inhibitor aurovertin B

Proc. Natl. Acad. Sci. USA Vol. 93, pp. 6913-6917, July 1996 Biochemistry

The structure of bovine Fl-ATPase complexed with the antibiotic inhibitor aurovertin B (crystal structure/ATP synthesis/ATP hydrolysis)

MARK J. vAN RAAIJ, JAN PIETER ABRAHAMS, ANDREW G. W. LESLIE, AND JOHN E. WALKER* Medical Research Council Laboratory of Molecular Biology, Hills Road, Cambridge CB2 2QH, United Kingdom

Communicated by Paul D. Boyer, University of California, Los Angeles, CA, March 20, 1996 (received for review February 12, 1996)

In the structure of bovine mitochondrial ABSTRACT Fl-ATPase that was previously determined with crystals grown in the presence of adenylyl-imidodiphosphate (AMPPNP) and ADP, the three catalytic a-subunits have different conformations and nucleotide occupancies. Adenylylimidodiphosphate is bound to one (8-subunit (j3p), ADP is bound to the second (8DP), and no nucleotide is bound to the third (PE). Here we show that the uncompetitive inhibitor aurovertin B binds to bovine F1 at two equivalent sites in 13rp and PE, in a cleft between the nucleotide binding and C terminal domains. In IDP, the aurovertin B pocket is incomplete and is inaccessible to the inhibitor. The aurovertin B bound to I3TP interacts with a-Glu399 in the adjacent aep subunit, whereas the aurovertin B bound to PE iS too distant from rE to make an equivalent interaction. Both sites encompass fBArg-412, which was shown by mutational studies to be involved in binding aurovertin. Except for minor changes around the aurovertin pockets, the structure of bovine FlATPase is the same as determined previously. Aurovertin B appears to act by preventing closure of the catalytic interfaces, which is essential for a catalytic mechanism involving cyclic interconversion of catalytic sites.

19, 18

Et/

1,2

FIG. 1. Structure of the aurovertin B. The molecule consists of a (Left) substituted pyrone ring linked by a rigid spacer containing conjugated double bonds to a (Right) substituted dioxabicyclo[3,2,1] octane, or aglycone, ring. The carbon atoms are numbered. The numbering of oxygen atoms follows that of the carbon atom to which they are attached. Where oxygen atoms bridge two carbons, the lower number is used.

Insensitive ATPases and their isolated 13-subunits do not bind aurovertin (16). In the atomic structure of F1-ATPase from bovine mitochondria (17), determined with crystals grown in the presence of adenylyl-imidodiphosphate (AMP-PNP) and ADP, the three chemically identical 1-subunits have different conformations and different occupancies by nucleotides. AMP-PNP is bound to one, ADP to the second, and no nucleotide at all is bound to the third ,B-subunit, despite the presence of excess nucleotides in the mother liquor. These three conformations are referred to as Prp, ,BDP, and ,BE, respectively. In the structure, the three (3-subunits and the three noncatalytic a-subunits are arranged alternately like six segments of an orange around the central anti-parallel coiled-coil of a-helices in the 'y-subunit. During the catalytic cycle, the three conformations of }3-subunits could be interconverted by the relative rotation of this central coiled-coil. Therefore, this structure supports a binding change mechanism that proposed a cyclic interconversion of three different catalytic sites during the

The aurovertins are a family of related antibiotics from the fungus Calcarisporium arbuscula (1). They inhibit oxidative phosphorylation in mitochondria (2) and in many bacterial species (3). Aurovertins B (Fig. 1) and D have identical biological properties and are more potent than aurovertin A. Citreoviridin (4, 5) (from Penicillium citreoviride and Aspergillus terreus) and asteltoxin (6, 7) (produced by Aspergillus stellatus) are metabolites with related structures and similar properties to aurovertins. The aurovertins inhibit the protonpumping F,Fo-ATP synthase by binding to (3-subunits in its F1 catalytic sector (8). This is a globular domain that can be released as a water-soluble ATP hydrolase by disruption of a slender stalk that joins it to the membrane domain of ATP synthase. In the membrane-bound enzyme, both ATP hydrolysis and synthesis are inhibited, as is ATP hydrolysis by isolated F1-ATPase (1). The inhibition by aurovertins is uncompetitive with nucleotides (9). Three (-subunits are present in the ATP synthase complex; each (3-subunit contains a nucleotide binding site that is directly involved in catalysis (10). The number of aurovertin binding sites per F1 assembly is uncertain, but there is at least one high affinity binding site (Kd -1 ,uM), and either one or two additional sites with lower affinities (Kd -4-6 ,uM) (8, 11). One site has low affinity in the presence of ADP but high affinity in the presence of ATP (11). Through the analysis of aurovertin-resistant mutants in Escherichia coli, (3-Arg398 (equivalent to bovine (3-Arg412) has been implicated in aurovertin binding (12, 13). In some other microorganisms that are naturally insensitive to aurovertins, this residue is also substituted by other amino acids (3, 14, 15).

enzyme's catalytic cycle (18). The sites at which aurovertin B binds to bovine F1-ATPase have now been determined by x-ray analysis of crystals soaked in the inhibitor.

MATERIALS AND METHODS Crystallization and Data Collection. Crystals of bovine heart F1-ATPase were grown by microdialysis in the presence of AMP-PNP and ADP (19). An ethanolic solution of aurovertin B (2 mM; Sigma) was then added to the solution outside the dialysis membrane to give a final concentration of 20 ,uM and 1% (vol/vol) ethanol. Immediately before data collection Abbreviations: AMP-PNP, adenylyl-imidodiphosphate; rms, rootmean-square. Data deposition: The atomic coordinates and structure factors have been deposited in the Protein Data Bank, Chemistry Department, Brookhaven National Laboratory, Upton, NY 11973 (reference

The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact.

RlCOWSF). *To whom reprint requests should be addressed.

6913

6914

Biochemistry: van Raaij et al.

at the Synochrotron Radiation Source (Daresbury, U. K.), the glycerol concentration was increased by dialysis from 0 to 20% (wt/vol) in steps of 5%, with at least 2 h of dialysis between steps. One crystal, transferred to a similar solution but containing 25% (wt/vol) glycerol, was frozen at 100 K, and data were collected to 3.1 A resolution (1 A = 0.1 nm). Reflections were integrated with MOSFLM (20) and processed further with programs from the Collaborative Computational Project Number 4 suite (21). Structure Solution and Refinement. The structure of the bovine F1-ATPase-aurovertin B complex was solved by difference Fourier analysis using the calculated amplitudes and phases from the bovine F1-ATPase coordinates (17). The starting model had an R factor of 17.2% and a free R factor (22) of 25.4% with respect to the F1-ATPase data, and an R factor of 34.9% and free R factor of 34.8% with respect to the data obtained from the aurovertin-F1 complex. It was subjected to rigid body refinement using the computer program TNT (23) and all data from 20 to 3.1 A. In this and in all subsequent refinement steps, 2% of the data were set aside for calculation of the free R factor, and refinement procedures were chosen so that the R factor decreased and the free R factor either decreased or remained constant. Two aurovertin molecules, one in PTP and the other in PE, were built into the structural model with the graphics program o (24), using bond lengths and angles from asteltoxin (6) and 3,6-anhydro-a-Dgalactoside (25). The pyrone ring and the spacer of conjugated double bonds were kept planar (6, 26). The shape of the positive difference density in both sites agreed with the absolute stereochemistry of aurovertin B (27). A few regions, particularly around the aurovertin binding sites, were rebuilt with o. The positions of 603 water molecules in the starting model were refined by real-space refinement using TNT. Those water molecules that were too close either to the aurovertins or to other parts of the model were removed, as were those incapable of making potential hydrogen bonds. Where appropriate, additional water molecules were introduced into the model. The temperature factors of all water molecules were refined, and those with temperature factors greater than 100 A2 were removed. The positions of individual atoms were refined with TNT, restraining the coordinates of the model to those of F1-ATPase (for details, see ref. 28). The individual temperature factors of the atoms and the occupancies of the aurovertins were refined in a separate run, in which temperature factors were restricted to less than 100 A2. The refined occupancies for the aurovertin molecules in ITP and PE were 0.8 and 0.7, respectively.

RESULTS AND DISCUSSION Quality of the Data and the Model. The final model has good stereochemistry, and 86.5% of the residues have main chain torsion angles within the most favored regions of the Ramachandran plot (29), and none is in the disallowed regions. In the resolution range between 5.5 and 3.1 A, the R factors are significantly lower than for all the data (R factor 21.0%, free R factor 25.6%; Table 1), which presumably reflects some inadequacy in the correction for bulk solvent scattering applied in TNT (23). The Aurovertin B Binding Sites. After initial rigid body refinement, only two regions of significant positive density were present in a difference electron density map, each of them with smaller negative features nearby (Fig. 2 A and B). The stronger positive density was in subunit PTP and the weaker was in the equivalent region of subunit IE. In the Fl-ATPase structure and in the Fl-aurovertin complex, the thermal mobility of the region involved in aurovertin binding is significantly higher in PE than in the equivalent regions of fTp and I3DP, and this may explain the relative weakness of the positive electron density difference associated with this site. There

Proc. Natl. Acad. Sci. USA 93

(1996)

Table 1. Crystallographic parameters Spacegroup P212121 283.4 x 107.6 x 140.2 A3 Unit cell 20.0-3.1 A Resolution 76,643 Reflections 0.8% Rejected measurements* 0.075 (0.23) Rmerget 0.976 (0.94) Completeness 3.0 (2.6) Multiplicity 43.4 (12.9) Mean (F/or(F)) 22,883 Fl-ATPase atomst 634 Water molecules* 66 Aurovertin B atomst 0.230 (0.29) R factors§ 0.286 (0.33) Free R factorl 0.013 A rms deviation bonds 2.40 rms deviation angles Values in parentheses are for the highest resolution bin. *Measurements with intensities differing more than 3.5orI) from the weighted mean were rejected. X2i I(h)i, where I(h) is the mean intensity tRmerge = E EiII(h) after rejections. The contributing reflections were weighted by their standard deviations, which were determined by adjusting the measured standard deviation to reflect the observed differences between symmetry related reflections. tHydrogen atoms were excluded. §The R factor is defined as Eh Fo - FCI/Xh Fo where Fo and F, are the observed and calculated structure factor amplitudes, respectively, and was determined using 98% of the data. IThe free R factor (22) was determined from the residual 2% of the data.

I(h)il/2

were no changes in electron density associated with aurovertin binding in subunit IDP. After superposition of the refined Fl.aurovertin B model on the F1 model (17), the root-meansquare (rms) difference in Ca atom positions was calculated to be 0.27 A. The values obtained by superimposing individual subunits from the two structures were 0.14, 0.18, and 0.16 A for aTP, aE, and aDP, respectively, 0.19, 0.22, and 0.18 A for ITP, PE, and IDP, respectively, and 0.37 A for the y-subunit. The absence of any features in the original difference density map other than those attributed to aurovertin binding, together with the low rms positional differences, suggest that the overall structure of F1-ATPase is essentially unchanged by binding aurovertin. However, with the exception of the y subunit (which has relatively high thermal parameters), the rms differences for the individual subunits are significantly smaller than for the whole enzyme, suggesting that their relative positions have changed slightly in the F1-aurovertin complex. These changes could have been caused either by aurovertin binding to F1-ATPase in the crystals, or by the freezing of the crystal of the F1-aurovertin complex before data collection. The aurovertin B binding sites in the ITP and PE subunits lie in clefts between the nucleotide binding and C terminal domains (Fig. 3), but without overlap between nucleotide and aurovertin B binding sites, which is consistent with the observed uncompetitive inhibition (9). In PTP, the adenine to pyrone ring distance and the adenine to ,B-Arg412 distance was 13 and 17 A, respectively. In E. coli FI-ATPase, the distance between lin-benzo-ADP (lin-benzoadenine is 8-aminoimidazo[4,5-g]-quinazoline) bound to the catalytic site and a tryptophan residue substituted for f3-Arg398 was estimated by fluorescence energy transfer to be 23 A (30). In the structure of Fl-ATPase, subunit PE has an open conformation in which parts of its nucleotide binding domain (residues 129-178 and 330-363) and the entire C-terminal domain (residues 364474) have rotated away from the sixfold axis of pseudosymmetry, relative to subunits ITP and /DP (17). This displacement is thought to arise from the interaction of the central a-helical coiled-coil in the y-subunit with the C-terminal domain of 1E.


Proc. Natl. Acad. Sci. USA 93 (1996)

6915

FIG. 2. Aurovertin B binding sites in bovine F1-ATPase. The two aurovertin B molecules are sandwiched between the nucleotide binding and the C-terminal domains of subunits f3Tp and ,BE. (A and B) The difference electron density in subunits P3p and 13E, respectively, following the initial rigid body refinement. Green and red represent positive and negative electron density, respectively, at a contour level of 2.5 Co. (C-F) Two orthogonal views of each site, with the C-terminal domain at the top and the nucleotide binding domain at the bottom. The final (2Fo - Fc) electron density map (contoured at 1 a) is shown in blue. (C and D) Aurovertin B bound to I3TP. (E and F) Aurovertin B bound to g9E. (G and H) The equivalent site in IDP, where aurovertin B is not bound.

Nonetheless, aurovertin B is bound to ,BE and its binding site is not compromised by this structural difference. The three a-13 pairs differ in their catalytic interfaces (17). In the region of the aurovertin binding site, the interface between aDp and ,BDP iS tightly packed, the one between aE and P3E is wide open, and the interface between aTp and J3TP iS partially open (Fig. 4). In both 13E and 13TP, aurovertin binds with the rigid spacer of conjugated double bonds lying almost parallel to the interface with the adjacent a-subunits, but at a slight inclination so that the pyrone ring projects into the interface region. The pyrone ring is close to aTp at the aTP-I3TP interface, but in the aE-13E interface the subunits are too far apart to allow an equivalent interaction between aE and the pyrone ring projecting from PE

Amino Acids that Interact with Aurovertin B. The interactions between aurovertin B and amino acids in the nucleotide binding domain of ,3E and ,BTP are mainly hydrophobic (Fig. 2 C-F). In both subunits, atom C8 of the aurovertin makes a van der Waals interaction with ,B-Leu342, atoms Cl1 and C12 interact with the side-chain of /3-11e344, atom C21 forms a hydrophobic contact with C,y and C8 of j3-Pro350, and atom C24 is close to the side-chain of 13-Leu351. There are also two potential hydrogen bonds to the C-terminal domain of the ,3-subunit between NE of 1-Gln411 and 025 of aurovertin B, and between NE of 3-Arg412 and 019. In addition, the pyrone ring makes a staggered stacking interaction with aromatic ring of P3-Tyr458. The 017 of the aurovertin B bound to I3TP iS in van der Waals contact with

6916


Proc. Natl. Acad. Sci. USA 93

(1996)

A nb

Ct

B

FIG. 3. Stereograms of ribbon diagrams of the (B-subunits of bovine Fl-ATPase with bound aurovertin B. InA and B, respectively, subunits TP and ,BE are shown in white, each with an associated coiled-coil of a-helices from the y-subunit (in black). An AMP-PNP molecule is bound to the nucleotide binding site of subunit rrp, and aurovertin B is bound between the nucleotide binding and C-terminal a-helical domains (indicated as nb and Ct, respectively). Both AMP-PNP and aurovertin are shown in ball-and-stick representation. No nucleotide is bound to ,BE, but a molecule of aurovertin B is bound in the same position as in f3rp.

0°2 of aTp-Glu399. This interaction is not present in the aurovertin binding site of IJE. The involvement of j-Arg412 in binding aurovertin B was anticipated from analysis of aurovertin resistant mutants of the E. coli F1-ATPase where the equivalent arginine had been replaced by histidine, cysteine, or tryptophan (12, 13). These substitutions probably result in local conformational changes that reduce the affinity for the antibiotic. The Arg to Cys substitution would prevent possible hydrogen bond formation with 019 of aurovertin. Similar explanations can be invoked in the naturally aurovertin insensitive Bacillus firmus OF4 (14) and Bacillus PS3 (3), where 13-Arg412 is replaced by phenylalanine, and in Mycoplasma gallisepticum (15), where it is substituted by asparagine. In Bacillus PS3 and M. gallisepticum, P3-Tyr458 is also changed to arginine and phenylalanine, respectively. Therefore, at least in Bacillus PS3, the stacking interaction with the pyrone cannot form. In M. gallisepticum, ,B-Gln411 is replaced by arginine (15), which would be expected to modify the interaction between the protein and 025 of aurovertin. Whether this would weaken or strengthen aurovertin binding is unclear. Although aurovertin B is not bound to I3DP, the binding sites for the aglycone ring and the attached spacer appear to be conserved (Fig. 2G). However, the interface with the adjacent aDP subunit is much tighter, and in consequence aDp-Gln396

is in van der Waals contact with 13Dp-Tyr458, effectively blocking the binding site for the pyrone ring (Fig. 2H). Fluorescence of Bound Aurovertin B. Aurovertins have been used as fluorescent probes of conformational changes during catalysis in Fl-ATPase (31). They fluoresce weakly in aqueous solution, presumably because the energy of the excited state is dissipated by surrounding water molecules. When aurovertins are bound to F1-ATPase, this dissipation is prevented and fluorescence occurs. The high thermal mobility of the residues involved in aurovertin binding in ,3E makes a detailed comparison with the site on /rp difficult. However, it is probable that the two sites are not structurally identical. In particular, the aurovertin bound to Tp is involved in a unique contact between the pyrone ring and the carboxylate group of aiTpGlu399. Thus, the fluorescence of a bound aurovertin would be expected, as observed (31), to change according to the conformation of the 13-subunit to which it is bound, and to depend on the amount and kind of bound nucleotide. Mechanism of Inhibition. It is clear from the F1-aurovertin structure that aurovertin cannot bind at the catalytic aDP-I3DP interface. If, as is thought, the crystal structure of bovine F1-ATPase represents the ADP-inhibited state of the enzyme (17), the catalytic site in (3DP occupied by ADP, can be considered to be equivalent to a"tight"-site in Boyer's binding change mechanism (18). Similarly, the catalytic site in TpP


FIG. 4. Locations of the aurovertin B binding sites in bovine F1-ATPase. The assembly is viewed from above (toward the membrane domain of intact F,Fo-ATPase) showing the central a-helical coiled-coil structure in the y-subunit and the C-terminal domains of a-subunits (residues 380-510) and (3-subunits (residues 364-474) arranged in alternation around the coiled-coil. Aurovertin B molecules are bound to subunits 1Tp and PE. In ITP, the pyrone ring is shown to interact with aTp, whereas in PE it is unable to make an equivalent interaction with aE, which is too far away. In I3DP, the aglycone pocket is present, but aurovertin B cannot bind to this subunit because of the closed interface between IDP and aDp.

(occupied in the crystals by AMP-PNP) is equivalent to a "loose" site, and the site in PE is an "open" site. In the active enzyme, the "tight" site would contain ATP and the "loose" site would contain ADP and phosphate; energy would be required for the conversion from "tight" to "open" with release of ATP and binding of ADP and phosphate. The hydrolytic cycle is assumed to be the reverse of the synthetic cycle, and therefore the "tight" site participates in both. The structure of the F1-aurovertin B complex suggests that during ATP synthesis, aurovertin could act by inhibiting the conversion of the "loose" site to the "tight" site. Conversely, during ATP hydrolysis, the aurovertin would inhibit the conversion from the empty "open" site to a "tight site." The inhibition of ATP hydrolysis driven by NAD+ reduction through succinate needs higher concentrations of aurovertin than inhibition of the reverse reaction, ATP synthesis driven by NADH oxidation (2, 32-35). The structure of the F1aurovertin B complex suggests a possible explanation. If the "loose" site (1Tp) has a higher affinity for aurovertin B than the "empty" site (PE), inhibition of ATP synthesis would be expected to be more effective than inhibition of ATP hydrolysis. We thank M. Montgomery for help with data collection and R. Henderson and P. R. Evans for their comments on the manuscript. M.J.v.R. was supported by a Medical Research Council Research

Studentship. 1. Linnett, P. E. & Beechey, R. B. (1979) Methods Enzymol. 55, 472-518.

Proc. Natl. Acad. Sci. USA 93 (1996)

6917

2. Lardy, H. A., Connelly, J. L. & Johnson, D. (1964) Biochemistry 3, 1961-1968. 3. Saishu, T., Kagawa, Y. & Shimizu, R. (1983) Biochem. Biophys. Res. Commun. 112, 822-826. 4. Linnett, P. E., Mitchell, A. D., Osselton, M. D., Mulheirn, L. J. & Beechey, R. B. (1978) Biochem. J. 170, 503-510. 5. Franck, B. & Gehrken, H.-P. (1980)Angew. Chem. Int. Ed. Engl. 19, 461-462. 6. Kruger, G. J., Steyn, P. S., Vleggaar, R. & Rabie, C. J. (1979) J. Chem. Soc. Chem. Commun. 441-442. 7. Satre, M. (1981) Biochem. Biophys. Res. Commun. 100, 267-274. 8. Verschoor, G. J., van der Sluis, P. R. & Slater, E. C. (1977) Biochim. Biophys. Acta 462, 438-449. 9. Issartel, J. P. & Vignais, P. V. (1984) Biochemistry 23,6591-6595. 10. Cross, R. L. (1981) Annu. Rev. Biochem. 50, 681-714. 11. Issartel, J. P., Klein, G., Satre, M. & Vignais, P. V. (1983) Biochemistry 22, 3492-3497. 12. Lee, R. S.-F., Pagan, J., Satre, M., Vignais, P. V. & Senior, A. E. (1989) FEBS Lett. 253, 269-272. 13. Lee, R. S.-F., Pagan, J., Wilke-Mounts, S. & Senior, A. E. (1991) Biochemistry 30, 6842-6847. 14. Hicks, D. B. & Krulwich, T. A. (1990) J. Bio. Chem. 265, 20547-20554. 15. Rasmussen, 0. F., Shirvan, M. H., Margalit, H., Christiansen, C. & Rottem, S. (1992) Biochem. J. 285, 881-888. 16. Douglas, M. G., Koh, Y., Dockter, M. E. & Schatz, G. (1977) J. Biol. Chem. 252, 8333-8335. 17. Abrahams, J. P., Leslie, A. G. W., Lutter, R. & Walker, J. E. (1994) Nature (London) 370, 621-628. 18. Boyer, P. D. (1993) Biochim. Biophys. Acta 1140, 215-250. 19. Lutter, R., Abrahams, J. P., van Raaij, M. J., Todd, R. J., Lundqvist, T., Buchanan, S. K., Leslie, A. G. W. & Walker, J. E. (1993) J. Mol. Biol. 229, 787-790. 20. Leslie, A. G. W. (1992) Joint CCP4 and ESF-EACMB Newsletter on Protein Crystallography (Daresbury Laboratory, Warrington, U. K.). 21. Collaborative Computational Project Number 4 (1994) Acta Crystallogr. D 50, 760-763. 22. Brunger, A. T. (1992) Nature (London) 335, 472-475. 23. Tronrud, D. E., Ten Eyck, L. F. & Matthews, B. W. (1987) Acta Crystallogr. A 43, 489-501. 24. Jones, T. A., Zou, J.-Y., Cow-an, S. W. & Kjeldgaard, M. (1991) Acta Crystallogr. A 47, 110-119. 25. Campbell, J. W. & Harding, M. M. (1972) J. Chem. Soc. Perkin Trans. 2, 1721-1723. 26. Norrestam, R. (1978) Acta Crystallogr. A 34, S79. 27. Nishiyama, S., Toshima, H., Kanai, H. & Yamamura, S. (1988) Tetrahedron 44, 6315-6324. 28. Abrahams, J. P. (1996) Macromolecular Refinement: Proceedings of the 1996 CCP4 Study Weekend (Daresbury Laboratory, Warrington, U. K.), in press. 29. Laskowski, R. A., MacArthur, M. W., Moss, D. S. & Thornton, J. M. (1993) J. Appl. Crystallogr. 26, 283-291. 30. Weber, J., Lee, R. S. F., Grell, E. & Senior, A. E. (1992) Biochemistry 31, 5112-5116. 31. Chang, T.-M. & Penefsky, H. S. (1973) J. Biol. Chem. 248, 2746-2754. 32. Lee, C.-P. & Ernster, L. (1968) Eur. J. Biochem. 3, 391-400. 33. Lenaz, G. (1965) Biochem. Biophys. Res. Commun. 21, 170-175. 34. Bertina, R. M., Schrier, P. I. & Slater, E. C. (1973) Biochim. Biophys. Acta 305, 503-518. 35. Roberton, A. M., Holloway, C. T., Knight, I. G. & Beechey, R. B. (1968) Biochem. J. 108, 445-456.