Interactions of Inositol 1,4,5-Trisphosphate (IP3) Receptors with Synthetic Poly(ethylene glycol)-linked Dimers of IP3 Suggest Close Spacing of the IP3-binding Sites*

The distances between the inositol 1,4,5-trisphosphate (IP3)-binding sites of tetrameric IP3 receptors were probed using dimers of IP3linked by poly(ethylene glycol) (PEG) molecules of differing lengths (1–8 nm). Each of the dimers potently stimulated45Ca2+ release from permeabilized cells expressing predominantly type 1 (SH-SY5Y cells) or type 2 (hepatocytes) IP3 receptors. The shortest dimers, with PEG linkers of an effective length of 1.5 nm or less, were the most potent, being 3–4-fold more potent than IP3. In radioligand binding experiments using cerebellar membranes, the shortest dimers bound with highest affinity, although the longest dimer (8 nm) also bound with almost 4-fold greater affinity than IP3. The affinity of monomeric IP3 with only the PEG attached was 2-fold weaker than IP3, confirming that the increased affinity of the dimers requires the presence of both IP3 motifs. The increased affinity of the long dimer probably results from the linked IP3 molecules binding to sites on different receptors, because the dimer bound with greater affinity than IP3 to cerebellar membranes, where receptors are densely packed, but with the same affinity as IP3 to purified receptors. IP3and the IP3 dimers, irrespective of their length, bound with similar affinity to a monomeric IP3-binding domain of the type 1 IP3 receptor expressed in bacteria. Short dimers therefore bind with increased affinity only when the receptor is tetrameric. We conclude that the four IP3-binding sites of an IP3 receptor may be separated by as little as 1.5 nm and are therefore likely to be placed centrally in this large (25 × 25 nm) structure, consistent with previous work indicating a close association between the central pore and the IP3-binding sites of the IP3 receptor.

Inositol 1,4,5-trisphosphate (IP 3 ) 1 receptors are intracellular Ca 2ϩ channels that are expressed in many cells and that mediate the release of Ca 2ϩ from intracellular stores evoked by receptors that stimulate IP 3 formation. The three subtypes of mammalian IP 3 receptor are closely related to each other and to the receptors expressed in birds, Xenopus, crayfish, Drosophila and Caenorhabditis elegans (2). For each of these receptors, the functional IP 3 -gated Ca 2ϩ channel is thought to be a tetramer, which may either be homomeric or, in those species that express more than one receptor subtype, heteromeric (3). Although the different subtypes of mammalian IP 3 receptor differ in their distribution (2), differ modestly in their affinity for IP 3 and their ability to recognize different inositol phosphates (4), and may be differentially modulated (5), the physiological significance of this diversity is unclear. More striking than the differences between IP 3 receptor subtypes are the similarities: the primary sequences of the subunits are closely related, each assembles to form a tetrameric IP 3 -gated Ca 2ϩ channel, they recognize similar ligands with broadly similar affinities, and most are biphasically regulated by cytosolic Ca 2ϩ (6).
Analyses of the relationships between structure and function have focused primarily on the mammalian type 1 IP 3 receptor, but the other subtypes are likely to be similar. Electron microscopy of IP 3 receptors in cerebellar Purkinje cells (7) or after purification from cerebellum (8,9) or smooth muscle (10) suggests that as viewed from the cytosol they exist as either square (7)(8)(9) or pinwheel (8,10) structures. Recent evidence suggests that Ca 2ϩ might regulate the transition between these states (8). In negatively stained images of pure receptors, the sides of the structures are about 25 nm in length (8 -10), but they appear smaller (about 15 nm) in single particle reconstructions of purified receptors in vitreous ice (11) and smaller still (about 12 nm) in images of native receptors (7).
The IP 3 -binding site (one on each subunit of the receptor) lies within the N-terminal between residues 226 and 576 (12) and is formed by two structurally distinct domains linked by a region that includes the S1 splice site (13). Within this region, conserved lysine and arginine residues are likely to be important for recognition of the phosphate groups of IP 3 . All four subunits of the tetrameric receptor contribute to the Ca 2ϩ pore, with residues lying close to the C terminus, namely the last two transmembrane regions (TMR-5 and TMR-6) and the intervening loop, forming the Ca 2ϩ channel (14). Although the IP 3binding site and the pore of the channel lie at opposite ends of the primary sequence and are separated by almost 1700 residues (the "modulatory domain"), there is evidence of direct contact between them (15), with the IP 3 -binding domain of one subunit perhaps interacting directly with the pore region of an adjacent subunit (16). Despite this considerable progress in establishing the structural determinants of IP 3 binding and of the pore, the location of the IP 3 -binding sites within the quaternary structure of the receptor is unknown.
A powerful strategy for the identification of potent and se-lective ligands for receptors with multiple binding sites is the use of multivalent ligands (17). Applications of this approach to multi-subunit proteins have generally employed synthetic bivalent constructs in which two molecules of a ligand are linked by a spacer in which the length is customized to span the distance between two binding sites within a protein multimer. Successful examples include the development of selective -opioid antagonists such as norbinaltorphine (18), potent muscarinic agonists (19), and inhibitors for ␤-tryptase (20) and the proteasome (21). Bivalent ligands may show enhanced affinity and selectivity for multimeric proteins, and they can also be used to investigate the separation of ligand-binding sites. In a recent example, potent bivalent ligands for tetrameric cyclic nucleotide-gated channels of photoreceptors and olfactory neurons were identified from a series of synthetic dimers of the natural ligand (cGMP) linked by PEG chains of various lengths (1). The separation of the binding sites for cGMP in the multimeric channels could then be estimated from the calculated mean separation of the two cGMP molecules in the most active dimers for each type of channel. Because previous work had established that even bulky additions to the axial 2-oxygen of IP 3 only slightly decrease affinity for the IP 3 receptor (22), we attached linkers to this position to produce dimers of IP 3 (23). In the present work, we use IP 3 dimers of varying lengths to examine the distances between IP 3 -binding sites within a tetrameric IP 3 receptor.
Equilibrium 3 H-IP 3 Binding-Membranes were prepared from rat cerebellum (26), and tetrameric IP 3 receptors were purified from cerebellum using heparin and concanavalin A columns (27). A membrane fraction enriched in IP 3 -binding sites was prepared from rat liver using a Percoll gradient exactly as reported previously (28). The N-terminal fragment of the rat type 1 IP 3 receptor (residues 224 -604 and containing the S1 splice site) tagged at the N terminus with hexa-His was expressed in Escherichia coli. The construct was transformed into E. coli strain BL21(DE3), and 1 ml of the culture was grown overnight in Luria-Bertani medium (29) with 50 g/ml ampicillin at 30°C. This inoculum was added to 100 ml of Luria-Bertani medium and cultured at 22°C, and when the A 600 had reached 1.0 -1.5 (about 7 h), isopropyl-1thio-␤-D-galactopyranoside (0.5 mM) was added. After a further 20 h at 15°C, cells were harvested by centrifugation (5000 ϫ g, 15 min) and washed in phosphate-buffered saline. The pellet was frozen rapidly in liquid nitrogen and stored at Ϫ80°C. Bacterial lysates were prepared by re-suspending the frozen pellet in 10 ml of Tris/EDTA medium (TEM, 50 mM Tris, 1 mM EDTA, pH 8.3) supplemented with 1 mM ␤-mercaptoethanol and a protease inhibitor mixture formulated for purification of poly-His-tagged proteins in bacteria (Sigma). The suspension was incubated with lysozyme (100 g/ml, Sigma) for 30 min on ice followed by five rapid freeze-thaw cycles using liquid nitrogen. The lysate was then sonicated for 20 s (maximal setting on an MSE Soniprep 150), and after centrifugation (30,000 ϫ g, 60 min), aliquots of the supernatant (typically 4 mg protein/ml) were frozen in liquid nitrogen and stored at Ϫ80°C. The major band detected after Western blotting of the final supernatant fraction with a hexa-His antibody had the expected size of 43.5 kDa (Fig. 4A); a very minor band (double arrowheads in Fig. 4A) probably represents dimeric fusion protein. Although we engineered an enterokinase cleavage site into the fusion protein, we were unable to remove the hexa-His tag using enterokinase without causing cleavage of the IP 3 -binding domain. All experiments with the bacterially expressed IP 3 -binding domain therefore used the hexa-Histagged protein.
All equilibrium binding incubations were performed at 4°C in TEM (final volume 200 l) containing 3 H-IP 3 (1-2 nM), membranes (typically 50 g), bacterial lysate (100 g) or purified IP 3 receptor (8 g), and appropriate concentrations of competing ligands. After 5 min, reactions were terminated either by centrifugation alone (membranes; 20,000 ϫ g, 5 min) or, for soluble proteins, by the addition of 200 l of cold TEM containing 30% PEG-8000 and 200 g of ␥-globulin followed by centrifugation. Pellets were solubilized in 1 ml of EcoScint A scintillation mixture, and their activity was determined by liquid scintillation counting. Total 3 H-IP 3 binding was usually more than 2500 dpm and nonspecific binding was Ͻ10% of total binding. Equilibrium competition binding curves were fitted to logistic equations using nonlinear curve fitting (Kaleidegraph, Synergy Software, Reading, PA) from which equilibrium dissociation constants were determined (4).

Short IP 3 Dimers Potently Stimulate Ca 2ϩ
Release from Permeabilized Hepatocytes-Despite attachment of PEG to the 2-position of IP 3 (Fig. 1), the IP 3 dimers potently stimulated the release of Ca 2ϩ from the intracellular stores of permeabilized hepatocytes, with maximally effective concentrations of IP 3 and each of the dimers releasing similar fractions of the intracellular Ca 2ϩ stores (Table I, Fig. 2A). To estimate the separation of the two IP 3 structures in each dimer, we followed the approach described by Kramer and Karpen (1) in which the effective separation is taken as the average (r.m.s.) length of the flexible PEG linker. The r.m.s. lengths can be calculated from previous determinations of the lengths of specific PEGs (25), assuming that the r.m.s. length is proportional to the square root of the number of ethylene glycol monomers. This method predicts r.m.s. lengths of ϳ1 and 1.5 nm, respectively, for the linkers in the two smallest dimers [1 and 2], increasing to 8 nm for the largest dimer [6]. A comparison of the potencies of the dimers relative to monomeric IP 3 indicates that dimers linked by the shortest linkers (r.m.s. length Յ 1.5 nm; 1 and 2) 2 were about (3-4)-fold more potent than IP 3 (Fig. 2C).
These results are consistent with short dimers of IP 3 achieving their increased potency because the separation between the pair of IP 3 molecules is sufficient to allow each IP 3 to simultaneously interact with a binding site on a tetrameric receptor. An obvious prediction would then be that short dimers of IP 3 would not bind with increased affinity to monomeric subunits of the receptor. Unfortunately, the only methods that have succeeded thus far in dissociating native IP 3 receptors into their subunits have also abolished their ability to bind IP 3 (30). The only effective way of examining IP 3 binding to monomeric IP 3 -binding sites is, therefore, by expression of recombinant proteins lacking the membrane-spanning regions that mediate oligomerization (30,31). Type 2 IP 3 receptors are the major subtype (ϳ80%) (32) expressed in hepatocytes, and although we have successfully expressed full-length recombinant type 2 IP 3 receptors (4), we have not succeeded in expressing the type 2 monomeric IP 3 -binding domain in bacteria. For subsequent analyses, we therefore used type 1 IP 3 receptors, where it was possible to express monomeric IP 3 -binding domains (33,34).
Using cerebellar membranes, the richest native source of type 1 IP 3 receptors, we determined the affinities of IP 3 and the IP 3 dimers for type 1 IP 3 receptors using 3 H-IP 3 in equilibrium competition binding assays (Fig. 3A). The results (Table II) reveal an obvious biphasic effect of the length of the PEG linker on the affinity of a dimer for the type 1 IP 3 receptor (Fig. 3B). The shortest dimer bound with the greatest affinity, and dimers of intermediate length were similar to IP 3 , but the longest dimer [6] had significantly greater affinity than IP 3 . The high affinity of short IP 3 dimers is consistent with the functional analyses in both hepatocytes (Table I) and SH-SY5Y cells, but the high affinity of the longest dimer was unexpected (see below).
Monomeric versions of a short dimer [2] in which IP 3 was linked to either (CH 2 ) 2 NH 3 ϩ [7] or to PEG with one end capped by a methyl group rather than IP 3 [8] (Fig. 1) bound with 4.5or 1.7-fold lower affinity than IP 3 to cerebellar membranes (Table II). These results established that the high affinity of the dimers for IP 3 receptors requires the presence of both IP 3 motifs; it is not a direct consequence of an interaction with the linking group. Indeed, the reduced affinity of these parent compounds [7,8] relative to IP 3 suggests that the increased affinity of IP 3 dimers may be rather greater than implied by our comparisons with IP 3 itself. We speculated that the very high density of IP 3 receptors in cerebellar membranes (7) might allow the longest dimer to span IP 3 -binding sites between neighboring IP 3 receptors. Subsequent experiments were designed to determine whether the high affinity of the shortest dimers reflects bivalent binding within a tetrameric receptor and to establish whether the longest dimers might simultaneously bind to sites on adjacent receptors.
Intra-and Inter-receptor Binding of IP 3 Dimers-We reasoned that if the high affinity of the long dimer [6] for cerebellar IP 3 receptors (Fig. 3B) resulted from its binding simultaneously to sites on adjacent receptors, then disrupting the close packing of IP 3 receptors in native membranes (7) would reduce its affinity. Whereas dimer 6 bound with 3.4 Ϯ 0.6-fold greater affinity than IP 3 to receptors in cerebellar membranes (Fig.  3B), its affinity for purified cerebellar IP 3 receptors (K d ϭ 2.3 Ϯ 0.5 nM) was not significantly different from that of IP 3 (K d ϭ 3.5 Ϯ 0.6 nM). We conclude that the high affinity of 6 for IP 3 receptors in cerebellar membranes probably results from an inter-receptor interaction (Fig. 4C, b). This interpretation is consistent with both the reported spacing of IP 3 receptors in native membranes (ϳ2-4 nm) (7) and with our observation that the relative potency of 6 is greater in the cerebellum, which has a very high density of IP 3 receptors, than in hepatocytes (Fig. 2B), where IP 3 receptors are present at lower density. Our results with 6 are therefore consistent with IP 3 receptors in the cerebellar membranes being close together, but they do not allow a precise estimate of their spacing. Although the estimated r.m.s. length of the linker in 6 is 8 nm, this dimer was synthesized from a polydisperse PEG containing a range of polymer lengths distributed about a mean molecular weight. Thus, the longer dimers present in 6, particularly in more extended conformations, will be capable of spanning larger distances.
We expressed a near minimal IP 3 -binding domain of the type 1 IP 3 receptor (residues 224 -604 with the S1 splice region residues 224 -604) in bacteria (Fig. 4A) and used it to establish whether the high affinity of IP 3 dimers for native receptors required a tetrameric receptor structure. In keeping with pre-vious results suggesting that residues toward the N terminus may inhibit IP 3 binding (12, 34), residues 224 -604 bound IP 3 with about 10-fold greater affinity than did the full-length receptor (Tables II and III). Previous work established that full-length IP 3 receptor or N-terminal fragments (residues 1-604) whether expressed alone or with a hexa-His tag had indistinguishable affinities for IP 3 (33), confirming that the tag does not affect IP 3 binding. Our results with residues 224 -604 (Table III, Fig. 4B), in contrast to those obtained with tetrameric receptors (Figs. 2 and 3), demonstrate that dimers of IP 3 , whether linked by long [6] or short linkers [1,2], bind with slightly lower affinity than IP 3 to monomeric IP 3 -binding domains. These results establish that short IP 3 dimers bind with high affinity only when the receptor exists in its tetrameric state. Their increased affinity must therefore result from simultaneous binding of the two linked IP 3 molecules to binding sites within a tetrameric receptor.
Conclusions-In both functional and radioligand binding assays using type 1 and type 2 IP 3 receptors, dimers of IP 3 linked by short spacers (Յ1.5 nm) 2 bind to tetrameric IP 3 receptors with significantly greater affinity than monomeric IP 3 . Molecular models show that even in the most extended conformations of the shortest dimer [1] the separation of the two IP 3 molecules does not exceed 2 nm. We conclude that the IP 3binding sites of the receptor are likely to lie within 2 nm of each other and must therefore be near the center of the large (12-25 nm) (7-10) tetrameric IP 3 receptor structure. Such a location FIG. 2. Interactions between IP 3 dimers and type 2 IP 3 receptors. A and B, the effects of the indicated concentrations of IP 3 (E) or 2 (q) on the Ca 2ϩ contents of the intracellular stores of permeabilized hepatocytes (A) and on specific 3 H-IP 3 binding to hepatic membranes (B) are shown. C, a summary of the effects of IP 3 dimers separated by linkers of different effective (r.m.s.) PEG lengths (footnote 2) on Ca 2ϩ release (q) and 3 H-IP 3 binding (E) in hepatocytes. In both cases the sensitivity of the response to the dimer (EC 50 or K d ) is shown relative to that for IP 3 , such that relative potencies greater than unity denote compounds that are effective at lower concentrations than IP 3 . Results (A-C) are the means Ϯ S.E. from at least three independent experiments. would place the IP 3 -binding sites close to the central pore of the channel, consistent with evidence suggesting a close association between the N terminus and channel region of the receptor (15,16). In a recent study of purified type 1 IP 3 receptors, heparin (a competitive antagonist of IP 3 binding) conjugated to gold via albumin was used to locate IP 3 -binding sites by electron microscopy (8), and the results suggested that the sites might lie toward the periphery of the receptor (i.e. Ն10 nm apart). However, because the albumin (66 kDa) and gold (Ͼ5 nm) attached to the heparin are themselves large, they may exaggerate the distance between IP 3 -binding sites. Alternatively, the binding sites may be paired such that the spacing between sites within a pair is less (Յ2 nm) than the spacing between pairs (Ͼ10 nm). 2 Because PEG-linked dimers of cGMP are membane-permeant (1), we assessed whether similarly modified IP 3 might, despite being more highly charged than cGMP, also cross the plasma membrane. However, in fura-2-loaded HEK cells, none of the PEG-modified IP 3 analogues tested (2, 5, and 8; 10 min, 30 M) caused either detectable Ca 2ϩ release or any diminution of the subsequent Ca 2ϩ release evoked by carbachol. 3 Our results established that dimeric molecules afford opportunities to develop high affinity ligands of IP 3 receptors (1 and 2 are the first inositol phosphates to bind to IP 3 receptors with significantly greater affinity than IP 3 ) and that they are a means of addressing the structural organization of the IP 3 3 J. E. Church and C. W. Taylor, unpublished observations.

FIG. 3. Interactions between IP 3 dimers and type 1 IP 3 receptors.
A, effects of the indicated concentrations of IP 3 (E) or 2 (q) on specific 3 H-IP 3 binding to cerebellar membranes are shown. B, affinity of the different IP 3 dimers for the IP 3 receptors of cerebellar membranes expressed relative to the affinity for IP 3 Results are the means Ϯ S.E. from at least three independent experiments.   receptor. Having established the utility of this approach, future studies with more rigid linkers should provide opportunities to develop very high affinity agonists and antagonists, also allowing more precise estimates of the spacing of IP 3 -binding sites within a tetramer and of IP 3 receptors in native membranes. Conversely, as the structure of the IP 3 receptor is resolved at higher resolution, it will become possible to tailor the spacing of bivalent ligands more precisely to match the spacing of the IP 3 -binding sites and perhaps thereby to produce very high affinity agonists and antagonists of IP 3 receptors.