Mitochondrial iron detoxification is a primary function of frataxin that limits oxidative damage and preserves cell longevity

Abstract

Friedreich ataxia is a severe autosomal-recessive disease characterized by neurodegeneration, cardiomyopathy and diabetes, resulting from reduced synthesis of the mitochondrial protein frataxin. Although frataxin is ubiquitously expressed, frataxin deficiency leads to a selective loss of dorsal root ganglia neurons, cardiomyocytes and pancreatic beta cells. How frataxin normally promotes survival of these particular cells is the subject of intense debate. The predominant view is that frataxin sustains mitochondrial energy production and other cellular functions by providing iron for heme synthesis and iron–sulfur cluster (ISC) assembly and repair. We have proposed that frataxin not only promotes the biogenesis of iron-containing enzymes, but also detoxifies surplus iron thereby affording a critical anti-oxidant mechanism. These two functions have been difficult to tease apart, however, and the physiologic role of iron detoxification by frataxin has not yet been demonstrated in vivo. Here, we describe mutations that specifically impair the ferroxidation or mineralization activity of yeast frataxin, which are necessary for iron detoxification but do not affect the iron chaperone function of the protein. These mutations increase the sensitivity of yeast cells to oxidative stress, shortening chronological life span and precluding survival in the absence of the anti-oxidant enzyme superoxide dismutase. Thus, the role of frataxin is not limited to promoting ISC assembly or heme synthesis. Iron detoxification is another function of frataxin relevant to anti-oxidant defense and cell longevity that could play a critical role in the metabolically demanding environment of non-dividing neuronal, cardiac and pancreatic beta cells.

INTRODUCTION

Micromolar concentrations of Fe(II) are required for heme synthesis and iron–sulfur cluster (ISC) assembly in the mitochondrial matrix (1,2). Storing this iron in a readily available form is critical for the biogenesis of a large number of heme- and ISC-containing proteins. At the same time, limiting iron-catalyzed generation of reactive oxygen species (ROS) is vital for the cellular anti-oxidant protection. Biochemical evidence suggests that the mitochondrial protein frataxin has the capacity of influencing both these fundamental mechanisms. Frataxin is able to serve as a Fe(II) donor for ferrochelatase in the final step of heme synthesis (3,4) or the scaffold protein Isu1 in the initial step of ISC assembly (5). Frataxin can also repair inactive [3Fe–4S]²⁺ aconitase to the active [4Fe–4S]⁺ enzyme (6). Moreover, frataxin assembles into homopolymers that convert Fe(II) to a stable protein-bound mineral (7–9), attenuating iron-catalyzed production of ROS (10–12). We have proposed that the ability to bind labile Fe(II) and to either make it available to other proteins or store it as a redox-inactive mineral may enable frataxin to promote iron metabolism and concurrently limit iron toxicity (13). Studies in living organisms have shown that frataxin does in fact promote (although it is not essential for) ISC biogenesis (14–16) and repair (6) as well as heme synthesis (17–19), suggesting that frataxin may serve as a general Fe(II) chaperone. Additional studies have shown that partial or complete loss of frataxin leads to oxidative damage in humans (20,21), mice with conditional frataxin knock-out in pancreatic β cells or hepatocytes (22,23), as well as Caenorhabditis elegans and yeast models (24,25). However, the inability to detect oxidative damage in mice with conditional frataxin knock-out in neurons or striatal muscle cells has recently been presented as evidence that frataxin deficiency is not a primary cause of oxidative stress (26). In addition, the hypothesis that frataxin is important for iron detoxification was challenged by a study in which a mutant form of yeast frataxin, unable to store iron, was shown to be phenotypically silent in vivo (27). These reports have suggested that the mitochondrial iron imbalance that accompanies frataxin deficiency is a secondary event that results mainly from the loss of ISC synthesis and does not represent a relevant source of cell injury.

Frataxin deficiency underlies Friedreich ataxia (FRDA), a relentless neurodegenerative disease associated with cardiomyopathy and diabetes (28). In addition, iron imbalance and oxidative stress are increasingly implicated in aging and age-related neurodegenerative disorders (29). Thus, whether frataxin does or not play a primary role in iron detoxification is an important question, relevant to the pathophysiology of FRDA and other more prevalent conditions. We, therefore, sought to identify mutations that specifically impair the ferroxidation or mineralization activity of yeast frataxin, which were predicted to be essential for iron detoxification (10,12). We show that although these activities are not required by frataxin to serve as a Fe(II) chaperone in vitro or in vivo, they are critical for anti-oxidant protection and cell survival in conditions of increased oxidative stress.

RESULTS

Iron storage by Yfh1p involves functionally distinct sites

In the presence of Fe(II) and atmospheric oxygen, yeast frataxin (Yfh1p) binds Fe(II) and oligomerizes into a homotrimer; if the Yfh1p-bound Fe(II) is not transferred to other ligands, the trimer catalyzes the oxidation of Fe(II) to Fe(III) and ultimately promotes the conversion of Fe(III) to a protein-bound ferrihydrite mineral core (3,7,9,10,30). Here, we screened for amino acid residues that are specifically involved in the ferroxidation or mineralization activity of Yfh1p. Frataxin proteins across species contain an extended acidic surface that has been previously implicated in iron binding via an as-yet-undefined mechanism (27,31,32). We replaced all carboxylate residues in the acidic patch of Yfh1p with alanine residues (Supplementary Material, Table S1) and first analyzed the iron uptake capacity of the Yfh1p variants. Each protein was incubated with Fe(II) at a saturating Fe(II)/protomer ratio of 75/1 and iron uptake was measured as a function of time as described previously (3). In the presence of wild-type Yfh1p, most iron was recovered in protein-bound form after 10 min of incubation and this fraction increased further after 60 min as expected (Table 1) (3). Multiple replacements of carboxylate residues affected overall iron uptake less severely than certain single replacements (Table 1, variant [E71A;E75A;E76A] or [D79A;D82A] versus [E93A]), indicating that the overall negative charge of the acidic surface is not a critical determinant in the iron uptake process of Yfh1p. In addition, replacements of adjacent residues had different effects on iron uptake (Table 1, variant [E89A] or [E90A] versus [E93A]), and a triple substitution, [E93A;D97A;E103A], caused a more severe defect than expected from the single [E93A], [D97A] or [E103A] mutations, further suggesting that iron uptake by Yfh1p is a cooperative process involving multiple residues with complementary functions. In particular, the ability to nucleate and grow a ferrihydrite mineral core was predicted to enable frataxin to compete with non-specific hydrolysis of Fe(III) in solution and to accumulate iron in a water-soluble form (9). The results in Table 1 indicate that several residues may be involved in this function. Compared with wild-type Yfh1p, the [E93A] and [E93A;D97A;E103A] variants exhibited a drastic reduction in the amount of protein-bound iron at 10 min (49 and 23% of wild-type) and 60 min (75 and 35% of wild-type), and correspondingly higher levels of labile+insoluble iron (Table 1). The [D86A], [E89A], [E90A], [D101A] and [E103A] variants exhibited lower levels of protein-bound iron and higher levels of labile+insoluble iron at 10 min, but accumulated essentially normal levels of protein-bound iron after 60 min (Table 1). These results suggest that E93 is a critical nucleation site required to incorporate iron to capacity, whereas D86, E89, E90, D101 and E103 represent accessory sites that cooperate with E93 in iron core formation. In addition, the ferroxidase activity of frataxin was predicted to promote iron uptake by providing a rapid local accumulation of Fe(III), thereby facilitating nucleation of the iron core (3,9), similar to ferritin (33). As will be shown below, residues D79 and D82 are required for ferroxidase activity, however, mutations in these residues did not significantly impair the ability of Yfh1p to accumulate iron to capacity (Table 1, [D79A;D82A] variant). We suggest that the high Fe(II)/protomer ratio used in this analysis promoted spontaneous Fe(II) autoxidation as observed previously (10) and that the presence of functional nucleation sites enabled the [D79A;D82A] protein to compete with non-specific Fe(III) hydrolysis. Thus, ferroxidase-deficient Yfh1p could form a water-soluble mineral core, although with slower kinetics (see belowFig. 2C), and incorporate iron to capacity.

Residues D79 and D82 are responsible for the ferroxidase activity of Yfh1p

To determine the ferroxidase activity of the Yfh1p variants, O₂ consumption was measured as a function of time upon addition of Fe(II) to buffer in the absence or presence of purified protein and the stoichiometric ratio of Fe(II) atoms oxidized per O₂ molecule consumed was determined at the end of each reaction (Supplementary Material, Table S1) as described previously in detail (10). A Fe(II)/O₂ ratio of ∼2 was obtained with wild-type Yfh1p, consistent with the ability to catalyze pair-wise oxidation of Fe(II), and a Fe(II)/O₂ ratio of ∼4 was otherwise measured in buffer without protein as expected for spontaneous iron autoxidation (Supplementary Material, Table S1) (10,34,35). Most variants showed a reaction rate as fast as that of wild-type Yfh1p and a final Fe(II)/O₂ stoichiometric ratio of ∼2 (Supplementary Material, Table S1 and data not shown). However, the double [D79A;D82A] variant exhibited a slower reaction rate (Fig. 1) and a final Fe(II)/O₂ ratio of 3.1±0.5 (n=5). Similar results were obtained with the single [D79A] or [D82A] variant (Supplementary Material, Table S1 and data not shown). Thus, the D79A and D82A mutations abolished most of the ferroxidase activity of Yfh1p, suggesting that these residues are part of the Yfh1p ferroxidase center.

Residue E93 is required for the mineralization activity of Yfh1p

At increasing Fe(II)/protomer ratios, the Yfh1p monomer assembles stepwise following the progression α→α₃→α ₆→α₁₂→α₂₄→α₄₈ (7,30). Formation of a stable α₄₈ multimer depends on the ability to form a stable iron core and can thus be used as a measure of frataxin mineralization activity (3,9). Assembly reactions were incubated at a Fe(II)/protomer ratio of 40/1 at 30°C for 2 or 60 min and immediately analyzed by gel filtration (30). At 2 min, most wild-type monomer was converted to a major peak corresponding to α₄₈ with a small shoulder corresponding to assembly intermediates (Fig. 2A) as described (30). The A₂₈₀ reading of the α₄₈ peak nearly doubled after 60 min, reflecting progressive accumulation of ferrihydrite within the assembled protein (Fig. 2A) as previously reported (3,9). In contrast, large proportions of the [E93A;D97A;E103A] protein were eluted from the column in monomeric form after 2 or 60 min (∼64 and 47% residual mutant monomer versus ∼32 and 20% residual wild-type monomer). A sharp high-molecular-weight peak was detected at both time points; however, unlike wild-type α₄₈, this species did not significantly increase in intensity over the time course of the experiment (Fig. 2B) and was prone to precipitate upon elution from the gel filtration column. Self-assembly of the [E93A] monomer was nearly as severely affected (58 and 41% residual monomer at 2 and 60 min), whereas the [E103A] and [D97A] monomers were able to form a stable α₄₈ multimer similar to wild-type Yfh1p (data not shown). Together with the iron uptake data shown in Table 1, these results suggest that E93 is critical for nucleation of the mineral core. In the absence of this residue, Yfh1p is unable to compete effectively with non-specific Fe(III) hydrolysis, which results in the formation of insoluble ferric iron oxides and unstable iron–protein complexes.

Ferroxidation deficit slows down Yfh1p self-assembly

The [D79A;D82A] variant assembled with slower kinetics compared with wild-type, as deduced from the presence of broader α₄₈ peaks and higher levels of residual monomer at 2 and 60 min (∼76 and 40% residual mutant monomer versus ∼32 and 20% residual wild-type monomer) (Fig. 2C). The slower assembly kinetics were consistent with the lack of ferroxidase activity slowing down iron oxidation and mineralization. Interestingly, in spite of the relatively high levels of residual monomer, the iron-loading capacity of the [D79A;D82A] variant was not affected, as already observed in Table 1. The α₄₈ multimer formed by this variant was stable upon elution from the column and actually accumulated ∼1.4 times more iron per subunit than the wild-type α₄₈. Thus, the loss of ferroxidase activity makes Yfh1p similar to L-subunit-rich ferritins, which have weak ferroxidase activity and mineralize iron more slowly but ultimately form larger iron cores than ferroxidation-proficient H-subunit-rich ferritins (33).

Ferroxidation or mineralization deficits do not alter the Yfh1p iron chaperone function

We have shown previously that frataxin can make Fe(II) bioavailable to other ligands in the presence of atmospheric O₂ at neutral pH, conditions that normally promote conversion of Fe(II) to insoluble ferric iron oxides (3,12). We used these previously developed assays to test the iron chaperone function of the [D79A;D82A] and [E93A;D97A;E103A] variants. Each protein was preloaded with 10 atoms of Fe(II)/subunit and incubated in aerobic buffer. Bipyridine (BIPY), a chelator that preferentially binds Fe(II) (36), or purified yeast ferrochelatase, which catalyzes the insertion of Fe(II) into protoporphyrin IX to yield heme (37), was added at different incubation times to measure the levels of bioavailable iron. Relative to buffer, all three Yfh1p proteins made significantly higher levels of Fe(II) available to either acceptor (Fig. 3). In the presence of the [D79A;D82A] variant, the availability of Fe(II) was slightly lower compared with wild-type (Fig. 3A and B). A small increase in Fe(II) availability was otherwise observed with the [E93A;D97A;E103A] variant (Fig. 3A and B). Thus, mutations that compromise the ferroxidation or mineralization activity of Yfh1p have no significant detrimental effect on the ability of Yfh1p to make Fe(II) available to external ligands in vitro. Different residues, which remain to be identified and probably reside outside the acidic patch of frataxin, are expected to be responsible for the iron chaperone function of the protein.

Ferroxidation or mineralization activity is not required for normal yeast growth

The identification of ferroxidation- and mineralization-deficient forms of Yfh1p provided a means to test the biological role of iron detoxification by frataxin independent of iron chaperone function. The precursor forms of the [D79A;D82A] and [E93A;D97A;E103A] variants (i.e. carrying the normal mitochondrial matrix-targeting peptide of yeast frataxin) (38) were expressed in YFH1-deleted yeast cells lacking wild-type Yfh1p (yfh1Δ). Relative to wild-type Yfh1p, each mutant protein exhibited a similar electrophoretic mobility whether it was produced in yeast (i.e. via mitochondrial import and processing of the cytoplasmically translated precursor form) or bacterial cells (i.e. via expression of the mature-size form), indicating that all proteins were processed to the mature form in yeast (Fig. 4A). The two mutant proteins migrated faster than wild-type Yfh1p on SDS/PAGE (Fig. 4A), but exhibited the expected molecular mass as determined by mass spectrometry (data not shown). Compared with yeast cells expressing wild-type Yfh1p (WT), the mutant strains (D79A/D82A and E93A/D97A/E103A) did not show any obvious growth defect on rich or synthetic defined medium supplemented with fermentable (dextrose or galactose) or non-fermentable (ethanol) carbon sources at 30°C (Fig. 4B and data not shown). In addition, during growth in rich medium, the D79A/D82A and E93A/D97A/E103A mutants exhibited levels of total mitochondrial iron and aconitase activity similar to those determined in WT cells (Supplementary Material, Table S2).

We further tested the ferroxidation and mineralization mutants during different growth phases on synthetic defined medium with glucose as the sole carbon source (SD). In these conditions, yeast cells initially grow logarithmically by fermentation of glucose and when glucose becomes limiting, transiently arrest growth and shift to respiration (diauxic shift) (39). During the subsequent post-diauxic phase, yeast cells grow more slowly using the ethanol produced by fermentation (39) and when the ethanol is exhausted enter a particular stationary phase in which a high metabolic rate is maintained (40,41). Survival in these post-diauxic and stationary phases depends on the cellular resistance to different forms of stress, including oxidative stress (41). The D79A/D82A and E93A/D97A/E103A strains grew at the same rate as WT in the logarithmic and post-diauxic phases, with all three strains achieving the same OD₆₀₀ of ∼1 at the beginning of the stationary phase (Fig. 5A, 48 h). These results were reproduced with independent isolates for each strain, demonstrating that the [D79A;D82A] and [E93A;D97A;E103A] proteins do not affect yeast growth even in conditions of high metabolic demand with respiration as the primary source of energy. This suggests that the loss of ferroxidation or mineralization activity does not affect the ability of Yfh1p to support mitochondrial iron usage, consistent with the nearly normal iron chaperone function exhibited by the [D79A;D82A] and [E93A;D97A;E103A] proteins in vitro (Fig. 3). Our data are in agreement with those of a previous report in which a triple substitution in the acidic patch of Yfh1p, [D86 N;E90Q;E93Q], which abolished iron-dependent oligomerization similar to the [E93A;D97A;E103A] variant, was found to have no apparent effects on yeast growth (27).

Ferroxidation and mineralization activities are required for iron detoxification

Others and we have shown that the ability to convert redox-active iron to an Fe(III) mineral enables frataxin to attenuate iron-catalyzed production of ROS in vitro (10–12). Therefore, we investigated if the [D79A;D82A] and [E93A;D97A;E103A] variants affected yeast resistance to iron-catalyzed oxidative stress. Wild-type and mutant strains were grown in liquid SD as described earlier, except that 100 µm iron was added to the medium. This concentration of iron induces transcriptional and post-transcriptional repression of the yeast high-affinity iron transport system (42), indicating that it is sufficient to increase intracellular iron levels. Relative to wild-type, the two mutant strains accumulated significantly higher levels of non-respiring colonies (petites) at all the time points analyzed (Fig. 5B). The data suggest that clearance of excess labile iron from the cytoplasm led to elevation of mitochondrial labile iron, which in the absence of ferroxidation- or mineralization-competent frataxin further led to an increase in oxidative damage to mitochondrial DNA and respiratory proteins, and possibly also damage to nuclear genes (24) involved in respiratory function. This scenario is consistent with the deficits exhibited by the [D79A;D82A] variant, which lost the ability to rapidly convert redox-active Fe(II) to the less reactive Fe(III) form (Figs 1 and 2C, Supplementary Material, Table S1), and the [E93A;D97A;E103E] variant, which was largely unable to incorporate Fe(III) into a stable mineral (Table 1, Fig. 2B), a process that normally limits reductive cycling of Fe(III) back to Fe(II) (43).

Ferroxidation and mineralization activities are required for anti-oxidant protection

We investigated the possibility that the presence of iron detoxification-competent frataxin might be important for anti-oxidant protection even at low iron concentrations. The WT, D79A/D82A and E93A/D97A/E103A strains were grown in liquid SD medium without iron supplementation and analyzed for the presence of non-respiring colonies as described earlier. The WT strain exhibited a basal level of petite formation of approximately 20% that increased slightly over time (Fig. 5C). The two mutant strains were not significantly different from WT after 12 h of growth (late logarithmic phase) but progressively accumulated more petites after entry in post-diauxic phase (Fig. 5C, >30 h). In agreement with this finding, after 30 h of growth in SD, the D79A/D82A and E93A/D97A/E103A mutants exhibited ∼2-fold higher levels of oxidized proteins in both the mitochondria and the post-mitochondrial supernatant as compared with WT (Fig. 5D). The increased damage observed in the D79A/D82A and E93A/D97A/E103A strains in the post-diauxic phase was not due to a deficiency of Yfh1p protein because after 30 h of growth in SD, both strains contained levels of Yfh1p either comparable to or higher than those detected in WT cells (Fig. 5E). It is important to note that both the wild-type and the two mutant Yfh1p proteins were expressed in the yfh1Δ strain from a centromeric low-copy plasmid under the control of the natural YFH1 gene promoter region (see Materials and Methods for details). In the post-diauxic phase, the mutant strains also contained wild-type levels of chelatable iron (Supplementary Material, Table S2). Thus, the oxidative damage exhibited by these strains in Figure 5C and D may have resulted from redox-active iron that was bound but inefficiently detoxified by the ferroxidation- or mineralization-deficient Yfh1p proteins. ROS cause a variety of DNA lesions including abasic sites (44). The alkylating agent, methyl methanesulfonate (MMS), is known to cause formation of abasic sites in both the mitochondrial and the nuclear DNA (45,46). The D79A/D82A and E93A/D97A/E103A strains exhibited higher sensitivity to MMS treatment compared with WT during the early logarithmic phase (Fig. 5F). Together, the results in Figure 5 suggest that the loss of frataxin ferroxidation or mineralization activity leads to chronically increased levels of oxidative damage, which accumulates slowly but progressively until it reaches a threshold for phenotypic expression in the post-diauxic phase. The threshold is reached earlier in the presence of an additional source of damage as seen upon treatment with iron or MMS.

Ferroxidation and mineralization activities are required for normal chronological life span

We further investigated if the [D79A;D82A] and [E93A;D97A;E103A] variants might influence yeast cell survival in the stationary phase. The WT, D79A/D82A and E93A/D97A/E103A strains were grown in SD medium as described earlier. The initial growth rate and the maximum cell density achieved at the beginning of the stationary phase (∼2 days) were the same for all strains (Fig. 6A). The OD₆₀₀ of these stationary cultures remained unchanged (Fig. 6A) and all three strains contained comparably low levels of Yfh1p for at least 12 days (Fig. 6B). At different times, starting from day 4, equal numbers of cells from the stationary cultures were plated on rich medium with glucose as the carbon source (YPD) and scored for viability after 5 days at 30°C. Viability is defined as the ability of a stationary-phase cell to resume growth and form a colony when plated on rich medium (47). Although all strains exhibited similar viability at day 4, the viability of the two mutants decreased drastically over time relative to WT (Fig. 6C). This effect was accelerated by treatment with MMS at the beginning of the stationary phase (Fig. 6D).

Loss of Yfh1p ferroxidation or mineralization activity is lethal in the absence of Cu,Zn superoxide dismutase

To further assess if the [D79A;D82A] and [E93A;D97A;E103A] proteins influenced yeast resistance to oxidative stress, the SOD1 gene, coding for the Cu,Zn superoxide dismutase (Sod1p), was deleted in the WT, D79A/D82A and E93A/D97A/E103A strains. The Saccharomyces cerevisiae Sod1p is localized to the cytoplasm and the mitochondrial intermembrane space (48) and has been shown to be essential for the detoxification of ROS in yeast mitochondria (49). Upon loss of Sod1p, the D79A/D82A and E93A/D97A/E103A strains did not show any appreciable growth defect compared with WT cells if grown in a microaerophilic chamber (Fig. 7A, low O₂). However, the viability of the mutant strains was dramatically reduced during growth in air (Fig. 7A, 21% O₂). These results demonstrate that in cells lacking ferroxidation- or mineralization-competent frataxin, there is a large potential for oxidative damage, which can be reduced through the action of Sod1p. The data further indicate that the ferroxidation and mineralization activities of frataxin have a protective role in conditions of increased superoxide production. We further analyzed the phenotype of our mutants in the absence of Isu1p, the main ISC scaffold protein that interacts with frataxin (50,51). Double deletion of the ISU1 and YFH1 genes or point mutations in both genes have been previously shown to be lethal to yeast cells (27,51), probably due to a severe defect in ISC synthesis that cannot be overcome by the other ISC scaffold protein, Isu2p, and other mitochondrial iron chaperones. Deletion of ISU1 (isu1Δ) did not cause any obvious effect on the growth or viability of the D79A/D82A and E93A/D97A/E103A strains as compared with WT (Fig. 7B). This indicates that the [D79A;D82A] and [E93A;D97A;E103A] mutations do not result in a synthetic negative effect with isu1Δ, further supporting the conclusion that the iron chaperone function of frataxin is normal in these mutants.

DISCUSSION

The function of frataxin has been the subject of intense analysis since a deficiency in this protein was found to be responsible for FRDA (28). Studies in patients as well as yeast and mouse models of the disease have shown that frataxin is required for mitochondrial iron balance, maintenance of ISC- and heme-containing proteins and protection from oxidative damage (reviewed in 52). We initially reported that frataxin binds iron and self-assembles (7). Monomeric and assembled frataxins have since been shown to be able to donate Fe(II) to other proteins (3–6,12,17,18). In addition, assembled frataxin has been shown to also possess ferroxidation and mineralization activities that enable the protein to detoxify surplus iron limiting iron-induced oxidative damage (3,8–12,30,53). Both in vitro and in vivo studies support a role for frataxin as a general Fe(II) chaperone. In contrast, a direct role in anti-oxidant protection has been more difficult to demonstrate in vivo because the oxidative damage associated with frataxin defects could either result from loss of iron detoxification or simply represent a secondary effect of impaired ISC synthesis affecting both mitochondrial iron balance and respiratory chain function (16,54,55). In addition, Aloria et al. (27) reported that a mutant form of frataxin, [D86N;E90Q;E93Q], unable to store iron in vitro, was phenotypically silent when expressed in yeast, which led the authors to conclude that the iron storage/detoxification function of frataxin is dispensable in vivo. To resolve this issue, we sought to identify point mutations that specifically affect the ability of frataxin to detoxify iron without impairing its iron donor function. We undertook a systematic alanine-scanning mutagenesis of potential iron ligands in the acidic patch of Yfh1p. None of the mutations analyzed caused a complete disruption of iron uptake by the protein (Table 1). Interestingly, certain triple or double mutations affected the overall iron uptake capacity less severely than certain single mutations (e.g. [E71A;E75A;E76A] or [D79A;D82A] compared with [E93A]). In addition, different single mutations affected iron uptake to varying degrees (e.g. [E89A] or [E90A] compared with [E93A]), and a triple substitution, [E93A;D97A;E103A], had a more severe effect than each individual mutation ([E93A], [D97A] or [E103A]) (Table 1). These results suggest that the topology of the various carboxylate residues in the acidic patch of frataxin is more important for iron uptake than the net negative charge of this region. Moreover, it appears that iron uptake by frataxin is a cooperative process involving functionally distinct sites, as it occurs in ferritin (35). Further elucidation of the mechanism of iron uptake must await the results of ongoing analyses of the three-dimensional structure of assembled frataxin.

Mutational analysis identified the [D79A;D82A] and [E93A;D97A;E103A] variants, which did not significantly impair frataxin ability to serve as Fe(II) donor for BIPY or ferrochelatase but nearly completely abolished the protein ability to catalyze iron oxidation or mineralization (Figs 1 and 2, Table 1, Supplementary Material, Table S1). These two activities are unique to the assembled form of frataxin and were predicted to be critical for iron detoxification (10,12), as they are in ferritin. Ferroxidase activity serves to rapidly convert redox-active Fe(II) to Fe(III), whereas mineralization enables the protein to store Fe(III) in a stable form that limits iron cycling between its two oxidation states (33,43). Thus, the [D79A;D82A] and [E93A;D97A;E103A] proteins offered an opportunity to test the physiological relevance of the iron-detoxifying properties of frataxin independent of its iron donor function. Yeast cells expressing these proteins did not show any obvious defect during logarithmic growth on rich media. However, impaired function became evident during growth on SD medium after entry in post-diauxic phase and the subsequent stationary phase. These are conditions in which yeast cells grow slowly and finally stop dividing, but continue to maintain a high metabolic rate, with respiration as the primary source of energy, leading to increased ROS production (39,41). In such conditions, yeast expressing the [D79A;D82A] or [E93A;D97A;E103A] protein accumulated more damage than wild-type cells, even when the culture medium was not supplemented with iron (Fig. 5C and D). Moreover, stationary D79A/D82A and E93A/D97A/E103A cells exhibited a progressive decrease in their chronological life span (Fig. 6C), a finding consistent with recent reports that frataxin preserves longevity in higher organisms (23,25).

Probably, mitochondrial DNA and mitochondrial proteins were the most immediate targets of the ROS produced when redox-active iron was inefficiently detoxified by frataxin. However, the ferroxidation- and mineralization-deficient mutants exhibited increased levels of carbonylated proteins in both the mitochondrial and post-mitochondrial fractions (Fig. 5D). Moreover, nuclear DNA damage was observed previously in frataxin-deficient yeast cells (24). Thus, we suggest that our mutants were subject to chronically elevated levels of oxidative damage in mitochondria and other cellular compartments. Although the mitochondrial damage most likely resulted from iron-catalyzed Fenton chemistry and immediate reaction of hydroxyl radicals with mitochondrial biomolecules, the radicals that may have led to cytoplasmic or nuclear damage remain to be elucidated. Importantly, it was only in slowly dividing or stationary cells that the damage could progressively accumulate and reach the threshold for phenotypic expression. This threshold was reached earlier in the presence of an additional source of damage, as observed upon addition of iron or MMS to the medium (Figs 5B, F and 6D) or upon deletion of the SOD1 gene (Fig. 7A). These effects were not secondary to defects in heme or ISC synthesis because the [D79A;D82A] or [E93A;D97A;E103A] protein interacted normally with ferrochelatase in vitro (Fig. 3) and did not cause any growth defect nor any synthetic negative effect upon deletion of the ISU1 gene (Fig. 7B). In addition, during growth on rich medium, mitochondrial iron content and aconitase activity were normal in the mutant strains (Supplementary Material, Table S2). These data support strongly the idea that frataxin is bifunctional and plays a primary role in iron detoxification independent of its iron chaperone function. Both aspects of the protein should be taken into consideration in addressing the pathophysiology of FRDA and in designing preventive and therapeutic strategies.

MATERIALS AND METHODS

Expression and purification of Yfh1p variants

Site-directed mutants of Yfh1p were created by polymerase chain reaction (PCR), cloned into vector pET24a(+) (Novagen) downstream of the T7 promoter and sequenced completely. The mature form of wild-type Yfh1p and its variants were expressed in Escherichia coli strain BL21(DE3) (Novagen) and purified as described (30) except that a HR16/50 Superdex 75 column (Pharmacia) was used for the final step of purification. Most variants were expressed in soluble form and at the same levels as wild-type Yfh1p. The [D78A] variant was expressed at very low levels and could not be characterized. The accurate molecular mass of variants exhibiting an electrophoretic mobility on SDS/PAGE different from that of wild-type Yfh1p was determined by isoelectrospray ionization mass spectrometry at the Mayo Proteomic Research Center. For each variant, protein concentration was determined from the absorbance and extinction coefficient (ε_280 nm=20 000 m⁻¹ cm⁻¹) and confirmed by amino acid analysis and SDS/PAGE. All purified protein concentrations were calculated per subunit. Iron concentration was directly measured by inductively coupled plasma mass spectrometry at the Mayo Metals Laboratory or deduced from the concentration of Fe[BIPY]₃²⁺ (ε_520 nm=9000 m⁻¹ cm⁻¹) (36).

Measurements of iron uptake, ferroxidase activity, self-assembly and heme synthesis

These procedures were performed as described previously in detail (3,12).

S. cerevisiae expression constructs and strains

The wild-type and mutant Yfh1p protein precursors were expressed in yfh1Δ yeast as follows. Plasmid YCplac22-YFH1 is a centromeric low-copy yeast expression vector that carries the wild-type YFH1 gene, including 625 bp of 5′-UTR (i.e. the natural YFH1 promoter region), the entire coding sequence and 22 bp of 3′-UTR, cloned in the KpnI–BamHI sites of a TRP1-based YCplac22 vector (38). Site-directed mutants of YFH1 were created by PCR, cloned into plasmid YCplac22-YFH1 and sequenced completely. The resulting plasmids, YCplac22-YFH1[D79A;D82A] and YCplac22-YFH1[E93A;D97A;E103A], were therefore identical to plasmid YCplac22-YFH1 except for the presence of the relevant mutations. Our YFH1-deleted strain, yfh1Δ[YFH1][ρ⁺] (MATα ura3-52 lys2-801^amber ade2-101^ochre trp1-Δ63 his3-Δ200 leu2-Δ1 yfh1Δ::HIS3+YCp50-YFH1-URA3 [ρ⁺]), is a haploid, respiratory competent derivative of strain YPH501 (56). It contains a disrupted chromosomal yfh1Δ::HIS3 allele and is fully complemented by a wild-type copy of the YFH1 gene on a URA3-based YCp50 plasmid. The strains used in this study, WT, D79A/D82A and E93A/D97A/E103A, were obtained in the same manner by transformation of strain yfh1Δ[YFH1][ρ⁺] with vectors YCplac22-YFH1, YCplac22-YFH1[D79A;D82A] or YCplac22-YFH1[E93A;D97A;E103A], followed by counter-selection with 5′-fluoroorotic acid to eliminate the YCp50-YFH1 plasmid (38). Upon counter-selection, several independent transformants for each strain were characterized by plasmid retrieval and DNA sequencing, and frozen stocks stored at −70°C. To delete the SOD1 or ISU1 gene, sod1Δ::kanMX4 and isu1Δ::kanMX4 targeting constructs were obtained by PCR amplification of genomic DNA isolated from the Open Biosystems sod1Δ and isu1Δ knock-out strains and transformed into the WT, D79A/D82A and E93A/D97A/E103A strains described earlier. Transformations were plated on YPD containing 200 µg/ml G418 (Mediatech, Inc.), colonies were allowed to form in a microaerophilic GasPak150 chamber (BBL) for 5 days at 30°C. Correct integration of each targeting construct was verified by PCR analysis of genomic DNA isolated from independent transformants. Positive clones (four per strain) were grown in low O₂ (57) and frozen stocks stored at −70°C. In the resulting strains (WT sod1Δ, D79A/D82A sod1Δ, E93A/D97A/E103A sod1Δ, WT isu1Δ, D79A/D82A isu1Δ and E93A/D97A/E103A isu1Δ), the coding sequence of the SOD1 or ISU1 gene is precisely replaced by the kanamycin resistance (kanMX4) cassette.

Measurements of oxidative damage and life span

The following liquid and solid media were used: YPD (2% peptone, 1% yeast extract and 2% dextrose); YPE (as YPD except with 2% ethanol instead of dextrose); SD (6.7% bacto-yeast nitrogen base without amino acids, 2% dextrose; supplemented with amino acids and other growth requirements as needed). All experiments were started with freshly streaked frozen stocks. Two or four independent isolates were analyzed for each strain in all experiments. For all the experiments in liquid SD medium, yeast cells were initially allowed to reach late logarithmic phase (OD₆₀₀∼1) by aerobic growth in SD medium at 30°C for ∼20 h and then diluted in fresh SD medium to an OD₆₀₀=0.1. For iron challenge, cultures were allowed to reach OD₆₀₀=0.3 (∼6 h), after which 100 µm FeCl₃ was added. To assess sensitivity to DNA damage, cells were directly diluted in SD medium containing 2 mm MMS (Sigma). All aerobic incubations were carried out with a culture volume equal to or smaller than one-fifth of the volume of the culturing flask, shaking at 225 rpm. In all cases, cell growth was continued aerobically at 30°C and at designated times aliquots of yeast cultures were plated or used for protein isolation. Western blot or OxiBlot analyses of protein extracts from whole cells or isolated mitochondria were performed as described (24,58).

SUPPLEMENTARY MATERIAL

Supplementary Material is available at HMG Online.

ACKNOWLEDGEMENTS

This work was supported by grants AG15709 (to G.I.) and GM49640 (to J.A.I.) from the National Institutes of Health and RSG-96-05106-TBE (to G.C.F.) from the American Cancer Society. Funding to pay the Open Access publication charges for this article was provided by a grant from the Friedreich Ataxia Research Alliance (FARA) (to G.I.).

Conflict of Interest statement. The authors have no conflicts of interest to declare.

Figure 1. Residues D79 and D82 are required for the ferroxidase activity of Yfh1p. Oxygen consumption curves were recorded with a MI-730 micro-O₂ electrode (Microelectrodes, Inc.) upon addition of 48 µm Fe(II) to buffer containing 96 µm protein or buffer without protein (10). Conditions were 10 mm HEPES–KOH, pH 7.0, at 25°C. The drift of the oxygen electrode was negligible (∼2 µm/h). At the end of the measurements, the final Fe(II)/O₂ stoichiometric ratio was 1.7 for wild-type, 3.7 for [D79A;D82A] and 4.3 for buffer.

Figure 2. Residue E93 is required for the mineralization activity of Yfh1p. Assembly reactions containing 40 µm (A) wild-type, (B) [E93A;D97A;E103A] or (C) [D79A;D82A] protein and 1.6 mm Fe(II) were aerobically incubated in 10 mm HEPES–KOH, pH 7.3, at 30°C for 2 or 60 min. Each sample was placed on ice to stop assembly, immediately centrifuged for 5 min at 20 000g and chromatographed through a gel filtration column (HR 10/30 Superdex 200) at 4°C (3,30). An equivalent amount of untreated purified monomer was also analyzed. The α peak represents iron-free monomer, whereas the A₂₈₀ of the α₄₈ peak is largely accounted for by the ferrihydrite mineral associated with the assembled protein (9). The Fe/protomer stoichiometry of α₄₈ was estimated at 60 min as follows. Fractions containing monomer were pooled, and the protein concentration determined from the absorbance and extinction coefficient. Fractions containing α₄₈ were also pooled, and the iron concentration determined by inductively coupled plasma mass spectrometry; because of the presence of iron oxides in these fractions, the protein concentration could not be determined directly and was deduced from the concentration of residual monomer. The wild-type α₄₈ and [D79A;D82A]-α₄₈ contained similar amounts of iron but the latter contained 25% less protein thus exhibiting a Fe/protomer stoichiometry ∼1.4 times higher than wild-type. Slower iron oxidation and mineralization explain why the A₂₈₀ reading of the [D79A;D82A]-α₄₈ peak is lower than that of wild-type α₄₈. The iron–protein complex formed by [E93A;D97A;E103A] was unstable and was not further analyzed.

Figure 3. Fe(II) bound to iron detoxification mutants of Yfh1p is bioavailable. Wild-type or mutant Yfh1p protein (3 µm) was aerobically loaded with 30 µm Fe(II) in 10 mm HEPES–KOH, pH 7.3, at 30°C. After each indicated period of incubation in aerobic buffer in the absence of any iron acceptor, iron-loaded Yfh1p was incubated with (A) BIPY (2 mm) for 5 min or (B) yeast ferrochelatase (FC) and protoporphyrin IX (PP) (2 and 120 µm, respectively) for 20 min. In each case, the concentration of the final product, Fe[BIPY]₃²⁺ or heme, was determined spectrophotometrically (3). The bars represent the mean±standard deviation of three independent reactions. We have shown previously that in the experimental conditions used in these assays, Yfh1p enhances Fe(II) availability relative to buffer or other protein controls (3).

Figure 4. Iron detoxification mutants of Yfh1p exhibit normal growth. (A) Protein extracts from yeast strains expressing the precursor forms of wild-type Yfh1p and its variants (250 µg/lane) and from E. coli strains overexpressing the mature forms of these proteins (residues 52–174) (40 ng/lane) were analyzed by 12% SDS/PAGE and western blotting with a polyclonal anti-Yfh1p antibody; wild-type mature Yfh1p runs as a 20 kDa protein (38). (B) Freshly streaked frozen stocks were grown aerobically on YPD and then SD plates. Ten-fold serial dilutions were plated on rich (YP) or synthetic (S) media with dextrose (D) or ethanol (E) as the carbon source, and grown for 5 days at 30°C. The red pigment is due to the ade2-101^ochre mutation in the strain genome and is a marker of robust respiration. Two independent isolates were analyzed for each strain.

Figure 5. Iron detoxification mutants of Yfh1p are hypersensitive to oxidative stress. (A) Yeast cultures were started from freshly streaked frozen stocks, synchronized to late logarithmic phase in SD medium, diluted in fresh SD medium to OD₆₀₀=0.1 and further cultured aerobically at 30°C. Cell growth was followed by OD₆₀₀ measurements. All cultures reached post-diauxic growth in ∼20 h. For each strain, data shown are the average of two independent cultures. (B) After dilution to OD₆₀₀=0.1, cells were allowed to reach OD₆₀₀=0.3, at which point 100 µm FeCl₃ was added to the medium (6 h, arrow), and incubation continued. At successive time points, equal numbers of cells (∼3×10² cells) were plated and the % of respiratory-deficient (petite) colonies was determined (24). Data shown are the means±standard deviation of two independent isolates per strain, two plates per isolate. (C) Cells were analyzed as described in (B) except that iron was not added to the culture medium. Data shown are the means±standard deviation of two independent isolates per strain, two plates per isolate. The same synchronized cultures were used to start the experiments in (B) and (C). (D) After 30 h of aerobic growth in SD, yeast cells were fractionated and aliquots of the mitochondrial fraction (Mito; 100 µg) and post-mitochondrial supernatant (PMS; 50 µg) were analyzed by OxiBlot to detect protein carbonyls (24). Different lanes of the same OxiBlot are shown. Coomassie blue staining of the membrane following OxiBlot shows that equal amounts of total protein were loaded in each lane. The brackets highlight the same region of the membrane as analyzed by OxiBlot or Coomassie blue staining. (E) Aliquots (100 µg) of the mitochondrial fraction described above were analyzed by western blotting with anti-Yfh1p antibody. Equal protein loading was verified by Coomassie blue staining of the membrane following western blotting. Different lanes of the same blot are shown. (F) Yeast cultures were synchronized in SD medium, diluted in fresh SD medium to OD₆₀₀=0.1 in the absence or presence of 2 mm MMS and cultured aerobically for another 3 or 6 h. At each time point, equal numbers of cells (∼3×10² cells) were plated on YPD and scored for viability after 5 days at 30°C. Data shown are the means±standard deviation of two independent isolates per strain, two plates per isolate.

Figure 6. Iron detoxification mutants of Yfh1p have shortened chronological life span. (A) Yeast cultures were started from freshly streaked frozen stocks, synchronized in SD medium, diluted in fresh SD medium to OD₆₀₀=0.1 and further cultured aerobically at 30°C. All cultures reached stationary phase in ∼2 days. (B) Protein extracts were prepared from yeast cells after 12 days of growth in SD and aliquots (25 µg) analyzed by western blotting as in Figure 4A. (C) In different days, equal numbers of cells from cultures similar to those shown in (A) were plated on YPD and scored for viability after 5 days at 30°C. For each strain, data show the mean±standard deviation of four independent isolates, one plate per isolate. (D) At the beginning of the stationary phase, 2 mm MMS was added to yeast cultures and viability was assessed after 3 or 6 h as in Figure 5F.

Figure 7. Synthetic lethality between iron detoxification-deficient Yfh1p and sod1Δ. The (A) SOD1 or (B) ISU1 genes were deleted in wild-type and mutant Yfh1p strains by homologous recombination. Freshly streaked frozen stocks were grown on YPD plates in a microaerophilic chamber. Ten-fold serial dilutions were spotted on different plates. Duplicates were grown at 30°C in the chamber or in atmospheric O₂ and photographed after 5 days. Two independent sod1Δ or isu1Δ isolates were analyzed for each strain.

Table 1.

Iron uptake by the acidic patch variants of yeast frataxin

	10 min		10 min	60 min		60 min
	Protein-bound (µm iron)	Labile+insoluble (µm iron)	Protein-bound (µm iron)	Labile+insoluble(µm iron)
Buffer	2±0.3a	123±6	2±2	124±6
Wild-type	95±6	38±7	120±5	11±3
[E71A;E75A;E76A]	91±2	46±2	114±8	16±3
[D78A]	n.d.	n.d.	n.d.	n.d.
[D79A;D82A]	90±9	51±7	124±1	18±6
[D86A]	83±13	54±10	124±5	8±4
[E89A]	75±4	53±4	116±2	10±3
[E90A]	58±4	71±5	117±3	11±2
[E93A]	47±3	81±6	90±5	37±5
[D97A]	91±5	40±6	116±2	10±2
[D101A]	78±4	54±4	120±4	8±1
[E103A]	78±5	49±6	116±1	9±1
[E93A;D97A;E103A]	22±2	109±2	42±4	87±5

Upon 10 or 60 min of aerobic incubation of protein (2 µm) with Fe(II) (150 µm) in 10 mm HEPES–KOH, pH 7.3, at 30°C, each sample was centrifuged in an ultrafiltration device (5 kDa nominal molecular weight cut-off) and iron concentration was determined in the retentate (protein-bound) and filterable (labile) fractions, whereas insoluble iron was stripped from the filter, as described previously in detail (12). n.d., not determined because of low protein yield. The iron concentration in the labile+insoluble fraction was underestimated because of the difficulty inherent in solubilizing iron oxides precipitated on the filter. This resulted in total iron recoveries ranging from ∼80 to 90% depending on the extent of iron precipitation. However, the levels of iron measured in the three fractions for any given variant were reproducible. The levels of protein-bound iron determined in the presence of wild-type mYfh1p at 60 min are in accord with the iron-loading capacity reported previously (3).

^aMean±standard deviation of 8 (buffer, wild-type) or 3–5 independent assays (mYfh1p variants).