Full-length P2X₇ structures reveal how palmitoylation prevents channel desensitization

SUMMARY

P2X receptors are trimeric, non-selective cation channels activated by extracellular ATP. The P2X₇ receptor subtype is a pharmacological target because of involvement in apoptotic, inflammatory, and tumor-progression pathways. It is the most structurally and functionally distinct P2X subtype, containing a unique cytoplasmic domain critical for the receptor to initiate apoptosis and not undergo desensitization. However, lack of structural information about the cytoplasmic domain has hindered understanding of the molecular mechanisms underlying these processes. We report cryo-electron microscopy structures of full-length rat P2X₇ receptor in apo and ATP-bound states. These structures reveal how one cytoplasmic element, the C-cys anchor, prevents desensitization by anchoring the pore-lining helix to the membrane with palmitoyl groups. They show a second cytoplasmic element with a unique fold, the cytoplasmic ballast, which unexpectedly contains a zinc-ion complex and a guanosine nucleotide-binding site. Our structures provide first insights into architecture and function of a P2X receptor cytoplasmic domain.

Graphical Abstract

In Brief Statement:

Structures of the full-length P2X₇ receptor provide the first insights into the architecture and function of a P2X receptor cytoplasmic domain, which includes a unique cytoplasmic ballast fold containing a zinc ion complex and a guanosine nucleotide binding site, as well as an understanding of how palmitoylation of the C-cys anchor prevents receptor desensitization.

INTRODUCTION

Extracellular ATP signals through two main classes of membrane proteins, known as purinergic receptors (Burnstock, 1976): ligand-gated P2X receptor ion channels (Brake et al., 1994; Valera et al., 1994) and G-protein coupled P2Y receptors (Webb et al., 1993). P2X receptors are particularly promising pharmacological targets (Burnstock and Knight, 2018; North and Jarvis, 2013), as they are found in a variety of eukaryotic cells (Fountain and Burnstock, 2009) and mediate processes such as platelet activation, smooth muscle contraction, synaptic transmission, nociception, inflammation, and apoptosis (Burnstock and Kennedy, 2011).

Seven mammalian P2X subunits (denoted P2X₁ to P2X₇) co-assemble in different combinations to form homo- and hetero-trimeric non-selective cation channels (Brake et al., 1994; North, 2002; Surprenant et al., 1996; Valera et al., 1994). All subunits contain intracellular N- and C-termini, two transmembrane (TM) helices that form the pore, and an extracellular domain containing the orthosteric ATP binding site (Habermacher et al., 2015; North, 2002). Although ATP activates each P2X receptor subtype, its affinity varies significantly from nanomolar (P2X_1,3), through low micromolar (P2X_2,4), to hundreds of micromolar (P2X₇). In addition, receptor subtypes have different rates of desensitization, ranging from fast (milliseconds in P2X_1,3), through slow (seconds in P2X_2,4), to complete lack of desensitization (P2X₇) (Jarvis and Khakh, 2009; Koshimizu et al., 1999). The significantly lower affinity for ATP and complete lack of desensitization set the P2X₇ receptor apart from any other P2X receptor family member. Moreover, P2X₇ is the only subtype that acts as a cytotoxic receptor in multiple signaling pathways (Costa-Junior et al., 2011; Surprenant et al., 1996).

Despite the importance of P2X receptors, the molecular principles governing their pharmacology, gating, and signal transduction are incompletely understood. The cytoplasmic termini of P2X receptors are thought to play a key role in receptor desensitization (Allsopp and Evans, 2011; Hausmann et al., 2014; Koshimizu et al., 1998; Mansoor et al., 2016; Smith et al., 1999; Werner et al., 1996), but the complete molecular mechanism by which they modulate desensitization is unclear due to a lack of structural information. Structures of several P2X receptor subtypes exist (Hattori and Gouaux, 2012; Karasawa and Kawate, 2016; Kasuya et al., 2017; Kawate et al., 2009; Li et al., 2019; Mansoor et al., 2016), however, the N- and C-terminal truncation constructs used in those studies as well as the flexibility of the cytoplasmic domain have precluded visualization of cytoplasmic residues. To date, the only structural insight into any P2X receptor cytoplasmic domain has been obtained from a minimally-truncated human P2X₃ (hP2X₃) receptor (Li et al., 2019; Mansoor et al., 2016). This structure revealed how N-and C-terminal residues form a dynamic structural domain termed the ‘cytoplasmic cap’, the stability of which dictates the lifetime of the open pore conformation and therefore sets the rate of receptor desensitization (Mansoor et al., 2016). In contrast to other P2X subtypes, P2X₇ has a unique cytoplasmic domain with two additional elements (Surprenant et al., 1996): a cytoplasmic cysteine-rich (C-cys) region (Allsopp and Evans, 2015) at the end of the second transmembrane domain (TM2) that we name the C-cys anchor; and an additional 120 residues in the C-terminus that we name the cytoplasmic ballast. The cytoplasmic ballast, specifically, has been implicated in modulating P2X₇ receptor’s ability to undergo pore dilation and initiate cytolytic signal transduction (Cheewatrakoolpong et al., 2005; Costa-Junior et al., 2011; Surprenant et al., 1996). Until now, no structural information has been available for either of these cytoplasmic elements.

To understand the distinct functional properties of the P2X₇ receptor and investigate the structure of its large and enigmatic cytoplasmic domain, we used single-particle cryo-electron microscopy to solve the structures of full-length rat P2X₇ (rP2X₇) in both apo (closed-pore) and ATP-bound (open-pore) states. Our structures suggest why the P2X₇ receptor subtype has a low affinity for ATP, demonstrate that palmitoylation of residues in the C-cys anchor prevents receptor desensitization, and reveal a novel fold for the cytoplasmic ballast. Unexpectedly, we find a dinuclear zinc ion complex and a high-affinity guanosine nucleotide binding site in the cytoplasmic ballast, features for which the function is yet unknown, but potentially provide a new framework for future experiments to explain the unique signaling properties of P2X₇ receptors (Cheewatrakoolpong et al., 2005; Costa-Junior et al., 2011; el-Moatassim and Dubyak, 1992; Humphreys and Dubyak, 1996; Surprenant et al., 1996; Ugur and Ugur, 2019).

RESULTS

Structure Determination.

To gain insight into the structure of mammalian P2X₇ receptors, full-length, wild type rat P2X₇ receptor (rP2X₇-WT) was reconstituted into detergent micelles for single-particle cryo-electron microscopy (Figure S1). As expected, rP2X₇-WT was activated by ATP and did not undergo desensitization when expressed in Xenopus oocytes (Figure S2A–B). The three-dimensional reconstructions of the apo (closed state, Figure 1) and ATP-bound (open state, Figure S2C–J) structures of rP2X₇-WT were resolved to an estimated 2.9 Å and 3.3 Å, respectively (Figures S3–S4). The final models had good stereochemistry (Table S1) that correlated well with the respective density maps (Figure S3C,E).

Figure 1. Overall Architecture of rP2X₇ in the Apo State.

(A-D) The three-dimensional reconstruction of rP2X₇ in the apo state viewed parallel to the membrane (A), perpendicular to the membrane from the extracellular surface (B), from the center of the transmembrane domain, toward the cytoplasm (C), and from the intracellular side of the membrane, toward the cytoplasm (D). (E) Ribbon representation of a receptor corresponding to the apo state reconstruction shown in (A). Each protomer is colored differently. (F) A close-up side view of the cytoplasmic portion of the transmembrane domain highlighting how the cytoplasmic cap is composed of domain-swapped β-strands (β₋₁, β₀, and β₁₅) from each protomer. To improve visual clarity of the cytoplasmic cap, the C-cys anchor and cytoplasmic ballast have been removed. (G) Top-down cross-sectional view of the cytoplasmic cap looking toward the cytoplasm. Residues 29–345 and 395–595 of the receptor have been removed to improve visual clarity. Each protomer is a different color. (H) To better visualize the spatial interactions between protomers, one protomer is shown in blue while the other two are shown in gray. (I) Ribbon representation of a single subunit of rP2X₇ receptor, highlighting P2X receptor domains, including two new domains found only in the P2X₇ subtype, the C-cys anchor and the cytoplasmic ballast. The major domains within a protomer are organized by color: extracellular domain in blue, transmembrane domain in green, cytoplasmic cap in cyan, C-cys anchor in purple, and the cytoplasmic ballast in red. The palmitate groups are colored in sand.

Overall Architecture of rP2X₇.

The rP2X₇ receptor shares the same trimeric architecture as other P2X receptor subtypes (Hattori and Gouaux, 2012; Kasuya et al., 2017; Kawate et al., 2009; Mansoor et al., 2016). Each protomer has a hydrophilic extracellular domain, two transmembrane-spanning α-helices, and intracellular termini (Figure 1). However, unlike other P2X receptor subtypes, P2X₇ possesses a large cytoplasmic domain that has been visualized here for the first time in both apo (Figure 1A,D) and open state (Figure S2C,F) conformations. The three protomers intertwine by extensive domain swapping such that each protomer undergoes a relative ~ 120° clockwise rotation from the extracellular domain to the transmembrane domain, followed by a ~ 120° counterclockwise rotation from the transmembrane domain to the cytoplasmic domain (Figure 1B–D, Figure S2D–F).

As previously observed in the open state of hP2X₃ (Mansoor et al., 2016), rP2X₇ contains a domain-swapped cytoplasmic cap formed by secondary structural elements from each of the three protomers: sequential β-strands (β₀ and β₋₁) in the N-terminus of two protomers and a β-strand (β₁₅) after TM2 in the third protomer (Figure 1E–I, Figure S2G–J). The cytoplasmic cap is located beneath the TM domain and its formation and stability are proposed to govern the kinetics of desensitization in the fast-desensitizing hP2X₃ receptor (Mansoor et al., 2016). However, unlike hP2X₃ in which the cytoplasmic cap dynamically forms during receptor activation and is visible only in the open state, the cytoplasmic cap of rP2X₇ is present and visible in both the apo (Figure 1E–I) and open states (Figure S2G–J) and is thus a permanent structural scaffold in this non-desensitizing P2X receptor subtype.

The P2X₇ protomers intertwine to form a trimeric receptor (Figure 1A–E,H, Figure S2C–G), such that the N-terminus of each protomer interacts with the C-cys region at the cytoplasmic end of TM2 from an adjacent protomer and the long C-terminal domain of each protomer (final ~ 200 residues) hangs under the cytoplasmic cap of an adjacent protomer, as if wrapped around the TM helices (Figure 1A,E,H,I, Figure S2C,G,H). We refer to the C-terminal domain of each P2X₇ protomer as the ‘cytoplasmic ballast’ (Figure 1I, Figure S2H) because of the way it hangs into the cytoplasm beneath the receptor’s transmembrane domain, as if providing a counterbalance to the large extracellular domain.

Our structures also reveal the palmitoylation of at least one residue in the N-terminus (Figure 1E,H,I) as well as multiple residues in the cytoplasmic C-cys region (Figure 1E,H,I, Figure S2G–H), supporting the previous proposal that the C-cys region modulates receptor function (Allsopp and Evans, 2015) by bridging the interaction between P2X₇ and membrane phospholipid rafts (Gonnord et al., 2009; Robinson et al., 2014). We refer to the C-cys region as the ‘C-cys anchor’ because the palmitoyl groups link it, and the cytoplasmic end of TM2, to the membrane.

The P2X₇ Channel Pore Resembles Other P2X Receptors.

To assess the functional state of each rP2X₇ structure, we analyzed the occupancy of the orthosteric ATP binding site and the conformation of the central pore, along with the cavities, vestibules and fenestrations throughout the receptor (Figure S5A–B). In the absence of ATP, the pore is closed and the structure represents an apo, closed state of rP2X₇ (Figure 2A,C,E). Addition of ATP results in an open pore structure in which ATP occupies the orthosteric binding pocket, representing an agonist-bound, open state (Figure 2B,D,F). The structural changes in the extracellular domain of rP2X₇ that occur upon agonist binding and channel opening are conserved with P2X₃ (Mansoor et al., 2016) and P2X₄ (Hattori and Gouaux, 2012; Kawate, 2017; Kawate et al., 2009).

Figure 2. Ion Channel Pore.

(A-F) Cartoon representation of rP2X₇ structures shown parallel to the membrane as a side view, top down view of the ion channel pore, and ion permeation pathway, respectively, are drawn comparing the apo, closed state (A,C,E) to the ATP-bound, open state (B,D,F). Binding of agonist induces conformational changes around the binding pocket that result in outward flexing of the extracellular domain’s core β-strands, pulling on TM2, causing outward helix expansion to open the pore. For the pore size plots, colors represent different radii: reddish pink < 1.15 Å, green between 1.15 – 2.30 Å, and purple > 2.30 Å. The N-terminus, TM1, and cytoplasmic ballast have been removed from panels (E) and (F) to improve visual clarity.

TM2 lines the pore lumen of rP2X₇, and residues Q332, V335, S339, and S342 face the pore in the apo state (Figure 2E). Residue S339 defines the extracellular boundary and S342 the cytoplasmic boundary of the TM gate, which has a maximum radius of 0.1 Å in the closed, apo state (Figure S5C). This is consistent with previous work suggesting that the ion selectivity filter of hP2X₇ occurs around residue S342 (Pippel et al., 2017). The radius of the gate is too narrow to pass dehydrated Na⁺ ions (Degrève et al., 1996) and therefore defines the ion channel as closed (Figure 2C,E). The size and shape of the apo state pore of rP2X₇, as well as the position of the TM gate, are similar to the apo state pore of hP2X₃ (Mansoor et al., 2016).

The ATP-bound structure of rP2X₇ has a continuous pore throughout the transmembrane domain with a minimum radius of 2.5 Å (Figure S5C). This pore is large enough to pass partially hydrated Na⁺ ions (Degrève et al., 1996), and thus defines the ion channel gate as open (Figure 2D,F). Compared to the apo structure, S339 is translated toward the extracellular surface and rotated away from the pore’s center in the open state structure, opening the pore via a mechanism that is reminiscent of the hP2X₃ receptor (Supplementary Movie M1–M2) (Mansoor et al., 2016). Residue S342, which defines the cytoplasmic boundary of the closed gate in the apo state, also defines the narrowest region of the pore in the open state of rP2X₇. The narrowest region of the open hP2X₃ pore is defined by a threonine residue at the analogous position (Mansoor et al., 2016). In the rat P2X₂ receptor, the threonine residue at this position has been implicated in ion selectivity (Migita et al., 2001), suggesting that S342 in rP2X₇ receptor might interact with permeating cations. Once beyond the TM gate, ions likely egress through the cytoplasmic fenestrations (Figure 2E–F) (Mansoor et al., 2016).

It has been proposed that the P2X₇ receptor forms a dilated pore after prolonged exposure to ATP, permitting passage of dyes up to ~ 1 kDa in size (Rokic and Stojilkovic, 2013; Surprenant et al., 1996; Virginio et al., 1999). Two mechanisms have been postulated for the apparent time-dependent increase in membrane permeability, referred to as “pore dilation” (Costa-Junior et al., 2011; Ugur and Ugur, 2019): (1) progressive dilation of the pore via a slow conformational change (Surprenant et al., 1996), and (2) induction of conformational changes that recruit other proteins that are permeable to large molecules (Pelegrin and Surprenant, 2006). Studies have challenged the progressive pore dilation paradigm (Harkat et al., 2017; Li et al., 2015; Pippel et al., 2017), thus the topic remains controversial (Peverini et al., 2018). We see no evidence in our 2D/3D classes for the existence of a receptor population with a dilated pore, although progression to a dilated pore may require a specific lipid environment absent in our conditions (Bernier et al., 2008a; Bernier et al., 2008b; Karasawa et al., 2017; Murrell-Lagnado, 2017; Robinson et al., 2014).

The P2X₇ Apo State Has a Narrow Entrance to the Orthosteric ATP Binding Pocket.

To attempt to understand the relatively low affinity of P2X₇ receptors for ATP – a property that distinguishes P2X₇ from other P2X receptor subtypes - the orthosteric binding pockets of rP2X₇ and hP2X₃ were compared. Differences found in the binding pockets of the ATP-bound, open state structures between these two receptor subtypes cannot readily explain the >1,000-fold lower potency of ATP on P2X₇ receptors compared to P2X₃ receptors (Jarvis and Khakh, 2009) (Figure 3A–D). Both the U-shaped pose of ATP and the shape of the occupied orthosteric binding pocket are virtually identical in the open state structures of both subtypes. In addition, nearly every amino acid that coordinates ATP by hydrogen bonding and ionic interactions (K64, K66, T189, N292, R294, and K311 in rP2X₇; Figure 3A–B) is conserved across P2X subtypes (Chataigneau et al., 2013; Kawate, 2017).

Figure 3. Extracellular Orthosteric Ligand-Binding Site.

(A-B) View of the orthosteric binding pocket comparing key interactions to ATP in rP2X₇ (A) and hP2X₃ (B). Two conserved phenylalanine residues (F188 and F293; rP2X₇ numbering), proposed to stabilize the ATP-binding pocket, are also similar in both structures (residues not shown). (C-F) Surface representation of the binding pocket for rP2X₇ in the open state (C) and apo state (E) compared to hP2X₃ in the open state (D) and apo state (F). The apo binding pocket of rP2X₇ contains a narrow channel (< 11 Å orifice) which shields the pocket from solvent, limiting ligand access. The apo binding pocket of hP2X₃ is significantly more solvent exposed (17 Å orifice). The orthosteric-binding pocket is at a subunit interface, with protomer A in green for open state of rP2X₇ and forest green for open state of hP2X₃ or red for the apo state of rP2X₇ and red-purple for the apo state of hP2X₃. Protomer B is in gray and protomer C is in white.

However, minor differences between how rP2X₇ and hP2X₃ receptors coordinate ATP are present. One difference is that S275 in hP2X₃ provides an additional hydrogen bond to the α-phosphate of ATP, whereas F288 in rP2X₇ (Y288 in hP2X₇) cannot (Figure 3A–B). We note, also, that F288 does not occlude the binding pocket. A second potential difference lies in the nonpolar residues that interact with the adenosine base of ATP: F174, M200, and I215 in hP2X₃ in the published ATP-bound structure (Mansoor et al., 2016) compared to residues L191, N213, and I228 in rP2X₇, respectively. These residues would be predicted to interact with ATP based on sequence alignment of hP2X₃ and rP2X₇. While a polar Asn residue in rP2X₇ receptor could potentially reduce ATP affinity by ~10-fold (Chataigneau et al., 2013), compared to a hydrophobic Met residue in hP2X₃ receptor, our ATP-bound structure of rP2X₇ reveals that I214, and not N213, occupies the same spatial location as M200 in hP2X₃ (Figure 3A–B). Thus, the hydrophobic interaction to ATP is maintained. Although minor differences in interaction with ATP exist between the open states of rP2X₇ and hP2X₃, these changes alone are unlikely to account for the orders of magnitude difference in affinity for ATP between the two receptor subtypes.

In contrast, comparison of the apo state structures reveals that the unoccupied, orthosteric ATP binding pocket in the apo, closed state of hP2X₃ is significantly more solvent exposed than that of rP2X₇ (Figure 3E–F, Supplementary Movie M3–M4). The apo binding pocket of rP2X₇ contains a narrow channel formed by the head domain and the left flipper domain, which shields the binding pocket from solvent (Figure 3E, Supplementary Movie M3). These differences in apo state conformations may arise from ‘pocket breathing’ (Stank et al., 2016), due to inherent flexibility in the head domain of P2X₇. Because affinity is influenced by the time scale of ligand accessibility to the binding site as well as the time scale of ligand binding (Stank et al., 2016), both the narrow entry to the binding pocket and any protein flexibility that opens and closes this entrance would decrease the time that ATP has to sample the binding pocket, potentially decreasing its affinity. While such a mechanism to explain the low affinity of P2X₇ receptor for ATP is consistent with and supported by our structures, more experiments are needed to confirm its validity.

Palmitoylation of the Cytoplasmic C-cys Anchor Prevents Receptor Desensitization.

The rP2X₇ structures provide the first snapshot of an entire P2X receptor cytoplasmic domain. In particular, they afford insight into the form and function of the 18 amino acid, cytoplasmic cysteine-rich (C-cys) region (Allsopp and Evans, 2015) that is specific to P2X₇ receptors and that we have named the C-cys anchor (residues ~ 360–377 in rP2X₇). The C-cys anchor begins as TM2 emerges from the plasma membrane and enters the cytoplasm. It forms a loop that flares away from the transmembrane domain before redirecting the peptide backbone to the axis of symmetry of the receptor and joining the β₁₅ strand of the cytoplasmic cap (Figure 1E,G,H,I, Figure 4A–B, Figure 5A, Figure S2G,H,J). By connecting TM2 to the cytoplasmic cap, the C-cys anchor of P2X₇ physically links the two domains known to play a role in hP2X₃ receptor desensitization (Mansoor et al., 2016) while anchoring both to the membrane, thereby acting as a molecular hinge.

Figure 4. P2X₇ Cytoplasmic Domain – the C-cys Anchor is Palmitoylated to Modulate Receptor Function by Preventing Desensitization.

(A) Surface representation of the apo state of rP2X₇ highlighting locations of the C-cys anchor and putative phospholipid binding site. (B, left panel) The C-cys anchor is a loop containing six cysteine residues and one serine residue that each face toward the plasma membrane. (B, right panel) At least four cysteines (C362, C363, C374, and C377) and one serine (S360) on the C-cys anchor are palmitoylated. Palmitoylation of the C-cys anchor is also present in the open state reconstruction but more easily visualized in the higher resolution apo state map. The higher resolution apo state reconstruction also reveals that at least one cysteine (C4) in the N-terminus is palmitoylated. Because of domain swapping within the context of a trimeric receptor, the N-terminus of one protomer is adjacent to the C-cys anchor of an adjacent protomer, grouping together the palmitoylated moieties of different protomers and effectively interlocking them to the membrane. (C-D) Deletion (rP2X₇-ΔCcys) of the C-cys anchor (C) or mutation of residues in the C-cys anchor to alanine (rP2X₇-CcysMut) to prevent their palmitoylation (D), both result in channels that, when expressed in Xenopus oocytes, nearly completely desensitize during a 30 second application of 100 μM ATP (black bar). Wild type rP2X₇ receptor (rP2X₇-WT) does not desensitize. (E) The putative phospholipid binding site, modeled as a PS, is near the middle of the plasma membrane. Each protomer is a different color.

Figure 5. P2X₇ Cytoplasmic Domain – the Cytoplasmic Ballast Contains a Novel Fold.

(A) Ribbon representation of one subunit of the open state structure of rP2X₇ receptor shown in orthogonal views. Each of the major P2X₇ domains is color-coded. The cytoplasmic cap, C-cys anchor, and cytoplasmic ballast domains are highlighted. (B) Ribbon representation of the cytoplasmic ballast of rP2X₇ receptor displaying the secondary structure elements, the two zinc ions, and the guanosine nucleotide. (C) Topology diagram of the cytoplasmic ballast, shown in the same orientation as (B), reveals a novel fold without structural homology in the PDB. Dashed lines denote un-modeled regions. (D-E) Additional ribbon representations of the cytoplasmic ballast of the rP2X₇ receptor, shown in two different orthogonal views from (B), highlighting the relative locations of the zinc ions and the guanosine nucleotide within the ballast fold. The models are missing residues S443 to R469 in the apo state and residues S443 to R471 in the open state.

A clue to the effect that the C-cys anchor has on P2X₇ gating is apparent from inspection of its cysteine residues. Every cysteine in the loop of the C-cys anchor points toward the TM domain (Figure 4B, left panel). Moreover, a striking feature of the density maps is the presence of contiguous density extending beyond the ends of the side chains of numerous residues in the C-cys anchor, consistent with post-translational modification (Gonnord et al., 2009). Focused refinements of the apo state data, using a mask that includes the TM and cytoplasmic domains but excludes the extracellular domain, results in a map with density features that confirm the presence of palmitoyl groups. At least four cysteine residues (C362, C363, C374, and C377) and one serine residue (S360) in the C-cys anchor, as well as one cysteine residue in the N-terminus (C4), are palmitoylated such that the aliphatic chains extend into the membrane (Figure 4B, right panel). Analysis of the structures suggests that the spatial arrangement of these palmitoyl groups secures the C-cys anchor (and the N-terminus) to the membrane. We speculate that this could serve to lock the cytoplasmic cap in place and restrict movement of TM2, thus serving to inhibit the disassembly of the cytoplasmic cap - a prerequisite to the recoiling of TM2 and subsequent receptor desensitization in hP2X₃ (Mansoor et al., 2016).

To experimentally validate our proposal that palmitoylation of the C-cys anchor is responsible for preventing receptor desensitization, we deleted the C-cys anchor in rP2X₇. The resulting construct (rP2X₇-ΔCcys) gives rise to a receptor that desensitizes when expressed in Xenopus oocytes (Figure 4C), consistent with previous findings for the human P2X₇ receptor (Allsopp and Evans, 2015). Importantly, keeping the C-cys anchor, but mutating all of its palmitoylatable residues (S360, C362, C363, C371, C373, C374, and C377; rP2X₇ numbering) to alanine in order to prevent palmitoylation, also results in a receptor construct (rP2X₇-CcysMut) that robustly (τ = 4820 ± 137 ms) and nearly completely desensitizes during a 30 second exposure to ATP (Figure 4D). Thus, the desensitization observed in rP2X₇-ΔCcys and rP2X₇-CcysMut must be due to the loss of membrane anchoring in these mutated receptors, as the C-cys anchor is either absent (rP2X₇-ΔCcys) or cannot be palmitoylated (rP2X₇-CcysMut).

The focused apo state map also enhances a density located near the center of the TM domain, just above the termination point of the aliphatic chains of the palmitoylation groups. The density has the distinctive forked appearance of a phospholipid moiety (Figure 4E, Figure S4J), consistent with a phosphatidylcholine (PC), phosphatidylethanolamine (PE) or phosphatidylserine (PS) molecule. This putative lipid is bound with its head group inserted between the TMs, near the middle of the plasma membrane, and its aliphatic chains directed nearly perpendicular to the TM helices (Figure 4A,E). Interestingly, activation of P2X₇ has been shown to result in a redistribution of PS from the inner to the outer leaflet of the plasma membrane (Elliott et al., 2005; Mackenzie et al., 2005), as part of the signaling pathway for immunocyte activation and apoptosis (Mackenzie et al., 2005; Rysavy et al., 2014). The mechanism for P2X₇-induced PS translocation is unknown, but has been predicted to involve either a flippase-like enzyme or through the receptor itself (Mackenzie et al., 2005). Because of the location and orientation of the putative lipid relative to the TM domain, it is tempting to imagine that this finding might represent a transition state for the translocation of PS from the inner to the outer leaflet of the plasma membrane, although this is not established in our current work.

The Cytoplasmic Ballast Forms a Novel Fold.

The second major cytoplasmic element that the structures reveal is the cytoplasmic ballast, which comprises the final ~ 200 residues of each protomer (Figure 5A). Within the context of a trimeric receptor, every P2X₇ receptor contains three cytoplasmic ballasts, and each ballast hangs beneath the TM domain of an adjacent protomer as a result of domain swapping (Figure 1A,E,H, Figure 4A, Figure 6A, Figure S2C,G). By querying the Protein Data Bank (PDB) using distance matrix alignment (DALI) (Holm et al., 2006), we established that the fold of this element (Figure 5B–E) is without significant structural homology and refer to its structure as the ‘ballast fold’.

Figure 6. The P2X₇ Cytoplasmic Ballast Contains a Dinuclear Zinc Ion Complex and a High-Affinity Guanosine Nucleotide Binding Site.

(A) The cytoplasmic ballast from each protomer lies beneath the TM domain from an adjacent protomer. (B) The entire cytoplasmic domain viewed perpendicular to the membrane from the cytoplasm. Residues S443 - R469 are missing from the apo state structure, visualized here, because no density is present in the map. (C) A bottom up view focused on the center of the cytoplasmic domain, perpendicular to the membrane from the cytoplasm reveals how, in the context of a trimer, helices α₁₂ and α₁₃ of each cytoplasmic ballast form a hexagonal hole (diameter ~ 14 Å) through which the α₉ helix from each protomer (labeled A, B, and C) emerges. This is referred to as the “cytoplasmic plug”. The entire cytoplasmic domain assembles as a trimer through, in part, the interactions at the interface formed by helices α₉, α_12, and α₁₃. (D) The cytoplasmic dinuclear Zn²⁺ ion complex is coordinated by seven cysteine residues in a tetrahedral geometry. (E-F) The cytoplasmic guanosine nucleotide binding site, occupied by a GDP molecule, shown in two views highlighting the density for GDP (E) and key interactions made with conserved P2X₇ residues (F). (G) Measurement of ³H-GDP saturation binding to purified, detergent solubilized rP2X₇ receptor using SPA. The K_d for GDP binding was directly measured as 40 ± 10 nM. Error bars represent the standard deviation of three measurements. (H) Inhibition of ³H-GDP binding to rP2X₇ receptor by guanosine nucleotides yields K_i values of 70 ± 10 nM and 150 ± 15 nM for GDP and GTP, respectively. Non-guanosine diphosphate nucleotides do not displace ³H-GDP and thus do not bind to the cytoplasmic domain of P2X₇ receptor. Error bars represent the standard deviation of three measurements. (I) The guanosine nucleotide binding pocket is exposed at the interface between two cytoplasmic ballasts, accessible as a docking site for protein-protein interactions. Each protomer is a different color.

The overall fold of the cytoplasmic ballast is globular and shaped like a wedge. Three β-strands (β₁₆ – β₁₈) form an antiparallel β-sheet, which is followed by eight α-helices of varying lengths (α₉ – α₁₆), separated by loops that form a helical bundle (Figure 5). Immediately after the β-sheet, the polypeptide backbone is directed to the axis of symmetry of the receptor, where there are interactions between a short α-helix (α₉) from each protomer (Figure 5, Figure 6B–C). Unfortunately, just as the α₉ helices come together, density is not visible in either of the maps, resulting in models without residues S443 to R469 in the apo state and without residues S443 to R471 in the open state. We reconnected the sequence so that each ballast is formed from a single chain, as this is the simplest arrangement (reconnecting the chain otherwise results in each cytoplasmic ballast being formed by residues from two different protomer chains).

The entire cytoplasmic domain assembles as a trimer of three ballasts through interactions at an interface formed by helices α₉, α₁₂, and α₁₃. Helices α₁₂ and α₁₃ from each protomer interact in an arrangement that creates a prominent, hexagonal hole down the center of the cytoplasmic domain (Figure 6B–C). It is through this hole that the α₉ helices emerge, creating a tight constriction called the ‘cytoplasmic plug’ (Figure S5A–B) and making it unlikely that the ion permeation pathway follows the central axis of the cytoplasmic domain.

The Cytoplasmic Ballast Contains a Dinuclear Zinc Ion Complex and a High-Affinity Guanosine Nucleotide Binding Site.

In addition to cysteine residues in the C-cys anchor, P2X₇ has numerous other cysteines in the cytoplasmic ballast (C477, C479, C482, C498, C499, C506, C572 and C573; rP2X₇ numbering). These cysteines are important for proper P2X₇ receptor trafficking as their mutation results in receptor retention in the endoplasmic reticulum and subsequent proteolytic degradation (Gonnord et al., 2009). It was concluded that palmitoylation of these cytoplasmic cysteine residues plays a key role in cell surface localization (Gonnord et al., 2009). Our structures, however, do not support the idea that these cysteine residues are palmitoylated, as they are buried deep within the fold of the cytoplasmic ballast and are therefore inaccessible to post-translational modification. Instead, our structures suggest that these cysteine residues interact to coordinate an ion complex. Seven of these cysteine residues face each other, surrounding two strong non-protein densities in the maps (Figure S6A,D). This is suggestive of two ions being coordinated by seven cysteine residues in a tetrahedral geometry, characteristic of a dinuclear zinc ion complex (Kochanczyk et al., 2015) (Figure 6D).

We tested for and confirmed the presence of Zn²⁺ binding using inductively coupled plasma mass spectroscopy (ICP MS). Although we added no exogenous Zn²⁺ during purification, ICP MS confirmed ~ 130 μM Zn²⁺ in ~ 25 μM of purified rP2X₇ receptor. This > 5:1 mole-to-mole ratio of Zn²⁺ to receptor (Table S2) is consistent with two Zn²⁺ ions being bound per protomer (Figure 6D). Because previous work showed that mutation of these cysteine residues results in a P2X₇ receptor that does not localize to the plasma membrane (Gonnord et al., 2009), this Zn²⁺ binding motif must play a role in proper receptor processing and trafficking. However, as multinuclear Zn²⁺ binding sites often participate in processes of Zn²⁺ redistribution rather than trapping of Zn²⁺ at a structurally inert site (Kochanczyk et al., 2015), it remains to be seen whether this dinuclear Zn²⁺ binding site plays a role in P2X₇ signaling, for example, by modulating effects on apoptosis (Franklin and Costello, 2009).

A second unexpected non-protein density identified in the cytoplasmic ballast of both the apo and open state rP2X₇ structures has a size and shape that is consistent with a nucleotide (Figure S6B–C, E–F). We therefore performed liquid chromatography tandem mass spectrometry analysis on a purified receptor preparation using multiple reaction monitoring (MRM) and unambiguously confirmed the presence of guanosine diphosphate (GDP) nucleotide (Figure S7A–D), despite no exogenous GDP being added during purification. No other nucleotide was present at detectable levels. Each of the residues surrounding the density makes interactions consistent with those expected for GDP and is also conserved in the human P2X₇ receptor (Figure 6E–F).

We developed a series of radioactive binding assays to biochemically characterize the affinity and specificity of the binding site occupied by GDP. Using scintillation proximity assays (SPA) (Quick and Javitch, 2007), we found that GDP binds to the cytoplasmic domain of rP2X₇ with an affinity of ~ 40 nM, which we confirmed both by direct saturation binding (Figure 6G) and by competition inhibition assays (Figure 6H). Guanosine triphosphate (GTP) nucleotide (Figure 6H, Figure S6G) and the non-hydrolyzable GTP analog, GTP-γ-S (Figure S6G), bind with a similarly high affinity. In contrast, competition inhibition assays suggest that non-guanosine diphosphate (Figure 6H) or non-guanosine triphosphate (Figure S6G–I) nucleotides have > 150-fold lower affinity than guanosine nucleotides or do not bind at all. These measured binding affinities alongside cellular concentrations of diphosphate and triphosphate nucleotides (Traut, 1994) strongly suggest that the binding site occupied by GDP in the cytoplasmic ballast of rP2X₇ is a high affinity guanosine nucleotide binding site, capable of binding either GDP or GTP. Thus, a trimeric P2X₇ receptor can bind up to three guanosine nucleotide molecules in its cytoplasmic domain. The guanosine nucleotide binding pocket is at an exposed interface between two cytoplasmic ballast protomers, and in the absence of any P2X₇ receptor accessory or adaptor protein partners, would be easily accessible to cytosolic pools of guanosine nucleotides (Figure 6I).

To investigate what, if any, functional roles the dinuclear zinc ion complex, the guanosine nucleotide binding site, and the cytoplasmic ballast structure might play in P2X₇ receptor gating, we generated a receptor construct missing the entire cytoplasmic ballast (rP2X₇-ΔBallast). The activation kinetics of P2X₇ receptors are known to speed up during repeated applications of ATP (Allsopp and Evans, 2015; Roger et al., 2010; Roger et al., 2008), in a process known as current facilitation (Figure S7E). The rP2X₇-ΔBallast receptor shows the same facilitation of current kinetics when expressed in Xenopus oocytes (Figure S7F), indicating that the cytoplasmic ballast does not dramatically influence current facilitation nor desensitization. Moreover, the ATP dose-response curve, recorded in the absence of divalent cations, for the rP2X₇-ΔBallast receptor (EC₅₀ = 52 ± 11 μM) is nearly indistinguishable from rP2X₇-WT (42 ± 3 μM), suggesting that the cytoplasmic ballast does not influence the receptor’s apparent affinity for ATP (Figure S7G). Finally, the reversal potentials for rP2X₇-ΔBallast (−10 ± 3 mV) and rP2X₇-WT (−8 ± 1 mV) are very similar, revealing that the cytoplasmic ballast does not affect ion selectivity (Figure S7H). Taken together, these functional experiments suggest that neither the cytoplasmic ballast itself, nor the dinuclear zinc ion complex and guanosine nucleotide binding site within it, modulate the basic ion channel properties of P2X₇ receptor.

DISCUSSION

Our work has revealed the architecture of a full-length mammalian P2X₇ receptor for the first time. The apo and ATP-bound structures of rP2X₇ provide a number of major structural and functional insights: 1) Compared to the apo state conformation of the hP2X₃ receptor, the apo state conformation of the P2X₇ receptor has a more narrowed channel through which ATP must access the orthosteric binding pocket; thus, the P2X₇ receptor binding pocket is more shielded from solvent and ligand. 2) A putative lipid-binding site in the center of the TM domain potentially provides insight into the mechanism by which P2X₇ receptor activation has been reported to redistribute PS across the plasma membrane. 3) The structures reveal two P2X₇-specific cytoplasmic elements that we name the C-cys anchor and the cytoplasmic ballast. 4) The structures uncover the molecular basis of P2X receptor gating, demonstrating how palmitoylation of residues in the C-cys anchor uniquely prevent the P2X₇ receptor subtype from undergoing desensitization, a channel property inherent to all the other P2X subtypes. 5) Finally, the cytoplasmic ballast of P2X₇ receptor has a novel fold containing two previously unappreciated features, a dinuclear zinc ion complex and a high affinity guanosine nucleotide binding site, the functional roles of which warrant future studies.

In the fast-desensitizing hP2X₃ receptor, the cytoplasmic cap is evident in the ATP-bound open state but residues that form the cap are not resolved in the apo resting state or ATP-bound desensitized state structures (Mansoor et al., 2016). This suggests that the cap-forming elements are flexible or disordered in the closed states and that the formation and stability of this domain provides a structural scaffold for the open state until it is disassembled. When disassembly of the cytoplasmic cap occurs, the pore-lining TM2 helix recoils upward, inducing channel desensitization (Figure 7A). The structural transition of the TM2 helix from apo to open to desensitized state mimics the recoiling of a spring that is stretched from above and released from below. In this ‘helical recoil’ model of receptor desensitization (Mansoor et al., 2016), the structural stability of the cytoplasmic cap tunes the rate and extent of desensitization such that fast-desensitizing P2X subtypes have a less stable cap whereas slowly desensitizing subtypes have a more stable cap.

Figure 7. Molecular Mechanism of P2X Receptor Gating.

(A-B) Cartoon models summarizing the gating mechanism for the desensitizing P2X₁₋₆ receptor subtypes (A) compared to the non-desensitizing P2X₇ receptor subtype (B).

In stark contrast, the structures of the non-desensitizing rP2X₇ receptor reveal a preformed cytoplasmic cap that persists in both the apo and open states, permanently stabilized by a cytoplasmic, structural element that is unique to P2X₇ receptor - the C-cys anchor (Figure 7B). The C-cys anchor is a physical link between TM2 and the cytoplasmic cap and contains at least five palmitoylated residues that ‘anchor’ the cytoplasmic portion of TM2 to the membrane. We propose that this anchoring restricts the ‘helical recoil’ movement necessary for receptor desensitization, a hypothesis that explains why either deletion of the C-cys anchor (Figure 4C) or mutation of cysteine/serine residues in the C-cys anchor to prevent their palmitoylation (Figure 4D), both result in a P2X₇ receptor ion channel that nearly completely desensitizes during a 30 second application of ATP. These experiments prove that the presence of the C-cys anchor is necessary, but not sufficient, to prevent desensitization. It is palmitoylation of the C-cys anchor, and not simply the presence of the C-cys anchor, that modulates P2X₇ function to prevent receptor desensitization.

While ion channels have long been known to be regulated by protein palmitoylation either directly or through adaptor proteins, this form of regulation for ion channels most commonly occurs in processes of protein assembly, maturation, trafficking, insertion into the membrane, localization into membrane microdomains, and internalization/recycling (Naumenko and Ponimaskin, 2018; Shipston, 2011, 2014). Examples of palmitoylation modulating intrinsic ion channel activity or gating kinetics are less common but do exist (Shipston, 2011, 2014). For example, palmitoylation of the voltage-sensitive potassium channel Kv1.1 increases intrinsic voltage sensitivity (Gubitosi-Klug et al., 2005), and palmitoylation of N-type calcium channels modulates voltage-dependent inactivation (Qin et al., 1998; Stephens et al., 2000). Here, we have uncovered a direct example of how palmitoylation dramatically changes the gating properties of a ligand-gated ion channel. Moreover, because we now have structures of P2X₇ receptor that allow direct visualization of the palmitoyl groups and explain how palmitoylation prevents receptor desensitization, this can be used as an example to understand how palmitoylation might affect the functional activity of other ion channels.

Identification of both a dinuclear zinc ion complex and a high-affinity guanosine nucleotide binding site in the cytoplasmic ballast was completely unexpected (Figure 7B). Why these features are present in the cytoplasmic domain of P2X₇ receptor and whether they modulate P2X₇ function is still not clear, but we have determined that completely disconnecting the cytoplasmic ballast by making a receptor that terminates at residue K395 of rP2X₇ (Figure S2H) results in an ion channel (rP2X₇-ΔBallast) that is essentially indistinguishable from rP2X₇-WT in terms of current facilitation, lack of desensitization, apparent affinity for ATP, and ion selectivity (Figure S7E–H). Thus, the large cytoplasmic domain containing the dinuclear zinc ion complex and the guanosine nucleotide is not involved in P2X₇ receptor gating. A similar conclusion was reached about the P2X₇ C-terminus from studies on a naturally occurring, alternatively spliced P2X₇ receptor which terminates at residue 431, P2X₇-13B (Cheewatrakoolpong et al., 2005), and on a P2X₇ receptor truncated at residue 418 (Surprenant et al., 1996).

Truncating the C-terminus of the P2X₇ receptor, however, is not without a functional consequence. P2X₇ receptors that lack the C-terminus have severe defects in their ability to undergo “pore dilation” as well as their ability to induce apoptosis through membrane blebbing (Wilson et al., 2002) and activation of caspases (Cheewatrakoolpong et al., 2005). Indeed, the C-terminal domain of the P2X₇ receptor is required for the cytolytic actions of ATP (Adinolfi et al., 2010; Surprenant et al., 1996). Thus, although the C-terminus of P2X₇ receptor is not directly connected to ion channel activity, it appears to be important for initiating the downstream signal transduction pathways following P2X₇ receptor activation, suggesting that the cytoplasmic ballast may serve as a link between the P2X₇ receptor and intracellular signaling proteins (Adinolfi et al., 2010). Activation of P2X₇ receptors is known to recruit protein kinases and activate phospholipase D (el-Moatassim and Dubyak, 1992; Humphreys and Dubyak, 1996) and it has been speculated that these actions are mediated through its unique cytoplasmic domain (Costa-Junior et al., 2011). Together, these data have led to the idea that the P2X₇ receptor not only acts as an ion channel but, like metabotropic receptors, may also directly interact with and directly activate intracellular signaling proteins (Ugur and Ugur, 2019). Although we have not defined the link between the cytoplasmic ballast of the P2X₇ receptor and its unique intracellular signaling mechanisms, our structures now provide a new framework from which to begin these important studies.

Our two structures of a full-length P2X₇ receptor have provided the first insight into the architecture of the cytoplasmic domain of this unique receptor. As well as revealing how palmitoylation directly modulates P2X₇ function to prevent receptor desensitization, the structures unexpectedly reveal a dinuclear zinc ion complex and a high-affinity guanosine nucleotide binding site in the cytoplasmic domain. More studies are needed to investigate how these unexpected findings might act to modulate P2X₇ signal transduction.

STAR METHODS

LEAD CONTACT AND MATERIALS AVAILABILITY

Further information and requests for reagents should be directed to the Lead Contact Steven E. Mansoor (mansoors@ohsu.edu). All unique/stable reagents generated in this study are available from the Lead Contact with a completed Materials Transfer Agreement.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell Lines

SF9 cells (ThermoFisher Scientific, product 12659017, lot 421973) were cultured in Sf-900 III SFM (ThermoFisher Scientific) at 27 °C. These cells were used for expression of Baculovirus, and are female in origin. HEK 293 GNTI^- cells (Reeves et al., 2002) were cultured in FreeStyle 293 Expression Medium (GIBCO) at 37 °C supplemented with 2% (v/v) fetal bovine serum. These cells were used for expression of receptor and are female in origin. Unfertilized Xenopus laevis oocytes were purchased from Ecocyte Biosciences and kept at 18 °C until injection.

METHOD DETAILS

Receptor Constructs

The full-length rP2X₇ gene was sub-cloned into the pEG BacMaM vector as a C-terminal EGFP fusion with an octa-histidine affinity tag, and an HRV 3-C protease cleavage site to facilitate removal of the GFP and affinity tag prior to structural studies. No mutations or truncations were made to the receptor for structure determination. For electrophysiology experiments, the rP2X₇-WT construct is unmodified, full-length, wild-type rat P2X₇ receptor, the rP2X₇-ΔCcys construct has residues S360 - C377 removed, the rP2X₇-CcysMut construct has 7 residues in the C-cys anchor mutated to alanine residues (S360A, C362A, C363A, C371A, C373A, C374A, and C377A), and the rP2X₇-ΔBallast construct is truncated after residue K395. No GFP or affinity tag was present on any of the constructs used for the electrophysiology experiments.

Receptor Expression and Purification

The full-length rP2X₇ protein expression was facilitated by baculovirus mediated gene transduction (Goehring et al., 2014). Briefly, HEK 293 GNTI^- cells (Reeves et al., 2002) in suspension were infected by P2 BacMam virus and grown at 37 °C for 12 hours, followed by addition of sodium butyrate to a final concentration of 10 mM. Cells were then moved to 30 °C and incubated for an additional 36 hours before being harvested, washed with PBS buffer, and resuspended in TBS (50 mM Tris, pH 8.0 and 150 mM NaCl). Cells were then broken by sonication in the presence of protease inhibitors (1 mM PMSF, 0.05 mg/mL aprotinin, 2 μg/mL pepstatin A, and 2 μg/mL leupeptin) and ultracentrifugation was used to isolate the membrane fraction.

Membranes were homogenized in TBS buffer containing 15% glycerol and solubilized in 40 mM dodecyl-β-D-maltopyranoside (C12M) and 8 mM cholesterol hemisuccinate tris salt (CHS). The soluble fraction was isolated using ultracentrifugation and stirred with TALON resin at 4 °C for 1.5 hours before the resin was packed into an XK-16 column. The column was washed with ~ 4 column volumes of buffer (TBS buffer plus 1 mM C12M, 0.2 mM CHS, 30 mM imidazole and 5% glycerol) before elution with buffer containing 250 mM imidazole, pH 8.0. Peak fractions were pooled, concentrated, and digested by HRV 3-C protease (1:50, w/w) at room temperature for 1–3 hours. HRV 3-C protease was added once more (1:50, w/w) and the protein was additionally digested at 4 °C for ~14 hours. Following ultracentrifugation, the supernatant was injected onto a Superdex 200 10/300 column with a mobile phase containing 20 mM HEPES, pH 7.0, 100 mM NaCl, and 0.5 mM C12M. Trimeric rP2X₇ was isolated by size-exclusion chromatography (SEC) and monodisperse fractions were collectively concentrated for cryo-EM grid preparation.

Electron Microscopy Sample Preparation and Data Acquisition

For grid preparation, purified rP2X₇ was used with or without the addition of 1 mM ATP. Quantifoil holey carbon grids (Cu or Au 1.2/1.3 μm size/hole space, 300 mesh) were glow discharged for 60 seconds. Then, 2.5 – 3.0 μL of protein sample was added to the carbon face of the grids and blotted. The grid was plunge-frozen in liquid ethane cooled by liquid nitrogen using a Vitrobot Mark III held at 4 °C and 100% humidity.

Images were taken using an FEI Titan Krios electron microscope operating at 300 kV with a nominal magnification of 130,000x for the ATP data set and 75,000x for the Apo data set. For the ATP data set, images were recorded using a Gatan K2 Summit with a binned pixel size of 1.070 Å/pixel. Each image was dose-fractionated at 100 frames with a total exposure time of 10 seconds with 0.1 s per frame. The dose rate was 4.4 e⁻ Å⁻² s⁻¹. The images were recorded using the automated acquisition program SerialEM (Mastronarde, 2005). Nominal defocus values varied from −0.8 to – 2.2 μm. For the apo data set, images were recorded using an FEI Falcon III detector with a pixel size of 1.045 Å/pixel. Each image was dose-fractionated at 160 frames with a total exposure time of 60 seconds with 0.375 s per frame. The dose rate was 0.45 e⁻ Å⁻² s⁻¹. The images were recorded using the automated acquisition program EPU. Nominal defocus values varied from −1.2 to – 2.8 μm.

Electron Microscopy Data Processing

Images were motion-corrected and summed using MotionCor2 (Zheng et al., 2017). Defocus values were estimated using Gctf (Zhang, 2016). Particles were picked using DoG Picker (Voss et al., 2009) and subjected to an initial reference-free 2D classification using RELION (Scheres, 2012). Representative 2D class averages were selected and particles were further cleaned up with several rounds of 2D classification using RELION. The initial ab initio reconstruction was obtained using cryoSPARC (Punjani et al., 2017). The particles were then classified into five classes using 3D classification in RELION, with the initial reconstruction low-pass filtered to 30 Å as a reference model. Multiple rounds of 3D classification were performed. The particles from classes showing high-resolution features were combined and refined with C3 symmetry using RELION. Map sharpening was performed in RELION. The final resolutions reported in Table S1 are based on the gold standard Fourier shell correlation (FSC) 0.143 criteria. Local resolution was estimated in RELION.

Model Building and Structure Determination.

The extracellular domain of the apo state model of rP2X₇ receptor was built in Coot (Emsley and Cowtan, 2004) using the chicken P2X₇ receptor structure as a guide (PDB ID: 5XW6) (Kasuya et al., 2017). The cytoplasmic domain was built in Coot de novo from the apo state data. All stages of model building involved manual adjustments based on the quality of the maps in Coot, followed by real space refinement in PHENIX (Adams et al., 2010; DiMaio et al., 2013). For the apo and open state structures, residues S443 to R469 and residues S443 to R471 in the cytoplasmic domain, respectively, are missing from the models because no density is present in the maps. We chose to reconnect the broken chains such that each of the three globular cytoplasmic ballast domains is formed from a single protomer chain, as this is the simplest arrangement. However, reconnecting the chains such that each cytoplasmic ballast domain is formed from two different protomers (i.e., domain swapping within each cytoplasmic ballast) is possible.

For the apo state data, focused refinements were performed on both the extracellular domain (mask including the TM and extracellular domain but excluding the cytoplasmic domain) and the cytoplasmic domain (mask including the TM and cytoplasmic domain but excluding the extracellular domain). For the open state data, a focused refinement was performed on the cytoplasmic domain (mask including the TM and cytoplasmic domain but excluding the extracellular domain). These focused refinement maps were used to guide model building but were not used during refinement. For validation of the refined structures, FSC curves were applied to calculate the difference between the final model and the electron microscopy map. The geometries of the atomic models were evaluated using MolProbity (Chen et al., 2010). All figures were prepared using UCSF Chimera (Pettersen et al., 2004) and PyMOL (https://pymol.org).

Two-Electrode Voltage Clamping.

rP2X₇-encoding RNA was transcribed from pCDNA 3.1x using the mMessage mMachine T7 Ultra kit. Xenopus oocytes were injected with 0.5 ng of RNA and incubated at 18 °C for 6 – 12 hours in Barth’s solution plus 250 μg mL⁻¹ amikacin. Electrodes (0.5 – 3 MΩ) were filled with 3M KCl, oocytes were voltage-clamped at −60 mV, and currents were recorded in buffer containing 100 mM NaCl, 2.5 mM KCl, 0.1 mM EDTA, 0.1 mM flufenamic acid, and 5 mM HEPES, pH 7.4. Traces were recorded with application of either 100 μM ATP or 100 μM ATP co-applied with 10 μM A-438079 antagonist. Data acquisition was performed using the Axoclamp 2B amplifier and pClamp 10 software (Molecular Devices).

For dose-response curves, traces were recorded with applications of 8 separate ATP concentrations in a dilution series (3 mM, 1 mM, 333 μM, 111 μM, 37 μM, 12 μM, 4 μM, and 1 μM). The maximal amplitude of each current in the series was normalized as a percentage of the overall maximal current amplitude of the series. Current-voltage data were recorded with application of 100 μM ATP by measuring the current during a 1 second voltage ramp from −60 to 60 mV. Data acquisition was performed using the Axoclamp 2B amplifier and pClamp 10 software (Molecular Devices). Dose-response and current-voltage experiments were performed in triplicate. Data were averaged and presented with standard deviation values.

Structural Homology Search Using Distance Matrix Alignment (DALI).

The structure of the rP2X₇ cytoplasmic ballast domain (residues 395–595) was compared to all existing protein structures in the Protein Data Bank (PDB) using a heuristic PDB search on the DALI server. No significant structural homology was observed, with all reported Z-scores < 3.6. As a general rule, a Z-score > 20 means that the two structures are “definitely homologous”, a Z-score between 8 and 20 means that the two structures are “probably homologous”, a Z-score between 2 and 8 is a “grey area”, and a Z-score < 2 is not significant (Holm et al., 2006). There were 12 entries in the PDB with Z-scores between 2 and 3.5. Displaying the structural alignment of these 12 entries with the cytoplasmic ballast made it clear there was no structural homology between the cytoplasmic ballast and any of them.

Inductively Coupled Plasma Mass Spectrometry (ICP MS).

The purified rP2X₇ receptor sample was concentrated to ~ 5.5 mg/mL (or ~ 25 μM receptor) and 10 μL of this protein solution were added to 10 μL of concentrated HNO₃ (trace metal grade, Fisher) in an acid-rinsed Sarstedt tube (55.516 series). The sample was heated on a heating block to 90°C for 30 min. After the sample was cooled to room temperature, 1 mL of 1 % HNO3 (trace metal grade, Fisher) was added, the volume was measured to be 1020 μl. A total of 500 μL of this solution was then used for the analysis, accounting for the dilution factor.

Inductively coupled plasma mass spectroscopy (ICP MS) analysis was performed using an Agilent 7700x equipped with an ASX 500 auto sampler. The system was operated at a radio frequency power of 1550 W, an argon plasma gas flow rate of 15 L/min, Ar carrier gas flow rate of 0.9 L/min. Elements were measured in kinetic energy discrimination (KED) mode using He gas (4.3 mL/min). Data were quantified using a 10-point (0, 0.5, 1, 2, 5, 10, 20, 50, 100, 1000 ppb (ng/g)) curve for Mn, Fe, Cu, and Zn using external standards. For each sample, data were acquired in triplicate and averaged. A coefficient of variance (CoV) was determined from frequent measurements of a sample containing ~ 10 ppb of Mn, Fe, Cu, and Zn. An internal standard (Sc, Ge, Bi) continuously introduced with the sample was used to correct for detector fluctuations and to monitor plasma stability. Elemental recovery was evaluated by measuring NIST reference material (water, SRM 1643f) and found to within 90 – 110% for all determined elements.

Buffer was treated with concentrated HNO₃ by the same method as the protein sample to correct for background from the buffer and digestion tubes. Calibration standard (‘CEM2’) was added to DI water and digested with concentrated HNO₃ by the same method as the protein sample to check for recovery. NIST SRM 1683f was digested with concentrated HNO₃ by the same method as the protein sample to check for recovery. A NIST SRM 1683f was prepared at a 5x dilution (4 mL of 1% HNO3 (trace metal grade, Fisher) + 1 mL of NIST SRM 1683f) to ensure accuracy of the standard calibration curve. Finally, a sample containing the provided buffer diluted 2x (500 μL buffer + 500 μL 1% HNO₃ (trace metal grade, Fisher)) was prepared and measured to determine the ion concentration in the buffer.

GDP Detection Method Using Multiple Reaction Monitoring (MRM).

All standards were obtained from Sigma-Aldrich (St. Louis, MO). Solvents for LC-MS/MS were from VWR. Standard mono, di- and tri-nucleotide stocks (1 mg/mL) were prepared in 50 mM Tris-HCl, pH 7.4 with 5 mM MgCl₂ and stored at −80 °C. Stocks were diluted in water and final standards were prepared at different concentrations in 85% acetonitrile. Protein was concentrated to a final concentration of ~ 5 mg/mL in 20 mM Hepes buffer, pH 7.0 with 100 mM NaCl and 0.5 mM C12M (dodecyl maltoside). A 10 μL aliquot of the protein solution or a buffer blank solution were added to 90 μL of methanol containing 0.1% formic acid. The tube was vortex mixed for 1 min and then centrifuged at 10,000xg for 10 min to remove any precipitate. The clear supernatant was removed, placed in an autosampler vial and analyzed by liquid chromatography tandem mass spectrometry (LC-MS/MS). A buffer spike was prepared in an identical manner after spiking the buffer with 4.3 ng/μL of GDP. Standards were prepared by dilution into methanol with 0.1% formic acid. A 5 μL sample was injected for analysis.

LC-MS/MS Analysis for Nucleotides:

Nucleotides were analyzed using a 4000 Q-TRAP hybrid/triple quadrupole linear ion trap mass spectrometer (Applied BioSystems) with electrospray ionization (ESI) in positive mode. The mass spectrometer was interfaced to a Shimadzu (Columbia, MD) SIL-20AC XR auto-sampler followed by 2 LC-20AD XR LC pumps. The instrument was operated with the following settings: source voltage 4500 V, GS1 40, GS2 50, CUR 15, TEM 550 and CAD gas Medium. Compounds were monitored with multiple reaction monitoring (MRM). Optimal compound parameters were obtained by infusion of each compound. The gradient mobile phase was delivered at a flow rate of 0.4 mL/min and consisted of two solvents, A: water:acetonitrile 95:5 with 20 mM ammonium carbonate and 0.26% ammonium hydroxide, pH 9.8 and B: acetonitrile. The initial concentration of solvent B was 85%, held for 1 min then decreased to 38% over 21 min, held for 3 min, then increased back to start conditions of 85% B over 0.1 minute, and then held for 5 minutes, for a total run time of 30 minutes. Separation was achieved using a Phenomenex Luna 3μ Amino 100 Å 150 × 2.0 mm column with a 4 × 2.0 mm pre-column with the same support. The column was maintained at 35 °C using a Shimadzu CTO-20AC column oven. Data were acquired and analyzed using Analyst 1.6.3 (SCIEX). Retention times (min) for the compounds under these conditions were: CMP 12.7; UMP 11.5; AMP 12.1; IMP 13.0; GMP 13.8; ADP 14.8; GDP 16.8; ATP 15.7; GTP 17.0.

Radioligand-Binding Experiments:

Scintillation proximity assays (SPA) were performed on detergent-solubilized rP2X₇ receptors, purified without tag cleavage from dialyzed membranes (to remove bound, endogenous GDP), as previously described (Mansoor et al., 2016). In brief, the GDP affinity saturation binding experiments were carried out in 100 μL volumes using polyvinyltoluene copper (PVT-Cu) beads at 0.5 mg/mL, increasing concentrations of ³H-labeled GDP, and 10 nM GFP-tagged rP2X₇ protein in PBS buffer, pH 8.0, 0.3% BSA, and 0.2 mM C12M. Once components were mixed together, the samples were nutated at room temperature for 30 minutes before being counted. The background, nonspecific counts were determined by measuring the SPA signal in the presence of 0.2 M imidazole. All data points are obtained from triplicate experiments with the error bars representing the standard deviation. The reported K_d value is an average from two such saturation bindings curves with the error representing the standard deviation between the two independent experiments. The background nonspecific counts, also done in triplicate, were subtracted from the total counts to yield the specific counts using standard error propagation.

K_i inhibition studies were carried out in 100 μL volumes using polyvinyltoluene copper (PVT-Cu) beads at 0.5 mg/mL, 30 nM total ³H-labeled GDP, and 10 nM GFP-tagged rP2X₇ protein in PBS buffer, pH 8.0, 0.3% BSA, and 0.2 mM C12M. Counts were recorded with increasing concentration of cold competitive antagonist. The SPA signal was counted at various time points after 20 minutes nutation at room temperature. Assay plates were read using a MicroBeta TriLux 1450 LSC & Luminescence counter. Data were fitted using a standard single site competition equation, and K_i values were calculated from the IC₅₀ values using the Cheng-Prusoff equation. All data points are obtained from triplicate experiments with the error bars representing the standard deviation. The reported K_i values are an average from two such bindings curves with the error representing the standard deviation between the two independent experiments.

QUANTIFICATION AND STATISTICAL ANALYSIS

Electrophysiology and Radioligand-Binding Experiments

The TEVC recordings were performed on three biological replicates. Traces were initially analyzed with Clampfit 10.3 (Molecular Devices) and then exported to Prism 8 (GraphPad) for graphical purposes. For each data point in the dose-response graph, the error bars represent standard deviation from triplicate experiments. No values were omitted from analysis. The reported EC₅₀ is an average ± standard deviation from three independent measurements. The current-voltage experiments were performed in triplicate and reversal potentials are an average of those three measurements ± standard deviation.

For the saturation radioligand-binding experiment, the background nonspecific counts were subtracted from the total counts to yield the specific counts with standard error propagation. Each data point in the graph is the average ± standard deviation from three separate measurements. No values were omitted from analysis. Data was fit to a one-site nonlinear regression model to determine binding affinity. The reported K_d value is an average ± standard deviation from the fits of two such graphs.

For the competitive inhibition radioligand-binding experiment, data points in the graph are obtained from triplicate experiments with the error bars representing the standard deviation. Data were fitted using a standard single site competition equation, and K_i values were calculated from the IC₅₀ values using the Cheng-Prusoff equation. No values were omitted from analysis. The final reported K_i values are an average ± standard deviation of two such graphs.

DATA AND CODE AVAILABILITY

Data Resources

The three-dimensional cryo-EM density maps for the full-length rat P2X₇ receptor in the apo, closed state have been deposited in the EM Database under the accession code EMDB: EMD-20702, and the coordinates for the structure have been deposited in Protein Data Bank under accession code PDB: 6U9V. The overall map was used for refinement. Maps resulting from focused refinements performed on both the extracellular domain (mask including the TM and extracellular domain but excluding the cytoplasmic domain) and the cytoplasmic domain (mask including the TM and cytoplasmic domain but excluding the extracellular domain) are also deposited under this EMDB accession code. These maps were used to help with model building and to help enhance visualization of numerous features (such as palmitoylation in the C-cys anchor) present in the overall map but were not used for refinement in PHENIX.

The three-dimensional cryo-EM density maps for the full-length rat P2X₇ receptor in the ATP-bound, open state have been deposited in the EM Database under the accession code EMDB: EMD-20703, and the coordinates for the structure have been deposited in Protein Data Bank under accession code PDB: 6U9W. A map resulting from a focused refinement performed on the cytoplasmic domain (mask including the TM and cytoplasmic domain but excluding the extracellular domain) is also deposited under this EMDB accession code. This map was used to help with model building and to help enhance visualization of numerous features (such as palmitoylation in the C-cys anchor) present in the overall map but was not used for refinement in PHENIX.

Supplementary Material

3

Movie M1. Structural Conformational Changes that Occur in the P2X₇ Receptor Pore During the Gating Cycle. Related to Figure 2. This video demonstrates the structural conformational changes that occur in the receptor’s pore during the gating cycle of rP2X₇, shown perpendicular to the membrane from the extracellular surface. It highlights the transition of the pore from the apo state to the open state before resetting back to the apo state. Each protomer subunit is shown in a different color.

4

Movie M2. Structural Conformational Changes that Occur in the P2X₇ Receptor Pore During the Gating Cycle, Highlighting the Residues that Define the Narrowest Region Across the Pore in Each Conformational State. Related to Figure 2. This video shows the structural conformational changes that occur in the receptor’s pore during the gating cycle of rP2X₇, shown perpendicular to the membrane from the extracellular surface, highlighting the residues that define the narrowest region across the pore in each conformational state: S339 in the apo state (red) and S342 in the open state (green). The movie steps through the transition of the pore from the apo state to the open state before resetting back to the apo state.

5

Movie M3. Unoccupied, Orthosteric ATP Binding Pocket in the Apo, Closed State of rP2X₇ Receptor. Related to Figure 3. This video demonstrates that the unoccupied, orthosteric ATP binding pocket in the apo, closed state of rP2X₇ contains a narrow channel formed by the head domain and the left flipper domain, which potentially shields the binding pocket from ligand. Each protomer subunit is shown in a different color.

6

Movie M4. Unoccupied, Orthosteric ATP Binding Pocket in the Apo, Closed State of hP2X₃ Receptor. Related to Figure 3. This video shows the unoccupied, orthosteric ATP binding pocket in the apo, closed state of hP2X₃. Compared to the binding pocket in the apo, closed state of rP2X₇ receptor, the hP2X₃ apo binding pocket is significantly more solvent exposed, as evidenced by the greater distance between the head domain and the left flipper domain. Each protomer subunit is shown in a different color.

7

Figure S1. The Cryo Electron Microscopy Data Processing Workflow, Using the Apo State Data of Full-Length rP2X₇ as an Example. Related to Figure 1. Particles were picked using DoG Picker and visually checked using RELION. Particles were subjected to multiple rounds of 2D classification in RELION to remove false positive picks. For 3D classification in RELION, an ab initio reconstruction was generated in CryoSPARC and used as the reference model. Multiple rounds of 3D refinements and the final map sharpening were carried out using RELION.

Figure S2. Functional Studies and the Overall Architecture of Full-Length rP2X₇ in the Open State. Related to Figure 1. (A) ATP-induced rP2X₇ currents (100 μM ATP, black bar) do not undergo desensitization. (B) Co-application of 10 μM antagonist A-438079 (red bar) inhibits the rP2X₇ receptor current induced by 100 μM ATP (black bar). (C-F) The three-dimensional reconstruction of full-length rP2X₇ in the ATP-bound, open state viewed parallel to the membrane as a side view (C), perpendicular to the membrane from the extracellular surface (D), from the center of the transmembrane domain, toward the cytoplasm (E), and from the intracellular side of the membrane, toward the cytoplasm (F). (G) Ribbon representation of a trimeric receptor structure corresponding to the ATP-bound, open state reconstruction shown in (C). Each protomer is represented by a different color. (H) Ribbon representation of a single subunit of the open state structure, highlighting various purinergic receptor domains, including two new domains, the C-cys anchor and the cytoplasmic ballast. The major domains within a protomer are organized by color: extracellular domain in blue, transmembrane domain in green, cytoplasmic cap in cyan, C-cys anchor in purple, and the cytoplasmic ballast in red. The palmitate groups are colored in sand. (I) A close-up side view of the cytoplasmic portion of the transmembrane domain highlighting the interactions that form the cytoplasmic cap. The cytoplasmic cap is composed of domain-swapped β-strands (β₋₁, β₀, and β₁₅) from each protomer. To improve visual clarity of the cytoplasmic cap, the C-cys anchor and cytoplasmic ballast have been removed. (J) Top-down cross-sectional view of the cytoplasmic cap looking toward the cytoplasm. Residues 29–345 and 395–595 of the receptor have been removed to improve visual clarity. Each protomer is a different color.

Figure S3. Cryo-Electron Microscopy Analysis of Full-Length rP2X₇ in the Apo and Open States. Related to Figure 1 and Figure S2. (A) Representative electron micrograph of the apo state rP2X₇ data. (B) Selected two-dimensional class averages of the electron micrographs of the apo state rP2X₇ data. (C, E) The gold-standard Fourier shell correlation curves for the electron microscopy maps of the apo state data (C) and the open state data (E) of rP2X₇ are shown in black for the masked map and blue for the unmasked map, and the Fourier shell correlation curves between the atomic model and the final electron microscopy maps are shown in red. (D, F) Angular distribution of particles used for the refinement of the apo state data (D) and the open state data (F).

Figure S4. Local Resolution Estimates and Representative Densities of the Reconstructions for the Apo State and Open State Data of rP2X₇. Related to Figure 1, Figure 4 and Figure S2. (A-F) Local resolution estimation of the reconstructions shown as a side view, a top down view from the extracellular surface, and a bottom up view from the cytoplasmic surface, respectively, for the apo state data (A-C) and open state data (D-F) of rP2X₇ receptor. (G, H) Representative protein densities from different regions of the apo state (G) and open state (H) maps of rP2X₇ receptor. (I, J) Densities for ATP (I) and putative PS phospholipid (J) in the open state map.

Figure S5. The Pore-Lining Surface of rP2X₇ for the Open and Apo States. Related to Figure 2. (A) Coronal section of a surface representation of the open state of rP2X₇ reveals five vestibules (upper, central, extracellular, intracellular vestibule I, and intracellular vestibule II) are located on the molecular three-fold axis. Beyond the transmembrane gate, two additional sites of constriction exist along the three-fold axis (cytoplasmic cap and cytoplasmic plug). (B) Pore-lining surface along the entire axis of rP2X₇ for open and apo states. The color of each sphere represents a different radius from the receptor center, as calculated by the program HOLE: reddish pink < 1.15 Å, green between 1.15 – 2.30 Å, and purple > 2.30 Å. (C) Plot of pore radius as a function of distance along the pore axis (focused around the TM gate) for the apo state vs. the open state of rP2X₇. The positions of the residues making up the narrowest radius for the TM gate in each conformational state are labeled. The C_α position of residue L353 is set as zero.

Figure S6. Dinuclear Zinc Ion Complex, GDP Nucleotide Densities, and Nucleotide Triphosphate Binding in the Cytoplasmic Ballast of rP2X₇ Structures. Related to Figure 1, Figure 5, Figure 6 and Figure S2. (A-F) The cytoplasmic ballast of rP2X₇ receptor contains density features in the reconstructions that represent the unexpected findings of a zinc ion complex with two Zn²⁺ ions (Zn1 and Zn2) and a GDP molecule, respectively, for both the apo state (A-C) and open state (D-F) data sets. (A, D) The zinc ion densities are still present when the maps are contoured to a threshold (σ = 25) where the protein density is no longer clearly visible. (B-C, E-F) The residues surrounding the density make interactions consistent with those expected for GDP. Specifically, R546, K583, and R578 make ionic interactions with the phosphates, R574 and the carbonyl of A567 make hydrogen-bonding interactions with the guanine base, and residues L569, Y550, and F591 contribute van der Waal’s interactions to the guanine base. Each of these residues is conserved in human P2X₇. The apo state structure is shown in red (B-C) and the open state structure is shown in green (E-F). The densities shown are from the apo and open state maps obtained using a mask that included the TM and cytoplasmic domain, but excluded the extracellular domain (focused refinement on the cytoplasmic domain). (G) Inhibition of ³H-GDP binding to rP2X₇ receptor by guanosine nucleotide triphosphates yields K_i values of 150 ± 15 nM and 100 ± 10 nM for GTP and GTP-γ-S, respectively. Non-guanosine nucleotide triphosphates show significantly lower binding affinity – nucleotides UTP and CTP inhibit ³H-GDP binding to rP2X₇ receptor with K_i values of 60 ± 10 μM and 130 ± 3 μM, respectively. Nucleotide TTP does not displace ³H-GDP and thus does not bind to the cytoplasmic domain of P2X₇ receptor. (H-I) Nucleotide ATP shows a time-dependent displacement of ³H-GDP from the cytoplasmic domain of P2X₇ receptor. While no displacement of ³H-GDP occurs initially (H), an ATP-induced displacement of ³H-GDP (K_i value of 22 ± 10 μM) can be measured after several hours (I). A time-dependent change in the displacement of ³H-GDP did not occur with any other nucleotide. Because ATP also binds to the extracellular domain of P2X₇ receptor in the orthosteric ATP binding pocket, further studies are needed to determine if the time-dependent displacement of ³H-GDP is due to direct displacement from ATP binding to the cytoplasmic domain or from allosteric modulation, whereby the affinity of ³H-GDP binding to the cytoplasmic domain is decreased after ATP binds to the extracellular, orthosteric ligand binding pocket. Error bars represent the standard deviation of three measurements.

Figure S7. GDP Detection Using Multiple Reaction Monitoring (MRM) and Functional Assessment of the Cytoplasmic Ballast. Related to Figure 5 and Figure 6. (A-D) Liquid chromatography tandem mass spectrometry analysis on a purified rP2X₇ receptor preparation using multiple reaction monitoring (MRM) of fragment ions (m/z 135 and m/z 152) produced from the parent ion (m/z 444) in positive ESI mode confirm the identification of GDP nucleotide. MRM monitoring ion fragment m/z 152 and ion fragment m/z 135, respectively, for buffer and GDP control (A, C) compared to the receptor sample (B, D). The retention time and ratio of the two ions are consistent with GDP nucleotide. No other nucleotides were detected. (E-H) The functional aspects of rP2X₇-ΔBallast are nearly identical to rP2X₇-WT with respect to kinetics of activation, lack of desensitization, affinity for ATP, and ion selectivity. (E-F) Both rP2X₇-WT (E) and rP2X₇-ΔBallast (F) show similar current facilitation, and no desensitization, with repeated exposure to 100 μM ATP (1 minute, black bar) at one minute intervals, for three exposures. (G) ATP-activated dose-response curves for rP2X₇-WT (black trace) and rP2X₇-ΔBallast (red trace), performed in the absence of divalent cations, show similar EC₅₀ values of 42 ± 3 μM and 52 ± 11 μM, respectively. Error bars represent the standard deviation of three measurements. (H) The ATP-induced (100 μM) two electrode voltage clamp currents for rP2X₇-WT (black trace) and rP2X₇-ΔBallast (red trace) reversed sign at −8 ± 1 mV and −10 ± 3 mV, respectively. Error bars represent the standard deviation of three measurements.

HIGHLIGHTS:

• Cryo-EM structures of P2X₇ receptor in apo- (closed) and ATP-bound (open) states

• P2X₇ structures reveal two cytoplasmic elements: C-cys anchor and cytoplasmic ballast

• C-cys anchor prevents desensitization by anchoring to membrane through palmitoylation

• Cytoplasmic ballast has a zinc ion complex and a guanosine nucleotide binding site