenesis (Stratagene, La Jolla, CA).
P2X Receptor Expression in X. laevis Oocytes-Defolliculated X. laevis oocytes injected with capped cRNAs as described previously (12) were kept at 19°C in sterile frog Ringer's solution (ORi: 90 mM NaCl, 1 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , and 10 mM Hepes, pH 7.4) supplemented with 50 g/ml gentamycin. Two days after cRNA injection, ATP responses were measured by two-electrode voltage-clamp recording at a holding potential of Ϫ60 mV as described previously (13). Capping the N-terminal end of the rP2X 5 subunit with a His tag for one-step affinity purification had virtually no effect on the electrophysiological phenotype of the corresponding receptor in X. laevis oocytes (results not shown).
P2X 5 receptor constructs at the plasma membrane were selectively labeled by incubating oocytes 2 days after cRNA injection with 125 I-labeled sulfosuccinimidyl-3-(4-hydroxyphenyl)propionate, a membrane-impermeant derivative of the Bolton-Hunter reagent (15) exactly as described previously (2,4). Proteins were purified from digitonin extracts of the oocytes by Ni 2ϩ -NTA-agarose chromatography as detailed above.
Trypsin Resistance Assay-P2X 5 receptor constructs, purified by non-denaturing Ni 2ϩ affinity chromatography, were treated with 10 -1000 g/ml of bovine trypsin (Sigma-Aldrich) for 15 min on ice in buffer 1. Reactions were terminated by the addition of a 5-fold excess of soybean trypsin inhibitor. Digested samples were then analyzed by Tricine-SDS-PAGE with PhosphorImager scanning (see below).
Blue Native PAGE and SDS-PAGE-Blue native PAGE (16,17) was carried out as described (2) using gradient gels (4 -20% acrylamide). For partial dissociation of natively purified P2X receptors into lower-order complexes down to monomers, samples were treated for 1 h at 37°C with 0.1% SDS or 4 -8 M urea as indicated. For SDS-PAGE or Tricine-SDS-PAGE (18), proteins were supplemented with the appropriate SDS-PAGE sample buffer containing 20 mM dithiothreitol followed by heating to 56°C for 15 min and electrophoresed in parallel with 14 C-labeled molecular mass markers (Rainbow TM , Amersham Biosciences). In some experiments samples were treated before SDS-PAGE for 2 h at 37°C with either endoglycosidase H or peptide:N-glycosidase F (PNGase F) (New England Biolabs, Frankfurt, Germany) in the presence of 1% Nonidet P-40 to decrease inactivation of PNGase F. Both SDS-PAGE gels and blue native PAGE gels were fixed, dried, exposed to a Phosphor-Imager screen, and scanned using a Storm 820 Phosphor-Imager (Amersham Biosciences). Individual bands were quantified with the ImageQuant software.

Functional Rat P2X 5 Receptors Possess a Homotrimeric
Architecture-All P2X 5 constructs in this study were N-terminal-tagged with a hexahistidyl sequence to allow for protein purification by metal affinity chromatography after expression in X. laevis oocytes. Oocytes expressing WT rP2X 5 subunits or His-rP2X 5 subunits responded to ATP with a non-desensitizing inward current of similar shape and magnitude. A typical current trace in response to ATP from His-rP2X 5 subunit-expressing oocytes is shown in Fig. 1A. The shape of the current trace was notably changed when rP2X 5 receptors were repeatedly challenged with ATP. We observed the development of a slowly activating inward current, which was most likely mediated by secondary activation of oocyte endogenous channels since it was greatly reduced by flufenamic acid, an inhibitor of Ca 2ϩ -activated Cl Ϫ channels (19). Flufenamic acid did not affect the shape of the current component attributable to the ATP-activated rP2X 5 channel (results not shown).
We have previously demonstrated that the blue native PAGE technique has the capacity to correctly display the quaternary state of receptor channels and transporters. These include the known pentameric structure of Cys-loop receptors (2, 20 -22) and the trimeric structure of glutamate transporters (23), which has also been visualized by x-ray crystallography (24). Like other oocyteexpressed P2X receptors except for hP2X 6 (4), the plasma membrane-bound rP2X 5 receptor migrated on the blue native PAGE gel as a distinct protein band (Fig. 1B, lane 1). Treatment with urea (lanes 2 and 3) resulted in a ladder-like pattern of three bands, each separated by the approximate mass of an rP2X 5 monomer. Using this ladder as a mass marker, a trimeric state can be assigned to the non-denatured rP2X 5 receptor.
Plasma membrane-bound rP2X 5 subunits bear three N-glycans, which add ϳ10 kDa of carbohydrate to the protein core (Fig. 1C). The mass calculated from the protein sequence is 52 kDa including the His tag, which is in good agreement with the experimental mass of ϳ54 kDa of the fully deglycosylated protein. The different glycosylated states obtained by combined deglycosylation with endoglycosidase H and submaximally effective concentrations of peptide:N-glycosidase F indicate that all three predicted N-glycosylation sites ( 77 NTT, 157 NST, 202 NFS) were occupied, one with a high mannose-type glycan and two with complex-type glycans (Fig. 1D).
Human P2X 5 Subunits Lacking Exon 10 Are Trimerizationdefective-Using gene-specific primer pairs, we PCR-amplified an hP2X 5 cDNA from a human brain cDNA library. The deduced amino acid sequence was identical to a previously reported hP2X 5 splice variant that lacked exon 10 (8). We refer to this splice variant as hP2X 5 ⌬328 -349 to indicate that codons 328 -349 of the full-length P2X 5 subunit are missing. The loca-tion of the exon 10-specified sequence relative to the membrane is schematically illustrated in Fig. 2A.
Consistent with previous studies (8), expression of hP2X 5 ⌬328 -349 in Xenopus oocytes did not lead to the formation of an ATP-gated cation channel. Because trimerization is essential for P2X receptor function, we performed a blue native PAGE analysis to investigate the assembly status of this splice variant. Fig. 2B, lane 1, shows that the [ 35 S]methionine-labeled His-hP2X 5 ⌬328 -349 protein was poorly expressed and migrated on the blue native PAGE gel as an amorphous mass of protein (indicative of aggregates) rather than as a distinct band. In contrast, full-length His-rP2X 5 subunits migrated largely as homotrimers (Fig. 2B, lanes 10 and 11). We conclude that the lack of exon 10 results in a severe assembly defect that prevents the formation of functional homotrimeric hP2X 5 receptor channels at the level of subunit assembly in the endoplasmic reticulum.
Not All Exon 10-Encoded Residues Are Necessary for the Trimerization and Plasma Membrane Localization of hP2X 5 Subunits-To examine which of the 22 amino acids encoded by exon 10 were necessary for proper trimer formation and ion channel function, we progressively rebuilt exon 10 by inserting codons for 4 -8 consecutive amino acids at a time, thus generating by stages a full-length hP2X 5 subunit cDNA (Fig. 2B). The inserted codons correspond to those previously identified by a TBLAST search of human genomic DNA corresponding to exon 10 (9). Expression of the various hP2X 5 constructs in X. laevis oocytes revealed that insertion of a minimum of 11 amino acids (residues 328 -338; construct 3) from the total of 22 amino acids encoded by exon 10 was required to mediate significant homotrimer formation.
Trimerization was associated with increased protein stability and with plasma membrane localization of properly assembled trimers ( Fig. 2D) but not with the formation of a functional receptor channel (Fig. 2C). ATP-gated inward currents could be elicited only from oocytes expressing the full-length His-hP2X 5 subunit (Fig. 2C, lane 8). The trimerization-defective constructs His-hP2X 5 ⌬328 -349 and His-hP2X 5 ⌬335-349 were neither functional (Fig. 2C, lanes 1 and 2) nor exported to the cell surface (Fig. 2D, lane 2). Accordingly, once the span of inserted residues specified by exon 10 is of adequate length to allow for trimer formation, the resultant protein complex must have a sufficiently native-like conformation to be no longer recognized and retained by the endoplasmic reticulum quality control system, which retains malfolded proteins.
Insertion of Alanines, but Not Serines, Supports Trimerization as Efficiently as the Insertion of Genuine Exon 10 Residues-Insertion of seven residues, 335 PTIINVG 341 , into the His-hP2X 5 ⌬335-349 construct resulted in a polypeptide that efficiently assembled into homotrimers (cf. Fig. 2B, lane 4). To determine whether this effect was sequence specific, we inserted a stretch of seven alanine or serine residues instead of the genuine 335 PTIINVG 341 sequence and examined whether any of these substitutions was able to support oligomerization (Fig. 2B). Strikingly, the polyalanine stretch (lane 5) supported trimerization equally well as the 335 PTIINVG 341 sequence, whereas the polyserine stretch (lane 6) did not. Trimerization Parallels the Formation of a Second Membrane-spanning Segment-A possible explanation of these results is that alanine but not serine residues are capable of complementing the missing outer portion of TM2, thus allowing for membrane anchoring of the C-terminal end of the ectodomain. To address this possibility, we tried to exploit the fact that the hP2X 5 subunit carries an N-glycosylation sequence, 430 NGS (numbering refers to the full-length hP2X 5 subunit), which is normally topologically inaccessible because it resides on the C-terminal endodomain (cf. topology model in Fig. 2A). This site may be expected to be N-glycosylated when TM2 is excluded from the lipid bilayer, thus forcing the C-terminal tail to remain aberrantly located in the endoplasmic reticulum lumen. Deglycosylation analysis, however, provided no evidence for hyperglycosylation of the hP2X 5 ⌬328 -349 polypeptide, which could be shown to carry two N-glycans (results not shown); this is fully consistent with the occurrence of two N-glycosylation sites in the ectodomain.
430 NGS may remain unused as an acceptor site because of its short distance of only 15 residues to the C-terminal end (25). Therefore, we mutated the normally cytoplasmic Glu 380 to Asn, thus generating an artificial glycosylation reporter sequence ( 380 NDS; numbering refers to the full-length hP2X 5 subunit) located at a more distal position of 65 residues from the C-terminal end. Fully efficient glycosylation has been shown to occur if 60 or greater residues separate the glycosylation acceptor site from the C terminus (25). Indeed, the corresponding hP2X 5 ⌬328 -349 mutant migrated with a 2-3-kDa higher mass (Fig. 3A, lane 2) than the parent construct (lane 1), demonstrating glycosylation of Asn 380 and, hence, a luminal orientation of most likely the entire C-terminal tail (cf. model in Fig. 3B). This result can best be reconciled with the exclusion of the incomplete TM2 sequence Gly 350 -Ile 362 from the membrane as a result of a negative hydrophobic mismatch that occurs when the length of a hydrophobic transmembrane segment is too short to span the thickness of the hydrophobic bilayer (26). This is supported by the observation that the identical acceptor site remained unused when engineered into the trimerization-competent hP2X 5 ⌬339 -349 construct (Fig. 3A, lane 3). Evidently, pre-TM2 residues Ala 328 -Ile 338 can combine with the TM2-authentic residues Gly 350 -Ile 362 to form a hydrophobic domain of sufficient length to be integrated into the membrane. The concurrency of correct transmembrane threading and appearance of homotrimers on the blue native PAGE gel suggests an important role for TM2 in P2X 5 subunit homotrimerization.
Hydrophobicity of TM2 and of TM2-embedded Aspartate 355 Contributes to Stable hP2X 5 Homotrimer Formation-To further examine the contribution of amino acid side chains specified by exon 10 and more distal sequences to trimerization in the context of the full-length hP2X 5 , we carried out a systematic mutagenesis. Five consecutive amino acids were replaced by single alanines throughout the sequence Lys 330 -Leu 361 (except for one double alanine mutation, 328 AGKF 331 to 328 AGAA 331 ) of the full-length His-hP2X 5 construct, comprising all exon 10-specified residues plus the subsequent 12 residues of the inner half of TM2 (Fig. 4). Overall, alanine block mutations were well tolerated. Only one of the mutants, 330 AA 331 -hP2X 5 , was more likely to aggregate than to homotrimerize (Fig. 4A,  lane 3). The ATP-gated inward current amplitude was similarly reduced to ϳone-third of that mediated by the WT hP2X 5 receptor (Fig. 4B, lane 3). Cell surface radioiodination verified that the 330 AA 331 -hP2X 5 mutant appeared exclusively as properly assembled trimers at the surface, albeit at lower levels than the WT hP2X 5 receptor (Fig. 4C). Strikingly, none of the pentaalanine block mutations covering the entire span of TM2 affected either subunit homotrimerization (Fig. 4A) or export to the plasma membrane (Fig. 4, B and C). Robust inward cur-   rents in response to ATP were mediated by most of the mutants. Only penta-alanine substitutions in the inner third of TM2 (residues 352-361) abolished channel function, but these mutants also appeared as homotrimers at the cell surface (Fig.  4, B and C).
To examine if trimer formation is also possible when bulkier residues than alanine are substituted, we performed a similar mutagenesis of TM2 with penta-leucine and penta-glutamine substitutions. Results obtained by leucine replacements of residues 337-361 resembled those obtained by equivalent alanine replacements except in one respect; the 352 FFCDL mutant trimerized only weakly (Fig. 5A, lane 5). All the other leucine replacement mutants assembled into homotrimers as efficiently (lanes 2-4 and 6) as the parent hP2X 5 subunit (lane 1). Even when 15 consecutive residues constituting the N-terminal two-thirds of TM2 (residues 337-351) were simultaneously replaced by leucines, homotrimers formed and appeared at the cell surface (Fig. 5B). However, inclusion of the 352 FFCDL sequence to replace 20 consecutive residues by leucines virtually abolished trimer formation (results not shown). Responses to ATP could be elicited from mutants carrying leucine replacements at the membrane entry (residues 337-341) or exit sites of TM2 (residues 357-361), but penta-leucine replacements in the center of TM2 resulted in nonfunctional trimers (Fig. 5A, lanes 1-6,  lower panels).
Penta-glutamine block replacements of the same TM2 residues resulted in all cases in severe assembly defects, as judged by the inability of these mutants to migrate as defined oligomers on blue native PAGE gels (Fig. 5A, lanes 9 -13). Moreover, all the penta-glutamine replacement mutants were nonfunctional in terms of mediating ATP-gated inward currents (lanes 9 -13, lower panels). Introducing the same glycosylation reporter sequence 380 NDS as described above (cf. Fig. 3B) into two of the glutamine mutants resulted in a 2-3-kDa increase in molecular mass (Fig. 5C). This mass shift indicates usage of Asn 380 and, accordingly, exclusion of the penta-glutamine block-containing TM2 sequence from the membrane.
To address the impaired trimerization of the 352 FFCDL leucine replacement mutant in more detail, single, double, and triple mutants were generated. All of the mutants that included the D355L mutation trimerized very weakly even if singly mutated (Fig. 6A, lanes 2-4). In contrast, leucine replacement of 352 FFC alone, thus excluding Asp 355 , did not lower the trimerization efficiency (Fig. 6A, lane 5). Quantification of radioactivity showed that about 10-fold fewer trimers were formed from the D355L-containing mutants than from the WT hP2X 5 subunit or from the 352 FFC leucine replacement mutant. In addition, a subset of the D355L-containing mutants  DECEMBER 22, 2006 • VOLUME 281 • NUMBER 51 migrated as dimers. This can best be seen in the quantitative scans of the various protein lanes (Fig. 6B). None of these mutants was capable of mediating ATP-gated inward current in oocytes, as assessed by two-electrode voltage-clamp measurements.

Role of TM2 and Asp 355 in P2X 5 Homotrimerization
The Entire TM2 Domain Is Required for Subunit Trimerization-To further examine the role of TM2 in subunit trimerization, we generated truncated forms of the full-length hP2X 5 subunit by inserting premature stop codons at C-terminal positions which excluded or included TM2 or portions thereof. Blue native PAGE showed that trimerization occurred only when the expressed polypeptide comprised the entire TM2, capped by three basic residues, 363 KKR 365 , at the extreme C-terminal end (Fig. 7A). Analysis of equivalent samples under denaturing conditions on a reducing SDS-PAGE gel verified that all truncation mutants were expressed with the expected molecular masses (Fig. 7B).
Assembly-competent TM2 Mutants Are as Tightly Folded as the WT hP2X 5 Receptor-The WT hP2X 5 subunit has 33 lysine and 24 arginine residues, potential cleavage sites for trypsin. The 61-kDa subunits of the non-denatured trimeric WT hP2X 5 receptor were cleaved at the lowest trypsin concentration to yield two major products of 57 and 51 kDa (Fig. 8A, lane 2).
These fragments arose evidently from proteolytic removal of the cytoplasmic N and/or C-terminal tails, which have calculated masses of 3.1 and 9.0 kDa, respectively. The 51-kDa fragment was still prominent at a 10-fold higher trypsin concentration (lane 3). At a 100-fold higher trypsin concentration (lane 4), extensive digestion occurred, with a 17-kDa band being the predominant species. In contrast, two mutants judged as completely assembly-incompetent by blue native PAGE (cf. Figs. 5A and 7A) were degraded almost to completion even at the lowest trypsin concentration used (Fig. 8A, lanes  5-12). The site of initial proteolytic scission of a native protein is mainly determined by its higher order structure and much less by its primary sequence (27). We, therefore, deduced from these results that particularly the ectodomain of the WT hP2X 5 receptor must be tightly folded to resist full digestion by a trypsin concentration sufficient to completely digest aggregated hP2X 5 polypeptides. This result further suggests that hP2X 5 polypeptides contained in aggregates exist in a substantially unfolded state.
Next, we systematically tested the trypsin susceptibility of five leucine replacement mutants that cover the entire TM2 region and that all assembled as trimers, although with distinct differences in efficiency. Three of the mutants covering Ser 342 -Leu 356 were non-functional. Under identical digestion conditions, identical patterns of tryptic fragments were produced from the WT hP2X 5 receptor (Fig. 8B, lanes  1-4) and from both the functional (lanes 5-8 and 21-24) and non-functional mutants (lanes 9 -20).
The initial proteolytic events are the most critical for unraveling structural features, since proteins fragments are more easily degraded as a result of their enhanced flexibility than the intact protein (27). Therefore, we quantified trypsin resistance of TM2 mutants by relating the amount of the two major products formed at the lowest trypsin concentration to the amount of the non-trypsin-treated intact polypeptide. Non-functional leucine replacement mutants were as resistant to trypsin digestion as either the WT hP2X 5 receptor or the functional mutants (Fig. 8C). The TM2 mutants that contained a D355L exchange, although exhibiting ϳ10 -15% trypsin resistance, were proteolyzed to a markedly greater extent than the other mutants (Fig.  8, B, lanes 17-20, and C). As shown above, the mutants harboring D355L existed mostly as aggregates and less as trimers (cf. Fig. 6). Accordingly, the intense non-trypsin-treated polypep-  tide band in the SDS-PAGE gel (Fig. 8B, lane 17) originates predominantly from aggregates and much less from trimers. Because trypsin treatment rapidly degrades hP2X 5 mutants contained in aggregates, the weak bands that are left (lane 18) must originate from the small fraction of trimers, which apparently resisted degradation.
The important contribution of Asp 355 to efficient trimerization and, accordingly, to overall trypsin resistance is also apparent from a leucine block mutant that explicitly retained Asp 355 (Fig. 8B, lanes 25-28). The strong increase in trimer formation (cf. Fig. 6) was accompanied by an equivalent increase in trypsin resistance (Fig. 8C).
To further strengthen our view that trimers represent the trypsin-resistant entity, we plotted the amount of trimers (nor-malized to monomers) against trypsin resistance (Fig. 8D). Data points originating from functional and non-functional mutants are both located on one straight line. Linear regression analysis yielded a significant correlation (r ϭ 0.76, p Ͻ 0.0001), indicating that trypsin resistance refers indeed to the fraction of P2X 5 polypeptides that are assembled as trimers. The close relationship suggests further that the overall folding of the WT hP2X 5 receptor and all the TM2 mutants studied must be very similar if not identical.

DISCUSSION
P2X subunits trimerize cotranslationally in the endoplasmic reticulum and reside in intracellular compartments and in the plasma membrane as permanently assembled trimeric receptor complexes (2) ("obligomers" (28)). Here we investigated the contribution of residues encoded by exon 10 and of the subsequent residues in TM2 to hP2X 5 subunit trimerization. We demonstrated that the hP2X 5 ⌬328 -349 splice variant forms aggregates instead of trimers, indicating that a severe defect in assembly is one reason for the known inability of this polypeptide to form functional ATP-gated receptors (8,9). Consistent with the view that the absence of a defined trimeric state on the blue native PAGE gel in general reflects a non-native conformation of a P2X receptor, aggregated hP2X 5 proteins, including the hP2X 5 ⌬328 -349 splice variant, were entirely retained in the endoplasmic reticulum by the quality control system. Moreover, structural probing by limited proteolysis revealed an extreme trypsin susceptibility of aggregates that is consistent with the view that aggregates consist of essentially unfolded polypeptides.
In contrast, trimerization-competent constructs, as judged by blue native PAGE, exhibited substantial trypsin resistance in the native state and were all exported to the cell surface. The virtually identical degree of resistance to trypsin digestion of the WT hP2X 5 receptor and the assembly-competent functional and non-functional mutants indicates that they all share a similar tightly folded structure. The presence of a trimeric state was not necessarily equivalent with an electrophysiologically functional receptor, as indicated by trimerization-competent hP2X 5 constructs that incorporated only 11 or 14 residues A, trimerization-incompetent P2X 5 receptor mutants are highly trypsin-sensitive. B, trimerization-competent P2X 5 receptor mutants exhibit trypsin resistance comparable with the WT hP2X 5 receptor. C, quantification of trypsin resistance. The trypsin resistance of the indicated P2X 5 receptor constructs was expressed as the percentage of radioactivity incorporated in the two bands a and b after 0.01 mg/ml trypsin treatment relative to that detectable in the corresponding intact P2X 5 polypeptide of the non-trypsin-treated sample. Data are from B except for the additional D355L-hP2X 5 single mutant, which was analyzed on a separate gel not shown. Filled columns, D355L-containing mutants. D, relationship between trimerization efficiency and trypsin resistance. Trypsin resistance data are from C. The relative amount of trimers was calculated for equivalent samples by normalizing the radioactivity incorporated into trimers as resolved on blue native PAGE gels to the radioactivity of the corresponding monomer resolved by Tricine SDS-PAGE (from B). Data of the D355L-hP2X 5 mutant not included in B are also shown. In C and D, error bars indicate S.D. values of duplicate determinations; where absent, single determinations are shown. from the total of 22 exon 10-encoded residues. These polypeptides appeared at the cell surface exclusively as fully assembled trimers but were unable to mediate an ATP-gated current.
TM2 Plays an Indispensable Scaffolding Role in P2X 5 Subunit Homotrimerization-An involvement of TM2 in the assembly of P2X receptors has previously been inferred from co-immunoprecipitation assays (7). A P2X 2 deletion mutant terminating 25 amino acids N-terminal to TM2 did not co-precipitate with other subunits or with itself, suggesting that the missing region of the protein participates in assembly. In contrast, a truncation located C-terminal to TM2 at position Tyr 362 of the P2X 2 subunit (corresponding to Tyr 368 of P2X 5 ) enabled co-isolation. These and additional data obtained by studies of chimeras led to the conclusion that TM2 is a critical determinant for productive P2X subunit assembly (7).
Our present data offer a substantially extended perspective on this issue by showing that TM2 contributes to assembly first of all merely as a hydrophobic membrane anchor rather than by providing specific subunit recognition information. This conclusion is supported by the following data. First, blocks of five consecutive amino acid residues could be replaced by alanines throughout the entire pre-TM2 and TM2 regions without impairing homotrimer formation. Second, the absence of the outer half of TM2 residues did not prevent efficient homotrimeric assembly if a sufficiently long alanine stretch was present in the pre-TM2 region in order to alleviate the hydrophobic mismatch, thus enabling formation of a second membranespanning segment. Third, all sequence manipulations that prevented the formation of a second membrane-spanning segment also led to subunit aggregation and accelerated protein degradation. This was true for the exon 10-lacking hP2X 5 ⌬328 -349 variant, for a series of pre-TM2 and TM2 mutants with original residues replaced by blocks of polar residues (serines or glutamines), and for C-terminal truncation mutants, which included all pre-TM2 and TM2 residues but lacked a positively charged cap, 363 KKR 365 , at their free C-terminal end. In summary, we note a close relationship between correct lipid bilayer insertion of a hydrophobic domain in N exo -C cyt orientation and proper P2X 5 subunit trimerization.
P2X 5 Subunit Assembly Is Not Driven by Specific Packing Interactions of TM2 Residues-Major driving forces for association of helices in membrane proteins are (i) van der Waals interactions such as those mediated by the membrane-spanning leucine zipper domain (29) or the abundant GXXXG motif of bitopic proteins (30) and (ii) hydrogen bonding between polar residues. TM domains of P2X receptors do not contain a leucine zipper-like motif, but small residues (Gly, Ala, Ser) are over-represented in the outer half of TM2 of all P2X subunits and may allow for tight van der Waals packing between neighboring helices. The outer TM2 half of the hP2X 5 subunit contains as many as 8 small residues (4 Gly, 3 Ala, 1 Ser) over a span of 11 residues. Of these, the sequence 340 VGSGVA 345 (GXXXA) is reminiscent of the common pattern defined by the GXXXG-containing assembly motif of glycophorin A; small residues are next to the larger side chains of the ␤-branched amino acids valine and isoleucine (31), enabling strong (SDSresistant) self-association by a ridge-into-groove mode of packing (32). However, in contrast to the exquisite specificity of TM-mediated dimerization of glycophorin A, which is destabilized significantly by small interfacial changes (33)(34)(35), homotrimerization of hP2X 5 subunits was entirely insensitive to such mutations. Even when 15 consecutive TM2 residues, Ile 337 -Ala 351 , including all eight small residues, were mutated to leucines, homotrimerization was unimpaired. Pure polyleucine-based coiled-coil helices mediate only relatively weak associations betweens TM domains in the absence of a polar residue (36). Thus, despite the fact that small residues occur abundantly in helix interfaces of membrane proteins and mediate tight helix-helix interactions (37), small residues in TM2 are not necessary for efficient P2X 5 subunit trimerization.
Intramembrane Hydrogen Bonding May Play a Role in Initiation of Homotrimerization-hP2X 5 homotrimer formation was remarkably tolerant to TM2 substitutions by small (alanine) and large (leucine) residues except in one position, Asp 355 , which represents the only charged TM2 residue. Recent evidence indicates that single strongly polar residues (Asp, Asn, Glu, Gln) in a transmembrane helix can be sufficient to drive self-association of transmembrane helices through the formation of side-chain/side-chain interhelical hydrogen bonds in the absence of detailed packing (36, 38 -40). The importance of Asp 355 and, accordingly, of hydrogen bonding as a hotspot in P2X 5 subunit assembly is limited insofar as that (i) the D355L mutation reduced, but did not abolish hP2X 5 subunit homotrimerization, and (ii) substitution of Asp 355 by alanine did not impair homotrimerization. To account for these observations, we propose that Asp 355 , by satisfying hydrogen bonding groups in the membrane, functions at an early stage of the assembly process (41) to drive intersubunit interactions to form higher order oligomers. A locally increased effective subunit concentration may provide an environment that promotes interactions between specific assembly interfaces, which are evidently located in the ectodomain. This model can explain the observation that Asp 355 increases the yield of homotrimers during biosynthesis but is not essential for the continued structural maintenance of the fully assembled homotrimer. Admittedly, the undisturbed assembly of the D355A mutant cannot be directly integrated in this model, and additional experiments are needed to assess the extent to which intramembrane hydrogen bonding contributes to hP2X 5 subunit homotrimerization.
Negative Hydrophobic Mismatch Results in hP2X 5 Subunit Aggregation-The inability of the splice variant hP2X 5 ⌬328 -349 to form a second membrane-spanning domain can be directly attributed to the absence of the exon 10-encoded pre-TM2 region and to the absence of the outer half of TM2. Using a helix translation of 1.5 Å per residue, the calculated hydrophobic length of the existing inner half of TM2, 350 GAFFCDLV-LIYLI 362 , is ϳ20 Å, which is clearly too short to span the 30-Å hydrophobic core of the membrane. Studies with synthetic hydrophobic peptides consisting of a polyleucine or polyleucine/alanine hydrophobic stretch have demonstrated that helices too short to span the phospholipid model membranes (referred to as negative hydrophobic mismatch) are excluded from the lipid bilayer. Because of the general insolubility of hydrophobic segments in the aqueous phase, peptides with a negative hydrophobic mismatch have a strong tendency to aggregate (42,43). Proteins such as bacteriorhodopsin (44,45) and Ca 2ϩ ATPase (46) behave in a similar fashion and aggregate when reconstituted in phospholipids with short chain fatty acids; the excluded hydrophobic helices are shielded by self-association from the energetically unfavorable contact with the aqueous solvent. By analogy, we suggest that the hydrophobic mismatch-induced exclusion of incomplete TM2 segments from the lipid bilayer accounts for the propensity of the TM2-defective mutants to aggregate.
The support of TM2 formation and, hence, subunit trimerization by a polyalanine sequence inserted in the exon 10-lacking splice variant deserves comment. Alanine exhibits a high propensity for ␣-helical conformations, but the hydrophobicity of lysine-flanked polyalanine peptides is just at the threshold required for transmembrane integration into pure lipid bilayers (47). However, introduction of a few leucines into a polyalanine scaffold is sufficient to stabilize an exclusively transmembrane orientation (48). Similarly, a stretch of 16 alanines and 5 leucines was determined as the threshold value of hydrophobicity required to function as a stop-transfer sequence and to anchor alkaline phosphatase, a water-soluble protein, in Escherichia coli membranes (49). We suggest that the highly hydrophobic inner half of TM2 is capable of compensating for the weak or absent hydrophobicity of inserted polyalanine blocks to keep the overall hydrophobicity above the threshold for stable membrane insertion.

P2X 5 Homotrimerization Is Further Supported by Residues That Anchor TM2 in the Cytoplasmic Membrane-Water
Interface-Truncation of the full-length P2X 5 subunit C-terminal to the entire TM2 region also resulted in aggregation as long as the C-terminal tail of the hydrophobic segment was not flanked by the naturally occurring residues 363 KKR 365 . Positive charges in the region flanking a transmembrane segment are known to represent topological determinants that tend to have cytoplasmic localizations (50). Moreover, flanking lysine residues are thought to electrostatically anchor the position of the edge of a TM helix in the acidic phospholipid head-group region of the membrane-water interface, thus securing a stable transmembrane orientation of the helix more or less perpendicular to the bilayer surface (51). The importance of flanking lysine residues as anchors for a stable transmembrane orientation is particularly obvious in experiments with synthetically designed transmembrane peptides, which also show that basic flanking residues disfavor extensive peptide aggregation (52). Interestingly, sufficiently long hydrophobic peptides that are flanked by lysine residues only at the C-terminal end have the peculiar feature of spontaneously adopting a transmembrane orientation (53). Overall, several mechanisms may account for the crucial role of the 363 KKR 365 sequence in the assembly of C-terminal-truncated P2X 5 subunits. However, within the context of the full-length P2X 5 subunit, the 363 KKR 365 sequence appears of limited importance for assembly. 4 TM2 Side-chain Chemistry and Channel Function-A detailed characterization of the electrophysiological properties of the various P2X 5 mutants was beyond the scope of the present study, and thus, only maximum current amplitudes elicited by a supersaturating concentration of ATP (1 mM) were recorded. hP2X 5 channel function was not abolished by pentaalanine substitutions in the outer half of TM2 (as summarized in the model in Fig. 9) even when the sole non-hydrophobic residue of this region, Ser 342 , was also replaced by alanine. Most of this tolerant region bears small residues including glycine, so their mutation to alanine (Gly 341 -Ala 351 ) caused only moderate changes in side-chain volume. Block substitutions by pentaleucines were less well accommodated and ablated function except at the outer and inner end of TM2 ( 337 IINVG 341 and 357 VLIYL 361 ), where bulky hydrophobic residues (Ile, Leu, Val) already prevailed in the WT sequence. Apparently, small side chains are important in the center of TM2. Of these, Gly 342 or Gly 344 has been suggested to constitute a part of the channel gate (54), similar to the glycine hinge of K ϩ channels (55).
Smooth helical faces that result from having small interfacial residues may provide the flexibility needed for the conformational change in the center of TM2 (56) or, alternatively, may ensure a wide pathway through the channel for ions to reach the selectivity filter.
The sole stretch of TM2 residues that could not be blocksubstituted to alanines or to leucines without abolishing function was 352 FFCDL 356 . Function was also abolished when Asp 355 was singly mutated to alanine or leucine, 4 indicating that Asp 355 has a dual role, being both structurally and functionally important. The aspartate corresponding to Asp 355 of the P2X 5 subunit is conserved among all seven P2X isoforms and has been shown by cysteine-accessibility scanning to be located on the intracellular side of the channel gate of the P2X 2 receptor (57). Our data fit well with those of a tryptophanscanning study on the P2X 4 receptor that showed that nonfunctional mutations clustered in the inner TM2 region (58). To account for the dual structural and functional roles of Asp 355 that emerge from our study, we propose that Asp 355 serves to control and stabilize the positioning of inner TM2 4 W. Duckwitz, R. Hausmann, and G. Schmalzing, unpublished results. Each cylinder corresponds to a block of five amino acid residues, the sequence of which is indicated in the left margin. Labeling of cylinders with A and/or L signifies that the corresponding penta-alanine (A) or penta-leucine (L) block mutations did not abolish ATP-gated cation channel function. The sole sequence block that could be neither mutated to alanines nor to leucines is highlighted in red and includes Asp 355 . Plus signs indicate positively charged residues involved in the interfacial anchoring of the C-terminal end of TM2. DECEMBER 22, 2006 • VOLUME 281 • NUMBER 51 JOURNAL OF BIOLOGICAL CHEMISTRY 39571 residues during the conformational transitions that accompany channel gating.

Role of TM2 and Asp 355 in P2X 5 Homotrimerization
TM2 and Asp 355 Function as Scaffolds in P2X 5 Subunit Assembly-Our data are most compatible with the view that TM2 makes two essential contributions to hP2X 5 subunit oligomerization. First, the entire TM2 domain exerts a role that seems to have its basis in constraining the spatial mobility of the polypeptide chain, when the C-terminal end of the ectodomain is tethered to the membrane to form a loop-like structure. As long as certain threshold values for the hydrophobicity and length of this domain meet the requirements for membrane insertion, homotrimerization occurs. Second, TM2 formation results in an intramembrane localization of Asp 355 , which may initiate homooligomerization by hydrogen bond-driven transmembrane helix-helix associations. We suggest that TM2 formation and the intramembrane presentation of Asp 355 constitute scaffolds, which drive productive assembly by assisting in the creation and correct positioning of specific recognition surfaces that are evidently located in the ectodomain.