Fe-S Cluster Biosynthesis Controls Uptake of Aminoglycosides in a ROS-Less Death Pathway

See allHide authors and affiliations

Science  28 Jun 2013:
Vol. 340, Issue 6140, pp. 1583-1587
DOI: 10.1126/science.1238328

Unreactive Death

A controversial proposal that all bactericidal antibiotics kill by reactive oxygen species (ROS) and not by their primary cell target has recently attracted high-profile refutations. The ROS-death pathway implicated overstimulation of the electron transport in respiratory chains; a malfunction that leads to ROS releasing Fe from Fe-S clusters and causing cell death via Fenton chemistry. Ezraty et al. (p. 1583) show that electron transport chains and Fe-S clusters are key to killing by aminoglycoside antibiotics but not for the reasons envisioned in the ROS theory. Fe-S clusters are essential for killing because they mature the respiratory chains that produce the necessary proton motive force for the energized uptake of aminoglycosides. Consequently, iron chelators protect against aminoglycosides, not because they scavenge the iron from Fenton chemistry, but because they block aminoglycoside uptake.


All bactericidal antibiotics were recently proposed to kill by inducing reactive oxygen species (ROS) production, causing destabilization of iron-sulfur (Fe-S) clusters and generating Fenton chemistry. We find that the ROS response is dispensable upon treatment with bactericidal antibiotics. Furthermore, we demonstrate that Fe-S clusters are required for killing only by aminoglycosides. In contrast to cells, using the major Fe-S cluster biosynthesis machinery, ISC, cells using the alternative machinery, SUF, cannot efficiently mature respiratory complexes I and II, resulting in impendence of the proton motive force (PMF), which is required for bactericidal aminoglycoside uptake. Similarly, during iron limitation, cells become intrinsically resistant to aminoglycosides by switching from ISC to SUF and down-regulating both respiratory complexes. We conclude that Fe-S proteins promote aminoglycoside killing by enabling their uptake.

Reactive oxygen species (ROS) have been recently proposed to be central to cell killing by all classes of bactericidal antibiotics (1). However, using a recently published high-throughput chemical-genetics screen in Escherichia coli, we did not detect any functional enrichment for ROS-defense genes in the profiles of two major classes of bactericidal antibiotics: β-lactams, which target the cell wall, and aminoglycosides, which cause mistranslation (fig. S1) (2). Instead, β-lactams and aminoglycosides cause cellular death through unrelated morphological defects (35) (fig. S2). We decided to further explore the proposed role of ROS in antibiotic killing by using a series of mutants altered in the protective response of E. coli against ROS and testing them with a β-lactam [ampicillin (Amp)] and an aminoglycoside [gentamicin (Gm)] antibiotic.

E. coli mutants, hypersensitive to O2.– (lacking both cytoplasmic superoxide dismutases, ΔsodA and ΔsodB) or to H2O2 (lacking the H2O2-sensing master activator, ΔoxyR), exhibited similar sensitivities to Gm and Amp as the wild type (WT) in a time-dependent killing experiment, with ΔoxyR being more resistant to Amp at the last time point, 4.5 hours after drug addition (Fig. 1, A and B). When tested in a concentration-dependent killing experiment, the two mutants were as sensitive as WT to Gm (Fig. 1C) but exhibited small differences to the WT at intermediate Amp concentrations, at levels that provided no support for a prominent role for ROS defense mechanisms during treatment with bactericidal antibiotics (Fig. 1D). In contrast and as expected, both strains were hypersensitive to their respective ROS source, a known O2.– generator (paraquat) and H2O2 (fig. S3). Similarly, an oxyRc strain constitutively expressing the OxyR regulon—which is significantly more resistant to H2O2 (fig. S3)—showed slight differences to WT in killing experiments with Amp and Gm (Fig. 1, A to D), tending to be more susceptible to both antibiotics than WT (Fig. 1, C and D).

Fig. 1 The ROS-stress response is not required for antibiotic killing in E. coli.

(A and B) Survival of WT and mutants with compromised oxidative stress resistance after Gm (5 μg/ml) and Amp (5 μg/ml) treatment was essentially the same (black, no antibiotic; red or blue, with antibiotic). Survival, measured by colony-forming units (CFU) per ml, was normalized relative to time zero at which the antibiotic was added (midexponential phase cells; ~5 × 107 CFU/ml) and was plotted as log10 of % survival. ΔoxyR lacks the H2O2 sensing master regulator OxyR, which positively controls the expression of ROS-defense genes, whereas oxyRc expresses ROS-defense genes constitutively. The ΔsodA ΔsodB strain lacks both cytoplasmic superoxide dismutases. (C and D) As above, but survival was measured as a function of antibiotic concentration. Log10 of % survival was measured at 1.5 hours after Gm addition (C) and at 3 hours after Amp addition (D). Mutants deviated from the WT behavior only in Amp, but the effects were inconsistent with a need for ROS defense mechanisms during the antibiotic treatment. Values are expressed as means (number of experiments n = 3 to 10), and error bars depict standard deviations. Asterisks indicate a statistically significant difference between mutants and the WT, apart from when two mutants are compared, and then this is clearly indicated in the figure. *P ≤ 0.05; **P ≤ 0.01; and ***P ≤ 0.001 (Mann-Whitney U test). When all the measurements within a curve share the same significance level, asterisks are shown in parentheses after the last measurement point.

The lack of evidence for a link between oxidative stress and bactericidal antibiotics also held true when testing the same strains for minimal inhibitory concentrations (MIC) and growth rates in subinhibitory antibiotic amounts (table S1 and fig. S4). Taken together, these results revealed no association between ROS and bactericidal antibiotic sensitivity, in agreement with two recent reports using complementary approaches (6, 7).

Kohanski et al. (1) proposed that protein-bound Fe-S clusters are required for killing by bactericidal antibiotics because they release Fe2+ ions that fuel ROS production by Fenton chemistry. This assumption was based on the fact that mutants lacking the major Fe-S cluster biogenesis system ISC were resistant to both Gm and Amp. The iscS gene codes for the ISC cysteine desulfurase that, in addition to Fe-S protein maturation, is involved in all sulfur trafficking pathways (8, 9). We found that the iscS mutant, as previously reported (1), was fully resistant to Gm killing and showed partial resistance to Amp in a time-dependent killing experiment using 5 μg/ml for both drugs (Fig. 2, A and B). However, the enhanced resistance of the iscS mutant was only recapitulated for Gm, but not for Amp at lower antibiotic concentrations (fig. S5, A and B) or when measuring MICs and growth rates in subinhibitory antibiotic concentrations (table S1 and fig. S5C).

Fig. 2 Mutations in the ISC, but not the SUF, Fe-S cluster biogenesis machinery makes E. coli resistant to bactericidal aminoglycosides.

Survival of WT and mutants defective in Fe-S cluster biogenesis after (A) Gm (5 μg/ml) and (B) Amp (5 μg/ml) treatment (black, no antibiotic; red or blue, with antibiotic). The ΔiscUA mutant was resistant to Gm, but the pleiotropic ΔiscS also showed some resistance to Amp. (C) The eSUF strain is resistant to Gm. The eSUF strain contains deletions of the endogenous iscUA genes and suf operon and expresses the suf operon from an ectopic chromosomal position under the control of the pBAD promoter. The eSUF isc+ strain is the parent of the eSUF strain before transduction of the ∆iscUA deletion. Cultures of eSUF and eSUF isc+ were grown in the presence of arabinose (0.2%). (D) CCCP, a PMF uncoupler, increased resistance to killing by Gm in a dosage-dependent manner. (E) Tritiated gentamicin (3H-Gm) uptake was compromised in cells that depend on the SUF rather than the ISC machinery. Uptake was measured by incubating early exponential-phase cultures (OD600 ~ 0.3) with 5 μg/ml 3H-Gm at 37°C. (F) Exogenous proteorhodopsin (pPR) restores the Gm susceptibility of cells depending on the SUF machinery for Fe-S cluster formation. For killing experiments, strains were grown, and log10 of % survival was measured as in Fig. 1. Values, error bars, and statistical significance were calculated and are indicated as in Fig. 1.

We then tested an iscUA mutant, because in contrast to the pleiotropic iscS mutant, it is specifically compromised in Fe-S cluster biogenesis, as it lacks both the scaffold for assembling the Fe-S cluster and the transport machinery that inserts the Fe-S cluster into apo-proteins (9, 10). Interestingly, the iscUA mutant was resistant to Gm and sensitive to Amp in all tests used (Fig. 2, A and B, table S1, and fig. S5). We conclude that Fe-S clusters are required for the bactericidal effect of aminoglycosides but not for that of β-lactams.

If killing by aminoglycosides is not caused by ROS, why does eliminating the ISC system render E. coli resistant to these antibiotics? Fe-S clusters are essential for growth, and E. coli has a second assembly system, called SUF (10). To dissect the role of SUF in aminoglycoside treatment, we made a mutant lacking endogenous ISC and SUF and instead expressing SUF from an arabinose-inducible ectopic copy (eSUF). This strain only grew in the presence of arabinose, confirming that E. coli needs at least one of the two Fe-S biogenesis systems to survive (11). Moreover, eSUF was more resistant to killing by Gm, indicating that Fe-S clusters are not per se detrimental to E. coli on aminoglycoside treatment. In contrast, the parental eSUF isc+ strain was Gm sensitive (Fig. 2C and fig. S6A). Thus, E. coli without the ISC machinery is resistant to Gm, not because it cannot synthesize Fe-S clusters but because it uses SUF to build them.

Aminoglycosides’ uptake into bacterial cells requires proton motive force (PMF) that is generated by electron flow through the respiratory chain (1215). We verified the link between the PMF and aminoglycoside uptake by using carbonyl cyanide-m-chlorophenylhydrazone (CCCP), a PMF uncoupling reagent. Increasing amounts of CCCP blocked Gm-mediated killing (Fig. 2D) and allowed E. coli to grow well under subinhibitory Gm concentrations (fig. S7). Gm uptake was also reduced in cells using exclusively SUF (iscUA and eSUF), but addition of ISC to eSUF restored Gm uptake (Fig. 2E). Heterologous expression of proteorhodopsin (16), a PMF generating system, restored Gm sensitivity for the iscUA mutant (Fig. 2F) but played no role in WT cells (fig. S8). We concluded that cells lacking ISC are resistant to Gm as a result of a PMF defect.

PMF is largely generated by the respiratory complex I, NADH (reduced form of nicotinamide adenine dinucleotide) dehydrogenase (Nuo) and, to some extent, by complex II, succinate dehydrogenase (Sdh) (17). Complex I, but not complex II, directly translocates protons, but both also indirectly contribute to PMF production by passing electrons to the proton-translocating cytochrome oxidases (17). Complexes I and II contain nine and three Fe-S clusters, respectively. Although only the single nuo mutant, and not the sdhB mutant, was significantly more resistant to Gm killing and exhibited lower Gm uptake, the nuo sdhB double mutant showed some aggravating effect for both killing and uptake (Fig. 3, A and B, fig. S6B, and fig. S9). This suggests that both systems are targeted by ISC, but they are not the only ISC-matured systems that are relevant for aminoglycoside resistance, because the nuo sdhB double mutant is still more sensitive to Gm than the iscUA mutant. Consistent with ISC maturing, these complexes—and, as previously reported (8)—an iscUA mutant showed close to background activity for respiratory complexes I and II (Fig. 3C and fig. S10). Because the levels of respiratory complex I polypeptides are similar in both WT and iscUA (fig. S11A), we concluded that the lack of ISC impairs efficient Fe-S cluster insertion in main PMF-producing aerobic respiratory complexes, leading to reduced Gm uptake and killing.

Fig. 3 The PMF-generating respiratory complexes I and II function in strain using ISC, but not in strain using SUF, Fe-S cluster biogenesis machinery.

(A) Survival of WT, complex I and/or II-deficient strains after 5 μg/ml Gm treatment. The Δnuo and ΔsdhB mutants cannot synthesize respiratory complexes I and II, respectively (black, no antibiotic; red, with antibiotic). (B) 3H-Gm uptake is defective in cells without respiratory complexes I and II (Δnuo ΔsdhB) and restored when both SUF and respiratory complex I are overexpressed (eSUF/pNuo+), indicating that SUF can mature sufficient amounts of respiratory complex I to create the necessary PMF. The uptake experiment was performed as described in Fig. 2E. (C) Activities of respiratory complexes I and II are altered in cells lacking ISC. Complex I activity can be partially restored when both SUF and respiratory complex I are overexpressed. Complex I (Nuo) activity was assessed by monitoring deamino-NADH consumption in whole-cell lysates. Complex II (Sdh) activity was assayed by monitoring dichloro-phenol-indo-phenol (DCPIP) reduction in membrane preparations (fig. S10 provides an explanation for the higher than basal succinate dehydrogenase activity of the ∆nuo ∆sdhB double mutant). (D) Cells without ISC become susceptible to Gm only when both SUF and respiratory complex I are overexpressed. For killing experiments, strains were grown and log10 of % survival was measured as in Fig. 1. Values, error bars, and statistical significance were calculated and are indicated as in Fig. 1.

The findings above also imply that SUF cannot adequately substitute for ISC in the maturation of respiratory complexes I and II. Indeed the ISC-deficient eSUF strain showed reduced activity levels of respiratory complexes I and II (~45% of WT) (Fig. 3C), despite the high SUF protein level (fig. S11A). Only a concomitant increase in the respiratory complex I protein level by a factor of about 5 (fig. S11B), using an inducible plasmid (pNuo+), yielded sufficient increase in complex I activity (~70% of WT) (Fig. 3C) to restore both Gm uptake (Fig. 3B) and sensitivity (Fig. 3D). This suggests that higher than normal levels of both SUF and its substrate are required for SUF to efficiently mature respiratory complexes I and II. E. coli has a second NADH dehydrogenase system, NdhII, which lacks Fe-S clusters and does not translocate protons. We found that NADH reduction mediated by NdhII is slightly increased in the iscUA mutant (fig. S12), but this increase in respiration was not sufficient to support Gm uptake (Fig. 2E).

Under iron limitation, cells use SUF instead of ISC (11, 18, 19). We suspected that this might be the reason Kohanski et al. reported that chelation of Fe eliminates aminoglycoside killing (1) and misinterpreted it as a role for Fenton chemistry in antibiotic killing. As predicted, adding the intracellular iron chelator 2,2′ dipyridyl (DIP) to growing E. coli resulted in low activity levels for respiratory complexes I and II, inhibited Gm uptake, with consequent resistance to Gm killing (Fig. 4, A to C, and fig. S6, C and D). Adding back exogenous iron to DIP-treated cells suppressed the protection to Gm (Fig. 4D) and confirmed that the DIP effect was due to iron limitation. Interestingly, DIP-treated cells were characterized by a decrease in respiratory complex I protein levels, in addition to the higher SUF–lower ISC protein levels (Fig. 4E), which may also have contributed to the low Gm uptake. This effect on protein level was also rapidly reversed after addition of exogenous iron to DIP-treated cells (Fig. 4E). Thus, iron limitation leads to both a switch from ISC to SUF and a reduction of respiratory complex I levels, thereby making cells aminoglycoside resistant.

Fig. 4 Iron limitation orchestrates a switch from ISC to SUF and prevents aminoglycoside uptake.

Complexes I and II activity assays (A), 3H-Gm uptake (B), and survival after Gm treatment (5 μg/ml) (C) of WT in the presence/absence of an iron chelator (250 μM DIP), and of three mutants affected in iron homeostasis control, Δfur, ΔryhB, and Δfur ΔryhB. When the iron-limitation response was turned on by the addition of the iron chelator DIP or deletion of the global repressor fur, complexes I and II activity and 3H-Gm uptake were low, which caused Gm resistance. Knocking out one of the central players of the iron response, ncRNA RyhB, restored antibiotic sensitivity, indicating that protection against aminoglycosides is mediated by RyhB. (D) Addition of exogenous iron restored sensitivity to Gm to iron-starved cells (250 μM DIP). FeSO4 (100 μM) was added 1.5 hours after treatment with 5 μg/ml Gm. (E) Iron limitation induced SUF protein levels and reduced that of ISC and complex I. The response was rapidly reversed on addition of exogenous FeSO4 (100 μM). Western blots were performed with antibodies against specific subunits of the respiratory complex I (NuoF), the SUF (SufC), and the ISC (IscS) machineries. Adding exogenously excess of iron to DIP-treated cells resulted in a 9-fold decrease in SufC protein levels and a 2-fold increase in IscS and NuoF levels after 90 min. Values, error bars, and statistical significance were calculated and are indicated as in Fig. 1.

Iron bioavailability is sensed by Fur, a global repressor of >100 genes involved in iron homeostasis (20, 21). Surprisingly, a fur-null mutant, which expresses the iron-limitation response constitutively and hence contains high intracellular iron levels, produced similar, albeit more modest, phenotypes as the DIP-treated cells in terms of respiratory complex I and II activities (Fig. 4A), complex I protein levels (fig. S10C), and Gm uptake (Fig. 4B) and sensitivity (Fig. 4C). Although loaded with intracellular iron, the fur mutant was still significantly more resistant to Gm than WT as a result of reduced Gm uptake, supporting the argument that it is not intracellular iron levels per se that control aminoglycoside resistance but rather the iron limitation response. We then wondered whether Fur was directly, or indirectly via one of its downstream targets, responsible for the aminoglycoside resistance. The small noncoding RNA (ncRNA), RyhB, seemed an ideal target, as it is Fur-dependent and itself inhibits the synthesis of nonessential Fe-utilization proteins (22), including Isc, Nuo, and Sdh (23). Consistent with this scenario, the ryhB fur double mutant restored the activities of respiratory complexes I and II (Fig. 4A), the complex I protein levels (fig. S11C), the uptake of Gm (Fig. 4B), and the Gm sensitivity to WT levels (Fig. 4C and fig. S6C). Thus, RyhB expression is the relevant iron-limitation feature that leads to aminoglycoside resistance.

In summary, we have shown that the Fe-S cluster biogenesis machineries play a key role in aminoglycoside resistance by affecting their PMF-energized uptake. Switching from ISC to SUF allows E. coli to maintain enough Fe-S clusters to survive, albeit with low PMF levels. As a consequence, aminoglycosides, whose uptake is strongly PMF-dependent, cannot reach the ribosome, their cytoplasmic target. As uptake and target lie in a positive-feedback loop for aminoglycosides (fig. S13) (24), small changes in the otherwise basal uptake of aminoglycosides has a severe impact on their action. At least for E. coli, iron limitation would be a physiological environment where the cell switches from ISC to SUF, becoming more resistant to aminoglycosides. Bacteria are iron-depleted in the host, but some pathogens induce local increases of iron concentration by lysing host cells and erythrocytes (25), which suggests that aminoglycosides may be to some degree selective against enterobacterial pathogens. In addition, aminoglycosides may be very effective in treating recurrent enterobacterial infections in patients with human genetic disorders associated with iron overload in the blood, such as thalassemia and hemochromatosis (26).

Supplementary Materials

Materials and Methods

Figs. S1 to S13

Tables S1 and S2

References (2740)

References and Notes

  1. Acknowledgments: We are grateful to C. Gross (University of California, San Francisco, USA) for critically reading the manuscript and providing feedback. We thank T. Friedrich (Freiburg University, Germany), T. Yagi and J. Torres Bacete (La Jolla, CA, USA), E. Bouveret (Laboratoire d’Ingénierie des Systèmes Macromoléculaires, Marseille, France), J. Armitage (Oxford, UK), and P. Moreau and T. Mignot (Laboratoire de Chimie Bactérienne, Marseille, France) for providing materials and strains. This work was supported by grants from Agence Nationale Recherche (ANR Blanc SPV 05511), the Institut Universitaire de France, the Fondation pour la Recherche Médicale, and European Molecular Biology Laboratory. Data are deposited in the Dryad Repository:
View Abstract

Stay Connected to Science

Editor's Blog

Navigate This Article