If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
* This work was supported, in whole or in part, by National Institutes of Health Grants GM101012 and GM49640 (to J. A. I.). This is the fourth article in the Thematic Minireview series “Metals in Biology 2014.” 2 The abbreviations used are:Rperibulose-5-phosphate 3-epimeraseSODsuperoxide dismutaseFurferric uptake regulatorDAHPS3-deoxy-d-arabinoheptulosonate-7-phosphate synthaseTCAtricarboxylic acid.
Mononuclear iron enzymes can tightly bind non-activating metals. How do cells avoid mismetallation? The model bacterium Escherichia coli may control its metal pools so that thermodynamics favor the correct metallation of each enzyme. This system is disrupted, however, by superoxide and hydrogen peroxide. These species oxidize ferrous iron and thereby displace it from many iron-dependent mononuclear enzymes. Ultimately, zinc binds in its place, confers little activity, and imposes metabolic bottlenecks. Data suggest that E. coli compensates by using thiols to extract the zinc and by importing manganese to replace the catalytic iron atom. Manganese resists oxidants and provides substantial activity.
Ferrous iron is an excellent surface catalyst, adeptly binding and activating anionic metabolites for non-redox reactions. Because the primordial Earth was anoxic, environmental iron was present in its reduced form, which is relatively soluble (
). Thus this metal was recruited to be a cofactor in numerous enzymes, and the emerging metabolic pathways were configured around the chemistry that iron can perform. That anoxic habitat persisted for at least two billion years, during which cellular biochemistry and metabolic networks became highly refined and integrated.
The subsequent appearance of photosystem II triggered a slow-moving cataclysm. Over the next 2 billion years oxygenic photosynthesis progressively filled the atmosphere with oxygen. Oxidized iron precipitated from the Earth's seas, forcing iron-dependent organisms to evolve complex strategies to obtain this necessary metal. Simultaneously, they struggled to find ways of coping with the threat of reactive oxygen species. There is a growing recognition that these problems tightly intertwine. Studies have revealed that oxidants primarily target iron enzymes and that cells adjust their metal usage to compensate. This review focuses upon these processes in Escherichia coli, a model organism in which knowledge of metal and stress metabolism is highly integrated.
How Do Cells Control What Metal Enters Which Enzyme?
Transition metals are often incorporated into larger cofactors such as heme, molybdopterin, cobalamin, and iron-sulfur (Fe-S) clusters. The metals bind first to assembly proteins and then are transferred to the ultimate ligands, and these two steps provide substantial metal specificity. In contrast, most mononuclear enzymes, proteins that use polypeptide residues to bind a single metal atom, appear not to employ any protein-based metal chaperone system. Therefore the identity of the bound metal must be dictated by the intrinsic metal binding properties of the protein and the availability of metals inside the cell. These determinants of metal specificity will be explored here because oxidative stress disrupts both the metalloproteins and the metal pools.
The primary divalent transition metals in the E. coli cytoplasm are iron, zinc, and manganese. (Nickel is delivered by protein chaperones to a single enzyme, hydrogenase (
); copper is restricted to the extra-cytoplasmic periplasm, where all copper-dependent enzymes reside; and cobalt is not used at all.) Although Fe2+, Zn2+, and Mn2+ favor similar coordination geometries and ligands, in a given enzyme one metal may provide far more activity than another. This is obvious for redox enzymes, such as superoxide dismutase and ribonucleotide reductase, which provide ligand spheres that poise only a particular metal at the correct potential for catalysis. Mismetallation can happen in vivo, leading to inactive enzyme, and this observation was the first to indicate that no highly gated chaperone system controls metal access to mononuclear enzymes. It appears that some of these redox enzymes use multistep loading strategies to minimize mismetallation (
). Epimerization entails deprotonation of a weakly acid carbon atom, and the key role of the Fe2+ is to electrostatically support this step by stabilizing the anion that is formed. Other mononuclear iron enzymes catalyze a variety of reaction classes, but they have in common the use of divalent iron to neutralize oxyanion intermediates (
). However, they plummet by up to 95% when the enzymes are charged with zinc. This poor activity may indicate that Zn2+ is slow to make the necessary shifts between distorted tetrahedral and octahedral geometries as substrates bind and products depart; or, because Zn2+ binds ligands with especially high avidity, it could reflect the sluggish release of product. In any case, the inactivity of zinc-loaded enzyme poses a real problem because these binding sites generally exhibit metal affinities in the order zinc ≫ iron > manganese. For example, manganese dissociated from purified Rpe with a half-time of a few minutes, iron did so in 50 min, and zinc did not detectably dissociate in 8 h (
) and therefore is difficult for evolution to circumvent. Clearly, the cell must do something special to overcome these preferences and get the metallation process right.
The key seems to be that cells carefully control the sizes of their cytoplasmic metal pools. These pools are composed of divalent metals that weakly and reversibly associate with various metabolites, membranes, protein surfaces, and nucleic acids. Whereas the total iron content of E. coli is ∼10−3m, most of the metal is tightly bound in dedicated enzyme sites, and the loose-iron pool is only ∼10−5m (
), but it is less clear what fraction is available to metallate apoproteins.
Metal pools arise from the balance between the activities of importer and exporter (or storage) proteins, and so the homeostatic point is set by metal-specific transcription factors that control their synthesis. The Fur transcription factor reversibly binds Fe2+; in the Fur:Fe form it represses synthesis of importers and activates synthesis of ferritin, an iron-storage protein (
). Zinc levels are set by two complementary regulators: Zur binds Zn2+ with high affinity and blocks further synthesis of the Znu importer, whereas ZntR binds Zn2+ with somewhat lower affinity and induces the ZntA exporter (
This scheme implies that cytoplasmic metal levels should match the binding constants of these transcription factors. These constants have been measured but can be misconstrued. For example, ZntR exhibits a dissociation constant for zinc of 10−15m (
) measured against fully hydrated zinc. At first blush this value seems untenable because as a concentration it represents <10−6 molecules per cell. Indeed, if the concentration of zinc available to bind nascent apoprotein were only 10−15m, then even a diffusion-limited (109m−1 s−1) binding process would require a week to activate half of an apoprotein population. One resolution of this conundrum is that virtually none of the zinc pool is fully hydrated; instead, a large (10−4m) pool of zinc is loosely associated with various biomolecules that can transfer it to zinc-requiring apoproteins via ligand exchange. In vitro studies support this idea (
). Another possibility is that the zinc pool is much larger than 10−15m and that ZntR tracks zinc levels according to its rate of zinc acquisition, without achieving equilibrium. Indeed, data suggest that the concentration of hydrated zinc inside E. coli is about 10−11m (
), whereas much enzyme metallation is thermodynamically determined.
Although these binding constants do not directly represent the real concentrations of available metal, they do enable one to estimate the metal occupancy of proteins, if one assumes that thermodynamic equilibrium is attained. For example, if Fur (KD = 10−6m (
)) establishes a loose-iron pool with a nominal concentration of 10−6m hydrated iron, then at equilibrium this pool would ensure that a protein with an iron-binding site of 10−8m affinity is 99% occupied. The problem becomes more interesting if one posits that this theoretical protein also has a dissociation constant for zinc of 10−10m. This affinity is substantially greater than that for iron, in accordance with the behavior of most metal-binding sites. However, if the nominal availability of hydrated zinc is 10−11m, zinc would populate only 10% of the enzyme population. Zinc metallation would predominate only if zinc levels rose or if iron became scarce. At the same time, an authentic zinc-dependent enzyme might have dissociation constants of 10−12m for zinc and 10−6 for iron and would be 99% zinc-bound.
Whether this scheme correctly describes the situation for mononuclear iron enzymes depends upon their affinities for competing metals. To date a full set of binding constants has not been determined for any of these proteins, and so this model must be regarded as tentative. Still, these enzymes have exposed metal sites and exhibit iron and manganese dissociation on a minute time scale in vitro; similar behavior in the cell should easily allow the protein to equilibrate with the loose-metal pool. Data described below indicate that even zinc dissociates from mononuclear iron enzymes in the relevant time frame in vivo (
). Thus a thermodynamic model of metal sorting is plausible for this particular class of enzymes. Note that this is less likely to be true for other groups of proteins that fold entirely around their metals, thereby prohibiting release and equilibration.
The Mechanisms of Oxidative Stress
These determinants of protein metallation turn out to be important in the context of oxidative stress. Superoxide (O2⨪) and hydrogen peroxide (H2O2) are continuously generated in aerated cells (
The toxicity of these species was proven by the growth defects of scavenger-deficient mutants of E. coli. Strains that lack cytoplasmic superoxide dismutase (SOD−) cannot catabolize TCA-cycle substrates, such as acetate, and they require supplementation with branched-chain (Ile, Leu, Val), aromatic (Phe, Trp, Tyr), and sulfur-containing (Met, Cys) amino acids (
). E. coli uses two catalases and an NADH peroxidase to scavenge H2O2, and mutants lacking these three enzymes (Hpx−) exhibit defects that overlap with those of SOD− strains: they cannot catabolize TCA substrates, and they require branched-chain and aromatic supplements (
), and the formation of hydroxyl radicals at that site leads to DNA lesions. It is the threat of oxidative DNA damage that forces E. coli to limit the concentration of intracellular iron. In mutants that lack Fur protein, the loose-iron pools swell and mutation rates rise (
Severe stress occurs when H2O2 flows into the cell from environmental sources. In this circumstance most microbes activate stress responses. The H2O2 directly oxidizes a sensory cysteine residue on the OxyR protein itself, creating a disulfide bond that activates it as a transcription factor (
). Three of these play important roles in reducing loose-iron levels and thereby suppressing DNA damage: Dps, Fur, and YaaA. Dps is a ferritin-like protein that slows Fenton chemistry by sequestering iron (
). Collectively, these adjustments enable E. coli to survive encounters with physiological (micromolar) doses of H2O2.
Inactivation of Iron-Sulfur Clusters
However, the primary effect of oxidative stress is not to kill cells but to block growth. The TCA-cycle and branched-chain biosynthetic defects arise when O2⨪ and H2O2 inactivate dehydratases that lie in these pathways (
). These enzymes are distinguished by their [4Fe-4S]2+ clusters. Unlike electron-transfer clusters, the dehydratase clusters are exposed within the active site so they can directly bind substrate, an arrangement that leaves them accessible to dissolved oxidants. The O2⨪ anion is electrostatically attracted to the cluster; presumably, it forms a complex, and if it is momentarily protonated, it exerts its oxidizing potential (+ 0.94 V) by abstracting an electron. The [4Fe-4S]3+ species that is thereby formed is unstable and releases an iron atom into the bulk solution. A [3Fe-4S]+ species is left behind. In this form the enzyme is inactive, and the pathway that it belongs to cannot operate. The TCA cycle employs two such enzymes, aconitase and fumarase, whereas the branched-chain biosynthetic pathways include dihydroxyacid dehydratase and isopropylmalate isomerase (
). Although one might expect a hydroxyl radical to be formed, evidence indicates that the transfer of a second electron from the cluster preemptively quenches the nascent ferryl radical. Nevertheless, the catalytic iron atom is again lost, leading to the same metabolic blocks that result from O2⨪ stress. The damaged enzyme can be repaired in vitro by the addition of reductant and ferrous iron. Enzyme reactivation occurs in vivo as well, with a half-time of 5–10 min (
). No particular proteins have been shown to be involved in repair of [3Fe-4S] clusters, and it is plausible that the process is solely chemical, with electron donation from reductants such as cysteine and metallation by the passive binding of Fe2+. Other metals apparently do not bind to the [3Fe-4S] form, so this class of enzyme is protected from mismetallation.
Damage to Mononuclear Iron Enzymes
The elucidation of the TCA and branched-chain phenotypes laid the foundation for understanding the aromatic biosynthetic defect of Hpx− and SOD− strains. In principle, the problem might arise from a block in either the aromatic pathway itself or the pentose phosphate pathway, which generates erythrose-4-phosphate as an aromatic precursor. Neither pathway employs Fe-S enzymes. Metabolic analysis ultimately showed that both pathways suffer bottlenecks. As little as 0.5 µm cytoplasmic H2O2 specifically inactivates Rpe of the pentose phosphate pathway (
). Both enzymes use Fe2+ to directly bind substrate. The concurrence of these two examples suggested that non-redox mononuclear iron enzymes as a class might be highly sensitive targets of H2O2. This prediction was validated through analysis of three additional E. coli enzymes, peptide deformylase, cytosine deaminase, and threonine dehydrogenase, all of which were inactivated by low-grade H2O2 stress both in vitro and in vivo (
). It is notable that although the five enzymes all catalyze different types of reactions, they each use lone Fe2+ atoms to bind substrate and to electrostatically stabilize an oxyanionic intermediate. This action requires that the iron atom be solvent-exposed; like the catalytic iron atom in [4Fe-4S] dehydratases, this exposure allows access to small oxidants. Oxidation by either H2O2 or O2⨪ converts the iron to its ferric form, which dissociates, leaving behind an inactive apoprotein. When this oxidation occurs in vitro, the enzymes can be reactivated by the addition of Fe2+.
The literature on these enzymes was long unclear as to which metal serves as cofactor. The ambiguity arose from the quick dissociation of metals during purification, plus the fact that enzyme activity is not necessarily greatest with the native metal. Rpe, for example, exhibits maximal activity in the order Co(II) > Fe(II) > Mn(II) ≫ Zn(II) (
Purification and properties of the 3-deoxy-d-arabino-heptulosonate-7-phosphate synthase (phenylalanine sensitive) of Escherichia coli K12. I. Purification of enzyme and some of its catalytic properties.
). However, only the iron-charged enzyme is sensitive to H2O2in vitro, and the recognition that H2O2 poisons the enzyme in vivo proved that iron is its native metal. One wonders, then, how many non-redox enzymes that are annotated as using single manganese, zinc, or cobalt atoms are actually cofactored by iron in vivo.
Measurements revealed that O2⨪ inactivates mononuclear enzymes with the same rate constant (1–5 × 106m−1 s−1) with which it inactivates [4Fe-4S] dehydratases (
), the upshot is that these enzymes will be inactivated by endogenous O2⨪ every 30 min or so. Repair processes countervail this effect, so that at any moment the majority of the enzymes are functional (
). However, there is a striking discrepancy in the timing of the auxotrophies that O2⨪ confers. If branched-chain amino acids are not provided, an SOD− strain stops growing almost immediately upon aeration. In contrast, the aromatic defect requires hours to be fully manifest (Fig. 2). What is the source of this difference?
Assays revealed that in SOD− mutants the activities of dehydratases declined within minutes, whereas the mononuclear enzymes retained substantial activity until much later (
). Importantly, although simple Fe2+ addition fully reactivated mononuclear enzymes that had been oxidized by O2⨪in vitro, it failed to restore any activity to extracts from SOD− cells. The scant residual activity of the extracts was fully resistant to H2O2, indicating that it originated from a metal other than iron. The Km value of the activity matched that of zinc-loaded enzymes, and sequential treatment with penicillamine, which can extract zinc, and then iron allowed full activity to be restored. This pattern was observed for DAHPS, Rpe, and threonine dehydrogenase.
Thus the current model (Fig. 3) is that oxidation in vivo starts as it does in vitro: O2⨪ oxidizes the ferrous iron and triggers its dissociation. However, in the cellular environment a pool of Fe2+ is immediately available to bind apoproteins, probably within seconds; thus the steady-state activity, which represents the dynamic equilibrium between oxidation and remetallation, remains high. In contrast, the repair of damaged [4Fe-4S] dehydratases is far slower, so their net activities quickly decline. However, in stressed cells the cycle of mononuclear-enzyme demetallation and remetallation repeats itself continuously, and with each iteration there is a finite chance that zinc will bind to the apoprotein in place of iron. When it finally does, it sticks tightly in the active site, and because the zinc-loaded enzyme possesses little activity, a pathway bottleneck develops.
This model implies that iron and zinc routinely compete to metallate apoproteins, which is a tenet of the thermodynamic metal-loading model. In fact, when ΔzntA mutants that cannot export zinc were grown in zinc-rich medium, they became zinc-overloaded and exhibited a similar mismetallation of mononuclear iron enzymes (
). Thus the basic competition between iron and zinc for these enzymes was affirmed. The competition is simply exacerbated by oxidants that repeatedly displace bound iron.
Do Cells Rescue Mismetallated Enzymes?
The thermodynamic model also requires that metal binding be reversible so that an equilibrium can be achieved. Superoxide stress was used to test that idea. When SOD− mutants were returned to an anoxic habitat, O2⨪ stress ended, and over the succeeding 30 min, the activities of zinc-loaded mononuclear iron enzymes rose back to wild-type levels, although new protein synthesis was blocked (
). Examination confirmed that these enzymes had exchanged zinc for iron. Spontaneous zinc dissociation requires days in vitro, so how does it happen so quickly in vivo?
Penicillamine, a cysteine analogue, can extract zinc from active sites, suggesting that cysteine might be able to do the same. In vitro experiments confirmed that physiological levels of cysteine (0.2 mm) can do so on the appropriate time frame (
). Interestingly, glutathione was ineffective. Cysteine binds metals in bidentate fashion using both its sulfhydryl and primary amine moieties; in glutathione the pertinent amine is derivatized, and so its ability to coordinate metals is suppressed. The important point, however, is that these experiments demonstrate not only that iron and zinc compete for mononuclear metal sites, but that dissociation happens on a time scale that enables the process to run toward equilibrium. At least in principle, no special machinery must be invoked to favor iron binding over zinc binding.
The Protective Effect of Manganese
The role of manganese in E. coli metabolism is intriguing. The bacterium has a single dedicated importer, MntH, that is minimally expressed in standard laboratory media (
) depend exclusively upon manganese, as they are redox enzymes reliant upon its reduction potential. These genes are induced during iron starvation, whereas during iron-replete growth their iron-dependent isozymes (SodB, NrdAB, HemN) suffice (
). Thus the overall picture is that a primary role of manganese is to compensate when iron is unavailable. By extension, one anticipates that manganese might simultaneously substitute for iron in mononuclear non-redox enzymes, sparing residual iron to be used in Fe-S cluster and heme enzymes.
A similar substitution may occur during oxidative stress. For many years workers recognized that manganese supplements protect all sorts of microbes against the static effects of oxidants (
). The effect is independent of O2⨪ or H2O2 degradation. These observations raised the prospect that manganese might be an oxidant-resistant substitute for iron in mononuclear enzymes. Indeed, even modest manganese supplements to Hpx− cells protect Rpe and DAHPS activities and fully suppress the blocks in the pentose-phosphate and aromatic pathways (
The emerging model is that during H2O2 stress E. coli averts iron-focused damage by shifting from iron- to manganese-centered metabolism. The sequestration of iron by Dps and the suppression of its import by Fur serve to diminish the pools of iron that might otherwise damage DNA through Fenton chemistry. Simultaneously, manganese import provides mononuclear enzymes with an oxidant-resistant cofactor that provides nearly as much activity as does iron. The diminution of iron pools is potentially problematic for Fe-S and heme synthesis, but this problem is averted when OxyR induces high titers of the Suf cluster assembly system (
). Thus the metal-swap idea is mechanistically reasonable and physiologically appealing. Solid proof would lie in demonstrating that mononuclear enzymes recovered from H2O2-stressed cells contain manganese rather than iron. However, the experiment is technically challenging because of the high manganese dissociation rate.
One alternative model might be that manganese displaces iron from its incidental associations with metabolites, so that the released iron can metallate enzymes. However, this scheme requires the improbable notion that manganese outcompetes iron for a reservoir of adventitious ligands, whereas iron outcompetes manganese for the mononuclear protein sites.
What Lies Ahead
Substantial work remains. It must still be verified that the metallation status of non-redox iron enzymes passively follows their metal binding affinities and the sizes of the cellular metal pools. The full number of non-redox iron enzymes remains unknown even for E. coli, and so the full imprint of iron deficiency and of oxidative stress upon metabolism is uncertain.
A middle ground between iron and manganese centrism is struck by Bradyrhizobium japonicum and Bacillus subtilis, aerobes that require both metals for optimal growth. They employ transcription factors that are configured to respond differentially to iron and manganese and thereby carefully balance the two metal pools (
). One presumes that these arrangements are dictated by the threat of enzyme mismetallation. The next step is to determine the tactics that higher organisms use solve the same problem, whether it be by coordinating metal pools, by replacing these enzymes with metal-free analogues, by compartmentalizing metals in organelles, by using chaperones, or by some combination of all of these.
I thank the members of my laboratory who have driven our investigations in this topic.
Purification and properties of the 3-deoxy-d-arabino-heptulosonate-7-phosphate synthase (phenylalanine sensitive) of Escherichia coli K12. I. Purification of enzyme and some of its catalytic properties.