Eur. J. Biochem. 209, 597-602 (1992) 0FEBS 1992

X-ray crystallographic characterisation of type-2-depleted ascorbate oxidase from zucchini Albrecht MESSERSCHMIDT ', Wolfgang STEIGEMANN ', Robert HUBER I, Gunter LANG' and Peter M. H. KRONECK3

' Max-Planck-Institut fur Biochemie, Martinsried, Federal Republic of Germany Boehringer Mannheim GmbH, Penzberg, Federal Republic of Germany Fakultat fur Biologie, Universitat Konstanz, Federal Republic of Germany (Received May 25/July 31, 1992) - EJB 92 0726

The type-2 depleted form of ascorbate oxidase from zucchini has been prepared in crystals and characterised by X-ray crystallography and EPR spectroscopy. The X-ray structure analysis by difference-Fourier techniques and refinement shows that, on average, about 1.3 Cu atomslascorbate oxidase monomer are removed. The copper is lost from the trinuclear site whereby the EPR-active type-2 copper is depleted most; type-1 copper is not affected. This observation indicates preferential formation of a 1 Cu-depleted form with the hole equally distributed over all three copper sites. Each of these 1 Cu-depleted species may represent an anti-ferromagnetically coupled copper pair which is EPR-silent and could explain the disappearance of the type-2 EPR signal.

Ascorbate oxidase is a blue multicopper oxidase that catalyses the four-electron reduction of dioxygen to water with concomitant one-electron oxidation of the organic substrate [1]. Copper-dependent ascorbate oxidase is only found in higher plants [2]. The enzyme from Cucurbita pep0 medullosa (green zucchini) is a dimer of 140 kDa containing eight copper ions of three different spectroscopic forms classified as type-1, type-2 and type-3 [3]. Primary structures of ascorbate oxidase from cucumber [4] and pumpkin [5] have recently been reported. The 0.25-nm X-ray structure of the oxidised form of ascorbate oxidase from zucchini showed the polypeptide fold and the coordination of the mononuclear blue copper site; in addition, an unprecedented trinuclear copper site has been discovered consisting of three Cu atoms within 0.37 nm of each other [6]. The structure has now been refined to 0.19 nm and its detailed description and implications for the catalytic mechanism have been the subject of a further publication [7]. The structural relationship to the other blue copper oxidases, laccase and ceruloplasmin, has been demonstrated by amino acid sequence alignment based on the spatial structure of ascorbate oxidase from zucchini [8]. All canonical copper ligands are strictly conserved with the exception of the methio-

nine ligand of type-1 copper which is replaced by a leucine in domain 2 of human ceruloplasmin [9] and in laccase from Neurospora crassa [lo]. Recently, it has been shown by a recombinant Met121 +Leu mutant of azurin from Pseudomonas aeruginosa that the methionine is not an essential constituent of the type-1 copper site to generate the distinct properties of this metal site [ll]. In ascorbate oxidase the non-blue EPR-active type-2 copper is an integral part of the trinuclear copper site [6, 71. This trinuclear cluster has eight histidine ligands symmetrically supplied from domains 1 and 3. It may be subdivided into a pair of copper atoms (designated CU2 and CU3) with six histidine ligands whose ligating N atoms (five NE2 atoms and one ND1 atom) are arranged in a trigonal prismatic manner. This pair may represent the type-3 copper. The remaining copper has two histidine ligands and is the putative EPRdetectable type-2 copper (CU4). Two oxygen atoms are bound to the trinuclear species. The first one bridges the putative type-3 copper pair and is either a OH- or 0'-. The second one is bound to the putative type-2 copper trans to the copper pair as OH- or HzO (Figs 1,2). The type-2 copper can be removed selectively. Several methods have been described for laccase from the fungus Polyporus versicolor [13, 141, and from the Japanese lacquer Correspondence tu A. Messerschmidt, Max-Planck-Institut fur tree Rhus vernicifera [15- 181, and for ascorbate oxidase from Biochemie, Am Klopferspitz I8 A, W-8033 Martinsried, Federal Re- zucchini [18-201. All procedures except one work under republic of Germany. ducing conditions with metal-chelating reagents, such as Fax. + 49-89-8578-3516. EDTA, dimethyl glyoxime, bathocuproine disulfonate or Abbreviations. T2D, type-2 depleted; FO,observed structure nitrilotriacetate. Reaction of N,N-diethyldithiocarbamate factor amplitude; FC, calculated structure factor amplitude; g,,, parallel component of g-tensor; g,, perpendicular component of g-tensor; with ascorbate oxidase under aerobic conditions in solution All, parallel component of the electron-nuclear hyperfine tensor; A,, gave the type-2-depleted (T2D) enzyme [20]. Many experiments were carried out on T2D multi-copper oxidases in perpendicular component of the electron-nuclear hyperfine tensor. Enzymes. Ascorbate oxidase (EC 1.10.3.3.); laccase (EC 1.10.3.2); the past and for the interpretation of these experiments it is ceruloplasmin (EC 2.16.3.1). important to know the actual occupation of the copper sites

598

INSERT \A552 PRO

SER

Fig. 1. Schematic structure of the ascorbate oxidase monomer. The figure was produced by the program RIBBON [12]. Insert: trinuclear copper site plus ligands.

\ TB

f

*

Fig. 2. Averaged FOTZD2 - FCTZD2difference electron density map plus atomic model around the trinuclear copper site. Contour levels: - 18.0 solid line, 18.0 dashed line.

in the depleted enzyme. An X-ray structure analysis of the depleted enzyme would provide reliable information about this. In this paper, we report on the preparation of T2D ascorbate oxidase in crystals, its characterisation by EPR spectroscopy and X-ray structure analysis. EXPERIMENTAL PROCEDURES Preparation of crystals of T2D ascorbate oxidase

Crystals of native ascorbate oxidase were prepared as described [21]. The protein material was obtained from Boehringer Mannheim GmbH. The crystals were grown by the microdialysis technique. The protein (15 mg/ml) in 50 mM sodium phosphate pH 5.5 was dialysed against 50 mM sodium phosphate pH 5.5 containing 20% (by vol.) 2-methyl-2,4-pentane-diol and 4% (by vol.) ethanol. The crystals grew in one week; they were transferred into a 25% (by vol.) 2-methyl-2,4pentane-diol solution buffered with 50 mM sodium phosphate pH 5.5. The crystals were anaerobically dialysed in microcells against 50 mM sodium phosphate pH 5.2 containing 25% (by vol.) 2-methyl-2,4-pentane-diol,1 mM EDTA, 2 mM dimethyl glyoxime, and 5 mM ferrocyanide for 7 h and 14 h.

Thereafter, crystals were brought back to the 25% 2-methyl2,4-pentane-diol solution buffered with 50 mM sodium phosphate pH 5.5. This procedure is based on the method of Avigliano et al. [19] to prepare T2D ascorbate oxidase in solution and was modified by Merli et al. [22] for use with ascorbate oxidase crystals. Data collection

Crystallographic intensity data were collected on a Micro Vax I11 controlled FAST television area detector diffractometer (Enraf-Nonius, Delft, Netherlands) at 10"C. The crystal was cooled by a stream of cold air to 4°C. CuKa radiation from a rotating anode generator (Rigaku, Tokyo, Japan), apparent focal spot size 0.3 mm x 0.3 mm, 5.4 kW was used. The detector-to-crystal distance was 49.2 mm, the detector tilt angle was 0" for all measurements. Data were recorded in frames of 0.1 O with 80-s exposure time and evaluated on-line using the program MADNES [23,24]. The structure factors were scaled and corrected for absorption effects using ABSCOR [25, 261 and merged and loaded with PROTEIN [27]. The data collection statistics for the two T2D crystals obtained after 7-h and 14-h dialysis (designated as T2D4 and T2D2 derivatives, respectively) are summarized in Table 1.

599 Table 1. Data collection statistics of T2D ascorbate oxidase. Parameter

Value for T2D2

T2D4

Crystal size

0.8 x 0.4 x 0.3 mm

0.6 x 0.3 x 0.25 mm

Dialysis time

14h

7 11

Space group

P21212 10.639 nm 10.519 nm 11.284 nm

10.632 nm 10.509 nm 11.300 nm

Crystal mosaicity

0.22

0.20"

Resolution

0.250 nm

0.259 nm

Measurements, z > 2.0ff (I)

164518

159603

Unique reflections

32909

32295

Non-rejected unique reflections

28 997

27 572

Data completeness:

co -0.259 nm 62.1% 0.259-0.250 nm 16.3%

co -0.259 nm 65.0% 0.264-0.259 nm 17.9%

Reflection statistics

R-merge" 13.71% RFb6.54%

17.16% 8.20

Unit cell constants a h c

a R-merge = CCIZ(h)i- ( I@)) I/CCZ(h)i, with observed intensity in the ith source and (I@)), mean intensity of reflection h over all measurements of I@). RF = R-merge after independent averaging of Friedel pairs.

Structure solution and refinement FO,rzD- FCTZD difference electron density maps were calculated using the phases of the X-ray crystal structure of the native ascorbate oxidase [7]. These maps were averaged about the local twofold axis using programs of Bricogne [28]. The T2D structures were subjected to crystallographic refinement (X-PLOR Program [29]) of atomic positions and B values. Individual B factors were refined but restrained to a standard deviation of 0.0225 nm2 for the difference of B factors (AB) of bonded atoms and 0.030 nm2 for atoms related by bond angles. The solvent molecules of the native ascorbate oxidase model were used. At the end, occupancies and B values of the copper atoms and oxygen atoms bound to the trinuclear copper site were refined alternating for each of six cycles. The copper sites were refined with weak energy restraints to target values. The geometry of the type-1 copper was averaged from poplar plastocyanin [30] and azurin from Alcaligenes denitrificans [31]. The geometry for the trinuclear copper site was derived from two binuclear copper model compounds with nitrogen and oxygen copper ligands [32, 331. Weak potential energies were applied to Cu-ligand bonds (33.472 kJ mol-' nm-' for Cu-equatorial ligand, 12.552 kJ mol-' nm-' for Cu-axial ligand, and 41.84 kJ mol-' deg-' for ligand-copper-ligand angles. Non-crystallographic symmetry restraints were not imposed on the dimer.

EPR spectroscopy and analysis of samples The X-band EPR spectra were recorded on a Bruker ESP 300 instrument equipped with a variable modulation frequency unit, a microwave frequency counter and a NMR

gauss meter (Bruker Analytische Messtechnik, Karlsruhe, FRG). The modulation frequency was 100 kHz and the modulation amplitude 1.O mT. The sample temperature was maintained at 10- 15 K using the Helitran LTD-110 C system (Air Products, Allentown, USA). The apparent EPR parameters ( g and A ) were determined by measuring the microwave frequency and the magnetic field, calibrated with diphenylpicrylhydrazyl and Mn(I1) in MgO [34]. The concentration of EPR-detectable copper was calculated according to Yanngard [35]. EPR spectra were simulated with the program QPOW obtained from Dr. R. L. Belford, University of Illinois, using literature parameters [34, 361. Five or six crystals of ascorbate oxidase, or of the T2D derivative, in buffer were transferred into quartz tubes (4.0 Ifr 0.1 mm inner diameter) and frozen in liquid nitrogen. Each crystal was selected under the microscope; upon freezing in buffer most of the crystals broke. Note that there are eight monomers of ascorbate oxidase in the unit cell. In each monomer, the copper sites have different spatial orientations. In summary, these samples can be regarded as polycrystalline. To obtain maximum signal intensity and optimum resolution of the EPR signals, the sample tube had to be rotated and shifted within the cavity. After the EPR measurements; these samples were thawed and analysed both by ultraviolet/visible spectroscopy on a Cary 210 spectrophotometer, and by atomic absorption spectroscopy on the AA-175 instrument equipped with a carbon rod atomiser (Varian, Darmstadt, FRG). Ascorbate oxidase (7.95 i-0.1 Cu/140 kDa) purified as described [34, 371 was used as the reference. Protein was determined with the bicinchoninic acid reagent [38]. Ultraviolet/ visible spectra of the crystals Tedissolved in 100 mM sodium phosphate pH 7.0 showed the absorption maximum at 610 nm and the shoulder at 330 nm; for the native enzyme the purity indices A z s o / A s l oand A 3 3 0 / A 6 1were 0 23 and 1 vs 30 and 0.9 for the T2D derivatives. Similarly, the EPR parameters of the crystals redissolved in phosphate buffer were in agreement with data reported earlier [34, 36, 391.

RESULTS AND DISCUSSION The difference electron density maps for the T2D2 and T2D4 derivatives around the trinuclear copper site are shown in Figs 2 and 3. The peaks at the copper positions of the trinuclear copper species correspond to the highest negative peaks in the difference maps (Table 2). The results of the individual refinement of the occupancies and B values of the copper ions and oxygen atoms bound to the trinuclear copper site are displayed in Table 3. The final crystallographic R factors for these refinements are 15.2% for the T2D2 derivative (included reflections from 0.80 nm to 0.259 nm) and 17.0% for the T2D4 derivative (included reflections from 0.8 nm to 0.250 nm). As expected, the type-I centre CU1 has not been removed (Table 3). The Cu atoms of the trinuclear site are partly removed whereby CU4, the putative spectroscopic 'type-2 copper' is depleted most. An average of about 1.3 Cu atoms/ascorbate oxidase monomer are removed similarly for both crystals and soaking times. In the crystal structure of native ascorbate oxidase [6, 71 we detected an additional structural metal positioned on the local diad between the crystallographically independent monomers. This metal can be removed by soaking the crystals in 0.5 mM N,N-diethyldithiocarbamate-containing buffer solution. The nature of this metal is still unknown but it was included in the X-ray model as copper. This metal (CU5

600

k

w Fig. 3. Averaged FOTZD4 - FCTZo4 difference electron density map plus atomic model around the trinuclear copper site. Contour levels as in Fig. 2.

Table 2. Largest holes in the averaged FOT2D-FCTZD maps using phases of the native hob-ascorbate oxidase. Derivatives T2D4 (7-h dialysis) and T2D2 (14-h dialysis). Noise level at -20.0. Atom

Peak height in T2D4

T2D2

Type-3 CU2

-26.6

-41.5

Type-3 CU3

- 36.7

- 35.9

Type-2 CU4 cu5

-48.2 -21.0

- 38.4 -31.1

Table 3. Averaged occupancies and B values for copper atoms and copper oxygen ligands for both T2D derivatives. Atom

Value in derivative T2D2 of

T2D4 of occupancy

B

occuPancY

B

nm2

nm2 Type-I CU1

0.99

0.162

0.99

0.150

Type-3 CU2

0.68

0.260

0.57

0.246 0.238 0.225 0.521 0.221 0.201

Type-3 CU3

0.58

0.239

0.65

Type-2 CU4 cu5 OH1 OH3 .Z Occupancy of c u 1 +c u 4

0.43 0.53 0.76 0.95

0.198 0.469 0.233 0.371

0.51 0.47 0.67 0.92

2.68

2.72

in Table 2) is also partly removed in the T2D crystals (see Table 2). The difference map of the T2D4 derivative shows a prominent positive peak (dashed lines) near His450 (see Fig. 3). In this derivative, the occupancy of the type-2 copper is low (0.43) and that of CU2 from the type-3 dimer is relatively high. This implies the possibility that one histidine ligand of the type-2 copper CU4 moves away and binds to CU2 by its ND3 atom (see Fig. 4). The positive peak is not only produced

by a rotation of the imidazole ring of His450 around its CBCG bond but also by a shift of CU2 towards His450. The EPR spectra of native and T2D ascorbate oxidase crystals are shown in Fig. 5. Double integration of the area under the first derivative of the native enzyme revealed approximately 50 f 5% of the total chemically determined copper to be EPR-detectable. By double integration of the first low-field line around 270 mT (mI= - 3/2), which solely arises from the spectroscopic type-2 Cu, and by computer simulation [34], a ratio of type-l/type-2 Cu of 1.3 & 0.15: 1 was obtained for the native enzyme in fair agreement with the stoichiometry from the X-ray structural analysis. The 3 : 1 ratio of type-l/ type-2 reported earlier for native ascorbate oxidase in frozen solution [34] could be due to partial reduction of the spectroscopic ‘type-2 copper’. The EPR spectrum of the T2D crystals reveals the characteristic resonances of the type-1 copper centre, as also observed for T2D ascorbate oxidase in frozen solution (Fig. 5b) and the nearly complete disappearance of the spectroscopic ‘type-2 copper’. X-ray diffraction provides the statistical average over the occupancies of the metal sites. The statistical average represents the sum of all possible species of the trinuclear copper site with all three Cu atoms removed, two Cu removed, one Cu removed, or no copper removed. Unfortunately, crystallography cannot discriminate between these various possibilities. The observation that, on average, 1.3 Cu atoms are removed from the trinuclear copper site indicates preferential formation of a one-Cu-depleted form with the hole equally distributed over all three copper sites. Each of these oneCu-depleted species may represent an anti-ferromagnetically coupled Cu pair which is EPR-silent and could explain the complete disappearance of the ‘type-2’ EPR signal. It is also conceivable that the residual Cu atoms in the trinuclear copper site remain reduced. This second explanation finds some support from X-ray absorption spectroscopic studies on tree laccase [40]. The very small residual ‘type-2’ signal in the EPR spectra of the T2D crystals may be due to the existence of a small amount of trinuclear copper species with no copper removed. The situation in T2D lacoase may be different. In this enzyme, about 25% of the copper is removed after extensive dialysis against the buffer containing the metal-chelating reagents [14, 171. Hereby, the amount of copper depleted is nearly independent of the dialysis time. It is very likely that, in contrast to ascorbate oxidase, the type-2 copper atom CU4 is removed preferentially compared to the type-3 copper atoms

601 b

b

Fig. 4. Averaged FOTZD4 - FCTID,difference electron density map plus atomic model around the trinuclear copper site. Contour levels as in Fig. 2. His450 is bound to CU2.

5. Esaka, M., Hattori, T., Fujisawa, K., Sakajo, S. & Asahi, T. (1990) Eur. J . Biochem. 191, 537-541. 6. Messerschmidt, A,, Rossi, A., Ladenstein, R., Huber, R., Bolognesi, M., Gatti, G., Marchesin, A,, Petruzzelli, R. & Finazzi-Agro, A. (1989) J. Mol. Biol. 06, 513 -529. 7. Messerschmidt, A., Ladenstein, R., Huber, R., Bolognesi, M., Avigliano, L., Petruzzelli, R., Rossi, A. & Finazzi-Agro, A. (1992) J . Mol. Bid. 224, 179-205. 8. Messerschmidt, A. & Huber, R. (1990) Eur. J. Biochem. 187, 341 - 352. 9. Takahashi, N., Ortel, T. & Putnam, F. W. (1984) Proc. Natl Acad. Sci. USA 81,390-394. 10. Germann, U. A., Muller, G., Hunziker, P. E. & Lerch, K . (1988) J. Biol. Chem. 263, 885 - 896. 11. Karlsson, G., Aasa, R., Malmstrom, B. G . & Lundberg, L. G. (1989) FEBS Lett. 253, 99-102. 12. Pricstlc, J. P. (1988) J. Appl. Crystallogr. 21, 572-576. 13. Malkin, R., Malmstrom, B. G. & Vanngard, T. (1969) Eur. J . Biochem. 7,253-259. 14. Hanna, P. H., McMillin, D. R., Pasenkiewicz-Gierula, M., Antholine, W. E. & Reinhammar, B. (1988) Biochem. J. 253, 561 -568. 15. Ciraziani, M. T., Morpurgo, L., Rotilio, G. & Mondovi, B. (1976) FEBS Lett. 70, 81 -9 1. 16. Reinhammar, B. & Oda, Y. (1979) J . Znorg. Biochem. 11, 115120. 17. Schmidt-Klemens, A. & McMillin, D. R. (1990)J. Inorg. Biochem. 38. 107-115. 18. Graziani, M. T., Loreti, P., Morpurgo, L., Savini, I. & Avigliano, L. (1990) Znorg. Chim. Acla 173, 261 -264. 19. Avigliano, L., Desideri, A,, Urbanelli, S., Mondovi, B. & Marchesini, A. (1979) FEBS Lett. 100, 318-320. 20. Morpurgo, L., Savini, I., Mondovi, B. & Avigliano, L. (1987) J . Inorg. Biochem. 29,25 - 31. 21. Bolognesi, M., Gatti, G., Coda, A,, Avigliano, L., Marcozzi. G. & Finazzi-Agro, A. (1983) J. Mol. B i d . 169, 351 -352. 22. Merli, A., Rossi, G. L., Bolognesi, M., Gatti, G., Morpurgo, L. & Finazzi-Agro, A. (1988) FEBS Lett. 231, 89-94. 23. Pflugrath, J. W. 81 Messerschmidt, A. (1985) Crystallography in molecular biology, NATO Advanced Study Institute and EMBO Lecture Course, p. 86, Bischenberg, France. 24. Messerschmidt, A. & Pflugrath, J. W. (1987) J. Appl. Crystallogr. 20,306-315. 25. Huber, R. & Kopfmann, G. (1969) d c t a Crystallogr. Sect. A 25, 143- 152. 26. Messerschmidt, A., Schneider, M. & Huber, R. (1990) J. Appl. Crystallogr. 23,436 -439. 27. Steigemann, W. (1974) PhD thesis, Technische Universitat, Miinchen.

f

I

V B

i

-,---

, \j

240.0

2601)

280.0

300D

320.0

I

I

340.0

360.0

-1 I 380.0

H (mT) Fig. 5. Electron paramagnetic resonance spectra of crystalline native (A) and T2D (B) ascorbate oxidase. The preparation of samples was as described in Experimental Procedures. Instrumenl settings: microwave frequency 9.25069 GHz; microwave power 2.0 mW; 16 scans. sweep time per scan 335 s; temperature 12 K. EPR parameters obtained from simulations for type-I Cu: gII 2.229, gi 2.053, All 5.9 mT, A, S 0.5 mT; for type-2 Cu: g,, 2.242, gI 2.053, AIl 19.0 mT. A I 5 0.5 mT.

CU2 and CU3. This could be explained by different solvent accessibilities of the type-2 and type-3 Cu atoms in laccase. In ascorbate oxidase, both type-2 and type-3 Cu atoms arc well accessible from the solvent. A. M. and P. K. thank the Deutsche Forschungsgemeinschaft (Schwerpunk tprogramm Bioanorganische Chemiej for financial support.

REFERENCES 1 . Lee, M. H. & Dawson, C. R. (1973) J . Biol. Chem. 248, 65966602. 2. Chichiriao, G., Ceru, M. P., D’Alessandro, A,, Oratore, A. & Avigliano, L. (1989) Plant Sci. 64, 61 -66. 3. Malkin, R. & Malmstrom, B. G. (1970) Adv. Enzymol. 33, 177243. 4. Ohkawa, J . , Okada, N., Shinmyo, A. & Takano, M. (1989) Proc. Natl Acad. Sci. U S A 86. 1239- 1243.

602 28. Bricogne, G. (1976) Acta Crystallogr. Sect. A 32, 832-847. 29. Rrunger, A. T. (1990) X - PLOR (version 2.1) manual, Yale University, New Haven CT. 30. Guss, J. M. & Freeman, H. C. (1983) J . Mol. Biol. 169,521 -563. 31. Baker, E. N. (1988) J . Mol. Biol. 203, 1071-1095. 32. Karlin, K. D., Hayes, J. C., Gultneh, Y.. Cruse, R. W.? McKnown, J. W., Hutchinson, J. P. & Zubieta, J. (1984) J . Am. Chern. Soc. 106,2121-2128. 33. Chaudhuri, P., Ventor, D., Wieghardt, K., Peters, E., Peters, K. & Simon, A. (1985) Angew. Chern. 97,55-56. 34. Marchesini, A. & Kroneck, P. M. H. (1979) Euv. J. Biochem. 101, 65 - 76. 35. Vanngard, T. (1972) Biological applications of electron spin resonance (Schwartz, H. M., Bolton, J. R., Borg, D. C., eds) pp. 41 1-447, J. Wilcy and Sons, Ncw York.

36. Dcinum, J., Reinhammar, B. & Marchesini, A. (1974) FEBS [,eft. 42,241 -245. 37. Avigliano, L., Vechini, P ,Sirianni, P., Marcozd, C., Marchesini, A. & Mondovi, B. (1983) Mol. Cell. Biochem. 56, 107- 112. 38. Smith, P. K., Krohn, R. I., Hermanson. G. T., Mallia, A. K., Gartner, F. H., Provenzano, M. D., Fujirnoto. E. K., Goeke, N. M., Olson, B. J. &Klenk, D. C. (1985) Anal. Biochem. 150, 76 - 85. 39. Cole, J. L., Avigliano, L., Morpurgo, L. & Solomon, E. I. (1991) .7. Am. Chem. Soc. 113, 9080-9089. 40. Kau, L.-S., Spira-Solomon, D. J., Penncr-Hahn, J. E., Hodgson, K. 0. & Solomon, E. I. (1987) J. Am. Chem. Soc. 109,64336442.