globin and, in the case of the latter two most of the drug was bound to protein rather than to the haem moiety; there was no binding to free globin (Yang et ai, 1994). Haem did react to form covalent adducts (see below under parasite molecules targetted by artemisinin), and artemisinin also reacted with isolated haemozoin although the incorporation of label was about 5-fold less than that found with intact parasites (Hong et al., 1994); it is suggested that this is because the drug acts more rapidly with haem (Fell) than with haem (Felll), (for evidence see below), as haem is maintained as haem (Fell) in intact red cells.
The action of artemisinin against P. falciparum in vitro was antagonised by iron chelators suggesting that free iron or haem was responsible (since some chelators e.g. desferrioxamine bind haem as well as iron). Desferrioxamine also antagonised the action of artemether, 3, and the action of this and other chelators was prevented by pre-saturation with iron (Meshnick et al., 1993). The binding of chelators to haem would prevent the interaction of haem-iron with the endoperoxide moiety. It has also been suggested that free iron could be released by the alkylation of haem with artemisinin. Iron chelators would then prevent this free iron from reacting with the endoperoxide group to produce more free radicals (Cumming et al., 1997 and references therein).
The necessity for haem-iron, (rather than free-iron) involvement in the mode of action of artemisinin is supported by the high degree of selectivity seen in the action of this drug against malaria parasites present in red blood cells while leaving other cells, with the possible exception of neuronal cells unaffected (see page 268). In addition chloroquine, which binds to haem (Felll) preventing its polymerisation into haemozoin antagonises the action of artemisinin (Chou et al., 1980). Interestingly, chloroquine resistant P. berghei which lacks visible haemozoin is resistant to artemisinin suggesting that haemoglobin breakdown is essential for the antimalarial action of artemisinin (Peters et al., 1986). Further support for this conclusion has been provided by electron micrographs which reveal that artemisinin, (Anon. 1979), and artesunate, 5, (Li et al., 1981), initially affect parasite food vacuoles. In addition, autoradiography studies have confirmed that artemisinin is located in the parasite food vacuole (Maeno et al., 1993).
Following the observation that artemisinin causes changes to membranous structures including the mitochondria, Zhao et al. (1986), investigated the possibility that the drug might act by inhibiting the mitochondrial haem-containing enzyme, cytochrome oxidase. The cytochrome oxidase of P. berghei trophozoites was completely inhibited by sodium artesunate at 1 mM in vitro and in vivo at 100 mg/kg when given intravenously to infected mice; however, these doses are much greater than are required to inhibit the growth of malaria parasites so that inhibition of this enzyme is unlikely to be the primary mode of action of artemisinin-like compounds.
Taken together, the above suggests that artemisinin-like compounds are likely to react most readily with haem (Fell), released from haemoglobin as a result of parasite digestion before its conversion into haemozoin.
The Nature of the Interaction between Haemin and Artemisinin
A number of research groups have investigated the ability of iron containing compounds to cleave the peroxide group of artemisinin and related compounds including the trioxanes which are compounds based on the 1,2,4-trioxane ring la of artemisinin 1 (Wu et al., 1998 and references therein). Ferrous (Fell) compounds readily react but with ferric (Felll) compounds the reaction is much slower under the same conditions; in a study of the reduction of artemisinin using cyclic voltammetry it was found that artemisinin cannot be reduced by Felll (Zhang et al., 1992). The rate of the reaction is also pH dependent, being much slower at pH 6-7 than at pH 4 (Haynes and Vonwiller, 1996). It is noteworthy, in this context that the pH of the malaria parasite food vacuole where haemoglobin is digested is estimated to be about 5.3 (Krogstad, et al, 1985).
Whereas ferric iron reacts only slowly with artemisinin, Meshnick et al. (1993), reported that artemisinin reacts strongly with haem in both Fell and Felll forms. However, in the case of artesunate, the reaction is reported to be much slower with haem (Felll) than with haem (Fell), (Adams and Berman 1996). Similar slowing of the reaction was found when artemisinin was incubated with haem (Fell) without a reducing agent. When haemoglobin is digested by the malaria parasite, haem (Fell) is released but this may readily oxidise to haem (Felll), although it has been suggested that the high level of glutathione present in red blood cells may maintain haem in its reduced (Fell) form (Chen et al., 1998). Questions remain concerning the the oxidation states of haem in parasitised erythrocytes and its reactivity towards artemisinin derivatives.
The interaction of artesunate, 5 with haem has been studied using cyclic voltam-metry (Chen et al., 1998). Note that in these experiments, haem (Felll) was initially solubilised in sodium hydroxide solution resulting in the formation of haematin in which a hydroxyl group is co-ordinated onto the iron; however, spectroscopic studies indicate that the molecules actually exist as dimers linked by an Fe-O-Fe bridge (Brown et al., 1980). In the presence of haem (Felll) at concentrations as low as 2xl0~8 M the reduction of artesunate, (1 mM), was facilitated (reduction potential reduced by 680 mV) indicating that the reduction was catalytic; this was supported by UV spectra which showed that no complex was formed between haem and artesunate. No reduction was observed when deoxydihydroartemisinin and succinic acid were used in place of artesunate, thus confirming that reduction of the peroxy bridge had occurred. During the course of the reaction haem (Felll) is electrolytically reduced to haem (Fell) and the latter then reduces artesunate with regeneration of haem (Felll).
The above results are consistent with those reported for artemisinin, (Zhang et al., 1992), suggesting that the modes of action of the two drugs are similar. For the above reductive process to occur in the parasite-infected red blood cell, a mechanism for the reduction of haem (Felll) would be needed; as suggested above, glutathione, may fulfill this role.
Paitayatat et al., 1997, examined the ability of a number of artemisinin derivatives to bind with haem (Fell); actually this was Fell haematin since it was prepared by dissolving haem (Felll) in sodium hydroxide and then adding sodium dithionite as a reducing agent. The absorption peak of haem (Fell) at 415 nm was immediately reduced by artemisinin and other derivatives which possess antimalarial activity; deoxyartemisinin, 7, as expected did not affect absorbance. With the exception of artesunate, there was a correlation between the dissociation constant and the log IC50 antiplasmodial activity.
Using molecular modelling techniques, Shukla et al. (1995), explored the mode of binding of artemisinin and of deoxyartemisinin, 7, with haem (Fell) and haem (Felll).
With haem (Felll) and artemisinin, the lowest energy docking configuration was found to be when the peroxide bridge oxygens were in close proximity to the haem iron (this projects slightly above the plane of the porphyrin ring and artemisinin binds on the same side); the oxygen at C-10 was also involved as well. However, binding could also occur with the 3 non-peroxide oxygens of artemisinin locating onto the iron since the energy of this configuration is only 1.6 kcal/mol above that for the peroxide-oxygen binding mode. A third possibility involving binding with one peroxide oxygen O-l and the non-peroxidic oxygen O-l3 was found to be of much higher energy. In the case of haem (Fell), three modes of binding are also possible but in this case the energy difference between them was small, only 0.2 kcal. The configuration in which the peroxide oxygens bound to the Fell iron had the lowest energy at 233.5 kcal/mol but this is significantly higher than that for peroxide binding with haem (Felll), (220 kcal/mol). It is interesting that the reaction of Fe II compounds with artemisinin is much faster than those of Fe III since the latter would appear to provide the most favourable complex. However, the binding of artemisinin to Fe II or Fe III haemin is only the initial step in a sequence of events leading to radical generation and it is likely that the subsequent step(s) are more favourable with Fell. Further studies are needed in order to explain the above observations.
When the above modelling was repeated with deoxyartemisinin, 7, the most stable arrangement was found to be one in which the oxygens 0-2, 0-13 and 0-19 interacted with haem (Felll); the lack of involvement of 0-18 indicates that the interaction of the latter with haem is different from that of artemisinin (Shukla et al., 1995), but in any case 7 is inactive as an antimalarial because it lacks the endoperoxide group.
In another study, Grigorov et al. (1997), examined the quantitative structure-activity relations (QSAR), of a series of synthetic trioxanes and showed that two hydrophobic features and hydrogen bonding ability are essential for antimalarial activity in these compounds. Molecular modelling of an active trioxane-haem (Fell) complex indicated that the peroxide bond of the trioxane lies close to the Fell atom of haem, thus supporting the above hypothesis.
In order to test the hypothesis that the action of artemisinin-like antimalarials requires binding of the peroxide group with haem iron, Zouhiri et al. (1998), syn-thesised tricyclic 1,2,4 trioxanes possessing an a-methyl group at C5 in place of the hydrogen normally found in this position in antimalarial trioxanes. As expected, antiplasmodial activity in the latter compounds was markedly reduced, presumably as a result of steric hindrance preventing close contact between the iron and the peroxide moiety.
Mechanistic Scheme for the Reaction of Artemisinin with Iron
The reaction of artemisinin with ferrous (Fell) compounds has been used as a model by several research groups in order to postulate the identity of products which may have a significant role in the parasite killing action of artemisinin. The use of various reactants and reaction conditions in different laboratories has led to some confusion as divergent results and a variety of reaction products have been reported. However, many of these results have been rationalised by Wu et al. (1998) who have proposed a "United mechanistic framework for the Fe(II) induced cleavage of qinghaosu (artemisinin) and its derivatives". Scheme 1 illustrates the pathways proposed in this scheme which explains the formation of the major products found in laboratory experiments (adapted from Wu et al., 1998). The main features of the scheme are discussed below, but the reader should consult the original work for further information.
It is generally agreed that the initial event is the transfer of a single electron from an Fe(II) ion to the peroxy bond of artemisinin 1. Two possible radicals (1A, IB, scheme 1), result according to whether the iron combines with O-l or 0-2, although for reasons which will be explained later, it is assumed that 1A and IB are rapidly interchangeable. Each of the two radicals may then undergo further reactions.
Radical 1A undergoes a 1,5-H shift to give 9 which leads to the formation of deoxyartemisinin, 11 via two possible pathways; firstly, C3-02 scission leads to the enol 10 which could easily form 11 by addition of the OH to the enol double bond. During this process the loss of Fe(IV)=0, (Fe2+=0) occurs which may be significant as this species has been suggested to be responsible for the parasiticidal affect of artemisinin (see page 260). In the second pathway leading to deoxyartemisinin formation, the intermediate 9 acquires a hydrogen atom to yield 12 and then loses Fe(III)=0, (Fe+=0)to form the anion, 13 which then deprotonates to form 11. Alternatively, intermediate 9 may undergo radical substitution at 0-2 with the loss of Fe2+ to give 14, which then rearranges to 15.
Unlike radical 1A, radical IB cannot abstract a hydrogen from C-4, (and no a-hydrogen is present on C-3), so that reaction takes place by /3-scission. C3-C4 scission yields the C-4 primary radical 16, which may undergo radical substitution at O-l with the loss of Fe2+ to give 17 (arteannuin G). It is also possible that C3-013 scission in IB may occur leading to 19 and/or 20; this will be discussed below (see page 258).
Evidence which supports the above routes of product formation (via 1A and IB respectively), was obtained by examining the reaction of simple 1,2,4-trioxanes (based on la as models of artemisinin) with iron (Posner and Oh, 1992 and reviewed in Cumming et al., 1997). Using lsO labelled compounds the former authors found that the major products formed were analogous to 10 and 11, (deoxyartemisinin), and proposed that these products could arise by mechanisms analogous to those shown in scheme 1. Interestingly, the formation of different products according to which oxygen atom is attached to the iron is paralleled in nature by the haem-induced prostaglandin-endoperoxide rearrangement (Ullrich and Brugger, 1994). In addition, 11 and 17 are also known microbial degradation products of artemisinin (Lee and Hufford, 1990) and analogous metabolites are similarly produced from artemether by microbial (Hu et al., 1992) and mammalian (Chi et al., 1991), metabolism as well as from deoxoartemisinin (Khalifa et al., 1995).
We will now consider how the above scheme is able to explain the apparently disparate published results of experiments in which the reaction of various iron compounds with artemisinin has been investigated. It is clear that the nature and the
Scheme 1 Proposed mechanistic framework for the Fe(II) induced cleavage of artemisinin. (Adapted from Wu et al., 1998).
relative proportions of the products arising in a particular reaction are very much dependent upon both the nature of the solvent used and on the anion of the iron salt.
For example, when tetrahydrofuran, (THF) is used as the solvent with ferrous bromide and artemisinin the major product is deoxyartemisinin, 11, whereas acetoni-trile with ferrous chloride gives 17 as the major product while 11 is not formed at all. In the less polar THF the Fe-O bond is stronger than it is in the more polar acetonitrile so that the radical substitution reactions which lead to products 14,15, and 17 are not favoured since the Fe-O bond must be broken. Instead, /3-scission (C3-02), is more likely and hence the pathway to deoxyartemisinin, 11 is encouraged. The reversibility of many of the reactions, especially the interchange between radicals 1A and IB is a key feature of the scheme. Hence, the formation of 11 will result in the removal of 9, which in turn will drive the reaction via radical 1A.
When acetonitrile is used with ferrous chloride, /3-scission (C3-02), which may be considered to be the elimination of an O atom radical, is not favoured because in more polar solvents the oxygen, (as Fe-OR) is more like a solvated alkoxide anion making this process very difficult. Conversely, the favouring of radical substitution results in the production of 17 along with a smaller proportion of 15. When aqueous acetonitrile with ferrous sulphate is used, the products are similar but 15 is the major product together with a smaller amount of 17 i.e. in this case the reaction proceeds mainly via radical 1A rather than via IB as in the former case where acetonitrile and ferrous chloride were used. Again, it is suggested that this finding can be explained in terms of the effects of the solvents and anions and provided that the conversions of 1A to 9 and of 16 to 17 are considered to be the rate limiting steps in the 1A and IB pathways repectively. With acetonitrile/ferrous chloride, the solvent will favour and therefore accelerate reactions 9 to 14 and 16 to 17. Since only the rate limiting step of the IB pathway is affected, the result is 17 as the major product.
When aqueous acetonitrile/ferrous sulphate is used, the solvent will have a greater accelerating effect than acetonitrile alone but this will be counteracted by the sulphate anion because reactions of artemisinin with ferrous sulphate have been shown to be much slower than those with ferrous chloride which is the most reactive salt. The sulphate anion reduces the ease with which Fe2+ can deliver an electron to the peroxy bridge to form 1A or IB, and that with which Fe3+ can receive an electron in the radical substitution reactions. In the presence of sulphate anion, reaction 16 to 17 will therefore be slowed as compared with when the chloride anion is used, while the conversion of 1A to 9 will not be affected; hence the reaction is able to proceed via 1A to give 15 as the major product.
Application of the Scheme to Artemisinin Derivatives
With dihydroartemisinin, 2, artemether, 3, and artesunate, 5, the major products formed by reaction of the above with ferrous sulphate in aqueous acetonitrile are analogous to those formed with artemisinin i.e. identical to 15 and 17 except for the substituent at C-10. In the case of artesunate, a significant amount of the dihydro-
analogue of 15 was also formed while artemether gave rise to only a very small amount of this derivative.
In contrast, when deoxoartemisinin, 8 (Figure 1), was reacted with ferrous bromide in THF, the major product was the C-10 deoxo-analogue of 19 together with a small quantity of deoxo-deoxyartemisinin i.e. 11 without the C-10 carbonyl (Avery et al., 1996). The pathway followed via 1A leading to 11 is therefore similar to that which occurs with artemisinin 1 under the same conditions (see above). However, with deoxoartemisinin the route via IB is favoured and the major product is the deoxo-analogue of 19. The formation of 17 will be disfavoured since radical substitution at oxygen is relatively slow in THF but, alternatively, 18 may arise by C3-013 scission. There are two possibilities for 18; C12~C12a scission with the loss of Fe2+ to give 19 or C12-011 scission to form 20, but only the latter appears to take place with artemisinin. With deoxoartemisinin, 8, the reverse is found, hence the deoxo-analogue of 19 forms rather than the deoxo-analogue of 20. A likely explanation for this is that a carbonyl at C-10 (as found in artemisinin), favours C12-OH scission because in the resulting radical, 20, the unpaired electron can be delocalised by the carbonyl and hence it more readily gives rise to other products than does 19.
It can now be seen that this scheme as proposed by Wu et al. (1998), provides an explanation for the experimentally observed finding that the nature of the substituent on C-10 affects the nature of the products even though it is some distance from the peroxy bridge of the molecule (although it is important to note that the replacement of the carbonyl at C-10 with a methylene group will affect the molecule as a whole as the ring system is strained). By analogy, other artemisinin derivatives which do not possess a C-10 carbonyl such as those mentioned above would be expected to behave similarly to deoxoartemisinin under the same reaction conditions. The reason why the above reactions follow the IB route may be that the C3-C13, C12-C12a and Fe-Ol scissions are able to occur rapidly at a rate significantly higher than the reactions of the 1A route leading to 11.
While the above experiments have provided a valuable insight into the interactions of iron compounds with artemisinin, the relationship between the latter and the reaction which takes place with haem is not well understood. Experiments in which artemisinin is reacted with haem (Fell) are complicated by the requirement to prepare haem Fell in situ from haem (Felll), e.g. by using a thiol. The latter, being a good hydrogen atom donor (i.e. radical scavenger), may enhance intermolecular H-abstraction thus increasing the formation of 11 via 12, hence changing the product ratio; in addition, the thiol as well as the haem may form adducts with the products, thus interfering with their recovery. The use of imidazole to complex the iron and thus mimic haem has been reported by Haynes and Vonwiller, (1996a); with ferrous chloride, imidazole and acetonitrile, 17 was the major product but a significant amount of 15 and a small quantity of 11 was also formed; Jefford et al. (1996), obtained a similar result without the imidazole. Haemin/benzylmercaptan in THF also gave rise to 17 as the major product with a little 15 and only a trace of 11, (Posner et al., 1995), whereas ferrous bromide in THF yields 11 as the major product (see above).
While the above evidence suggests that the reaction of artemisinin-like antimalarials with haem generates one or more cytotoxic species which are responsible for the antiplasmodial action of these agents, the identity of the "killer" molecule(s) remains speculative. Species postulated to fulfill this role are carbon-centred radicals, the high valence iron species Fe(IV)=0 and the epoxide 14.
The involvement of free radicals in the antimalarial action of artemisinin is suggested by the finding that free-radical scavengers such as ascorbic acid and vitamin E («-tocopherol) antagonise its action both in vitro (Krungkrai and Yuthavong, 1987, Meshnick et ah, 1989) and in vivo (Levander et ah, 1989), while oxidant drugs such as miconazole and doxorubicin have been shown to potentiate its action (Krungkrai and Yuthavong, 1987). In addition, a-tocopherol inhibited the oxidation of free thiol groups in isolated erythrocyte membrane proteins which occurs during the reaction of artemisinin with haem (Felll) (Meshnick et al., 1993). Further, lipid peroxidation end products have been detected in artemisinin treated parasites (Meshnick et al., 1989). Recently, Wu et al. (1998), have shown that a secondary radical is formed when artemisinin is reacted with ferrous sulphate in aqueous acetonitrile (see below). The radical was trapped by adding the spin trapping reagent 2-methyl-2-nitropropane to the reaction mixture and recording the electron spin resonance spectrum. This finding is consistent with the formation of a C-4 radical although there is no evidence to show that the trapped radical contained iron.
As discussed previously and shown in scheme 1, the nature and relative proportions of the degradation products arising from the reaction of iron with artemisinin or its derivatives is dependent upon which of the two main pathways is followed. Both of the postulated pathways involve the formation of carbon-centred radicals, but these are secondary in the case of the route via 1A, (compound 9) and primary in one branch only of the route via IB (compound 16). From the experiments described above, it is clear that the route taken is greatly influenced by the nature of the iron salt and the conditions of the reaction.
In order to determine which of the two routes is likely to be more important in vivo, Posner et al. (1994), have examined the structure activity relationships of a number of derivatives of the trioxane alcohol, 21, which were designed so that the 1,5-H shift necessary for the route via 1A is either blocked or enhanced.
In artemisinin-like compounds the 1,5-H shift can only take place if there is an a-hydrogen at C-4, i.e. on the same side of the ring as the (cleaved) peroxide group. Inverting the configuration of C-4 in 21 to give 22 which has a C-4a methyl group, resulted in more than 100-fold loss of in vitro activity against P. falciparum suggesting that the route via 1A is far more important in the mode of action of trioxane alcohols than the route via IB. Further, iron degradation experiments showed that an analogue in which both C-4 hydrogens were replaced by methyl groups, 23, (and which also had low activity against P. falciparum), gave rise only to the ring contracted ester analogous to 17, i.e. degradation occured only via the IB route (Posner et al., 1994), whereas the potent trioxane alcohol, 21 gave rise to analogues of both 15 and 17 as expected (Posner and Oh, 1992). The loss of the in vitro activities of 22 and 23 were not, therefore due to steric effects preventing degradation via the IB pathway.
The synthesis and structure-activity relationships of a further series of C-4 substituted 1,2,4-trioxanes, (Posner et al., 1995b) and reviewed in Cumming et al. (1997), confirmed the above; these compounds were C-4/3 substituted so that the 1,5-H shift results in the formation of tertiary C-4 radicals which are more stable than the corresponding secondary C-4 radicals. Some of these compounds, e.g. 24, showed enhanced antiplasmodial activity compared to the parent (unsubstituted at C-4), but surprisingly, those predicted to yield the most stable C-4 radicals, e.g. 25, were less active and appeared to give rise to degradation products via IB. Taken together however, the above results suggest that the pathway via 1A is important for the activity of artemisinin-like compounds in malaria parasites and therefore that the 1,5-H shift which leads to secondary carbon centred radicals may be essential for the generation of highly active derivatives.
In the route via IB, the formation of a primary C-4 radical 16 resulting from the cleavage of C3-C4 in radical IB is proposed. However, no evidence of primary radical formation was found in the artemisinin/ferrous sulphate/aqueous acetonitrile reaction, but in this system the 1A route predominates so that 9 rather than 16 would be expected to be the main radical produced. However, Robert and Meunier, (1997), have shown that a primary C-4 radical is likely to be formed when artemisinin reacts with meso-tetraphenylporphyrin co-ordinated to manganese (II). (This haem model was chosen since, like artemisinin it is non-polar, and the manganese ion can easily be removed so that characterisation of adducts formed is less difficult). The isolation of a covalent adduct in which the C-4 of artemisinin was bonded to a porphyrin ring provides evidence for the existence of a primary C-4 radical analogous to 16 in scheme 1.
Evidence for the involvement of high-valent Fe(IV)=0 species
Experimental results which support the formation of high-valent Fe(IV)=0 species have been reviewed by Cumming et al. (1997). As shown in scheme 1, the formation of epoxide 14 from 9 may occur via two pathways. Radical substitution at 0-2 with the release of Fe2+ leads directly to 14 while C3-02 scission gives 10 with the release of Fe(IV)=0, (Fe2+=0). Epoxide 14 may then arise by "rebound epoxidation" by the high valent iron species which is itself reduced to Fe2+. Support for the existence of high valent iron species has been obtained by the use of "reporter" reactions. In the presence of iron and artemisinin (but not one or the other), hexamethyl Dewar benzene was rearranged to hexamethylbenzene, (Traylor and Miksztal, 1987), oxidation of methylphenylsulphide to the corresponding sulphoxide and oxidation of tetralin to 1-hydroxytetralin were observed (Groves and Viski, 1990). The oxidation reactions were not affected by the removal of oxygen from the solvents, eliminating the possibility that molecular oxygen was responsible for the oxidations.
When artemisinin is degraded with iron, 4-hydroxydeoxyartemisinin, 15, is formed as a single 4a-stereoisomer by the action of the a-hydroxyl group on the a-epoxide, 14 which could arise by one or both of the above routes (Posner et al., 1995). In contrast, Posner and Oh, (1992), found that the degradation products of trioxane tosylate, 27 contained both C-4a and C-4/3 hydroxytrioxanes. This mixture could only have arisen via a mixture of a- and ^-epoxides which in turn must have been produced by an intermolecular reaction since the 3a-iron oxy bond cannot access the jS-face of the molecule i.e. the high valent iron species can react on either face. Cummings et al. (1997), consider that, "taken as a whole the above strongly implicate involvement of a high-valent iron-oxo intermediate in the iron-induced degradation of artemisinin and its analogs". However, by analogy with trioxane 27, we would expect artemisinin to give rise to both C-4 a and C-4/3 hydroxy isomers if a high valent iron-oxo intermediate is involved, but, as stated above only the 4a isomer has been reported.
In support of the involvement of high valent Fe(IV)=0, haem iron-oxo species are known to be involved in the action of horse radish peroxidase and cytochrome P-450 enzymes; also, a number of non-haem iron containing mono-oxygenases have been shown to effect the epoxidation of olefins which suggests that high valent iron-oxo species could be functioning as intermediates (Cumming et al., 1997, and references therein).
In order to provide further evidence in support of high-valent Fe=0 as a reaction intermediate Posner et al. (1996), synthesised a number of 1,2,4-trioxane analogues with various substituents at C-3 and C-4 which were designed either to facilitate or inhibit the release of Fe(IV)=0 following the reaction of the trioxane with Fe(II). The substitution into the molecule of a (trimethyltin)methyl group at C-4 (compound 26), as a better radical leaving group than Fe(IV)=0 resulted in a ten-fold reduction of antiplasmodial activity compared to an analogue with a C-4/3-(trimethylsilyl)methyl group, 25. When the former was degraded in the presence of iron the product contained 15% of a 4-methylene deoxytrioxane derivative showing that trimethyltin was lost by at least a proportion of the compound and no rearrangement of hexamethyl Dewar benzene was observed. However, it is possible that the presence of the tin in the molecule may have reduced the antiplasmodial activity due to other reasons; to assess this possibility, Posner et al. (1996), prepared a tin-containing ether of dihydroartemisinin, 28, and reported that, as this compound has measurable antiplasmodial activity this indicates that the presence of tin does not necessarily destroy a trioxanes activity; however, it was more than 100-fold less active than artemisinin so that, on the contrary, it is quite possible that the presence of tin per se may well reduce antiplasmodial activity, especially as dihydroartemisinin, 2, is more active than artemisinin itself.
Better evidence for the formation of Fe(IV)=0 was obtained with two C-3 substituted analogues, 29, 30, which were designed so that the postulated C-4 radical intermediate (analogous to 9 in scheme 1) could undergo a second 1,5-H shift leading to a tertiary radical, (from 29), or an even more stable benzylic radical, (from 30) so that the release of Fe(IV)=0 would be less favoured (scheme 2). Unexpectedly, these compounds were found to be about 6-fold more active than their 3/3-methyl and 3/3-ethyl analogues against P. falciparum in vitro but, significantly, on reaction with ferrous bromide in tetrahydrofuran rearrangement of hexamethyl Dewar benzene was seen suggesting the presence of a high-valent iron oxo intermediate. The potential of the above compounds to undergo a second 1,5-H shift appears to have failed to prevent /3-scission of Fe(III)-0- (<-» Fe(IV)=0) perhaps because the former is too slow to compete with the latter; the enhanced antiplasmodial activity is suggested to be due to another effect e.g. increased transport across membranes.
Derivatives with either vinyl or phenyl substituents at C-3 designed to conjugate with, and hence stabilise the double bond formed when Fe(IV)=0 leaves were found to have similar and markedly increased activity (respectively) compared to an analogue possessing a C-3/3 ethyl group, but on reaction with iron only the vinyl trioxane produced rearrangement of hexamethyl Dewar benzene to hexamethyl-benzene (Posner et al., 1996).
Wu et al. (1998), have argued against the above evidence for the involvement of high-valent Fe=0, by suggesting that the "reporter" reactions do not necessarily require Fe(IV)=0 as an oxidant. For example, the rearrangement of hexamethyl Dewar benzene may be initiated by the loss of one electron which could be received by radical 1A; this would lead to an increased proportion of 12 and hence 11 as is actually observed (Posner et al., 1995). Wu et al. (1998), also propose that since, in aqueous media, product 11 is not formed, this is evidence that the C3-02 /3-scission required for the release of Fe(IV)=0 does not occur; however, as discussed above, and shown in scheme 1, C3-02 /3-scission may lead to 15 so that absence of 11 does not preclude C3-02 scission. In addition, the latter authors point out that nothing is known about the oxidative ability of Fe(IV)=0 which itself is a speculated species; in the parasite, by analogy, haem Fe(IV)=0 would be present and it cannot be assumed that this would behave in the same way as the iron in non-haem containing sytems.
Scheme 2 Proposed products resulting from the Fe(II) induced cleavage of compounds 29 and 30. (Adapted from Posner et al., 1996).
Taken together, the above indicates that while there is some experimental evidence in support of the role of high valent iron-species in the antiplasmodial action of artemsinin-like compounds, this is by no means proven and there are some anomalous results which have yet to be explained.
Posner et al. (1995), proposed that 14 may act as an alkylating agent but Wu et al., (1998) consider that rearrangement e.g. to 15 is more likely as O-l is well placed to compete with other nucleophiles which may attack C-3. If the non-peroxidic 0-13 of artemisinin is replaced by a methylene group, the resulting carba-analogue is 25fold less active against P. falciparum in vitro (Avery et al., 1995). A possible explanation for this finding is that the carba-analogue of 10 would be less susceptible to epoxidation by Fe(IV)=0 since it is not as electron-rich as 10 itself (Cumming et al., 1997); thus the reduced activity of the carba-analogue of 10 may be consistent with a reduced ability to form the epoxide (i.e. the carba-analogue of 14). On the other hand, Avery et al. (1996), have shown that the presence of an epoxide does not necessarily enhance antimalarial activity since the epoxide, 31, was devoid of anti-
Was this article helpful?