The Structure of the Polycystic Kidney Disease Channel PKD2 in Lipid Nanodiscs

SUMMARY

The Polycystic Kidney Disease 2 (pkd2) gene is mutated in autosomal dominant polycystic kidney disease (ADPKD), one of the most common human monogenic disorders. Here, we present the cryo-EM structure of PKD2 in lipid bilayers at 3.0 Å resolution, which establishes PKD2 as a homotetrameric ion channel and provides insight into potential mechanisms for its activation. The PKD2 voltage-sensor domain retains two of four gating charges commonly found in those of voltage-gated ion channels. The PKD2 ion permeation pathway is constricted at the selectivity filter and near the cytoplasmic end of S6, suggesting that two gates regulate ion conduction. The extracellular domain of PKD2, a hotspot for ADPKD pathogenic mutations, contributes to channel assembly and strategically interacts with the transmembrane core, likely serving as a physical substrate for extracellular stimuli to allosterically gate the channel. Finally, our structure establishes the molecular basis for the majority of pathogenic mutations in pkd2-related ADPKD.

Graphical abstract

Structural characterization of the polycystic kidney disease 2 (PKD2) ion channel in lipid environment reveals how pathogenic mutations perturb the channel structure and function.

INTRODUCTION

Autosomal dominant polycystic kidney disease (ADPKD) is a common, life threatening, multisystem genetic disorder that afflicts >600,000 Americans and ~12.5 million people worldwide (Ong and Harris, 2015). ADPKD is characterized by progressive accumulation of fluid-filled bilateral renal cysts, leading to gross kidney enlargement, relentless loss of normal renal tissues, and ultimately end-stage renal disease that requires dialysis or transplantation. Apart from renal insufficiency, patients with ADPKD may develop hepatic and pancreatic cysts, hypertension, and intracranial aneurysms (Harris and Torres, 1993).

ADPKD is caused by loss-of-function mutations in a single copy of the pkd1 or pkd2 genes (Mochizuki et al., 1996; The European Polycystic Kidney Disease Consortium, 1994). Complete genetic knockout of either pkd1 or pkd2 in mice results in embryonic lethality due to structural defects in the cardiovascular system, pancreas, and kidneys (Boulter et al., 2001; Kim et al., 2000; Lu et al., 1997). PKD1 (or PC1) is predicted to adopt an 11-transmembrane topology with a remarkably large auto-cleaved amino-terminal ectodomain (>3,000 residues) that comprises an array of putative adhesive and ligand binding modules (Burn et al., 1995; Hughes et al., 1995; Qian et al., 2002). PKD2 (or PC2, TRPP1, formerly TRPP2) is a member of the large, 6-transmembrane spanning transient receptor potential (TRP) ion channel family (Ramsey et al., 2006; Semmo et al., 2014) and has been observed to form a biochemical complex with PKD1 (Qian et al., 1997; Tsiokas et al., 1997). ADPKD is currently an untreatable disease in large part due to a gap in understanding the biochemical, structural, and functional properties of PKD1, PKD2, and the PKD1/PKD2 complex, which impedes the development of therapeutic strategies. Current treatments only alleviate some symptoms, without altering disease progression (Harris and Torres, 1993, 2009).

Human TRP channels are a family of 27 related members that have been subdivided into Canonical (TRPC), Vanilloid (TRPV), Melastatin-like (TRPM), Ankryin repeat (TRPA1), Mucolipins (TRPML), and Polycystins (TRPP) based on sequence homology alone (Ramsey et al., 2006). Venkatachalam and Montell divided all 27 into two broad groups (group 1 and group 2) based on similarities in sequence and topology (Venkatachalam and Montell, 2007). Recent structures of several group 1 TRP channels, including TRPV1 (Cao et al., 2013; Liao et al., 2013), TRPV2 (Huynh et al., 2016; Zubcevic et al., 2016), TRPV6 (Saotome et al., 2016), and TRPA1 (Paulsen et al., 2015), demonstrate that they adopt a tetrameric architecture resembling that of voltage-gated ion channels (VGICs), wherein transmembrane segments 5 and 6 (S5 and S6) and the intervening pore loops contributed by four individual subunits form a central ion conduction pore that is flanked by four S1–S4 voltage sensor domains (Long et al., 2005). Although all TRP channels are assumed to share a certain degree of structural similarity, PKD2 is classified among the more distantly related group 2 channels and differs significantly from group 1 TRP channels and related VGICs by having a large, intervening extracellular sequence (>200 residues) between transmembrane segments 1 and 2 (S1 and S2). Importantly, this extracellular domain harbors 15 ADPKD-associated missense mutations, suggestive of a critical, but as-yet-unidentified, structural and/or functional role for this enigmatic ectodomain. Notably, PKD2-like channels exist in organisms ranging from yeast to humans, indicating a primitive and fundamental role for this TRP subfamily (Palmer et al., 2005). Structures of an intact PKD2 channel should therefore address how structural similarities and differences between group 1 and group 2 TRP channels account for their unique functions.

Our understanding of ADPKD pathogenesis is hampered by disagreements about the properties of the putative PKD2 current. The PKD2 ion channel initially was reported to conduct calcium (Hanaoka et al., 2000) and later shown to be blocked by calcium (Cai et al., 2004). Most recently, a gain-of-function mutation, but not the wild-type PKD2, underlies a measurable current when heterologously expressed in Xenopus oocytes (Arif Pavel et al., 2016). Another study claims that it is indirectly activated by Wnt ligands via PKD1 (Kim et al., 2016). While all or some of these findings could be true, it is worrisome that the biophysical properties of the currents are inconsistent between reports. Moreover, there are relatively few recordings of the current since the discovery of PKD2 in 1996. These inconsistencies and our own negative results in reproducing this literature over many years suggest to us that native PKD2 currents have not been measured to date (see below). Nonetheless, there is no ambiguity that PKD2 is genetically linked to formation of cysts in kidney and other tissues to cause significant morbidity and mortality in humans (Mochizuki et al., 1996; Ong and Harris, 2015). Moreover, genetic studies in mice demonstrate that PKD2, together with PKD1-Like 1 (PKD1-L1), is essential for establishment of left-right asymmetry of internal organs and vasculature during development (Field et al., 2011; Kamura et al., 2011; Yoshiba et al., 2012). These points further highlight the necessity of understanding PKD2 structure and function.

Here, we determined the structure of human PKD2 channel in lipid nanodiscs at 3.0 Å resolution by single-particle electron cryo-microscopy (cryo-EM). We also show that PKD2 is apparently a Na⁺/K⁺ conducting channel with lower permeability and smaller single-channel conductance to Ca²⁺. More broadly, our study, together with the recently determined TRPV1-nanodisc structures (Gao et al., 2016), illustrates the powerful applicability of combining nanodisc technology with single-particle cryo-EM to determine structures of membrane proteins in a lipid environment similar to native cells.

RESULTS AND DISCUSSION

Structural Determination of the Human PKD2 Channel

We identified three PKD2 truncation constructs that enable structure determination (Figures S1A–S1D). These constructs exhibit enhanced biochemical stability and structural homogeneity over the full-length channel. The longest construct (hPKD2:53–792) retains most functional domains of the channel, including glycogen synthase kinase phosphorylation sites (Streets et al., 2006), transmembrane segments S1–S6, the entire extracellular domains, and the EF hands (Allen et al., 2014; Petri et al., 2010). The construct lacks a cilia transport motif (Geng et al., 2006), the endoplasmic reticulum retention sequence (Cai et al., 1999), and the coiled-coil domain (Tsiokas et al., 1997) located at N or C termini. Given that truncation of most of the C-terminal domains in the related PKD2-Like 1 (PKD2-L1) channel has negligible effects on channel activity (DeCaen et al., 2016), this PKD2 construct should preserve most functional aspects of the native channel. The shortest construct (hPKD2:198–703) represents the minimal biochemically stable channel core and lacks all cytoplasmic domains. Importantly, all three constructs retain 22 of 26 documented sites of missense mutations that cause ADPKD (discussed below).

In order to visualize PKD2 in a native lipid-like environment, we assembled nanodiscs for hPKD2:198-703 and hPKD2:198–792, which each consist of the channel protein and membrane scaffolding proteins (Figures S1B and S1C). We also obtained hPKD2:53–792 protein stabilized in amphipols for structural determination (Figure S1D). 2D class averages of each PKD2 sample showed a broad distribution of views in which the distinct channel features of tetrameric architecture, transmembrane helices, and extracellular domains are clearly discernible (Figures S1E–S1G). Class averages of nanodisc samples also revealed the less-ordered lipid bilayer and membrane scaffolding proteins, seen as a fuzzy disc-shaped density surrounding the channel (Figures S1E and S1F).

Architecture of the PKD2 Channel

We determined cryo-EM structures of the aforementioned PKD2 channel of different lengths in either nanodiscs or amphipols (Figures 1, S1, and S2). Surprisingly, all three of our structures were superimposable and no extra density was observed to account for cytoplasmic domains (e.g., EF hands), suggesting that the intracellular regions of PKD2 either are unstructured or are connected to the stable channel core via flexible linkers (Figures S1I and S1J). This idea is further supported by reference-free 2D class averages, which show that all three PKD2 constructs exhibit indistinguishable channel features and lack densities attributable to cytoplasmic domains (Figures S1E–S1G). Our structural analysis shows that in the absence of cellular regulatory factors, the cytoplasmic domains of PKD2 are structurally uncoupled from the ion conduction pore, providing a plausible structural explanation for why the EF hands and distal coiled-coil domain are dispensable for channel activity in PKD2-L1 (DeCaen et al., 2016). Most strikingly, removal of all PKD2 cytoplasmic domains does not perturb its structure, which stands in contrast to group 1 TRP channels, such as TRPV1 (Liao et al., 2013), TRPA1 (Paulsen et al., 2015), TRPV2 (Huynh et al., 2016; Zubcevic et al., 2016), and TRPV6 (Saotome et al., 2016), in which cytoplasmic domains contribute to channel assembly.

Figure 1. PKD2 Structure Determined in Lipid Nanodisc.

(A) Side and bottom-up views of PKD2 in nano-discs. Individual channel subunits are color coded. Densities of the nanodisc (gray) and well-resolved lipids (purple) are also shown.

(B and C) Ribbon representations of a PKD2 sub-unit (B) with different domains denoted and the full tetrameric channel (C) with each of four identical subunits color coded as in (A).

See also Figures S1, S2, and S3 and Table S1.

We focus our discussion on the structure of the minimal channel (hPKD2:198–703) determined at 3 Å resolution in lipid nanodiscs (Table S1.). At 3.0 Å resolution, the PKD2 map showed well-resolved densities for the channel protein and three of five known N-glycans in an extracellular domain (Hofherr et al., 2014) (Figures 1A and S3A–S3D). Notably, a 5-nm-high, belt-shaped density attributable to the nanodisc lipid bilayer, encircles the channel protein (Figures 1 and S2E). Side chains of most residues were resolved, permitting unambiguous assignment of primary sequence into the density map and de novo construction of an atomic model for the PKD2 channel (Figures 1B, 1C, and S2F–S2I). In addition, several ordered annular lipids are observed to intercalate into the crevices at the subunit interface (Figure 1A), including 12 tightly bound lipids (three lipids per subunit) retained in hPKD2:53–792 stabilized with amphipols (Figure S3E). In GluCl and TRPV1 channels, phospholipids compete with the partial agonist ivermectin and agonist resiniferatoxin, respectively, for the same inter-subunit site (Althoff et al., 2014; Gao et al., 2016), indicating that high-affinity lipid binding sites are potential targets that can be exploited for drug discovery. Indeed, an antagonist of TRPA1 is observed to target a membrane pocket formed by S5, S6, and the pore helix (Paulsen et al., 2015), which is analogous to a lipid-binding site seen in our PKD2 density maps (Figure S3E).

The PKD2 channel adopts a tetrameric architecture with each subunit consisting of two distinct units within the trans-membrane region. The voltage-sensing domain (VSD) (which, in VGICs, harbors the positively charged S4 voltage sensor; Long et al., 2007) consists of the S1 to S4 helices. The pore domain includes S5 and S6 helices and an intervening re-entrant pore loop (Figure 1B). These two domains are connected by the S4–S5 linker, a short helix that runs nearly parallel to the inner leaflet of the lipid bilayer (Figure 1B). The central ion permeation pathway is formed by the tetrameric assembly of pore modules contributed by four individual subunits (Figure 1C). At the perimeter of the channel, each of four VSDs engages in ‘‘domain swapping’’ interactions with the pore domain of a neighboring subunit, reminiscent of TRPV1 (Liao et al., 2013), TRPV2 (Huynh et al., 2016; Zubcevic et al., 2016), TRPA1 (Paulsen et al., 2015), and VGICs (Long et al., 2005) (Figure 1C).

Aside from the conserved transmembrane core, PKD2 exhibits distinct structural features compared to group 1 TRP channels and VGICs, particularly within extracellular regions. For instance, a helix-turn structure protrudes upright toward the extracellular space and appears to serve as an extension of the S3 and S4 helices (Figure 1B). In addition, PKD2 features an extracellular domain (²⁴²Met-Tyr⁴⁶⁵) that is covalently connected to S1 and S2 helices, and which likely defines PKD2s (PKD2, PKD2-L1, and PKD2-L2) as a unique subtype within the TRP channel family (Figure 1B). Given that this domain adopts a novel protein fold and is restricted to PKD2s, PKD1, and PKD1-Like proteins (Li et al., 2003), we hereby refer to this as the polycystin domain. The polycystin domain not only contributes to channel assembly by virtue of engaging in extensive homotypic interactions with an adjacent polycystin domain of a neighboring subunit, it also strategically resides atop the VSD and the outer pore region and thus appears to be well positioned to allosterically regulate channel gating (Figure 1C).

PKD2 Is a Na⁺/K⁺ Conducting, Nonselective Cation Channel

Electrophysiology studies have suggested that heterologous PKD2 currents can be measured from the plasma membrane when it is either expressed with PKD1 in mammalian cells or when mutated (F604P) to render it constitutively active in Xenopus laevis oocytes (Arif Pavel et al., 2016; Hanaoka et al., 2000). However, we find that currents from transfected HEK293T and CHO cells were not different from untransfected cells when measured under physiological calcium or divalent-free sodium conditions (Figures S4A and S4B). In addition, currents from LLC-PK1 cells overexpressing PKD2 were no different from untransfected cells (data not shown). In other reports, PKD2 channels can be directly activated by triptolide or indirectly by activating PKD1 via Wnt (Wingless-related integration site) 3a or 9b (Kim et al., 2016; Leuenroth et al., 2007). However, we did not observe significant current activation by these reagents when expressed in either CHO or HEK293T cells (Figures S4C and S4D). Since the related PKD2-L1 channel currents can readily be measured on the plasma membrane, we generated a chimera of PKD2 and PKD2-L1 by replacing the pore region of the PKD2-L1 channel (⁵¹¹Lys-Pro⁵³⁸) with that of PKD2 (⁶³¹Glu-Pro⁶⁵⁸) (Figure 2A). When heterologously expressed in HEK293T and CHO cells, the chimera channel generates robust outwardly rectifying currents (Figure S4A). Since reproducible PKD2 currents could not be measured, we utilized the functional chimera in all electrophysiology experiments.

Figure 2. The PKD2-2L1 Chimera Ion Selectivity and Voltage Dependence.

(A) Top, topology and alignment of PKD2s. Note that the interconnecting pores of PKD2 and PKD2-L1 are 70% identical; residues that differ are underlined. Bottom, side (only two subunits are shown) and extracellular views of the PKD2 structure indicating the residues from PKD2 that can be grafted into the PKD2-L1 channel to create the chimeric channel (red).

(B) PKD2/2-L1 chimera ion selectivity and voltage-dependence. Top, a voltage ramp applied at 0.5 Hz to activate whole-cell currents from HEK293T cells. Middle, representative PKD2-L1 and chimera currents captured in the presence of 140 mM extracellular monovalent ions (divalent free: 1 mM EGTA and 0.5 mM EDTA) or 100 mM CaCl₂. Internal (pipette) saline contained 90 mM NaMES, 10 mM Na₄BAPTA, and 5 mM EGTA. The voltage transition is expanded in the inset green boxes to clearly show the differences in reversal potential. Right, time course of outward peak current, tail current density, and reversal potential (E_rev). Color shading corresponds to the ionic conditions (black, Na⁺; blue, K⁺; gray, NMDG; red, Ca²⁺ shown in left traces).

(C) The relationships between relative permeability, inward single-channel conductance, and tail current density of PKD2-L1 and the chimera channels. Data are summarized from the experiments shown in (A) and Figure S5 (n = 4–6 cells; error, SD).

See also Figures S4, S5, and S6.

Since the chimera channel houses the selectivity filter and the pore helices of PKD2 (Figure 2A), it should share the same ion selectivity as native PKD2. In whole-cell voltage clamp of cells, intracellular and extracellular cations were altered, and reversal potentials were measured to determine their relative permeability (Figure 2B). The PKD2 chimera is more selective for K⁺ and Na⁺ than Ca²⁺ (P_x/P_Na = 2.2 K⁺; 1 Na⁺; 0.5 Ca²⁺), in contrast to native PKD2-L1, which is 13 times more selective for Ca²⁺ (Figure 2B). Consistent with the relative Na⁺ and K⁺ permeability over Ca²⁺, the largest tail currents were measured in high K⁺ or Na⁺ conditions (Figures 2B and 2C). From the inside-out patch experiments (Figure S5), the inward single-channel conductance was higher for K⁺ and Na⁺ than for Ca²⁺ (γ_K = 218 ± 3; γ_Na = 139 ± 3; γ_Ca = 8 ± 2). Interestingly, in symmetrical Ca²⁺ conditions, only inward single-channel events were detected, suggesting that calcium does not conduct outward through the filter. This one-way Ca²⁺ permeation is also found in PKD2-L1 and was attributed to the most inner aspartate residue (D643 in PKD2 and D523 in PKD2-L1) of the selectivity filter (DeCaen et al., 2016). Together, these data indicate that the PKD2 pore preferentially conducts monovalent K⁺ and Na⁺, while PKD2-L1 contains an additional aspartate (D525) at a position analogous to N645 of PKD2 at the extracellular mouth of the filter (Figures 2A and S6), which contributes to its enhanced selectivity for Ca²⁺ over monovalent ions (DeCaen et al., 2016).

Structure of the Ion Permeation Pathway

The central ion permeation pathway of the PKD2 channel exhibits two major constrictions (or gates) that are sufficiently narrow to restrict ion flow, suggesting that our PKD2 structure represents the channel captured in a closed, non-conductive state (Figures 3A and 3B). One such gate is located at the selectivity filter region (⁶⁴¹IGD⁶⁴³) framed between pore helices 1 and 2, where carbonyl oxygen atoms from diagonally apposed Ile641 form the narrowest constriction (5.5 Å) at the outer pore region. Moreover, the adjacent Gly642 carbonyl oxygen atoms are only 7.0 Å apart and therefore likely restrict permeation of hydrated cations (hydrated Ca²⁺ and Na⁺ appear to be between 8 and 10 Å in diameter when passing through permeable selectivity filters; CavAb [Tang et al., 2014] and NavMs [Naylor et al., 2016]). Moving toward the extracellular space, Asp643 residues point their negatively charged side chains into the central canal and could coordinate permeating cations while repelling anions from passing through the selectivity filter. Notably, replacing this conserved aspartate with a serine or hydrophobic residue in PKD2-L1 abolishes channel activity (DeCaen et al., 2013, 2016; Fujimoto et al., 2011) (Figure S6). TRPV1 (Liao et al., 2013), TRPV2 (Zubcevic et al., 2016), and TRPA1 (Paulsen et al., 2015) also harbor an aspartate or glutamate residue at the equivalent position of the selectivity filter, suggesting an important role for this residue in specifying cation selectivity among different subtypes of TRP channels. Finally, the PKD2 outer pore region is notably rich with negatively charged residues, likely serving as a cation sink that concentrates cations within the extracellular vestibule and enhances ion conduction, as first observed in the KcsA K⁺ channel (Doyle et al., 1998) (Figure 3C).

Figure 3. The Architecture of the Trans-membrane Core of PKD2.

(A) Solvent-accessible pathway along the ion permeation pore mapped using the HOLE program. Residues located at the selectivity filter and lower gate are rendered as sticks.

(B) Radius of pore calculated with the HOLE program.

(C) Enrichment of negatively charged residues at the outer pore region of PKD2. Left: the PKD2 outer pore region iscolored by surface electrostatic potential. Right: negatively charged residues (glutamate and aspartate) are shown as sticks.

(D) An interaction network that couples pore helix 1 and selectivity filter of one subunit with pore helix 2 and S6 of a neighboring subunit.

(E) The structure of PKD2 VSD domain. Two lysine residues at the cytoplasmic end of S4 are likely stabilized by aromatic and negatively charged residues.

See also Figure S6.

Continuing down the pore, Leu677 and Asn681 are one helix turn apart along the pore-facing side of the S6 helix. Together, they form the most constricted sites of the lower gate above the inner leaflet of the lipid bilayer (Figure 3A). In our structure, the distances between diagonally apposing Leu and Asn measure only 5.0 and 6.5 Å between side chains, respectively, and are sufficiently narrow to block conduction of hydrated cations across this gate. Taken together, PKD2 resembles TRPV1 in exhibiting two gates (Cao et al., 2013), and activation of the PKD2 channel necessitates conformational changes that lead to expansion of the pore at both constriction points.

The upper and lower gates are likely allosterically coupled via an interaction network that involves residues at the selectivity filter (Leu641) and pore helix 1 (Arg638) of one subunit and pore helix 2 (Phe646) and the pore-lining S6 helix (Phe661 and Val665) of a neighboring subunit (Figure 3D). Of note, the ADPKD mutation R638C (http://pkdb.mayo.edu) is anticipated to abolish a cation-π interaction between Arg638 and Phe646 and thus likely weakens coupling between the pore helices of two adjacent subunits. This mutation might also perturb the structure of the pore helix 1 since Arg638 additionally forms a hydrogen bond with Thr635 via side chains (Figure 3D).

Structure of the VSD

While PKD2 grossly resembles TRPV1 and TRPA1 in the trans-membrane core, these channels nevertheless exhibit substantial structural differences that may account for their distinct gating mechanisms. Most notably, both TRPV1 (Liao et al., 2013) and TRPA1 (Paulsen et al., 2015) lack positively charged arginine or lysine residues along the S4 helix (gating charges) that are characteristic of VGICs (Long et al., 2007; Payandeh et al., 2011; Wu et al., 2016; Zhang et al., 2012). Furthermore, the S1–S4 domain of TRPV1, including its constituent S4 helix, remains stationary during channel gating (Cao et al., 2013), in sharp contrast to VGICs in which voltage-driven movement of the S4 helix gates channel conduction. In contrast to TRPV1 and TRPA1, PKD2 harbors two lysine residues (Lys572 and Lys575; conserved in both PKD2-L1 and PKD2-L2; Figure S6) at the cytoplasmic end of the S4 helix (Figure 3E). In PKD2, the last three turns of S4 adopt a 3₁₀-helical configuration, placing Lys572 and Lys575 on the same side of S4 with their side chains pointing into the interior of the S1–S4 domain. Such an unusually long 3₁₀-helix represents a hallmark feature of VGICs (Long et al., 2007; Payandeh et al., 2011; Vieira-Pires and Morais-Cabral, 2010; Wu et al., 2016; Zhang et al., 2012) and is not observed in TRPV1 and TRPV2 (Liao et al., 2013; Zubcevic et al., 2016). Lys572 and Lys575 are likely stabilized by Asp511, as well as two tyrosine residues (Tyr487 and Tyr227) protruding from other helices via a salt bridge and cation-π interactions, respectively, indicative of a voltage-sensor-like arrangement that may regulate PKD2 activity. Like PKD2-L1, the chimera channel is voltage dependent and conducts large tail currents upon membrane repolarization (Figures 2B and S4A) (DeCaen et al., 2013, 2016). Of note, the D511V mutation represents a frequent pathogenic missense variant found in ADPKD patients (Reynolds et al., 1999). Our structure now suggests that this mutation neutralizes a critical negatively charged counter ion for lysine residues and thus is expected to perturb the VSD structure. Taken together, these observations suggest that conformational changes in the PKD2 VSD, especially within the high-energy and dynamic 3₁₀-helical region of S4, might gate the ion conduction pore as in VGICs (Long et al., 2007; Zhang et al., 2012).

Structure of the Extracellular Polycystin Domain

The extracellular polycystin domain adopts a novel three-layered fold: the top layer consists of three α helices, the middle layer is a five-stranded anti-parallel β sheet, and the bottom layer can be analogized as two extended fingers (finger 1 and finger 2) (Figure 4A). The polycystin domain features two prominent grooves that are sites of intra- and inter-subunit interactions. One such groove is formed between finger 1 and the β4–β5 turn (⁴⁴⁴Glu-Val⁴⁵¹) and wraps around and interacts extensively with another β-turn structure (⁴³⁰Asn-Leu⁴³⁵ between β3 and β4) protruding from an adjacent polycystin domain of a neighboring subunit (Figure 4B). At an exterior side of this groove, the β4–β5 turn also contacts the H1 helix of an adjacent polycystin domain via a hydrogen bonding interaction mediated by Thr448 and Tyr248. Notably, a disulfide bond (C331–C344) likely constrains finger 1, allowing it to assume a rigid conformation that could be essential for efficient association with a neighboring polycystin domain (Figures 4A and S2I). Interestingly, both C331S and T448K mutations are associated with ADPKD (http://pkdb. mayo.edu), and our structure now suggests that these mutations likely perturb this essential inter-subunit interface and disrupt channel assembly. The extensive homotypic interactions mediated by the extracellular polycystin domain contrasts with channel assembly seen in group 1 TRP channels, such as TRPV1 and TRPA1, which form inter-subunit interactions via cytoplasmic domains (Liao et al., 2013; Paulsen et al., 2015).

Figure 4. Role of Polycystin Domain in Channel Assembly and Its Coupling with the Channel Transmembrane Core.

(A) Ribbon representation of the polycystin domain with specific structural elements denoted.

(B) Channel assembly mediated by the polycystin domain. Left, a groove formed between finger 1 and β4–β5 turn of one polycystin domain (surface) interacts with β3–β4 turn and H1 helix of a neighboring polycystin domain (ribbon). Right: zoomed view indicating detailed interactions.

(C) The polycystin domain (surface) interacts with the transmembrane core of the channel at two critical sites as indicated. (D and E) Zoomed views of (C) showing detailed interactions at helix turn (D) and pre-S6 loop (E).

See also Figures S3 and S6.

Another prominent groove within the polycystin domain interacts with the PKD2 transmembrane core at two classic allosteric regulatory sites: a helix-turn structure (⁵³⁰Ser-Phe⁵⁴⁸) that extends from the S3 and S4 helices at the peripheral VSD, and the pre-S6 loop that connects the pore helix 2 and porelining S6 helix at the outer pore region (Figure 4C). At the peripheral site, Arg320 on the β1 strand of the polycystin domain participates in cation-π interaction with Phe545 in the turn preceding S4 (Figure 4D). A series of hydrophobic residues on the β sheet layer of the polycystin domain additionally pack against their counterparts on a short helix extension following S3 (Figure 4C). At the outer pore region, Arg654 on the pre-S6 loop interacts with residues residing at the tip of finger 2, participating in cation-π interaction with Trp380 and forming a hydrogen bond with the carbonyl oxygen atom of Gly381 (Figure 4E). Given that equivalent sites are frequently targeted by gating modifier toxins and physiological modulators in related VGICs (Catterall et al., 2007) and TRPV1 (Bohlen et al., 2010; Cao et al., 2013), we speculate that the polycystin domain serves as a physical ‘‘lid’’ through which extracellular physical and/or chemical stimuli can allosterically modulate pore gating.

Molecular Basis of Pathogenic Mutations

Of the 26 known missense mutations identified in ADPKD patients (http://pkdb.mayo.edu), 22 mutations can be mapped onto the resolved core structure of PKD2 (Figure 5A). The other four sites, located in the C terminus, were not included in our constructs. Interestingly, no pathogenic missense mutations have been identified in the N terminus to date, suggesting this unstructured region has less relevance to human disease. In contrast, the extracellular polycystin domain and pore helix 1 are two mutation hotspots with 15 and three pathogenic missense mutations, respectively, likely reflecting their critical roles in channel assembly and/or gating. Our PKD2 structure now rationalizes how clinically relevant pathogenic variants, especially those within the polycystin domain, might perturb channel structure and function. For instance, three aromatic residues (W280, Y292, and W414) are buried inside the hydrophobic interior of the polycystin domain and six surface residues (R322, R325, Y345, S349, T419, and R420) all participate in hydrogen bonding interactions with neighboring residues (Figure 5B). Mutations of these residues are predicted to disrupt the hydrophobic core or abolish essential packing interactions of the polycystin domain, resulting in loss of channel function.

Figure 5. Structural Annotation of Pathogenic PKD2 Mutations.

(A) Mapping of human disease-associated missense mutations (red) onto the structure of a PKD2 subunit. Note that the polycystin domain (blue) and pore helix (purple) are mutation hotspots.

(B) Missense mutations within the polycystin domain. Note that the majority of these mutated residues (labeled in red) participate in hydrogen bonding interactions denoted with black dash lines or are buried inside a hydrophobic interior.

Potential Mechanisms of Channel Gating

This closed state PKD2 structure only hints at conformational changes that might occur upon channel opening. We suspect that two high-energy helices within the transmembrane core may play a role in channel gating (Figure 6). First, PKD2 harbors a π-helix in the middle of S6 where Phe670 forms a conspicuous bulge. The π-helix is a rare secondary structural element often found in functional sites of proteins due to its dynamic and destabilizing nature as compared to canonical α helices (Cooley et al., 2010). It is thus conceivable that this π-helix might function as a flexing point, around which the bottom half of the S6 helix can bend to open the lower gate of PKD2. Second, as discussed above, the last nine residues (⁵⁷¹Ile-Phe⁵⁷⁹) of S4 adopt a long, stretched 3₁₀-helical conformation, which similar to π-helices is also rare and structurally dynamic (Vieira-Pires and Morais-Cabral, 2010). We therefore suspect that structural rearrangements within this 3₁₀-helical region can accommodate or drive gating-associated movements of the S4–S5 linker, which is directly coupled to the pore-lining S6 helix via physical interactions (e.g., a hydrogen bond formed between Asn585 and Lys688; conserved in PKD2-L1 and PKD2-L2; Figure S6).

Figure 6. A Hypothetical Model for PKD2 Gating.

(A) Structural features that may play a role in PKD2 gating are shown, including high-energy 3₁₀- and π-helices (cyan) and a hydrogen bonding interaction that couples the S4–S5 linker (purple) with the pore-lining S6 helix. Residues that form the lower gate (L677 and N681) and lysine residues residing at the 3₁₀ helical region of S4 are also shown.

(B) A hypothetical model depicting conformational changes at the 3₁₀ and π-helices, shown, respectively, as helices and stars in cyan, drive or accommodate movements of the S4–S5 linker and bending of the S6 helix to open the lower gate. See also Figure S6.

In summary, we present the structure of PKD2, an apparent Na⁺/K⁺ conducting nonselective cation channel. PKD2 features a large extracellular polycystin domain, a potential ‘‘lid’’ that sits atop and interacts with the transmembrane domain and contributes to channel assembly via extensive homotypic interactions. This architecture differs significantly from VGICs and other TRPs, but shares interesting parallels with various ligand-gated ion channels (e.g., P2X, AMPA, and 5-HT channels). We speculate that this feature affects access by unknown physiological ligand(s) that modulate gating of the pore. Our structures also reveal that the cytoplasmic domains of PKD2, in the absence of regulatory cellular factors, are unlikely to directly affect channel gating owing to a lack of structural coupling with the ion conduction pore. The PKD2 structure will aid molecular diagnoses by helping to differentiate pathogenic mutations from naturally occurring neutral variants. Finally, successful biochemical purification of the PKD2 channel will enable reconstitution into liposomes for discovery of pharmacological agents (Su et al., 2016) to serve as physiological tools and lead compounds for the treatment of ADPKD.

STAR★METHODS

Detailed methods are provided in the online version of this paper and include the following:

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, Peptides, and Recombinant Proteins
n-Dodecyl-β-D-Maltopyranoside (DDM)	Anatrace	D310S
Soy extract polar	Avanti	541602C
Amphipols A8–35	Antrace	A835
Sodium butyrate	Alfa Aesar	A11079
FreeStyle 293 expression medium	Invitrogen	12338018
Triptolide	Sigma Aldrich	T3652-1MG
Wnt-3a	R&D	5036-WN/CF
Wnt-9b	R&D	3669-WN/CF
Critical Commercial Assays
Superose 6, 10/300 GL	GE Healthcare Life Sciences	17-5172-01
Amylose resin	New England Biolab	E8021L
Bio-beads	Bio-Rad	1523920
Deposited Data
The coordinate of human PKD2 channel	This study	PDB: 5T4D
Cryo-EM map of human PKD2:198–703	This study	EMDB: EMD-8354
Cryo-EM map of human PKD2:198–792	This study	EMDB: EMD-8355
Cryo-EM map of human PKD2:53–792	This study	EMDB: EMD-8356
Experimental Models: Cell Lines
HEK293S GnTr−/−	ATCC	CRL3022
HEK293T	ATCC	CRL-3216
CHO	ATCC	CCL-61
LLC-PK1	ATCC	CL-101
Recombinant DNA
human PKD2:198–703 cloned onto a modified pFastbac1 vector suitable for expression in mammalian cells	This study	N/A
human PKD2:198–792 cloned onto a modified pFastbac1 vector suitable for expression in mammalian cells	This study	N/A
human PKD2:53–792 cloned onto a modified pFastbac1 vector suitable for expression in mammalian cells	This study	N/A
pMSP2N2	Addgene	Plasmid #29520
Human PKD1 pTracer IRES	This study	N/A
Human PKD2 pTracer IRES	This study	N/A
Human PKD2 F604P pTracer IRES	This study	N/A
Sequence-based Reagents
Primer:hPKD2–53 Forward: GAGATCGAGATGCAGCGCATC	N/A	University of Utah HSC DNA peptide core
Primer:hPKD2–198 Forward: CGAGGTCTCTGG GGAACAAGA	N/A	University of Utah HSC DNA peptide core
Primer:hPKD2–792 Reverse: TCAATCCAAATCCAGGTCCTCCCT	N/A	University of Utah HSC DNA peptide core
Primer:hPKD2–703 Reverse: TCAAAGATCTGAGAGTTCCATTTC	N/A	University of Utah HSC DNA peptide core
Primer:hPKD2-F604P Forward:GGCTTTGCTAT TATGCCCTTCATTATTTTCCTA	N/A	University of Utah HSC DNA peptide core
Primer:hPKD2-F604P Reverse: TAGGAAAATAATGAAGGGCATAATAGCAAAGCC	N/A	University of Utah HSC DNA peptide core
Software and Algorithms
UcsfDfCorr	Zheng et al., 2016	http://biorxiv.org/content/early/2016/07/04/061960
RELION 1.4	Scheres, 2012	http://www2.mrc-lmb.cam.ac.uk/relion
SamViewer	Maofu Liao	http://liao.hms.harvard.edu
HOLE	Smart et al., 1996	http://www.smartsci.uk/hole
CTFFIND4	Rohou and Grigorieff, 2015	http://grigoriefflab.janelia.org/ctf
COOT	Emsley et al., 2010	http://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
Molprobity	Chen et al., 2010	http://molprobity.biochem.duke.edu/
EMringer	Barad et al., 2015	http://fraserlab.com/2015/02/18/EMringer
ResMap	Kucukelbir et al., 2014	http://resmap.sourceforge.net
phenix.real_space_refine	Adams et al., 2010	https://www.phenix-online.org/documentation/reference/real_space_refine.html
UCSF CHIMERA	Pettersen et al., 2004	http://www.cgl.ucsf.edu/chimera/
Other
Grids: Cu 400 mesh	Quantifoil	Q20474

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for reagents may be directed to, and will be fulfilled by Erhu Cao (erhu.cao@biochem.utah.edu).

METHOD DETAILS

Protein expression, purification, and nanodisc reconstitution

Three truncated biochemically stable constructs (hPKD2:198–703, hPKD2:53–792, and hPKD2:198–792) were expressed with an N-terminal maltose binding protein (MBP) fusion in HEK293S GnTI^−/− cells using the BacMam system as described, with slight modifications (Goehring et al., 2014; Liao et al., 2013). In brief, HEK293S GnTI^−/− cells, grown in suspension in Freestyle 293 expression medium (Invitrogen) at 37°C in an orbital shaker, were transduced with baculovirus when cell density reached ~2 × 10⁶/ml. 8–24 hr post transduction, sodium butyrate was added to the culture to a final concentration of 5 mM to boost protein expression; temperature was reduced to 30°C. Cells were harvested 72 hr post-transduction for preparation of crude membrane and subsequent affinity purification with amylose resin (New England BioLabs). hPKD2:198–703 protein was eluted from amylose resin with buffer composed of 50 mM HEPES (pH 7.4), 150 mM NaCl, 2 mM TCEP, 0.5 mM DDM, 20 mM maltose, and 0.1 mg/ml soybean lipids. Membrane scaffold protein MSP2N2 was expressed and purified from E. coli as described (Civjan et al., 2003; Ritchie et al., 2009). Nanodisc reconstitution was performed as described (Gao et al., 2016). In brief, soybean polar lipid extract (Avanti) dissolved in chloroform was dried under an argon stream and residual chloroform evaporated by vacuum desiccation (~3 h). Lipids stock [10 mM] was prepared by resuspending dried lipids in a buffer composed of 20 mM HEPES (pH 7.4), 150 mM NaCl, and 2 mM TCEP via bath sonication. Purified PKD2 channel (2–3 mg/ml) solubilized in 0.5 mM DDM was mixed with MSP2N2 (~170–250 μM) and the soybean lipid stock at a 1:1:200 molar ratio and incubated on ice for 30 min. Four batches of Bio-beads SM2 (30 mg per 1 ml reconstitution mixture; Bio-Rad) were added to remove detergents from the system at 1 hr intervals at 4°C with gentle rotation. TEV protease (40 μg/mg PKD2) was added together with the final batch of Bio-beads and the sample was incubated at 4°C overnight. Bio-beads were then removed and the reconstitution mixture was cleared by centrifugation before subsequent separation on a Superose 6 column (GE Heath Care) in buffer (10 mM HEPES, 150 mM NaCl, [pH 7.4]). The peak corresponding to tetrameric PKD2 channel reconstituted in lipid nanodiscs was collected for electron cryo-microscopy analyses.

A similar protocol was used to reconstitute hPKD2:198–792 into nanodiscs, except that 2 mM CaCl₂ was included in all buffers throughout purification since calcium was reported to stabilize the isolated EF hands of PKD2 (Allen et al., 2014; Petri et al., 2010). hPKD2:53–792 was reconstituted into amphipols A8–35 (Anatrace) as described (Liao et al., 2013).

EM data acquisition

3.5 μl of sample (PKD2 nanodiscs at ~1–3.5 mg/ml or hPKD2:53–792 stabilized in amphipols at ~1 mg/ml) was applied to glow-discharged Quantifoil 1.2/1.3 holey carbon 400 mesh copper grids. Grids were plunge frozen in liquid ethane using a Vitrobot Mark III (FEI) set to 4°C, 75% relative humidity, 20 s wait time, −1 mm offset, and 7 s blotting time.

hPKD2:198–703, hPKD2:198–792 in nanodiscs, and hPKD2:53–792 in amphipols were automatically collected at University of Utah using SerialEM (Mastronarde, 2005) on a Tecnai TF20 TEM (FEI) operating at 200 kV using a Gatan 626 side entry cryo-holder. Movies were recorded using a K2 Summit direct detector (Gatan) at a corrected magnification of 41,911 ×, corresponding to a pixel size of 1.193 Å, and at a dose rate of ~8 e/pixel/s. Each movie was recorded as a stack of 60 subframes, each of which was accumulated for 0.2 s, totaling ~65–70 electrons per Ų. Defocus values ranged between 0.8 to 2.0 μm.

To achieve higher resolution, we collected an additional dataset for hPKD2:198–703 on a Tecnai TF30 Polara (FEI) operating at 300kV equipped with the K2 Summit direct electron detector at UCSF. Images were recorded using UCSFImage4 (Barad et al., 2015), with a defocus range between 0.6 to 2.4 μm. Specifically, we recorded movies in super-resolution counting mode at a magnification of 31,000 3, which corresponds to a physical pixel size of 1.2156 Å. The beam intensity was set to ~8.2 counts/physical pixel/s, corresponding to a dose rate of ~10 electrons per physical pixel per second. Movies were recorded as a stack of 40 sub-frames, each of which was accumulated for 0.2 s, resulting in an average of ~1.4 counts per physical pixel in each subframe and a total specimen dose of ~54 electrons per Ų.

Image processing and 3D reconstruction

Movie frames were aligned, dose weighted, and then summed into a single micrograph using UcsfDfCorr (program written by Shawn Zheng, UCSF). CTF parameters for micrographs were determined using the program CTFFIND4 (Rohou and Grigorieff, 2015). Micrographs with poor CTF cross correlation scores were excluded from downstream analyses.

Particle selection was performed in SamViewer (program written by Maofu Liao, Harvard Medical School) using 2D class average templates generated from an initial round of manual boxing. All 2D and 3D processing steps were performed using RELION (Scheres, 2012). Particles were sorted into 2D classes, followed by rejecting ‘bad’ particles (i.e., those sorted into incoherent or poorly resolved classes) from downstream analyses.

For the initial TF20 dataset collected for hPKD2:198–703, ~55,000 particles were extracted; following 2D classification, ~45,000 particles were used for 3D reconstruction with C4 symmetry imposed given that the PKD2 channel exhibits 4-fold symmetry as other TRPs and related voltage-gated ion channels. The structure of TRPV1 (low-pass filtered to 60 Å) was used as an initial reference, resulting in a 4.2-Å map via gold-standard refinement (Chen et al., 2013) (data not shown).

We collected a cryo-EM dataset of hPKD2:198–792 with intact EF hands in the presence of 2 mM CaCl₂. ~156,000 particles were extracted from 1003 micrographs recorded on our TF20 equipped with a K2 Summit direct electron detector. After extensive 2D and 3D classification to reject bad particles, we retained 35,462 particle images for 3D reconstruction with C4 symmetry imposed. The resulting 4.0-Å reconstruction was indistinguishable from our initial reconstruction of PKD2:198–703 without the EF hands (Figure S1). Of note, our 2D classes reveal fuzzy densities at the expected locations of the EF hands, suggesting that the domain is likely separated from the rest of the particle by a flexible linker (Figure S1).

We collected a cryo-EM dataset of hPKD2:53–792, a construct that retains most functional domains of the channel, in amphipols A8–35 in the absence of Ca²⁺. ~290,000 particles were extracted from 794 micrographs recorded on our TF20 microscope equipped with a K2 Summit direct electron detector. After 2D and 3D classification to reject bad particles in Relion, we retained 66,988 particles for 3D auto-refinement in RELION with C4 symmetry imposed. The resulting 4.0-Å map was again superimposable with our initial reconstruction of hPKD2:198–703.

The TF30 Polara data collected for hPKD2:198–703 was similarly processed as described above. Specifically, 368,032 particles were selected from 1,500 micrographs through automatic picking followed by interactively removing ice and junk particles in SamViewer. A round of 2D classification followed by another round of 3D classification was performed to reject bad particles from downstream analyses (i.e., in silico purification), resulting in a final dataset of ~93,805 particles that was subsequently used for 3D reconstruction. For 3D classification and refinement, the initial 4.2-Å PKD2:198–703 structure (low-pass filtered to 40 Å) from the TF20 dataset was used as the starting model with C4 symmetry imposed. The auto-refinement procedure in RELION converged at an unmasked resolution of 3.38 Å. RELION post-processing using unfiltered half maps with automatic mask and b factor settings produced a 3.0-Å resolution map sharpened with a b factor of −87 Ų.

UCSF Chimera was used to visualize and segment density maps, as well as to generate figures (Pettersen et al., 2004). Local resolution of each map was computed using ResMap (Kucukelbir et al., 2014), using unfiltered map halves as outputted by RELION. Pore radii were calculated using the HOLE program (Smart et al., 1996).

Model building, refinement, and validation

At 3.0 Å resolution, the cryo-EM map was of sufficient quality for de novo atomic model building in Coot (Emsley et al., 2010) except for the disordered N terminus (198–215), a loop (296–301) within the extracellular polycystin domain, S2–S3 loop (494–503), and C terminus (695–703). Amino acid assignment was achieved based mainly on the clearly defined densities for bulky residues (Phe, Trp, Tyr, and Arg) and the absence of side densities for glycine residues. The PKD2 model was then subjected to global refinement and minimization in real space using the module ‘phenix.real_space_refine’ in PHENIX (Adams et al., 2010). The geometries of the model were assessed using and MolProbity in ‘comprehensive model validation’ section in PHENIX (Chen et al., 2010) and EMRinger (Barad et al., 2015), and detailed information listed in Table S1.

The atomic model was tested for overfitting using previously described methods (Amunts et al., 2014; Zhao et al., 2015). In brief, the coordinates were randomly displaced by 0.2 Å using PHENIX. The displaced model was refined in PHENIX against one of the half maps derived from half of the dataset used in RELION auto-refinement (half-map 1, ‘work’). FSC curves were calculated between the refined model and half-map 1, half-map 2 (‘free’), and the summed map. The general agreement between the half-map FSC curves (work versus free) suggests that the model was not overfitted (Figure S2D).

Voltage clamp experiments

Electrophysiological recordings were performed as described with modifications (DeCaen et al., 2016). HEK293T, CHO and LLC-PK1 cells were transiently transfected with mammalian cell expression vectors pTracer IRES Cherry and/or GFP subcloned with the human versions of PKD2 and/or PKD1 genes. Cells were seeded onto glass coverslips and placed in a perfusion chamber, enabling changes in extracellular conditions. Data were generated from cells patched in the whole-cell configuration with < 1 GΩ of resistance. Cell-attached single channel data with seal resistance < 8 GΩ were not analyzed. Whole-cell and excised inside-out patch currents were digitized at 25 kHz and low pass filtered at 10 kHz. For divalent free (DVF) symmetrical sodium conditions, the pipette electrode solution contained (in mM): NaMES (120), HEPES (10), Na₄-BAPTA (5), EGTA (0.5) and pH was adjusted to 7.4 with NaOH. The standard bath solution contained, in mM: NaCl (140), HEPES (10) and pH was adjusted with NaOH. The same solutions were used in the physiological Ca²⁺ conditions, but CaCl₂ was added to achieve 1.8 mM Ca²⁺ external and 90 nM free-Ca²⁺ internal (MaxChelator). When testing the relative permeability of monovalent cations, the bath solution contained (in mM): X-Cl (150), HEPES (10) EGTA (2) and the pH was adjusted with X-OH where X is the indicated monovalent cation. The extracellular NMDG solution was formulated with NMDG-MES (150), HEPES (10), EGTA (2) and the pH was adjusted with CH₃SO₃H. When testing the relative permeability of Ca²⁺, the bath solution contained: CaCl₂ (100), HEPES (10), and pH was adjusted with Ca(OH)₂. All saline solutions were adjusted to 300 mOsm (±5) with mannitol, if needed. Data were analyzed by Igor Pro 7.00 (Wavemetrics, Lake Oswego). The reversal potential, E_rev, was used to determine the relative permeability of monovalent cation X to Na⁺ (P_X/P_Na) according to the following equation:

$$\frac{P_x}{P_{Na}}=\frac{{\alpha }_{Na}}{{\alpha }_x}\left[\exp \left(\frac{\Delta E_{\mathit{rev}}}{RT/F}\right)\right]$$

where E_rev, α_x, R, T and F are the reversal potential, effective activity coefficients for cation x, the universal gas constant, absolute temperature, and the Faraday constant, respectively. The effective activity coefficients (α_x) were calculated using the following equation:

$${\alpha }_x={\gamma }_x[X]$$

where α_x is the activity coefficient and [X] is the concentration of the ion. For calculations of membrane permeability, activity coefficients (γ) were calculated using the Debye-Hückel equation: 0.74, 0.72, and 0.29 correspond to Na⁺, K⁺, and Ca²⁺, respectively. To determine the relative permeability of calcium cations to Na⁺, the following equation was used:

$$\frac{P_{Ca}}{P_{Na}}=\frac{\left\{{\alpha }_{Na}\left[\exp \left(\frac{E_{\mathit{rev}}F}{RT}\right)\right]\left[\exp \left(\frac{E_{\mathit{rev}}F}{RT}\right)+1\right]\right\}}{4{\alpha }_{Ca}}$$

E_rev for potassium and calcium conditions were corrected by the measured liquid junction potentials, −2.4 and 1.4 mV, respectively.

Effect of Triptolide, Wnt-3a and Wnt-9b, and F604P mutation

Whole cell patch clamp recordings were obtained in physiologicalcalcium conditions. Triptolide (Sigma Aldrich)and carrier-free versions of Wnt-3a and Wnt-9b (R&D) were formulated in water and diluted 1000–3000 × to concentrations of 300 nM and 1 μg/ml (~27 mM). Percent change of the outward current was determined by the following equation: % Change I_out = (I_drug-I_control/I_control) × 100 where I_control is the average current measured during the 30 s prior to drug application and I_drug is the average current after 60–120 s of drug application. The human PKD2 F604P mutant (Arif Pavel et al., 2016) was recorded in DVF symmetrical 140 mM Na⁺ conditions. Non-transfected HEK293T cells exhibit small outwardly rectifying currents in physiological calcium conditions, which become larger and ohmic (linear) in divalent free conditions. These intrinsic currents are largely attributed to native chloride and non-selective cationic TRPM7 channels in these cells (Nörenberg et al., 2016).

Single channel recordings

Single channel events used to determine the conductance of monovalent ions (Figure S5) were recorded in the inside-out configuration, digitized at 25 kHz and low pass filtered at 10 kHz. Single channel monovalent and Ca²⁺ currents were analyzed after low pass Gaussian filtering (1 KHz and 0.4 Hz respectively). Pipette and bath conditions contained (in mM): XCl (140), HEPES (10) and EGTA (2) where X is the indicated monovalent cation. Single channel calcium currents were measured using symmetrical 100 mM CaCl₂ and 10 mM HEPES. Inward single channel events were recorded in divalent-free symmetrical sodium conditions with holding potential = −90 mV.

QUANTIFICATION AND STATISTICAL ANALYSIS

Quantification and statistical analyses pertain to the analysis of cryo-EM data are integral parts of algorithms and software used. Two tailed Student’s t tests were used to determine the lack of effect of Wnt-3a, Wnt-9b and triptolide on current magnitude from transfected and untransfected HEK293T and CHO cells. The same statistical test was used to evaluate the difference in current magnitudes in response to changes in voltage from transected and transfected cells.

DATA AND SOFTWARE AVAILABILITY

Data Resources

The accession numbers for the data reported in this paper are EMD: EMD-8354, EMD-8355, EMD-8356 and PDB: 5T4D.

Highlights.

• 3.0-Å resolution cryo-EM structure of PKD2 in lipid nanodiscs

• Two gates regulate ion permeation

• The extracellular polycystin domain contributes to channel assembly

• The structure reveals the molecular basis of the majority of pathogenic mutations