|Home | About | Journals | Submit | Contact Us | Français|
Thapsigargin (Tg), a specific inhibitor of sarco/endoplasmic Ca2+-ATPases (SERCA), binds with high affinity to the E2 conformation of these ATPases. SERCA inhibition leads to elevated calcium levels in the cytoplasm, which in turn induces apoptosis. We present x-ray crystallographic and intrinsic fluorescence data to show how Tg and chemical analogs of the compound with modified or removed side chains bind to isolated SERCA 1a membranes. This occurs by uptake via the membrane lipid followed by insertion into a resident intramembranous binding site with few adaptative changes. Our binding data indicate that a balanced hydrophobicity and accurate positioning of the side chains, provided by the central guaianolide ring structure, defines a pharmacophore of Tg that governs both high affinity and access to the protein-binding site. Tg analogs substituted with long linkers at O-8 extend from the binding site between transmembrane segments to the putative N-terminal Ca2+ entry pathway. The long chain analogs provide a rational basis for the localization of the linker, the presence of which is necessary for enabling prostate-specific antigen to cleave peptide-conjugated prodrugs targeting SERCA of cancer cells (Denmeade, S. R., Jakobsen, C. M., Janssen, S., Khan, S. R., Garrett, E. S., Lilja, H., Christensen, S. B., and Isaacs, J. T. (2003) J. Natl. Cancer Inst. 95, 990–1000). Our study demonstrates the usefulness of a simple in vitro system to test and direct development toward the formulation of new Tg derivatives with improved properties for SERCA targeting. Finally, we propose that the Tg binding pocket may be a regulatory site that, for example, is sensitive to cholesterol.
Understanding protein-membrane interaction and the activity of lipophilic drugs and specific lipids like sterols attracts increasing attention and represents a challenge to our current models of protein-solvent interactions and protein dynamics. Sarco/endoplasmic Ca2+-ATPase (SERCA)6 offers an attractive model for the analysis of such interactions due to favorable biochemical properties with sensitive assays of function, a large resource of inhibitors, and a wealth of crystal structures related to the functional cycle. In conjunction with plasma membrane Ca2+-ATPases and Ca2+ channel proteins, SERCA orchestrates spatiotemporal changes in the concentration of Ca2+ inside the cell, providing the background essential for the function of Ca2+ as a second messenger in the regulation of numerous cellular processes (1,–3). At the same time, high cytosolic (>1 μm) concentrations of Ca2+, induced by depletion of the endoplasmic Ca2+ stores and uptake of extracellular Ca2+ by the endoplasmic store-regulated mechanism (4), are involved as essential factors in the induction of apoptotic processes, leading to mitochondrial disruption with release of cytochrome c, activation of intracellular caspases, and cell death (5). As a consequence, interference of cellular function by inhibition of SERCA may lead to apoptotic cell death irrespective of the proliferative status of the cell. In this connection, thapsigargin (Tg, compound I, see Fig. 1) is of particular interest as a specific inhibitor of SERCA. Tg is a sesquiterpene lactone, present in the Mediterranean umbelliferous plant, Thapsia garganica. Tg has been shown to bind to the E2 form of SERCA (6,–9) with very high affinity at a transmembrane site located at the protein-lipid interphase between the M3, M5, and M7 (10, 11). The target specificity and potency make Tg very valuable for studies on the cell physiological role of SERCA (12) and also has enrolled the compound in the development of a prodrug strategy for killing prostate cancer cells (13). Targeting is based on the ability of prostate cells to express the protease prostate-specific antigen on the cell surface (14). By conjugating Tg, substituted with a 12-aminododecyl linker at O-8 to a peptide possessing the C-terminal amino acid sequence His-Ser-Ser-Lys-Gln-Leu, a prodrug has been obtained that is unmasked by specific cleavage of the Gln-Leu peptide bond. The released Tg derivative is immediately taken up by the prostate cancer cells, leading to their demise over a few days (13). We previously described how a related Tg analog with a tert-butoxycarbonyl-12-aminododecanoyl linker (Boc-12-ADT) interacts with the ATPase after cocrystallization and determination of the structure by x-ray diffraction at low resolution (15). We found that this compound is bound in the Tg binding cavity of SERCA 1a but, quite surprisingly, with the hydrophobic methylene groups of the linker penetrating the membrane sector of the protein by insertion between the transmembrane helices 3 and 5, with the hydrophilic Boc group ending close to the region of the second Ca2+-binding site (15). Here, we investigate the basis for the high affinity interaction of SERCA 1a with Tg by comparison with the structure of the binding site and affinity of many Tg analogs with modified or desubstituted side chains. In this way, our study has pinpointed the factors that are crucial for the interaction of these compounds with SERCA and resolved a number of questions concerning the role of precise location and orientation of the side chains and of the hydrophobic properties of the compounds in the formation of the Tg pharmacophore as well as for the interaction with the lipid phase of the membrane. This has led not only to a deeper understanding of the affinity and specificity of the Tg-SERCA interaction, but our data may also serve as a rationale for development of new drugs with improved properties for targeting of Ca2+ ion pumps.
Tg, selectively radiolabeled in the butanoate side chain of the molecule, was prepared by 3H-catalyzed reduction after substitution of the 8-O-butanoyl group of Tg with but-3-enoyl (16). The resulting product was found to contain ~10% impurities that, unlike authentic Tg, did not bind to Ca2+-ATPase membranes or hydrophobic surfaces. A scheme of the Tg analogs used is shown in Fig. 1. Tg analog dOTg (II) (Fig. 1) and dATg (III) were synthesized as described previously (17). DTB (VI) was prepared by esterification of 8-O-debutanoylnortrilobolide with dodecanoic acid using dicyclohexylcarbodiimide as a catalyst. Boc-ϕTg (VIII), was prepared by esterification of 8-O-debutanoyl-Tg (IV) with 3-(5-tert-butoxycarbonylaminopentyl)-4-methoxymethoxyphenylpropanoate. The synthesis of this acid involved a Friedel-Craft reaction between ethyl 4-methoxyphenylpropionic acid and 5-phthalimidopentanoyl chloride to give the carbon skeleton. A series of protections and deprotections of the functional groups was needed to get the target compound. These reactions are described in detail in supplemental Scheme 1, a–c. Linkers containing one oxygen atom (compounds IX and X) were prepared using 1,6-hexanediol and ethyl 5-bromopentanoate as starting materials (see supplemental Scheme 3). The ethyl ester in both compounds was selectively hydrolyzed with lithium hydroxide in methanol. Esterification with debutanoyl-Tg (IV), according to the above mentioned method, nicely produced compounds IX and X. The same strategy was used for preparation of the linker containing three oxygen atoms using triethylene glycol and ethyl bromoacetate as starting materials to give compounds XI and XII (supplemental Scheme 3).
For Ca2+-ATPase, cocrystallization with inhibitor was achieved by solubilization in 35 mm octaethylene glycol dodecyl ether (C12E8) in 100 mm MOPS (pH 6.8), 20% glycerol (v/v), 80 mm KCl, 3 mm MgCl2, and 2 mm EGTA in the presence of 125 μm Tg analog. The solubilization was followed by ultracentrifugation for 35 min at 50,000 rpm in a Beckman TLA 110 rotor to remove insoluble residues. The supernatant, with a protein concentration around 12 mg/ml, was stored overnight at 4 °C and subjected to another ultracentrifugation for 15 min at 70,000 rpm prior to the crystallization setup. The vapor diffusion method with hanging drops was used for crystallization at 12 °C with 2 + 2-μl volumes of protein solution and reservoir buffer. For cocrystallization of dOTg and E2-ATPase, the reservoir buffer contained 18% w/v PEG 2000 monomethyl ether, 4% v/v glycerol, 0.05 m NaCl, and 4% v/v 2-methyl-2,4-pentanediol, whereas for E2-ATPase crystals with DTB or Boc-ϕTg inhibitors were obtained with 14% w/v PEG 2000 monomethyl ether, 10% v/v glycerol, 0.1–0.2 M MgSO4, and 4% v/v tert-butanol. Single squared crystals appeared after a few days and grew to a maximum size of 250 × 250 × 50 μm within a couple of weeks. The crystals were mounted with a bent litho loop (Molecular Dimensions) and stabilized by transferring them shortly to a solution consisting of 2 μl of reservoir buffer and 1 μl of 50% v/v glycerol before flash-cooling in liquid nitrogen.
Diffraction data for the crystals were collected at the Swiss Light Source beamline X06SA and the European Synchrotron Radiation Facility at beamline ID23-1. The E2(dOTg) diffraction data were processed with the HKL2000 package and for the E2(DTB) and the E2(Boc-ϕTg) data with the XDS package to 3.1, 2.65, and 3.1 Å resolution, respectively, all in the space group P41212.
The crystallographic data obtained from the E2(dOTg), E2(DTB), and E2(Boc-ϕTg) complex were isomorphous to data of the E2(Tg)-AMPPCP complex (PDB code 2C8K). The structures were accordingly determined on the basis of the E2(Tg)-AMPPCP structure (with Tg omitted). For initial positioning of dOTg and DTB, energy-minimized starting models obtained from the program Macromodel (Schrödinger) were used. Inhibitors (dOTg, DTB, and Boc-ϕTg) were located by unbiased Fobs − Fcalc electron density maps, further guided by Fobs − Fcalc isomorphous difference Fourier maps between data sets of dOTg, DTB, Boc-ϕTg, and Tg complexes. The molecular graphics programs O (18) and Coot (19) were used for model building. The resulting structures were refined with CNS (20) with the topology and parameter files for the inhibitor derived from the PRODRG server (21), and the parameter files for a bound lipid molecule obtained from the HIC-Up server (22). The final structures were evaluated with the Molprobity server (23).
The functional effects of binding of Tg and analogs were primarily assessed by activity measurements, adapted for evaluation of high affinity (down to subnanomolar) binding constants of the inhibitor. For this, the Ca2+-ATPase at a protein concentration of 1 mg/ml was pre-equilibrated with different inhibitor concentrations for 2.5–6 min, in a medium containing 1 mm EGTA, 1 mm Mg2+, 10 mm Tes (pH 7.5), and 100 mm KCl. Activities were then measured spectrophotometrically by addition of 10 μl of the pre-equilibrated mixture to a cuvette, containing 3 ml of an ATP-regenerating assay medium in 0.1 mm Ca2+, 1 mm Mg2+, 5 mm MgATP, 1 mm phosphoenolpyruvate, and other additions for spectrophotometric registration of enzyme activity (24). The inhibitor (Tg or Tg derivative), solubilized in DMSO, was added cumulatively to the ATPase pre-equilibration mixture in increments of usually 0.2–0.4 mol/mg protein. Examples of the inhibition curves obtained and their evaluation in terms of Kd (50% inhibition of activity) are shown in supplemental Fig. S4. We found that the gradual increase in inhibitor concentration sample, effectuated by the incremental addition, facilitated the attainment of binding equilibrium after each step, which especially for analogs with long hydrocarbon tails often was a slow process, requiring up to 6 min of preincubation after each addition (as also evidenced from slow inhibitor-induced changes in intrinsic fluorescence). The preincubation of ATPase with inhibitor and the use of glassware instead of plastic was also done to avoid as much as possible unspecific adsorption of inhibitor during the assay.
Changes in intrinsic (tryptophan) fluorescence were recorded on a Shimadzu RF5301 spectrofluorometer with continuous stirring in a thermostated cuvette at 23 °C. The protein concentration was 50 μg/ml, and measurements were made with the monochromators set at 290 and 335 nm as the excitatory and emission wavelengths, respectively. Phosphorylation by[γ-32P]ATP and 45Ca2+ binding measurements was as performed previously (25) by Millipore filtration.
In Fig. 2, we have explored the effect of Tg binding on the structure of Ca2+-ATPase in the E2 state. A detailed comparison was possible, because several crystal forms of this state are available, with and without bound Tg (supplemental Fig. S3), and the comparison of the Tg-binding site applies also to the E2 forms described below in this paper. As can be seen from Fig. 2A, the Tg tricyclic guaianolide ring is bound in a cavity that faces the protein/lipid interphase, corresponding to the cytoplasmic leaflet of the membrane. Within the binding site, Tg adopts a favorable energy minimized conformation within a preformed cavity with the tricyclic guaianolide ring grasped between the M3 and M7 transmembrane helix, and with M5 forming the back of the cavity. To illustrate the small differences in the E2 conformation with and without Tg, in Fig. 2, B and C, we have shown the appearance of the binding cavity in the E2-P-like transition state solved in the presence of Tg (E2(Tg)-AlF4, 1XP5, 3.0 Å) or in the absence of Tg (E2-AlF4−-AMPPCP, 3B9R, 3.0 Å). Overall, other available structures representing the Tg-bound forms are only slightly different with root mean square deviations of 1.5 Å when superimposed on all Cα positions. In Fig. 2D, we have superimposed two of our E2 structures in the E2-P-like transition state solved in the presence and absence of Tg. In the presence of Tg, the binding cavity is slightly larger because of movements of some of the amino acid side chains to accommodate the bound inhibitor. Thus, the side chain of Phe-256, which from mutational evidence has been identified as a key residue in the binding of Tg (26, 27), is rotated backward into the transmembrane region, to establish the correct position for stacking of the phenyl group against the central cycloheptane and γ-lactone ring of the Tg tricyclic guaianolide nucleus. Furthermore, Gln-259 is rotated 180° outward to allow for accommodation of Tg inside the cavity, although Phe-834 undergoes a slight rotation in the opposite direction, combined with a small cytosolic shift, probably to realize a favorable interaction between the phenyl ring and the carbonyl group of the acetoxy group of Tg at O-10. On the back, the angeloyl substituent of Tg at O-3 interacts with Ile-765, Asn-768, and Val-769 of M5 and with Val-263 of M3. For Ile-765, this requires rotation of the side chain, a movement that, together with immobilization of M3 and of Asn-768 (one of the Ca2+ ligand residues at site I), effectively prevents the conformational changes leading to Ca2+ binding in the E1 state. At the upper pole Tg is capped by Pro-827, Leu-828, and Ile-829 of the 825KEPLIS motif present at the end of the L6–7 loop that like a lid covers the binding site from exposure to the cytosol. At the lower end of the binding cavity, the octanoyl substituent at O-2 extends into the lipid phase of the membrane with only weak contacts to the ATPase via Met-838 and Val-769. Although the binding cavity is tightly fitted for binding of Tg, there are few specific contacts (<3.2 Å) with the surrounding amino acid residues (mainly between the acetyl group at O-10 and Phe-834 and the butanoyl group at O-8 and Ile-828). Furthermore, as noted previously (17), the carbonyl and hydroxyl groups of Tg are not in register for hydrogen bond formation with hydrogen bond donating or accepting groups on the protein. Thus, the binding bears all the hallmarks of being hydrophobic in nature. However, a theoretical GRID analysis performed with a methyl probe suggests a localized nature of the hydrophobic interactions, hydrophobic molecular interaction fields mainly being present corresponding to the O-3 angeloyl, the C-4 methyl, the O-8 butanoyl, and O-10 acetyl substituents, although similar interactions appear to be absent from the tricyclic ring structure (17).
We have experimentally approached these questions by activity measurements where we have tested the effect of the side chains on the binding affinity after their systematic removal from Tg. These desubstituted derivatives were prepared by previously described methods (17, 28, 29), but we note that the unavailability of desoctanoyl Tg forced us to use as a model compound the previously described dihydronortrilobolide, which differs from desoctanoyl Tg by having a saturated bond between C-4 and C-5 (30). From the activity/inhibitor concentration curves, we find for all desubstituted compounds substantial reductions in the affinity, from a subnanomolar value (0.2 nm) for unmodified Tg to 3.5–113 nm after loss of the side chains. The 3rd and 4th rows of Table 1 indicate the associated changes in Gibbs free energy (ΔG0′) and those estimated for binding of the side chains from the difference in binding energy (ΔΔG0′) of each of the Tg derivatives and that of Tg. It can be seen from Table 1 that the effect of the substituents on binding affinity increases in the order O-2 octanoyl < O-8 butanoyl < O-10 acetyl < O-3 angeloyl. Furthermore, the total effect of the side chains, calculated as the sum of their ΔΔG0′ values (−45 kJ/mol), would amount to about 80% of the total binding energy for Tg (−55 kJ/mol). These calculations provide support for the previously suggested pharmacophore model (17), according to which a major part of the binding energy comes from the interaction of Ca2+-ATPase with the separately spaced Tg side chains. According to these calculations, less contribution will come from the guaianolide tricyclic ring, its major role being to serve as a scaffold to bring the side chains in correct position for hydrophobic interactions inside the binding cavity (17, 30). In a previous study, Wictome et al. (32) assigned an important role of the lactone ring to the binding mechanism based on a drastic (micromolar) reduction in binding affinity upon dehydration of the OH-7 and OH-11 groups. This is a conclusion that is only partly confirmed by our studies, because although we do find a decreased binding affinity by ring opening and other modifications of the lactone ring, these changes will also affect the spatial configuration of the side chains and are generally not larger than observed here by their removal.
Due to its hydrophobic properties, Tg can be expected to partition into the membranous lipid phase. Would this occur independent of protein binding (i.e. in a kinetic scheme by a parallel reaction pathway) or does binding occur in series with a preceding uptake by the lipid phase? With the aid of tritiated Tg, we have attempted to resolve questions concerning the role of the lipid phase for the interaction with Tg as shown by the chromatographic data in Fig. 3. As can be seen from Fig. 3A, Tg, when reacted with stoichiometric amounts of Ca2+-ATPase in the E2 conformation (presence of 1 mm EGTA), remains firmly bound to the membranes, being only slowly eluted by passage through a long Sephadex column. However, almost the same amount of Tg remains attached to membranes in the Ca2E1 conformation, following chromatography in the presence of 0.5 mm Ca2+ (Fig. 3B). Because Tg does not react with Ca2+-ATPase under the conditions of the latter experiment, this is an indication of binding by the lipid phase. In accordance with this conclusion, Tg was also retained by lipid after addition of Tg to unilamellar 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) liposomes (Fig. 3C), irrespective of whether these were eluted in the presence of EGTA or Ca2+. Thus, it is clear that Tg regardless of its interaction with Ca2+-ATPase is avidly taken up by the lipid membrane phase, from which it is only removed with difficulty by chromatography. Therefore, the first step in binding of Tg at the protein site can be proposed to take place by interaction with the membrane lipid phase from which the compound in a following step moves into the protein-binding site, obviating a much more difficult access via the cytoplasmic interphase. Furthermore, we found that the ATPase is capable of interacting with the tritiated Tg after solubilization by C12E8 and HPLC in the presence of C12E8 (Fig. 3D). Under these conditions, the tritiated label partitioned between the detergent-solubilized and -delipidated E2-ATPase monomer and the mixed micelles of C12E8 and sarcoplasmic reticulum lipid. The molar binding ratio of Tg to Ca2+-ATPase under these conditions was estimated to be 1.2 ± 0.3. Activity tests with C12E8 ATPase indicated a dissociation constant of 10 nm, i.e. around a 50 times decrease in binding affinity. Thus, these data substantiate the important role of the membrane lipid not only in mediating contact with the binding site but also in modulating the affinity of Tg at the binding site.
Overall, the decreased affinity of the desubstituted Tg analogs (Table 1) fits well with the result of the previous GRID analysis (17), although with one exception: for dOTg, subnanomolar affinity was predicted, similar to that of Tg, based on the assumption that the octanoyl group would not contribute to the interaction with the protein. However, it is clear from the Tg-SERCA structure that the octanoyl group is located in a position where instead of interacting with the protein it is expected to react with the lipid at the protein-lipid interphase. In agreement with these considerations, x-ray diffraction analysis of cocrystals of ATPase with dOTg shows that binding of the central ring structure and remaining side groups is only slightly affected in comparison with Tg (Fig. 2, E–G). Thus the additional affinity with which Tg is bound, compared with dOTg (Table 1), can probably be attributed to interaction with membrane lipid. As will be shown below, dOTg also reacts with Ca2+-ATPase with different kinetics.
In support of an adaptable interaction of Tg with Ca2+-ATPase, we have found that high affinity is retained when the butanoyl group at O-8 is replaced with a wide variety of flexible hydrophobic acyl groups (Table 2). Most of these compounds represent new syntheses, with the exception of DTB and Boc-12ADT. DTB is an O-8 dodecanoyl derivative of the naturally occurring nortrilobolide which, like dOTg, is deoctanoylated at O-2, whereas Boc-12ADT, with a 12-aminododecane side chain, is a model compound for the Tg released from the peptide prodrug for targeting prostate cells (17) and whose structure with Ca2+-ATPase we published previously (15). By cocrystallization, we have now at 2.65 and 3.1 Å resolution also determined the structure of DTB and Boc-ϕTg (compound VIII, with a phenol group incorporated into the side chain), respectively. Unbiased (Fobs − Fobs) isomorphous difference and model-based (Fobs − Fcalc) difference Fourier maps combined with comparisons of refined structures (supplemental Figs. S1 and S2) provide our basis for discussing the structural details of the binding of these analogs as illustrated in Fig. 4. We find that the central ring structure of both analogs is bound in basically the same way as Tg between the M3, M5, and M7 transmembrane segments. However, by superimposition of the structures with Tg (Fig. 4, A and B), it can be seen that the upper puckered 7-membered ring and the lactone ring is slightly pushed outwards, although the lower five-membered ring is slightly moved inwards as compared with Tg. The overall structures that Ca2+-ATPase adopt in complex with these compounds are practically unchanged, as indicated by root mean square deviations of 0.52–0.8 Å for Cα atoms, compared with a number of other Tg-bound structures (PDB 2C8K, 2DQS, and 2AGV, see supplemental Fig. S3). But the most remarkable feature of these structures is that the long side chain at O-8 is internally located, penetrating between transmembrane helices of Ca2+-ATPase membrane helices, starting at a crevice forming between M3 and M5 as was previously reported for Boc-12ADT (15). For both DTB and Boc-ϕTg, the first part of the O-8 side chain, like the butanoyl group of Tg, moves upwards in the cytosolic direction toward the above-mentioned crevice. Here, the further course of the long chain linkers is changed toward the inside of the ATPase transmembrane domain. For DTB, the hydrocarbon chain during the intramembranous course gradually disappears in the electron density maps, probably due to the inherent flexibility of the dodecyl chain in the hydrophobic environment between M3, M4, and M5 segments (supplemental Figs. S1B and S2B). However, for the less flexible Boc-ϕTg (supplemental Figs. S1C and S2C), we find that the phenol ring (incorporated at its 2 and 4 position) forms a rigid and defined bend at the crevice that by displacing the side chain of Val-253 inserts between M3 and M5 and directs the linker toward the N-terminal region of the pump. During this traverse, the intervening hydrocarbon chain proceeds in front of M4 close to Glu-309 of the 308PEGLP motif that is involved in Ca2+ binding at site II and ends close to the kink region of M1. This region forms part of the putative entrance for cytosolic Ca2+ binding (Fig. 4, C–F). The location of the last part of the chain with the Boc group is also close to the head group of the phospholipid present in a cavity between M2 and M4 and overlaps with the binding site of BHQ (Fig. 4E) (33) and CPA (Fig. 4F) (34,–36), other well known inhibitors of Ca2+ transport. This closely defined course, obtained with a less flexible linker, is consistent with the insertion previously deduced for the linker with the Boc-12ADT analog (15).
The dissociation constants (defined as the concentration of free inhibitor giving 50% inhibition of activity) for the seven analogs with long side chains at O-8 range from 1.3 to 48 nm as shown in the 3rd column of Table 2. In particular, the side chain linkers where one ether group has replaced two methylene groups in the dodecyl linker (compounds IX and X) retain high affinity. Surprisingly, the inhibition of Ca2+-ATPase by Boc-12ADT, the structure for interaction with Ca2+-ATPase that we previously published (15), is much less efficient (48 ± 4 nm) than that of Tg and the other analogs. The much smaller effect of Boc-12ADT was somewhat unexpected, because the analog with Leu-12ADT linker previously was reported to be as efficient as Tg in producing apoptosis in prostate cancer cell lines (29). The sluggishness of Boc-12ADT may have to be considered in connection with the slow kinetic features of the compound in the interaction with SERCA as indicated by the data presented below.
The kinetics of the binding of the compounds were registered by the decrease in the intrinsic tryptophan fluorescence of Ca2+-ATPase accompanying the interaction (Fig. 5). Despite that Tg is bound with extraordinarily high affinity, the time course of the binding can easily be followed, because it proceeds slowly over a period of minutes with a rate constant around 1 min−1 to ATPase in the E2 conformation (Fig. 5A). The interaction is slightly longer for binding to ATPase in the Ca2E1 state (3.5 min−1) where it is accompanied by the concomitant release of two Ca2+ ions (Fig. 5C). By contrast, the decrease in fluorescence observed by interaction of dOTg with E2 is instantaneous, whereas the interaction of the compound with Ca2E1 is biphasic, consisting of one instantaneous and one slow phase, each of which is accompanied by the release of one Ca2+ ion (cf. Fig. 5, C with B). Most of the O-8 linkers (with the exception of DTB and Boc-12ADT) react with E2 in a biphasic reaction modus but are inefficient in displacing more than one bound Ca2+ from Ca2E1, findings which we attribute to the existence of an intermediate CaE form that is only incompletely converted to E2, as can be rationalized with the aid of the ramified reaction scheme shown in Fig. 5D. DTB is bound slowly both to E2 and Ca2E1, effects that are probably accounted for on the basis of slow incorporation of the unsubstituted dodecyl chain into the membranous domain. An even slower interaction is obtained with Boc-12ADT, which requires 1–2 h for completion, and where the interaction can only be detected as slow reductions in the Ca2+ binding (Fig. 5C), although no changes in intrinsic fluorescence are observed, apart from a tiny increase occurring immediately after the addition, a feature that is also revealed with the aid of the other slowly reacting analog DTB (Fig. 5B). With Boc-12ADT, a slow decrease in the fluorescence response, similar to that of the other analogs is probably not recognized, because it is masked by other time-dependent changes in fluorescence such as photolysis.
The structures of E2 ATPase that we have obtained with or without Tg reveal how Tg with few adaptations fits into a preformed cavity between the third, fifth, and seventh transmembrane segment, resulting in stabilization of the protein in a non-Ca2+ binding E2 conformation. The interaction is mainly of a hydrophobic nature, where high affinity and specificity is achieved by interaction with hydrophobic amino acid residues that, in accordance with a previous GRID analysis (17), interact with the side chain groups of Tg, brought in accurate position by the tricyclic guaianolide nucleus. By these interactions and by stacking with Phe-256, the lactone ring is distorted and, together with the upper part of the molecule pressed against the M5 and M3 helices, thereby probably preventing the movements that are necessary for Ca2+ binding and the E2 → Ca2E1 transition. The high affinity binding of Tg is an exclusive property of E2 conformations leaving the cation-binding sites in a proton-occluded and cytoplasmically oriented state. In accordance with the simulations of Paula and Ball, Jr. (37), we find that attempts to dock Tg in the Ca2E1 structure lead to stereochemical clashes, mainly because the cavity is slightly narrower and the transition to the E1 conformation drives the M3 helix with Phe-256 and other interacting amino acid side chains 5 Å in the cytoplasmic direction. Obviously, these features emphasize the role of complementarity, as a result of which the binding site does not interact well with the slightly different conformation of the binding cavity in the E1 states.
Kinetically, the binding reaction appears to occur in at least three stages, during which Tg and analogs are first taken up by the membrane lipid, then form a presteady state complex at the phospholipid/protein interphase, before finally being adapted within the binding cavity. The extent of hydrophobic interaction with membrane lipid probably is a significant kinetic factor which e.g. accounts for the slow rate (kon ~1 min−1) with which Tg interacts with the ATPase. This is also suggested by the rapidity with which the less hydrophobic dOTg reacts with the protein (Fig. 5A). The formation of the presteady state complex can take place both in the E2 and Ca2E1 state and is accompanied by a slight increase in intrinsic fluorescence, a change that is most easily observed with the slowly complexing analogs Boc-12ADT and DTB (Fig. 5), probably as the result of slow accommodation of their long hydrocarbon chains inside the membrane. The formation of the genuine complex with Ca2+-ATPase in the E2 conformation requires some adaptation of the binding site and is accompanied by a distinct decrease in intrinsic (tryptophan) fluorescence.
It is of interest that although binding cavities, similar to those for Tg in Ca2+-ATPase, are also present in the crystalline structures of Na+,K+-ATPase (38) (PDB code 3KDP) and plant H+-ATPase (39) (PDB code 3B8C), only Ca2+-ATPase can be inhibited by Tg. In Na+,K+-ATPase, the homologous Phe-285 occupies almost the same position within the binding cavity as Phe-256 in Ca2+-ATPase. However, in Na+,K+-ATPase, the binding cavity is narrower, mainly because M7, with an intramembranous kink at Gly-848, imposed by interaction with the β-subunit, is situated closer to M3 than it is in Ca2+-ATPase. Thus, we can conclude that the exact match of the hydrophobic binding cavity in the E2 conformation and the pharmacophore of Tg are the main factors behind the high affinity of Tg for binding to the E2 state of the Ca2+-ATPase.
We find that our new analogs of Tg provided with flexible linkers, substituting for the butanoyl group at O-8 of Tg, are bound in basically the same way as Tg, except that the guaianolide ring structure is less distorted and tilts outward. Thus, some adaptation has taken place within the binding site that probably results in a less intense interaction with amino acid residues in the upper part of the binding pocket. In all cases, the linkers are inserted between the M3, M5, and M4 transmembrane segments in a similar way as the aminododecyl side chain of Boc-12ADT (15). For Boc-ϕTg, we demonstrate that the chain proceeds toward the putative N-terminal Ca2+ entrance region close to the Ca2+-binding site II in such a way that the terminal Boc group overlaps with that of other well documented Ca2+-ATPase inhibitors, CPA (34,–36) and BHQ (33). Our experiments indicate that Boc-12ADT, which has been used as a model for the active component of a prostate cancer cell-killing prodrug (13, 40), is a very slowly working inhibitor of Ca2+-ATPase. Boc-12ADT was also much slower than DTB, the nortrilobolide analog where the side chain is an unmodified dodecyl group. To account for the slow behavior of Boc-12ADT, we consider that the dodecyl chain due to its strongly hydrophobic and flexible properties is prevented from extending beyond the hydrophobic barrier separating the terminal Boc group bound at Ca2+-binding site II from the Tg-binding site. For this purpose, the other modified linkers as used here (Table 2) may be more appropriate. In particular, we found that linkers where one or more oxygen atoms replaced methylene groups were capable of interacting much more rapidly and efficiently with the E2 conformation of Ca2+-ATPase than Boc-12ADT. The faster complex formation may be caused by the more amphipathic or other chemical properties of the oxygenated side chain.
Our study with SERCA 1a and Tg emphasizes the importance of the lipid phase of membranes as an entrance port for uptake of amphipathic and lipophilic compounds that after their uptake may affect the function of membrane proteins. Present concepts of membrane protein-lipid interaction suggest that Ca2+-ATPase is surrounded by a layer (an annulus) of nonspecifically associated lipid as well as endowed with binding of lipid at a few specific (nonannular) sites of probably regulatory significance for the protein (41). The cavity present between M3, M5, and M7 is a prime candidate for a nonannular site, which by its close relation to the Ca2+-binding sites would be capable of exerting a regulatory effect on Ca2+-ATPase activity. Although the cavity par excellence is adapted for binding of Tg, it could probably in the absence of this inhibitor be the site for binding of lipid and other hydrophobic compounds, including drugs and detergents that in low nonsolubilizing concentrations perturb the functional properties of Ca2+-ATPase (42,–44). So far, an endogenous regulator of the activity of SERCA interacting with the ATPase at the Tg-binding site has not been identified. Theoretical docking studies are compatible with the possibility that sterols such as cholesterol could fulfill a role as modulator of ATPase at this site.7 It has been found that cholesterol interacts with Ca2+-ATPase at annular sites after reconstitution of Ca2+-ATPase with cholesterol/dioleoylphosphatidylcholine mixtures, although with a somewhat lower affinity than phospholipids (41, 45). However, there is also evidence from tryptophan quenching studies with a brominated cholesterol derivative that cholesterol is bound more firmly at a specific nonannular binding site (46). In shark Na+,K+-ATPase (PDB code 2ZXE) (47), a cholesterol molecule has been found associated with the enzyme, although at a slightly different position between the α- and β-subunit where the pig kidney enzyme structure also shows a bound bilayer component (47).
With respect to the implications of our study for prostate cancer, it will be important to use the detailed knowledge accruing from the binding mechanism of Tg as a guide in the design of new inhibitors, amenable to commercially interesting chemical syntheses, to replace T. garganica as the limited plant resource of potent SERCA inhibitors. This may be feasible with the aid of a fragment-based approach (31) by synthesizing compounds where the interacting pharmacophore is built on a central skeleton placing the side chains in approximately the same topological position as in Tg (30). Furthermore, a linker at O-8 may allow for more flexibility in drug design to target prostate cancer cells by providing inhibitors that in addition to binding at the Tg site also interact at the same location as CPA and BHQ. These developments should go hand in hand with theoretical docking studies from which a number of the conclusions on the interaction of Tg and analogs, including the presteady binding reported here, were actually previously predicted by theoretical computations without prior experimental evidence (37). Finally, it should be borne in mind that similar cavities as in SERCA are present in other P-type ATPases. These may also be amenable to the elaboration of specific and individual inhibitors by fragment-based approaches.
We are grateful to Dr. E. Doris, SBCM, Commissariat à l'Energie Atomique, Saclay, France, for synthesizing [3H]Tg.
*This work was supported in part by the Danish Medical Research Foundation, the Danish Natural Science Foundation (through the Dansync Program), the Danish Cancer Society, the Lundbeck Foundation, the Aarhus University Research Foundation, the Novo Nordisk Foundation, and the Carlsberg Foundation.
The on-line version of this article (available at http://www.jbc.org) contains supplemental Figs. S1–S4, Tables S1 and S2, and Schemes S1–S3.
7Y. Sonntag, A. M. L. Winther, C. Olesen, J. V. Møller, and P. Nissen, unpublished data.
6The abbreviations used are: