Irreversible steps in the ferritin synthesis induction pathway.

The ability of cells to re-repress ferritin synthesis after removal of an inducing agent (iron or heme) was investigated. Re-repression was found to be a slow process, requiring approximately 4 (after iron removal) to 10 h (after heme removal) for completion. Desferrioxamine mesylate (Desferal) had only a slight effect on the rate of re-repression, whereas cycloheximide was strongly inhibitory, indicating that new protein synthesis is required for re-repression. Re-repression occurred at a slow but significant rate in the presence of both Desferal and cycloheximide. These results indicate that, in the absence of an iron chelator, the induction of ferritin synthesis is essentially irreversible. The kinetics of the previously reported covalent modification of IRE-binding protein (IRE-BP) were then examined, to see whether this phenomenon might account (at least in part) for the irreversibility of induction. It was found that the heme- or iron-dependent disappearance of 98-kDa IRE-BP occurred rapidly (within 1 h), and was equally rapidly reversed upon removal of heme after a 1-h exposure. By contrast, after a 4-h exposure to heme, little 98-kDa IRE-BP could be regenerated after heme removal. These results suggest that the slow, irreversible covalent modification of IRE-BP correlates closely over time with the induction of ferritin synthesis. The covalent modification of IRE-BP depends on cell growth rate, and is most readily detected in rapidly growing cells.

The presence of iron can also trigger the covalent modification, and subsequent disappearance of IRE-BP . This reaction is stimulated by porphyrin precursors (such as Gaminolevulinic acid (ALA)) and is inhibited by antagonists of porphyrin synthesis (such as succinylacetone), which suggests the involvement of heme. This reaction was presumed to be irreversible, since it coincided with the appearance of small amounts of low molecular weight products. In contrast, the formation of the Fe4.S4 center has been proposed to be rapidly and easily reversible (Constable et al., 1992;Tang et al., 1992;Klausner et al., 1993).
In order to determine whether either of these two iron-dependent reactions (iron-sulfur center formation or covalent modification) actually regulates ferritin synthesis in vivo, we have sought to determine whether induction of ferritin synthesis itself is reversible. To this end, we have studied the conditions under which repression is restored following a period of induction (i.e. "re-repression"). Both heme and iron (as ferric ammonium citrate (Fe3+) plus transferrin) were used as inducers. In both cases, it was found that cycloheximide inhibits re-repression in a dose-dependent manner, even in the presence of the iron chelator, desferrioxamine mesylate (Desferal). This result indicates that induction of ferritin synthesis is either not thermodynamically reversible, or is kinetically blocked under physiological conditions, and that de novo synthesis of some protein, probably IRE-BP itself, is required for re-repression.
These observations prompted a further investigation of the IRE-BP covalent modification pathway. In particular, it was of interest to determine which of the two steps previously described  is irreversible. Thus the initial fonpation of a high molecular weight species (HMS), as well as its subsequent disappearance, were examined for reversibility. We also wished to determine whether either of these steps correlates in time with the induction of ferritin synthesis. The results of these experiments indicate that the formation of HMS, which is very fast, is reversible, whereas the slower disappearance of HMS is not. Moreover, it is evident that the induction of ferritin synthesis correlates closely in time with the second, irreversible, step.
Finally, we have sought to determine why others have reported diffticulties in observing the IRE-BP covalent modification phenomena (Tang et al., 1992). A possible explanation arises from our observation that the degree of IRE-BP modification is dependent on cell growth rate. Thus while IRE-BP disappearance triggered by either heme or iron is readily apparent in growing cells, this phenomenon is reduced in nongrowing cells. These and other results suggest that the variation in results reported by different laboratories may be explained by differences in cell culture conditions. The implications of these observations for the biochemical mechanism of ferritin induction are discussed.

4343
Cells were allowed to grow for 2 days, by which time they had reached from two-thirds to three-fourths confluency, or for 7 days by which time all growth had ceased, prior to conducting the experiments described.
Analysis of Ferritin Synthesis Rate by Immune Precipitation-Ferritin synthesis was induced by treatment of cells with heme or FeS+ plus transferrin. In some experiments, the inducing medium was replaced with fresh media containing Desferal, cycloheximide, or no supplement, and incubation was continued for the times indicated. Newly synthesized proteins were then labeled for 15 min to 1 h with ["Slmethionine plus [35S]cysteine (cycloheximide was removed from the indicated cultures prior to labeling), cells were lysed in immune buffer or band shift buffer, and equal counts per min of labeled lysate were immunoprecipitated with anti-ferritin antibody and analyzed by SDS-PAGE and fluorography, all as previously described .
Analysis of Total IRE-BP by Immunoblotting-After lysis in immune buffer or band-shift buffer, equal quantities of protein were analyzed by SDS-PAGE, and then transferred by electroblotting for 36 h a t 100 mA onto nitrocellulose (Schleicher & Schuell) membranes. IRE-BP was then detected using anti-IRE-BP antibody and alkaline phosphataseconjugated to goat anti-rat I& antibody, as described by Blake et al. (1984).
RNA Band Shift Assays-Cells were lysed in band shift buffer, and equal quantities of protein (usually 10 pg) were then analyzed by the technique of Leibold and Munro (1988) as modified by Walden et al. (1989). Transcripts of a L-ferritin 5'-untranslated region (from the 5' cap to the AuaI site 31 nucleotides into the open reading frame) labeled with [32P]GTP were prepared as previously described (Brown et al., 1989). Rat IRE-BP and IRE-BP I1 standards were generously provided by Dr. E. Leibold (Salt Lake City, UT).
All experiments presented here were performed a t least three times.

RESULTS
Ferritin synthesis in rabbit fibroblasts (RAB-9) was induced by addition of 100 1.1~ Fe3+ plus 0.2 mg/ml transferrin to the culture media as previously described . At various times thereafter, samples of cells were labeled with [35S]methionine plus [35S]cysteine for 1 h. Labeled ferritin was then analyzed by cell lysis, immunoprecipitation with antiferritin antibody, SDS-PAGE, and fluorography. The results of this experiment (Fig. lA) show that induction of ferritin synthesis is relatively slow, peaking at 4-5 h after iron addition. Interestingly, the induction process did not always stop immediately upon removal of iron, especially at early times. Thus after 2 h of iron treatment, ferritin synthesis continued to rise for at least 1 h after iron removal, prior to declining. This behavior is consistent with the interpretation that iron does not immediately inactivate IRE-BP, but rather potentiates a much To label proteins currently being synthesized, L3WMet plus P5S]Cys was included during the final hour of incubation. Cells were then lysed, and labeled proteins were analyzed by immunoprecipitation with anti-ferritin antibody, followed by SDS-PAGE and fluorography. Panel B, cells were treated with ferric ammonium citrate plus transferrin for 5 h to induce ferritin synthesis. Subsequently, cells were washed in Earle's balanced saline solution, and then incubated in normal growth media. This media was supplemented with 200 p~ Desferal or 3 pg/ml cycloheximide for the times indicated. After the indicated periods cells were then washed three times with minimal essential medium lacking Met or Cys and placed in media containing ["SlMet plus [35SJCys for 15 min for labeling of cell proteins. (For cells that had been exposed to Desferal, this was included in the labeling media.) After labeling, cells were lysed and processed as described for Panel A. Panel C , cells were induced with ferric ammonium citrate plus transferrin as described above. m e r 5 h, cells were washed and (where indicated) incubated for 3 h with 200 p~ Desferal. Cells were then incubated overnight in media supplemented with increasing concentrations of cycloheximide (units are pg/ml). On the next day, all cells were washed three times with minimal essential medium minus Met and Cys to remove cycloheximide, labeled for 1 h with [Y31Met plus [36SlCys, lysed, and processed as described for Panel A. slower inactivation process. %-repression of ferritin synthesis was then studied by removing excess iron and returning cells to the original media. In some cases, Desferal (at 200 1.1~) or cycloheximide (at 3 pg/ml) was included in this re-repression media (cycloheximide was removed from the media prior to pulse labeling; see "Experimental Procedures"). In pression to be achieved (Fig. 1C). Similar results were obtained when ferritin synthesis was induced with heme instead of iron (Fig. 2). However, in this case re-repression is slower than after induction by iron. This observation may reflect the fact that heme is a considerably more effective inducer of ferritin synthesis than iron in growing cells, and causes a more extensive loss of IRE-BP; moreover, the effects of heme are substantially more resistant to Desferal treatment than those of iron .
Prolonged treatment of cells with cycloheximide alone sometimes produced detectable induction of ferritin synthesis (Fig.  2B, lanes 5 and 6). A similar effect was observed with actinomycin D (Daniels-McQueen et al., 1992). Both effects may be a manifestation of the slow turnover of IRE-BP that is due to low levels of iron present in normal growth media.
These results suggest that protein synthesis is ordinarily required for the re-repression of ferritin synthesis following induction. This conclusion is consistent with previous studies which suggested that IRE-BP may be degraded in the presence of iron during induction . We next sought to obtain more direct evidence that this apparent loss of IRE-BP is in fact irreversible. (If it were not, then our current results might be explained by a requirement for de nouo synthesis of some protein other than IRE-BP, which is necessary for re-repression.) It was also of interest to determine whether either of the two steps previously described, the rapid conversion of IRE-BP to HMS, and the slow disappearance of HMS, might be reversible. (The presence of only one irreversible step in a pathway could be sufficient to render the overall process irreversible.) Indeed, the results in Fig. 3 show that approximately one-half of the prelabeled IRE-BP that is apparently lost during a 1-h exposure to heme can be recovered by subsequent incubation in the absence of heme.
A similar result was observed when total IRE-BP was detected by immunoblotting (Fig. 4). Thus IRE-BP that had apparently been lost can be partially regenerated. The source of this recoverable IRE-BP may be the HMS. (This material is difficult to remove from the SDS-PAGE gel, so it frequently is not detected by immunoblotting. Since the technique employed allows electrophoretic transfer of a 200-kDa marker protein, it seems likely that HMS may possess a branched-chain structure.) In any event, these results indicate that a covalent modification of IRE-BP can be partially reversed if the exposure to heme is sufficiently brief. This reversal process is quite rapid, being virtually complete in 1 h (Fig. 4).
By contrast, when the period of heme treatment was extended to 4 h, little of the lost IRE-BP could be recovered in the 98-kDa region (Fig. 5, A and B ) . Similar results were obtained when the samples were analyzed by immunoblotting (Fig. 5C). This confirms the ultimate irreversibility of the heme-dependent loss process.
Quantitative comparison of immunoblot with immunoprecipitation data sometimes revealed a component of the IRE-BP population that is resistant to heme-dependent degradation. This "refractory component" represents up to 25% of the total IRE-BP in Fig. 5, and appears to be poorly labeled, as it is not detected by immunoprecipitation of prelabeled IRE-BP. This suggests that the refractory component represents a species that is slowly modified post-translation. The nature of this refractory component will be considered further below.
The irreversible loss of prelabeled IRE-BP was also observed when ferritin synthesis was induced with iron instead of heme (Fig. 6). This iron-stimulated loss of IRE-BP is also detectable by immunoblotting (Fig. 7A). Thus loss of IRE-BP is not simply an artifact due to exogenous heme. The effect of iron is stimulated by ALA (Fig. 6). This stimulation is counteracted by succinylacetone, as previously reported . These results are consistent with the suggestion that newly synthesized heme is necessary for this process. By contrast, in the absence of cycloheximide, de novo synthesis of IRE-BP occurred after iron removal, which ultimately restored IRE-BP to its initial level (Fig. 7B).
These data confirm that the loss of IRE-BP is at least a two-step process, the first being fast and reversible, the second being slower and irreversible. Comparison of the data in Figs.  3-7 with the data shown in Figs. 1 and 2 indicates that the rapid, reversible step occurs well before any ferritin synthesis can be detected, whereas the slower irreversible step correlates closely in time with the induction of ferritin synthesis. This suggests a causal relationship between IRE-BP loss and induction; however, other interpretations are still possible.

Irreversibility of Ferritin Induction
In control experiments, we have confirmed the loss of IRE-BP detected by immune techniques by using the band shift assay to  Fig. 1C (lanes 2, 3, 8, 7, 6, 5, 4 in that figure, rearranged in the order shown here) were also analyzed by immunoblotting with anti-IRE-BP antibody. Band intensities were quantitated by densitometry. monitor RNA binding activity. These experiments were done with and without pretreatment with 2% P-mercaptoethanol (Leibold and Munro, 1988;Haile et al., 1992a). The results shown in Fig. 8 indicate that this loss of activity is specific to IRE-BP, as little activity ascribable to IRE-BP I1 could be de-

tected. Similar results have recently been obtained by others2
In contrast to the results described here, others have reported no loss of IRE-BP activity or protein in response to iron or heme treatment (Tang et al., 1992). A comparison of these authors' results with ours suggested that the difference observed may be due to differences in cell growth conditions. (In our system, pretreatment of cells for 16 h with 50 p heme or 100 p Desferal, as described by Tang et al. (19921, would inhibit cell growth.) To test this hypothesis, we compared the heme response of growing cells to non-growing cells (Figs. 8 and   9). The differences are striking: while a loss of IRE-BP is clearly evident in growing (two-thirds confluent) cells, this loss is difficult to detect after 5 additional days in culture without refeeding. Similarly, labeling experiments followed by immunoprecipitation reveal only a slight loss of newly synthesized * E. Leibold, personal communication.  FIG. 10. Induction of ferritin synthesis by iron or heme in cells grown either 2 or 7 days in culture. Panel A, RAB-9 cells were grown for 2 or 7 days, then treated with 100 p~ FeS+ plus 0.2 mg/ml transferrin ( T f ) or 20 y heme for 4 or 8 h as indicated. To identify newly synthesized proteins, the cells were then placed in labeling media (minimal essential medium lacking Met and Cys,plus [35SlMet and [3sSlCys) with Fea* plus transferrin or heme during the last hour of incubation. Cells were then lysed, and equal countdmin of labeled proteins were analyzed by immunoprecipitation with anti-ferritin antibody, followed by SDS-PAGE and fluorography. Panel B , densitometric quantitation of rabbit ferritin heavy chain synthesis as detected by immunoprecipitation (Panel A), normalized for total protein instead of counts per min incorporated. than non-growing cells; (ii) induction of ferritin synthesis in non-growing cells is a slower process than in growing cells.
These observations not only may account for variability in published results, but they also suggest the existence of biochemically distinct iron-and heme-dependent pathways which may act in variable proportions to regulate iron metabolism as a function of cell growth.

DISCUSSION
The results presented here indicate that the induction of ferritin synthesis is ordinarily an irreversible process in growing cells: synthesis of new protein is required for full re-repression. While addition of Desferal after iron removal can cause a partial restoration of the repressed state, this process is much slower than the normal re-repression mechanism. This Desferal-dependent re-repression may be due to the removal of iron from cytoplasmic aconitase, thereby regenerating the apo form of IRE-BP, which is an active repressor, as shown by others Haile et al., 1992b).
Results presented here also demonstrate that IRE-BP is irreversibly lost during the induction process, and that the time course of this phenomenon inversely parallels that of ferritin synthesis induction. These results are consistent with the hypothesis that IRE-BP loss can be a direct cause of induction, which would explain why new IRE-BP ordinarily must be synthesized in order to restore the repressed state.
Previous experiments have suggested that the irreversible loss of IRE-BP is due to its proteolytic degradation. However, more detailed information about this process has been difficult to obtain, due to the small amounts and short half-lives of putative degradation intermediates. We have considered the possibility that some of the apparently "lost" IRE-BP protein is simply covalently modified so as to mask antigenically active epitopes, thereby rendering it invisible to our antibody preparations. However, in this connection it should be noted that a polyclonal mouse anti-rabbit IRE-BP antibody, generously donated by Dr. W. Walden, has recently been tested in our laboratory. Results obtained with this preparation were indistinguishable from those reported here, where a polyclonal rat anti-rabbit IRE-BP antibody was used.
Our results also show that the rapid formation of a previously identified high molecular weight intermediate in the loss process is reversible upon removal of heme. This reversal reaction is also rapid, being essentially complete in 1 h. Thus the formation of this intermediate, and its reversion to form the 98-kDa species of IRE-BP, can both occur without having any apparent effect on ferritin synthesis.
The results described here also show that the irreversible effects of heme on IRE-BP are strongly dependent on cell growth rate. In non-growing cells, the heme-dependent loss of IRE-BP is greatly diminished relative to that seen in growing cells; similarly, the effectiveness of heme as an inducing agent is also diminished in non-growing cells. By contrast, induction by iron is moderately enhanced in non-growing cells. These disparate effects of iron and heme support the view that they can trigger distinct biochemical modifications of IRE-BP.
In summary, our results are consistent with a model in which IRE-BP reacts with iron and/or heme in more than one way. One reaction leads to the covalent modification and apparent irreversible loss of the protein. A second reaction leads to aconitase formation. Either reaction can, in principle, result in induction of ferritin synthesis. In any case, both reactions must be irreversible under physiological conditions, as deduced by the fact that induction of ferritin synthesis itself is irreversible. The relative impact of these two reactions on the induction of ferritin synthesis may depend on the cellular growth rate, as well as on cell type (Eisenstein and.