Synthetic partial agonists reveal key steps in IP₃ receptor activation

Abstract

Inositol 1,4,5-trisphosphate receptors (IP₃R) are ubiquitous intracellular Ca²⁺ channels. IP₃binding to the IP₃-binding core (IBC) near the N-terminal initiates conformational changes that lead to opening of a pore. The mechanisms are unresolved. We synthesized 2-O-modified IP₃ analogues that are partial agonists of IP₃R. These are like IP₃ in their interactions with the IBC, but they are less effective than IP₃ in rearranging the relationship between the IBC and N-terminal suppressor domain (SD), and they open the channel at slower rates. IP₃R with a mutation in the SD occupying a position similar to the 2-O-substituent of the partial agonists has a reduced open probability that is similar for full and partial agonists. Bulky or charged substituents from either the ligand or SD therefore block obligatory coupling of the IBC and SD. Analysis of ΔG for ligand binding shows that IP₃ is recognised by the IBC and conformational changes then propagate entirely via the SD to the pore.

Inositol 1,4,5-trisphosphate receptors (IP₃R) are ligand-gated channels. They are expressed in most animal cells and mediate release of Ca²⁺ from the endoplasmic reticulum in response to the many stimuli that evoke IP₃ formation. IP₃R are tetrameric, and each subunit of about 2700 residues has an IP₃-binding site near the N-terminus and six transmembrane domains (TMD) towards the C-terminus (Fig. 1a)¹. The pore is formed by the last pair of TMD and the intervening loop, the pore-loop (“P-loop”), from all four subunits¹. The structure of the pore is predicted to be broadly similar to the pores of other tetrameric P-loop channels, like bacterial K⁺ channels, for which high-resolution structures are available². IP₃ binds to a discrete part of the IP₃R, the IP₃-binding core (IBC, residues 224-604, Fig. 1a)³. Although the extreme N-terminus (residues 1-223) is not required for IP₃ binding, it decreases the affinity for IP₃ and has therefore been called the suppressor domain (SD)⁴. The SD is thought to be required for channel gating because IP₃ binds to IP₃R without an SD, but it no longer opens the pore^4,5. However, the links between IP₃ binding and gating are not understood, and nor do we have a structure of the entire IP₃R at sufficient resolution to provide insight into these gating mechanisms^1,6.

Figure 1.

Structure of the IP₃R and its ligands. (a) Key domains of IP₃R (numbering from rat IP₃R1, accession code NM_001007235). Pink denotes the SD (residues 1-223), red the IBC (224-604), and black vertical lines represent TMD. The SD (β₁) and IBC (β₂ and armadillo-like repeat, ARM) comprise 3 stably folded domains connected by flexible linkers (L1 and L2)³⁴. Crystal structures are shown below³,³³. (b) Agonist binding (yellow) to a discrete site on the receptor (IBC for IP₃R, red) evokes conformational changes that propagate through the receptor and which then affect (arrows) the binding site. Removal (boxed diagrams) of a domain through which conformational changes must pass prevents this energetic interplay between conformational changes and binding. (c) Structures of the ligands used.

Activation of ligand-gated ion channels begins with agonist binding to a stable closed state and proceeds via many short-lived intermediates to a state in which the pore is open⁷. This activation may proceed entirely through a sequence of incremental changes in the receptor⁸ or be dominated by a single concerted transition between two stable conformations⁹. Agonists differ in both the strength of their binding (affinity) and in their ability to drive the receptor to its open state (efficacy)¹⁰. A ligand with reduced efficacy must occupy more receptors than a full agonist to evoke the same cellular response. Such partial agonists, by occupying receptors, diminish the response to a full agonist¹¹. Partial agonists are particularly useful for exploring the mechanisms of receptor activation because they lie between full agonists and antagonists in their ability to activate receptors^7,10. This is true for all receptors, but ligand-gated ion channels are uniquely amenable to such analyses because single-channel recording allows key conformational changes of single receptors to be determined with outstanding temporal resolution⁷.

For these ligand-gated ion channels, full and partial agonists may differ in either the frequency with which they cause the receptor to visit the fully active open state or they may stabilize different open states that mediate lesser ion fluxes. In both cases, a partial agonist evokes lesser activation. Two subtypes of ionotropic glutamate receptors (iGluR) illustrate the distinction. These receptors, which mediate most excitatory neurotransmission in the brain, are those for which the structural basis of efficacy has been most thoroughly explored^12,13. For the AMPA subtype of iGluR, a series of partial agonists, which differ from each other by only a single atom, close the clam-like binding site to varying degrees, but less completely than do full agonists. Partial agonists thereby preferentially open the pore to states with lesser conductance¹². For full and partial agonists of the NMDA subtype of iGluR, the conformational changes in the binding site are more subtly different. Both cause similar closure of the clam, and they fully open the pore, but the conformational changes proceed more slowly for partial agonists^14,15.

Affinity and efficacy are distinguishable, but the two properties are not independent because energy provided by agonist binding drives the conformational changes that cause channel opening^10,16,17. This “binding-gating problem” is a fundamental issue in pharmacology¹⁰, but it can also be turned to advantage because the interplay depends upon both the efficacy of the ligand and the presence of the parts of the receptor through which conformational changes must pass. The former because partial agonists divert less binding energy than full agonists into effective conformational changes; and the latter because receptors lacking essential domains are expected to be less able to divert binding energy into conformational changes (Fig. 1b). Analyses of the free energy changes for ligand binding (ΔG = -RTlnK_d, where K_d is the equilibrium dissociation constant) can thus provide insight into the conformational changes evoked by agonist binding. Comparisons of ΔG for full and partial agonists, and for agonist binding to normal and truncated IP₃R, can therefore contribute to defining the links between IP₃ binding and opening of the pore.

Here we synthesize a new series of partial agonists of the IP₃R, and, in defining their properties, we identify a novel form of partial agonism that allows us to define key steps in IP₃R activation.

RESULTS

Synthesis of 2-O-modified analogues of IP₃

All high-affinity agonists of all IP₃R have structures equivalent to the vicinal 4,5-bisphosphate and 6-hydroxyl of IP₃ (Fig. 1c), but the axial 2-hydroxyl is not required¹⁸. The essential phosphate moieties interact predominantly with opposite sides of the clam-like IBC (P-4 with the β₂-domain, and P-5 with the ARM domain)³ (Fig. 1a), suggesting that agonists might close the clam in a manner reminiscent of glutamate binding to iGluR^6,13.

In seeking to develop novel high-affinity ligands of IP₃R that might differ in efficacy, we focused on the 2-OH group of IP₃ because earlier structure-activity analyses had suggested that analogues modified at this position retain activity¹⁸. The X-ray structure of the IBC with IP₃ bound subsequently confirmed that the 2-OH group of IP₃ makes no significant contacts with the IBC³.

We began by preparing homo-dimers of IP₃ with linkers of various lengths (2, 3 and 4 in Fig. 1c), aiming initially to define the separation of IP₃-binding sites within a tetrameric IP₃R. However, informed by our initial results^19,20 and cognizant that dimeric cGMP is a partial agonist of a cGMP-gated cation channel²¹, we extended our work to include syntheses of additional 2-O-modified analogues (Fig. 1c) and an assessment of their efficacy.

The shortest IP₃ dimer (2) is a symmetrically substituted N,N’-diethyl urea, synthesized by cross-linking of a protected D-2-O-(2-aminoethyl)-IP₃ building block using bis(4-nitrophenyl) carbonate¹⁹. A modification of this synthetic method was used to synthesize hetero-dimers such as 5, 6 and 7, in which the second IP₃ moiety is replaced by a different inositol phosphate or by inositol (Scheme 1). We also synthesised an L-IP₃ homo-dimer (8, the enantiomer of 2) (Scheme 1) and an IP₃-adamantane conjugate (9). The syntheses of 2-deoxy-IP₃ (10)²² and 2-O-(2-aminoethyl)-IP₃ (11)²³ were reported previously. Details of the synthetic procedures and compound characterizations are provided in Supplementary Methods online.

Scheme 1.

Syntheses of hetero-dimers 5, 6 and 7, and L-IP₃ dimer 8. Dimers 5 and 8 were synthesized from L-IP₃-based building block 15, obtained from diol 14​⁴⁸ (see Supplementary Methods online). The N-trifluoroacetyl protecting group was removed, generating an unstable amine which was reacted with 0.5 equivalents of bis(4-nitrophenyl) carbonate, giving protected L-IP₃ dimer 16. Hydrogenolytic deprotection of 16 gave L-IP₃ dimer 8. When 1 equivalent of bis(4-nitrophenyl) carbonate was used, the product was 4-nitrophenyl N-alkylcarbamate 17, which could be isolated and conjugated with D-IP₃ component 11. The conjugation reaction was carried out in CD₃OD and monitored by ³¹P NMR spectroscopy. Deprotection followed by anion-exchange chromatography then gave IP₃–L-IP₃ hetero-dimer 5. Dimers 6 and 7 were synthesized from alcohol 18​⁴⁹. Nitrile 19 was reduced, and the amine product was temporarily protected as the N-trifluoroacetamide (20). Acid-labile protecting groups were then removed, giving pentaol 21, which was converted, via 22, into carbamate 23. Carbamate 23 was then conjugated with 11, and deprotection followed by anion-exchange chromatography gave IP₃–IP₅ hetero-dimer 6. Alternatively, conjugation of carbamate 24 with 11 gave IP₃–Ins dimer 7. Reagents and conditions: (i) LiOH, THF, MeOH, H₂O; (ii) bis(4-nitrophenyl) carbonate (0.5 equiv), THF; (iii) bis(4-nitrophenyl) carbonate (1 equiv), THF; (iv) H₂, Pd(OH)₂/C, MeOH, H₂O; (v) 11, CD₃OD, Et₃N; (vi) NaH, BrCH₂CN, CH₃CN; (vii) LiAlH₄, THF; (viii) EtOC(O)CF₃, THF; (ix) TFA, H₂O; (x) (BnO)₂PNⁱPr₂, 1H-tetrazole, CH₂Cl₂ then 3-chloroperoxybenzoic acid; (xi) Et₃N, H₂O, reflux; (xii) bis(4-nitrophenyl) carbonate (1 equiv), DMF, Et₃N. Bn, benzyl; PMB, 4-methoxybenzyl. All experimental procedures are described in detail in Supplementary Methods online.

2-O-modified IP₃ analogues are high-affinity agonists of IP₃R

We used a cell line that expresses only recombinant rat IP₃R1 (DT40-IP₃R1 cells)²⁴ to measure Ca²⁺ release from intracellular stores, and IP₃R1 purified from rat cerebellum to measure IP₃ binding. These analyses show that homo-dimers of IP₃ linked through the 2-O-positions of the inositol rings are high-affinity agonists of IP₃R. The shortest dimer (2) (Fig. 1c) binds to IP₃R1 with greater affinity than IP₃ (Table 1 and Supplementary Fig. 1a online) and stimulates Ca²⁺ release from intracellular stores at lower concentrations than does IP₃ (Table 1 and Supplementary Fig. 1b online). 2 is the most potent inositol phosphate-based agonist so far identified.

Table 1.

Responses to IP₃ analogues. The effects of each analogue on Ca²⁺ release from the intracellular stores of permeabilized DT40-IP₃R1 cells and on ³H-IP₃ binding to full-length purified IP₃R1 (FL), its N-terminal (NT, residues 1-604) or the IP₃-binding core (IBC, residues 224-604) are summarized (n ≥ 4). For dimers, the estimated separation of the two moieties is shown calculated as in⁵⁰. Results are means ± SEM. *No detectable Ca²⁺ release with 30μM 8. ND, not determined.

		Ca2+ release		FL		NT	IBC
		EC50 (nM)	Release %	Kd (nM)	EC50/Kd	Kd (nM)	Kd (nM)
1	IP3	20±2	77±5	12.5±1.06	1.6±0.2	2.82±0.26	0.21±0.03
13	AdA	1.5±0.1	75±6	1.26±0.11	1.2±0.01	ND	ND
2	(IP3)2 0.8nm	2.7±0.4	76±6	0.45±0.04	6.0±1.0	0.41±0.03	0.18±0.01
3	(IP3)2 1.5nm	4.9±0.3	74±6	0.86±0.19	5.7±1.3	0.47±0.09	0.18±0.01
4	(IP3)2 8nm	13±1	73±7	1.39±0.23	9.3±1.6	1.37±0.20	0.48±0.02
5	IP3–L-IP3 0.8nm	5±1	70±7	0.90±0.12	5.6±1.3	0.42±0.02	0.14±0.02
6	IP3–IP5 0.8nm	5±0	72± 5	0.89±0.11	5.6±0.7	0.36±0.03	0.20±0.01
7	IP3–Ins 0.8nm	21±1	76±4	7.82±1.73	2.7±0.6	2.81±0.22	0.33±0.01
8	(L-IP3)2 0.8nm	Inactive*	ND	ND	ND	ND	ND
9	2-adamantane-IP3	30±3	71±1	7.62±0.51	3.9±0.5	1.63±.022	0.20±0.02
10	2-deoxy-IP3	32±7	65±3	17.1±1.1	1.9±0.4	4.01±0.38	0.22±0.03
11	2-aminoethyl-IP3	42±5	69±6	43.6±4.6	0.95±0.02	13.5±2.0	0.52±0.04

A homo-dimer of L-IP₃ (8) is, as expected, inactive because L-IP₃ does not bind to the IBC¹⁸. However, homo-dimers of IP₃ with longer linkers (3, 4) also bind to IP₃R with greater affinity than IP₃, as do hetero-dimers in which IP₃ is linked to inositol (inositol-IP₃​, 7), an unrelated bulky hydrophobic group (adamantane-IP₃​, 9), or to an inositol phosphate that does not itself bind to the IBC (IP₃-IP₅​, 6; or IP₃​-L-IP₃​, 5) (Table 1). The latter (5-7 and 9) demonstrate that high-affinity binding of IP₃ dimers does not result from an interaction with a second specific IP₃-binding site, nor does it result from alternating association of the two IP₃ moieties with the IBC¹⁹. Furthermore, the two components of the dimer must be linked, because a high concentration of IP₅ (12, 10μM) had no effect on the Ca²⁺ release evoked by IP₃ or 2-deoxy-IP₃ (10) (Supplementary Fig. 1c online). We conclude that addition of bulky or charged groups to the 2-O-position of IP₃ produces high-affinity agonists of the IP₃R.

A new family of partial agonists of IP₃R

Because a partial agonist less effectively activates its receptor than a full agonist, it must occupy more receptors to evoke the same cellular response¹¹. In our Ca²⁺ release assays, where each ligand caused the same maximal Ca²⁺ release as IP₃ (Table 1), we can therefore gain some insight into the efficacy of a ligand by comparing the concentration that causes 50% of the maximal response (EC₅₀) with that which occupies 50% of the binding sites (K_d).

For each ligand, we compared the EC₅₀/K_d ratio using DT40-IP₃R1 cells for the functional assays²⁴ (EC₅₀) and purified IP₃R1 to measure IP₃ binding (K_d). Our results suggest that 2 occupies more IP₃R than IP₃ to evoke the same Ca²⁺ release (2 has a higher EC₅₀/K_d ratio, Fig. 2a and Table 1). This indicates that 2 may be a partial agonist. These characteristics, an increase in both affinity and EC₅₀/K_d ratio, are shared by very different 2-O-modified IP₃ analogues (Fig. 2a and Table 1). They do not, therefore, depend upon precise structural features: an IP₃ moiety that binds to the IBC and a 2-O-substitutent larger than adamantane (9) are sufficient to increase both the affinity and EC₅₀/K_d ratio.

Figure 2.

2-O-modified IP₃ analogues are partial agonists of IP₃R. (a) From experiments similar to those shown in Supplementary Fig. 1a,b online, EC₅₀/K_d ratios (n ≥ 4) were calculated for each ligand. (b) Traces (each typical of at least 3 similar records) show excised nuclear patch-clamp recordings from DT40-IP₃R1 cells with the pipette solution containing ATP (0.5mM), a free [Ca²⁺] of 200nM and the indicated ligands (10μM, except where shown otherwise). The holding potential was +40mV. C denotes the closed state. (c) Cells were treated with IP₃ alone, or for 30s with either 0.1nM AdA or 2 and then with the indicated concentrations of IP₃. Results (n = 3) show the concentration-dependent release of Ca²⁺ by IP₃. (d) Current-voltage (i-V) relationship for patches stimulated with IP₃ or 2 (means ± SEM, n = 3). (e) Summary data showing P_o and mean closed time (τ_c) for IP₃R1 stimulated as shown, n = 3-11 (further details in Supplementary Table 1 online). The simplified activation scheme for IP₃R shows the transition between closed (C) and open (O) states determined by rate constants, β and α (see Supplementary Methods online). All results (a,c-e) are means ± SEM.

The EC₅₀/K_d ratio for adenophostin A (AdA, 13, Fig. 1c)²⁵, another high-affinity agonist of IP₃R, is similar to that for IP₃ (Fig. 2a and Table 1). This is consistent with single channel analyses, where IP₃ and AdA cause the IP₃R to open to the same maximal single channel open probability (P_o, the fraction of time that each channel spends in its open state) (Fig. 2b and Supplementary Table 1 online)²⁶. We conclude that IP₃ and AdA are full agonists of the IP₃R, whereas 2-7 and 9 appear to be partial agonists.

The results shown in Fig. 2c confirm that 2 must occupy more IP₃R than the full agonist, AdA, to evoke the same Ca²⁺ release. DT40-IP₃R1 cells were first pre-treated with concentrations of AdA or 2 that caused the same Ca²⁺ release (5.7 ± 1.0% and 5.4 ± 1.3% of the intracellular stores, respectively) and then stimulated with IP₃. More IP₃ is required to evoke further Ca²⁺ release after treatment with 2 than after AdA (EC₅₀ = 44.4 ± 3.19 and 4.37 ± 0.18nM, respectively). This confirms that 2 occupies more IP₃R than AdA to evoke the same Ca²⁺ release.

Because the nuclear envelope is continuous with the endoplasmic reticulum²⁶, we can use patch-clamp recording from the outer nuclear envelope of DT40-IP₃R1 cells to resolve the behaviour of single IP₃R. To both maximize the amplitude of the currents recorded and to avoid the complexity of feedback regulation of IP₃R by Ca²⁺ passing through them¹, we used K⁺ as the charge-carrier in these experiments^26,27. Both the single channel K⁺ conductance (γ_K) and the mean channel open time (τ_o) were the same for all agonists examined (Fig. 2b,d and Supplementary Table 1 online). The open state of the IP₃R thus appears to be similar whether it is evoked by binding of a full (AdA, IP₃ and 10) or partial agonist (2, 6 and 9).

However, for IP₃R activated by maximal concentrations of AdA, IP₃ or 10, P_o was higher than with 2, 6 or 9 (Fig. 2b,e and Supplementary Table 1 online). Increasing the concentration of 2 (from 0.5 to 10μM, Fig. 2b,e) did not further increase P_o, and after stimulation with a mixture of IP₃ and 2 (10μM of each) P_o was significantly less than with IP₃ alone (Fig. 2b and Supplementary Table 1 online). These analyses of single IP₃R confirm and extend the results obtained with Ca²⁺ release and binding assays (Table 1 and Fig. 2a). Analogues of IP₃ with bulky additions to the 2-O-position (2-7 and 9) are high-affinity partial agonists of IP₃R1. Subsequent analyses of the mechanisms underlying these properties of the 2-O-substituted analogues focus on 2 (Fig. 1c).

Our single channel analysis shows that whereas τ_o is similar for all agonists (Supplementary Table 1 online), mean channel closed times (τ_c) were longer for the partial agonists (2, 6 and 9) than for full agonists (IP₃, AdA and 10) (Fig. 2e and Supplementary Table 1 online). The latter, assuming a simplified activation scheme (Fig. 2e, Supplementary Methods online), reveals that the rate constant for channel opening (β = 1/τ_c) with partial agonists is less than with full agonists. The 2-O-modified analogues are the first partial agonists of IP₃R for which the basis of their reduced efficacy has been established. They open the channel fully (Fig. 2b,d), the channel closes at the same rate whether it has a partial agonist or IP₃ bound (α in Fig. 2e), but the rate constant for channel opening (β) is lower for partial agonists (Fig. 2e and Supplementary Table 1 online). We conclude that these full and partial agonists drive the IP₃R into a similar open state, but the partial agonists do so less effectively.

Full and partial agonists differ in how they rearrange the IBC-SD

For IP₃R, conformational changes evoked by IP₃ binding to the IBC near the N-terminal must be transmitted to the pore formed by residues close to the C-terminal (Fig. 1a). Some of the energy provided by IP₃ binding is used to drive the opening of the pore. The K_d (ΔG = -RTlnK_d) measured in a binding assay is therefore determined by both the strength of the contacts between IP₃ and the IBC (“intrinsic binding affinity”)¹⁶ and the ensuing conformational changes¹⁰.

The IBC includes all the amino acid residues that contact IP₃ (Fig. 1a)^3,28 and each 2-O-modified agonist (2-7 and 9-11) (Fig. 1c) retains the groups within IP₃ that interact with the IBC. Furthermore, each of these ligands binds with similar affinity to the IBC alone (Fig. 3a and Table 1). Because the full (IP₃) and partial agonists (2-7, 9) are both expected to make the same contacts with the IBC and are also observed to bind to it with similar affinity, we suggest that they do not differ in the binding energy they divert into changing the conformation of the IBC. This contrasts with AMPA receptors, where the clam-like binding site closes more fully with more efficacious agonists^12,13. The distinction highlights two fundamentally different ways of reducing efficacy, a defining feature of all ligand-receptor interactions¹⁰. A partial agonist may fail to make optimal contacts with the binding site and so less effectively activate the receptor (e.g., AMPA receptors¹²), or it may impair onward transmission of conformational changes. Subsequent experiments demonstrate that our partial agonists (2-7 and 9) belong to the second category. They are thereby useful in defining the steps that follow IP₃ binding.

Figure 3.

Partial agonists are IP₃-like in their interactions with the IBC. (a) Equilibrium-competition binding to IBC (top) and NT (bottom) with ³H-IP₃ and either IP₃ or 2, n ≥ 17. (b) ΔΔG (ΔG^IBC-ΔG^NT), reflecting ΔG used to rearrange the IBC-SD relationship, is shown for each ligand, and compared with the efficacy of each (EC₅₀/K_d, with K_d determined for full-length IP₃R1). Results (a,b) are means ± SEM. (c) Estimated ΔG for conformational changes associated with IP₃R activation. The affinity (K_d) of IP₃ for IP₃R1 truncated as shown was measured herein (Table 1) or by others: by⁴ for ΔSD (IP₃R1 lacking residues 1-223), and by³² for Δloop (IP₃R1 lacking residues 2428-2437); ΔG was then calculated from ΔG = -RTlnK_d. The K_d for IP₃ was not directly measured in³², but under the conditions used the 4-fold increase in IP₃ binding after deletion of residues 2428-2437 (ie Δloop) is likely to reflect a 4-fold decrease in K_d. We assume that deletion of IP₃R fragments through which conformational changes must pass increases IP₃ affinity because less binding energy is diverted into re-arranging the protein (Fig. 1b). Deletions of many other regions (shown in blue) do not increase IP₃ affinity⁴, suggesting that the IP₃-evoked conformational changes do not pass through them. This analysis is consistent with each IP₃ binding event diverting ~9kJ/mol into conformational changes of the IP₃R, of which ~6kJ/mol rearranges the SD-IBC relationship, and ~3kJ/mol is used by the SD to gate the pore.

For all three IP₃R subtypes, IP₃ binds to the IBC with greater affinity than to either full-length IP₃R or the NT (Fig. 3a, Table 1 and Supplementary Table 2 online)^28,29. The SD reduces the IP₃ binding affinity through its intramolecular interaction with the IBC²⁸ and appears also to mediate communication between the IBC and pore^4,5. We therefore examined the contribution of the SD to the conformational changes initiated by IP₃ via analysis of ΔG for ligand binding.

Removal of the SD increases the affinity of the NT for IP₃, but it has lesser effects on binding of the partial agonists (Table 1). Efficacy (reported by the EC₅₀/K_d ratio) and the difference in ΔG (ΔG = -RTlnK_d) for binding to the IBC and NT (ΔΔG) are inversely correlated (Fig. 3b). Because we suggest that each agonist contributes similar “intrinsic binding energy”^16,17 through the similar interactions that each makes with the IBC (Table 1 and Supplementary Table 2 online), the different ΔΔG for binding of full and partial agonists to the NT must reflect the extent to which each uses binding energy to rearrange the relationship between the IBC and SD^16,17,30. We conclude that full and partial agonists differ minimally in their interactions with the IBC, but radically in how they rearrange its relationship with the SD.

Conformational changes pass from the IBC entirely via the SD to the pore

IP₃ binds only to a small contiguous sequence within the IP₃R, the IBC (Fig. 1a). Truncations of the IP₃R might therefore disconnect IP₃ binding from downstream conformational changes without directly perturbing the IP₃-binding site. These truncated IP₃R might then reveal, via analysis of ΔG for ligand binding, the parts of the IP₃R through which IP₃-evoked conformational changes must pass (Fig. 1b).

All full-length IP₃R subtypes bind IP₃ with only slightly lower affinity than the NT (ΔΔG ca. -3kJ/mol)²⁸, whereas the NT and IBC differ more substantially in their affinities for IP₃ (ca. -6kJ/mol) (Table 1 and Supplementary Table 2 online). This suggests that the most costly conformational changes evoked by IP₃ occur within the NT (~6kJ/mol) with downstream events requiring less energy (~3kJ/mol) (Fig. 3c). Removing the SD from full-length IP₃R increases its affinity for IP₃ by an amount (≤ ca. - 9kJ/mol)⁴ consistent with uncoupling IP₃ binding from all the conformational changes downstream of the IBC (Fig. 3c). These analyses suggest that the IBC communicates with the rest of the IP₃R entirely via the SD.

A site within the first 340 residues of the IP₃R, which includes the SD, appears to interact with a short cytosolic loop linking TMD 4 and 5 (Fig. 3c). This interaction has been proposed to open the pore directly^31,32. Disruption of this loop increases the affinity of the IP₃R for IP₃ by an amount (ca. -3kJ/mol)³² that matches the estimated cost of all conformational changes downstream of the SD (Fig. 3c).

These analyses corroborate our suggestion that conformational changes pass directly and exclusively from the IBC to the SD, and then perhaps directly to the TMD4-5 loop^31,32.

Point mutations within the SD mimic partial agonists

Removal of the SD and additions to the 2-O-position of IP₃ similarly increase binding affinity (Table 1). The latter, we suggest, because the analogues evoke lesser conformational changes in the IP₃R. Both modifications also uncouple ligand binding from gating, although removal of the SD does so more completely^4,5 than do the 2-O modifications to IP₃. We therefore speculated that 2-O-modified analogues partially mimic removal of the SD by disrupting its interaction with the IBC and that this causes both a decrease in efficacy and an increase in affinity.

The SD has a structure reminiscent of a hammer with a large head and short handle (described earlier as an “arm”)³³ (Fig. 4a). Others³³ have shown that removing the handle of the SD (residues 67-108) minimally affects IP₃ binding to the NT. But mutation of highly conserved residues on the surface of the head domain, most notably within the β2-β3 loop (loop 2)³³, increases the affinity of the NT for IP₃. We therefore tested our hypothesis that 2-O-modified analogues of IP₃ disrupt the IBC-SD interface by mutagenesis of residues in the β2-β3 loop and of other residues nearby in the 3D structure of the SD (Fig. 4a). As reported³³, several mutations increased the affinity of the NT for IP₃, with the most effective (V33K) almost mimicking the effect of removing the entire SD. Another mutation (K52E) had no effect (Supplementary Table 3 online)³³. Furthermore, and consistent with our suggestion that 2-O-substituents of IP₃ disrupt the IBC-SD interaction, the effective mutations had lesser effects on binding of 2 to the NT (Fig. 4b and Supplementary Table 3 online). From these non-additive effects, we conclude that binding of 2 displaces the SD in a manner that mimics its removal or displacement by appropriate mutations.

Figure 4.

Point mutations within the SD mimic partial agonists. (a) The structure of the SD³³ is shown highlighting the residues mutated in this study. (b) Relative affinities (K_d) of IP₃ and 2 for IBC, NT, and NT with the indicated mutations (Supplementary Table 3 online); n ≥ 5. The dashed line shows K_dIP₃/K_d2 = 1. (c) Potency (EC₅₀) of IP₃ relative to 2 in releasing Ca²⁺ from permeabilized DT40 cells stably expressing mutant IP₃R1 (Supplementary Table 4 online); n ≥ 5. The dashed line shows EC₅₀IP₃/EC₅₀2 = 1. (d) Typical recordings from excised nuclear patches of DT40-IP₃R1^V33K cells with 10μM IP₃ or 2 in the patch pipette. The holding potential was +40mV. C denotes the closed state. (e) Summary data showing P_o and γ_K for IP₃R1 and IP₃R1^V33K stimulated with 10μM IP₃ or 2; n ≥ 3. Results (b,c,e) are means ± SEM.

Our results so far establish that the 2-O-substituents of the IP₃ analogues and appropriate point mutations within the SD cause similar increases in binding affinity. These effects mimic removal of the SD, leading us to conclude that they result from disrupted communication between the IBC and SD. Given that the 2-O-substituted analogues are partial agonists, and that the SD is required for IP₃ to gate the pore^4,5, we speculated that the point mutations might further mimic the analogues and give IP₃R that even full agonists are unable to activate fully.

In DT40 cells expressing IP₃R1 mutated within the SD (Fig. 4a and Supplementary Fig. 2a,b online), IP₃ and 2 evoke Ca²⁺ release from permeabilized cells and activate IP₃R in nuclear patch-clamp recordings (Fig. 4c-e, Supplementary Fig. 2c online, and Supplementary Table 4 online). The properties of these interactions are consistent with our prediction that disrupting the IBC-SD interaction decreases efficacy and increases agonist affinity by blocking propagation of conformational changes from the IBC. In permeabilized DT40 cells expressing IP₃R1 with the V33K mutation (IP₃R1^V33K), IP₃ and 2 are equipotent (Supplementary Fig. 2c online), and in single channel recordings each has the same P_o (Fig. 4d,e). This P_o is similar to that observed for normal IP₃R stimulated with 2, but lower than the P_o with IP₃ (Fig. 2b,e). The less effective mutations have lesser effects (Fig. 4c), consistent with our suggestion that they cause lesser disruption of the IBC-SD interaction.

The structures of the IBC-IP₃ and SD are known^3,33 (Fig. 1a), but not the relationship between them³⁴. We used protein-protein docking to identify a likely relationship between them (Supplementary Methods online). The three IP₃R subtypes differ in their affinities for IP₃, but their IBC share similar sequences and bind IP₃ with the same affinity²⁸. A subtype-specific interaction between the IBC and SD determines the different affinities of the three full-length IP₃R²⁸. Because the residues within the SD that confer these subtype-selective interactions^28,33 are likely to lie at an IBC-SD interface, this criterion was used to select between possible models of the IBC-SD complex. Our proposed model (Fig. 5a,b and Supplementary Fig. 3 online) is consistent with the radius of the NT-IP₃ complex from small-angle X-ray scattering³⁴. In this structure, four of the loops (loops 2 and 5, and part of loops 3 and 7)³³ that link the β-strands of the SD interact primarily with loops from the β₂-domain of the IBC (Supplementary Fig. 3 online). Within this IBC-SD structure, the second IP₃ moiety of 2 lies close to several point mutations in the SD (V33K, D34R, R36E) that reduce efficacy (Supplementary Table 4 online), each lying on the putative IBC-SD interface (within loop 2). The same interface includes the other effective mutation (K127E, within loop 5), but not the ineffective one (K52E) (Fig. 5a,b and Supplementary Fig. 3c,d online).

Figure 5.

IP₃ binding to the IBC activates IP₃R entirely via the SD. (a,b) Predicted relationship between the SD (pink) and IBC (red) with 2 bound. Residues within the SD that affect efficacy (V33, D34, R36 and K127) are shown in black (see Fig. 4a and Supplementary Fig. 3 online for details). The ineffective residue K52 is shown in pink. Panel b is an enlargement of the boxed area in panel a, with the IBC-bound IP₃ moiety indicated by an arrow. (c) IP₃ (yellow) rearranges the 2 domains of the IBC (β₂ and ARM, red) around its L2 loop causing rearrangement of the SD (= β₁, pink) around the L1 loop. The SD is then entirely responsible for transmitting conformational changes towards the pore, probably by directly interacting with the TMD4-5 loop of an adjacent subunit³¹,³². ΔG associated with rearranging the SD and its subsequent communication with the pore region is shown. Partial agonists effectively rearrange the IBC, but the inositol 2-O-substituent (or point mutations in the SD; black circle) disrupt the IBC-SD interface and so block communication with the SD. The latter is now less likely to contact the TMD4-5 loop, but once it makes contact the channel gates normally.

We conclude that bulky or charged groups introduced into the IBC-SD interface by either the ligand or the SD disrupt essential communication between the IBC and SD and thereby reduce efficacy.

DISCUSSION

We have synthesized and characterized a family of partial agonists of IP₃R that differ minimally from full agonists in their interactions with the binding site (IBC), but which have reduced efficacy because they block an obligatory communication between the IBC and SD. These results define two fundamentally different routes to reduced efficacy. A partial agonist may fail to make optimal contacts with the ligand-binding site^12,13,35. Alternatively, it may, as we have shown for our partial agonists of IP₃R, bind normally and then, through additional interactions, block onward transmission of essential conformational changes. These novel properties of our partial agonists allow us to show that the conformational changes initiated at the IBC pass entirely via the SD to the pore (Fig. 5c).

Our activation scheme is consistent with an earlier proposal that IP₃ minimally affects the structures of the three domains of the NT, but rearranges their relationships via flexible linking loops³⁴ (Fig. 5c). We suggest that IP₃ first stabilizes interaction of the β₂ and ARM domains of the IBC by interacting with residues in each^3,36. These interactions require the 4- and 5-phosphate groups of IP₃. The IBC then interacts with the SD (=β₁ in Fig. 5c) to give a compact structure³⁴ that allows the SD alone to signal onwards to the pore, probably via its interaction with the TMD4-5 loop (Fig. 5c)³².

IP₃R are close relatives of ryanodine receptors (RyR), sharing most sequence similarity within their N-termini and pores. The likely structural similarities between the SD of IP₃R and the N-terminal of RyR suggests these regions may have similar functions in both families of intracellular Ca²⁺ channels³³. Mutations that cause RyR to become dysfunctional in malignant hyperthermia, central core disease (RyR1) and catecholaminergic polymorphic ventricular tachycardia (RyR2) cluster in four regions that include the N-terminal and a region close to the pore³⁷. Furthermore, 3D reconstructions of RyR have shown that activation is associated with major conformational changes within a region that includes the N-terminus³⁸. For RyR1, the same region includes residues that interact with the dihydropyridine receptor, which is the major physiological regulator of RyR1. From structure-based sequence alignment³⁶, it has been suggested that the SD surface opposite to that which we suggest contacts the IBC (Supplementary Fig. 3e,f online) is most conserved between IP₃R and RyR. We speculate that this may be the surface that communicates with the conserved pore region for both IP₃R and RyR.

The SD of an IP₃R activated by a partial agonist fully engages the structures that open the pore because an open IP₃R is the same whether activated by a full or partial agonist (Fig. 2b,d and Supplementary Table 1 online), but it does so less frequently than when activated by a full agonist (Fig. 5c). The many additional proteins that interact with the SD^1,33 may exert their effects on IP₃R by targeting this essential link between IP₃ binding and channel opening.

In conclusion, we have synthesized a family of 2-O-modified analogues of IP₃ and shown they are partial agonists of IP₃R. IP₃ and these partial agonists interact similarly with the IBC, but the 2-O-substituents of the analogues block transmission of essential conformational changes from the IBC to the SD. The partial agonists thereby open the channel less effectively. This unusual form of partial agonism allows us to define two means whereby a ligand may have reduced efficacy: it may either fail to make optimal contacts with the binding site, or it may bind like a full agonist but then interfere with subsequent conformational changes. By combining mutagenesis of IP₃R with analyses of the effects of these novel partial agonists, we have shown that the major conformational changes evoked by IP₃ occur within the N-terminal and they pass to the pore entirely via the SD (Fig. 5c).

METHODS

Synthesis of ligands

Adenophostin A (AdA, 13)³⁹, inositol 1,3,4,5,6-pentakisphosphate (IP₅​, 12)⁴⁰ , IP₃ dimers¹⁹ 2, 3 and 4, D-2-deoxy-IP₃ (10)²², and 2-O-(2-aminoethyl)-IP₃ (11)²³ were synthesized as previously reported. Details of the syntheses of compounds 5–9 are given in Supplementary Methods online. IP₃ was from American Radiolabeled Chemicals. [³H]-IP₃ (18-23Ci/mmol) was from Amersham Biosciences.

Stable expression of IP₃R1 in DT40 cells

Rat IP₃R1 were stably expressed in DT40 cells in which the genes for all endogenous IP₃R had been disrupted⁴¹. The open reading frame⁴² of rat IP₃R1 was amplified by PCR using primers P6 and P7 and cloned as an EcoRI fragment into pcDNA3. The CMV promoter was replaced by the chicken β-actin hybrid promoter, excised from the vector pAneo⁴¹, to produce the construct pcDNA3-IP₃R1. QuikChange II XL site-directed mutagenesis kit (Stratagene) was used to introduce point mutations in rat IP₃R1, which had been previously cloned into the pENTR1A vector. The primers are listed in Supplementary Table 5 online. Mutated IP₃R1 was subcloned into pcDNA3.2 by recombination (Gateway, Invitrogen). The sequences of all full-length IP₃R constructs were confirmed. DT40 cells stably expressing IP₃R1 and its mutants were generated and cultured as described²⁴. Expression of mutant IP₃R in DT40 cell lines was quantified by immunoblotting (Supplementary Fig. 2a,b online).

Functional assay of IP₃R1 in DT40 cells

A low-affinity Ca²⁺-indicator (Magfluo-4) trapped within the intracellular Ca²⁺ stores was used to measure IP₃-evoked Ca²⁺ release²⁴.

Cloning and mutagenesis of N-terminal fragments of IP₃R1

Appropriate regions of rat IP₃R1 were amplified by PCR from the full-length receptor clone lacking the S1 splice region (S1^-). Fragments are numbered by reference to the full-length (S1⁺) rat IP₃R1 (Accession number NM_001007235). PCR used P1 and P2 primers for the fragment including residues 1-604 (NT), and P3 and P2 for residues 224–604 (IBC). Both P1 and P3 insert a thrombin-cleavage site. Fragments were ligated into the pTrcHisA vector at the XhoI/EcoRI sites (Invitrogen) to allow expression of N-terminally tagged His₆ proteins. Insertion of the S1 splice region into the IBC fragment used QuikChange mutagenesis kit with P4 and P5 primers. Mutagenesis of residues within the SD used the same kit. The primers are listed in Supplementary Tables 5 and 6 online. The sequences of all constructs were confirmed by DNA sequencing.

Expression of IP₃R1 fragments in bacteria

Constructs were transformed into E. coli BL21(DE3)⁴³ and 1ml of the culture was grown overnight at 37°C in Luria-Bertani medium (LBM) with 50μg/ml ampicillin. The inoculum was cultured at 22°C in 100ml of LBM until the OD₆₀₀ reached 1.0–1.5, isopropyl β-D-thiogalactoside (0.5mM) was added, and after 20h at 15°C, cells were harvested (5000xg, 5min). The pellet was resuspended in Tris/EDTA medium (TEM: 50mM Tris, 1mM EDTA, pH 8.3) supplemented with 10% PopCulture (Novagen), 1mM 2-mercaptoethanol and protease inhibitor cocktail (Sigma). The suspension was incubated with lysozyme (100μg/ml) and RNAase (10μg/ml) for 30min on ice, and the lysate was sonicated for 20s. After centrifugation (30,000xg, 60min), aliquots of supernatant were frozen in liquid nitrogen and stored at -80°C.

For immunoblotting, samples were loaded onto SDS-PAGE gels, transferred to Immobilon membranes (Millipore) and His₆-tagged proteins were identified using an anti-His₆ antibody. Proteins were cleaved from their His₆ tags by incubating bacterial lysates with biotinylated thrombin (Novagen), and thrombin was removed with streptavidin-agarose (Novagen). Cleavage was monitored by immunoblotting using anti-His₆, and Ab1⁴² or Ab1.1 antisera for the NT and IBC fragments, respectively (Supplementary Fig. 4 online and Supplementary Methods online).

Purification of IP₃R1 from rat cerebellum

IP₃R1 was purified at 4°C from cerebella of adult rats using heparin-affinity chromatography⁴⁴. Frozen cerebella were homogenized in homogenization medium (HM: 1M NaCl, 1mM EDTA, 50mM Tris, 1mM benzamidine, protease inhibitor cocktail tablet (Roche), pH 8.3) and centrifuged (100,000xg, 30min). The pellet was solubilized in HM without NaCl and supplemented with 1.2% CHAPS. After centrifugation (100,000xg, 1h), the NaCl concentration of the supernatant was increased to 250mM before loading onto heparin-agarose beads (Sigma). After 30min, the beads were washed twice in glycerol-containing medium (250mM NaCl, 50mM Tris, 10% glycerol, 1mM 2-mercaptoethanol, 1mM benzamidine, 1mM EGTA, 1% CHAPS, protease inhibitor cocktail, pH 8.0). IP₃R were then eluted with elution medium (500mM NaCl, 50mM Tris, 10% glycerol, 1mM 2-mercaptoethanol, 1mM benzamidine, 1mM EGTA, 50mM Tris, 1% CHAPS, pH 8.0), and aliquots frozen in liquid nitrogen before storage at -80°C.

³H-IP₃ binding

Equilibrium-competition binding assays were performed at 4°C for 5min in TEM containing ³H-IP₃ (18-23Ci/mmol, 0.2-1.5nM), bacterial lysate (5-10μg) or purified IP₃R (2.5μg), and competing ligands. Results were analysed by fitting to a Hill equation (GraphPad Prism) from which the IC₅₀, and thereby the K_d, were calculated. The variance of the ratios of mean values (a and b) were calculated from the variances (var) of each⁴⁵: var(a/b) = (a/b)²[(var(a)/a²)+(var(b)/b²)].

Single channel recording

Patch-clamp recording from excised nuclear patches of DT40 cells used the methods reported previously^26,27. IP₃R are relatively non-selective cation channels (P_Ba/P_K ~6)¹. K⁺ Ba was therefore used as charge-carrier to increase single channel current amplitudes²⁶ and avoid feedback regulation of IP₃R by permeating Ca²⁺. QuB (http://www.qub.buffalo.edu) was used for analysis of all channel records (Supplementary Methods online).

Molecular modelling

We developed a model of the IBC-SD relationship from the coordinate files for the IBC (1N4K) and SD (1XZZ) using protein-protein docking. Coarse-grained models of the complex were first produced using the program Hex5.1 (http://www.csd.abdn.ac.uk/hex/)⁴⁶. From these models we selected those in which the linked termini of the SD and IBC were appropriately separated, and then considered only those models in which residues from the SD known to affect binding of IP₃ to the IBC^28,33 were located at an IBC-SD interface. A representative structure was further refined using a local docking search with RosettaDock⁴⁷. Detailed methods are given in Supplementary Methods online. Our predicted structure of the IBC-SD complex (Fig. 5a,b and Supplementary Fig. 3 online) has an inertial radius of gyration (26.1Å), which is compatible with the Guinier radius of gyration (30.7Å) obtained by small angle X-ray scattering³⁴.