Ferritin is a symmetric, 24-subunit iron-storage complex assembled of H and L chains. It is found in bacteria, plants, and animals and in two classes of mutations in the human L-chain gene, resulting in hereditary hyperferritinemia cataract syndrome or in neuroferritinopathy. Here, we examined systemic and cellular ferritin regulation and trafficking in the model organism Drosophila melanogaster. We showed that ferritin H and L transcripts are coexpressed during embryogenesis and that both subunits are essential for embryonic development. Ferritin overexpression impaired the survival of iron-deprived flies. In vivo expression of GFP-tagged holoferritin confirmed that iron-loaded ferritin molecules traffic through the Golgi organelle and are secreted into hemolymph. A constant ratio of ferritin H and L subunits, secured via tight post-transcriptional regulation, is characteristic of the secreted ferritin in flies. Differential cellular expression, conserved post-transcriptional regulation via the iron regulatory element, and distinct subcellular localization of the ferritin subunits prior to the assembly of holoferritin are all important steps mediating iron homeostasis. Our study revealed both conserved features and insect-specific adaptations of ferritin nanocages and provides novel imaging possibilities for their in vivo characterization.
IRON is an essential element of aerobic life. Cells have evolved highly regulated molecular pathways that ensure iron incorporation into heme (Ponka 1997) or formation of iron-sulfur clusters (Rouault and Tong 2005). Cellular and systemic iron levels are tightly regulated to ensure bioavailability and protect from the hazards of iron overload (Hentze et al. 2004). Ferritin, a heteropolymer composed of H and L subunits, acts as the primary iron-storage molecule (Harrison and Arosio 1996). The H subunit contains a ferroxidase center, which enables the mature heteropolymer to oxidize soluble ferrous iron, whereas the L chain provides the nucleation centers for deposition of the ferrihydrite mineral (Santambrogio et al. 1996).
In mammals, transcriptional control of the ferritin genes influences the relative ratio of H to L chains in different cell types (Torti and Torti 2002; Pham et al. 2004). Translation of ferritin proteins is regulated by the binding of either of the two iron regulatory proteins (IRPs) to an iron responsive element (IRE) located on the 5′-untranslated region (UTR) of the respective mRNAs (Pantopoulos 2004; Rouault 2006). Hereditary hyperferritinemia cataract syndrome, a disease in which ferritin L-chain IRE mutations interfere with appropriate translational repression, illustrates the physiological importance of the IRP/IRE system (Cazzola 2002; Rouault 2006). Moreover, adult mice lacking IRP2 overexpress ferritin and develop variable degrees of late-onset neurodegeneration (Lavaute et al. 2001; Smith et al. 2004; Galy et al. 2006) and anemia (Cooperman et al. 2005; Galy et al. 2005). Negative consequences of chronic ferritin H-chain overexpression have been verified in aging mice (Kaur et al. 2006). Conversely, protection from oxidative stress has been shown in young mice that overexpress ferritin H-chain, because of the iron-chelating properties of ferritin (Kaur et al. 2003; Wilkinson et al. 2006). A complete null for ferritin H-chain has been generated in mice; homozygous animals die in utero, whereas heterozygotes exhibit signs of mild iron deficiency (Thompson et al. 2003). Finally, mutations in the human ferritin L-chain lead to neurodegeneration in a condition described as neuroferritinopathy (Levi et al. 2005). Altogether, these results underscore the importance of ferritin regulation in mammalian health.
In Drosophila, Ferritin 1 heavy chain homolog (Fer1HCH) and Ferritin 2 light chain homolog (Fer2LCH) encode the ferritin subunits that compose the major, secreted form of ferritin (Charlesworth et al. 1997; Georgieva et al. 1999, 2002). The crystal structure of secreted ferritin from Trichoplusia ni revealed a symmetrical arrangement of H and L chains (Hamburger et al. 2005). Inter- and intrasubunit disulfide bonds were shown to be important for the folding/assembly of T. ni ferritin, and the respective cysteine residues mediating these bonds were also conserved in Drosophila melanogaster, suggesting that the ferritins of the two species share the same mode of assembly (Hamburger et al. 2005). The Fer1HCH amino acid residues that are required for ferroxidase activity in mammals were conserved in the insect ferritin structure, and a predicted Fer2LCH ferrihydrite nucleation site formed by the L-chains was also found (Hamburger et al. 2005).
As in vertebrates, the IRE/IRP system functions in Drosophila (Rothenberger et al. 1990; Missirlis et al. 2003). A functional IRE is present in the 5′-UTR of the Fer1HCH mRNA, but only in certain splice variants that are preferentially encoded under iron-limiting conditions (Lind et al. 1998; Georgieva et al. 1999). In contrast, no IRE is present in Fer2LCH mRNA (Georgieva et al. 2002). IRP homologs are expressed in the fly (Muckenthaler et al. 1998), and one homolog (IRP-1A) has been shown to bind to IREs from both Drosophila and mammals (Lind et al. 2006).
Intracellular localization of ferritin in many insects also differs from mammals. Ultrastructural studies, combining electron microscopy and energy electron-loss spectroscopy, have revealed the presence of Calpodes ferritin in intracellular membrane compartments (Locke and Leung 1984). Drosophila Fer1HCH and Fer2LCH subunits contain signal peptides that direct them to the endoplasmic reticulum upon translation. Fer1HCH also contains a predicted N-glycosylation site (Nichol et al. 2002). The two subunits are predominantly expressed in the midgut and are also abundant in hemolymph, where ferritin may transport iron for nutritional needs of Drosophila tissues (Georgieva et al. 2002).
This article shows that mutational inactivation of either Fer1HCH or Fer2LCH in Drosophila, as well as the disruption of the ferroxidase center of Fer1HCH, results in developmental arrest and fly embryonic lethality. We characterize a novel fly strain expressing GFP-tagged Fer1HCH and show that GFP-Fer1HCH is incorporated into endogenous functional ferritin. We use this strain to study induction and trafficking of ferritin in the fly midgut, the major iron-storing organ in the insect.
MATERIALS AND METHODS
D. melanogaster stocks:
Fly strain Fer1HCHG188/TM3 was sent to us by Lynn Cooley and is available at http://flytrap.med.yale.edu (Kelso et al. 2004). This line was generated from a screen utilizing a mobile exon encoding GFP carried in a P-transposable element (Morin et al. 2001), which landed in the second intron of the Fer1HCH gene (Figure 2A). Fer1HCH451/TM3, Sb, ry and Fer2LCH35/TM3, Sb, ry were generated during a large-scale mutagenesis screen (Spradling et al. 1999) and were obtained from the Bloomington Drosophila Stock Center (Indiana University) (nos.11497 and 11483, respectively). To generate UAS-Fer1HCH and UAS-Fer2LCH flies, expressed sequence tags LD03437 and LD01936 were obtained from the Berkeley Drosophila Genome Project and subcloned into pUAST. Clone LD03437 contains the Fer1HCH IRE, but we have omitted it from the transgenic construct using a downstream XhoI restriction site present within the 5′-UTR for cloning. To generate ferroxidase-inactive UAS-Fer1HCH* we have mutated E87 to K87 and H90 to G90 (Cozzi et al. 2000). The resulting plasmids were verified by sequencing and injected into embryos using conventional techniques (Rubin and Spradling 1982). Overexpression was accomplished by means of the UAS/Gal4 binary expression system (Brand and Perrimon 1993). All phenotypes reported were confirmed with independent transgene insertions and we generated recombinant chromosomes (on the X and the third) carrying both UAS-Fer1HCH and UAS-Fer2LCH. Levels of ferritin expression were assessed by Western blotting and were found increasingly elevated in the following genotypes: (1) induction from the third chromosome recombinant +/+; Actin-Gal4/+; UAS-Fer1HCH, UAS-Fer2LCH/+; (2) induction from the X chromosome recombinant UAS-Fer1HCH, UAS-Fer2LCH/+; Actin-Gal4/+; +/+; and (3) induction from the X and third chromosomes UAS-Fer1HCH, UAS-Fer2LCH/+; Actin-Gal4/+; UAS-Fer1HCH, UAS-Fer2LCH/+.
Biochemical assays and RNA in situ hybridization:
Antibodies, separation of proteins by SDS–PAGE, and Western blots (for reducing conditions, we add β-mercaptoethanol, boil the samples for 10 min, and separate on 12% acrylamide; for nonreducing conditions, we omit β-mercaptoethanol and do not heat the samples prior to separating on 6% acrylamide), as well as the assay for in vivo loading of ferritin with 55FeCl3, were all performed in triplicates as described before (Missirlis et al. 2006). Methods for mRNA localization and photographic imaging are also provided in a previous publication (Brody et al. 2002).
Prussian blue staining and immunohistochemistry:
Midguts from larvae were dissected in PBS and quickly transferred to 4% formaldehyde in PBS for 30 min. Following washes with PBS, the tissue was permeabilized by treating with 1% Tween in PBS for 15 min. For detection of ferric iron, the samples were incubated in the dark with Prussian blue stain [2% K3Fe(CN)6, 2% HCl] for 45 min. Five washes with water followed and the preparations were mounted in water on glass slides (mounting in PBS will result in loss of the blue stain due to the change in pH) and imaged on a Nikon microscope (see below). For immunohistochemistry, we used the primary antibodies mouse anti-GFP (1:1000) from Molecular Probes (Eugene, OR) (A-11120) in combination with rabbit anti-Lava lamp (1:250) (a gift from John Sisson; Sisson et al. 2000) or rabbit anti-Fer2LCH (1:500) (Missirlis et al. 2006), followed by secondary fluorescent antibodies Alexa Fluor 488-conjugated goat anti-mouse (1:1000) from Molecular Probes (A-11029) and Alexa Fluor 546-conjugated goat anti-rabbit (1:1000) from Molecular Probes (A-11035). Preparations were mounted on slides with Fluoromount-G (Southern Biotechnology Associates, Birmingham, AL) in preparation for microscopy.
Images of GFP-ferritin in freshly dissected intestines mounted in PBS were captured using a Nikon ECLIPSE E600 microscope linked to a Nikon digital camera DXM1200F and were further processed by ACT-1, the application program for the DXM1200F digital camera and Adobe Photoshop. GFP was visualized using a Nikon mercury lamp and the B-2E/C FITC filter. Confocal microscope images of dissected tissues were captured with a 40×/1.2 NA water objective on a Zeiss 510 Meta inverted microscope using the 488-nm line of an argon laser with a 505-550 emission filter for GFP and the 543-nm helium-neon laser line with a 560–615 emission filter for Alexa Fluor 546-conjugated antibodies.
Conservation of the IRE in the Drosophila genus:
Two types of mutations in the IRE of the human L-ferritin gene cause hereditary hyperferritinemia cataract syndrome: mutations that disrupt base pairing of the stem loop and mutations that directly affect the specific CAGUG sequence of the loop and the cytosine of the IRE bulge (Cazzola 2002; Rouault 2006). The mutations that cause disease are detrimental to the defining features of IREs that are conserved from humans to flies (Figure 1). A survey of many sequenced Drosophila species confirmed that Fer2LCH mRNAs lack IREs, which are present only in the 5′-UTR of Fer1HCH transcripts. Comparison of the new IREs identified several nonconserved nucleotides between species, but in all cases the overall structure of the stem loop remained intact and the nucleotides implicated in disease were conserved or altered in ways that were compatible with base pairing (Figure 1). Considering the finding that Fer2LCH mRNA lacks the IRE and is not predictably regulated by IRP-1A, we asked whether other post-transcriptional control mechanisms govern Fer2LCH expression.
Characterization of P-element mutants in the ferritin genes:
To study post-transcriptional regulation of Fer1HCH and Fer2LCH in Drosophila, we utilized flies with P-element insertions in these genes. We focused on three P elements: Fer1HCH451 and Fer2LCH35, which result in a loss-of-function mutant for each of the respective ferritin chains, and Fer1HCHG188, which adds GFP to the H subunit (Figure 2A). Fer2LCH35 is predicted to cause a genetic null mutation because the transposable element disrupts the open reading frame, whereas in Fer1HCH451 the P element resides in intronic sequences and the mechanism by which it disrupts gene function is unknown (Figure 2A). When homozygous, these insertions are embryonic or first instar larval lethal. Each insertion present over a deficiency chromosome lacking both genes also resulted in embryonic lethality, indicating that both ferritin-subunit chains perform essential functions in Drosophila.
To investigate possible homeostatic interactions between the two proteins, their expression in heterozygous adult flies was examined by Western blot analysis (Figure 2B). As expected, Fer1HCH levels were markedly reduced in Fer1HCH451/+ heterozygotes and the same was true for Fer2LCH levels in Fer2LCH35/+ flies. Unexpectedly, endogenous levels of Fer2LCH were also low in Fer1HCH451/+, and Fer1HCH levels were low in Fer2LCH35/+ flies. Because trans-heterozygous (Fer1HCH451/Fer2LCH35) flies are fully viable, we hypothesized that neither of these insertions would directly affect the transcriptional activity of both genes simultaneously. To address the potential role of the P-element insertions on transcription at the locus, we ubiquitously overexpressed Fer1HCH in Fer1HCH451/451 homozygous flies and restored adult viability (Figure 2C; see also materials and methods). When Fer1HCH was overexpressed in homozygous Fer2LCH35/35 flies, viability was not restored. The converse was true for overexpression of Fer2LCH (Figure 2C). These results indicate that the respective P elements interfere specifically with Fer1HCH and Fer2LCH expression and that reduction in levels of the alternate chain in each mutant is most likely due to post-transcriptional regulation.
We also generated flies that express Fer1HCH with a mutation that inactivates the ferroxidase activity, on the basis of analogy to the ferroxidase-null human allele (Cozzi et al. 2000). Expression of the mutated transgene was not able to functionally substitute for Fer1HCH and rescue the lethality of Fer1HCH451/451 flies. Thus, the ferroxidase activity of Fer1HCH provides an essential function in vivo and is likely required for iron loading of the ferritin shell.
Genetic and biochemical characterization of the GFP-ferritin trap line:
A GFP-containing P element integrated into the second intron of Fer1HCH was used to study ferritin expression (Figure 2A) (Morin et al. 2001). Intron/exon donor and acceptor splice sites flank the GFP sequence on the P element; consequently, all Fer1HCHG188 mRNA types (IRE +/−) are predicted to contain the GFP exon, which should encode GFP in frame with the Fer1HCH protein (Figure 2D). Therefore, post-transcriptional regulation of GFP-Fer1HCH mediated by the IRP-1A (Lind et al. 2006) would remain unaffected by the GFP exon. GFP is inserted two amino acids downstream of the predicted cleavage site for the Fer1HCH signal peptide (Charlesworth et al. 1997), which is encoded from the second exon of the gene (Figure 2D). Thus, the full-length mature polypeptide is predicted to translocate to the endoplasmic reticulum and contain the GFP attached to the N terminus, after the signal peptide is cleaved.
To test whether the GFP-Fer1HCH fusion polypeptide was generated as predicted from the locus containing the G188 P element, extracts from adult wild-type (+/+) flies or flies heterozygous for the GFP-insertion line (G188/+) were separated by SDS–PAGE electrophoresis under reducing conditions. Western blots were probed with antisera raised against Fer1HCH or Fer2LCH peptides (Figure 2E). Results indicated the presence of two immunoreactive bands with the Fer1HCH antibody in Fer1HCHG188/+ lysates (Figure 2E, right blot). The higher-molecular-weight species migrated at the predicted size of 50 kDa, consistent with a 27-kDa GFP addition to the 23-kDa Fer1HCH chain. Levels of the Fer2LCH did not change compared to wild-type lysates (Figure 2E, left blot).
To address whether the expressed GFP-Fer1HCH polypeptide contributed to the formation of the holoferritin in the transgenic insect, we separated the same extracts used in Figure 2E by SDS–PAGE electrophoresis under nonreducing conditions, which have previously been shown to preserve higher-molecular-weight ferritin heteropolymers (Missirlis et al. 2006). In extracts from wild-type flies we observed a single ferritin species that migrated close to mouse liver ferritin (Figure 2F, left blots). As expected from the lower levels of both ferritin subunits in Fer1HCH451/+ heterozygotes, these flies expressed less total ferritin. Conversely, in extracts from heterozygous Fer1HCHG188/+ flies that expressed GFP-tagged ferritin, several higher-molecular-weight bands could be detected. Thus, ferritin polymers are formed in these flies and the different molecular-weight species most likely reflect varying ratios of Fer1HCH and GFP-Fer1HCH chains in polymers. Importantly, no ferritin composed solely of native Fer1HCH and Fer2LCH was detected in the heterozygous Fer1HCHG188/+ flies, suggesting that the GFP-tagged ferritins are the functional ferritins of these animals. However, homozygous Fer1HCHG188/G188 and heteroallelic Fer1HCHG188/451 embryos failed to develop into larvae, indicating that the fly cannot survive when all Fer1HCH subunits are tagged with GFP. We wondered if GFP tagging of all Fer1HCH subunits hindered ferritin assembly or whether it obstructed access of an iron carrier to the assembled ferritin. If failure to assemble ferritin with 12 GFP-Fer1HCH subunits was the cause of lethality, we expected to rescue the lethality of Fer1HCHG188/451 flies by overexpression of the mutant ferroxidase construct UAS-Fer1HCH*. This expectation arose from the fact that GFP-Fer1HCH should provide intact ferroxidase centers to the holomer, and mutant Fer1HCH* should allow normal assembly. However, overexpression of wild-type UAS-Fer1HCH rescued the lethality of Fer1HCHG188/451 flies, but UAS-Fer1HCH* did not (Figure 2C), suggesting that a more likely explanation for the lethality of embryos expressing only GFP-Fer1HCH subunits is that the GFP tag blocks Fer1HCH-interacting proteins from accessing the ferroxidase centers of holoferritin.
GFP-ferritin is iron loaded:
We sought to demonstrate that the GFP-tagged ferritin in the Fer1HCHG188/+ heterozygous flies was indeed able to sequester iron and substitute for nontagged holoferritin. To this end, we fed adult flies with radioactive iron in the form of 55FeCl3. Nitrilotriacetate was added to maintain 55Fe3+ in solution at the neutral pH of food, and whole-fly homogenates were prepared after the insects were allowed to feed for 24 hr. Autoradiographs of the nondenaturing SDS–PAGE gels clearly indicated that iron was incorporated in ferritin from wild-type or Fer1HCH451/+ flies (Figure 2F, right, lanes 1 and 2) but also in GFP-tagged ferritin (Figure 2F, right, lane 3).
The observation that the same amount of iron was associated with half the amount of ferritin protein in the heterozygous Fer1HCH451/+ strain (Figure 2F, middle lanes) suggests that the iron load per ferritin holomer can increase when total ferritin decreases. Also, the absence of 55Fe incorporation in the heterozygous Fer1HCHG188/+ strain at the same molecular weight where putative residual normal ferritins would run suggests that these flies survive on ferritin containing both GFP-tagged and nontagged Fer1HCH subunits. Our speculation that GFP may partially inhibit iron loading is consistent with our findings that assembled ferritin molecules with fewer GFP tags (distinguished by their faster migration on Figure 2F) are more heavily iron loaded than assembled ferritin molecules with many GFP-Fer1HCH subunits (migrating more slowly on Figure 2F). Thus, although the GFP-Fer1HCH in solo is not fully functional, it coassembles in vivo with Fer1HCH to form functional fluorescent ferritin.
Overexpression of ferritin requires coexpression of both H and L subunits:
Since iron overload is known to induce ferritin expression in vertebrates and flies alike, we sought to test the consequences of overexpression of either ferritin H or L genes under normal or low iron levels. Immunoblotting of lysates prepared from whole fly extracts revealed that ubiquitous overexpression of either of the two single chains individually was not sufficient to significantly alter total ferritin amounts or the relative ratio between the two chains (Figure 3A). In contrast, when both UAS-Fer1HCH and UAS-Fer2LCH were simultaneously activated with the same Actin-Gal4 driver, robust overexpression was achieved in both sexes. Expression of the ferroxidase inactive UAS-Fer1HCH* was also demonstrated when coexpressed with UAS-Fer2LCH (Figure 3A, right). Overexpression of ferritin was also revealed when the proteins were separated in nonreducing gels (Figure 3C, left). We confirmed successful expression from the transgenes by performing RT–PCR using primers that were specific to the transgenic mRNA (Figure 3B).
Phenotypes associated with ferritin overexpression:
Ferritin overexpression in mammalian cells causes functional iron deficiency due to iron chelation (Cozzi et al. 2000; Wilkinson et al. 2006). However, no significant alteration of total radioactive iron associated with overexpressed ferritin was detected after 24 hr of feeding (Figure 3C, right). If the feeding period was extended for 5 days, a slight increase was seen in sequestration of radiolabeled iron into the ferritin of overexpressors (data not shown), but overall (see also Figure 2F) the results show that, at least in Drosophila, ferritin protein levels are not the determining factor controlling iron storage into ferritin. Overexpression of ferroxidase-inactive UAS-Fer1HCH* together with UAS-Fer2LCH in the presence of functional endogenous Fer1HCH subunits results in ferritin heteropolymers that are still potent iron-storage complexes.
We have recently shown that overexpression of Fer3HCH, a homopolymeric mitochondrial ferritin composed entirely of Fer3HCH chains, caused female-specific resistance to paraquat (Missirlis et al. 2006). Here we show that overexpression of Fer1HCH or Fer2LCH alone is not sufficient to confer paraquat resistance, but coexpression of either Fer1HCH or Fer1HCH* in concert with Fer2LCH confers greater survival to a paraquat challenge (Figure 3D), underscoring the functional cooperation of the ferritin subunits in vivo. As with flies overexpressing mitochondrial ferritin, resistance to oxidative stress is not observed in males that overexpress Fer1HCH and Fer2LCH (Figure 3D, right). These results were corroborated by using different driver lines (FB-Gal4 and Elav-Gal4) and also by using hydrogen peroxide as a stressor (data not shown). We speculate that overexpression of ferritin triggers a developmental signal, possibly related to iron availability, that is specific in females and could relate to the complex nutritional regulation that allocates resources toward reproduction, energy storage, or metabolic activity, all adaptive traits in females (Bauer et al. 2006).
In contrast to Fer3HCH overexpression, which did not affect development, flies overexpressing Fer1HCH and Fer2LCH are at a developmental disadvantage compared to their siblings (Figure 3E). We scored the progeny from the cross Actin-Gal4/Cyo × UAS-Fer1HCH, UAS-Fer2LCH/UAS-Fer1HCH, UAS-Fer2LCH by gender and by presence or absence of the Actin-Gal4 driver. We used two independent sets of recombinant chromosomes (one on the X and one on the third) for UAS-Fer1HCH and UAS-Fer2LCH and also flies that carried both recombinant chromosomes, allowing for testing if the phenotype was dosage sensitive. We show the results in male flies (Figure 3E), where the competitive disadvantage of ferritin overexpression is more pronounced than in females. Importantly, the effects of ferritin overexpression are more dramatic under iron-limiting conditions induced by addition of the iron chelator Bathophenanthroline disulfate in the food, whereas the lethal effects can be rescued by dietary iron supplementation. Therefore, our results indicate that ubiquitous ferritin overexpression in the absence of iron overload can be deleterious, by a mechanism that implicates the iron-chelating properties of ferritin.
Ferritin transcripts have similar expression patterns during embryogenesis:
The genomic proximity between Fer1HCH and Fer2LCH could facilitate similar expression patterns to coordinate biosynthesis of the ferritin heteropolymer. In situ hybridizations of antisense RNA probes against the two genes in whole mount preparations of embryos at different stages of development were performed. We reasoned that if the predominant sites of Fer1HCH and Fer2LCH expression were different, it would be unlikely that they shared common regulatory enhancers, as previously suggested (Dunkov and Georgieva 1999, 2006). Nevertheless, we found that the different transcripts were expressed in highly specific, yet similar, patterns during embryogenesis (Figure 4). Fer1HCH and Fer2LCH were expressed during oogenesis and the maternal transcripts became evenly distributed in the egg and early embryo (data not shown), where they could be detected in the blastoderm (Figure 4, A and B). Tissue-specific transcripts were first detected during germ-band elongation in cells of the mesoderm (Figure 4, C and D), which are specified to give rise to the fat bodies and amnioserosa (Figure 4, E and F). Staining was also seen in cells destined to become macrophages in the anterior head region of embryos (Figure 4, C and D). During germ-band retraction and dorsal closure, the amnioserosa retained ferritin mRNAs (Figures 4, G and H). At late stages of embryogenesis, cells in the developing midgut initiated ferritin transcription (Figure 4, I and J). The similar embryonic expression patterns of Fer1HCH and Fer2LCH suggest that both ferritin subunits are expressed in each cell type where ferritin is required.
Ferritin expression and iron homeostasis in the larval midgut:
Ferritin is abundant in the midguts from several different insect species and its expression is induced by dietary iron (Capurro Mde et al. 1996; Dunkov et al. 2002; Georgieva et al. 2002; Kim et al. 2002). We showed earlier that GFP-tagged ferritin stores iron in Fer1HCHG188/+ flies. We next wanted to determine if the expression pattern of GFP-tagged ferritin was similar to that of the endogenous nontagged protein. Indeed, GFP-tagged ferritin was most prominently expressed in a cluster of cells of the middle midgut (Figure 5, A and B). These cells have been identified as the iron region of the insect midgut on the basis of their positive stain with Prussian blue and the accumulation of exogenously administered radioactive iron (Poulson and Bowen 1952). Mid-third instar larvae were administered a diet containing 5 mm ferric ammonium citrate and their intestines were dissected for imaging. GFP-tagged ferritin was inducible only in cells of the anterior midgut, but was constitutively expressed in the iron region of the middle midgut and no expression was detected in the copper cells that are present in between the two regions (Figure 5, A–D).
Midguts from larvae subjected to the same treatment were also stained with Prussian blue (Figure 5, E–H). A light blue staining indicative of the presence of ferric iron was observed in the iron region of both iron-fed and control larvae and was largely unchanged by feeding on an iron-enriched diet (Figure 5, F and H). In contrast, the anterior region was not stained in control larvae (Figure 5E), but stained dark blue in iron-fed individuals (Figure 5G), indicating that the ferritin that accumulates in the anterior midgut upon iron feeding is rich in iron content.
To assess the time frame of ferritin induction during iron feeding and to further demonstrate that a specific subset of cells in the anterior midgut shows a strong response to iron levels, we determined endogenous ferritin levels in a time course following iron feeding. For this experiment we used wild-type third instar larvae (100 hr old at 25°) and dissected their anterior and middle midguts for Western blot analysis. The results showed a clear induction of the ferritin heteropolymers 3 hr postfeeding in the anterior midgut (Figure 5K). Consistent with our imaging and iron-staining results, the induction of ferritin was largely restricted to the anterior midgut and was much less pronounced in the iron region (Figure 5L).
Another essential transition metal and dietary nutrient for Drosophila is copper (Zhou et al. 2003; Selvaraj et al. 2005). Copper-containing (cuprophilic) cells function in the acidification of the midgut and are present at different sites than the iron region (Poulson and Bowen 1952; Hoppler and Bienz 1994; Dubreuil et al. 2001). Copper-metallothionein complexes in these cells fluoresce orange–red upon ultraviolet illumination (Mcnulty et al. 2001). We imaged simultaneously the copper-metallothionein fluorescence and GFP-ferritin in midguts from larvae fed a diet containing 1 mm Cu2+. The results showed that ferritin-expressing cells indeed form a distinct cellular population from cells that contain copper (Figure 5, I and J; see also discussion).
Ferritin induction at subcellular resolution:
To investigate which cellular compartment accumulates iron-loaded ferritin in midgut cells, we costained GFP-Fer1HCH-expressing cells with an antibody raised against the Golgi-associated protein Lava lamp (Sisson et al. 2000). Confocal images from a single cell in the iron region clearly showed that all GFP-ferritin localized within the Golgi compartment (Figure 6, A–C), consistent with the images obtained by electron microscopy (Locke and Leung 1984; Nichol and Law 1990). A few Golgi bodies in cells from the iron region were devoid of ferritin (Figure 6C, red). We also imaged the same cells stained with antibodies against the Fer2LCH subunits. As expected from their tight association revealed by structural studies, biochemical analysis, and electron microscopy, there was complete colocalization with GFP-Fer1HCH in the Golgi complex of these specialized cells (Figure 6, D–F). Finally, we focused on the cells of the anterior midgut that do not normally express ferritin, but potently do so in the presence of high iron levels (see Figure 5, A–D). The cells were stained for both ferritin chains during the time course of ferritin induction (Figure 6, G–I). Prior to iron feeding, GFP-Fer1HCH was not expressed and low levels of Fer2LCH were detected (Figure 6G). At 1 hr postinduction GFP-Fer1HCH was detected in a compartment resembling the endoplasmic reticulum, but we could also detect Fer2LCH-positive Golgi that were devoid of GFP-Fer1HCH subunits (Figure 6H). In contrast, at 4 hr postinduction the two subunits were strongly induced and were seen only in complex with one another within the Golgi (Figure 6I).
Collectively, our results have identified different specialized intestinal sites for iron and copper metabolism and showed that ferritin synthesis is differentially regulated along the intestine. Significantly, our results also validate the use of the Fer1HCHG188 line as a faithful reporter of endogenous ferritin expression.
Several novel findings on Drosophila ferritin are described in this work. We show that the absence of either Fer1HCH or Fer2LCH results in embryonic lethality and that modified Fer1HCH subunits (mutant in the ferroxidase center or GFP tagged) cannot substitute for lack of Fer1HCH. However, if the same modified subunits are expressed in the presence of wild-type subunits, they can be integrated into ferritin holomers without inducing dominant-negative effects. Analysis of heterozygous loss-of-function ferritin fly mutants or flies overexpressing ferritin subunits revealed that a constant ratio of Fer1HCH and Fer2LCH is maintained, independent of their internal transcriptional expression levels. The structural cooperation of the two subunits that is secured via disulfide bonds (Hamburger et al. 2005) likely explains these observations. A post-transcriptional mechanism, possibly involving the degradation of subunits that are present in excess, ensures the presence of equal amounts of the two subunits. Such a mechanism can explain the absence of the IRE in Fer2LCH mRNAs in insects, since Fer1HCH translational repression by IRP-1A under iron-limiting conditions (that favor the IRE-containing transcripts; Georgieva et al. 1999) would then be sufficient to reduce levels of Fer2LCH.
In contrast to results from mammalian cell or animal models (Thompson et al. 2003; Wilkinson et al. 2006), but consistent with what is known from human patients with hereditary hyperferritinemia cataract syndrome (Cazzola 2002), experimental reduction or increase of ferritin levels through genetic manipulation in Drosophila caused only very mild alterations in the insect's iron homeostasis. These results point toward an independent regulatory system that controls iron sequestration into ferritin. The nature of this system is currently unknown, but could involve a putative iron chaperone that delivers iron to ferritin (Napier et al. 2005). However, the hypothesized chaperone's function does not completely override the need for ferritin regulation, as shown by the phenotype of ferritin overexpressing flies that was lethal under low iron conditions, but was rescued with iron supplementation (Figure 3E). Alternatively, localization of ferritin in the Golgi apparatus of insect cells may prevent it from contact with the cytosolic and mitochondrial iron pools. It is currently not known how iron is delivered to the ferritin that resides in the secretory pathway of cells.
We provided evidence that the ferritin genes are coexpressed during embryogenesis. We wondered whether their genetic proximity is conserved in other Drosophila species with fully sequenced genomes. To this end, we used the EvoPrinter, a new multigenomic DNA sequence analysis tool that facilitates the rapid identification of evolutionarily conserved sequences within the context of a single species (Odenwald et al. 2005). In silico analysis of the Fer1HCH and Fer2LCH genomic locus by the EvoPrinter produced an output of the combined mutational histories of six Drosophila species, superimposed on a reference sequence from D. melanogaster (supplemental Figure 1 at http://www.genetics/supplemental/). The results not only showed that clustering of the two genes is conserved, but also identified potentially shared regulatory regions. From the 17,670 bp spanning the two genes that were analyzed, only 914 (or 5.2%) were conserved in an identical position in all six species tested. One-third of these conserved sequences (306) were contained within the open reading frames. Surprisingly, over half of them (491) were not contained in the cDNA sequence, but rather were clustered in two regions: downstream of the Fer2LCH-transcribed region or in the second intron of Fer1HCH (supplemental Figure 1). A particular stretch of 10 bp, TTTGCACACG, was found three times in the second intron of Fer1HCH and could represent a binding site for an unidentified factor. The remaining conserved base pairs (117) were mostly concentrated in the 5′-UTR of Fer1HCH, including the IRE itself. The conservation of the IRE over ∼160 million years of collective evolutionary divergence underscores the functional significance of the IRE/IRP control system in all Drosophila species.
Ferritin regulation is a critical aspect of the organism's iron economy. Our finding that ferritin expression in the iron region is constitutive (and remains so even under iron-deficient conditions), but is inducible in the anterior midgut, has a parallel with the expression of metallothioneins, which are the copper storage proteins (Egli et al. 2006). Interestingly, cuprophilic cells that are present anterior to the iron region show constitutive metallothionein expression (Mcnulty et al. 2001), whereas cells in both the anterior and posterior midgut induce metallothionein expression and copper/metallothionein fluorescence at higher copper concentrations (Poulson and Bowen 1952; Mcnulty et al. 2001) (Figure 5J). Results presented in this article, combined with previous reports, lead to the following fundamental conclusion with respect to metal metabolism in the midgut of Drosophila: the insect midgut contains two sets of specialized cells, one of which constitutively expresses metallothionein and a second constitutively expresses ferritin. If either metal is present in great abundance, a third and fourth set of distinct midgut cells have the potential to induce transcription of the genes that encode the two metal storage proteins. Importantly, there are also cells in the midgut that do not respond significantly to high concentrations of these metals. Whether similar cellular populations exist in the mammalian intestine remains unknown, but cellular populations with specific metal contents were recently described in plant seeds (Kim et al. 2006).
The GFP-tagged ferritin that we have described in Drosophila revealed the complex physiologic orchestration of intestinal metal responses. In addition, it has enabled visualization of subcellular ferritin dynamics upon iron-mediated induction and should facilitate the dissection and subcellular localization of ferritin biosynthesis and trafficking. Serum ferritin is measured in clinical practice as a measure of total-body iron stores (Beutler et al. 2002) and is an acute phase reactant (Tran et al. 1997) but few studies have addressed how ferritin is secreted in the circulatory system of humans (Ghosh et al. 2004; Renaud et al. 1991). Importantly, ferritin tagged with GFP could function and elucidate trafficking mechanisms in human cells (De Domenico et al. 2006) and transgenic mice (Cohen et al. 2007).
The authors thank Foteini Mourkioti and Herbert Jäckle for help with the generation of the UAS-Fer1HCH and UAS-Fer2LCH transgenic flies and Genetic Services for the UAS-Fer1HCH* transgenic flies. We especially thank Mary Lilly for offering lab space and fly-pushing support to members of the Cell Biology and Metabolism Branch of the National Institute of Child Health and Human Development (NICHD). We thank Wing-Hang Tong and Kuanyu Li for intellectual inputs. The intramural programs of the NICHD and the National Institute of Neurological Disorders and Stroke supported this work.
Note added in proof: The conserved stretch of 10 bp, TTTGCACACG, identified here by EvoPrinter analysis of the Fer1HCH and Fer2LCH genomic locus has been independently described by others (H. Yepiskoposyan, D. Egli, T. Fergestad, A. Selvaraj, C. Treiber et al., 2006, Transcriptome response to heavy metal stress in Drosophila reveals a new zinc transporter that confers resistance to zinc. Nucleic Acids Res. 34: 4866–4977). These authors have previously characterized the binding of metal transcription factor 1 to this sequence.
Communicating editor: T. Schüpbach
- Received April 28, 2007.
- Accepted June 22, 2007.
- Copyright © 2007 by the Genetics Society of America