Ferritin : A Cytoprotective Antioxidant Strategem of Endothelium *

Phagocyte-mediated oxidant damage to vascular endothelium is likely involved in various vasculopathies including atherosclerosis and pulmonary leak syndromes such as adult respiratory distress syndrome. We have shown that heme, a hydrophobic iron chelate, is rapidly incorporated into endothelial cells where, after as little as 1 h, it markedly aggravates cytotoxicity engendered by polymorphonuclear leukocyte oxidants or hydrogen peroxide (H20z). In contrast, however, if cultured endothelial cells are briefly pulsed with heme and then allowed to incubate for a prolonged period (16 h), the cells become highly resistant to oxidant-mediated injury and to the accumulation of endothelial lipid peroxidation products. This protection is associated with the induction within 4 h of mRNAs for both heme oxygenase and ferritin. After 16 h heme oxygenase and ferritin have increased approximately SO-fold and 10-fold, respectively. Differential induction of these proteins determined that ferritin s probably the ultimate cytoprotectant. Ferritin inhibits oxidant-mediated cytolysis in direct relation to its intracellular concentration. Apoferritin, when added to cultured endothelial cells, is taken up in a dose-responsive manner and appears as cytoplasmic granules by immunofluorescence; in a similar dose-responsive manner, added apoferritin protects endothelial cells from oxidant-mediated cytolysis. Conversely, a sitedirected mutant of ferritin (heavy chain Glu” 4 Lys; Hisas + Gly) which lacks ferroxidase activity and is deficient in iron sequestering capacity, is completely ineffectual as a cytoprotectant. We conclude that endothelium and perhaps other cell types may be protected from oxidant damage through the iron sequestrant, ferritin.

Phagocyte-mediated oxidant damage to vascular endothelium is likely involved in various vasculopathies including atherosclerosis and pulmonary leak syndromes such as adult respiratory distress syndrome. We have shown that heme, a hydrophobic iron chelate, is rapidly incorporated into endothelial cells where, after as little as 1 h, it markedly aggravates cytotoxicity engendered by polymorphonuclear leukocyte oxidants or hydrogen peroxide (H20z). In contrast, however, if cultured endothelial cells are briefly pulsed with heme and then allowed to incubate for a prolonged period (16 h), the cells become highly resistant to oxidant-mediated injury and to the accumulation of endothelial lipid peroxidation products. This protection is associated with the induction within 4 h of mRNAs for both heme oxygenase and ferritin. After 16 h heme oxygenase and ferritin have increased approximately SO-fold and 10-fold, respectively. Differential induction of these proteins determined that ferritin is probably the ultimate cytoprotectant. Ferritin inhibits oxidant-mediated cytolysis in direct relation to its intracellular concentration. Apoferritin, when added to cultured endothelial cells, is taken up in a dose-responsive manner and appears as cytoplasmic granules by immunofluorescence; in a similar dose-responsive manner, added apoferritin protects endothelial cells from oxidant-mediated cytolysis. Conversely, a sitedirected mutant of ferritin (heavy chain Glu" 4 Lys; Hisas + Gly) which lacks ferroxidase activity and is deficient in iron sequestering capacity, is completely ineffectual as a cytoprotectant. We conclude that endothelium and perhaps other cell types may be protected from oxidant damage through the iron sequestrant, ferritin.
Aerobic organisms are well endowed with enzymatic oxidant defense systems which protect against direct assault by activated oxygen species such as superoxide and hydrogen peroxide (H202). However, much of the cellular damage caused by activated oxygen involves the collaboration of intracellular iron. For example, the amount of Hz02 required to kill Stuphylococcus uureus decreases 1,000-fold if the bacteria R 0 1 HL33793, R32, HL28935, and R 0 1 AI25625. The costs of * This work was supported by National Institutes of Health Grants publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 5 To whom correspondence should he addressed Division of Hematology, Dept. of Medicine, Box 480, University of Minnesota Hospital, Harvard St. at East River Rd.,Minneapolis, are raised in iron-rich media (1). Conversely, iron chelators such as deferoxamine and dipyridyl protect eukaryotic and prokaryotic cells against challenge by oxidants such as H202 (2,3). This implies that oxidative reactions, when catalyzed by intracellular iron, are especially potent and important damaging events and, further, that endogenous iron sequestrants might be critical in antioxidant strategy.
Heme, a ubiquitous iron-containing compound, is present in large amounts in many cells and is also inherently dangerous, particularly when it escapes from intracellular sites (4). Its toxicity may reflect the fact that, unlike "free" iron, heme readily enters the hydrophobic domain of biologic membranes. The vascular endothelium, because of its continuous contact with circulating red blood cells, might be at risk from exogenous heme exposure. In previous studies we showed that hemin rapidly intercalates into cultured endothelial cells resulting in their marked hypersusceptibility to subsequent oxidant-mediated cytolysis by H202 or adherent polymorphonuclear leukocytes (PMNs)' ( 5 ) .
The foregoing considerations prompted us to hypothesize that endothelial cells might synthesize a natural iron chelator to limit the reactivity of heme-derived intracellular iron. Within most cells the major depot of nonmetabolic iron is ferritin, a high molecular mass (450 kDa), multimeric (24subunit) protein (heavy or H chain M , 21,000, light or L chain M , 20,700) with a very high capacity for storing iron (4,500 mol of iron/mol of ferritin). In the ferritin shell the proportion of H and L subunits depends on the iron status of the cell or tissue and varies between organs and species (6). Despite earlier reports that under certain chemical circumstances ferritin can release catalyticallyactive iron (7) which can actually foster peroxidation of lipids (8), recent studies suggest that such release is slight under more physiologic circumstances, less than 2 of 4,500 potential iron atoms released per ferritin molecule (9). Alternatively, ferritin might beneficially sequester intracellular iron, limiting the pro-oxidant hazard posed by this reactive metal; moreover, the fact that the H chain of ferritin manifests ferroxidase activity (10,ll) implies that ferritin-stored iron might resist cyclical reduction/oxidation reactions which tend to propagate and amplify oxidative damage.
The following studies, which have been presented in preliminary form elsewhere (12), demonstrate that endothelial exposure to heme induces sequential synthesis of heme oxygenase followed by ferritin; the latter whether endogenously produced or added directly as the apoprotein, protects endo- using the thiobarbituric acid calorimetric assay as described previously (5).

RESULTS
As shown in Fig. 1A (and reported previously in another context (5)), brief (l-4 h) but not prolonged (16 h Fig. lB), the cells were pulsed with a brief preincubation of hemin (10 pM) and then washed and incubated for a further 15 h; these cells then resist the usually very damaging combinations of newly added hemin (5 pM) plus H202 (compare third and fourth bars) or stimulated PMNs (compare seventh and eighth bars). Similar results were observed when human umbilical vein or human aortic endothelial cells were used instead of porcine aortic cells (data not shown). hemin-catalyzed injury of endothelial cells induced by hemin pretreatment. Treatment of endothelial cells was as described under "Experimental Procedures" with the indicated exception: namely that endothelial cells were exposed to hemin a t two separate times. The induction phase was accomplished by 10 p~ hemin treatment for 60 min, the culture medium replaced, and the cells incubated a further 15 h before the sensitization phase. Endothelial cell sensitization was provided by the addition of 5 PM hemin for 60 min 2 h before the cytotoxicity phase (addition of H202 or PMNs) as described under "Experimental Procedures." Results represent mean percent specific cytotoxicity f S.E. of a t least seven experiments performed in duplicate.

Iron Sequestration
Since incubation of hepatocytes and fibroblasts with hemin is known to cause induction of both heme oxygenase (28,32) and ferritin (28, 33), we wondered whether one or both of these constituents might provide oxidation resistance. As shown in Fig. 2 A , porcine aortic endothelial cells, when incubated with 10 p~ hemin, are rapidly induced to synthesize heme oxygenase mRNA. In addition, as shown in Fig. 2B, light chain ferritin mRNA increases, albeit gradually. Conversely, endothelial ferritin H chain mRNA did not increase following exposure to hemin (data not shown); this coincides with previous studies demonstrating that iron-mediated regulation of ferritin synthesis occurs primarily by post-transcriptional/translational modulation (6, 29). An associated increased synthesis of the subsequent intact heme oxygenase and ferritin proteins eventuates (Fig. 2C), consistent with the hypothesis that the protective effect of hemin preincubation might reflect induction of one or both of these substances. That ferritin, rather than heme oxygenase, is the ultimate protectant is implied by data in Table I. We developed two experimental conditions that induced endothelial ferritin but not heme oxygenase; both were cytoprotective. Thus, preincubation of endothelial cells with a cell permeant Fe-PIH chelate or with a combination of hemin and tin mesoporphyrin IX (an inhibitor of heme oxygenase) causes substantial increases in intracellular ferritin without any increment in heme oxygenase activity. In both of these cases, an associated marked protection against subsequent hemin + H202 challenge is noted. However, under converse circumstances (if cells were incubated with hemin + deferoxamine or with sodium arsenite) high levels of heme oxygenase, but not ferritin, accrue; no associated protection against later challenge by hemin and H202 is afforded. Not surprisingly, inhibition of all new protein synthesis with cycloheximide also ablates induction of this protective mechanism. In studies not shown, addition of the potent heme-binding protein, hemopexin, to our system prevented both heme uptake by endothelial cells and subsequent ferritin synthesis; no cytoprotection was induced under these circumstances.
Evidence favoring a critical antioxidant role for intracellular ferritin remains, nonetheless, circumstantial. Cells preincubated with heme and other protective compounds might be making numerous, subtle and unsuspected adjustments involving neither heme oxygenase nor ferritin synthesis. Therefore, to examine the consequences of elevated intracellular ferritin per se, we directly loaded preformed ferritin into  4.3 Hemin + Sn-meso-P (25 p M ) 6.7 & 2.9 522.6 k 17.6 0.88 f 1.1 Hemin + DF (1 mM) 343 . 3.9 Sodium arsenite (50 pM) 77.5 f 5.2 68.9 f 6.3 59.50 f 1.3 Hemin + CHX (10 Ng/ml) 10.0 k 1.36 28.4 k 2.3 58.89 k 2.3 " Endothelial cells were induced with hemin or other reagents as described under "Experimental Procedures." At the time points of the H,O,-mediated cytotoxicity assays endothelial cell heme oxygenase activities and ferritin contents were also determined.
For PAEC cytotoxicity, 100 p~ H,O, was used after the 60-min 5 p~ hemin sensitization period as described in Fig. 1B. ' The results represent mean f S.E. of at least three experiments done in duplicate.  1.5 Hemin (  " PAECs were treated with hemin or apoferritins as describedunder "Experimental Procedures." At the time points of the H202-mediated cytotoxicity assays endothelial cell ferritin contents were also determined. 5 p~ Hemin sensitization and 100 p M H,O, injury of PAECs were carried out as described in Fig. 1B. ' The results represent mean f S.E. of at least three experiments done in duplicate except at recombinant human ferritin treatments which were done two times in duplicate. cultured endothelial cells. This was accomplished via simple incubation of the cells with increasing concentrations of exogenous apoferritin. Diverse cells have been shown to pinocytose extracellular ferritin (34), and as validated by chemical and immunofluorescent assays, our cultured endothelial cells also accumulate substantial ferritin during such incubations (see below). As shown in the middle portion of Table 11, these ferritin-loaded cells become resistant, in a dose-responsive fashion, to the oxidant stress imposed by the combination of H202 + heme or to H202 alone (not shown). Accompanying this resistance was a parallel reduction in the incubated endothelial cells of lipid peroxidation products, measured as thiobarbiturate-reactive material ( 5 ) . For instance, in control heme/H202-exposed cells undergoing approximately 60% cytolysis, thiobarbiturate-reactive material accumulates ( FIG. 3. Endothelial cell ferritin inversely correlates with HzOz/hemin-mediated cytotoxicity. Ferritin was determined as described under "Experimental Procedures" in PAEC whose levels were either increased by exposure to hemin, Fe-PIH, or apoferritin at varying concentrations (Table 11) or decreased with cycloheximide or deferoxamine (Table I). H20, endothelial cytotoxicity was measured after sensitization with hemin as in Fig. 1B. Each point represents at least two experiments done in duplicate. The closed squares represent cells loaded with exogenous apoferritin.
negligible following heme/H202 exposure (0.2 k 0.1 nmol/mg of cell protein). Protection from heme/H202 assault is afforded specifically by apoferritin, and endothelial cells preincubated with other proteins (including apotransferrin, apolactoferrin, and albumin) fail to exhibit any increase in resistance to H202 or heme + H202 (results not shown).
Indeed, under various experimental conditions a close inverse linear correlation ( r --0.97) between induced intracellular ferritin content and the susceptibility of endothelial cells to oxidant challenge can be appreciated (Fig. 3). Ferritin has ferroxidase activity located on the H but not the L subunit (18). This activity catalyzes the oxidation of ferrous iron under aerobic conditions to ferric iron to allow intracellular iron storage in biological systems. In our studies of endothelial cells, the protection provided by ferritin is evidently attributable either to iron storage and/or to the intrinsic ferroxidase activity of the H chain. This is supported by experiments shown in the bottom portion of Table I1 in which human recombinant wild type or mutant H chain ferritin (18) was loaded into cultured endothelium. As expected, the wild type ferritin H chain protected as well as intact ferritin. However, the mutant H chain, which lacks both ferroxidase activity and iron storage capability, fails to provide any protection.
By immunofluorescent assay ferritin induced in human umbilical vein endothelial cells by exposure to hemin is found

Iron Sequestration
widely distributed in the cytoplasm in a finely granular fashion in almost every cell (Fig. 4B); this contrasts with the sparse and sporadic distribution noted in control endothelial cells (Fig. 4A). When exogenous apoferritin is added, nearly every cell contains ferritin but in a coarser granular pattern (Fig. 4C). Fig. 4 0 demonstrates morphologically that hemininduced ferritin synthesis is inhibited by deferoxamine, corroborating the fluorometric enzyme immunoassay results and cytotoxicity data presented in Table I. In nonpermeabilized (3.6% paraformaldehyde-fixed) endothelial cells, no immunoreactive ferritin was noted on the cell surface of hemetreated cells, whereas there was slight and variable surface staining of endothelial cells exposed to apoferritin (data not shown). This suggests that the ferritin complex is primarily localized intracellularly.

DISCUSSION
Ferritin has generally been thought to function as a "housekeeper" storage protein (6) which can release iron required for cellular proliferation (e.g. for ribonucleotide reductase) and metabolic renewal (e.g. for cytochrome synthesis). Although some studies have suggested that ferritin can amplify oxidative phenomena (7,8), evidence can also be marshalled to suggest that ferritin may play a protective role against toxic effects of iron overload in cells. For instance, rat liver nuclei from iron-intoxicated animals excessively transcribe ferritin (L subunit) mRNA as compared with control animal nuclei (29). We believe the present studies more directly support this latter notion by demonstrating a dose-responsive cytoprotective, antioxidant effect of ferritin in iron-loaded cells (Fig. 3). Further correlative support may derive from the time-dependent dichotomous effects of heme exposure on endothelial cells. Thus, heme amplifies oxidant-mediated cytotoxicity during brief (2 h) exposure, yet markedly protects oxidant-exposed target cells if provided for longer durations. These results correlate with the 1-2-h time lag that is then followed by rapid synthesis of ferritin in endothelium exposed to heme (Fig. 2C). Finally, the likelihood that ferritin per se is the critical cytoprotectant in our study is buttressed by its significant (albeit less potent) efficacy even when exogenously added to cultured cells (compare solid and open symbols, Fig.

3).
Usually cells take up iron in a highly regulated fashion dependent on transferrin binding and cellular transferrin receptor modulation; however, cells can evidently also incorporate iron by transferrin receptor-independent pathways with potential deleterious outcome. For example, studies by others which seem particularly relevant to the present work have demonstrated that the uptake of heme with resultant iron loading can occur in cultured cell lines via specific nontransferrin receptors (35). Perhaps as a countermeasure, diverse species possess avid heme-binding proteins, such as hemopexin, which efficiently inhibit cellular heme uptake and heme-catalyzed oxidation reactions (36,37). Moreover, many uni-and multicellular organisms also appear purposefully to destroy free intracellular heme by enzymatic means. For instance, heme oxygenase, a ubiquitous enzyme which opens the porphyrin ring and releases heme iron, is induced in several tissues by oxidant stress (and by free heme) (27, 38). In fact, it has been suggested that the other product of heme oxygenase activity, bilirubin, may itself be an important antioxidant. Nevertheless, the salutary effects of enhanced heme oxygenase activity are not intuitively obvious. Although catabolism of heme may rid the cell of a membrane-permeant form of iron, the resultant non-heme iron would seem to represent a potential hazard unless sequestered in some way. Indeed, the present studies demonstrate that heme oxygenase activity, unlike ferritin, does not correlate with cytoprotection. Instead, we suggest that heme oxygenase serves mainly to provide free intracellular iron from added heme; this iron, in turn (unless chelated by deferoxamine; Table I), drives the synthesis of ferritin, which is likely the proximate protectant against oxidant damage. Alternatively, heme itself might enhance ferritin synthesis directly by increasing RNA translation as recently shown by Lin and co-workers (39).
The protective effects of elevated intracellular ferritin reported herein presumably add to those of antioxidant enzymes such as catalase, glutathione-related enzymes, and superoxide dismutase. Induction of these enzymes has been linked by others to subsequent protection of cells from oxidant challenge. In that regard, in our hands hemin-exposed endothelium does not increase its catalase or glutathione peroxidase levels over 16 h (data not shown).
Our results using the recombinant mutant 222 demonstrate that ferritin's potent antioxidation efficiency probably depends on one or both of its unique molecular characteristics: namely, its high sequestering capacity for inorganic iron and the ferroxidase activity of its H chain, both lacking in the mutant. Presumably the latter prevents significant accumulation of cellular Fez+, the ultimate catalyst of oxidative damage. New H chain recombinants will soon be available which should allow further dissection of molecular characteristics critical for cytoprotection. Thus, endothelial protection studies are planned which will utilize a mutant that manifests ferroxidase activity but which cannot incorporate iron because of deletion of 22 amino acids forming the hydrophobic channel of the molecule (40). It is becoming increasingly clear that the susceptibility of eukaryotic and prokaryotic cells to oxidants is powerfully influenced (if not dictated) by levels of reactive intracellular iron. Relatively iron-poor cells (and many inherently ironpoor organisms) are highly resistant to oxidants and oxidant drugs (41). By contrast, cells containing excess iron (for instance, bacteria grown in iron-rich conditions or malaria parasites which generate ferruginous deposits from digestion of host red cell hemoglobin) are easily destroyed by oxidants (1,42). Germane to the potential role of endothelial oxidant damage in atherogenesis are recent observations: namely, that vessel wall cells from atherosclerotic lesions are remarkably iron-rich as compared with neighboring unaffected vascular tissue from the same individuals (43).
We suggest from our studies that under iron-loading con-ditions the facile synthesis of ferritin, an intracellular protein present in species ranging from bacteria to humans, may be a n important mechanism to control reactive iron and thereby suppress oxidant damage. Studies in press (44) further support this notion. We have shown that oxidant-mediated and universally fatal renal failure results from rhabdomyolysis in glycerol-injected rats but not if animals are preinfused with hemoglobin. Protection parallels the induction of high levels of renal ferritin; conversely, ablating this response with a competitive inhibitor of heme oxygenase, tin protoporphyrin, markedly accelerates kidney dysfunction in these animals (44).