Structural and functional conservation of key domains in InsP3 and ryanodine receptors.

Inositol-1,4,5-trisphosphate receptors (InsP(3)Rs) and ryanodine receptors (RyRs) are tetrameric intracellular Ca(2+) channels. In each of these receptor families, the pore, which is formed by carboxy-terminal transmembrane domains, is regulated by signals that are detected by large cytosolic structures. InsP(3)R gating is initiated by InsP(3) binding to the InsP(3)-binding core (IBC, residues 224-604 of InsP(3)R1) and it requires the suppressor domain (SD, residues 1-223 of InsP(3)R1). Here we present structures of the amino-terminal region (NT, residues 1-604) of rat InsP(3)R1 with (3.6 Å) and without (3.0 Å) InsP(3) bound. The arrangement of the three NT domains, SD, IBC-β and IBC-α, identifies two discrete interfaces (α and β) between the IBC and SD. Similar interfaces occur between equivalent domains (A, B and C) in RyR1 (ref. 9). The orientations of the three domains when docked into a tetrameric structure of InsP(3)R and of the ABC domains docked into RyR are remarkably similar. The importance of the α-interface for activation of InsP(3)R and RyR is confirmed by mutagenesis and, for RyR, by disease-causing mutations. Binding of InsP(3) causes partial closure of the clam-like IBC, disrupting the β-interface and pulling the SD towards the IBC. This reorients an exposed SD loop ('hotspot' (HS) loop) that is essential for InsP(3)R activation. The loop is conserved in RyR and includes mutations that are associated with malignant hyperthermia and central core disease. The HS loop interacts with an adjacent NT, suggesting that activation re-arranges inter-subunit interactions. The A domain of RyR functionally replaced the SD in full-length InsP(3)R, and an InsP(3)R in which its C-terminal transmembrane region was replaced by that from RyR1 was gated by InsP(3) and blocked by ryanodine. Activation mechanisms are conserved between InsP(3)R and RyR. Allosteric modulation of two similar domain interfaces within an N-terminal subunit reorients the first domain (SD or A domain), allowing it, through interactions of the second domain of an adjacent subunit (IBC-β or B domain), to gate the pore.

The essential role of the SD in linking InsP3 binding to InsP3R gating highlights the need to define the structural consequences of InsP3 binding to the NT (residues 1-604 of InsP3R1) (Supplementary Fig. 1). Because our attempts to crystallize the NT yielded poorly diffracting crystals, we expressed a Cys-less form of the NT (NTCysless). Native and Cys-less forms of the NT and IBC behaved indistinguishably (Supplementary Fig. 2 and Supplementary Tables 1-2), but NTCysless provided crystals with much improved diffraction (Supplementary Table 3). We determined crystal structures of NTCysless with (3.6 Å) and without (3.0 Å) InsP3 bound, showing three subdomains: the SD, IBC-β (residues 224-436) and IBC-α (residues 437-604) (Fig. 1a). The structures of these subdomains were nearly identical to those of isolated native SD and IBC[2,3] (Supplementary Fig. 3).

Figure 1

Structure of the N-terminal region of InsP3R1 without InsP3 bound

a, Structure of NTCysless at 3 Å resolution showing SD (blue), IBC-β (green) and IBC-α (yellow). Dashed lines show invisible regions in electron density. Positions of the three domains within a single InsP3R subunit are shown. b, c, Interfaces between SD/IBC-β (β-interface) (b) and SD/IBC-α domains (α-interface, with the hydrophobic core boxed and the 2Fo-Fc electron density map of key residues (contoured at 1.0 σ) shown as mesh) (c). d, Superposition of apo-NTCysless and RyR1-ABC (grey)[9] structures by overlaying IBC-β and RyR1 B-domain. ‘Hot spot’ (HS) loop in RyR1 and corresponding region in InsP3R1 are highlighted (red). e, Close-up views of HS regions of InsP3R1 (blue) and RyR1 (grey, black lettering) with conserved residues depicted as sticks. Structure-based DALILite alignment of rat InsP3R and rabbit RyR show conserved residues in yellow, RyR1 disease-associated mutations in red, and hydrophobic residues implicated in activation of InsP3R in blue[8].

The SD, IBC-β and IBC-α form a triangular structure, with the SD behind the InsP3-binding site (Fig. 1a). The SD interacts via two interfaces with the IBC, one with IBC-β (β-interface) and another with IBC-α (α-interface). A 3[10]-like turn between the last strand of the SD and the first strand of IBC-β positions the IBC relative to the SD (Supplementary Fig. 4e). Within this connecting turn, a salt bridge (K225/D228) stabilizes the backbone conformation and so positions residues that form the β-interface. These interactions in the connecting turn and β-interface are augmented by a network of hydrophobic interactions within IBC-β (Fig. 1b). The α-interface forms a long ‘Velcro’-like structure that also involves a network of hydrophobic and electrostatic interactions (Fig. 1c). Intimate hydrophobic interactions between V33, and to a lesser extent L32, from the SD; and V452, F445, A449 and L476 from IBC-α are supported by bidentate salt bridges between R54/K127 in the SD and D444 in IBC-α (Fig. 1c). The V33K mutation at the α-interface almost abolished inhibition of InsP3 binding by the SD[3,4] and reduced channel open probability[4], confirming its importance. Mutation of neighbouring residues that contribute less to the α-interface (L32K, D34K, R36E, K127E) had lesser effects on InsP3 binding, while mutation of residues that do not contribute to the interface (D35K, K52E) had no effect (Supplementary Table 4)[3,4]. Hydrophobic and electrostatic interaction networks at the α- and β-interfaces contribute to a buried surface between the SD and IBC (~2040 Å2) that forms a hub connecting InsP3 binding to channel activation.

The structure of the NT is remarkably similar to that of the N-terminal of RyR1[9]. The three NT domains of InsP3R1 (SD, IBC-β and IBC-α) can be individually superposed to corresponding domains of RyR1 (A, B and C) (Supplementary Fig. 3), and the relative orientation of the domains is nearly identical (Fig. 1d). Mutation of Y167A, located on an exposed loop of the SD opposite the IBC interfaces (‘HS-loop’[11], residues 165-180, boxed in Fig. 1d), attenuates InsP3-evoked Ca2+ release[8]; and Ca2+, a co-regulator of InsP3R[14], causes the loop to become accessible.[15] The disease-associated ‘hot spot loop’ of RyR1[11] sits at the same location within the ABC structure[9] (Fig. 1d) and a mimetic peptide causes RyR2 to become leaky[16]. Furthermore, the backbone and side chain conformation of this loop region superposes well in the two receptors (Fig. 1e). The HS-loop provides a critical link between InsP3 binding and gating.

The domain interfaces of InsP3R1-NT and RyR1-ABC are also similar. The bidentate salt bridges between R54/K127 and D444 at the InsP3R1 α-interface are preserved in RyR1 ABC, albeit in a reversed-charge manner between D40/D61 and R402 (Supplementary Fig. 4a). In RyR1, mutation of these residues (R402C, D61N) is associated with malignant hyperthermia and central core disease[9], suggesting that disruption of the interaction perturbs RyR gating, as it does for InsP3R. The structural similarities extend also to the β-interface of InsP3R1 and corresponding A/B interface in RyR1 (Supplementary Fig. 4b-d).

Our structures of NTCysless with and without InsP3 bound, together with that of the InsP3-bound IBC[2] (Supplementary Fig. 5) reveal the structural changes evoked by InsP3 (Fig. 2). Side chains of nine residues become organized around InsP3 (Supplementary Fig. 5a), and the domain orientation angle between IBC-β and IBC-α is reduced (by ~8°) after InsP3 binding (Fig. 2 and Supplementary Fig. 5a). This InsP3-evoked ‘clam closure’, which is consistent with earlier predictions[17] and small-angle X-ray scattering[18], causes the distance across the entrance to the InsP3-binding pocket to decrease (Supplementary Fig. 5b, c). A similar agonist-evoked domain closure occurs in some glutamate receptor channels[19]. The SD and IBC remain associated after closure of the IBC (Fig. 2). InsP3 binding hardly changes the interactions across the extensive α-interface, but at the β-interface the SD residues move away from IBC-β (Supplementary Fig. 5d-f). With the SD glued to IBC-α by the α-interface, and the β-interface serving as a lubricant, InsP3 binding causes the SD to twist (by ~9°) and move closer to the top of the IBC (Fig. 2). This causes an amplified translational movement of the conserved HS-loop in the SD (Supplementary Fig. 5g). While our work was under review, 3.8 Å structures of apo- and InsP3-bound NT derived from a single crystal grown in excess InsP3 were published, showing similar InsP3-induced allostery in the interfaces between domains[13]. This confirms our observations, but our higher resolution structures reveal more detail of the α- and β-interfaces associated with this conformational change (Supplementary Discussion).

Figure 2

InsP3-evoked conformational changes

Superposition of apo-NTCysless (SD, blue; IBC-α, yellow; IBC-β, green) and InsP3-bound NTCysless (3.6 Å resolution, magenta) by overlaying IBC-β domain. InsP3 binding causes the SD to rotate towards the IBC accompanied by a swing approximately perpendicular to the IBC ‘clam closure’. This twist (curved arrow) is measured as the angular difference between the SD arm helices in the apo- and bound states (~9°). Movement of the HS-loop (boxed) shows the distance between α-carbons of Y167 (~3.7 Å). A view rotated 90° about the x-axis is shown at right with only IBCs represented. The interdomain (IBC-β and IBC-α) angular difference between the free and bound states is ~8° (black arrow). Further details of InsP3 binding and its effects on the IBC and α- and β-interfaces in Supplementary Figure 5.

Docking the ABC structure into cryo-EM maps of RyR1 showed that the N-terminal domains form a central ring at the top of the mushroom-like RyR1[9]. Rigid-body docking of our apo-NTCysless structure into a cryo-EM 10 Å structure of a closed InsP3R1[10] reveals an arrangement remarkably similar to that of RyR1 with a high docking contrast (Fig. 3 and Supplementary Fig. 6). The three domains of the four NTs, which form the upper cytoplasmic surface of the mushroom-like InsP3R, are arranged as four hillocks around a central bowl. This arrangement allows InsP3 unrestricted access to the IBC from the side of the cap (Fig. 3), and it is consistent with accessibility studies and binding sites for regulatory proteins (Supplementary Fig. 6c and Supplementary Table 5). Within the tetrameric InsP3R, the only contacts between NT subunits are via the critical HS-loop of the SD and a flexible loop (β20-β21, Supplementary Fig. 7) in IBC-β (Fig. 3c, d). The latter is longer in RyR and it lies ~10 Å further from the neighbouring hot spot loop[9] (Fig. 3c, d). In InsP3R, the arm domains (residues 67-109) of each SD are the only NT structures that extend beyond the cap towards the pore (Fig. 3a), but these domains are neither essential for InsP3R activation[8] nor conserved in RyR[9,11].

Figure 3

Docking of the apo-NTCysless structure into the cryo-EM map of InsP3R1

a, b, Side (a) and top (b) views of apo-NTCysless structure docked into the cryo-EM map (grey mesh) of InsP3R1 in a closed state[10]. Contour level corresponds to mass of InsP3R1 tetramer of 1.3 MDa (protein density 0.8 Da/Å3). Four molecules of the NT (SD, blue; IBC-β, green; IBC-α, yellow) are located at top of the cytoplasmic portion of the InsP3R1 tetramer. c, d, Dockings of the apo-NTCysless (coloured as in a) and ABC[9] (grey) structures into cryo-EM structures of InsP3R1[10] and RyR1[9], respectively, are overlaid and presented to show only N-terminal structures. HS-loops of InsP3R (magenta) and RyR (orange) are highlighted. Enlargement of boxed area (d). Locations of other binding sites within NT of InsP3R1 are shown in Supplementary Figure 6.

The structural similarities between the N-termini of RyR and InsP3R prompted us to examine whether the domains are functionally interchangeable. In a chimeric N-terminal fragment comprising the A-domain of RyR2 and IBC from InsP3R1 (RyR2A-IBC), the A-domain mimicked the SD by inhibiting InsP3 binding (Fig. 4a, b). Mutations within the A-domain loop that forms the A-B interface in RyR[9] or the equivalent InsP3R loop in the SD attenuated this inhibition of binding (Fig. 4c, Supplementary Table 6 and Supplementary Fig. 8). InsP3 stimulated Ca2+ release via InsP3R1 or a chimeric InsP3R1 in which the SD was replaced by the A-domain of RyR1 (RyR1A-InsP3R1 (Fig. 4a, d and Supplementary Fig. 9). Both InsP3Rs were similarly expressed and they released similar fractions of the Ca2+ stores and with similar sensitivity to InsP3 (Supplementary Table 7). Opening of native InsP3R or RyR is restrained by interactions between cytosolic domains[20,21]. It is therefore significant that expression of InsP3R1 or RyR1A-InsP3R1 affected neither the Ca2+ content of the ER nor the Ca2+ leak from it (Supplementary Fig. 10), confirming that InsP3R and RyR1A-InsP3R1 have no detectable spontaneous activity. This demonstrates that the SD of InsP3R can be functionally substituted by the A-domain of RyR.

Figure 4

Functional chimeras of InsP3R and RyR

a, Proteins used. b, Specific binding of 3H-InsP3 in presence of adenophostin A. c, Inhibition of 3H-InsP3 binding to IBC by SD or A-domain, and effects of mutations within equivalent loops. Affinities shown relative to IBC (ΔpKD). Structures show key residues within SD or A-domain at α-interface. d, Ca2+ release from DT40 cells expressing InsP3R1, RyR1A-InsP3R1 or lacking InsP3R (KO). e, f, Effect of ryanodine (10 μM) on Ca2+ release from DT40 cells expressing InsP3R1-RyR1 (e) or InsP3R1 (f). Ryanodine (≤10 μM) did not stimulate Ca2+ release via InsP3R1-RyR1 suggesting that TMDs may not alone mediate stimulation of RyR[25]. Results (d-f) are percentages of ATP-dependent Ca2+ uptake. g, Specific 3H-ryanodine binding (dpm, disintegrations/min) to membranes of DT40 cells expressing InsP3R1 or InsP3R1-RyR1 with caffeine (10 mM) or InsP3 (1 μM). Non-specific binding was 2245 ± 211 dpm. Results (b-g) are means ± s.e.m., n ≥ 3.

An InsP3R1 in which residues downstream of TMD1 were replaced by the equivalent region of RyR1 (InsP3R1-RyR1) also responded to InsP3 (Fig. 4a,e). Expression of InsP3R1-RyR1 increased Ca2+ leak from the ER, and this was reversed by ryanodine, which blocks the RyR pore[22]. However, the increased leak was insufficient to affect the steady-state Ca2+ content (Supplementary Fig. 10), suggesting that InsP3R1-RyR1 has minimal spontaneous activity. Expression of InsP3R1-RyR1 matched that of other InsP3Rs, but cells expressing InsP3R1-RyR1 were ~20-fold less sensitive to InsP3 (Supplementary Table 7). Because the TMDs minimally affect InsP3 binding[23], this diminished response probably reflects a decrease in InsP3 efficacy. The increased Ca2+ leak and reduced efficacy of InsP3 suggest that within InsP3R1-RyR, communication between the SD and channel are slightly less effective than in native InsP3R. Nevertheless, it is remarkable that cytosolic domains of an InsP3R should so effectively regulate the pore of a RyR when the two receptors share only modest sequence identity and differ in the number of residues separating the NT from TMDs (Fig. 4a), and in the lengths and sequences of their C-terminal tails and the loops linking TMDs (Supplementary Fig. 11).

Ryanodine (10 μM) had no effect on InsP3R1 or RyR1A-InsP3R, but it abolished InsP3-evoked Ca2+ release via InsP3R1-RyR1 (Fig. 4e, f). Because ryanodine binds selectively to active RyR[24], 3H-ryanodine binding is stimulated by agonists of RyR, like caffeine. Whereas caffeine had no effect on specific 3H-ryanodine binding to InsP3R1-RyR1, InsP3 stimulated it (Fig. 4g). InsP3 therefore causes conformational changes to the channel of InsP3R1-RyR1 that mimic those of native RyR in allowing binding of 3H-ryanodine.

Conservation of structure-function relationships between InsP3R and RyR (Fig. 1-4) allows comparisons between them to suggest possible mechanisms of InsP3R activation. For both receptors, gating requires that conformational changes in the large cytoplasmic structures pass to the TMDs[10,22], but the N-terminal domains of InsP3R and RyR are at least 60 Å from these TMDs[1,9,22] (Fig. 3a and Supplementary Fig. 6). Despite some evidence implicating direct interactions between the SD and TMD4-5 loop in gating InsP3R (Supplementary Discussion), we suggest, and in keeping with results from RyR[20,26,27], that additional cytosolic domains couple the NT to opening of the pore. The exposed HS-loop in the SD (Fig. 1d and 3c,d) (hot spot-loop of RyR)[9,11] is arranged similarly within the isolated N-terminal structures of InsP3R and RyR (Fig. 1d) and it reorients after InsP3 binding (Fig. 2). When the NT is docked into the InsP3R structure[10], the HS-loop forms (with an exposed loop of IBC-β) the only interface between adjacent NT, however, the equivalent loop is displaced in the RyR[9] (Fig. 3c, d). InsP3 binding closes the clam-like IBC, disrupting the β-interface and re-orienting the HS-loop (Fig. 2 and Supplementary Fig. 5). This, we suggest, disrupts interaction of the HS-loop with a neighbouring NT to cause a coordinated re-arrangement of the apical InsP3R structure (Fig. 3). The open state of RyR1 is associated with outward movement of protein density in regions that match the locations of docked ABC structures[9,22] and with larger movements of peripheral ‘clamp domains’[9,22] that are absent from InsP3R[10]. Movement of these apical domains in RyR is accompanied by re-arrangements within regions that taper towards the pore[22] and which, in InsP3R, include the most flexible parts of its structure[10]. We suggest that similar rearrangements of the apical surface of InsP3R and RyR couple by shared mechanisms to additional cytosolic domains to gate the pore of each channel.

METHODS SUMMARY

The N-terminal (NT, residues 1-604) of rat InsP3R1 in which all Cys were replaced by Ala (NTCysless) was expressed in E. coli and purified. Crystals of NTCysless were grown by the hanging-drop vapour diffusion method in 0.1 M Hepes pH 7.0, 0.8-1.0 M (NH4)2SO4, and 3% (v/v) trimethylamine N-oxide for apo-state crystals, or 0.1 M Na citrate (pH 6.0), 8% (w/v) PEG-6000, 70 mM Li2SO4, 3% dimethyl sulfoxide for InsP3-bound crystals. Diffraction data were collected at 100 K on the 19-ID (apo-crystals) or 19-BM (InsP3-bound crystals) beam lines at the Advanced Photon Source Synchrotron facility (Argonne, IL) and processed with HKL2000[28]. Structures of apo-NTCysless at 3.0 Å resolution and InsP3-bound NTCysless at 3.6 Å resolution were determined by molecular replacement using structures of the SD (PDB code: 1XZZ)[3] and the IBC (1N4K)[2] as search models with the program Phaser[29]. Iterative refinement and model building were performed with Refmac5 and Coot, respectively. Numbering of secondary structure motifs is in accord with Supplementary Figure 7. Binding of 3H-InsP3 or 3H-ryanodine to full-length InsP3R1, chimeras of InsP3R1 and RyR, and to related N-terminal fragments was defined using equilibrium-competition binding assays[4]. Functional properties of InsP3R1 and chimeras were characterized after stable expression in DT40 cells lacking endogenous InsP3R[4]. A luminal Ca2+ indicator was used to record InsP3-evoked Ca2+ release from the intracellular stores of permeabilized DT40 cells[4].