Locating the thapsigargin-binding site on Ca2+-ATPase by cryoelectron microscopy1

doi:10.1006/jmbi.2001.4558 available online at http://www.idealibrary.com on

J. Mol. Biol. (2001) 308, 231±240

Locating the Thapsigargin-binding Site on Ca2‡-ATPase by Cryoelectron Microscopy Howard S. Young, Chen Xu, Peijun Zhang and David L. Stokes* Skirball Institute of Biomolecular Medicine Department of Cell Biology NYU School of Medicine, New York, NY 10016, USA

Thapsigargin (TG) is a potent inhibitor of Ca2‡-ATPase from sarcoplasmic and endoplasmic reticula. Previous enzymatic studies have concluded that Ca2‡-ATPase is locked in a dead-end complex upon binding TG with an af®nity of <1 nM and that this complex closely resembles the E2 enzymatic state. We have studied the structural effects of TG binding by cryoelectron microscopy of tubular crystals, which have previously been shown to comprise Ca2‡-ATPase molecules in the E2 conformation. In particular, we have compared 3D reconstructions of Ca2‡-ATPase in the absence and presence of either TG or its dansylated derivative. The overall molecular shape of Ca2‡-ATPase in the reconstructions is very similar, demonstrating that the TG/Ca2‡-ATPase complex does indeed physically resemble the E2 conformation, in contrast to massive domain movements that appear to be induced by Ca2‡ binding. Difference maps reveal a consistent difference on the lumenal side of the membrane, which we conclude corresponds to the thapsigargin-binding site. Modeling the atomic structure for Ca2‡-ATPase into our density maps reveals that this binding site is composed of the loops between transmembrane segments M3/M4 and M7/M8. Indirect effects are proposed to explain the effects of the S3 stalk segment on thapsigargin af®nity as well as thapsigargin-induced changes in ATP af®nity. Indeed, a second difference density was observed at the decavanadate-binding site within the three cytoplasmic domains, which we believe re¯ects an altered af®nity as a result of the long-range conformational coupling that drives the reaction cycle of this family of ATP-dependent ion pumps. # 2001 Academic Press

2‡

*Corresponding author

Keywords: Ca -ATPase; sarcoplasmic reticulum; ion transport; inhibitor-binding site; electron microscopy

Introduction Ca2‡-ATPase belongs to the family of P-type ion pumps, all of which use ATP to generate ion gradients across a wide variety of cellular membranes.1 The best-studied pumps appear to use the same, basic reaction cycle, the hallmark of which is transient phosphorylation of an aspartate residue in the catalytic site and cycling between two main conformations, dubbed E1 and E2.2 More speci®Present address: P. Zhang, National Institutes of Health, NIDDK, LCBB, Bethesda, MD 20892, USA. Abbreviations used: EM, electron microscopy; SR, sarcoplasmic reticulum; SERCA, Ca2‡ pumps from sarcoplasmic and endoplasmic reticula; TG, thapsigargin; DTG, dansylated TG; PLB, phospholamban; CPA, cyclopiazonic acid. E-mail address of the corresponding author: [email protected] 0022-2836/01/020231±10 $35.00/0

cally, these conformational changes serve to couple the energy of the aspartyl phosphate with translocation of ions across the membrane. Recent evidence suggests that these conformational changes involve dramatic rearrangement of the cytoplasmic domains with surprisingly small changes amongst the ten transmembrane helices. Originally, these Ê by cryostructural dynamics were revealed at 8 A electron microscopy (cryoEM) by comparing structures of Ca2‡-ATPase3 and H‡-ATPase from Neurospora,4 which were crystallized in different conformations.5 A more recent, X-ray structure of Ca2‡-ATPase revealed the atomic architecture of the E1 conformation and comparison with the cryoEM map indicated large movements of the cytoplasmic domains in response to Ca2‡ binding.6 For cryoEM of Ca2‡-ATPase, the particular conformational state required for crystallization was stabilized by adding the inhibitor thapsigargin (TG). This potent inhibitor is highly speci®c for the # 2001 Academic Press

232 Ca2‡ pumps from sarcoplasmic and endoplasmic reticula, known as SERCA, and was discovered as the major source of skin irritation from root extracts of Thapsia garganica, which were originally used for treatment of muscle and joint in¯ammation.7 Today, TG is widely used by cell biologists wishing to empty the internal Ca2‡ stores in eukaryotic cells, a result of total inhibition of the SERCA pumps. TG inhibits Ca2‡-ATPase by binding to the E2 conformation and by preventing further cycling of the enzyme.8,9 Normally, the E2 reaction intermediate is converted to the E1 intermediate by the binding of cytoplasmic Ca2‡ to their high-af®nity sites, which in turn activates the ATP-binding site to produce the phosphoenzyme intermediate (E1  P). In the presence of Ca2‡, TG does not inhibit the enzyme immediately, but allows a single cycle of phosphorylation and ion transport, indicating that TG can bind to only the E2 state. This interaction is of very high af®nity (Kd < 1 nM) and full inhibition is obtained with a 1:1 molar stoichiometry in what has been termed a dead-end complex. Tubular crystals require the E2 state, as indicated by their requirement for low Ca2‡ concentrations,10 as well as decavanadate,11 which mediates intermolecular contacts as well as intramolecular interactions between the three cytoplasmic domains.12 TG stabilizes these crystals,9 presumably by making them insensitive to Ca2‡ and by irreversibly trapping all Ca2 ‡-ATPase molecules in a conformation that is favorable for crystallization. For our earlier cryoEM structure of Ca2‡ATPase,3 we used the dansylated form of TG (DTG),13 in order to stabilize the tubular crystals, and with the hope that its larger size would increase our chances of locating its binding site. We have now solved several other structures in the Ê resolution presence and absence of TG at 8-10 A and calculated difference maps. These maps reveal a consistent, high-density difference on the lumenal side of the membrane. Based on the X-ray structure, this site is sandwiched between two lumenal loops connecting M3/M4 and M7/M8. We propose a mechanism for TG inhibition that involves increased rigidity amongst the transmembrane helices, thus preventing movements necessary for binding Ca2‡.

Results Enhanced crystallization by TG In previous work,9,10 TG was shown to stabilize decavanadate-induced 2D crystals of Ca2‡-ATPase against the disruptive effects of Ca2‡. In the current work, we also found that both TG and DTG greatly increased the number of tubular crystals formed under standard, Ca2‡-free crystallization conditions. In the absence of these inhibitors, a minority of vesicles in our SR preparation produced tubular crystals, even though 2D arrays were visible in virtually all the vesicles regardless

Thapsigargin-binding Site on Ca2 ‡-ATPase

of their shape (Figure 1(a)). In the presence of either TG or DTG, the great majority of vesicles gave rise to a tubular crystal (Figure 1(b) and (c)). The situation was more complicated for reconstituted proteoliposomes, which we originally reported as requiring thapsigargin for tube formation.14 Further investigation has revealed that the propensity of reconstituted proteoliposomes to form tubes is strongly in¯uenced by the lipid composition, especially phosphatidyl ethanolamine. Nevertheless, reconstituted preparations are comparable to native SR in that TG greatly increases the proportion of tubular crystals in a given preparation. TG also had a systematic effect on tube diameter. In general, the narrowest tubes in a given preparation were chosen for analysis, in order to restrict the range of symmetries obtained. Despite this selection, the SR treated with TG consistently Ê diameter) than yielded narrower tubes (540 A Ê ). This differthose that were untreated (565-580 A ence was re¯ected in the helical symmetry, which is characterized by Bessel orders (n) along principal layer-lines (1,0 and 0,1), which were systematically larger for untreated tubes compared to treated tubes (Table 1). Although tubes from reconstituted proteoliposomes are generally wider, the effects of TG were entirely consistent with those on SR. Finally, the resolution obtained in images of frozen-hydrated tubes was somewhat better from TG or DTG treated tubes, as seen in the Fourier transforms in Figure 2 and the phase statistics in Table 1. Although the resolution varies between 8 Ê , smaller numbers of tubes were suf®cient and 10 A for a given resolution when TG or DTG was used. Part of this improvement may be due to the larger number of tubes present in these treated samples and the concomitant ability to be more selective in imaging TG and DTG tubes, but as a qualitative observation, TG de®nitely appears to improve crystalline order. A plausible explanation for these effects is that TG strengthens intermolecular crystal contacts, thus helping to generate the high radius of curvature characterizing these tubes. Indeed, the preponderance of spherical vesicles in the absence of TG suggests that most arrays were not able to overcome this energy barrier and, even when they did, they tended to have lower curvature (i.e. larger diameter). The lipid composition will certainly affect the energy required to bend the surface and, in particular, the destabilizing effect of phosphatidyl ethanolamine on lipid bilayers15 is consistent with its role in promoting tube formation. In addition, co-reconstitutions with the cardiac regulatory protein phospholamban (PLB), more readily produced tubular crystals, which could re¯ect either the documented properties of PLB in oligomerizing Ca2‡-ATPase16 or uncharacterized effects of this small membrane protein on the physical stability of the bilayer. On the other hand, it is unlikely that TG and DTG are altering the physical chemistry of the bilayer. The af®nity of TG for

233

Thapsigargin-binding Site on Ca2 ‡-ATPase Table 1. Statistics of data analysis Control Native Number of crystals Helical symmetriesc

a

6 ÿ25,7

The 2-fold phase statistics (deg.)d Ê 40 A 2.8 Ê 30 A 9.7 Ê 20 A 19.5 Ê 14 A 35.6 Ê 10 A Ê 8A Overall 17.2

Thapsigargin PLB

TG

DTG

16 ÿ26,7 ÿ26,8 ÿ27,7

32 ÿ26,9 ÿ27,9

7 ÿ22,6

23b ÿ21,6 ÿ22,6 ÿ23,6

1.2 3.9 16.5 31.0 43.5 21.3

0.7 1.7 3.8 12.5 23.8 38.8 11.6

3.9 5.5 11.2 35.2 41.1 24.1

1.4 2.2 6.2 18.0 32.2 41.8 20.4

a

Structure previously published by Yonekura et al.17 Structure previously published by Zhang et al.3 c Helical symmetry characterized by helical start number (Bessel order) of the (1,0) and (0,1) layer-lines, respectively. d Amplitude-weighted phase residual relative to 0  or 180  including 100 % of the layer-line data. Because of the 2-fold symmetry, a phase residual of 45  indicates random phases. b

Ca2‡-ATPase is exceedingly high (Kd < 1 nM)8 and stoichiometric amounts are suf®cient for full inhibition (i.e. binding). The 10 mM TG used for crystallization corresponds to 1 % of the concentration of lipid, but most of this will be bound to Ca2‡-ATPase (present at 7-8 mM). The addition of a dansyl moiety to TG reduces its af®nity to 2 mM13 and the 30 mM DTG added for crystallization should theoretically bind to 92 % of Ca2‡ATPase molecules, leaving an excess corresponding to 2 % of the lipid concentration. Even this excess is far less than the 10-20 % phosphatidyl ethanolamine required to promote tube formation in reconstituted preparations, and is therefore unlikely to have a direct effect on bilayer stability. Averaged structures In addition to two previously published maps of Ca2‡-ATPase in the presence3 and absence17 of DTG, we have calculated three new maps in the presence and absence of TG, aligned the individual molecules and determined pair-wise difference maps. Two of these data sets (DTG and PLB) Ê resolution, two (TG and native) to extend to 8 A Ê , whereas an older native data set17 stops at 10 A Ê (see Table 1 for a summary of data sets). 14 A These reconstructions were completely independendent and involved the work of several different individuals, thus any correlations in the difference densities can be considered highly signi®cant. General inspection of the structures (Figure 3) reveals that neither TG nor DTG have changed the overall architecture of the molecule, which is composed of a compact group of cytoplasmic domains that are connected to the membrane by a stalk, a group of ten transmembrane helices that are capped by two, small lumenal domains. This can be contrasted with the effects of Ca2‡-binding, which caused the three cytoplasmic domains to separate dramatically, apparently providing free

access to the site of phosphorylation (Figure 3(d)). Ê reconstructions was determined One of the 8 A from co-crystals of Ca2‡-ATPase and PLB; however, the presence of PLB appears not to have affected the structure of Ca2‡-ATPase, and for the purposes of this report this data set is used to represent the native molecule (i.e. in the absence of TG). The differences attributable to PLB will be the subject of another report (H.S.Y., L.R. Jones, D.L.S., unpublished results). Difference maps Inspection of the structures from treated and untreated tubes in Figure 3 does reveal one obvious difference in the lumenal domain. The untreated structure consists of two lobes, one of which corresponds to the M7/M8 loop and the other to the M1/M2 and M3/M4 loops (see Figure 5). In the treated structure, these lobes are bridged by rather strong density that cannot be accounted for easily with the X-ray coordinates. To address these differences in a quantitative way, we calculated difference maps after masking and aligning individual molecules from the various maps (Figure 4). Most of these difference maps Ê resolution, except for those were calculated at 10 A including the older native data set from SR, which Ê . In addition to adjusting the were limited to 14 A physical scale of the molecules, we adjusted the density scale by comparing the maps pixel by pixel and ®nding the best least-squares solution for the scale factor. Subtraction of maps produced Gaussian distributions of difference densities (Figure 4(g)) and the difference maps were contoured at 3.0 times the standard deviation of this Gaussian distribution (3s). At this level, two consistent, positive differences are revealed, one between the two lobes of the lumenal domain and a second at the intersection of the three cytoplasmic domains. Similar difference maps

234

Thapsigargin-binding Site on Ca2 ‡-ATPase

Ê data set for calculated with an older, 14 A untreated Ca2‡-ATPase reveal a single difference density above 3s in the lumenal domain (Figure 4(d)). A few other positive differences are seen in the various maps; however, these were mostly at the perimeter of the cytoplasmic domain and are not consistent from one map to the next and therefore are likely to re¯ect noise or slight misalignment of molecules. In contrast to the consistent positive densities, no signi®cant differences were found from the following ``control'' maps: negative difference densities from these same difference maps (Figure 4(e)) and difference maps calculated from pairs of treated (DTG ÿ TG) or untreated (PLB ÿ native) crystals (Figure 4(f)). In these cases, there are almost no difference densities above 3s and differences above 2s are uncorrelated between the various maps, suggesting that they are due to noise.

Discussion Site of thapsigargin-binding

Figure 1. TG and DTG promote formation of tubular crystals. (a) Incubation of SR with decavanadate-containing solutions causes array formation in virtually every vesicle, but only a minority of vesicles elongate into the tubular crystals used for structure determination. When (b) TG or (c) DTG is added during crystallization, virtually all vesicles are converted into these tubular crystals.

Of the two consistent density differences that we observed, we believe that the lumenal density corresponds to TG itself and that the cytoplasmic site re¯ects a variable occupancy of the crystallization agent decavanadate. Indeed, two independent modeling efforts have concluded that decavanadate binds to a site between the three cytoplasmic domains, precisely where our cytoplasmic difference is located (compare Figure 3(b) with Figure 4(a) and (b)). This conclusion was initially based on the presence of a very high, spherical density12 and was later con®rmed by the inability to account for this density after ®tting atomic coordinates from X-ray crystallography to our EM structure.6 In fact, decavanadate appears to have two binding sites in these crystals, the other mediating an intermolecular contact at a crystallographic 2-fold axis.12 The existence of two sites is consistent with the measured decavanadate:Ca2‡ATPase stoichiometry of 1.5:1, with the further observation that labeling of Lys515 near the ATPbinding site with ¯uorescein isothiocyanate reduces this stoichiometry to 0.5 (i.e. this labeling displaces the intramolecular decavanadate) without preventing crystallization.18 The difference density between the two lumenal lobes is most likely to correspond to the TGbinding site. By ®tting the atomic coordinates for Ca2‡-ATPase to the transmembrane portion of our density map, it is clear that these two lobes are formed by the M7/M8 loop on the left and the M3/M4 loop on the right (red loops in Figure 5); the M1/M2 loop also contributes to the right-hand lumenal lobe, but is further back behind the plane of the difference density. This suggests that TG is intercalated in between these two lumenal loops and its binding speci®city may result from interaction with both. Indeed, when TG was docked into the extra density in the TG and DTG maps, its

Thapsigargin-binding Site on Ca2 ‡-ATPase

235

Figure 2. Computed Fourier transforms from tubular crystals. Data from (a) TG and (b) DTG tubes generally extend to higher resolution than those from untreated (c) PLB and (d) native tubes. Values in parentheses indicate the Miller indices assigned to layer-lines (h,k) followed by their helical start number or Bessel order (n). The smaller Bessel orders for (1,0) and (0,1) layer-lines from treated tubes correlate with their generally smaller diameter. Red circles indicate the zeros in the CTF, which range in these examples from (c) 690 nm to (d) 1.76 mm; the layer-line Ê resolution. designated (3,11) is at 10.5 A

size and shape were consistent with an interaction between these loops (Figure 5). Using the constraint that azido-TG (azido group at the C-8 position) crosslinks to the M3/M4 helical hairpin,19 we tentatively propose that the aliphatic tail at the C-2 position of TG interacts with the M7/M8 loop of Ca2‡-ATPase and that the more hydrophilic end of TG near C-7 and C-8 interacts with residues in the M3/M4 loop. The location of the TG-binding site has been investigated by a variety of indirect methods. These studies have built a consensus for TG binding along the stalk helix 3 (S3) near the cytoplasmic membrane interface just before M3 (Figure 5). In

particular, ¯uorescence energy transfer between DTG and either ¯uorescein isothiocyanate (in the cytoplasmic domain) or tryptophan residues (mostly in transmembrane domain) suggests that DTG binds within or near the membrane.13 As mentioned, photoactivated crosslinking of Ca2‡ATPase with an azido derivative of TG labeled a peptide corresponding to S3, M3, M4, and S4.20 This labeling is consistent with studies of Ca2‡ATPase and Na‡/K‡-ATPase chimeras, which associated TG sensitivity to the sequence including S3, M3 and the M3-M4 loop.21 Finally, the S3 stalk helix itself was implicated by site-directed mutagenesis19 and by natural mutants at Phe25622

236

Thapsigargin-binding Site on Ca2 ‡-ATPase

Figure 3. Averaged maps. The best-de®ned maps came from (a) PLB and (b) DTG data sets, which represent Ê resolution, the resolution was trununtreated and treated Ca2‡-ATPase, respectively. Although the data extend to 8 A Ê for comparison with other data sets. Most structural features are well conserved, with the most obvious cated to 10 A difference being a bridging density between the two lumenal lobes. The white, spherical density in (b) is at a very high density threshold (6s) and indicates the location of the intramolecular decavanadate-binding site. Atomic coordinates from Toyoshima et al.6 are shown in (d) and have been remodeled to ®t the EM map in (c); transmembrane domains are in equivalent orientations in (c) and (d), though cytoplasmic domains have obviously undergone a large conformational change (unpublished results). Note that the decavanadate ®ts into a space between the yellow and green domains in (c).

that confer resistance to TG. The fact that our difference map shows no positive peak at the cytoplasmic membrane surface (Figure 4) suggests that the effects of mutagenesis at this site are indirect. This could explain why these mutations caused an only 100-fold decrease in TG sensitivity (from subnM to 100 nM),19 whereas the exchange of the larger region including the M3/M4 loop resulted in chimeras that were insensitive even to 2 mM TG.21 An indirect role for the S3/M3 interface may explain the results of recent studies of the chemically unrelated inhibitor cyclopiazonic acid (CPA). Like TG, this inhibitor traps Ca2‡-ATPase in the E2 conformation and is able to promote formation of tubular crystals. A recent mutagenesis study showed that many of the same residues at this S3/M3 interface reduced CPA sensitivity significantly.23 The idea that the binding site for CPA should overlap with that of TG was surprising, given that their chemical nature is so different. In light of our difference maps, we propose the alternative explanation, that this S3/M3 interface mediates the effects of inhibitor binding by relaying the corresponding conformational changes from one part of the molecule to the other.

Effects of TG inhibition Previous studies have concluded that, after binding, TG holds Ca2‡-ATPase in a conformation similar to E2.9 In fact, TG binds only when Ca2‡ATPase is in the E2 conformation and, thereafter, Ca2‡-ATPase is unable to move either forward (i.e. bind Ca2‡ to produce E1
Ca2) or backward (i.e. react with Pi to produce E2-P) in its reaction cycle. In the former case, TG prevents Ca2 ‡ binding, as measured either directly with 45Ca or by intrinsic tryptophan ¯uorescence.9 Because most tryptophan residues occur in the membrane, the well-documented increase in ¯uorescence during Ca2‡ binding24,25 is considered to re¯ect some restructuring of the transmembrane domain. Given that TG has its own effect, namely a slight decrease in tryptophan ¯uorescence relative to the E2 conformation, Sagara et al.9 suggested that TG induces a unique conformation of Ca2‡-ATPase with corresponding global changes to the transmembrane architecture. An alternative explanation is that TG directly perturbs one or more nearby tryptophan residues without signi®cant global changes. Indeed, our structural results indicate that TG does not alter the basic shape of the Ca2‡-ATPase molecule in the E2 conformation. This can be

237

Thapsigargin-binding Site on Ca2 ‡-ATPase

Figure 4. Difference densities at 3s overlaid on the average maps from Figure 3. (a) Differences of DTG data set relative to native (yellow) and PLB (green) data sets. (b) Differences of TG data set relative to native (orange) and PLB (cyan) data sets. (c) All four difference maps are superimposed after turning the molecule 90  about the vertical axis (cytoplasmic nose pointing away from the reader). Note that the lower-resolution native structure gives rise to Ê resolnoisier difference maps. (d) DTG (white) and TG (cyan) differences relative to the older native data set at 14 A ution. (e) Negative densities from the four difference maps in (a) to (c). (f) Control differences between DTG-TG (gray) and PLB-native (magenta). (g) The representative distribution of densities from one of the difference maps; the density thresholds of 3s were taken relative to the mean, despite the fact that it was slightly different from zero.

contrasted with the drastic structural changes associated with the Ca2‡-induced transition from the E2 to the E1
Ca2 conformation, which involves major rearrangement of the three cytoplasmic domains (Figure 3). Furthermore, there is a tryptophan residue (Trp288) quite near this binding site in the M3/M4 loop with another (Trp272) at the bottom of M3; one of these residues could respond to TG binding, thus explaining its effects on intrinsic ¯uorescence. Given the presence of an intermolecular crystal contact in our tubular crystals at the lumenal end of M3, the added rigidity imparted to this loop by TG binding may explain its stabilization of these crystals.

Mechanism of TG inhibition Given the exceedingly high af®nity of TG, our proposed site is likely to prevent movement of these two lumenal loops relative to one another. Such movement may be required to achieve the cooperative binding of two calcium ions and the transition to the E1 state. Indeed, to optimize the ®t of the atomic coordinates from the E1
Ca2 state to the EM map in the E2 state, the M7/M8 loop was Ê closer to the M3/M4 loop. Theremoved 10 A fore, TG may be acting as a tether to prevent the conformational transition to the E1 state. Furthermore, since one full transport cycle can occur

238

Figure 5. (a) Fitting of TG and Ca2‡-ATPase coordinates to the DTG map. After rigid body docking of the atomic coordinates of the transmembrane domain from the E1
Ca2 state to the EM map in the E2 state, individual helices were manually adjusted to optimize the ®t (unpublished results). Similarly, the M7/M8 loop from the X-ray crystal coordinates (cyan in (b)) was moved Ê closer to the M3/M4 loop to ®t within the den10 A sity envelope. An atomic model for TG36 is shown at the bottom bridging the density between the M7/M8 loop on the left and the M3/M4 loop on the right (red). We have not attempted to model the side-chain interactions with TG, because the con®guration of these ¯exible loops is likely to change signi®cantly in response to TG binding. According to site-directed mutagenesis, a different group of residues in¯uence TG af®nity,19 the side-chains of which are shown at the top of M3 (red). (b) Structural changes to M3 and the two lumenal loops that are associated with the E2 to E1 transition and with determining TG af®nity. The upper panel is in the same orientation as (a), whereas the lower panel is rotated slightly to show the kink in M3.

before TG inhibition, the two loops may not come in close proximity until after Ca2‡ is released to the lumen. Thus, movement of these loops in a bellows-like fashion may be one of the key structural changes during transport. In addition to preventing Ca2‡ binding, TG prevents the backward reaction with inorganic

Thapsigargin-binding Site on Ca2 ‡-ATPase

phosphate9 and drastically reduces the af®nity of the E2 conformation for ATP,26 suggesting that TG effects more than just the lumenal and transmembrane domains. Such ATP binding by the E2 conformation is known to accelerate the transition to the E1 conformation and to affect the ion-binding properties of Ca2‡-ATPase27 and of Na‡,K‡ATPase. These ATP effects are certain to involve some sort of conformational change, which may well in¯uence the disposition of the lumenal loops. By resisting these conformational changes, TG could effectively lower the af®nity for ATP and other ligands in distant domains. In particular, the increased occupancy of decavanadate within the cytoplasmic domain may re¯ect these long-range effects of TG on the ATP-binding and phosphorylation domains. The effect of mutations in the S3 segment on TG af®nity implicates this region in relaying the structural changes between the distant binding sites. In the E1
Ca2 structure, this S3 segment is hydrogen bonded to both the M6/M7 loop and S4, and is certainly affected by the 90  rotation of the cytoplasmic nose (yellow domain in Figure 3(c) and (d)) during the E2 to E1 transition. The density for M3 and S3 are quite well de®ned in the cryoEM map of the E2 conformation and our ®tting of the atomic coordinates required the continuous M3/S3 helix to be bent near the cytoplasmic membrane surface. If S3 were acting as a lever arm to connect cytoplasmic and lumenal domains, then mutations that alter the local structure of S3 near this bend could in¯uence the lumenal domains, thereby indirectly affecting TG af®nity. Thus, TG inhibition represents a speci®c example of the structural coupling that generally characterizes Ca2‡-ATPase and, in fact, all P-type ATPases. Analogous changes are likely to coordinate ion-binding in the transmembrane domain with speci®c chemical changes at the phosphorylation site, thus driving the cycle of ATP-dependent ion transport.

Materials and Methods Crystallization SR membranes were prepared from the white, fasttwitch muscle of rabbit hind legs by the method of Meissner et al.28 Protein concentrations were determined by the Lowry assay and ATPase activities were measured by the coupled enzyme assay.29 Tubular crystals of Ca2‡-ATPase formed spontaneously after incubating SR membranes (1 mg/ml) for one to three days in a medium containing 100 mM KCl, 5 mM MgCl2, 0.5 mM EGTA, 0.5 mM decavanadate, 20 mM imidazole (pH 7.27.4).30 In some cases, 10 mM TG (LC Services Corp., Woburn MA) or 30-60 mM DTG (a gift from G. Inesi) was added during crystallization. Decavanadate was prepared from a stock solution of 50 mM Na3VO4 at 4  C ®rst by adding HCl to obtain pH 2.0 and then by neutralizing the pH with KOH immediately before dilution into the crystallization solution.

239

Thapsigargin-binding Site on Ca2 ‡-ATPase Electron microscopy Tubular crystals were negatively stained by applying 5 ml onto glow-discharged, carbon-coated grids and then by rapidly rinsing with three drops of ice-cold, 2 % (w/v) uranyl acetate. Frozen-hydrated specimens were prepared by diluting crystal samples 10-20-fold with excess crystallization buffer, and then by applying 5 ml to 400-mesh Cu grids that had been coated with perforated-carbon ®lms and glow-discharged in either amyl amine vapor or air. After blotting excess solution from the backside of grids, they were immediately plunged into an ethane slush and stored in liquid N2. Images of frozen-hydrated samples were obtained with a Philips (Eindhoven, Holland) CM200 FEG electron microscope equipped with an Oxford (Oxford, UK) CT-3500 specimen holder, which kept the specimen at ÿ185  C. Electron micrographs were recorded at 0.7-1.7 mm under-focus and at 50,000 nominal magni®cation, Ê 2 on Kodak SO163 with doses of 10-16 electrons/A ®lm, which were developed in full-strength D19 for 12 minutes.

Image processing Ê Images of straight tubes with a diameter of 550-600 A suspended over holes in the carbon substrate were selected by optical diffraction and digitized with either a PDS1010GM microdensitometer (Orbital Sciences Corp., Pomona CA) at 13mm intervals or a Zeiss SCAI microdensitometer (Carl Zeiss GMBH, Oberkochen, Germany) at 14 mm intervals. The helical symmetry was characterized by assigning Bessel orders (n) to the principal layerlines,31 i.e. those indexed as (1,0) and (0,1), as shown in Table 1. Established methods31,32 were used for determining the orientation of tubes, their repeat distance, the position of the 2-fold axis and for extracting and averaging layer-line data. The averaged data were then used as reference for the distortion corrections of Beroukhim & Unwin,33 which involve dividing individual tubes into Ê ) sections and determining independent short (1000 A orientational parameters for individual sections. During averaging, phase data were weighted by the contrast transfer function (CTF) and, after averaging, amplitude data were fully compensated assuming 4.6 % amplitude contrast34 before calculation of a 3D map. An initial estimate of image defocus was made from Fourier transforms either of the tubes themselves or from carbon ®lm adjacent to the tubes as described by Tani et al.,35 these defocus values were later re®ned by comparing individual images to the averaged data set.

Difference maps Pairwise difference maps were calculated between each of the various different structures in real space, because the helical symmetry of tubes varied. In most Ê resolution to match cases, data were truncated to 10 A Ê the maximal resolution from native tubes, though 14 A was used for the older native data set. After masking a single molecule in each map, magni®cation adjustment and spatial alignment were done by maximizing the cross-correlation coef®cient. Thereafter, the density scale factor was determined by linear regression of corresponding densities in 3D maps. After alignment and scaling, averaged 3D maps were simply subtracted, pixel by pixel.17

Acknowledgments The authors thank Giuseppe Inesi for the kind gift of dansyl-thapsigargin. We acknowledge Rameen Beroukhi and Nigel Unwin as well as Chikashi Toyoshima and Koji Yonekura for developing and sharing the computer programs we used for image analysis and determination of difference maps. This work was supported by NIH grants AR40997 and GM56960 to D.L.S. and a Scientist Development Grant from the American Heart Association, Heritage Af®liate to H.S.Y.

References 1. Axelsen, K. B. & Palmgren, M. G. (1998). Evolution of substrate speci®cities in the P-type ATPase superfamily. J. Mol. Evol. 46, 84-101. 2. Moller, J. V., Juul, B. & le Maire, M. (1996). Structural organization, ion transport, and energy transduction of ATPases. Biochim. Biophys. Acta, 1286, 1-51. 3. Zhang, P., Toyoshima, C., Yonekura, K., Green, N. M. & Stokes, D. L. (1998). Structure of the Ê calcium pump from sarcoplasmic reticulum at 8 A resolution. Nature, 392, 835-839. 4. Auer, M., Scarborough, G. A. & KuÈhlbrandt, W. (1998). Three-dimensional map of the plasma membrane H‡-ATPase in the open conformation. Nature, 392, 840-843. 5. Stokes, D. L., Auer, M., Zhang, P. & KuÈhlbrandt, W. (1999). Comparison of H ‡-ATPase and Ca2 ‡ATPase suggests that a large conformational change initiates P-type ion pump reaction cycles. Curr. Biol. 9, 672-679. 6. Toyoshima, C., Nakasako, M., Nomura, H. & Ogawa, H. (2000). Crystal structure of the calcium Ê resolution. pump of sarcoplasmic reticulum at 2.6 A Nature, 405, 647-655. 7. Rasmussen, U., Broogger Christensen, S. & Sandberg, F. (1978). Thapsigargine and thapsigargicine, two new histamine liberators from Thapsia garganica L. Acta Pharm. Suecica, 15, 133-140. 8. Sagara, Y. & Inesi, G. (1991). Inhibition of the sarcoplasmic reticulum Ca2‡ transport ATPase by thapsigargin at subnanomolar concentrations. J. Biol. Chem. 266, 13503-13506. 9. Sagara, Y., Wade, J. B. & Inesi, G. (1992). A conformational mechanism for formation of a dead-end complex by the sarcoplasmic reticulum ATPase with thapsigargin. J. Biol. Chem. 267, 1286-1292. 10. Stokes, D. L. & LacapeÁre, J.-J. (1994). Conformation of Ca2‡-ATPase in two crystal forms: effects of Ca2‡, thapsigargin, AMP-PCP, and Cr-ATP on crystallization. J. Biol. Chem. 269, 11606-11613. 11. Maurer, A. & Fleischer, S. (1984). Decavanadate is responsible for vanadate-induced two-dimensional crystals in sarcoplasmic reticulum. J. Bioenerg. Biomembr. 16, 491-505. 12. Stokes, D. L. & Green, N. M. (2000). Modeling a Ê density map for dehalogenase fold into the 8-A Ca2‡-ATPase de®nes a new domain structure. Biophys. J. 78, 1765-1776. 13. Hua, S., Malak, H., Lakowicz, J. R. & Inesi, G. (1995). Synthesis and interaction of ¯uorescent thapsigargin derivatives with the sarcoplasmic reticulum

240

14.

15. 16. 17.

18.

19.

20.

21.

22.

23.

24.

Thapsigargin-binding Site on Ca2 ‡-ATPase

ATPase membrane-bound region. Biochemistry, 34, 5137-5142. Young, H. S., Rigaud, J.-L., LacapeÁre, J.-J., Reddy, L. G. & Stokes, D. L. (1997). How to make tubular crystals by reconstitutioin of detergent-solubilized Ca2‡-ATPase. Biophys. J. 72, 2545-2558. Gruner, S. (1985). Intrinsic curvature hypothesis for biomembrane lipid composition: a role for nonbilayer lipids. Proc. Natl Acad. Sci. USA, 82, 3665-3669. Voss, J., Jones, L. R. & Thomas, D. D. (1994). The physical mechanism of calcium pump regulation in the heart. Biophys. J. 67, 190-196. Yonekura, K., Stokes, D. L., Sasabe, H. & Toyoshima, C. (1997). The ATP-binding site of Ca2 ‡-ATPase revealed by electron image analysis. Biophys. J. 72, 997-1005. Csermely, P., Varga, S. & Martonosi, A. (1985). Competition between decavanadate and ¯uorescein isothiocyanate on the Ca2‡-ATPase of sarcoplasmic reticulum. Eur. J. Biochem. 150, 455-460. Zhong, L. & Inesi, G. (1998). Role of the S3 stalk segment in the thapsigargin concentration dependence of sarco-endoplasmic reticulum Ca2‡ATPase inhibition. J. Biol. Chem. 273, 12994-12998. Hua, S. & Inesi, G. (1997). Synthesis of a radioactive azido derivative of thapsigargin and photolabeling of the sarcoplasmic reticulum ATPase. Biochemistry, 36, 11865-11872. Norregaard, A., Vilsen, B. & Andersen, J. P. (1994). Transmembrane segment M3 is essential to thapsigargin sensitivity of the sarcoplasmic reticulum Ca2‡-ATPase. J. Biol. Chem. 269, 26598-26601. Yu, M., Zhang, L., Rishi, A. K., Khadeer, M., Inesi, G. & Hussain, A. (1998). Speci®c substitutions at amino acid 256 of the sarcoplasmic/endoplasmic reticulum Ca2‡ transport ATPase mediate resistance to thapsigargin in thapsigargin-resistant hamster cells. J, Biol. Chem. 273, 3542-3546. Ma, H., Zhong, L., Inesi, G., Fortea, I., Soler, F. & Fernandez-Belda, F. (1999). Overlapping effects of S3 stalk segment mutations on the af®nity of Ca2 ‡ATPase (SERCA) for thapsigargin and cyclopiazonic acid. Biochemistry, 38, 15522-15527. Bigelow, D. J. & Inesi, G. (1992). Contributions of chemical derivatization and spectroscopic studies to the characterization of the Ca2‡ transport ATPase of sarcoplasmic reticulum. Biochim. Biophys. Acta, 1113, 323-338.

25. Dupont, Y. (1976). Fluorescence studies of the sarcoplasmic reticulum calcium pump. Biochem. Biophys. Res. Commun. 71, 544-550. 26. DeJesus, F., Girardet, J. L. & Dupont, Y. (1993). Characterisation of ATP binding inhibition to the sarcoplasmic reticulum Ca2 ‡-ATPase by thapsigargin. Fed. Eur. Biochem. Soc. Letters, 332, 229-232. 27. Stahl, N. & Jencks, W. P. (1984). Adenosine 50 -triphosphate at the active site accelerates binding of calcium to calcium adenosinetriphosphatase. Biochemistry, 23, 5389-5392. 28. Meissner, G., Conner, G. E. & Fleischer, S. (1973). Isolation of sarcoplasmic reticulum by zonal centrifugation and puri®cation of Ca2‡-pump and Ca2‡binding proteins. Biochim. Biophys. Acta, 298, 246269. 29. Warren, G. B., Toon, P. A., Birdsall, N. J. M., Lee, A. G. & Metcalfe, J. C. (1974). Reconstitution of a calcium pump using de®ned membrane components. Proc. Natl Acad. Sci. USA, 71, 622-626. 30. Dux, L. & Martonosi, A. (1983). Two-dimensional arrays of proteins in sarcoplasmic reticulum and puri®ed Ca2‡-ATPase vesicles treated with vanadate. J. Biol. Chem. 258, 2599-2603. 31. Toyoshima, C. & Unwin, N. (1990). Three-dimensional structure of the acetylcholine receptor by cryoelectron microscopy and helical image reconstruction. J. Cell. Biol. 111, 2623-2635. 32. DeRosier, D. J. & Moore, P. B. (1970). Reconstruction of three-dimensional images from electron micrographs of structures with helical symmetry. J. Mol. Biol. 52, 355-369. 33. Beroukhim, R. & Unwin, N. (1997). Distortion correction of tubular crystals: improvements in the acetylcholine receptor structure. Ultramicroscopy, 70, 57-81. 34. Toyoshima, C., Yonekura, K. & Sasabe, H. (1993). Contrast transfer for frozen-hydrated specimens. II. Amplitude contrast at very low frequencies. Ultramicroscopy, 48, 165-176. 35. Tani, K., Sasabe, H. & Toyoshima, C. (1997). A set of computer programs for determining defocus and astigmatism in electron images. Ultramicroscopy, 65, 31-44. 36. Christensen, S. B., Larsen, I. K. & Rasmussen, U. (1982). Thapsigargin and thapsigargicin, two histamine liberating sesquiterpene lactones from Thapsia garganica. X-ray analysis of the 7,11-epoxide of thapsigargin. J. Org. Chem. 47, 649-652.

Edited by W. Baumeister (Received 1 December 2000; received in revised form 19 February 2001; accepted 19 February 2001)