Structural and molecular basis of the assembly of the TRPP2/PKD1 complex

Abstract

Mutations in PKD1 and TRPP2 account for nearly all cases of autosomal dominant polycystic kidney disease (ADPKD). These 2 proteins form a receptor/ion channel complex on the cell surface. Using a combination of biochemistry, crystallography, and a single-molecule method to determine the subunit composition of proteins in the plasma membrane of live cells, we find that this complex contains 3 TRPP2 and 1 PKD1. A newly identified coiled-coil domain in the C terminus of TRPP2 is critical for the formation of this complex. This coiled-coil domain forms a homotrimer, in both solution and crystal structure, and binds to a single coiled-coil domain in the C terminus of PKD1. Mutations that disrupt the TRPP2 coiled-coil domain trimer abolish the assembly of both the full-length TRPP2 trimer and the TRPP2/PKD1 complex and diminish the surface expression of both proteins. These results have significant implications for the assembly, regulation, and function of the TRPP2/PKD1 complex and the pathogenic mechanism of some ADPKD-producing mutations.

Autosomal dominant polycystic kidney disease (ADPKD), one of the most common genetic diseases in humans, is caused by mutations in PKD1 or TRPP2 (1–3). PKD1 (also known as polycystin-1 or PC1) is a 4,302-aa, 465-kDa integral membrane protein containing 11 putative transmembrane regions (4) (Fig. S1). Its large extracellular N terminus contains a number of well-recognized repeats and domains, some of which are known to interact with extracellular matrix proteins. The short intracellular C terminus contains a G protein activation site. Thus, PKD1 is generally thought to function as a cell surface receptor for extracellular ligands and matrix proteins that couples extracellular stimuli, including mechanical stimuli, to intracellular signaling (1, 4–6). TRPP2 (also known as PKD2, polycystin-2 or PC2) is a member of the transient receptor potential (TRP) channel family (7). TRPP2 is a 968-aa, 110-kDa integral membrane protein with 6 putative transmembrane segments and a pore-forming loop (7) (Fig. S1) and forms a Ca²⁺-permeable nonselective cation channel (8–10).

How mutations in PKD1 and TRPP2 lead to ADPKD is unclear. The 2 proteins likely participate in the same molecular and cellular processes in some kidney cells, because their mutations produce similar pathological manifestations. Although both proteins are widely distributed, they show high levels of expression in kidney cells (1, 5–7), where they are present on the plasma membrane and in the Golgi and endoplasmic reticulum (1, 4–6, 11–13). They also colocalize in the primary cilia of kidney epithelium (14). The 2 proteins associate physically (15–19) and form functional complexes (18, 20, 21). However, the subunit composition of this heteromeric complex and the molecular mechanism underlying its assembly are unknown.

In this work we investigated these issues by using a multifaceted approach that includes biochemistry, X-ray crystallography, and single-molecule optical imaging. Our study provides the structural basis of the homomeric assembly of a functionally important TRPP2 coiled-coil domain, uncovers the crucial role of this domain in the assembly of the TRPP2/PKD1 complex, and reveals an unexpected subunit stoichiometry of this complex.

Results

Subunit Stoichiometry of the TRPP2 Homomultimer Expressed in a Cell Line.

To study the molecular mechanism of the assembly of the TRPP2/PKD1 complex, we first examined the stoichiometry of the TRPP2 complex in the absence of PKD1. TRPP2 purified from a HEK 293T cell line stably expressing HA-TRPP2 [TRPP2 tagged with hemagglutinin (HA) on the N terminus] was analyzed by blue native PAGE (BN-PAGE). Undenatured and unreduced TRPP2 protein migrated as a complex with an apparent molecular mass of ≈850 kDa (Fig. 1A, lane 1). Denatured and reduced protein migrated as a complex with an apparent molecular mass of ≈280 kDa (Fig. 1A, lane 2). In contrast, denatured but unreduced protein showed 3 products, with an apparent molecular mass of ≈280, ≈600, and ≈850 kDa (Fig. 1A, lane 3). The relative size of the 3 products suggests that they correspond to the TRPP2 monomer, dimer, and trimer. The apparent molecular mass of these products is much larger than their predicted molecular mass (110, 220, and 330 kDa, respectively). Glycosylation, detergent binding, and/or anomalous mobility in BN-PAGE probably account for their much higher observed molecular mass.

Fig. 1.

Subunit stoichiometry of full-length homomeric TRPP2 and heteromeric TRPP2/PKD1 complexes. (A) Western blot analysis after BN-PAGE of purified HA-TRPP2. In lanes 2 and 3, the sample was incubated at 70 °C for 10 min with 2% SDS plus 100 mM DTT or with 2% SDS. TRPP2 was detected with an anti-HA antibody, and putative TRPP2 monomer, dimer, and trimer are indicated. (B) Western blot analysis after SDS/PAGE showing the cross-linking products of TRPP2. Cross-linking was carried out with 1 mM DST. The putative subunit composition of the bands is indicated. (C and F) TIRF images of EGFP and mCherry fluorescence from an oocyte expressing the indicated constructs, showing spots that exhibited EGFP and mCherry dual fluorescence and were immobile (circles). Other spots do not fit analysis criteria. (Scale bars: 2 μm.) (D) Time course of photobleaching of 2 representative PKD1-mCherry and TRPP2-EGFP dual-fluorescence spots, showing 3 EGFP bleaching steps (arrows). mCherry and EGFP excitation is indicated by red and green bars, respectively. (E) Distribution of observed EGFP bleaching steps (green bars) for PKD1-mCherry and TRPP2-EGFP dual-fluorescence spots compared with calculated distribution based on 83% of EGFPs being fluorescent (white bars). No events with 5 or more bleaching steps were observed. (G) Time course of photobleaching of 2 representative PKD1-EGFP and TRPP2-mCherry dual-fluorescence spots showing 1 EGFP bleaching step (arrow). (H) Distribution of EGFP bleaching steps for PKD1-EGFP and TRPP2-mCherry dual-fluorescence spots.

We next used chemical cross-linking to further determine the subunit stoichiometry of the TRPP2 homomeric complex. Treatment of the lysate of the HA-TRPP2-expressing cells with the amine-reactive cross-linker DST resulted in 3 products, with an apparent molecular mass of 120, 260, and 400 kDa (Fig. 1B), which likely correspond to the TRPP2 monomer, dimer, and trimer (the largest band is unlikely to be the tetramer because the extrapolated molecular mass of the tetramer would be ≈520 kDa). These and the preceding results indicate that the prevalent form of TRPP2 in the cell lysate is a trimer.

We also produced 2 other HEK 293T stable lines, expressing either FLAG-PKD1 (PKD1 tagged with FLAG on the N terminus) or FLAG-PKD1 plus HA-TRPP2. Because of technical challenges, our attempts to cross-link TRPP2 and PKD1 produced inconclusive results.

Subunit Stoichiometry of the TRPP2/PKD1 Heteromeric and TRPP2 Homomeric Complex in the Membrane of Live Cells.

Next, we used a single-molecule optical approach (22) to determine the stoichiometry of the TRPP2/PKD1 complex and the homomeric TRPP2 complex in the membrane of live cells. This method uses total internal reflection fluorescence (TIRF) microscopy to visualize single, GFP-tagged proteins exclusively on the plasma membrane of Xenopus oocytes. Measurements were made simultaneously for many complexes at low levels of protein expression, where the formation of non-native, high-order complexes and aggregates is avoided. The number of GFP photobleaching steps was used to determine the number of subunits of one of the proteins in a complex and a mCherry tag was used to reveal the presence of the second subunit type within the same complex.

We fused EGFP or mCherry to the cytosolic C terminus of TRPP2 or PKD1 (the resultant subunits were named TRPP2-EGFP, TRPP2-mCherry, PKD1-EGFP, and PKD1-mCherry). Xenopus oocytes were injected with the cRNAs of PKD1-mCherry and TRPP2-EGFP or PKD1-EGFP and TRPP2-mCherry. Surface fluorescent spots were visualized by first exciting mCherry and then EGFP. Spots showing dual fluorescence were identified as TRPP2/PKD1 complexes (Fig. 1 C and F). Although many spots diffused laterally in the membrane, a large portion (≈40%) was immobile and their EGFP bleaching steps could be counted (e.g., circle spots in Fig. 1 C and F). In oocytes expressing PKD1-mCherry and TRPP2-EGFP, most dual-fluorescence spots showed 2 or 3 EGFP bleaching steps, with a small minority of the spots bleaching in 1 or 4 steps (Fig. 1 D and E). The 1- to 3-step distribution is well fit by a binomial distribution (Fig. 1E) that assumes that each spot contains 3 EGFPs and that the probability of EGFP to be fluorescent is 83%, similar to what was observed earlier on other membrane proteins (22). The presence of a small number of spots (3.5% of the dual-fluorescence spots) that showed 4 EGFP bleaching steps can be accounted for by the occasional colocalization of 2 complexes. The distribution of EGFP bleaching steps fits poorly to a binomial distribution that assumes that each spot contains 4 EGFPs (Fig. S2). In the complementary experiment in oocytes expressing PKD1-EGFP and TRPP2-mCherry (Fig. 1F), most dual-fluorescence spots showed only 1 EGFP bleaching step (Fig. 1 G and H). A small minority of spots showed 2 EGFP bleaching steps (Fig. 1H), probably because they contained 2 complexes. mCherry showed a gradual decrease in fluorescence, indicating the presence of several subunits (Fig. 1G), but because of mCherry's weak fluorescence emission and large fluorescence fluctuations, steps could not be counted reliably. Taken together, these results indicate that the TRPP2/PKD1 full channel complex on the surface membrane of Xenopus oocytes contains 3 TRPP2 and 1 PKD1.

Numerous PKD1 fluorescence spots were observed on the plasma membrane of oocytes expressing PKD1-EGFP and TRPP2-mCherry (Fig. 1F and Fig. S3A); however, very few spots were observed in oocytes expressing PKD1-EGFP alone (Fig. S3B). This finding indicates that, on its own, PKD1 reaches the surface membrane very inefficiently. In contrast, abundant fluorescence spots were detected on the plasma membrane of oocytes expressing TRPP2-EGFP alone (Fig. S3C). This suggests that TRPP2 reaches the surface membrane either on its own or, perhaps, in complex with an unknown endogenous oocyte protein. One such candidate protein is TRPC1, which is endogenously expressed in oocytes (23) and directly associates with TRPP2 (17, 21, 24). Fluorescence spots from oocytes expressing TRPP2-EGFP alone showed the same EGFP bleaching pattern (Fig. S3 D and E) as did the dual-fluorescence spots from oocytes expressing TRPP2-EGFP and PKD1-mCherry (Fig. 1 D and E), indicating that the complex containing TRPP2-EGFP alone has 3 subunits of TRPP2. This is in agreement with the stoichiometry of the TRPP2 homomeric complex in HEK 293 cell lysates determined biochemically (Fig. 1 A and B).

Stoichiometry of a TRPP2/PKD1 C-Terminal Coiled-Coil Domain Complex.

We next investigated the molecular basis for the 3:1 stochiometry of the TRPP2/PKD1 complex. Previous studies show that C-terminal fragments of TRPP2 and PKD1 associate with each other in vitro (15–17). The interaction site has been mapped to amino acids H822 to G895 in TRPP2 (17) and a putative coiled-coil domain (L4214 to R4248) in PKD1 (15). We examined the stoichiometry of complexes formed by TRPP2 and PKD1 C-terminal fragments encompassing these regions, which were coexpressed in and copurified from bacteria. The measured molecular mass of these complexes, determined by static light scattering (Fig. S4), closely matches the predicted molecular mass of 3 TRPP2 and 1 PKD1 (Fig. 2, bars 1 and 2). Likewise, the measured molecular mass of 2 different TRPP2 fragments corresponds nicely with the calculated molecular mass of 3 TRPP2 (Fig. 2, bars 3 and 4). These results indicate that the C-terminal domain of TRPP2 forms a homotrimer and that this trimer can associate with the C terminus of 1 PKD1 to form a 3:1 complex.

Fig. 2.

The complex formed by TRPP2 and PKD1 C-terminal interacting domains has 3 TRPP2 and 1 PKD1. Bar graph compares the calculated and measured molecular masses (determined by static light scattering) of the indicated proteins. The calculated molecular mass was obtained according to the stoichiometry schematized on the right. The protein fragments (indicated at the bottom) were tagged with either hexahistidine (His₆), MBP, or His₆ plus SUMO. MBP and SUMO also served to increase the resolution and accuracy of the molecular mass measurement. MBP-His₆ was used as a control to demonstrate the accuracy of the equipment. Results are shown as mean and SD. n = number of measurements.

Crystal Structure of a TRPP2 C-Terminal Coiled-Coil Domain.

To examine how the TRPP2 C-terminal domain forms a homotrimer, we solved the crystal structure of a C-terminal fragment of TRPP2 harboring the PKD1 coiled-coil domain interaction site. This fragment, from amino acids G833 to G895, is conserved across species (Fig. 3A) and contains a predicted coiled-coil domain (from F839 to A873) (Fig. 3B). The structure, solved at 1.9-Å resolution (Fig. 3C, Table S1, and Fig. S5), shows that this fragment forms a continuous α helix (from Y836 on) and assembles into a trimer. The N-terminal portion containing the coiled-coil domain (from Y836 to R872) is tightly bundled together via extensive hydrophobic interactions (Fig. 3D). The C-terminal portion of the α-helices, from A873 to G895, splays open (Fig. 3C). In the crystal lattice, this region interacts with the same region of another trimer, forming a hexameric complex containing 2 interacting trimers (Fig. 3E). This hexameric complex is likely the result of construct design and crystal packing because, in solution, longer TRPP2 fragments encompassing the aforementioned α-helix clearly form a trimer, not a hexamer (Fig. 2, bars 3 and 4).

Fig. 3.

Crystal structures of the TRPP2 coiled-coil domain reveal that it forms a trimer. (A) Amino acid sequence alignment of a TRPP2 C-terminal fragment (G833–G895 in human) from the indicated species. National Center for Biotechnology Information accession numbers for sequences used here are: EAX06011 (human), NP_032887 (mouse), NP_001026311 (chicken), NP_001002310 (zebrafish), and NP_999827 (sea urchin). Stpu*, Strongylocentrotus purpuratus. Red: residues identical to those in human. Green: similar residues. Blue bar indicates a predicted coiled-coil domain; green bar indicates the open region seen in the crystal structure in C. (B) Heptad repeats of the predicted coiled-coil domain in A. Residue number is given on both sides of the sequence. Letters a–g designate the position of the amino acids within the heptad. Positions a and d are invariantly hydrophobic, a characteristic of coiled-coil domains. (C) Crystal structure of TRPP2_G833–G895. G833–R872 is colored blue and A873–G895 is green. (D) Side-chain of amino acids involved in the trimeric coiled-coil interactions. For clarity, 1 subunit is removed. Residues in red were mutated to alanine to generate mut6. (E) Packing of TRPP2_G833–G895 trimers in crystal. Two trimers interact with each other through their C termini to form a hexamer, as exemplified by the pair colored in blue and orange. The unit cell is defined by the purple box. (F) Crystal structure of TRPP2_G833–R872. Residues V834–M870 were resolved in the structure. (G) Superposition of the structures of TRPP2_G833–G895 and TRPP2_G833–R872.

We also solved the structure of a fragment (from G833 to R872) containing only the coiled-coil domain (Fig. 3F and Table S1). Superposition of this structure with that of the longer fragment shows a high degree of overlap (Fig. 3G), with an rmsd of 0.55 Å when the Cα atoms of the coiled-coil are superimposed. This finding indicates that the C-terminal interaction observed in the crystal between 2 trimers (Fig. 3E) did not perturb the structure of the coiled-coil domain.

Disruption of the TRPP2 Coiled-Coil Domain Trimer Abolishes TRPP2/PKD1 Coiled-Coil Domain Interaction.

Pull-down experiments showed that the TRPP2 G833–G895 fragment, tagged with small ubiquitin-like modifier (SUMO), was able to bind the maltose binding protein (MBP)-tagged PKD1 coiled-coil domain, MBP-PKD1_R4213–R4248 (Fig. 4A, lane 1). However, neither the trimer-forming “bundled” TRPP2_G833–R872 fragment nor the downstream “open” E871–G895 fragment could, on its own, interact with the PKD1 coiled-coil domain (Fig. 4A, lanes 2 and 3). Indeed, even TRPP2_V857–G895, which contains more than one-third of the trimer-forming residues and the open region, could not bind the PKD1 fragment (Fig. 4A, lane 4). Thus, the TRPP2/PKD1 coiled-coil domain interaction requires both the bundled and open regions of TRPP2. The physical/chemical nature of this interaction remains to be elucidated.

Fig. 4.

The TRPP2 trimer is essential for the formation of the TPPP2/PKD1 coiled-coil domain complex. (A) SDS/PAGE showing the interaction or the lack thereof between the indicated PKD1 and TRPP2 fragments (schematized in the lower right corner). The MBP-tagged PKD1 fragment was used to pull down the indicated SUMO-tagged TRPP2 fragments (upper gel), which showed similar level of expression (lower gel; arrows indicated the corresponding protein bands). (B) Gel filtration profile of MBP-tagged WT TRPP2_G821–S926 and TRPP2_G821–S926_mut6 (schematized on top, with stars denoting the mut6 mutation). The right shift of the latter indicates a molecular mass decrease, as confirmed by static light scattering measurements (molecular mass shown in graph). (C) SDS/PAGE showing that mut6 abolishes interaction between the indicated PKD1 and TRPP2 fragments. The MBP-tagged PKD1 fragment was used to pull down the indicated SUMO-tagged TRPP2 fragments (upper gel), which showed a similar level of expression (lower gel).

We next tested whether the TRPP2 coiled-coil domain trimer was necessary for forming the TRPP2/PKD1 (3:1) C-terminal complex. Six hydrophobic residues (L842, V846, M849, I860, V863, and L867; Fig. 3D) critical for forming the trimer were simultaneously mutated to alanine in an MBP-tagged TRPP2 C-terminal fragment, MBP-TRPP2_G821–S926 (this hexa-alanine mutation was named “mut6”). Gel filtration profile and static light scattering both indicated that this mut6 fragment did not form a trimer; instead, it remained as a monomer (Fig. 4B). Pull-down experiments showed that a slightly shorter TRPP2 fragment (G833–S918) carrying mut6 could not bind the PKD1 coiled-coil domain (R4213–R4248), whereas the corresponding WT TRPP2 fragment was fully competent (Fig. 4C). These results indicate that the PKD1 coiled-coil domain binds the TRPP2 trimer but does not bind the TRPP2 monomer.

Disruption of the TRPP2 Coiled-Coil Domain Trimer Disrupts the Assembly of the Full-Length TRPP2/PKD1 Complex.

To further examine the functional importance of the TRPP2 coiled-coil domain trimer, we engineered the mut6 mutation into full-length TRPP2 and created 2 additional HEK 293T stable cell lines, expressing either HA-TRPP2_mut6 or HA-TRPP2_mut6 plus FLAG-PKD1. Two lines of evidence demonstrate that full-length TRPP2_mut6 does not interact with each other to form homomeric complexes: (i) TRPP2_mut6 could not be cross-linked (Fig. 5A). Cross-linking of WT TRPP2 with DST resulted in the presumed TRPP2 dimer and trimer, but the same treatment of TRPP2_mut6 failed to produce these cross-linked adducts (Fig. 5A). (ii) Whereas WT TRPP2 migrated as a trimer in BN-PAGE, TRPP2_mut6 migrated as a monomer (Fig. 5B).

Fig. 5.

The TRPP2 coiled-coil domain trimer is essential for the formation of TRPP2 homomeric and TRPP2/PKD1 heteromeric complexes. (A) Western blot analysis after SDS/PAGE showing the lack of cross-linking of TRPP2_mut6 subunits. Cross-linking reaction (with the indicated concentration of DST) was carried out in the lysate of HEK 293T cells stably expressing either HA-TRPP2 or HA-TRPP2_mut6. Here, and in B and C, TRPP2 was detected with an anti-HA antibody. Putative TRPP2 monomer, dimer, and trimer are indicated. (B) Western blot analysis after BN-PAGE of purified HA-TRPP2 (WT) or HA-TRPP2_mut6 (mut6). Putative TRPP2 monomer and trimer are indicated. (C) Western blot analysis after SDS/PAGE showing that WT TRPP2 but not TRPP2_mut6 can be coimmunoprecipitated with PKD1. Immunoprecipitation (IP) was carried out with either an anti-HA or an anti-FLAG antibody using the lysates of HEK 293T cells stably expressing either FLAG-PKD1 and HA-TRPP2 (lanes 1 and 3) or FLAG-PKD1 and HA-TRPP2_mut6 (lanes 2 and 4). (D and E) TIRF image of EGFP and mCherry fluorescence from an oocyte expressing the indicated combinations of constructs. (Scale bar: 2 μm.)

Our experiments also showed that the interaction between full-length TRPP2 and PKD1 is abolished when the TRPP2 coiled-coil domain trimer is disrupted. WT TRPP2 could be coimmunoprecipitated with PKD1 from HEK293T cells stably expressing FLAG-PKD1 and HA-TRPP2, with either an anti-HA or anti-FLAG antibody (Fig. 5C, lanes 1 and 3). In contrast, TRPP2_mut6 could not be coimmunoprecipitated with PKD1 from HEK293T cells stably expressing FLAG-PKD1 and HA-TRPP2_mut6 (Fig. 5C, lanes 2 and 4). These results indicate that formation of the TRPP2 coiled-coil domain trimer is necessary for full-length TRPP2 to form homotrimers and the TRPP2/PKD1 complex.

Disruption of the TRPP2 Coiled-Coil Domain Trimer Diminishes the Surface Expression of TRPP2 and PKD1.

Previous studies indicate that the association of TRPP2 and PKD1 enhances the surface expression of both proteins (18, 20, 25). Abundant surface EGFP fluorescence spots were observed in oocytes expressing TRPP2-EGFP (Fig. S6A); however, few such spots were seen in oocytes expressing TRPP2_mut6-EGFP (Fig. S6B). Furthermore, whereas numerous TRPP2 fluorescence spots were observed on the plasma membrane when TRPP2 was coexpressed with PKD1 (Figs. 1C and 5D and Fig. S3A), very few TRPP2_mut6 fluorescence spots were detected when TRPP2_mut6 was coexpressed with PKD1 (Fig. 5E). These results indicate that both basal and PKD1-stimulated TRPP2 surface expression was abolished by the mut6 mutation. Conversely, few surface PKD1 fluorescence spots were observed in oocytes expressing PKD1-EGFP and TRPP2_mut6-mCherry (Fig. 5E), in contrast to the abundance of such spots in oocytes expressing PKD1-EGFP and TRPP2-mCherry (Figs. 1F and 5D and Fig. S3A), indicating that TRPP2-dependent PKD1 surface expression was also abolished by the mut6 mutation.

Discussion

Many proteins assemble or interact with their partners through coiled-coil domains, which can associate homophilically and/or heterophilically to form dimers, trimers, or tetramers (26–28). Our work shows that TRPP2 forms a homotrimer through a coiled-coil domain (amino acids F839–A873) in the C terminus. This domain overlaps with the one (F839–D919) recently reported to be involved in homophilic TRPP2 interactions (29), but it differs from the one (E772–L796) previously postulated to play such a role (16). Our work further shows that a trimeric TRPP2 interacts with a single PKD1 to form a cell surface TRPP2/PKD1 complex, and that this 3:1 stoichiometric association is determined by an interaction between the newly identified TRPP2 coiled-coil domain trimer and a previously identified (15) C-terminal coiled-coil domain of PKD1 (amino acids L4214 to R4248). The functional importance of these domains and their interactions is underscored by many naturally occurring ADPKD pathogenic mutations, including R4227X in PKD1 and R742X, R807X, E837X, and R872X in PKD2 (2), which delete the coiled-coil domains or the downstream open region and, hence, abolish the assembly of the TRPP2/PKD1 complex.

It has been shown recently that a region in the C terminus of TRPM7 coassembles to form an antiparallel 4-stranded coiled-coil (30). In contrast, our results demonstrate that the newly identified TRPP2 coiled-coil domain assembles to form a trimer, in both solution (Fig. 2) and crystal (Fig. 3 C and F). This coiled-coil has many of the hallmarks of a canonical coiled-coil (26–28), in particular, the characteristic heptad repeat with hydrophobic residues at the first (i.e., a) and fourth (i.e., d) positions, and with charged residues at the fifth (i.e., e) and seventh (i.e., g) positions (Fig. 3B). Crystal structures show that this coiled-coil domain forms a 3-stranded, parallel bundle (Fig. 3 C and F), conforming to the propensity that opposite charges at the e and g positions (Fig. 3B) favor a parallel polarity of the α-helical chains (26–28). Interestingly, in the crystal lattice of the long fragment (i.e., TRPP2_G833–G895), the C-terminal regions downstream of the coiled-coil domain of one trimer interact with the same regions of another trimer (Fig. 3E). Even though this interaction appears to be an artifact caused by construct design and crystal packing, as alluded to above, the ability of this region to interact with other α-helices suggests that it might be the interacting site for the PKD1 coiled-coil domain. This notion and the physical/chemical nature of the TRPP2/PKD1 interaction await further studies.

Although our results indicate that the newly identified TRPP2 coiled-coil domain is necessary for the assembly of full-length homomeric TRPP2 complex, they do not exclude contributions from other regions. Indeed, an N-terminal region was recently shown to be involved in TRPP2 oligomerization (31). Furthermore, disulfide bonds between TRPP2 subunits may also play a role, because full-length TRPP2 remains dimerized and trimerized under nonreducing but denaturing conditions (31) (Fig. 1A).

The 3:1 stoichiometry of the TRPP2/PKD1 complex is surprising, given that PKD1 has not generally been thought of as a channel-forming subunit (refs. 1, 4–6, 11, and 12, but see ref. 32). If the channel pore of the TRPP2/PKD1 complex is formed by 4 subunits, the 3:1 stoichiometry would suggest either that PKD1 contributes to form the channel pore, in addition to its putative receptor functions, or that an endogenous oocyte protein, such as TRPC1 (23), associates with TRPP2 in a 1:3 stoichiometry to form the ion conduction pathway. These possibilities remain to be investigated. On first thought, PKD1 seems unlikely to be a channel-forming subunit because the original transmembrane topology prediction did not identify a pore-forming region (4). In that prediction, the region between the 10th and 11th transmembrane (TM) segments, which presumably correspond to the S5 and S6 TM segments of TRPP2, was only 7 aa long and was therefore too short to constitute a pore-forming loop. However, using 6 different contemporary programs for predicting TM segments, we find that the location of the last (i.e., 11th) TM segment is uncertain. Four predictions place this TM segment ≈25–36 aa away from the 10th TM segment, long enough to form a pore loop. More intriguing, these 25–36 aa share ≈50–55% similarity with the homologous putative pore-forming region in TRPP2. Thus, this region might constitute a pore-forming loop in PKD1.

TRPP2 forms spontaneously active nonselective cation channels in lipid bilayers (8, 9, 24). Functional expression of TRPP2 in Xenopus oocytes has also been reported (33). However, we failed to detect TRPP2-conducted currents when TRPP2 was expressed in oocytes (Fig. S7), even though plentiful TRPP2 proteins were detected on the plasma membrane (Fig. S6A). Also, in contrast to previous observations in Chinese hamster ovary cells (18), we were unable to obtain spontaneously active currents that could be conclusively ascribed to TRPP2/PKD1 complexes in HEK 293T cells or Xenopus oocytes expressing TRPP2 and PKD1 (Fig. S7). This lack of current was not caused by the lack of expression of the complexes on the surface membrane because both proteins were abundantly expressed (Figs. 1 C and F and 5D and Fig. S3A). Finally, it has been reported that the TRPP2/PKD1 complex can be activated by an anti-PKD1 antibody raised against an epitope (E2939–N2956) in the extracellular N terminus in human PKD1 (20, 21). We produced 2 different batches of anti-mouse PKD1 serum raised against the corresponding region (E2931–N2948) in mouse PKD1. Both sera induced a current in HEK 293T cells or Xenopus oocytes expressing TRPP2 and PKD1; however, this current was also evoked by a control serum and was mostly likely conducted by channels unrelated to the TRPP2/PKD1 complex (Fig. S8). Thus, it remains a challenge to functionally characterize homomeric TRPP2 and heteromeric TRPP2/PKD1 complexes in these expression systems.

Several TRP channel subunits have been shown to form tetramers on their own (34–36). Single-channel recording and atomic force microscopy imaging suggest that homomeric TRPP2 complexes incorporated into lipid bilayers are tetramers (24). Yet, our results indicate that TRPP2 predominantly forms a trimer when expressed in HEK 293 cells and oocytes, not only in the cell lysate (Fig. 1 A and B) but also in the plasma membrane (Fig. S3 D and E). These results suggest that the association of the fourth subunit is significantly weaker than the trimeric assembly and that this association is potentially dynamic and regulated. A tightly bound TRPP2 trimer may associate with other channel-forming subunits, such as TRPC1 (17, 21, 24) and TRPV4 (37), to form heterotetrameric channels, thereby greatly increasing its functional spectrum and versatility. It is of interest to note that the cyclic nucleotide-gated channels of rod photoreceptors also have a 3:1 stoichiometry, with 3 A1 and 1 B1 subunits (38–42), and that this stoichiometry is determined by a trimer-forming leucine-zipper domain in the C terminus of the A1 subunit (40).

Materials and Methods

Chemical Cross-Linking.

Cell lysates were treated with freshly prepared disuccinimidyl tartarate (DST; Pierce) on ice for 4 h.

TIRF and Determination of Bleaching Steps.

Xenopus oocytes were enzymatically treated to enable close contact to the coverslip (22). Movies were acquired with a back-illuminated EMCCD camera (Andor iXon DV-897 BV). Gaussian profile was fitted to the raw images to determine the time course of the emission intensity. Fluorescent spots that stayed immobile were selected, and bleaching steps were counted manually.

Light Scattering.

Purified proteins were run through a gel filtration column and the eluates were examined by static light scattering (Wyatt Technology).

Crystallization, Data Collection, and Structure Determination.

Crystallization was carried out by using the hanging-drop vapor diffusion method at 20 °C. Crystals were rinsed in Paratone-N (Hampton Research) and flash-frozen in liquid nitrogen for data collection at 100 K. Molecular replacement method was used to determine the structures. Data collection and refinement statistics are summarized in Table S1.

Other Materials and Methods.

Details for the methods described above and for construct cloning, cell culture, transfection and stable line generation, SDS/PAGE, BN-PAGE and Western blot analysis, protein fragment expression and purification, coimmunoprecipitation, anti-PKD1 serum generation, and electrophysiology are provided in SI Text.