Nuclear translocation of ferritin in corneal epithelial cells

In eukaryotic cells, a unique set of proteins is segregated to the nucleus by selective transport from the cytoplasm through the nuclear pore complex (NPC). The NPC is a large supramolecular structure, composed of more than 100 polypeptides, with a predicted molecular mass of 120 MDa (reviewed by Davis, 1995; Allen et al., 2000; Ryan and Wente, 2000). Most large nuclear molecules are transported through the NPC by signal and energy-dependent pathways. The best characterized signal for nuclear import is the ‘basic type’ nuclear localization signal (NLS) (reviewed by Dingwall and Laskey, 1991). This NLS consists of a cluster of basic amino acids or bipartite basic amino acids. During the translocation process, the NLS becomes associated with an NLS receptor. Subsequently, other molecules become involved, including importin β (P97), which mediates the docking of the NLS-NLS receptor complex with the NPC, and the GTPase Ran/TC4 and NTF2/P10 (reviewed by Melchior and Gerace, 1995; Pante and Aebi, 1996). Recently, a second nuclear translocation pathway has been discovered that does not require a consensus NLS for nuclear translocation (Pollard et al., 1996). Thus, multiple mechanisms for nuclear translocation most likely exist.

Ferritin is an iron-binding molecule that consists of 24 subunits, and has a molecular mass of ∼450 kDa and a diameter of 12 nm (Theil, 1987). The molecule is normally cytoplasmic - presumably because its large size prevents diffusion into the nucleus. Recently, however, we have observed that, in avian corneal epithelial cells, ferritin is predominantly, if not exclusively, a nuclear component (Cai et al., 1997). This nuclear ferritin is a homopolymer, consisting of only heavy chain, ferritin H (Passanitti and Roth, 1989), and is assembled into a supramolecular form that is indistinguishable from the cytoplasmic ferritin found in other tissues (Cai et al., 1997). Functionally, our studies suggest that this nuclear ferritin protects corneal epithelial cells from UV-induced DNA damage, most likely by sequestering iron (Cai et al., 1998).

Previously, some studies have examined the subcellular localization of ferritin. Microinjection of ferritin into the cytoplasm of most cells resulted in retention of the molecule within the cytoplasm (Lanford et al., 1986; Paine and Feldherr, 1972). In some cells, however, ferritin could cross into the nucleus, although it did so less efficiently than low molecular weight proteins (Stacey and Allfrey, 1984). However, if the microinjected ferritin was conjugated with a synthetic peptide containing the basic NLS of SV-40 large T antigen, efficient nuclear translocation occurred (Lanford et al., 1986). Thus, under certain conditions, even multimeric ferritin can undergo nuclear translocation. Studies also showed that some ferritin could be found within the nucleus of hepatocytes in animals treated with high, pathological amounts of iron (Smith et al., 1990; Richter, 1961). However, the lag time for this nuclear localization was in the order of weeks or months, and even then there was much less ferritin in the nucleus than in the cytoplasm.

Because, in normal tissues, the nuclear localization of ferritin is highly selective, if not specific, for corneal epithelial cells, it seems that a tissue-specific mechanism for nuclear transport must exist. To examine this mechanism further, we employed transfection analyses of myc-tagged ferritin deletion constructs in corneal epithelial cells to determine: (1) which portion(s) of the molecule are required for translocation; and (2) whether monomeric or polymeric ferritin is the form that undergoes nuclear translocation. For comparison, we performed the same analyses on corneal fibroblasts - a cell type that has a low level of ferritin. In the cultured fibroblasts the ferritin is predominantly cytoplasmic with only an occasional cell showing a low level of signal within the nucleus. Our results suggest that for nuclear transport in corneal epithelial cells, almost all (>83%) of the ferritin H chain must be intact; thus there seems to be no specific region that serves as an NLS. Also, the monomer seems to be the form that undergoes transport in these cells. For fibroblasts, the cytoplasmic localization of the ferritin seems to be regulated predominantly by supramolecular assembly within the cytoplasm, and it is this that prevents diffusion into the nucleus.

Transient transfections were performed on primary cultured cells grown on 35 mm dishes. For the cultures, corneas from 14-day chicken embryos were separated into their epithelial and stromal components after treatment in 0.5% Dispase in PBS for 1 hour at 4°C. The epithelia were rinsed with PBS and further digested in 0.25% trypsin at 37°C for 5 minutes. Then, the cells were cultured in medium containing DMEM:F-12 (1:1), 20% fetal calf serum (heat-inactivated), 1% chicken serum, 5 μg/ml insulin, 10 ng/ml each of human recombinant EGF, penicillin and streptomycin. Fibroblasts were obtained after further digestion of the stromas with type I collagenase (200 U/ml) for 2 hours at 37°C. They were grown in medium containing DMEM, 10% fetal calf serum and penicillin and streptomycin. For studies on the effects of iron, 100 μM ferrous sulfate (Sigma) was added to the normal iron medium, producing a high-iron medium.

All of the cDNA constructs were made by overlapping PCR (Ho et al., 1989) and cloned into the eukaryotic expression vector PCR™-3 (Invitrogen). The cDNA encoding the open reading frame of ferritin-H was tagged by PCR at the 3′-end with a sequence encoding a 12 amino acid human c-myc epitope (Chen et al., 1995). A series of deletion constructs was also generated (shown in Fig. 2). The NLS-containing constructs were made by tagging at the 5′-end of ferritin with the NLS signal from SV-40 large T antigen (Roberts et al., 1987). Plasmid DNAs were purified using a Qiagen maxi prep kit and their identities were confirmed by sequencing. Transfections of primary cell cultures were performed using the calcium phosphate method (Sambrook et al., 1989), Lipofectamine reagent (Gibco BRL) or Fugene-6 reagent (Boehringer-Mannheim) according to the manufacturer’s protocol to optimize the transfection. Cells were harvested 24-72 hours after transfection for evaluation by indirect immunocytochemistry or immunoblotting.

Transfected cells were fixed in ice-cold 4% paraformaldehyde for 10 minutes, washed in PBS, and permeabilized in ice-cold 100% methanol for 10 minutes. The samples were incubated with primary antibody for 45-90 minutes. The primary antibodies used were purified monoclonal antibody 6D11 (Zak and Linsenmayer, 1983), now identified as against ferritin (Cai et al., 1997); monoclonal antibody 9E10 against the human c-myc epitope (Evan et al., 1985); and, as a negative control, monoclonal antibody X-AC9 against type X collagen (Schmid and Linsenmayer, 1985). The samples were then washed thoroughly with PBS and incubated with a rhodamine-conjugated goat anti-mouse IgG secondary antibody for 45-90 minutes. After washing in PBS, samples were mounted in glycerol/PBS (95:5) containing 1 ug/ml of Hoechst dye No. 33258 (Sigma B2883) to stain nuclei. Samples were viewed with a Nikon fluorescence photomicroscope equipped with an epi-illuminator. Photographs were taken at a 1 minute exposure using Kodak Tri-X film or TMAX 400, except for the Hoechst dye, which was exposed for 12 seconds. Pictures were also taken with a digital camera (Spot RT).

Cell lysates from the transfected cells were prepared as described previously (Harlow and Lane, 1988). Briefly, the cells were washed in ice-cold PBS and harvested with a cell scraper. The cell pellets were resuspended in a five-fold volume of ice-cold lysis buffer (1% Triton X-100, 1 mM PMSF, 5 mM iodoacetamide, 0.2 U/ml aprotinin, 1 mM EDTA, 50 mM Tris-HCl, 150 mM NaCl, 1% sodium deoxycholate and 0.5% SDS). The extracts were sonicated five times for 30 seconds each.

Gel electrophoresis was performed under reducing conditions containing 1% β-mercaptoethanol. For denaturing conditions, the samples were boiled for 5 minutes, followed by centrifugation at 12,000 g for 5 minutes at 4°C; for non-denaturing conditions the samples were centrifuged without boiling. Total protein was quantified with a BCA kit (Pierce). 2-4 μg of proteins were solubilized in reducing sample buffer (2% SDS, 10% glycerol, 50 mM Tris-HCl, 1% β-mercaptoethanol and 0.005% Bromphenol blue, pH 6.8) and separated on a 4-20% acrylamide Ready Gel (Bio-Rad) in Tris-glycine-SDS running buffer. The gels were then electrotransferred to ECL^TM nitrocellulose membrane (Amersham). Immunodetection was performed by incubating each blot with monoclonal antibody 9E10 (Evan et al., 1985) for myc-tagged ferritin followed by incubation in 1:10,000 dilutions of horseradish peroxidase-conjugated secondary antibody (Pierce). Positive signals were detected by luminol-based enhanced chemiluminescence (ECL) (Pierce).

To begin to determine the mechanism for the nuclear transport of ferritin, cultured corneal epithelial cells were transfected with a series of deletion constructs for the heavy chain of ferritin (ferritin-H), tagged with a human c-myc epitope for identification. For comparison, corneal fibroblasts were also transfected. The subcellular localization of the products expressed from the transfected constructs was determined by immunofluorescence with antibodies for the myc-tag.

As described previously (Cai et al., 1997), corneal epithelial cells in primary culture maintained a characteristic epithelial appearance, growing in sheets and having a polygonal cell morphology (Fig. 1A,C, in which two nuclei are demarcated by asterisks). In these cells, immunofluorescence shows that the signal for the endogenous ferritin is intense and predominantly, if not exclusively, nuclear (Cai et al., 1997) (Fig. 1B). It is distributed throughout the nucleus, except for the nucleoli (Fig. 1B, arrow).

In cells transfected with a myc-tagged (Chen et al., 1995) full-length ferritin construct, in the majority of the cells, the myc-tagged molecule is also localized to the nucleus (Fig. 1C,D, asterisks). Occasionally, a cell will show some cytoplasmic signal. Preferentially, such cells are located near the edge of an epithelial sheet where outward expansion is occurring. The untransfected cells showed no immunofluorescence signals by anti-myc antibody (not shown).

Because the ferritin sequence does not contain any known consensus NLS, we examined the possibility that a ‘non-standard’ NLS sequence was present within the molecule. To test for this, we transfected corneal epithelial cells with a series of deletion constructs (summarized in Fig. 2). The constructs consisted of deletions from the ends of the molecule, and internal deletions removing portions of the highly-conserved helices and loop structures predicted by structural analysis to be present within the molecule (Harrison and Arosio, 1996a). The results obtained suggest that, in order for nuclear transportation to occur, most of the molecule must remain intact.

The deletion constructs at the C-terminus showed that the last 10 or the last 30 amino acids could be deleted without affecting the nuclear localization (Fig. 2, constructs D14 and D11, respectively). Most of the corneal epithelial cells transfected with these constructs exhibited a nuclear signal for the truncated proteins (see, for example, Fig. 3, D11), similar to that of the full length construct (Fig. 3, top row, HM). (In corneal epithelial cells the endogenous ferritin is virtually all nuclear (designated ‘N’ in Fig. 2), as is most of the ferritin produced by transfections of a full-length ferritin construct (HM) and those deletion constructs that produce nuclear localization (N>C in Fig. 2; constructs D11, D14 and D13).) However, deletion of an additional 20 amino acids abolished the nuclear localization (Figs 2, 3, construct D10). Of potential importance, the additional amino acids (130-150) deleted in this construct remove a portion of a major helical domain (Helix D) (see Fig. 2), raising the possibility of an involvement of molecular conformation in their nuclear transport (see below).

At the N-terminus, again, a small deletion consisting of the first 10 amino acids could be made without affecting the nuclear localization of the molecule (Figs 2, 3 construct D13). However, removal of an additional 30 amino acids (11-41) resulted in a predominantly cytoplasmic localization of the truncated molecule (Fig. 2, construct D12). This larger deletion removes a major portion of a helical domain (Helix A), which again is consistent with an involvement of molecular conformation in nuclear transport.

To test this potential requirement for molecular conformation in nuclear transport further, we made deletion constructs for the other helical domains (Fig. 2: construct DB for Helix B; construct DC for Helix C; construct DD for Helix D). We also made a deletion for another region involved in molecular conformation, the loop region (construct DL). Deletions in any of these regions resulted in a failure of nuclear transportation of the mutant proteins (summarized in Fig. 2). Most of the transfected cells showed cytoplasmic localization of these truncated proteins (see, for example, Fig. 3, DL, DD).

As a control, we examined whether the cytoplasmic localization of any of the constructs that did not undergo nuclear translocation might be an artifact and might be caused by, for example, their insolubility and precipitation in the cytoplasm. For this we tested whether the constructs could undergo nuclear transport if a consensus NLS were attached to them. When the NLS from the SV-40 large T antigen (Roberts et al., 1987) was attached to the 5′-end of these constructs, most of them underwent nuclear transport, for example, Fig. 3, constructs D10-NLS, DL-NLS, DD-NLS.

Taken together, these results show that most, if not all, of the sequence between amino acids 11 and 150 (>83% of the molecule) is necessary for nuclear transport. This suggests that no specific NLS sequence is involved. Instead, the data emphasize the importance of the overall molecular structure of the molecule in its nuclear transport, because the process is inhibited by deletions in helices A, B, C and D, and in the loop region. The only regions of the molecule that do not seem to be required are helix E (see Figs 2, 3, D11), and the N- and C-terminal ends.

To examine the structural form of ferritin that undergoes nuclear transport, we examined whether the transfected proteins that underwent nuclear transport (determined by immunofluorescence) had undergone supramolecular assembly (determined by gel electrophoresis). Supramolecular assembly was evaluated by gel electrophoretic separation of lysates of the transfected cells run under conditions in which polymeric ferritin remains intact (β-mercaptoethanol+SDS, but without boiling) (Passanitti and Roth, 1989), followed by immunoblotting with the anti-myc-tag antibody (Fig. 4). Three of the four constructs that underwent nuclear transport (Fig. 2, HM, D14, D13) had their myc-tagged protein assembled into a ferritin-like supramolecular complex (Fig. 4, top bands). The supramolecular complex(es) that contained the truncated proteins from D13 and D14 showed sizes slightly smaller than that of the full-length construct (HM), whose size itself is similar to that of the endogenous ferritin complex (not shown). In samples boiled before electrophoresis, only the monomeric form of each construct was detectible (not shown) - a feature characteristic of the endogenous nuclear ferritin (Cai et al., 1997) as well as chicken ferritin in general (Passanitti and Roth, 1989).

However, the product of one nuclear-localized deletion construct, D11, remained entirely monomeric, at least within the limits of detection by immunoblotting (Fig. 4, D11, lower band). (The faint high molecular weight band (arrow) seen in D11 is nonspecific, probably resulting from the nonspecific binding of the labeled secondary antibody as described previously (Cai et al., 1997).) This suggests that supramolecular assembly is not necessary for nuclear transport to occur, and raises the possibility that the monomer is one form of the molecule that undergoes transport (but see Discussion). All of the products from the other deletion constructs (see Fig. 2) that remained cytoplasmic were monomeric, as detected by immunoblotting (see, for example, Fig. 4, D10).

These data from immunoblotting suggest that supramolecular assembly from full-length monomers is more efficient than it is from those containing deletions. As can be seen in Fig. 4, in the cells transfected with the full-length ferritin construct (HM), almost all of the myc-tagged product has been assembled into the supramolecular complex. However, in the deletion constructs D13 and D14, in addition to the supramolecular complex, truncated myc-tagged monomers were also present (Fig. 4, lower molecular weight bands), suggesting slower assembly of the complex.

Unlike the epithelial cells, embryonic corneal fibroblasts in vivo have little detectible endogenous ferritin, if any at all (Zak and Linsenmayer, 1983). In vitro, however, they initiate synthesis of the molecule. This is found in the cytoplasm - except for an occasional cell that has some in the nucleus (Cai et al., 1997). Thus, we reasoned that transfections of this cell type would serve as a control for the transfections of corneal epithelial cells and might also uncover additional parameters involved in determining the intracellular localization of the ferritin.

Transfections of corneal fibroblasts were performed using the same series of myc-tagged ferritin constructs used for the transfections of the corneal epithelial cells. Representative data are shown in Fig. 5 and are summarized in Fig. 2. In fibroblasts, the full-length myc-tagged ferritin from construct HM behaved similar to the endogenous ferritin. In the majority of cells, it was predominantly, if not exclusively, cytoplasmic (Figs 2, 5, HM).

Deletion of 10 amino acids from either the N-terminal (Fig. 5, D13) or the C-terminal (D14, not shown) ends of the molecule did not appreciably alter the cytoplasmic localization of the protein from that observed after transfection with the full-length construct (HM). The only difference noted was a slight increase in the number of cells showing both nuclear and cytoplasmic localization.

Transfections of constructs with larger deletions from the ends of the molecule showed an appreciable increase in the nuclear signal (summarized in Fig. 2). When 30 amino acids (Fig. 5, D11) or 50 amino acids (D10, not shown but summarized in Fig. 2) were deleted from the C-terminal end, or 40 amino acids (D12) were deleted from the N-terminal end (not shown but summarized in Fig. 2), most of the transfected cells now showed equivalent nuclear and cytoplasmic signals, with only an occasional one having a solely cytoplasmic localization. Likewise, deletions of the internal helical and loop regions of the molecule resulted in equivalent nuclear and cytoplasmic signals (DB, DC, DL and DD, not shown but summarized in Fig. 2) (see also Discussion). The untransfected cells showed no immunofluorescence signals by anti-myc antibody (not shown).

We confirmed that the constructs that remained cytoplasmic (e.g. HM and D13), or partitioned between the cytoplasm and nucleus (e.g. D11), were, in fact, capable of undergoing nuclear transportation if given a proper signal. This was shown by transfections of fibroblasts with constructs HM, D13 and D11, to which the NLS of SV-40 large T antigen was added, yielding constructs HM-NLS, D13-NLS and D11-NLS, respectively. All of these NLS-containing constructs produced a nuclear localization (Fig. 5, bottom panels).

Combining these data obtained from the fibroblast transfections with the various constructs with the ability of the products to undergo supramolecular assembly, as determined by gel electrophoresis (Fig. 4 and data not shown), a common property emerges. The constructs whose products remain largely cytoplasmic (e.g. D14 and D13) are capable of undergoing supramolecular assembly, whereas those that partition between the cytoplasm and nucleus (e.g. D11) remain monomeric (see Discussion).

Our previous studies (Cai et al., 1997), as well as those presented here, demonstrate that ferritin is a nuclear component of avian corneal epithelial cells. Other cell types, at least in vivo, contain little if any of the molecule in this subcellular location. We have also observed that the ferritin in this location is strongly protective against UV-induced damage to nuclear DNA (Cai et al., 1998).

To our knowledge, the only other reports of nuclear ferritin in vertebrates are in hepatocytes from animals overloaded with pathological doses of iron (Smith et al., 1990; Richter, 1961). Even under these conditions, the level of the nuclear ferritin was an order of magnitude lower than that of the cytoplasmic ferritin, and it did not appear until 24 weeks after iron-overloading (Smith et al., 1990). In such pathological conditions, the ferritin in the nucleus might result from slow, passive diffusion (see also Stacey and Allfrey, 1984).

The in vitro transfection studies presented here confirm that corneal epithelial cells have a mechanism for the rapid transportation of ferritin into their nuclei. Ferritin does not contain the standard basic monopartite or bipartite NLS found in most nuclear proteins (Dingwall and Laskey, 1991). Nor does it contain the recently described M9 NLS (Pollard et al., 1996) that has been implicated in the receptor-mediated nuclear importation of the ribonucleoprotein, hnRNP A1 (Siomi and Dreyfuss, 1995). Obviously, even though there is no known NLS within the ferritin sequence, this does not in itself preclude a non-standard NLS. However, the transfection studies of corneal epithelial cells with the deletion constructs suggest that this is unlikely. These transfections failed to detect any single region of the molecule that might function as a NLS, because deletions throughout ∼80% of the molecule curtailed nuclear transport. Therefore, if an NLS sequence did exist, it would encompass almost the entire molecule.

However, other possible mechanisms exist for the tissue-specific nuclear localization for ferritin, including: (1) a specialization of the corneal epithelial nucleus itself; (2) a non-standard NLS within the ferritin chain that is cryptically masked in cell types other than the corneal epithelium; and (3) a nuclear chaperone that could carry ferritin piggy-back into the nucleus. That the nucleus of corneal epithelial cells has transport characteristics different from those of other cell types, such as the corneal fibroblast, is suggested by the behavior of the truncated proteins from certain of the transfected ferritin deletion constructs. It is generally believed that proteins <40 kDa are capable of free diffusion across the nuclear pore complex. Therefore, because the truncated proteins that are incapable of undergoing supramolecular assembly are approximately half this size, they should be expected to diffuse into the nucleus. For the corneal fibroblasts, this is what we generally observed, with the truncated proteins partitioning between the nucleus and cytoplasm. Corneal epithelial cells, however, behave differently. In corneal epithelial cells the truncated ferritin chains that have lost the ability to undergo nuclear transport remain cytoplasmic - with little if any in the nucleus. Therefore, it seems that the nucleus/nuclear pore complex of the corneal epithelial cell does not allow passive diffusion of these monomers to occur. Whether the nuclear pore complex of the corneal epithelial cell is also specialized for the nuclear transport of the intact ferritin chain remains to be tested.

The second alternative - that ferritin does have a non-standard NLS that is cryptically masked in all cell types except for those of the corneal epithelium - is not addressed and cannot be definitively eliminated by the studies reported here. However, we think it unlikely that this is the mechanism used by the corneal epithelial cells for their nuclear localization.

Instead, we currently favor the involvement of a corneal epithelial-specific nuclear chaperone, a possibility with which certain of our preliminary observations (J. Fitch et al., unpublished) is consistent. A partial cDNA for a multi-domain molecule has been isolated with characteristics indicative of the putative nuclear chaperone, including specificity for corneal epithelial cells (by in situ hybridization) and potential ferritin-binding and nuclear localization domains (from analysis of the sequence). Studies are underway to examine functionally whether this molecule is capable of transporting ferritin into the nucleus.

Whichever mechanism is involved in the nuclear transport, the transfection studies suggest that the ferritin monomer is potentially one form that may be involved. They also suggest that this transportation seems to be an active process.

That the monomer can undergo nuclear transport is suggested by the transfections with deletion construct D11 which is missing 30 amino acids from its N-terminal end (see Fig. 2). By immunofluorescence, this truncated protein undergoes transport into the nucleus. However, by immunoblotting it remains monomeric, suggesting that it is unable to participate in supramolecular assembly. This result does not, however, address whether, in the normal situation, it is the monomer that undergoes nuclear transport, versus the supramolecular ferritin complex or possibly even some intermediate in the supramolecular assembly process. Conceptually it is possible that the supramolecular form can undergo nuclear transport. In a previous study on the NLS of SV-40 large T antigen, Lanford et al. used the supramolecular ferritin complex as a test protein (Lanford et al., 1986). They observed that conjugating this NLS to the ferritin complex resulted in its nuclear transport. Whether this can also occur with the naturally occurring nuclear transport in the corneal epithelial cell is not yet known.

Lastly, it seems that the epithelial cell nuclear transport involves an active mechanism rather than passive diffusion. If diffusion were the sole mechanism involved, in the transfection studies the smaller truncated proteins would be expected to preferentially enter the nucleus. Instead, nuclear localization was found only with the largest constructs (i.e. the one for the intact chain and the two with small N- and C-terminal deletions). Therefore, active transport is likely to be involved. Precedents for the active nuclear transport of small molecules exist in histone H1 (Breeuwer and Goldfarb, 1990) and calmodulin (Pruschy et al., 1994). These two nuclear proteins are similar in size to the ferritin monomer; neither has a classical NLS, and their translocation seems to be mediated by a cytosolic factor.

The immunoblotting analyses of the transfected corneal epithelial cells confirm previous work on the molecular domains required for ferritin supramolecular assembly (Levi et al., 1989; Levi et al., 1988). The data showed that the myc-tagged protein from the transfected full-length construct could participate in supramolecular assembly, as could the deletion constructs from which the first 10 or the last 10 amino acids were removed. Thus, the N-terminal end up to helix A, and the C-terminal end up to helix E are not required. Further deletions from the ends, or from the internal helical domains or the ‘loop’ domain, abolished the ability of the chain to participate in assembly. This is consistent with other, previous studies (Levi et al., 1989; Levi et al., 1988) showing that deletions of the first 13 or the last 22 amino acids from the human ferritin-H did not affect assembly, whereas deletion of the last 28 amino acids (which included six amino acids in helix D) did.

The results with corneal fibroblasts show that cell-specific differences exist in the behavior of the newly synthesized ferritin chains. In cultured fibroblasts the signal for the endogenous ferritin was cytoplasmic and much less than that in corneal epithelial cells. Even in the occasional fibroblast which had some nuclear ferritin, the signal was relatively weak.

In the fibroblast, one factor that appears to be involved in retention of the ferritin within the cytoplasm is supramolecular assembly. In transfectants of this cell type, we observed that the ferritin constructs whose products are retained within the cytoplasm are also able to assemble into the supramolecular complex. Conversely, those truncated proteins that partition between the cytoplasm and nucleus cannot undergo such assembly. From these results, it seems most likely that, in fibroblasts, supramolecular assembly results in a complex that is too large to allow passive diffusion through the nuclear pore complex. However, when such assembly cannot occur, as for the constructs with larger deletions, the ferritin monomers can enter the nucleus by passive diffusion, resulting in the observed nuclear/cytoplasmic partitioning.

We thank John Fitch and Marion Gordon for their helpful discussions and comments on the manuscript. We also thank Raquell Holmes, Eileen Gibney, Kevin Murphy and Emmanuel Zycband for technical assistance. This work was supported by grant EY13127 from The National Eye Institute of the NIH.