Ferritin couples iron and fatty acid metabolism

Abstract

A physiological relationship between iron, oxidative injury, and fatty acid metabolism exists, but transduction mechanisms are unclear. We propose that the iron storage protein ferritin contains fatty acid binding sites whose occupancy modulates iron uptake and release. Using isothermal microcalorimetry, we found that arachidonic acid binds ferritin specifically and with 60 μM affinity. Arachidonate binding by ferritin enhanced iron mineralization, decreased iron release, and protected the fatty acid from oxidation. Cocrystals of arachidonic acid and horse spleen apoferritin diffracted to 2.18 Å and revealed specific binding to the 2-fold intersubunit pocket. This pocket shields most of the fatty acid and its double bonds from solvent but allows the arachidonate tail to project well into the ferrihydrite mineralization site on the ferritin L-subunit, a structural feature that we implicate in the effects on mineralization by demonstrating that the much shorter saturated fatty acid, caprylate, has no significant effects on mineralization. These combined effects of arachidonate binding by ferritin are expected to lower both intracellular free iron and free arachidonate, thereby providing a previously unrecognized mechanism for limiting lipid peroxidation, free radical damage, and proinflammatory cascades during times of cellular stress.—Bu, W., Liu, R., Cheung-Lau, J. C., Dmochowski, I. J., Loll, P. J., Eckenhoff, R. G. Ferritin couples iron and fatty acid metabolism.

Fatty acids serve a plethora of physiological functions, including structural, signaling, and metabolic functions, and are therefore tightly regulated and transported via a series of both intracellular and extracellular fatty acid-binding proteins. These proteins also maintain free fatty acid concentration in the low micromolar range (1) to reduce their toxicity and potential for perpetuating oxidative injury because many are unsaturated. At least 9 such intracellular proteins have been identified with differential tissue distribution (2) and fatty acid selectivity. Most intracellular fatty acid-binding proteins use a β-barrel motif to provide a hydrophobic tunnel where the long acyl chain resides and a helix-loop-helix “portal.” Electrostatic contacts, hydrogen bonds, and π-cation (with acyl double bonds) interactions are formed with arginine and tyrosine residues inside the tunnel (3). This combination of interactions typically results in ∼1 μM affinities and a binding stoichiometry of 1:1, although in some examples, 2 fatty acids can be bound (1). The principal extracellular fatty acid-binding protein is serum albumin, a largely α-helical protein providing lower affinity sites, but with a higher stoichiometry of 5 to 7 sites/albumin molecule (4). Its role is primarily transport.

In times of tissue stress, such as ischemia/reperfusion, free fatty acids released by the action of calcium-activated phospholipases go on to serve primarily signaling and substrate roles (5). Because many released fatty acids are highly unsaturated (e.g., arachidonic acid), they also can become oxidized and facilitate peroxidation chain reactions. Among the intermediates is hydrogen peroxide, which in the presence of free iron (Fe²⁺) results in the far more damaging hydroxyl radical and the hydroxyl anion via Fenton chemistry. These products perpetuate the peroxidation cycle, further damaging lipids, proteins, and nucleic acids, and ultimately activating apoptosis or other pathways to cell death (6, 7).

A functional coupling between iron and fatty acid metabolism has been previously demonstrated. For example, application of free fatty acids to cells in culture causes an up-regulation in ferritin expression, interpreted as part of a generalized stress response (8). Down-regulation of L-ferritin in melanoma cells in culture produced an enhanced response to oxidative injury and the resulting apoptosis, whereas the most oxidatively resistant cells displayed the highest ferritin expression (9, 10). Finally, a correlation between serum ferritin and obesity has been demonstrated (11, 12). Thus, it would seem that iron and fatty acid metabolism are somehow coupled, but whether it is a direct interaction or is transduced indirectly remains unclear.

Our previous structural work with ferritin identified a U-shaped, intersubunit pocket that binds the amphiphilic general anesthetics with high affinity (13, 14). This site lies between bundles of α helices and is referred to herein as the “intersubunit pocket.” Further, this ferritin “anesthetic” site has two arginine residues at the mouth, which do not appear to interact with the anesthetic but might interact with ligands with charged groups that engage the pocket more fully. Although other ligands can bind (15, 16), an endogenous ligand has not yet been identified as occupying this well-defined pocket. We hypothesized that fatty acids are endogenous ligands for this site, especially those involved with signaling and free radical interactions, such as arachidonic acid.

MATERIALS AND METHODS

Isothermal titration calorimetry (ITC)

Measurements were made at 30°C using a MicroCal VP-ITC system (MicroCal, Northampton, MA, USA) as described previously (14). In the ITC cell was 10 μM horse spleen apoferritin (HSAF) in 20 mM Tris and 130 mM NaCl (pH 8.5), and in the 0.28-ml syringe was 700 μM fatty acid in the same buffer. Injections were 15 μl. Raw data were corrected for control titrations (fatty acid into buffer and buffer into protein) and integrated using Origin 5.0 (MicroCal). Thermodynamic parameters were fit to the enthalpogram using a single-class binding site model.

Crystallography

With the use of HSAF (Sigma-Aldrich, St. Louis, MO, USA) with no other purification, apoferritin was crystallized by hanging-drop vapor diffusion at 18°C. Arachidonate was added to the degassed reservoir and protein solution at a final concentration of 200 μM. Crystals grew to a final size of 200–400 μm in <1 wk and were cryoprotected with glycerol and flash-cooled in liquid nitrogen. Diffraction data were collected at 100 K in-house on a MicroMax-007 rotating copper anode X-ray generator with an R-AXIS IV++ imaging plate detector (Rigaku, The Woodlands, TX, USA). The structures were determined by molecular replacement using 1XZ1 as the starting model, and other aspects of refinement were as reported previously (14).

Thiobarbituric acid reactive substances (TBARS) assay

The TBARS assay was performed by well-established methods (17). In brief, 200 μM arachidonate in 20 mM Tris and 130 mM NaCl (pH 8.5), with or without 10 μM HSAF, was preincubated for 15 min with 0.01 mM CuSO₄. The reaction was started with 2 mM H₂O₂ and incubated for 3 h at 37°C. Then 1% w/v thiobarbituric acid in 50 mM NaOH and 10% trichloroacetic acid were added, the mixture was boiled for 15 min, and TBARS were quantified by absorption at 532 nm.

Mineralization assay

Iron mineralization was measured as described previously with minor modifications (18, 19). HSAF (2 μM, 24-mer) in 50 mM Tris and 50 mM NaCl (pH 7.0) was incubated with or without 700 μM arachidonate or 2 mM caprylate, and the assay was started with 50 Eq of Fe(II) per HSAF 24-mer, at 25°C. We monitored the increase in absorbance at 350 nm every second over the course of ∼5 min.

To verify that the increased mineral was localized to the ferritin protein, we performed 2 assays. First, to ensure that mineral was associated with the ferritin 24-mer, we injected samples from the above assay into an HPLC size-exclusion column (G4000SW; Tosoh Bioscience, King of Prussia, PA, USA),and eluted with 50 mM Tris (pH 7.0) and 50 mM NaCl at 1 ml/min while monitoring at 280- and 350-nm absorbance simultaneously. To further verify that the protein remained intact during the iron mineralization process, we used transmission electron microscopy (TEM). In brief, samples from the above assays were diluted to 1 nM HSAF and drop-dried on a carbon-coated copper grid (300 lines/inch; Electron Microscopy Sciences, Hatfield, PA, USA). Dried grids were rinsed with water to remove buffer salts and then were negative stained with 1% uranyl acetate. Grids were imaged with a JEOL 1010 transmission electron microscope at 80 kV (Electron Microscopy Resource Laboratory, University of Pennsylvania, Philadelphia, PA, USA).

Iron release assay

Iron release from 1 μM horse spleen ferritin partially loaded with Fe(III) per 24-mer (Sigma-Aldrich) was monitored at 522 nm in the presence of 2.5 mM FMN/NADH and bipyridyl in 0.1 M N-morpholino propanesulfonic acid (MOPS) buffer (pH 7) with or without 0.7 mM arachidonate or 2 mM caprylate.

RESULTS

We used ITC to evaluate the fatty acid/ferritin interaction. After correction of the data for heats of dilution using titrations of ligand into buffer, buffer into protein, and buffer into buffer, we found strong evidence of specific binding (Fig. 1A). Fitting to a single-class, noninteracting site model revealed ∼60 μM affinity (Table 1). The energetic basis for binding was dominantly enthalpic with a small entropic penalty, most likely due to ligand restraint. To test the specificity of fatty acid binding, we also injected the much shorter saturated fatty acid, caprylic acid (C8:0), which revealed 5-fold lower apparent affinity (Fig. 1B and Table 1).

Figure 1.

ITC of arachidonate (A) and caprylate (B) injections into 10 μM HSAF. Parameters derived from fits to these data are given in Table 1.

Table 1.

Isothermal titration calorimetry parameters

Parameter	C20:4/HSAF	C8:0/HSAF
N	1.0	1.0
KA (M−1)	15,627 (381)	3049 (305)
ΔH (cal/mol)	−8675 (608)	−6231 (878)
ΔS (cal/mol · K)	−2.5 (1.91)	−3.04 (1.06)

Numbers in parentheses are se of 3 separate titrations. N is fixed at 1 arachidonate molecule/HSAF dimer.

To confirm binding at the predicted intersubunit 2-fold pocket (13) and characterize the underlying protein-ligand interactions, we cocrystallized HSAF and arachidonic acid as described previously (ref. 14 and Table 2). Our 2.18-Å structure reveals clear electron density in the intersubunit site that is well fit by the fatty acid ligand (Fig. 2). We are confident that this electron density corresponds to arachidonic acid and not to some endogenous ligand present in the protein preparation, because crystals grown under the same conditions but lacking the fatty acid fail to show this distinctive density. The ferritin binding pocket is ideally suited to accommodate an unsaturated long-chain fatty acid, in that it is a long, U-shaped tunnel lined mostly by apolar and noncharged polar residues. At one entry to this tunnel (which opens onto the large interior cavity of the ferritin 24-mer), 2 arginine residues (Arg-59, contributed by each subunit at the dimer interface) provide polar interactions with the carboxylate head of the fatty acid. Water molecules and cadmium ions (present in the crystallization solution) are also seen to bind near the entry of the tunnel, but no direct interaction with the fatty acid headgroup is seen; the nearest ordered cadmium ion is >6 Å away from the fatty acid molecule, whereas the nearest water is >5 Å away. Away from the entrance of the tunnel, π-π interactions may exist between aromatic residues deep in the pocket (e.g., Tyr-28) and the distal double bonds on the arachidonate molecule (Fig. 2B). It is difficult to evaluate the energetic contribution of these potential interactions, because the fatty acid without these double bonds (arachidic acid, C20:0) has a water solubility much less than the arachidonate/HSAF K_D, and thus reliable binding energetics are difficult to measure. The hydrophobic tail of arachidonate protrudes well out of the opposite tunnel mouth into the apoferritin interior. The complex of HSAF and caprylic acid is likely to be similar, based on our published complexes with small amphiphilic molecules (13, 14), except that the fatty acid should be entirely accommodated within the pocket, because the position of carbon 8 in the arachidonate molecule is at the bottom of the U-shaped pocket.

Table 2.

Crystallographic data collection and refinement statistics

Statistic	C20:4
Data collection
Space group	F432
Resolution range	25.32–2.18 (2.25–2.18)
Cell constants a = b = c (Å)	180.83
Unique reflections	12,950
Mean redundancy	20.6 (20.5)
Completeness	99.3 (99.3)
Rmerge	0.094 (0.458)
Mean I/σ(I)	19.6 (2.8)
Refinement
Rwork	0.19
Rfree	0.22
Atoms
Total	1508
Protein	1355
Ligand	22
Water	319
Metal	7
Other solvents	5
RMS deviation from ideality
Bonds (Å)	0.07
Angles (deg)	0.93

Values in parentheses correspond to the outermost resolution shell.

Figure 2.

Location and structure of the arachidonate binding site. A) Left: cutaway view of the entire apoferritin 24-mer, showing the interior cavity. Two of the subunits are colored yellow and magenta. Right: these two subunits are shown in an expanded view. Arachidonic acid is shown as a dark blue ball-and-stick representation; its binding site lies at the interface of the yellow and magenta subunits. B) The intersubunit 2-fold pocket (showing two L chains), showing the arachidonate molecule (blue net electron density map). The carboxylate headgroup interacts with the two Arg-59 residues, which are in turn interacting with at least two of the adjacent glutamate residues (Glu-60 and Glu-63). The poor definition of the arachidonate tail is probably due to disorder because it is projecting well out of the pocket. C) The arachidonate molecule fully occupies the pocket with the terminal 5 carbons projecting out into the apoferritin interior. The volume of the binding pocket is shown as a transparent surface. Crystallographic data are given in Table 2; structural data have been deposited to Protein Data Bank (identification code 4DE6).

It is not clear how the fatty acid gains access to the intersubunit pocket. This site has 2 openings, both of which lead to the large interior cavity of the apoferritin 24-mer, and it is probable that the fatty acid enters the pocket via these openings. How the fatty acid gains access to the large interior cavity is less clear, because the 2-fold and 4-fold channels that connect the outside of the apoferritin “sphere” with this interior cavity appear to be too small to allow passage of bulky molecules such as fatty acids. Because these and larger molecules clearly do access this pocket (15), it is necessary to invoke some breathing motion by the oligomer that allows the ligand to negotiate the known 3- and 4-fold channels and perhaps other passages that can be seen in Fig. 2A.

Sequestration of fatty acids in a protein binding site is expected to protect them from oxidation and thereby limit participation in lipid peroxidation chain reactions. Accordingly, we found that HSAF inhibited the production of oxidized products (TBARS; ref. 17) from solutions of arachidonate by 41 ± 4% (n=3) compared with an absence of protein. Beyond protection through sequestration, we hypothesized that occupancy of this pocket might alter ferritin iron uptake or release properties. We tested 2 ferritin activities: ferrihydrite formation (mineralization) and iron release. The addition of arachidonate to HSAF significantly accelerated iron mineralization compared with either HSAF alone or the same concentration of arachidonate alone (Fig. 3A). On the other hand, the shorter fatty acid, caprylate, slightly reduced mineralization rates. That the enhanced mineral formation was associated with the HSAF 24-mer was clearly demonstrated by association of the 350-nm absorbance with the 24-mer peak on size exclusion chromatography (SEC) and by the intact sphere of correct diameter and core density on TEM (Fig. 4). In addition, we monitored iron release of purified horse spleen ferritin (partially loaded with iron mineral) exposed to arachidonate and found it to be consistently decreased, whereas caprylate had no significant effect (Fig. 3B). Thus, these data provide compelling evidence that long-chain fatty acid occupancy of the intersubunit 2-fold pocket is accompanied by changes in ferritin iron uptake and release that would be predicted to decrease free iron concentration.

Figure 3.

Ferritin iron uptake and release. A) Kinetics of mineralization for 2 μM HSAF alone (black line), 0.7 mM arachidonate alone (blue line), HSAF + arachidonate (green line), and HSAF + 2 mM caprylate (red line). Absorbance was monitored at 350 nm, and traces shown are the means ± se; n = 3. B) Iron release from 1 μM horse spleen ferritin without (black) and with 200 μM arachidonate (green) or 2 mM caprylate (red). Symbols represent means ± se; n = 3.

Figure 4.

Mineral formation is associated with the intact HSAF 24-mer. Elution profiles of HPLC SEC were monitored simultaneously at 280 nm (top panel) and 350 nm (bottom panel). Compared with the HSAF/Fe data, it is clear that the addition of arachidonate [HSAF/Fe/fatty acid (FA)] enhances the 350-nm absorbance associated with the intact 24-mer peak, as represented by the 280-nm absorbance. The peak at 15 min most likely represents small arachidonate micelles, which have broad absorbance. Inset: TEM micrograph of HSAF and Fe incubated with arachidonate as above. Uniform ferritin shells of ∼12 nm outer diameter are observed with a core density consistent with the maximum 50 Eq of iron.

DISCUSSION

These data indicate that ferritin is a novel fatty acid-binding protein of moderate affinity and that the 2-fold symmetric intersubunit pocket is the dominant binding site. This protects unsaturated fatty acids, such as arachidonate, from further oxidation, and reduces their participation in downstream proinflammatory cascades. Moreover, physiological relevance is indicated by the fact that arachidonate binding alters ferritin iron uptake and release. Mineralization is dramatically increased in the presence of this fatty acid, and iron release is decreased, suggesting that intracellular free fatty acid concentration might serve a signaling role to reduce free iron via ferritin binding, presumably to decrease perpetuation of free radical generation during times of oxidative stress.

Normal resting intracellular arachidonate concentrations are probably much lower than 60 μM, but labile pools exist in many cell types that, during times of cell activation or stress, could raise arachidonate concentrations well above this value (7).

Clues for the structural basis underlying these changes in iron uptake and release are revealed by the crystal structure (see Table 2 for data collection and refinement statistics). Mineralization in mammalian ferritin is a multistep process that relies on H-subunits for ferroxidase activity (20–22) and L-subunits for mineralization in the ferrihydrite site (23, 24). The H-subunit has similar interior surface sites, but these sites are apparently not essential for mineralization (25). The predominance of L-subunits in the purified HSAF protein (26) makes H impossible to visualize with crystallography, but a very similar 2-fold pocket exists at the H:L and H:H interface in the human protein. Comparison of the arachidonate/HSAF complex with unliganded HSAF (3F32) reveals only minor differences in the side-chain positions for residues known to form the ferrihydrite site (Glu-56, Glu-57, Glu-60, and Glu-63) on the interior surface of the L-subunit. However, it is relevant to note that both the head and tail of the fatty acid are located in direct proximity to this L-subunit ferrihydrite site (Fig. 2B). In addition, ≥8 water molecules appear to be displaced from the intersubunit pocket and immediate mouth regions. This might make water more available to hydrolytically couple ferric oxo mineral precursors emerging from channels after leaving the ferroxidase sites. Regardless of the precise mechanism, the structural features we observe make it plausible that the environment of the L-subunit ferrihydrite site is altered sufficiently to explain the enhancement in mineralization. This is supported by the absence of enhanced mineralization with the very short caprylic acid, because, being 12 carbons shorter, it is unlikely to protrude into the ferrihydrite site at all. The enhanced mineralization may also be functionally linked to iron release, in that caprylic acid had no effect on either, despite occupancy of the intersubunit pocket. It should be noted that the sequences of both human ferritin subunits are highly homologous to the horse protein (87% identity for L and 93% for H) and the crystal structures contain the same intersubunit 2-fold pockets of essentially identical character (23, 26).

This observation reconciles a large body of literature connecting oxidative stress, fatty acids, iron, and ferritin. For example, cells with siRNA knockdown of ferritin are significantly more susceptible to oxidative stress, and those overexpressing ferritin are relatively protected (9, 10). Up-regulation of ferritin expression is a well-known component of the stress response, with free iron being the assumed trigger through the iron-regulatory protein but also through a calcium-activated pathway (27). However, it is also possible that these pathways are sensitive to oxidant-triggered increases in free fatty acid or that free fatty acids themselves activate specific pathways to increase ferritin expression, because an antioxidant-response element has been identified in the 5′ region of the H-ferritin promoter (28). Prostaglandin A₁, derived from arachidonate, also enhances the expression of ferritin (8). Consistent with free fatty acids having a regulatory role in ferritin expression, the ferritin concentration has been noted to be elevated in adipocytes (29), in obesity (11, 12), hyperlipidemia (30), and type II diabetes (31), and in myelinated neuronal tissue (32). Interestingly, total blood iron is not increased nearly as much as the ferritin concentration in type II diabetes (31), suggesting that the lipid, and not iron, is driving ferritin expression. Our finding that caprylate also binds ferritin makes it likely that a range of fatty acids or fatty acid-like endogenous substances bind this same ferritin site; it is certainly possible that it is not optimized for arachidonate. Drugs also bind to this ferritin site (13, 14), raising the possibility of competition and the consequent alterations in signaling pathways mediated by the interaction between fatty acid and ferritin. Targeting this site with small molecules might be an intriguing approach to limiting tissue injury in ischemia reperfusion events.

Plasma ferritin concentration has often been used as a surrogate biomarker for total iron levels. A prevailing hypothesis is that high total iron is a risk factor for stroke and cardiovascular morbidity (33, 34), because of the likelihood that total iron reflects free iron. Interestingly, the searched-for associations have been inconsistent at best. For example, cardiovascular risk in type II diabetes was found to be paradoxically lower in those with high ferritin levels (31) or simply unassociated with serum ferritin levels (34). Finally, the results of an interventional clinical trial of the “iron-heart” hypothesis of cardiovascular disease was largely negative (33). Our observation of an interplay between fatty acids and free iron through ferritin demonstrates that additional complexity must be considered in constructing these hypotheses.