Phorochemisrry and Phorobiology Vol. 49, No. 3, pp. 331-336, 1989 Printed in Great Britain. All rights reserved

0031 -8655/89 $03.00+0.00 Copyright @ 1989 Pergamon Press plc

A STEREOCHEMICAL MODEL FOR THE ACTIVE SITE OF PHOTOSYSTEM I1 HERBICIDES*

GARY GARDNERt E I. du Pont dc Nemours and Company. Agricultural Products Department, Experimental Station,

Bldg. 402, Wilmington, DE 19898, USA

(Received 26 August 1988; accepted 21 September 1988)

Abstract-Photosystem I1 herbicides act by blocking electron transport at the secondary electron acceptor ‘Oe’, thought to be a non-covalently bound plastoquinone. Recent evidence suggests that these compounds work by displacing the plastoquinone from its site in the thylakoid. Since the herbicides cannot act as electron carriers, electron transport is then blocked. In this report a model is presented for the site of action of Photosystem I1 herbicides that encompasses biochemical, biophysical, and structure-activity considerations. The essence of the model is that Photosystem I1 herbicides are non-reducible analogues of plastoquinone or its semiquinone anion. As examples of the ways in which known herbicidal classes fit the model, the possible interactions of diuron, atrazine, the putative urea-triazine hybrid MBAT (the a-methylbenzyl analogue of atrazine), and dinoseb with the active site are discussed. This model provides a stereochemical basis for herbicidal activity and offers a qualitative approach for the design of novel Photosystem I1 herbicides.

INTRODUCTION

Herbicides that inhibit photosynthesis have been the subject of numerous structure-activity studies for over two decades. Despite some success in understanding the structural requirements for activity within a given chemical class, no quantitat- ive or mechanistic basis exists for applying infor- mation learncd from one class to the design of novel compounds of a different class. Recent information has shed new light on the mechanism of action of Photosystem I1 (PS 11)s herbicides. The purpose of this report is to review this new information and propose a stereochemical model for the site of action that will allow the rational design of novel biologically active molecules.

Inhibition of electron transport and herbicide binding

Many of the world’s commercially important herbicides act by inhibiting photosynthesis, and most of these act at the same site in photosynthetic electron transport. These compounds inhibit light- induced reduction of the secondary electron acceptor, called QB, B (Bouges-Bocquet, 1973), or R (Velthuys and Amesz, 1974), in PS 11 of higher plants. The secondary electron acceptor has been thought to be a bound form of plastoquinone, and

‘Dedicated to Professor Winslow R. Briggs on the occasion of his sixtieth birthday.

tPresent address: Abbott Laboratories, Chemical and Agricultural Products Division, North Chicago, IL 60064, USA.

$Abbreviations: DCMBAT, the N,a-methyl-3,4-dichloro- benzyl analogue of atrazine; MBAT, the N,a-methyl- benzyl analogue of atrazine; PS 11, Photosystem 11; QSAR, quantitative structure-activity relationship.

it was suggested that compounds such as diuron act between the primary acceptor QA and the secondary acceptor by lowering the midpoint potential of QB

relative to QA (Velthuys and Amesz, 1974). This formalism is consistent with the biophysical obser- vations in the literature, but it does not encompass some recent biochemical observations.

Biochemical study of the PS I1 herbicide binding site began with the work of Tischer and Strotmann (1977). They carried out binding studies with radio- labeled herbicides and demonstrated that the PS I1 herbicides bind competitively at the same site on the thylakoids. The advantage of this technique for structure-activity studies is that it allows effects at this site to be distinguished from effects on other sites in the electron transport chain. Use of this approach enabled Pfister et al. (1979) to determine that the biochemical basis for triazine resistance was an alteration in the binding site so that it no longer bound triazines such as atrazine, although affinity for other herbicide classes, such as ureas, still appeared normal. Photoaffinity studies with an azi- do-analogue of atrazine indicated that the receptor site contained a protein ( D l ) of 32 000 mol wt (Gardner, 1981). This peptide was not labeled in thylakoids of triazine-resistant pigweed (Pfister et a l . , 1981). Similar studies with an azido-analogue of dinoseb labeled different polypeptides, in the range of 41-53 kDa (Oettmeier et al . , 1980), indicat- ing the probability that the active site of PS I1 inhibitors consists of more than one polypeptide.

Quinone displacement

In recent years, experimental evidence has led to a drastically altered concept of the way in which PS I1 inhibitors act. Rather than being a tightly bound plastoquinone, QB in either the fully oxidized or

33 1

332 GARY GARDNER

fully reduced form is thought to be loosely bound to its apoprotein and in equilibrium with the plasto- quinone pool. QB is a two-electron carrier, and the semiquinone anion is quite long-lived and appears not to be protonated, implying that the semiquinone anion has a much higher affinity for the binding site than either the fully oxidized or fully reduced forms. Using bacterial photosynthesis as a model, Wraight (1981) suggested that inhibitors act by displacement of the Q, quinone from the binding site, i.e. PS I1 herbicides are competitive inhibitors of plastoqui- none binding. Velthuys (1981) has proposed a simi- lar model based on spectral studies with higher plant thylakoids. In this model, the action of the inhibitors has nothing to d o with an effect on the midpoint potential of Qe. Because PS I1 herbicides prevent the binding of plastoquinone to the Qe site, they also prevent the reduction of QB.

Several reports have offered direct experimental evidence for this new model of PS I1 herbicide action. Vermaas et al. (1983) have shown that an azidoquinone replaced the native plastoquinone at QB. This synthetic azidoquinone also competed with radiolabeled atrazine and ioxynil for binding. In another study, a short-chain plastoquinone analogue was found to compete for radiolabeled diuron bind- ing sites in thylakoids that had been depleted of native plastoquinone (Oettmeier and SOU, 1983). Thus there is now direct evidence that quinones and PS I1 inhibitors can compete for the same binding site.

Another line of experimentation indicates that herbicide binding is dependent on the redox state of the chloroplasts. Jursinic and Stemler (1983) showed that decreased binding of atrazine corre- sponds to an increase in the semiquinone form of Qe. Similar results were reported for the binding of diuron (Laasch et al . , 1983). Because these effects were on the affinities of the herbicides, rather than on the number of sites, they are quite consistent with the view that the semiquinone anion has a much greater affinity for the site than the oxidized or fully reduced forms.

Structure-activity relationships

Structure-activity studies on PS I1 herbicides have largely ignored this biochemical information about the binding site. It has long been recognized that a common chemical element essential for the inhi- bition of PS I1 is an sp2 hybrid -C-N- bound to a lipophilic carrier (Trebst and Draber, 1979). This element is found in the chemical classes of ureas, carbamates, anilides, triazines, trifluorobenzimida- zoles, triazinones, pyridazinones, thiadiazolones, pyrazolones, uracils, and pyrimidinones. The one major class of PS I1 inhibitor that does not contain this element (at least as described) is the nitrophen- 01s. Recently, a second type of essential element has been formulated to account for the activity of

the phenolic PS I1 inhibitors (Trebst and Draber, 1986). Any comprehensive model of the PS I1 bind- ing site must be able to account for the activity of the nitrophenols as well as for the other classes.

Quantitative structure-activity relationship (QSAR) studies have primarily focused within a single class of PS I1 inhibitor. Considerable progress has been made in understanding structural requirements for activity within a class, for example, the phenylureas (Cross et af., 1983; Kakkis et a l . , 1984). Mitsutake et al. (1986) have examined QSAR for the triazines as well as the phenylureas, although most of their data relate the phenylureas to close analogues, the anilides and carbamates. They note that the site of action of the phenylureas is tolerant to steric bulk- iness of para substituents on the phenyl ring. They also note steric tolerance in the N-alkyl region of the triazine receptive surface, and infer a relationship between these two regions. However, a specific way in which the two classes might be related was not proposed.

THE MODEL

The essential feature of the model is that PS I1 inhibitors are non-reducible analogues of plastoqui- none and its semiquinone anion. This simple state- ment has several very explicit structure-activity implications. Firstly, plastoquinone is asymmetric; therefore, the receptor should be. The strongly hydrophobic regions of the inhibitors should corre- spond to the isoprenoid tail of the plastoquinone. Secondly, since the quinone has two carbonyl groups, the receptor should have two functional regions capable of interacting with these carbonyl groups. I suggest that the sp2 -C- of the essential element of the herbicide corresponds to the sp’ carbons of the quinone carbonyls, and thus there are two possible sites to which this element can bind. Consequently, not all PS I1 herbicides bind to the receptor in the same way. Thirdly, there are two endogenous ligands for the site-plastoquinone and its semiquinone anion. Some herbicides may be ana- logues of plastoquinone; others may be analogues of the semiquinone anion. In order to illustrate details of this model, I would like to discuss how four types of PS I1 inhibitors might interact with the site.

Diuron

Figure 1 shows the structure of plastoquinone and how three types of herbicides might act as plastoquinone analogues. For clarity, only three of the ten isoprenoid units of the plastoquinone are indicated. In Fig. 1B the urea herbicide diuron is oriented as it might occupy the plastoquinone site. The diuron carbonyl overlaps the carbonyl at C-4 of the quinone, and the substituted phenyl ring occupies the same region in space adjacent to C-5

A stereochemical model of Photosystem I1 herbicides 333

Figure 1 . Diuron, atrazine, and MBAT as analogues of plastoquinone. (A) The structure of plastoquinone, with only three of the ten isoprenoid units shown. (B) The orientation of diuron as it might occupy the plastoquinone site. ( C ) The suggested orientation of atrazine, occupying a different region of the plastoquinone site. (D) The way in which the N,a-methylbenzyl analogue of atrazine (MBAT) might occupy both the triazine and phenylurea regions of

the plastoquinone binding site.

as the isoprenoid tail. Nitrogens at positions 3 and 5 allow the herbicide to be planar in the region of the quinone ring. Moderate alkyl substitutions are allowable at the nitrogen corresponding to the methyl at C-3 of the quinone, as well as methoxy (as in linuron and ubiquinone).

Atrazine

I suggest here that triazine herbicides occupy a different region of the receptor than the ureas. Fig- ure 1C shows the way in which atrazine might occupy the plastoquinone site. In this case, the sp2 carbon is in the triazine ring and occupies the same region in space as the C-1 carbonyl of the quinone. The nitrogen of the essential element occupies the same region as C-6 of the quinone. It is probable that the alkyl substitution at C-6 can be sterically larger than at position C-3, since the optimal triazine substitution at that position is t-butyl. The most important consequence of orienting the structures in this way is that much of the triazine structure does not overlap with either the quinone or with a urea. The sector of space moving out from the regions between C-1 and C-2 of the quinone must allow quite a large hydrophobic substitution. Since the azido group replaced the chlorine in atrazine in the photoaffinity studies discussed above (Gardner,

1981), it is probable that the 32 kDa protein is found in this region of the receptor. Given this relationship (Figs. 1B and C ) , it is feasible to see how a mutation in this region of the binding site could result in a receptor that could not bind a triazine but could still bind a urea or plastoquinone.

M B A T

We have previously proposed that the N,a- methylbenzyl analogue of atrazine (MBAT), and its 3,4-CI2 derivative, DCMBAT, are hybrid molecules incorporating features of both triazines and ureas (Gardner et al., 1987). In that same report we also pointed out that MBAT is not a strict analogue of a phenylurea because there is a carbon atom between the nitrogen of the essential element and the phenyl ring. Figure 1D shows the way in which MBAT might act as an analogue of plastoquinone. The triazine ring occupies the same region of the receptor as in atrazine, and the extra carbon allows the phenyl ring to occupy the same region as the phenyl ring in diuron. Additional binding strength because of this hydrophobic pocket could account for the activity of this compound in chloroplasts from triazine-resistant pigweed (Gardner et a[., 1987).

Dinoseb

Whereas the three compounds above were dis- cussed as analogues of plastoquinone, I suggest that the nitrophenol herbicides are analogues of the semiquinone anion. Figure 2 shows the way in which dinoseb occupies the site in relation to the semiqui- none anion, with the hydroxyl of the phenol in the same region as C-4 of the semiquinone. The nitro group para,to the hydroxyl would then be in the same region as the C-1 carbonyl, and the alkyl substitution,corresponds to the isoprenoid tail at C-

BLQwsfs

Figure 2. Dinoseb as an analogue of the semiquinone anion of plastoquinone.

334 GARY GARDNER

5. The nitro group at C-3 is not essential, since halogen can substitute (Trebst and Draber, 1979). I t is this nitro that was replaced by azido in the photoaffinity studies discussed above (Oettmeier er al. , 1980), and since the azido group in azido-dino- seb and azido-atrazine occupy quite different regions in space, it is not unreasonable that they would label proteins of different molecular weight.

DISCUSSION

This model suggests that plastoquinone, diuron, and dinoseb occupy a similar region of the receptor but that atrazine is somewhat removed from that area. This suggestion is consistent with the recent experiments of Diner and Petrouleas (,1987), in which diuron, o-phenanthroline, and dinoseb modi- fied the EPR spectrum of non-heme Fe(II1) relative to controls. Atrazine (and terbutryn) had no effect, suggesting that atrazine, in contrast to plastoqui- none and the other inhibitors, binds at a site with no direct interaction with a ligand of the iron.

Shipman (1981) has pointed out that PS I1 herbi- cides have a flat polar component surrounded by hydrophobic substitution and that there should be a locally-strong electric field in the binding site to interact with this dipole. perhaps a salt bridge. We have shown that arginine residues, a candidate for the cationic end of such a salt bridge, are involved in the binding of atrazine and diuron (Gardner et a / . , 1983). If such a positively charged group is interacting with one (or both) of the carbonyls of plastoquinone, then it is easy to see why the semiquinone anion would have a greater binding affinity (Jursinic and Stemler, 1983; Laasch et al . , 1983), especially if the anion is localized at one of the carbonyls. As D. A . Kleier has pointed out (personal communication), nitrophenols such as dinoseb would probably be ionized at physiological pH. If the semi-quinone anion is localized at C-4, the analogy between it and dinoseb becomes more obvious.

One question still unexplained is the driving force that causes the fully reduced quinol to leave the binding site. Perhaps the increase in net negative charge leads to instability and shifts the equilibrium drastically toward dissociation. Alternatively, per- haps the second electron leads to a delocalization of the charge around the ring (if, in fact, the semiquinone is localized), and this places a negative charge in apposition to the negative pole of the electric field in the binding site. (This region would interact with the essential -NH- in ureas and tria- zines and thus be near carbons 5 and 6 of the quin- one.) Such a situation would also probably lead to rapid dissociation. Because there is no experimental evidence as to this question, these suggestions should be regarded as pure conjecture at this time.

The model presented here attempts to relate structure-activity principles that govern the binding

of ligands to the QB site in PS 11. It is not a model of the D1 protein itself. Without a three-dimensional structure of the PS I1 reaction center, one cannot be certain about the placement of the ligands within the protein. However, the recent X-ray structure of the bacterial reaction center of Rhodopseudomonas viridis (Deisenhofer et al., 1985; Michel et al., 1986; Michel and Deisenhofer, 1988) provides a good analogy. In R. viridis crystals, QB was lost. When the crystals were infused with ubiquinone-1, eight possible binding sites were found (Michel er a / . , 1986). Infusion with the triazine terbutryn yielded one binding site. Possible hydrogen bonds were observed between a ring nitrogen of the triazine and the peptide nitrogen of isoleucine L224 and between the ethylamino nitrogen of terbutryn and the side chain oxygen of serine L223 of the protein. Since a mutation in serine 264 of the higher plant protein confers triazine resistance (Hirschberg and Mclntosh, 1983), the analogy between ser 264 and ser L223 of the bacterial sub-unit is obvious. In R. viridis, o-phenanthroline binds close to the non- heme iron and hydrogen bonds to the imidazole nitrogen of histidine L190. By analogy, and with reference to the work cited above by Diner and Petrouleas (1987), one would expect diuron, dino- seb, and plastoquinone to bind also to a histidine, perhaps his 215 in the plant protein. The quinone head group of QA interacts with the indole ring of tryptophan M250 in the bacteria and phenylalanine L216 occupies an analogous position in the QB

pocket (Michel et al., 1986). It is tempting to suggest that phenylalanine 255 plays a similar role in the plant site; however, there are several aromatic amino acid residues in that region of the protein, and it is not possible to be precise about the interac- tions without structural data.

It must be stressed that the bacterial reaction center is only an analogy to the higher plant system, albeit an elegant one. With regard to the Qa site, the fine structure of the plant and bacterial systems differs in two very important respects. First, the loop connecting helices 4 and 5, that which defines the QB pocket, is much longer in the plant protein (Kleier pf al., 1987) than in the bacterial L subunit (Deisenhofer et a[ . , 1985). Second, the struc- ture-activity relationships for herbicides are very different in the two systems. For example, diuron is one of the most active inhibitors of PS I1 but is extremely weak in inhibiting bacterial electron transport. And atrazine is at least two to three orders of magnitude less active in the bacterial reac- tion centers than in plants. [Compare values in Gardner et af. (1987) with those in Brown et a/. (1984).] Because the model presented in the present paper seeks to explain herbicidal structure-activity relationships, one must be extremely cautious in extrapolating precise details from the bacterial data. I would, however, venture to speculate about three of the molecules featured in the present model. The

A stereochemical model of Photosystem I1 herbicides 335

two carbonyl oxygens of plastoquinone, as sug- gested above, interact with two regions of the bind- ing site. By analogy with QA in the bacterial system, these would be a histidine, perhaps his 215 in D1, and a peptide nitrogen on the opposite side of the binding site. This could be analogous to isoleucine L224 of R. viridis, but without more structural data, the exact analogy cannot be drawn. Diuron also probably interacts strongly with his 215, but structure-activity differences would suggest that the fine structure of the entrance to the hydrophobic pocket is different in the two organisms. Finally, atrazine may form three hydrogen bonds to the QB

pocket of D1, as opposed to two in the bactzria. The much higher activity could be ascribed to an additional H-bond, and it is known (Ebert and Dumford, 1976) that substitution of either of the aminoalkyl nitrogens results in lowered activity, implying that both are required for interaction with the protein. Again, because the details of the region between helices 4 and 5 are very different in plants and bacteria, with the exception of the involvement of serine 264 it is not realistic to speculate about the precise nature of the atomic interactions.

CONCLUSION

A model has been presented for the site of action of PS I1 herbicides that encompasses biochemical, biophysical, and structure-activity considerations. The essence of the model is that PS I1 herbicides are non-reducible analogues of plastoquinone or its semiquinone anion. Wraight (1981) noted that competition between herbicide and quinone at the same site implies a structural homology between the two. He pointed out that a stereochemical basis for herbicide activity had not been forthcoming. The model presented here does provide that stereochem- ical basis and offers a qualitative approach for the design of novel PS I1 herbicides.

Acknowledgement-I thank D. A. Kleier for valuable discussions.

REFERENCES Bouges-Bocquet, B. (1973) Electron transfer between

the two photosystems in spinach chloroplasts. Biochim. Biophys. Acta 314, 250-256.

Brown, A. E., C. W. Gilbert, R. Guy and C. J. Arntzen (1984) Triazine herbicide resistance in the photosyn- thetic bacterium Rhodopseudomonas sphaeroides. Proc. Natl. Acad. Sci. USA 81, 631G-6314.

Cross, B., P. P. Hoffman, G. T. Santora, D. M. Spatz and A. R. Templeton (1983) The design of postemergence phenylurea herbicides using physicochemical par- ameters and structure-activity analyses. J . Agric. Food Chem. 31, 260-264.

Deisenhofer, J., 0. Epp, K. Miki, R. Huber and H. Michel (1985) Structure of the protein subunits in the photosynthetic reaction centre of Rhodopseudornonas viridis at 3A resolution. Nature 318, 618-624.

Diner, B. A. and V. Petrouleas (1987) Light-induced oxidation of the acceptor-side Fe(I1) of Photosystem I1

PhP 49/3-G

by exogenous quinones acting through the QB binding site. 11. Blockage by inhibitors and their effects on the Fe(II1) EPR spectra. Biochim. Biophys. Acta 893,

Ebert, E. and S. W. Dumford (1976) Effects of triazine herbicides on the physiology of plants. Residue Rev. 65,

Gardner, G. (1981) Azidoatrazine: Photoaffinity label for the site of triazine herbicide action in chloroplasts. Science 211, 937-940.

Gardner, G., C. D. Allen and D. R. Paterson (1983) Modification of Photosystem I1 by phenylglyoxal: Evi- dence for the involvement of arginine in atrazine binding to chloroplasts. FEBS Lerr. 164, 191-194.

Gardner, G., J. R. Sanborn and J. R. Goss (1987) N- Alkylaryltriazine herbicides: A possible link between triazines and phenylureas. Weed Sci. 35, 763-769.

Hirschberg, J . and L. McIntosh (1983) Molecular basis of herbicide resistance in Amaranthus hybridus. Science 222, 13461348.

Jursinic, P. and A. Stemler (1983) Changes in [l4C]atra- zine binding associated with the oxidation-reduction state of the secondary quinone acceptor of Photosystem 11. Plant Physiol. 73, 703-708.

Kakkis, E., V. C. Palmire, Jr., C. D. Strong, W. Bertsch, C. Hansch and U. Schirmer (1984) Quantitative structure-activity relationships in the inhibition of Pho- tosystem I1 in chloroplasts by phenylureas. J . Agric. Food Chem. 32, 133-144.

Kleier, D. A., T. A. Andrea, J. K. J. Hegedus, G. M. Gardner and B. Cohen (1987) The topology of the 32 kDa herbicide binding protein of Photosystem I1 in the thylakoid membrane. 2. Naturforsch. 42c, 733-738.

Laasch. H., W. Urbach and U. Schreiber (1983) Binary flash-induced oscillations of [ ''C]DCMU binding to the Photosystem I1 acceptor complex. FEBS Leu. 159, 275-279.

Michel, H., 0. Epp and J. Deisenhofer (1986) Pigment-protein interactions in the photosynthetic reac- tion centre from Rhodopseudomonas viridis. EMBO J .

Michel, H. and J. Deisenhofer (1988) Relevance of the photosynthetic reaction center from purple bacteria to the structure of photosystem 11. Biochemistry 27, 1-7.

Mitsutake, K.-i., H. Iwamura, R. Shimizu and T. Fujita (1986) Quantitative structure-activity relationship of Photosystem I1 inhibitors in chloroplasts and its link to herbicidal action. J . Agric. Food Chem. 34, 725-732.

Oettmeier, W., K. Masson and U. Johanningmeier (1980) Photoaffinity labeling of the Photosystem I1 herbicide binding protein. FEBS Lett. 118, 267-270.

Oettmeier, W. and H.-J. Sol1 (1983) Competition between plastoquinone and 3-(3,4-dichlorophenyI)-l,l- dimethylurea at the acceptor side of Photosystem 11. Biochim. Biophys. Acta 724, 287-290.

Pfister, K., S. R. Radosevich and C. J. Arntzen (1979) Modification of herbicide binding to Photosystem I1 in two biotypes of Senecio vulgaris L. Plant Physiol. 64, 995-999.

Pfister, K., K. E. Steinback, G . Gardner and C. J. Arntzen (1981) Photoaffinity labeling of an herbicide receptor protein in chloroplast membranes. Proc. Narl. Acad. Sci. USA 78, 981-985.

Shipman, L. L. (1981) Theoretical study of the binding site and mode of action for Photosystem I1 herbicides. J . Theor. Biol. 90, 123-148.

Tischer, W. and H. Strotmann (1977) Relationship between inhibitor binding by chloroplasts and inhibition of photosynthetic electron transport. Biochim. Biophys. Acta 460, 113-125.

Trebst, A. and W. Draber (1979) Structure activity correlations of recent herbicides in photosynthetic reac- tions. In Advances in Pesticide Science (Edited by H.

138-148.

1-103.

5, 2445-2451.

336 GARY GARDNER

Geissbuehler), Part 2, pp. 223-234. Pergamon Press, Oxford.

Trebst, A. and W. Draber (1986) Inhibitors of Photosy- stem I1 and the topology of the herbicide and OR binding polypeptide in the thylakoid membrane. Photosyn. Res.

Velthuys, 8. R. (1981) Electron-dependent competition between plastoquinone and inhibitors for binding to Photosystem 11. FEBS Left. 126, 277-281.

Velthuys, B. R. and J . Amesz (1974) Charge accumu-

10, 381-392.

lation at the reducing side of System 2 of photosynthesis. Biochim. Biophys. Acta 333, 85-94.

Vermaas, W. F. J., C. J. Arntzen, L.-Q. Gu and C.-A. Yu (1983) Interactions of herbicides and azidoquinones at a Photosystem I1 binding site in the thylakoid mem- brane. Biochim. Biophys. Acta 723, 266-275.

Wraight, C. A. (1981) Oxidation-reduction physical chemistry of the acceptor quinone complex in bacterial photosynthetic reaction centers: Evidence for a new model of herbicide activity. Israel J. Chem. 21, 348-354.