Iron-sulphur clusters and the problem with oxygen


  • James A. Imlay

    Corresponding author
    1. Department of Microbiology, University of Illinois, Urbana, IL 61801, USA.
      *E-mail; Tel. (+1) 217 333 5812; Fax (+1) 217 244 6697.
    Search for more papers by this author

*E-mail; Tel. (+1) 217 333 5812; Fax (+1) 217 244 6697.


During the first billion years of life on the Earth, the environment was anaerobic. Iron and sulphur were plentiful, and they were recruited in the formation of iron-sulphur (Fe-S) clusters within ancient proteins. These clusters provided many enzymes with the ability to transfer electrons; to others they offered a cationic feature that tightly bound oxyanionic and nitrogenous metabolites. Still others acquired a crystallizing surface around which polypeptide could fold to establish a three-dimensional structure. However, the subsequent oxygenation of the Earth's atmosphere by photosynthetic organisms created a threat to cluster-dependent proteins that still has not been fully resolved. By oxidizing environmental iron, oxygen limits its bioavailability, requiring that organisms employ complex schemes with which to satisfy their iron requirement. More directly, oxygen species convert exposed Fe-S clusters to unstable forms that quickly decompose. Some microbes responded to this dilemma by retreating to anaerobic habitats. Others abandoned the use of low-potential electron-transfer pathways, which rely upon the least stable cluster enzymes, and developed antioxidant strategies to protect the remainder. These adjustments were only partially successful: largely because of their reliance upon Fe-S clusters, aerobes remain vulnerable to iron restriction and oxidative stress, features that higher organisms exploit in defending themselves against bacterial pathogens. Thus, the history of Fe-S clusters is an unusual one that has profoundly shaped contemporary microbial ecology.

Iron-sulphur cluster structures

In contemporary organisms, iron-sulphur (Fe-S) clusters are perhaps the most abundant and the most diversely employed enzymatic cofactor. The simplest Fe-S centre is comprised of a single iron atom liganded within a polypeptide by four cysteine residues (Fig. 1). The more common Fe-S clusters have two, three or four iron atoms coordinated to polypeptide residues and bridged by inorganic sulphide. More complex structures are assembled in specialized redox enzymes through metal substitution and/or bridges between the simpler cluster modules (Rees, 2002). Cysteine is by far the most common protein ligand, in accordance with the strong affinity of iron for thiolate residues, but a variety of others – including histidine, aspartate and even arginine (Berkovitch et al., 2004) – have been observed.

Figure 1.

Structures of common iron-sulphur clusters.
A. Rubredoxin-style iron centre. CS represents coordinating cysteine residues from the polypeptide backbone. The iron atom may be in the +2 or +3 valence.
B. [2Fe-2S] cluster. Typical stable cluster oxidation states are +1 and +2 (with the charges of coordinating cysteinate residues not being considered).
C. [3Fe-4S] cluster. Stable oxidation states are 0 and +1.
D. [4Fe-4S] cluster. Stable oxidation states are +1 and +2 for ferredoxin-type clusters and +2 and +3 for ‘HiPIP’ clusters. Electrons may be delocalized, so that the valences of individual iron atoms lie between the ferrous and ferric forms.

The structures depicted in Fig. 1 have substantial inherent stability in anaerobic solution, and analogous structures can be chemically created from ferrous iron, sulphide salts and organic thiolate compounds (Rao and Holm, 2004). Primordial clusters likely assembled spontaneously on protein templates. Facile ligand-exchange reactions, coupled to electron transfers from biological reductants, enable the interconversion of [2Fe-2S], [3Fe-4S] and [4Fe-4S] clusters, as dictated by the spacing of amino-acid ligands and packing constraints imposed by the surrounding polypeptide (Plank et al., 1989; Golinelli et al., 1998). Thus, unlike most contemporary organic cofactors, Fe-S clusters are constructed of simple compounds that were abundant in primordial environments and that could assemble spontaneously into extant polypeptide structures.

The biochemical utility of clusters rests upon two features: their ability to accept and donate electrons, and their tendency to bind the electron-rich oxygen and nitrogen atoms of organic substrates. Both behaviours are influenced by the solvent exposure and electrostatic environment of the cluster, and so it is easy to imagine that their catalytic function in ancestral proteins was rapidly improved and diversified through relatively simple changes in local polypeptide context. The variety of structures and uses of clusters that we see today is the outcome of that facile evolutionary process.

In contemporary organisms, cluster assembly and insertion into apoproteins is catalysed by dedicated enzymatic machinery. The Nif, Isc and Suf cluster-building systems are dispersed through the microbial biota, with many organisms possessing more than one (Tokumoto et al., 2004). The Isc machinery has been most closely studied. It appears to construct nascent clusters on a scaffold protein and then transfer them into recipient apoproteins. The biochemical mechanism of this process, and the distinct roles of the three systems, are currently the subjects of intense investigation. (For a review, see Johnson et al., 2005.)

The utility of clusters in an anaerobic world

Electron-transfer reactions

Iron-sulphur clusters serve most prominently in redox enzymes. In these proteins the clusters comprise a wire that delivers electrons one at a time between redox couples that are physically separated. Figure 2A depicts the arrangement of three redox clusters in fumarate reductase (Frd), an anaerobic respiratory enzyme that transfers electrons from membrane-bound menaquinone to cytosolic fumarate. As is typical, the clusters are spaced 10–14 Å apart, which is sufficiently close to enable rapid cluster-to-cluster electron hopping (Page et al., 2003) but far enough to minimize the number of clusters that are needed to cover the distance.

Figure 2.

Roles of clusters in redox enzymes.
A. Fumarate reductase (Iverson et al., 1999). Electrons move from menaquinone (blue, QH2) through the Fe-S cluster wire (red) to FAD (green), from which an electron pair is transferred to fumarate. Succinate dehydrogenase (Sdh) has a very similar structure, but its cluster and flavin potentials are altered to favour electron movement from succinate towards ubiquinone. Sdh also has a haem (not shown) that may facilitate electron exchange between the [2Fe-2S] cluster and the quinone.
B. Pyruvate:ferredoxin oxidoreductase (Chabriere et al., 1999). Two electrons move consecutively from the thiamine moiety (blue) through the cluster wire to a ferredoxin (Fdx) binding site, at which they hop to a ferredoxin cluster. In the Desulfovibrio africanus enzyme, a C-terminal extension (depicted here only on the right subunit) occludes the superficial cluster of the partner subunit and protects it against adventitious oxidation by molecular oxygen.

Comparative studies suggest that Frd may have arisen from modular components that evolved independently (Bossi et al., 2002). Aspartate:fumarate oxidoreductase is a soluble protein consisting of a single domain homologous to the flavoprotein subunit of Frd. Thus, the Fe-S subunit can be construed as a module that attached an ancestral fumarate-reducing domain to the membrane so that it could utilize dihydromenaquinone as a new electron donor. It seems likely that this modularity lent itself to the rapid evolution of complex redox enzymes: because the Fe-S wire need not interact intimately with either active site, its recruitment did not require that they be remodelled in a way that would compromise their catalytic efficiency.

Iron-sulphur clusters have several attributes that ensured that they be the redox moieties of choice in the evolution of such enzymes. Chief among these is their unusually wide range of reduction potentials, from −0.6 V to +0.45 V (Capozzi et al., 1998). The rate of electron transfer is optimal if the electron affinities of the connecting carriers are close to those of the enzyme substrates, so that endergonic transfers are minimized. Fe-S potentials are influenced by modest changes in protein structure that establish local residue charges, dipole interactions with the polypeptide chain, and hydrogen bonds between nearby residues and the cluster sulphur ligands (Capozzi et al., 1998; Babini et al., 1999). Thus, simple evolutionary steps could fine-tune Fe-S clusters for roles in many disparate redox pathways.

Pyruvate:ferredoxin oxidoreductase (PFOR) (Fig. 2B) exploits the ability of clusters to operate at low potentials. The clusters of this enzyme deliver electrons from its buried thiamine cofactor to the protein surface (Chabriere et al., 1999), where they are transferred to ferredoxin, a small soluble Fe-S protein. Ferredoxin carries them in turn to the surface of hydrogenase (Peters et al., 1998), from which another Fe-S wire leads to a dinuclear iron site, at which protons are reduced to molecular hydrogen. All the clusters involved in this chain are poised at low potential, so that the electrons can be readily transferred to the ultimate acceptor, protons (Em ∼ −0.440 V). Most other redox moieties that are found in contemporary organisms – NAD(P), haems, quinones, manganese, copper and flavins – function at higher potentials and would not be suitable for this chain. Carbon dioxide, dinitrogen and sulphur species are among other low-potential electron acceptors that were central to anaerobic metabolism in the ancient world, and their complementary reduction systems employed Fe-S clusters.


A completely different role for Fe-S clusters is manifested by a family of dehydratases, of which aconitase (Fig. 3A) is the most-studied member. In these enzymes, only three of the four iron atoms have cysteine thiolate ligands; the fourth iron atom is solvent-exposed within the active-site pocket and has a water molecule loosely bound at its fourth coordination site (Lauble et al., 1992). Binding of substrate occurs via additional coordination of this iron atom by both a carboxylate residue and the hydroxyl group that is to be abstracted. This association relies upon the ability of iron to shift smoothly from tetrahedral to octahedral (six-coordinate) geometry. This feature allows strong bidentate ligand binding without the energetic expense of bond-breaking, and it is the non-redox property of iron that is most widely exploited in catalysis. A nearby base then deprotonates a methylene group at the same time that the cationic iron atom, acting as a Lewis acid, withdraws the anionic hydroxyl substituent. In tandem these two steps accomplish the net dehydration of substrate. Thus, the role of the cluster is not to transfer electrons at all; instead, it both assists in substrate binding and provides a local positive charge to effect catalysis. Enzymes of this type are widespread in catabolic and biosynthetic pathways.

Figure 3.

Roles of clusters in substrate binding.
A. Aconitase (Lauble et al., 1992). Cluster geometry shifts from 4- to 6-coordinate upon citrate ligation, and the coordinating iron atom subsequently abstracts hydroxide anion during dehydration.
B. Proposed S-adenosylmethionine binding and scission in lysine-2,3-aminomutase (Chen et al., 2003). The [4Fe-4S] cluster coordinates SAM, transfers an electron to it, and then may coordinate the liberated thiolate. The figure does not show the cosubstrate that the adenosyl radical attacks.

Radical enzymes

The aconitase-family enzymes can only dehydrate substrates that contain an activating carbonyl adjacent to the site of deprotonation. Aliphatic substrates are much more resistant to derivatization, and most aerobic metabolic pathways are configured to circumvent the need for such reactions. However, aliphatic metabolites are unavoidable in fermentation pathways that are designed either to degrade or generate highly reduced substrates. To solve this problem, anaerobic microbes often employ radical-based enzyme mechanisms. Buckel and colleagues have identified two distinct dehydratase families that use this strategy to dehydrate aliphatic substrates (see Kim et al., 2004; Martins et al., 2004). Interestingly, both classes utilize Fe-S clusters – in one case as a low-potential electron donor to create the catalytic radical, and in the other case as a substrate-binding Lewis acid to abstract the hydroxyl group.

Aliphatic substrates can also be activated for substitution reactions by the abstraction of a hydrogen atom, a high-energy process catalysed by adenosyl radicals that are formed from either B12- or S-adenosylmethionine (SAM). The activating enzymes of pyruvate:formate lyase and anaerobic ribonucleotide reductase, biotin and lipoate synthases, coproporphyrinogen III oxidase and lysine 2,3-aminomutase are all well-studied members of the SAM radical superfamily. Their mechanism requires that an exposed iron atom of the [4Fe-4S] cluster shift towards octahedral geometry as it ligands the amino nitrogen and carboxylate group of SAM (Fig. 3B) (Layer et al., 2003; Berkovitch et al., 2004). An electron is then transferred from the low-potential cluster onto SAM, a step that is energetically difficult and is believed to be driven in part by the bonding of the liberated sulphur atom to the remaining coordination site of the iron (Chen et al., 2003). Thus, this cluster combines the two roles shown previously, serving both as a facile ligand for substrate and as a redox catalyst.

Biotin synthase and lipoate synthase are SAM radical enzymes that exhibit one final twist to Fe-S biochemistry. Their role is to catalyse the insertion of sulphur atoms into aliphatic substrates. After the adenosyl radical activates their organic substrates by hydrogen-atom abstraction, a second Fe-S cluster on the enzymes apparently donates its inorganic sulphur atoms for insertion (Jarrett, 2005). The mechanism of this cannibalization process, and the method by which the synthase cluster is subsequently regenerated, are not yet understood.

Reformatting anaerobic metabolism to exploit oxygen

The preceding discussion was intended to emphasize the traits of Fe-S clusters that ensured the dispersion of Fe-S-based enzymes throughout the anaerobic world. Then, approximately 2.75 billion years ago, cyanobacteria took the epochal step of evolving photosystem II. The immediate effect of this invention was to liberate these bacteria from a need for external electron donors. Oxygen concentrations are believed to have remained very low over the subsequent two billion years, limited both by the paucity of oceanic phosphorus to support oxygenic photosynthetic bacteria and by oxygen removal through reaction with dissolved ferrous iron and sulphides (Bjerrum and Canfield, 2002). Subsequent changes in ocean conditions allowed oxygen to accumulate. The evolutionary stresses that resulted were among the most profound since early biotic history.

On the plus side, microbes were presented with an opportunity to use oxygen as a terminal oxidant. This adaptation required surprisingly little molecular evolution: the creation of cytochrome oxidase was sufficient. When implanted into extant respiratory chains, both the quinone- and cytochrome c-dependent oxidases redirected electron flow to oxygen, while the ancestral NADH dehydrogenases, hydrogenases and bc1 complexes served as upstream electron donors. In some cases, this new metabolism resulted in a reversal of the physiological direction of electron flow through these enzymes. For example, Frd now served to transfer electrons from succinate to oxidized quinones (Fig. 2A). While contemporary Frd can catalyse this reaction, its catalytic efficiency was enhanced by modifications that elevated the potentials of its Fe-S clusters, favouring electron movement from the flavin towards the quinone-binding site and creating the enzyme that we now designate as succinate dehydrogenase (Yankovskaya et al., 2003).

Thus, the adaptation of anaerobic electron-transport chains for an aerobic habitat required remarkably little de novo evolution. For this reason, Fe-S clusters were retained as the primary carriers of electrons within extant redox enzyme complexes. Furthermore, aerobes continue to use the non-redox catabolic and biosynthetic pathways that they inherited from their anaerobic ancestors, ensuring the maintenance of the other Fe-S enzyme families, too.

The trouble with oxygen (I): iron availability

Molecular oxygen is a reactive chemical, and its essential chemical behaviour is the oxidation of other molecules. Molecular-orbital rules dictate that molecular oxygen accept electrons one at a time rather than in pairs (Naqui and Chance, 1986). This restriction ensures that oxygen does not react with most organic biomolecules, but it allows it to oxidize transition metals, because they are good univalent electron donors. A consequence is that oxygen chemically oxidizes ferrous iron in the environment to its ferric form, which rapidly precipitates (as ferric hydroxide) or forms insoluble complexes with anionic salts. The upshot is that as oxygen accumulated, iron became a limiting nutrient in many aerobic habitats. Because bacteria require near-millimolar concentrations of intracellular iron (Outten and O’Halloran, 2001) – primarily for Fe-S assembly, although also for haem synthesis – diminishing iron levels posed a serious challenge for early aerobes.

Microorganisms learned to tackle this problem by excreting siderophores, soluble organic molecules that avidly bind iron and can leach it off mineral precipitates (Wandersman and Delepelaire, 2004). The resultant iron-siderophore chelates are large and cannot pass through porins; therefore, gram-negative bacteria took the additional step of evolving dedicated outer-membrane iron-siderophore transporters that are coupled to the protonmotive force by the TonB system (Wiener, 2005). Once inside the cell, the tight iron-siderophore chelate can only be dissociated by siderophore hydrolysis, making them a uniquely expensive single-turnover delivery system (Brickman and McIntosh, 1992). To minimize the cost, these microbes inactivate their siderophore system and employ simpler low-affinity transport systems whenever they enter the few habitats where iron is plentiful. This control occurs through activation of the Fur protein, which represses the expression of siderophore biosynthesis and uptake genes (Neilands, 1993). Under iron-replete conditions, Fur also activates the synthesis of ferritins, which store excess iron in anticipation of future iron shortages. Collectively, these strategies comprise a remarkably complex and unprecedented adaptation to a unique problem.

Interestingly, when Bacillus subtilis and Escherichia coli cannot acquire enough iron to insert an Fe-S cluster into aconitase, the aconitase apoenzyme apparently serves as an additional RNA-binding protein that further modulates the cellular response (Alen and Sonenshein, 1999; Tang and Guest, 1999). This strategy was first discovered in eukaryotes (Kaptain et al., 1991), but evidently it initially evolved in bacteria.

What happens when even siderophore systems cannot deliver enough iron to satisfy the cellular demand? Recent discoveries in E. coli indicate that this organism responds by suppressing the synthesis of its most abundant Fe-S enzymes (McHugh et al., 2003). In iron-replete cells metallated Fur blocks transcription of a small RNA called RyhB. When cells are starved for iron, Fur is demetallated and RyhB is synthesized. Acting as an antisense RNA, RyhB then stimulates the degradation of transcripts that encode Fe-S enzymes such as succinate dehydrogenase and NADH dehydrogenase I (Masse and Gottesman, 2002). Iron demand is thereby reduced, but at a price: TCA-cycle flux diminishes – the cell can no longer catabolize succinate, for example – and respiration is re-directed through enzymes that do not couple electron flux to the generation of a membrane potential. Thus, this metabolic strategy is less energy-efficient, but it has the virtue of allowing iron-poor cells to grow as long as fermentable carbon sources are available. The benefit is that, by suppressing the synthesis of high-titre Fe-S enzymes in central metabolism, the cell re-directs what little iron it can acquire to the indispensable iron enzymes that belong to biosynthetic pathways. In this way E. coli begins to resemble lactic acid bacteria, whose fermentative style of metabolism seems wasteful but allows them to thrive in habitats in which iron is scarce. An analogous control system has recently been found in yeast (Puig et al., 2005).

Thus, the high iron demand that modern microbes inherited from their anaerobic ancestors does not suit the aerobic world. The many adjustments that have been made are expensive, and they still bestow only a limited capacity to tolerate iron deprivation. Most famously, the struggle to import sufficient iron is crucial to the success of pathogens: mammalian hosts employ proteins that sequester iron and bacterial siderophores as a key tactic to suppress the growth of invading bacteria (Ward and Conneely, 2004; Flo et al., 2004).

The price of Fe-S clusters (II): vulnerability to oxidants

Virtually all organisms struggle to grow when the ambient oxygen concentration is higher than that which they normally encounter in their native habitats. It turns out that Fe-S clusters are a big part of the problem.

Oxygen toxicity in aerobes

Early studies revealed that hyperoxia specifically blocks the ability of E. coli to synthesize branched-chain amino acids (Boehme et al., 1976). However, it is likely that molecular oxygen was not the direct toxin. Because oxygen can adventitiously steal electrons from the reduced flavins of redox enzymes, high oxygen concentrations favour the rapid formation of intracellular superoxide and hydrogen peroxide. In fact, branched-chain auxotrophy was also manifested at normal oxygen levels by E. coli mutants that cannot scavenge superoxide or hydrogen peroxide (Carlioz and Touati, 1986; S. Jang and J. A. Imlay, unpubl. data). These mutants additionally failed to catabolize carbon sources that are normally assimilated by the TCA cycle. By pursuing these clues, the Fridovich and Flint labs discovered that superoxide rapidly inactivates the [4Fe-4S] family of dehydratases, including key enzymes of the branched-chain and TCA pathways: dihydroxyacid dehydratase, aconitase and fumarase (Kuo et al., 1987; Gardner and Fridovich, 1991; Liochev and Fridovich, 1992; Flint et al., 1993). The damage occurs when superoxide directly oxidizes the Fe-S cluster, converting the [4Fe-4S]2+ form to an unstable [4Fe-4S]3+ state, which releases iron (Fig. 4). The resultant [3Fe-4S]1+ cluster lacks the catalytic iron atom, so that the enzyme is inactive and the pathway fails. Hydrogen peroxide oxidizes these clusters in similar fashion (Varghese et al., 2003).

Figure 4.

Cluster destruction by oxidation. Univalent oxidants, including superoxide, hydrogen peroxide, molecular oxygen and peroxynitrite [collectively denoted (O)], convert the exposed [4Fe-4S]2+ cluster of aconitase-family enzymes to the unstable +3 oxidation state. The cluster then spontaneously decomposes to the inactive [3Fe-4S]1+ form. Further degradation gradually occurs in vitro and in vivo, although the rate and ultimate products are not well established and probably vary from protein to protein.

It is not surprising that small oxidants can enter the active sites of dehydratases and make contact with the Fe-S cluster – which, after all, is positioned to bind dissolved solutes. What is less obvious is why the oxidized 3+ cluster is unstable. After all, the clusters of high-potential [4Fe-4S] ferredoxins (HiPIPs) normally shuttle between 2+ and 3+ states without decomposing. The answer seems to be that the HiPIP clusters are sequestered from solvent. The key evidence is that their clusters tolerate polypeptide unfolding by guanidinium hydrochloride only when they are reduced, as exposure of the oxidized cluster to water results in rapid solvolysis (Bertini et al., 1997). Thus, the exposure of aconitase-dehydratase-class clusters to solvent, which is necessary for their function, endangers them for two reasons: it allows oxidants to contact them directly, and it destabilizes the resultant [4Fe-4S]3+ species.

The rate constants with which dehydratase clusters react with superoxide and hydrogen peroxide are extremely high: 3 × 106 M−1 s−1 and 4 × 103 M−1 s−1 respectively (Flint et al., 1993; S. Jang and J. A. Imlay, unpubl. data). Consequently, E. coli must synthesize enough superoxide dismutase, catalase and peroxidase to restrict superoxide to 10−10 M (Gort and Imlay, 1998) and H2O2 to 10−8 M (Seaver and Imlay, 2001). Even when oxidants are at such very low concentrations, the half-time of a dehydratase cluster is only about an hour, which is substantially less than the likely doubling time of the microbe in natural aerobic habitats. A further complication is that the iron atoms that are released upon cluster destruction can react with hydrogen peroxide to generate hydroxyl radicals, which cause substantial DNA damage (Liochev and Fridovich, 1994; Keyer and Imlay, 1996).

This vulnerability to oxidants has not gone unnoticed by competitors. Macrophages blast captive bacteria with H2O2, with phagosomal concentrations probably approaching 10−4 M. Plants produce H2O2 along the margins of wounds to deter microbial invaders. Many lactic acid bacteria – which themselves eschew the use of Fe-S enzymes – gain a competitive advantage by releasing H2O2 into their habitat to poison potential competitors. Finally, a variety of plants and microbes excrete antibiotics that, when ingested by their competitors, produce toxic doses of superoxide and hydrogen peroxide through redox-cycling reactions. Juglone (Inbaraj and Chignell, 2004) which is produced by walnut trees, and pyocyanin (Ran et al., 2003), which is excreted by Pseudomonas aeruginosa, are characteristic examples.

Target bacteria, of course, have in turn evolved measures to defend themselves against such assaults. Fittingly, they detect superoxide through its oxidation of a [2Fe-2S] cluster on SoxR protein (Ding and Demple, 1997; Gaudu et al., 1997). This transcription factor then activates a response that includes the induction of superoxide dismutase and the syntheses of a cluster-free fumarase isozyme and of an oxidant-resistant aconitase isozyme, thereby restoring some degree of TCA-cycle function. A drug-export system is also activated, presumably because redox-cycling antibiotics are the usual environmental sources of superoxide stress. H2O2 stress is sensed by the OxyR or PerR systems, which direct activation of about two dozen genes (Zheng et al., 2001; Helmann et al., 2003). Among these genes are ones encoding the Suf cluster-assembly complex, suggesting that Suf is important for Fe-S cluster assembly or repair during H2O2 stress (Outten et al., 2004). Simultaneously, Dps protein sequesters the iron that spills from damaged clusters, thereby minimizing the formation of hydroxyl radicals (Park et al., 2005). The roles of other Sox-, PerR- and OxyR-regulated genes have not yet been identified.

The SAM-superfamily enzymes – which use a solvent-exposed [4Fe-4S] cluster to bind SAM – are, predictably, rapidly inactivated when they are exposed to oxygen in vitro. However, both the lipoate and bioin synthases, at least, continue to function normally inside aerobic cells – even in mutants that cannot scavenge endogenous superoxide or hydrogen peroxide (A. Wu and J. A. Imlay, unpubl. obs.). This fact raises the possibility that cells have some mechanism that either protects or quickly repairs this particular class of Fe-S enzyme. Oxidants may also damage biomolecules other than dehydratase Fe-S clusters (Benov and Fridovich, 1999). Nevertheless, these clusters appear to be the primary targets of oxidative stress, and by employing them microbes have endangered their ability to thrive in aerobic habitats.

Oxygen toxicity in anaerobes

At one time it was suspected that the sensitivity of anaerobes to oxygen might be due to an inability to scavenge superoxide and H2O2– and therefore that the mechanism by which oxygen damages anaerobes might be the same as aerobes, only more so. However, that view now seems to be incomplete at best. For one thing, although early surveys suggested that anaerobes are deficient in catalase and superoxide dismutase activities, it is now recognized that many of these organisms use peroxidases and superoxide reductases (Jenney et al., 1999; Lombard et al., 2000) to accomplish the same purpose. We must look elsewhere to explain their sensitivity to oxygen.

Obligate anaerobes are qualitatively different from aerobes in that their metabolism operates upon highly reduced substrates. As we have seen, anaerobes are therefore compelled to use enzymes that contain low-potential redox moieties to deliver electrons to low-potential acceptors, and many employ organic-radical mechanisms to activate aliphatic substrates. Oxygen wreaks havoc upon these specialized chemistries.

Pyruvate:ferredoxin oxidoreductase (Fig. 2B) optimizes anaerobic sugar fermentations by allowing anaerobes to deliver excess reducing equivalents to protons, rather than dumping them back upon growth substrates. This strategy allows more growth substrate to be utilized for ATP production rather than for redox balancing. Yet PFOR is abruptly poisoned when cells are exposed to oxygen. How? A variety of data suggests that the [4Fe-4S] cluster nearest the enzyme surface is oxidized to the +3 state and destroyed. The most compelling evidence comes from the PFOR of Desulfovibrio africanus, which is unique in retaining its activity in aerobic solutions. This PFOR is structurally exceptional in having an extra domain that is positioned to occlude the terminal cluster. Deletion of this domain restores the degree of oxygen sensitivity that is typical of most other enzymes (Pieulle et al., 1997). Presumably this domain either blocks access of oxygen or diminishes solvent accessibility enough to suppress solvolysis of the overoxidized [4Fe-4S] cluster until it is rereduced during its catalytic cycle.

Importantly, PFOR is more easily oxidized by oxygen than by superoxide (Pan and Imlay, 2001), in marked contrast to the [4Fe-4S] dehydratases. Hydrogen peroxide and superoxide oxidize metals by inner-sphere mechanisms (Goldstein et al., 1993), which require that the oxidant directly bind the cluster in order to receive electrons from it. In contrast, it is likely that electrons can hop from the slightly buried PFOR cluster to nearby molecular oxygen, as they normally do to the cluster of ferredoxin. In this way molecular oxygen may oxidize a fully coordinated cluster, while H2O2 and O2 cannot. The physiological significance is that anaerobes that use PFOR – or other ferredoxin-interacting enzymes – cannot protect themselves from oxygen merely by synthesizing scavengers of superoxide and H2O2.

The Fe protein of nitrogenase and the ‘archerase’ enzymes that drive electrons onto flavin-radical dehydratases also rely upon Fe-S clusters near the protein surface (Kim et al., 2004). All these enzymes are rapidly inactivated by oxygen. While oxygen is not inherently a strong univalent oxidant (Em = −0.16 V), [4Fe-4S] clusters that operate with low +2/+1 midpoint potentials apparently also have low +3/+2 potentials that leave them vulnerable to over-oxidation, and enzymes that use these clusters are stable in aerobic environments only if the clusters are deeply buried within polypeptide. Many anaerobic respiratory pathways using superficial clusters are therefore oxygen-sensitive, while aerobic pathways, which employ enzymes like succinate dehydrogenase (Fig. 2A) with fully occluded clusters, are oxygen-tolerant.

A full accounting of obligate anaerobiosis should note that clusters are not the sole determinants of oxygen sensitivity. Some anaerobes also employ glycyl-radical enzymes, including pyruvate:formate lyase, as specialized catalysts of difficult reactions. Because oxygen itself is a radical species, it reacts rapidly with these exposed active-site radicals, forming a peroxy radical species that then cleaves the enzymes. Anaerobes have learned to anticipate this threat, and they use deactivating enzymes to reduce the radical to a stable (but inactive) glycyl residue whenever oxygen is sensed. SAM-dependent Fe-S enzymes reactivate these enzymes when anaerobiosis is restored.

Interestingly, facultative aerobes learned to use the reactivity of surface-associated Fe-S clusters as a mechanism to detect the presence of molecular oxygen. Fnr protein is a transcription factor that activates expression of anaerobic respiratory enzymes whenever E. coli is in an anaerobic habitat. When oxygen is present, however, a [4Fe-4S] cluster that holds the enzyme in its active dimeric structure is oxidized and decays to a [2Fe-2S] form (Khoroshilova et al., 1997). The enzyme then dissociates into monomers, and transcriptional activity is lost. This represents a third example – after SoxR and aconitase – in which evolution exploits the instability of clusters in order to detect iron restriction or oxidative stress.

Is evolution done?

This discussion has emphasized that aerobic organisms rely heavily upon a cofactor that evolved in an old environment and is not well suited for their new one. It is doubtful that Fe-S clusters could have emerged as central catalysts of metabolism had life evolved in an aerobic environment. However, the accumulation of oxygen occurred gradually (Canfield et al., 2000), perhaps over a billion-year period, allowing time for numerous incremental accommodations to be made: the acquisition of iron siderophore, transport and storage systems; the creation of regulatory systems that adjust metabolism in response to iron deprivation; and the evolution of enzymes that scavenge reactive oxygen species, repair damaged enzymes and sequester free iron.

Still, a more fundamental adjustment would be the replacement of Fe-S enzymes with cluster-free isozymes or alternative metabolic strategies. Is this possible? In one sense it has already happened: pyruvate dehydrogenase, for example, has displaced PFOR and pyruvate:formate lyase in aerobes. This change makes sense not only because Pdh is resistant to oxidants, but also because its formation of NADH is metabolically helpful, rather than detrimental, to a respiring cell. But one is left with the sense that evolution has not yet completed the process of making microbes oxygen-tolerant. In recent years workers have discovered that in some organisms a handful of unstable Fe-S enzymes have been replaced by relatively resistant isozymes, including fumarase C (Liochev and Fridovich, 1992), aconitase A (Gruer and Guest, 1994), a [2Fe-2S] dihydroxyacid dehydratase (Flint and Emptage, 1990), Mo-nitrogenase (Eibbe et al., 1997), NiFe hydrogenase (Frey, 2002), 2-methylcitrate dehydratase (Grimek and Escalante-Semerena, 2004) and an oxygen-resistant PFOR (Pieulle et al., 1997). The fumarase example may be instructive: whereas anaerobes employ the oxidant-sensitive [4Fe-4S] enzyme, E. coli replaces that enzyme during oxidative stress with a cluster-free isozyme, and mammals have converted fully to the latter enzyme. This serves to remind us that contemporary microbial biochemistry represents only a snapshot of an ongoing evolutionary process.


Work in the author's laboratory is supported by GM49640 from the National Institutes of Health.