Differences in the regulation of messenger RNA for housekeeping and specialized-cell ferritin. A comparison of three distinct ferritin complementary DNAs, the corresponding subunits, and identification of the first processed in amphibia.

The ferritin family is a widespread group of proteins that maintain iron in a soluble form and also protect against the toxic effects of excess iron. The structure and sequence of the proteins are highly conserved. However, the cell-specific features of structure which occur within the same organism indicate cell specificity of gene expression and may be related to variations in types of iron storage, i.e. specialized-cell ferritin (stored iron is for other cell types) versus housekeeping ferritin (stored iron is for intracellular purposes related to normal or stress metabolism); the protein structure may also affect rates of iron turnover. Iron induces ferritin synthesis and accumulation by recruiting stored ferritin mRNA that is efficiently translated in cells specialized for iron storage. For the first time we show the occurrence of three different cDNAs from bullfrog tadpoles, corresponding to three subunits of the protein: H, M, and L. Thus, ferritin can be encoded by at least three different mRNAs and probably three different genes, in contrast to the older idea of two, H and L; the subunits maintain the conserved sequences of known ferritins and have similar predicted masses, 20.5, 20.6, and 19.9 kDa, but have distinct mobilities in denaturing gels. Ferritin subunit expression is cell specific; more of the H and L chain mRNAs are expressed in red cells than in liver. Ferritin expression is regulated by transcription (or mRNA stability) in adult red cells; cellular levels of ferritin mRNA were 20% that of embryonic red cells, and L subunit mRNA increased 2.5 times with excess iron. Ferritin expression is also regulated during translation in adult red cells; iron recruits stored ferritin mRNA, but only during certain stages of red cell maturation, in contrast to embryonic red cells. The developmental differences in ferritin expression are discussed in relation to the shift from specialized-cell ferritin to housekeeping ferritin in red cells of the embryonic versus adult lines.

The ferritin family is a widespread group of proteins that maintain iron in a soluble form and also protect against the toxic effects of excess iron. The structure and sequence of the proteins are highly conserved. However, the cell-specific features of structure which occur within the same organism indicate cell specificity of gene expression and may be related to variations in types of iron storage, Le. specialized-cell ferritin (stored iron is for other cell types) uersus housekeeping ferritin (stored iron is for intracellular purposes related to normal or stress metabolism); the protein structure may also affect rates of iron turnover. Iron induces ferritin synthesis and accumulation by recruiting stored ferritin mRNA that is efficiently translated in cells specialized for iron storage. For the first time we show the occurrence of three different cDNAs from bullfrog tadpoles, corresponding to three subunits of the protein: H, M, and L. Thus, ferritin can be encoded by at least three different mRNAs and probably three different genes, in contrast to the older idea of two, H and L; the subunits maintain the conserved sequences of known ferritins and have similar predicted masses, 20.5, 20.6, and 19.9 kDa, but have distinct mobilities in denaturing gels. Ferritin subunit expression is cell specific; more of the H and L chain mRNAs are expressed in red cells than in liver. Ferritin expression is regulated by transcription (or mRNA stability) in adult red cells; cellular levels of ferritin mRNA were 20% that of embryonic red cells, and L subunit mRNA increased 2.5 times with excess iron. Ferritin expression is also regulated during translation in adult red cells; iron recruits stored ferritin mRNA, but only during certain stages of red cell maturation, in contrast to embryonic red cells. The developmental differences in ferritin expression are discussed in relation to the shift from specialized-cell ferritin to housekeeping ferritin in red cells of the embryonic uersus adult lines.
Iron, which is required for proteins involved in DNA synthesis, electron transfer, and oxygen activation, is so insoluble (10"' M) (1) under physiological conditions that special proteins, the ferritins, are needed to maintain iron in a soluble available form (Ref. 2 is a review). Control of the cellular concentration of ferritin is often effected by excess iron which induces the translation of stored ferritin mRNA ( e g . . Ferritins are composed of three components, an outer protein coat, an inner iron core (4500 iron atoms as hydrous ferric oxide with varying amounts of phosphate), and an ironprotein interface; the interface appears to be the site of nucleation for the iron core. Conservation of the overall structure is observed among all organisms; in the case of vertebrates, blocks of conserved amino acids are distributed throughout the entire peptide chain (e.g. Refs. 1,5,[7][8][9]. Apoferritin, the protein coat, is composed of 24 similar or identical subunits of about 20 kDa. Ferritin may be classified as one of two types based on whether the iron is stored for other kinds of cells (specializedcell ferritin) or for intracellular purposes (housekeeping ferritin); housekeeping ferritin plays a role in normal metabolism or during stress, e.g. heat shock or inflammation (2). Specialized-cell ferritin is found, e.g. in hepatocytes which store iron for the entire organism for the long term, in erythrophagocytosing macrophages that recycle the iron daily, and in the erythrocytes of the embryonic cell line which provide iron during the red cell switch of early development (2, 10, 11); specialized-cell ferritin is the most abundant type of ferritin and, as a consequence, is the best characterized. Housekeeping ferritin, found in small amounts in most cell types, maintains the iron reserves required for intracellular metabolism. In general, housekeeping ferritin has been little studied except in the cells of the adult erythroid line where the need for iron is amplified to support hemoglobin synthesis. At the end of erythroid maturation in the adult cell line, the ferritin concentration drops to between %5 and '/a00 that of the embryonic 7901 erythrocyte (10-13); in the embryonic red cell, ferritin actually accumulates at the end of maturation (14,15). Another role of housekeeping ferritin is to protect the cell from the toxic effects of excess iron. For example, during the stress of transfusional iron overload in thalassemics, the cellular ferritin content increases in the heart muscle and adult red cells.
Cell-specific differences in ferritin are not limited to the type of iron storage function but also occur in the structure of apoferritin and include differences in amino acid sequence, subunit mobility during electrophoresis in denaturing gels, immunoreactivity, and heterogeneity of charge (e.g. Refs. 2,[16][17][18][19][20][21][22].
Cell specificity of apoferritin structure indicates cell specificity of gene regulation, which can be readily examined using subunit-specific hybridization probes and induction by excess iron. Previous studies showed that iron-induced accumulations of ferritin of up to 40-fold occurred in cells specialized for iron storage, with no change in either the concentration or the subunit ratio of ferritin mRNA in hepatocytes or red cells of the embryonic cell line (4-6), but resulted from the recruitment of stored ferritin mRNA and competitive translation (6,14,23). We now report that for housekeeping ferritin (the bullfrog adult red cell line') induction by iron is accompanied by changes in the ratios of ferritin subunit mRNA and protein; such differences in the mode of induction may be related to the different role of storage iron in the ferritin of the cell types studied. In addition, we show for the first time that there can be a t least three apoferritin subunits encoded in three distinct but related mRNAs, which contrasts with the two subunits, H and L, that had been characterized previously (e.g. ; the three ferritin subunits are of similar size and sequence but differ in the cell specificity of expression and regulation by iron.

Characterization of cDNAs for Three Red Cell Apoferritin
Subunits-The number of subunits in ferritin has been thought to be one or two (2). However, in the ferritin from tadpole red cells a minor third subunit can be detected after electrophoresis of the native protein or the protein synthesized under the direction of poly(A+) RNA from red cells in wheat germ extracts ( Fig. 1). Previously we described a ferritin subunit encoded in cDNA clone pJD5F12 from a tadpole reticulocyte library (5). We now report the isolation and characterization of two other cDNAs and account for the three protein subunits. In the earlier description of pJD5F12 (5) the subunit ident.ity could not be determined because the cloned DNA hybridized to the mRNA for two apparently similar subunits, even though the hybridization was carried out at the T,,, for the cDNA. By increasing the stringency of hybridization (elution at 85 and 95 "C), it was possible to show that pJD5F12 encoded the H(igher) subunit (Fig. 3). In addition, we correct a transversion of the base sequence at  ' lDlO is a cDNA which corresponds to a processed pseudogene, clone G1-10; the homology between the processed pseudogene cloned cDNA, i.e. the 5' UT region, the coding region, and 3' UT regions, is 97%.
241-249 in the coding region, which occurred during typing of the sequence(s) (see Figs. 5 and 6). The cDNA library used to isolate pJD5F12 was also used to isolate clone pJDlD10, which encodes the subunit that is similar to but distinct from pJD5F12; Fig. 2 depicts the sequencing strategy for clone pJDlD10. The information encoded in cDNA clone pJDlDlO is also encoded in genomic clone G1-10, a processed (intronless, with poly(A) tail) pseudogene which is the first observed in amphibia, to our knowledge. cDNA clone pJDlDlO has a coding sequence 85% homologous to that in pJD5F12 ( Table I ing to the 3' UT region, Genomic clone G1-10, the pseudogene, is 97% homologous to pJDlDlO over the 784 nucleotides of the cDNA and 98% homologous in the coding region; the base changes correspond to the conservative substitution of two amino acids in a region of variation of other ferritin sequences (Fig. 6). Of the 1106 bp of genomic clone G1-10 that have been sequenced, 869 bp are flanked by a 16-bp direct repeat which, at the 3' end, includes a stretch of 4 of the 25 As in the poly(A) tail that follow 177 bp, corresponding to the 3' UT region of an mRNA; the direct repeat at the 5' end of the genomic clone occurs at a position -141 bp from the start codon and corresponds to the position of one of the apparent transcription initiation sites for the mRNA encoded in clone pJD5F12 (5). Identification of the protein encoded in cDNA clone lDlO (predicted mass 20,594) and G1-10 (predicted mass 20,345) as the M(iddle) subunit ( Fig. 1) was made after electrophoresis and autoradiography of the peptide produced under the direction of the transcript of GI-10 subcloned flush with a T3 promoter (Fig. 4).
The third apoferritin subunit observed in tadpole red cell ferritin ( Fig. 1) is encoded in cDNA clone pJDlD8, which was isolated from the same reticulocyte cDNA library (5) used for the other cDNA clones described above. pJDlD8 was sequenced by the strategy described in Fig. 2 and identified as a L(ower) subunit by hybrid-select translation (Fig. 3). The DNA insert contains 695 bp, 519 of which correspond to the coding region, a partial 5' UT region (29 bp), and a complete 3' UT region (a polyadenylation signal at -21). The protein encoded in cDNA clone pJDlD8 is distinct, containing 16 serine residues compared to the usual eight or nine, and may contain specific phosphorylation sites previously observed (21).
The subunits encoded by the three cDNAs contain blocks of amino acids (Fig. 4), which are now known to be conserved in the apoferritins of higher vertebrates as well (2), with variable sequences distributed so frequently that the separate genes are indicated. In contrast to the apoferritin subunits in

I1
Cell specificity of expression of ferritin subunit mRNA The results were obtained using "dot blots" of total poly(A') RNA and hybridization of 32P-labeled cDNA probes (see "Experimental Procedures"). The cDNA used for H was pJD5F12, for M was pJDlD10, and for L was pJDlD8 (see Table I  Up to 2% could have been detected. human liver (8), for which electrophoretic mobility in denaturing gels (with discontinuous pH) coincides with differences in mass, the number of amino acids (175, 175, and 172) and the masses predicted from the cDNA sequence (20.5, 20.6, and 19.9 kDa) are very similar for the three apoferritin subunits from tadpole red cell ferritin ( Table I). When tadpole red cell ferritin was analyzed using phosphate buffers at pH 7 in SDS gels (241, the apparent mass corresponded closely to that predicted from the cDNA sequences, i.e. a single subunit of mass 19.6 f 1.0 kDa. Because ferritin mRNA has unusual features of regulation (storage and competition during translation) and because the structure of the 5' UT end appears to be important during translation (e.g. Refs. 25-28), secondary structure was predicted and compared for the 5' UT regions of the mRNA encoded in cDNA clones pJD5F12, pJDlD10, and genomic clone G1-10 using the same parameters in the Zucker program (BIONET). No common features of structure for the pJD5F12 were observed compared to pJDlDlO or G1-10, which were very similar to each other. However, the 5' UT regions are, in general, longer (about 140 nucleotides) compared to globin mRNA (60 nucleotides)* with which the ferritin mRNAs compete in uiuo and in uitro. The physiological significance of the differences in both the nucleotide sequence of the UT regions of the three ferritin mRNAs and the coding regions (Fig. 5) is not clear at this time, but the differential patterns of expression (Tables I1 and 111) indicate that the differences are not trivial.
Cell Specificity of Expression of the Three Ferritin Subunit mRNAs-To determine whether the mRNAs for the three ferritin subunits encoded in the three cDNA clones were differentially expressed in different cell types, the ratios of the subunit mRNAs in the cytoplasm were determined using "dot blot" analyses with labeled plasmid DNA as hybridization probes as previously described (5), using increased stringency (during washing, 0.075 X SSC, 65 "C) to prevent crosshybridization (see Fig. 7).
The results in Table 11 show that mRNA encoded in pJDlDlO (the M subunit) predominates in liver, accounting for about 66% of the ferritin mRNA, with no change in the subunit mRNA composition of liver during metamorphosis. The results with mRNA confirm those obtained for the protein (10, 21). Neither mRNA nor protein encoded by clone pJDlD8 (the L subunit) was detectable.s In contrast to liver, the H subunit encoded by pJD5F12  Effect of iron on reticulocyte ferritin mRNA and protein during animal development of ferritin by RNA hybridization ("dot blots") used total poly(A+) RNA and pJD5F12 under low stringency (5) for Tadpoles were premetamorphic; all animals were treated with iron as previously described (5). The measurement total ferritin mRNA and pJD5F12, pJDlD10, and pJDlD8 under stringent conditions (see "Experimental Procedures" and Fig. 7) for subunit mRNA; measurement by translation used wheat germ and immunoprecipitation of ferritin as described (4). Ferritin concentrations were determined with a polyclonal antiserum by double diffusion in agar (14,39) and subunit composition by electrophoresis in denaturing gels followed by staining and densitometry (21). The values are the mean of 2-6 determinations with the error presented as the standard deviation. Erythrocytes of tadpoles may also be induced to accumulate ferritin by excess iron with an increase from 0.37 f 0.14 to 1.2 f 0.01 mg/100 mg of protein; in contrast, iron has no effect on the ferritin content of erythrocytes of frogs (14). Note the reduction in the hybridizable and translatable ferritin mRNA in frog (housekeeping ferritin) reticulocytes compared to tadpole (specialized-cell ferritin) reticulocytes and the iron-induced increase in L subunit mRNA and protein in frog cells.
The ferritin content was too low in uninduced cells to characterize.
One determination only.
predominates in red cells, accounting for 57% ferritin mRNA, with the mRNA encoded by pJDlD8 accounting for 5% and that by pJDlDlO accounting for 39% of the total ferritin mRNA. However, during metamorphosis the shift in the role of red cell ferritin from that of a specialized cell (storing iron for other cells) to housekeeping (storing iron for intracellular purposes) is apparently accompanied by a reduction in transcription or stability of ferritin mRNA, since the amount of ferritin mRNA is much lower in frog reticulocytes than in tadpole reticulocytes (Table 111). The relative amount of functional (translatable) ferritin mRNA is 21% and of structural (hybridizable) ferritin mRNA is 24% in frog reticulocytes compared to tadpole reticulocytes, with the ratios of ferritin mRNA for the different subunits maintained (Table  11). The Effect of Iron on Expression of the Three Ferritin Subunits and Subunit mRNAs-The induction of ferritin in cells specialized for iron storage, first observed 40 years ago (29), has been shown to depend upon the recruitment of stored ferritin mRNA (3-6), which is then translated very efficiently (4,14,23), since no change in ferritin subunit mRNA was observed during the induction (6). (The apparent difference in hybridizable ferritin mRNA from tadpoles treated with iron (Table 111) is due to variations among experiments because the average of the ratio with iron-treated/control, computed as in Ref. 5 , is 0.92 f 0.24.) T o examine the induction of housekeeping ferritin, red cells of the adult (frog) cell line were studied and compared to embryonic (tabpole) red cells which are specialized for iron storage. Several differences were observed (Table 111). First, the constitutive level of ferritin was lower in frog reticulocytes compared to tadpole red cells. Second, iron could induce accumulations of ferritin to similar levels in reticulocytes of frogs (Table 111) and tadpoles as well as erythrocytes of tadpoles (14). In contrast, iron had no effect on ferritin concentrations in erythrocytes of frogs (14), which emphasizes the distinction in the role of ferritin in the maturing cell of each cell line.6 Third, the ratio of the ferritin Erythrocytes of frogs also have reduced iron uptake compared to the cells of tadpoles (30). subunit mRNAs changed during the induction of ferritin in the reticulocytes of frogs; the ratio of cpm hybridized to poly(A+) RNA in "dot blots" for pJDlD85F12 increases 300% in iron-treated anemic frogs compared to no change (ratio = 112%) in iron-treated anemic tadpoles (Table 111). Finally, the iron-induced change in ferritin subunit mRNAs is reflected in the subunit ratio of the protein; the L subunit mRNA (encoded by pJDlD8) increased from 6 to 16% of the total ferritin mRNA, and the amount of L subunit increased from 4 to 16% of the protein in iron-treated frog red cells (Table 111).

DISCUSSION
Cell specificity of structure among proteins of large families is well known and is exemplified by the lactate dehydrogenases and myosins (31); ferritins also display cell specificity of structure. Variations among the structures of members of protein families in different cell types indicate cell-specific variations in the regulation of gene expression as well (31,32). Cell specificity of ferritin gene expression has been examined in differentiating cells in culture (malignant macrophages, granulocytes, and erythroid cells of the adult cell line (33,34)). At the time of the studies, the number of apoferritin units was thought to be two, designated H and L, and differences in the relative amounts of the two subunits were thought to encompass most, if not all, of the structural variations in apoferritin. In addition, because of the fact that ferritin mRNA is stable and stored (3-6, 23), it appeared that cellular variations in the ferritin of a particular cell type could be explained by variations in the utilization of the stored ferritin mRNA. The results in Figs. 1-4 and Tables 1-111 indicate that a greater complexity of gene regulation is required to account for the cell specificity of ferritin because there are more than two ferritin subunits and because changes in the ferritin subunit mRNA composition can occur under different physiological conditions in the same cell type. The distribution of differences in the coding regions of the three cDNAs is such (Fig. 6) that each mRNA was probably encoded by a separate gene; if not, alternate splicing of a gene with many small exons would be required.
The three subunits encoded in the cDNAs isolated from a tadpole red cell library share sequences with ferritin subunits from humans, rats, and horses (2) that preserve the subunit interactions at the dimeric, trimeric, and tetrameric levels (Fig. 6). Although the three subunits can be distinguished by their mobilities during electrophoresis in SDS gels (Fig. 1) and are designated H(igher), M(iddle), and L(ower), the sizes are very similar (predicted number of amino acid residues: 175, 175, 172; predicted masses: 20.5, 20.6, and 19.9 kDa, respectively). Such variations between mass of the amino acid sequence and the apparent mass determined by SDS-gel electrophoresis occur for other proteins (35, 36) and may represent incomplete binding of SDS. Thus, the use of additional structural information about the ferritins is required before interpreting the results of analysis by electrophoresis in denaturing gels; the absence of such information in the past may be the basis for some apparent discrepancies in the literature.
The subunit composition encoded in ferritin mRNA is different in red cells and liver. In the liver the M subunit ferritin mRNA predominated (66% of the ferritin mRNA) and the L subunit ferritin mRNA was not detectable (<2% ,  Table 11). In contrast to the liver, all three ferritin subunit mRNAs were expressed in red cells with the H subunit ferritin mRNA predominating (57% of the ferritin mRNA) and the L subunit ferritin mRNA present at a low (4%) level (Table  111). In addition, red cells displayed differences related to the developmental stage of the animal. For example, the amount of ferritin mRNA (translatable or hybridizable) of frog reticulocytes was 20% that of tadpole reticulocytes. The constitutive level of ferritin was also proportionately lower (Table  111).
Differences between tadpole and frog reticulocytes were accentuated by iron, even though both cell types used stored mRNA to make the additional ferritin (Table 111). As discussed in the following paragraphs four differences were observed.
1. Accumulation of ferritin induced by iron was restricted to certain stages of erythroid maturation in frogs, i.e. reticulocytes responded to iron (Table 111) but erythrocytes did not (20). In contrast, iron induced the accumulation of ferritin in both erythrocytes and reticulocytes of tadpoles (14).
2. The utilization of ferritin mRNA during iron-induced accumulation of ferritin appeared to be more efficient in reticulocytes of frogs, since one-fifth the amount of ferritin mRNA produced the same amount of accumulated ferritin (Table 111). (Note that the protein is stable (14) so that changes in turnover are unlikely.) 3. The composition of the subunit ferritin mRNAs changed during the induction of ferritin by iron in reticulocytes of frog adults in contrast to reticulocytes of tadpoles (Table 111) and to liver of adult rats (6). In frog reticulocytes the amount of mRNA encoding the L subunit increased from 6 to 16% of the total (the ratio to H subunit increased from 0.11 f 0.03 to 0.31 f 0.11). The results suggest that a change in transcription or stability occurred for the ferritin mRNA encoding the L subunit during the accumulation of ferritin in frog reticulocytes.
4. Ferritin in reticulocytes of iron-treated frogs contained more of the L subunit (from 4 to 16%) than that of tadpoles and proportionately less of the H and M subunits (Table 111). The results in the adult red cell indicate a 60 X increase in the amount of the H and M subunits and a 230 X increase in the L subunit, in contrast to the reticulocytes of iron-treated tadpoles where all three subunits appear to increase 8 X.?
Not only does the regulation of red cell ferritin by iron change during animal development, but the role of iron storage in red cell metabolism changes as well (10). The change in red cell ferritin and ferritin mRNA induced by iron in frog red cells can be explained in terms of the change in the type of red cell iron stores. For example, red cells of tadpoles behave as typical cells specialized for iron storage and provide iron for other cells, i.e. the first generation of adult red cells (10); the constitutive level of ferritin and ferritin mRNA is high (Table 111), and ferritin concentrations increase during red cell maturation in both tadpoles and mice (14,15). On the other hand, iron in the red cells of frogs or red cells of other adults is a housekeeping protein, storing iron for intracellular use (12, 13, 37); the level of ferritin and ferritin mRNA is relatively low (Table 111). Loading tadpole reticulocytes with excess iron to store merely activates a specialized property of the cell. Loading frog reticulocytes, which are at the end of rapid hemoglobin synthesis and iron need, with excess iron to store and for which the cell has no use requires the cell to protect itself by sequestering the excess iron; the superfluous iron appears to stress the cell. Later in maturation the cell will protect itself further by reducing iron uptake (12) and preventing the accumulation of ferritin (14). Thus, the ferritin synthesized in the iron-loaded frog red cells appears to be a stress housekeeping ferritin, and the shift in the mechanism of regulation may be related to the shift in the type of iron storage by the protein. Note that preliminary data show that heat-shock stress also induces ferritin synthesis (38).
In summary, cell-specific features of ferritin gene expression include differences in the concentration of stored mRNA and in the subunit composition of the stored ferritin mRNA. In addition, iron induces a change in relative amounts of the ferritin subunit mRNA for housekeeping ferritin where the mRNA concentration is low and extra iron must be sequestered to prevent toxicity; iron thus appears to influence the transcription or stability of ferritin mRNA for certain subunits. The regulation of housekeeping ferritin contrasts with that of specialized-cell ferritin where the mRNA concentration is high and iron appears only to recruit ferritin mRNA from a large pool for translation without changing the ferritin subunit mRNA composition. The difference between the effect of iron on the regulation and structure of ferritin in different cell types may relate to the role of the extra iron in the cell, i.e. specialized-cell iron stores or stress housekeeping iron stores.
Acknowledgments-The enthusiastic technical assistance of Beverly Parker and Donna N. Freeman and the editorial assistance of Joann Fish is gratefully acknowledged. We thank R. Ann McKenzie for the identification of clones pJDID8 and pJDlDlO as ferritin clones.
'The total accumulation of ferritin in the reticulocytes of irontreated frogs is 47 X compared to 8.2 X in tadpoles. Since the cellular content of ferritin mRNA is comparatively low in frog reticulocytes, translation (and/or stabilization of the protein) appears to be unusually efficient. Calculation of the accumulations of each individual ferritin subunit can be made if the subunit composition is assumed to be the same for ferritin in uninduced frog and tadpole reticulocytes as it is in induced tadpole reticulocytes. The justification for such an assumption is the fact that the subunit mRNA ratios are the same for the cell types considered and that comparison of the subunit composition of the protein and the mRNA for iron-induced tadpole red cells indicates translation of the subunit ferritin mRNAs in proportion to their relative concentration; the direct analysis of ferritin from uninduced cells is difficult because of the lower (%-I&) amounts of protein.