Mapping of the ATP-binding Sites on Inositol 1,4,5-Trisphosphate Receptor Type 1 and Type 3 Homotetramers by Controlled Proteolysis and Photoaffinity Labeling*

Submillimolar ATP concentrations strongly enhance the inositol 1,4,5-trisphosphate (IP 3 )-induced Ca 2 1 release, by binding specifically to ATP-binding sites on the IP 3 receptor (IP 3 R). To locate those ATP-binding sites on IP 3 R1 and IP 3 R3, both proteins were expressed in Sf9 insect cells and covalently labeled with 8-azido-[ a 32 P]ATP. IP 3 R1 and IP 3 R3 were then purified and sub- jected to a controlled proteolysis, and the labeled proteolytic fragments were identified by site-specific antibodies. Two fragments of IP 3 R1 were labeled, each containing one of the previously proposed ATP-binding sites with amino acid sequence G X G XX G (amino acids 1773–1780 and 2016–2021, respectively). In IP 3 R3, only one fragment was labeled. This fragment contained the G X G XX G sequence (amino acids 1920–1925), which is conserved in the three IP 3 R isoforms. The presence of multiple interaction sites for ATP was also evident from the IP 3 -induced Ca 2 1 release in permeabilized A7r5 cells, which depended on ATP over a very broad concentration range from micromolar to millimolar. Inositol

Inositol 1,4,5-trisphosphate (IP 3 ) 1 is an intracellular second messenger that mediates the release of Ca 2ϩ from internal stores by binding to the IP 3 receptor (IP 3 R), an intracellular Ca 2ϩ -release channel (1). The IP 3 R is composed of three functionally different domains: an N-terminal IP 3 -binding region, a large transducing domain, and a C-terminal channel region (2). The transducing domain contains interaction sites for several modulators of IP 3 -induced Ca 2ϩ release such as Ca 2ϩ , calmodulin, kinases, phosphatases, ATP, and FKBP12 (reviewed in Refs. 1 and 3). IP 3 Rs are encoded by three different genes, resulting in the existence of IP 3 R1, IP 3 R2, and IP 3 R3, and the various IP 3 R isoforms are distributed in a tissue-specific manner (4 -7). Nearly all cell types coexpress at least two IP 3 R isoforms (4,5,8), which are mostly co-organized in heterotet-rameric structures (9 -12). The different IP 3 R isoforms show functional differences in their regulation by IP 3 and by several modulators of IP 3 -induced Ca 2ϩ release (13)(14)(15).
ATP regulates the IP 3 R in a concentration-dependent manner: Submillimolar concentrations enhance IP 3 -induced Ca 2ϩ release (16 -20), whereas millimolar levels of ATP inhibit IP 3induced Ca 2ϩ release by competing with IP 3 for the IP 3 -binding site (18 -23). The stimulatory effect of ATP is likely to occur via binding to one or more sites on the IP 3 R, because purified IP 3 Rs bind [␣-32 P]ATP in a specific manner (17,20,24). The number and the localization of these sites have, however, not yet been determined. Based on the glycine-rich amino acid sequence GXGXXG (25), two ATP-binding sites were postulated on the neuronal form of IP 3 R1 (aa 1773-1780 and 2016 -2021). The former is only present in IP 3 R1, whereas the latter is common to the three IP 3 R isoforms (2, 26 -28). In a previous study, we have expressed the cDNA domains of IP 3 R1 containing these glycine-rich motifs as glutathione S-transferase (GST) fusion proteins in bacteria and showed that they both were able to bind ATP (29). The aim of the present study was to determine the location and number of ATP-binding sites on the intact IP 3 R. We did this by photoaffinity labeling with 8-azido-[␣-32 P]ATP of microsomes of Sf9 insect cells expressing recombinant IP 3 R1 or IP 3 R3 homotetramers, followed by purification and controlled proteolysis with chymotrypsin. We found that controlled proteolysis of IP 3 R1 and IP 3 R3 yielded roughly the same major fragments, indicating that both isoforms have a similar general structure. Moreover, the results indicated that the IP 3 R1 contained two ATP-binding sites, because two separate fragments obtained by proteolysis were labeled. These two fragments each contained one of the presumed ATP-binding sites (aa 1773-1780 and 2016 -2021, respectively), as identified using site-specific antibodies. In IP 3 R3, only one proteolytic fragment was labeled. This fragment contained the proposed ATP-binding site that is conserved in all IP 3 R isoforms (aa 1920 -1925). The unequal number of ATP-binding sites in IP 3 R1 and IP 3 R3 could have implications for the modulation of these isoforms by ATP. Recently, we found that IP 3 R1 and IP 3 R3 have a different ATP affinity (EC 50 values of 1.6 and 177 M, respectively) (24). In this study, the ATP dependence of IP 3 -induced Ca 2ϩ release measured in permeabilized A7r5 cells, which express both IP 3 R1 and IP 3 R3, extended over a very broad range from micromolar to millimolar concentrations. This finding confirms the presence of multiple nucleotide-binding sites with different affinities for ATP in IP 3 R1 and IP 3 R3.
Expression of IP 3 R1 and IP 3 R3 in Insect Sf9 cells-The full-length mouse IP 3 R1 and the full-length rat IP 3 R3 were expressed in insect Sf9 cells as described by Sipma et al. (30) and by Maes et al. (24), respectively.
Purification of Recombinant IP 3 R1 and IP 3 R3 from Sf9 Microsomes-The purification of IP 3 Rs was based on the method described by Parys et al. (32). Microsomes of Sf9 cells expressing IP 3 R1 and IP 3 R3 in a concentration of 10 mg of protein/ml were centrifuged, and the pellet was solubilized (at 5 mg/ml) for 1.5 h at 4°C in buffer A (50 mM Tris-HCl, pH 7.4, 1 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride, 0.83 mM benzamidine, and 10 mM 2-mercaptoethanol) with addition of 200 mM NaCl, 77 nM aprotinin, 1.1 M leupeptin, 0.7 M pepstatin A, 2.5% CHAPS, and 1% L-␣-phosphatidylcholine. After centrifugation, the supernatant was diluted with an equal volume of buffer A with addition of 400 mM NaCl. The diluted supernatant was incubated for 30 min with heparin-agarose beads (112.5 l/mg of protein). The eluate obtained in buffer A with 600 mM NaCl, 0.75% CHAPS, and 0.3% L-␣-phosphatidylcholine, was incubated for 2 h with wheat germ agglutinin-Sepharose (75 l/mg of protein). After wash steps in high (600 mM) and low (100 mM) salt conditions, the specifically bound proteins were eluted in low salt conditions with 300 mM N-acetyl-D-glucosamine. All centrifugation steps were for 17 min at 35,700 ϫ g at 4°C.
Controlled Proteolysis-Purified IP 3 R was partially digested with chymotrypsin (0.05 g/ml) for 2, 5, 10, or 30 min on ice as described previously (32). The digestion was stopped by the addition of 100 g/ml N-tosyl-L-phenylalanine chloromethyl ketone and by boiling the samples for 5 min in sample buffer for SDS-PAGE.
Antibodies and Western Blotting-The polyclonal antibody against the C terminus of mouse IP 3 R1 (Rbt03), the mouse monoclonal antibody against an N-terminal epitope of human IP 3 R3 (MMAtype3) (Transduction Laboratories, Lexington, KY) and the polyclonal antibody against the Ca 2ϩ -binding domain cytI3b (amino acids 378 -450) in the IP 3binding domain of mouse IP 3 R1 (33) were characterized earlier (4,30,34). A novel antibody was raised against the luminal Ca 2ϩ -binding fragment LoopI17a of mouse IP 3 R1 (aa 2463-2528) (35). Two rabbits were injected subcutaneously and intramuscularly with Freund's complete adjuvant containing 0.5 mg of LoopI17a fused to GST. Animals were boosted 2 weeks later with the same antigen in Freund's incomplete adjuvant and regularly thereafter. After three boost injections, both rabbits produced high titers of antibody. Both antibodies (named anti-loopI17a-1 and anti-loopI17a-2) reacted with mouse, rat, human, hamster, and rabbit IP 3 R1. They also recognized rat IP 3 R3, although with lower sensitivity. 2 A polyclonal antibody directed against residues 1829 -1848 of human IP 3 R1 was purchased from Alexis Corp. (Lä ufelfingen, Switzerland). A polyclonal antibody against the C terminus of human IP 3 R3 was from Santa Cruz Biotechnology (Santa Cruz, CA). The various microsomal preparations were analyzed on 3-12% Laemmli-type gels and transferred to Immobilon-P (Millipore Corp., Bedford, MA). Immunodetection of the proteins on the transfers was exactly as described previously (30,36). 45 Ca 2ϩ Fluxes-IP 3 -induced Ca 2ϩ release from permeabilized A7r5 monolayers was described elsewhere (16). The added ATP concentrations are indicated in the legend to Fig. 4. The medium for the challenge with IP 3 contained 120 mM KCl, 30 mM imidazole-HCl, pH 6.8, and 1 mM EGTA.

RESULTS AND DISCUSSION
Purification of IP 3 Rs from Sf9 Insect Cells-To allow an accurate analysis involving controlled proteolysis and immunostaining, purification of the IP 3 R is needed. IP 3 R1 has been purified from cerebellum (37,38), smooth muscle (39,40) and oocytes (32). Until now, no IP 3 R3 has been purified due to the lack of a known cell type that abundantly expresses this isoform. We therefore expressed IP 3 R1 or IP 3 R3 in Sf9 insect cells, resulting in a 2.5 times higher expression of IP 3 R1 and a Ͼ50 times higher expression of IP 3 R3 as compared with rabbit cerebellar and 16HBE14o-microsomes, respectively (24,30). In the purification procedures described by Chadwick et al. (39) and Parys et al. (32), microsomes were first solubilized by a detergent followed by chromatography on heparin-and lectinbased matrices. This method was based on the ability of heparin to bind to the IP 3 R with high affinity (41)(42)(43) and on the presence of N-glycosylation sites on 2 asparagine residues present in IP 3 R1 (44). It has been suggested that IP 3 R3 is also a glycoprotein (45), although only one N-glycosylation site is predicted based on the primary sequence (46).
In this study, we have used an identical approach to purify recombinant IP 3 R1 and IP 3 R3 overexpressed in Sf9 insect cells. It had to be verified whether the glycosylation patterns of these proteins were the same as in mammalian cells. Briefly, microsomes from Sf9 cells overexpressing either IP 3 R1 or IP 3 R3 were solubilized with 2.5% CHAPS. Subsequently, the solubilized microsomes were incubated with heparin-agarose. After elution of the bound fraction, the latter was incubated with wheat germ agglutinin-Sepharose. Both receptors could be purified with high efficiency and were recognized by isoformspecific antibodies (Fig. 1, A and B, first lane of each blot), confirming that they are both glycoproteins and that the posttranslational glycosylation of the IP 3 Rs in insect cells is similar to that in mammalian cells. The purified IP 3 R1 migrated on SDS-PAGE with a molecular mass of 273 kDa, which deviated from the molecular mass of 313 kDa predicted from the primary structure (Fig. 1A, first lane of each blot). The purified IP 3 R3 also migrated with a lower apparent molecular mass (248 kDa) than predicted (304 kDa) (Fig. 1B, first lane of each blot). Because a similar behavior is also found for endogenous IP 3 Rs from, e.g. cerebellar or 16HBE14o-cells (20,47), this discrepancy is likely due to aberrant mobility of higher molecular mass proteins on SDS-PAGE.
Controlled Proteolysis and Identification of Proteolytic Fragments-The purified IP 3 Rs were subjected to a controlled proteolysis with chymotrypsin (0.05 g/ml, up to 30 min on ice), and the digestion fragments were detected by a panel of different site-specific antibodies (Fig. 1, A and B, and Table I). No degradation of the intact IP 3 R was observed during incubation without chymotrypsin (Fig. 1, first lane of each blot).
For IP 3 R1, four site-specific antibodies were used, of which the epitopes were spread over the whole sequence. The anti-cytI3b-2 antibody (30) is directed against a Ca 2ϩ -binding site in the IP 3 -binding domain (33). The anti-(1829 -1848) antibody (Alexis Corp.) recognized an amino acid stretch (residues 1829 -1848) located in the regulatory domain between the two putative ATP-binding sites (residues 1773-1780 and 2016 -2021) (2, 26 -28). A third antibody, anti-loopI17a-2, was raised against the luminal Ca 2ϩ -binding fragment (35). Finally, Rbt03 (30,34) recognized the C terminus of IP 3 R1. The proteolytic pattern, resulting from up to 30 min of incubation with chymotrypsin, and as detected by the four antibodies against IP 3 R1, is shown in Fig. 1A. We determined the length of the fragments using Rainbow molecular mass markers. Based on these data, we were able to localize the chymotrypsin-sensitive sites on IP 3 R1 (Fig. 2A). The sum of the molecular mass of the five major proteolytic fragments (40,65,80,40, and 90 kDa) was close to the molecular mass of the intact IP 3 R1 (313 kDa). This result was in complete agreement with the study of Yoshikawa et al. (48), where trypsin was used to digest cerebellum-purified IP 3 R1 and where five similar major proteolysis-insensitive fragments were found. Although we were able to recognize most of the intermediate digestion products with site-specific antibodies, two proteolytic fragments, which were predicted based on Fig. 2A, could not be detected when the digestion was performed for 30 min. Particularly, a 145-kDa fragment, precursor of the 65-and 80-kDa fragments should be recognized by the anti-cytI3b antibody and a 185-kDa fragment, precursor of the successive fragments of 65, 80, and 40 kDa should be recognized by the anti-(1829 -1848) antibody. Because it is possible that these intermediate fragments have a short life time, we decreased the time of proteolysis to 2, 5, and 10 min, respectively (Fig. 1A, insets of blots 1 and 2). Upon staining with the anti-cytI3b-2 antibody (blot 1 and inset), we detected a proteolytic band corresponding to a molecular mass of 145 kDa, which is most intense at 2 and 5 min of incubation with chymotrypsin. This fragment is rapidly degraded into smaller fragments, because it is poorly or not visible in the proteolytic patterns representing 10 and 30 min of incubation of IP 3 R1 with chymotrypsin. Upon staining with the anti-(1829 -1848) antibody, no clear fragment with a mass of 185 kDa was visible, even at shorter time points (blot 2 and inset). Because the corresponding predicted fragment was also not detected in the study of Yoshikawa et al. (48), it is conceivable that the latter intermediate fragment is rapidly degraded into smaller subfragments during proteolysis and has therefore a steady-state level below the detection limit for the antibodies. All identified fragments, with indication of their molecular mass and the recognizing antibodies, are represented in Table II.
The same type of experiment was performed for IP 3 R3. Only two site-specific antibodies are available for this isoform: the MMAtype3 antibody (Transduction Laboratories) (4) directed against the N terminus, and the anti-CIII antibody (Santa Cruz Biotechnologies) against the C terminus. However, the anti-loopI17a-2 antibody could also recognize IP 3 R3, 2 although with lower sensitivity. The proteolytic pattern as stained by the three antibodies against IP 3 R3 is shown in Fig. 1B. The time dependence of the occurrence of the proteolytic fragments was also investigated for IP 3 R3, but incubation with chymotrypsin for shorter times (2-10 min) revealed the same pattern of proteolytic fragments (data not shown). All identified fragments, with their molecular mass and the recognizing antibodies, are represented in Table III. In addition we have verified the N-terminal boundaries of some of the major proteolytic fragments by N-terminal amino acid microsequencing (data not shown). A schematic presentation of IP 3 R3 with the major proteolytic fragments (105, 70, 35, and 95 kDa) is shown in Fig.  2B. The sum of the molecular mass of the fragments was close to the molecular mass of the intact receptor (304 kDa) as calculated from the cloned rat IP 3 R3. The general structure of IP 3 R3 resembled that of IP 3 R1: Both receptor isoforms were sensitive to proteolysis at similar sites. Only the chymotrypsinsensitive site that is present in the IP 3 -binding domain of IP 3 R1, could not be detected in IP 3 R3. This could however be due to the lack of an antibody that recognized the relevant part of the IP 3 -binding domain. Alternatively, it is also possible that IP 3 R3 lacks the chymotrypsin-sensitive site in the IP 3 -binding domain. It is conceivable that the proteolysis-sensitive sites represent regions that are exposed on the surface of the protein and thereby accessible to the proteolytic enzymes as well as to different modulators of IP 3 -induced Ca 2ϩ release. Because functional IP 3 Rs are mostly organized in heterotetramers (9 -12), it can be expected that corresponding regions of the different IP 3 R isoforms are exposed at the surface of the receptor protein so that they can be properly regulated.
To prove this, we incubated microsomes from Sf9 cells expressing recombinant IP 3 R1 with the photoaffinity label 8-azido-[␣-32 P]ATP. Covalent labeling of the ATPbinding sites by UV irradiation was followed by purification and controlled proteolysis of IP 3 R1 and identification of the labeled proteolytic fragments by site-specific antibodies. The two smallest labeled proteolytic fragments of IP 3 R1 (90 and 40 kDa, Fig. 3A) were recognized by the Rbt03 antibody and the anti-(1829 -1848) antibody, respectively (Table II), indicating that they represented the proteolytic fragments containing the previously proposed ATP-binding sites ( Fig. 2A). IP 3 R3 contained only one of these proposed ATP-binding sites, which is conserved in all IP 3 R isoforms and which is also located near a chymotrypsin-sensitive site (Fig. 2B). To confirm this, we performed the same photoaffinity labeling experiment for the IP 3 R3 isoform. The smallest labeled band of 95 kDa

FIG. 2. Schematic representation of the proteolytic fragments of IP 3 R1 (A) and IP 3 R3 (B).
The top line indicates the molecular mass and the number of amino acids of the IP 3 R. The horizontal bar represents a scheme of the IP 3 R with the chymotrypsin-sensitive sites (indicated by the scissors), the size of the proteolytic fragments (in kDa), the proposed ATP-binding sites, and the epitopes of the site-specific antibodies.

TABLE II
Identification of proteolytic fragments of IP 3 R1 by site-specific antibodies and 8-azido-[␣-32 P]ATP labeling IP 3 R1 was purified from Sf9 insect cells and subjected to a controlled proteolysis. The proteolytic fragments were identified with site-specific antibodies. The proteolytic fragments (represented in kDa), obtained after 2-30 min of incubation with chymotrypsin and recognized by a particular antibody are indicated by an "x." "x o " indicates that the used antibody detected the particular fragment with a very weak intensity, or that the fragment was very weakly labeled with 8-azido-[␣-32 P]ATP. The same method was used for IP 3 R1, which was labeled with 8-azido-[␣-32 P]ATP prior to purification and controlled proteolysis. The presence of two different labeled sites (on fragments of 90 and 40 kDa, respectively) follows from their specific identification with different antibodies (Fig. 2A). The details of the photoaffinity labeling, purification, and controlled proteolysis are described under "Experimental Procedures." kDa Anti-cytI3B-2 Anti-(1829-1848) Anti-loopI17a-2 Rbt03 32 P label 273 x x 90 x x x 65 x 40 x x TABLE III Identification of proteolytic fragments of IP 3 R3 by site-specific antibodies and 8-azido-[␣-32 P]ATP labeling IP 3 R3 was purified from Sf9 insect cells and subjected to a controlled proteolysis. The proteolytic fragments were identified with site-specific antibodies. Proteolytic fragments (represented in kDa) recognized by a particular antibody are indicated by an "x". "x o " indicates that the used antibody detected the particular fragment with a very weak intensity. The same method was used for IP 3 R3, which was labeled with 8-azido-[␣-32 P]ATP prior to purification and controlled proteolysis. As illustrated in Fig. 2B, the smallest labeled fragment was the 95-kDa C-terminal fragment. The details of the photoaffinity labeling, purification, and controlled proteolysis are described under "Experimental Procedures." kDa MMAtype3 Anti-loopI17a-2 Anti-CIII 32 P label 248 x x x x 210 x 200 x x x 175 x 130 x o x x 105 x 95 x x x ( Fig. 3B) was recognized by the anti-CIII antibody (Table III), indicating that this band represented the proteolytic fragment containing the putative ATP-binding site of IP 3 R3 (Fig. 2B). In summary, covalent labeling with 8-azido-[␣-32 P]ATP occurred at two different proteolytic fragments of IP 3 R1 and only at one proteolytic fragment of IP 3 R3. The labeled fragments contained the two previously proposed ATP-binding sites, one of which is conserved in all IP 3 R isoforms. The unequal number of ATP-binding sites found in IP 3 R1 and IP 3 R3 may explain the differential modulation of these isoforms by ATP. IP 3 R1 showed a higher affinity for ATP than IP 3 R3 (13,14,24), suggesting that the upstream ATP-binding site, which is only present in IP 3 R1, is a high-affinity binding site. Moreover, IP 3 R3 displayed a broader nucleotide specificity than IP 3 R1 (24), because it bound equally well ATP and GTP. The latter property can be assigned to the ATP-binding site present in IP 3 R3 and conserved in all IP 3 R isoforms. The ATPbinding site that is only present in IP 3 R1 was more specific for adenine nucleotides like ATP and ADP (24).
ATP Dependence of IP 3 -induced Ca 2ϩ Release-In permeabilized A7r5 cells, which express IP 3 R1 and IP 3 R3 in a 3 to 1 ratio (5), ATP dependence of IP 3 -induced Ca 2ϩ release was found over a very broad concentration range (Fig. 4). ATP stimulated IP 3 -induced Ca 2ϩ release from the low micromolar range up to 1 mM. At still higher ATP concentrations, the release was inhibited probably due to competition of ATP for the IP 3 -binding site (18 -23). The broad concentration dependence in A7r5 cells is in very good agreement with the different ATP affinities described previously for recombinant IP 3 R1 and IP 3 R3 (EC 50 values of 1.6 M and 177 M, respectively) (24). This difference in ATP affinities between IP 3 R1 and IP 3 R3 was also observed by other groups: recent findings of Hagar and Ehrlich (49) demonstrated that IP 3 R3 incorporated in lipid bilayers was activated by ATP with an EC 50 of about 3 mM, whereas a much lower EC 50 (40 M) was observed for IP 3 R1 (19). Moreover, IP 3induced Ca 2ϩ release in genetically engineered DT40 B cells that express a single IP 3 R subtype was also found to respond differently to ATP. In IP 3 R1-expressing cells, the rate of Ca 2ϩ release was enhanced by ATP with an EC 50 of 0.39 mM, whereas IP 3 R3expressing cells were much less sensitive to ATP (14).
Because it was not possible to resolve the ATP concentration dependence in A7r5 cells by curve-fitting procedures, the pres-ent data do not allow a determination of whether the presence of two IP 3 R isoforms in A7r5 is reflected in two separate stimulatory ATP-binding sites. However, for preparations from rat cerebellum containing nearly exclusively IP 3 R1, the maximum stimulation by ATP was found at 50 M. At 1 mM ATP, IP 3induced Ca 2ϩ release in cerebellar preparations was close to control values (50), whereas 1 mM ATP was the maximum stimulatory concentration in A7r5 cells. The much higher maximum for ATP stimulation found for A7r5 cells is therefore very probably a reflection of the presence of IP 3 R3. It was also not possible to decide whether these properties of IP 3 R3 are inferred in A7r5 cells by homo-or heterotetramers. Coimmunoprecipitation experiments indicated that a significant fraction of IP 3 R1 and IP 3 R3 expressed in A7r5 cells is present as heterotetramers. 2 Our data clearly showed that the presence of different ATP-binding sites on IP 3 R1 and IP 3 R3 resulted in a nucleotide sensitivity of IP 3 -induced Ca 2ϩ release that extended over a broad concentration range. The ATP concentration that yielded maximum stimulation seems very variable and to be dependent on the IP 3 R isoform composition in the particular cell type.  45 Ca 2ϩ , were incubated in efflux medium for 10 min, at which time 1 M IP 3 plus the indicated ATP concentration was added for 2 min. The stimulation of the Ca 2ϩ release by ATP is expressed as the percentage increase in the Ca 2ϩ release above the control value, obtained in the absence of ATP. Data are the means of four independent experiments. The error bars smaller than the data symbol are not indicated.