IP₃ receptors – lessons from analyses ex cellula

Inositol 1,4,5-trisphosphate receptors (IP₃Rs) and ryanodine receptors (RyR) are the two major families of intracellular Ca²⁺-release channels in animal cells (Fig. 1A). IP₃Rs are expressed in most cells, whereas RyRs have a more restricted distribution. RyRs are most abundant in excitable cells, notably in striated muscle, where they contribute to excitation–contraction coupling (Fig. 1A) (Van Petegem, 2014). In this Review, we focus on IP₃Rs, and how methods applied to IP₃Rs removed from intact cells have contributed to our understanding of IP₃R behaviour. Progress in our understanding of IP₃Rs and RyRs has advanced in parallel, and with this progress it became clear that the two families share structural and functional features (Baker et al., 2017; Seo et al., 2012). Hence, despite our focus on IP₃Rs, we draw also on evidence from analyses of RyRs.

Classic work by Sydney Ringer demonstrated that cardiac muscle contraction requires extracellular Ca²⁺ (Ringer, 1883). This was, with the benefit of hindsight, the first of many studies to show that the contributions to physiological responses of extracellular Ca²⁺ and Ca²⁺ held within intracellular stores are entangled. For cardiac muscle, depolarization of the plasma membrane (PM) causes voltage-gated Ca²⁺ channels (Ca_v1.2, also known as CACNA1C) to open, and the local increase in cytosolic free Ca²⁺ concentration ([Ca²⁺]_c) is then amplified by Ca²⁺-induced Ca²⁺ release (CICR) through type 2 ryanodine receptors (RyR2) in the sarcoplasmic reticulum (Bers, 2002) (Fig. 1A). CICR and the local Ca²⁺ signalling that is required to avoid CICR from becoming explosive have become recurrent themes in the field of Ca²⁺ signalling (Rios, 2018). Fluorescent Ca²⁺ indicators and optical microscopy now allow Ca²⁺ sparks, local Ca²⁺ signals evoked by a small cluster of RyRs, to be measured with exquisite subcellular resolution in cardiac muscles (Cheng and Lederer, 2008). However, it was studies of permeabilized cells (‘skinned’ fibres) that provided the first evidence for CICR in muscle (Endo et al., 1970; Fabiato and Fabiato, 1979). Analyses of RyRs that were reconstituted into planar lipid bilayers first showed that RyRs form large-conductance cation channels that are biphasically regulated by cytosolic Ca²⁺ (Lai et al., 1988; Meissner, 2017). Finally, analyses of RyR fragments by X-ray crystallography (Van Petegem, 2014) and of complete RyRs by cryo-electron microscopy (des Georges et al., 2016; Efremov et al., 2015; Peng et al., 2016; Yan et al., 2015; Zalk et al., 2015) are revealing the structural basis of RyR behaviour.

Progress towards understanding the second major family of intracellular Ca²⁺-release channels, the IP₃Rs, began with an influential review in which a causal link between receptor-stimulated turnover of phosphatidylinositol and Ca²⁺ signalling was proposed (Michell, 1975). Subsequent work established that many receptors stimulate phospholipases C, which cleave phosphatidylinositol 4,5-bisphosphate to produce IP₃ and diacylglycerol (Berridge, 1993) (Fig. 1A). IP₃ provides the link to Ca²⁺ signalling; not, as first envisaged, by directly stimulating Ca²⁺ entry across the PM (Michell, 1975), but by stimulating Ca²⁺ release from the endoplasmic reticulum (ER) through IP₃Rs (Berridge and Irvine, 1984; Streb et al., 1983). Another influential review suggested the link between IP₃-evoked Ca²⁺ release and Ca²⁺ entry across the PM, and proposed that loss of Ca²⁺ from the ER stimulated Ca²⁺ entry (Putney, 1986). The workings of this store-operated Ca²⁺ entry (SOCE) pathway are now clear: dissociation of Ca²⁺ from the luminal EF-hand motif of a protein embedded in the ER membrane, stromal interaction molecule 1 (STIM1), causes STIM1 to oligomerize and expose a cytosolic domain, through which it stimulates opening of a Ca²⁺-selective channel in the PM (Feske et al., 2006; Prakriya and Lewis, 2015). The Ca²⁺ channel that mediates SOCE is a hexameric assembly of Orai subunits (Hou et al., 2012; Yen and Lewis, 2018), grandiloquently named from Greek mythology after the keepers of heaven (Feske et al., 2006).

IP₃Rs and RyRs are biphasically regulated by cytosolic Ca²⁺ (Bezprozvanny et al., 1991). For IP₃Rs exposed to IP₃, a modest increase in [Ca²⁺]_c stimulates opening, whereas a higher [Ca²⁺]_c is inhibitory (Foskett et al., 2007; Iino, 1990). Hence IP₃Rs, at least once they have bound IP₃ (Alzayady et al., 2016), can, like RyRs, mediate CICR (Fig. 1B,C). As with RyRs, IP₃Rs assemble into clusters, within which opening of one IP₃R ignites the activity of its neighbours to generate local ‘Ca²⁺ puffs’ (Fig. 1C) (Smith and Parker, 2009; Thillaiappan et al., 2017), analogous to Ca²⁺ sparks in muscle. These behaviours illustrate some of the many similarities between IP₃Rs and RyRs, which include their close structural relationship (Baker et al., 2017; Seo et al., 2012; Van Petegem, 2014). Although IP₃Rs and RyRs are the major intracellular Ca²⁺-release channels and the focus of this Review, they are not the only intracellular Ca²⁺ channels. Brief descriptions of additional intracellular Ca²⁺ channels are provided in Box 1.

The productive interplay between studies of minimally perturbed tissue, facilitated by a plethora of Ca²⁺ indicators (Lock et al., 2015), fluorescent proteins (Rodriguez et al., 2017) and fluorescence microscopy techniques (Thorn, 2016), alongside analyses of cellular components, has shaped our understanding of Ca²⁺ signalling. Here, we consider how analyses of IP₃Rs conducted outside their normal intracellular environment (ex cellula) have advanced our understanding of IP₃-evoked Ca²⁺ signals. We begin by considering how analyses of permeabilized cells established that the ER is the major intracellular Ca²⁺ store and that IP₃ releases Ca²⁺ from it. Radioligand binding analyses then both identified the sites to which IP₃ binds to activate IP₃Rs, and paved the way to structural studies, which we show are now coming close to revealing how IP₃ binding causes the pore of the IP₃R to open. We conclude by considering the contributions of electrophysiological recordings to our understanding of IP₃R gating.

Permeabilized cells allow the Ca²⁺ content of intracellular organelles to be measured under conditions where the intracellular environment can be precisely controlled. To achieve this control, the PM must be disrupted without unduly perturbing organelles (Schulz, 1990). The permeabilized cells are then bathed in medium that mimics cytosol, notably in its low [Ca²⁺]_c (∼100 nM). Electroporation (Knight, 1981; Xie et al., 2013) and a variety of chemical means have been used to selectively permeabilize the PM. The chemicals achieve their PM-selectivity by interacting with cholesterol (e.g. saponin, digitonin, β-escin), which is enriched in the PM (Wassler et al., 1987), or as pore-forming toxins (e.g. α-toxin, streptolysin-O) that are too large to pass through the pores that they induce (Schulz, 1990).

After a protracted controversy (Babcock et al., 1979; Dehaye et al., 1980), analyses of permeabilized cells established that the ER, rather than mitochondria, is the major intracellular Ca²⁺ store in animal cells (Burgess et al., 1983). In an elegant study, saponin-permeabilized hepatocytes were bathed in cytosol-like medium with Ca²⁺ buffered to mimic the [Ca²⁺]_c of an unstimulated cell. Each permeabilized cell was then shown to have the same Ca²⁺ content as an intact cell, and critically all of that Ca²⁺ was in the ER (Burgess et al., 1983). Hence, it is the ER from which most extracellular stimuli evoke Ca²⁺ release.

Analyses of insect salivary glands demonstrated that phosphoinositide turnover was required for extracellular stimuli to evoke Ca²⁺ signals (Berridge and Fain, 1979), and showed that IP₃ was the first cytosolic product of receptor-stimulated phosphoinositide hydrolysis (Berridge, 1983). Hence, IP₃ emerged as the likely messenger that links receptors in the PM to Ca²⁺ release from the ER (Fig. 1A). Permeabilized cells again were the subject of the decisive experiment: addition of IP₃ to permeabilized pancreatic acinar cells stimulated release of Ca²⁺ from a non-mitochondrial Ca²⁺ store (Streb et al., 1983). It is now universally accepted that most IP₃Rs reside in ER membranes, but IP₃Rs can also mediate Ca²⁺ release from the Golgi apparatus (Aulestia et al., 2015; Pinton et al., 1998), the nuclear envelope (Foskett et al., 2007; Rahman et al., 2009; Stehno-Bittel et al., 1995) and perhaps from a nucleoplasmic reticulum (Echevarria et al., 2003). In some cells, a few IP₃Rs (typically only 2–3 IP₃Rs per cell) are also expressed in the PM, where they mediate Ca²⁺ entry (Dellis et al., 2006, 2008). In many studies, though not in all (Watras et al., 1991), the ER Ca²⁺ release evoked by IP₃ was shown to be positively cooperative (e.g. Champeil et al., 1989; Marchant and Taylor, 1997; Meyer et al., 1988), suggesting a need for IP₃ to bind to several IP₃R subunits before the channel can open. A recent study using concatenated IP₃R subunits showed that a defective IP₃-binding site in only one of the four subunits prevents IP₃R activation (Alzayady et al., 2016), leading to the conclusion that all four subunits of an IP₃R must bind IP₃ before the channel can open.

But IP₃ binding is not alone sufficient to stimulate Ca²⁺ release through IP₃Rs. Instead, IP₃ binding primes IP₃Rs to bind Ca²⁺, and Ca²⁺ binding then causes the channel to open (Adkins and Taylor, 1999; Marchant and Taylor, 1997) (Fig. 1B). Hence, IP₃Rs require binding of two ligands, IP₃ and Ca²⁺, to open. This dual regulation endows IP₃Rs with their capacity to mediate regenerative Ca²⁺ signals through CICR. Again, it was analyses of permeabilized cells that provided the first evidence that Ca²⁺ release through IP₃Rs is regulated by [Ca²⁺]_c (Iino, 1987). High-resolution optical analyses of Ca²⁺ signals later revealed that within intact cells, IP₃-evoked Ca²⁺ signals originate from elementary units that comprise a small cluster of IP₃Rs (Smith and Parker, 2009; Thillaiappan et al., 2017). Opening of the first IP₃R within a cluster is proposed to rapidly ignite the activity of some of its neighbours by CICR to generate a Ca²⁺ puff (Fig. 1C). As the stimulus intensity increases, Ca²⁺ spreading from one Ca²⁺ puff to another IP₃R cluster can initiate further Ca²⁺ puffs, allowing the signal to spread across the cell as a regenerative Ca²⁺ wave (Marchant et al., 1999). The frequency of these global signals then increases with stimulus intensity (Thurley et al., 2014).

Structure-activity relationships (SAR), established by comparing the activities of a range of structurally-related chemical stimuli, are often used to probe the recognition properties of receptors. SAR analyses of the effects of IP₃ analogues on Ca²⁺ release from permeabilized cells provided the first evidence that dephosphorylation of IP₃ to (1,4)IP₂ terminates IP₃ activity (Burgess et al., 1984). The (1,3,4,5)IP₄ that is produced when IP₃ is phosphorylated by IP₃ 3-kinases was proposed to regulate IP₃Rs (Loomis-Husselbee et al., 1996), but it is now clear that this phosphorylation also inactivates IP₃ signalling through IP₃Rs (Bird and Putney, 1996; Saleem et al., 2012). Hence, both endogenous pathways for IP₃ metabolism effectively inactivate IP₃ signalling to IP₃Rs (Fig. 1A). SAR analyses of many analogues of IP₃ and adenophostin A, a fungal metabolite that binds with high affinity to IP₃Rs (Takahashi et al., 1994), established that a key feature of IP₃R agonists is the presence of a vicinal 4,5-bisphosphate moiety (Fig. 1D) (Rossi et al., 2010, 2009; Saleem et al., 2012). All active inositol phosphate analogues have this 4,5-vicinal bisphosphate moiety (Fig. 1D).

There are no wholly selective antagonists of IP₃Rs. Some ligands [heparin, 2-aminoethoxydiphenylborane (2-APB), xestospongin C and caffeine] have utility, but they all lack selectivity. Furthermore, heparin is not membrane-permeant, and results with xestospongin C are inconsistent (see Saleem et al., 2014). Addition of large substituents to the 2-O-position of IP₃ produces partial agonists. Partial agonists are ligands that, once they have bound to IP₃R, are less effective in causing the channel to open than full agonists like IP₃ (Rossi et al., 2009). These SAR analyses of IP₃ analogues modified at the 2-position, which again relied heavily on analyses of permeabilized cells, confirmed the importance of the extreme N-terminal region of the IP₃R (the suppressor domain, SD; Fig. 1E) in IP₃R activation (Rossi et al., 2009). They further suggest systematic strategies towards developing high-affinity antagonists of IP₃Rs.

There is, therefore, a long history of experiments using permeabilized cells illuminating our understanding of IP₃-evoked Ca²⁺ release. These studies first identified ER as the major intracellular Ca²⁺ store, they showed that IP₃ evokes Ca²⁺ release from the ER, and that IP₃Rs are regulated by Ca²⁺. Furthermore, they defined the biochemical steps that inactivate IP₃ and, through SAR analyses, they have revealed ligands that have contributed to understanding the mechanisms of IP₃R activation.

Binding of IP₃ to the four binding sites of the IP₃R initiates the conformational changes that culminate in opening of the Ca²⁺-permeable pore (Alzayady et al., 2016; Chandrasekhar et al., 2016). These IP₃ binding events are usually analysed by means of radioligand binding, which allows determination of binding affinities (as equilibrium dissociation constants, K_D) for ³H-IP₃ or any competing ligand, but a variety of other methods have also been successfully applied (Fig. 2, Box 2). K_D values are important for comparison with functional analyses in revealing how ligands activate IP₃Rs. Such analyses were, for example, critical in showing that the vicinal 4,5-bisphosphate of IP₃ is essential for activity, whereas the 1-phosphate improves binding affinity (Fig. 1D) (Nahorski and Potter, 1989). Comparisons of SAR with binding analyses can also establish which bound ligands most effectively open the channel. Our comparisons of functional and ³H-IP₃ equilibrium-competition binding analyses, for example, established that whereas IP₃ is a full agonist that effectively gates the IP₃R, other modified analogues of IP₃ bind with high affinity to IP₃R, but they are much less effective in causing the channel to open (Rossi et al., 2009). These partial agonists provide insight into the mechanisms of IP₃R activation by demonstrating how large moieties at the 2-position of IP₃ attenuate IP₃R activation, (Rossi et al., 2009). They further suggest strategies for development of analogues that bind without activating IP₃Rs (i.e. antagonists).

Binding analyses also allow IP₃R properties to be addressed under conditions where IP₃-evoked Ca²⁺ release is not retained. This opportunity is particularly important during purification of IP₃Rs for structural studies using either IP₃R fragments for X-ray crystallography (Bosanac et al., 2002, 2005; Hamada et al., 2017; Lin et al., 2011; Seo et al., 2012) or, after detergent-solubilization of complete IP₃Rs, for single-particle analysis by cryo-EM (Fan et al., 2015; Paknejad and Hite, 2018). In subsequent sections, we review progress towards understanding how IP₃ binding leads to opening of the IP₃R pore.

The route to determining IP₃R structures began with the identification of specific, high-affinity, intracellular ³²P-IP₃-binding sites with recognition properties that matched those expected of the receptor through which IP₃ evoked Ca²⁺ release (Baukal et al., 1985; Spät et al., 1986). Subsequent studies established that heparin competed with ³H-IP₃ for these binding sites (heparin is a competitive antagonist of IP₃), and that the sites were abundant in Purkinje cells of the cerebellum (Worley et al., 1987). Together, these observations allowed IP₃Rs to be purified from cerebellum using heparin chromatography (Maeda et al., 1988; Supattapone et al., 1988). Functional reconstitution of the purified protein then established that it was alone sufficient to form an IP₃-gated Ca²⁺ channel (Ferris et al., 1989; Maeda et al., 1991). Many additional proteins were later shown to associate with IP₃Rs and modulate their responses to IP₃ (Prole and Taylor, 2016). Screening of cDNA libraries from cerebellum then provided the complete primary sequence of IP₃R1 (also known as ITPR1) (Furuichi et al., 1989; Mignery et al., 1989), and soon afterwards the other two IP₃R subtypes, IP₃R2 (ITPR2) (Südhof et al., 1991) and IP₃R3 (ITPR3) (Blondel et al., 1993) were identified. Subsequent studies established that the three IP₃R subunits (IP₃R1–IP₃R3) assemble to form homo-tetrameric and hetero-tetrameric channels (Monkawa et al., 1995), and confirmed that the core properties of all IP₃R subtypes are similar: each forms a large-conductance Ca²⁺-permeable channel that is gated by binding of IP₃ and Ca²⁺ (Foskett, 2010), and each generates Ca²⁺ puffs (Mataragka and Taylor, 2018). The subtypes are, however, differentially expressed, and they differ in their affinities for IP₃ (Iwai et al., 2007), sensitivity to Ca²⁺ regulation (Foskett, 2010) and in whether they are modulated by additional regulators (Prole and Taylor, 2016). Furthermore, the functional consequences of mutant or defective IP₃Rs differ among subtypes (see Terry et al., 2018). IP₃R1 has so far been the major focus of the structural studies.

Deletion analyses (Mignery and Südhof, 1990) and expression of IP₃R fragments in bacteria (Yoshikawa et al., 1996) established that each IP₃R subunit has a single IP₃-binding site, the IBC, formed by residues 224–604 towards the N-terminal of the primary sequence (with a total of ∼2750 residues) (Fig. 1E). The identification of four IP₃-binding sites in each IP₃R tetramer, and the demonstration that all four are required for IP₃ to evoke Ca²⁺ release (Alzayady et al., 2016), provided an explanation for the widely observed cooperative responses to IP₃ (Champeil et al., 1989; Meyer et al., 1988; Parker and Miledi, 1989). Subsequent studies identified residues within the IBC that are required for IP₃ binding, notably the residues that bind to the critical 4- and 5-phosphate groups of IP₃ (Furutama et al., 1996). These residues are conserved in IP₃Rs, but not in RyRs (Bosanac et al., 2002; Seo et al., 2012). It was also shown that the SD inhibits IP₃ binding (Uchida et al., 2003), which aligns with the importance of the SD in coupling IP₃ binding to channel gating (Rossi et al., 2009): IP₃Rs without an SD bind IP₃ with high affinity, but they do not release Ca²⁺ (Uchida et al., 2003).

Examination of high-resolution crystal structures of N-terminal fragments of the IP₃R directly revealed both the determinants of IP₃ binding and the initial steps in IP₃R activation. The two domains (α and β) of the IBC form a clam-shaped structure, within which conserved residues bind to IP₃ (Bosanac et al., 2002). The 1- and 5-phosphates of IP₃ interact predominantly with residues in IBC-α, whereas the 4-phosphate interacts with IBC-β (Fig. 1D,E). Interaction of the critical 4- and 5-phosphates with opposing sides of the ‘clam’ allows IP₃ to partially close the clam and initiate IP₃R activation (Hamada et al., 2017; Lin et al., 2011; Paknejad and Hite, 2018; Seo et al., 2012). This interpretation, which elegantly reveals the structural basis of the SAR, is supported by results with an adenophostin A analogue in which an alternative contact with the α-domain substitutes for loss of the usual phosphate (Sureshan et al., 2012).

In the isolated N-terminal domain, the SD is firmly anchored to IBC-α by an extensive interface and more loosely associated with IBC-β (Fig. 1E). Hence, when IP₃ causes the IBC clam to close, the SD moves with IBC-α, which was predicted to disrupt interaction of an exposed SD loop, the ‘hot spot’ loop (Yamazaki et al., 2010) with IBC-β of a neighbouring subunit (Seo et al., 2012). In RyR, too, these inter-subunit interactions between N-terminal domains are weakened during receptor activation (des Georges et al., 2016). The resulting weakening of interactions between subunits may contribute to channel gating. This is supported by evidence that Ca²⁺-binding protein 1 (CABP1), which inhibits IP₃R gating, rigidifies these interactions between IP₃R subunits (Li et al., 2013). However, within the constraints of a full-length IP₃R, strong inter-subunit interactions between the SD and IBC-β might constrain the SD, such that IBC-α moves when IP₃ closes the clam (Paknejad and Hite, 2018). Identification of the sites to which IP₃ binds, which relied heavily on radioligand binding analyses, set the scene for the structural analyses that allow researchers to seek to understand how IP₃ binding opens the pore of the IP₃R. We consider recent progress with such structural analyses in the next section.

Single-particle analysis of cryo-EM images has allowed determination of the structures of the complete IP₃R1 in a closed state (Fan et al., 2015), of IP₃R3 with and without IP₃ and Ca²⁺ bound (Paknejad and Hite, 2018), and of RyRs in different states (des Georges et al., 2016; Efremov et al., 2015; Peng et al., 2016; Yan et al., 2015; Zalk et al., 2015), and has begun to reveal the workings of the pore regions of these related channels. The results also tentatively suggest how IP₃ binding might lead to opening of the IP₃R pore.

The IP₃R has a structure reminiscent of a square mushroom. Much of the stalk is embedded in the ER membrane and the cap, with a diameter of ∼25 nm, extends at least 13 nm into the cytosol (Fan et al., 2015). The large size is significant because it might exclude IP₃Rs from the narrow junctions between ER and the PM (Thillaiappan et al., 2017), whereas at other junctions, between ER and mitochondria for example (Csordás et al., 2018), it places the head of the IP₃R, from which Ca²⁺ exits, very close to the neighbouring organelle.

The cytosolic entrance to the central cavity of the IP₃R is surrounded by the N-terminal domains (SD and IBC-β, Fig. 3). IBC-α forms part of a larger domain (ARM1) that curves to the edge of the cap and interacts with two large curved domains (ARM2 and ARM3) that comprise most of the remaining cytosolic structure and form the underside of the mushroom cap (Fig. 3). Within the ER membrane, there are 24 IP₃R transmembrane domains (TMDs), six from each subunit (Fan et al., 2015). However, recent structural analyses of both IP₃R (Paknejad and Hite, 2018) and RyR1 (des Georges et al., 2016) identified a pair of additional helices (between TMD1 and TMD2 of IP₃R3) that challenge the accepted view that there are six TMDs per subunit. The TMD region, similar in structure to voltage-gated ion channels, is very similar (though not identical) (Baker et al., 2017) in RyRs and IP₃Rs. The ion-conducting path is lined by the four tilted TMD6 helices and a short (∼1 nm) ‘selectivity filter’ at the luminal end through which hydrated cations must pass in single file. The selectivity filter, its supporting pore-loop helix and a flexible luminal loop are all formed by residues linking TMD5 to TMD6. Near the cytosolic end of TMD6, a narrow hydrophobic constriction blocks the movement of ions in the closed channel (Fan et al., 2015) (Fig. 3). The hydrophobic side chains of these residues must move for the pore to open. Opening of the RyR pore is associated with splaying and bowing of TMD6, such that the hydrophobic side-chain of a residue that occludes the cytosolic end of the closed pore is displaced, opening a hydrophilic path that allows passage of a hydrated Ca²⁺ ion. Similar mechanisms may be associated with opening of the IP₃R pore.

TMD6 is supported by TMD5, which in turn is buttressed by the TMD1–TMD4 bundle of the adjacent subunit. The short cytosolic TMD4–TMD5 helical linker aligns along the ER membrane behind the TMD6 helices, holding them in place. In the closed RyR1 channel, this linker tightly encircles the cytosolic end of the TMD6 bundle, restricting its movement, but this grip is relaxed as the channel opens, freeing TMD6 to move and allowing the pore to dilate (des Georges et al., 2016). In both IP₃R and RyR, TMD6 extends well beyond the ER membrane (∼1.5 nm in IP₃R) and then terminates in a pair of short α-helices (the linker domain, LNK, in IP₃R) that includes a Zn²⁺-finger motif that aligns parallel with the ER membrane (des Georges et al., 2016; Fan et al., 2015; Paknejad and Hite, 2018). In IP₃R, but notably not in RyR, the entwined TMD6 helices then continue beyond the LNK domain to the cap of the mushroom, where each contacts the IBC-β domain of a neighbouring subunit (Fan et al., 2015). Hence, structures formed by the TMD5–TMD6 loop guard the luminal entrance to the pore, whereas the cytosolic vestibule is formed by the extended TMD6. Each of these regions is enriched in acidic residues that probably contribute to the cation selectivity of IP₃R and RyR (des Georges et al., 2016; Fan et al., 2015; Paknejad and Hite, 2018).

A conserved Ca²⁺-binding site is present in both RyR (des Georges et al., 2016) and IP₃R (Paknejad and Hite, 2018). The site is formed, in the case of IP₃R, by residues near the C-terminal end of ARM3 and by another residue contributed by the LNK domain (Fig. 3). In RyR, the equivalent residues are proposed to coordinate the Ca²⁺ required for stimulation (des Georges et al., 2016). The same may hold true for IP₃Rs, but this has yet to be tested. A conserved glutamate residue on the bottom surface of the ARM3 domain (Glu²¹⁰¹ in IP₃R1) previously suggested to mediate Ca²⁺ regulation of IP₃R (Miyakawa et al., 2001) and RyR (Fessenden et al., 2001), does not contribute to Ca²⁺ binding to this site, but it does stabilize the interaction between the cooperating domains in RyR1 (des Georges et al., 2016). A second Ca²⁺-binding site was identified in the structure of IP₃R3, and again it is formed by residues that are contributed by different domains (ARM3 and the α-helical domain linking ARM1 to ARM2) (Paknejad and Hite, 2018) (Fig. 3B). Formation of both Ca²⁺-binding sites requires movement of the contributing domains from their positions in the apo-state (i.e. IP₃R without bound IP₃), so as to bring the Ca²⁺-coordinating residues into register (Paknejad and Hite, 2018). This important observation is consistent with evidence that IP₃ controls IP₃R gating by regulating Ca²⁺ binding (Fig. 1B).

Taken together, analyses of the structures of full-length IP₃Rs have allowed researchers to define where IP₃ binds, identify Ca²⁺-binding sites that may mediate Ca²⁺ regulation, and establish that hydrophobic residues projecting into the pore must move to allow Ca²⁺ to pass.

The only contacts between the large cytosolic structures of the IP₃R and its channel region are the C-terminal end of ARM3 and the LNK domain (Fig. 3) (Fan et al., 2015). There are similar contacts in RyR (des Georges et al., 2016). In both IP₃R and RyR, this critical nexus comprises a platform of interleaved structures: the C-terminus of the ARM3 domain (the intervening lateral domain, ILD) forms a ‘thumb-and-fingers’ arrangement of an upper thumb abutting the bulk of ARM3, and an α-helical pair of fingers lying below and forming a cavity into which the LNK domain inserts (Fig. 3) (Fan et al., 2015). Mutations within the thumb disrupt IP₃R function (Hamada et al., 2017). The LNK domain also wraps around the thumb and contributes a residue to the Ca²⁺-binding site at the base of the ARM3 domain.

How, then, does IP₃ binding to the IBC cause hydrophobic pore residues some 7 nm distant to move and allow Ca²⁺ to pass from the ER lumen to the cytosol (Fan et al., 2015)? Recalling that IP₃ primes IP₃Rs to bind Ca²⁺, which then triggers channel opening (Adkins and Taylor, 1999) (Fig. 1B), it seems reasonable to speculate that IP₃ binding to the IBC is communicated to the Ca²⁺-binding site at the ILD–LNK nexus and thence to the pore (Paknejad and Hite, 2018). IP₃ binding closes the clam-like IBC, and, with IBC-β held firmly in place by inter-subunit interactions at the top of the mushroom, IBC-α moves and initiates conformational changes throughout the associated ARM domains. These changes include disruption of inter-subunit interaction between ARM1 and ARM2 domains, and rotation of the LNK domains (Paknejad and Hite, 2018). Here, the need for the SD is attributed to its role in stabilizing inter-subunit interactions to provide a fixed structure against which movement of IBC-α can leverage conformational changes through the ARM domains (Paknejad and Hite, 2018). Given the essential role of the SD in IP₃R activation, an alternative possibility was that the direct contact between the SD and ARM3 might mediate communication between N-terminal regions and the ILD. However, the SD–ARM3 interaction occurs through the handle of the hammer-like SD, which can be deleted without impairing IP₃R function (Yamazaki et al., 2010). Another possibility was that interaction between IBC-β and the C-terminal α-helical domain (CTD), which is unique to IP₃R, might communicate IP₃ binding to the LNK domain. However, this scheme is difficult to reconcile with functional IP₃R/RyR chimeras (Seo et al., 2012) since the RyR structure does not have an extended CTD, and with evidence that deletion of residues within the CTD that interact with IBC-β do not prevent IP₃-evoked Ca²⁺ release (Hamada et al., 2017; Schug and Joseph, 2006). Whatever the exact path from IBC to the ILD–LNK nexus is, IP₃-evoked conformational changes appear to reconfigure the Ca²⁺-binding site formed at the LNK–ARM3 interface to allow Ca²⁺ binding (Paknejad and Hite, 2018), thereby providing a plausible mechanism for IP₃ priming IP₃Rs to respond to Ca²⁺ (Fig. 1B).

We conclude that analyses of IP₃ binding contributed to defining the SAR for IP₃Rs and to quantitative comparisons of the relationship between binding and channel activation, but most significantly they allowed IP₃Rs to be identified during their purification, which paved the way to cloning and molecular manipulation of IP₃Rs, and to structural studies. The latter have allowed investigators to establish that IP₃Rs are huge tetrameric structures, wherein IP₃ binding closes a clam-like IBC. This conformational change is communicated to a critical nexus between interleaved structures from the cytosolic and channel domains. IP₃ binding probably stabilizes Ca²⁺ binding to this nexus, leading to rearrangement of the pore, such that occluding hydrophobic residues are displaced to allow the passage of Ca²⁺ from the ER lumen to the cytosol.

Electrical recordings from ion channels, most often by means of patch-clamp recording (Box 3) (Lape et al., 2008; Neher, 1992), allow the opening and closing of single channels to be recorded with sub-millisecond resolution, and they allow their ion permeation properties to be defined. Because the intracellular location of RyR and IP₃R in the ER presents a formidable barrier to such recordings (Jonas et al., 1997), two alternative approaches have been used.

The first approach, which involves reconstitution of ER vesicles or solubilized IP₃Rs into planar lipid bilayers, provided the first measurements of currents through IP₃R (Bezprozvanny et al., 1994, 1991; Bezprozvanny and Ehrlich, 1994; Ehrlich and Watras, 1988; Maeda et al., 1991). These analyses established that IP₃R, like RyR, are large-conductance cation channels with relatively low selectivity for Ca²⁺. Both features are important in allowing IP₃R to generate large local cytosolic Ca²⁺ signals: the large conductance allows an open IP₃R to pass ∼500,000 Ca²⁺ ions per second (Foskett et al., 2007), whereas the weak selectivity might allow a counter-flux of K⁺ to dissipate the electrical gradient that is formed as Ca²⁺ leaves the ER and which would otherwise rapidly terminate Ca²⁺ release (Zsolnay et al., 2018). The short, wide selectivity filter and large vestibules with abundant acidic residues probably provide the structural basis of these ion permeation properties (Fan et al., 2015). Bilayer analyses also allowed confirmation of the biphasic regulation of IP₃R1 by cytosolic Ca²⁺ (Bezprozvanny et al., 1991). A potential problem with recordings from planar lipid bilayers is that solubilization and/or reconstitution could lead to loss of accessory proteins or perturbation of structure. Maximal open probabilities recorded from IP₃Rs in bilayers, for example, are much lower than in patch-clamp recordings, and bilayer recordings of IP₃R2 and IP₃R3 failed to capture the inhibitory effect of cytosolic Ca²⁺ (Hagar et al., 1998; Ramos-Franco et al., 2000).

The second approach to obtaining electrical recordings from IP₃R exploits the fact that the ER is continuous with the outer nuclear membrane (ONM) (Box 3) (Dingwall and Laskey, 1992). Hence, patch-clamp recording from the ONM allows analysis of IP₃R in a native membrane, albeit not the ER (Mak et al., 2013; Rahman and Taylor, 2010) (Box 3). Examination of these recordings, which have been applied to both native and heterologously expressed IP₃Rs (Betzenhauser et al., 2008; Cheung et al., 2010; Foskett et al., 2007; Marchenko et al., 2005; Rahman et al., 2009), confirmed the ion permeation properties of IP₃Rs and the biphasic regulation of all IP₃R subtypes by Ca²⁺. The recordings have also suggested complex gating schemes wherein IP₃ drives bursts of IP₃R activity by extending the duration of sequences of openings and shortening the gaps between the bursts (Gin et al., 2009; Ionescu et al., 2007).

Another application of nuclear patch-clamp recording is provided by our work, where we showed that IP₃Rs within patches that fortuitously contained several IP₃Rs behave differently to patches with only a single IP₃R (Rahman et al., 2009). This led to our proposal that low concentrations of IP₃, perhaps arising from occupancy of only some of the four IP₃-binding sites, trigger IP₃R clustering (Rahman et al., 2009). The clustered IP₃Rs, we suggest, are better placed than lone IP₃Rs to benefit from CICR when a near neighbour opens to release Ca²⁺ and so provide the second stimulus that is needed for IP₃R opening (Fig. 1B). Effects of clustering on the IP₃ and Ca²⁺ sensitivity of IP₃Rs reinforce the propensity of clustered IP₃Rs to amplify Ca²⁺ signals by CICR. These proposals have been challenged (Rahman et al., 2011; Smith et al., 2009; Vais et al., 2011) and our own recent work suggests that even in unstimulated cells there are pre-existing clusters of IP₃Rs, each typically comprising about eight IP₃Rs (Thillaiappan et al., 2017). Our revised proposal therefore envisages that IP₃Rs are, as we have shown, loosely clustered in unstimulated cells (Thillaiappan et al., 2017) and that IP₃ might then cause IP₃Rs within the cluster to huddle more closely and so be more likely to respond to Ca²⁺ released by a neighbour.

Throughout the long history of analyses of intracellular Ca²⁺ signalling, there has been a productive interplay between studies of intact tissues and of biological systems extracted from intact cells (ex cellula). These approaches have revealed that the ER is the major intracellular Ca²⁺ store and allowed identification of the enormous channels (RyR and IP₃R) that mediate Ca²⁺ release from the ER. We are now fast approaching an understanding of how IP₃ binding leads, through its interactions with Ca²⁺ binding, to opening of the IP₃R. In parallel with these approaches, developments in optical microscopy have provided opportunities to examine IP₃-evoked Ca²⁺ release with exquisite temporal and spatial resolution in intact cells. We can surely look forward to these analyses converging with structural analyses in situ to provide a comprehensive understanding of IP₃Rs in living cells.