Subcellular localization of iron regulatory proteins to Golgi and ER membranes

Iron is a crucial nutrient for all cells but excess iron can result in cellular damage through the production of free radicals via the Fenton reaction (Gutteridge and Halliwell, 1989). In addition, iron deficiency in rodents causes dramatic behavioral and motor effects (Beard and Connor, 2003; Glover and Jacobs, 1972; Hunt et al., 1994; Youdim et al., 1989). Consequently, it is essential for cells to maintain intracellular iron homeostasis. The model for cellular iron homeostasis has become a classic example of post-transcriptional regulation of protein synthesis. In this model, mRNA of critical proteins involved in the movement, storage and utilization of iron contain iron response elements (IREs). These are specific sequences that, by interacting with cytoplasmic iron regulatory proteins (IRP1 and IRP2), regulate the fate of the mRNAs (Crichton et al., 2002). Also, iron levels in the cell regulate IRP1 and IRP2 via different mechanisms. IRP1 activity changes from IRE binding to cytosolic aconitase with increasing levels of iron, while IRP2 is degraded through the proteosomal pathway (Guo et al., 1994; Samaniego et al., 1994). Control of the interaction between the IREs and the IRPs has been studied under different conditions and has been described in detail elsewhere (Drapier et al., 1993; Eisenstein, 2000; Hanson and Leibold, 1998; Pantopoulos and Hentze, 1995; Schalinske and Eisenstein, 1996).

Functionally, cytoplasmic IRPs exist in the cell in three major pools: (1) bound to an IRE, (2) free and available to bind to an IRE and (3) free but unavailable to bind IRE (for example, as part of the cytoplasmic aconitase activity pool). Variations in the levels of intracellular iron produce a change in the IRP/IRE binding activity and, as a consequence, a change in the levels of the proteins involved in iron homeostasis, such as ferritin and transferrin receptor (Eisenstein, 2000). We have previously reported that IRPs are associated with an intracellular membrane fraction (Pinero et al., 2001). In this study, we explore in greater detail the intracellular distribution of IRPs using both a microscopic and biochemical approach. We identify the ER and Golgi membranes as the sources of the IRP activity in the microsomal membrane fraction and reveal that only IRP1 can be detected. Finally, the localization of IRP1 to the membranes involves phosphorylation and is responsive to cellular iron status and agents of stress.

In order to visualize the distribution of IRPs in cells, we used immunocytochemistry and epitope tagging (c-myc). The microscopic analysis was performed on the following cell lines: NIH 3T3 and SW1088. Both cell lines were purchased from American Type Culture Collection (ATCC; Rockville, MD, USA).

The initial immunocytochemical studies were performed using antibodies against IRP1 and IRP2 that were provided by Betty Leibold (University of Utah, USA). The epitope-tagged antibody (c-myc) was purchased from Ambion.

For co-localization studies, ∼5×10⁴ cells were plated on poly-D-lysine-treated glass slides. Cells were allowed to attach for 1 hour at 37°C under a humid 5% CO₂ atmosphere, then the slides were submerged in either Dulbecco's modified Eagle's medium (DMEM, Gibco, Rockville, MD, USA), DMEM supplemented with ferric ammonium citrate (100 mg/l), or DMEM supplemented with desferrioxamine (DFO) (100 μM) and cultured at 37°C under a humid 5% CO₂ atmosphere. After 72 hours, the slides were removed, washed once with Hank's buffered salts solution and twice with PBS and fixed with 4% paraformaldehyde for 30 minutes. Slides were blocked as described above and then incubated with primary anti-rat IRP1 (Alpha Diagnostics, International; San Antonio, TX, USA), the ER marker anti-human calnexin (AF18; Abcam, Cambridge, MA, USA), or the Golgi marker anti-human 58K Golgi protein (58K-9; Abcam; in 5% normal goat serum) for 90 minutes at room temperature, as per the manufacturer's protocols. Free antibodies were removed by three washes with PBS and slides were then incubated with Alexa Fluor 555-conjugated goat anti-mouse (organelle markers) and Alexa Fluor 488-conjugated goat anti-rabbit secondary antibody (for IRP1) and DAPI for a further 60 minutes. The slides were washed, mounted and covered with a coverslip. Negative controls for antibody staining were prepared as described above, except that the primary antibody was omitted. Confocal microscopy was performed using a Leica TCS confocal laser scanning unit equipped with a DMR inverted microscope and a 63/1.4 objective. An argon/krypton laser was used to generate light at 488 and 555 nm for Alexa excitation and 360 nm for DAPI excitation. A high-pass filter with wavelength cut-off at 500 nm was used to recover Alexa fluorescence and a low-pass filter with wavelength cutoff at 460 nm was used to detect DAPI fluorescence. Each fluorophore was excited sequentially to insure the lowest interference between the detector channels.

The protocol for isolation of ER-rich fractions was adapted from the methods of Zhang et al. (Zhang et al., 1998). Briefly, twenty 150 cm² culture flasks were washed in Hank's buffered salts and once in the homogenization medium (0.25 M sucrose, 1 mM EDTA, 10 mM Hepes-NaOH, pH 7.4 with protease inhibitor cocktail). The cells were suspended in homogenization medium and harvested by scraping. The cells were centrifuged at 2.5 g for 3 minutes, resuspended in homogenization medium that contained protease inhibitors and then disrupted by Dounce homogenization. The homogenate was centrifuged at 3000 g for 10 minutes to remove cell debris and the resulting supernatant was centrifuged again at 100,000 g for 40 minutes. The pellet was washed with homogenizing medium and resuspended in the same medium. A 12 ml linear Optiprep gradient (0-20%) was prepared and the vesicle solution was layered on top of the gradient and this was centrifuged at 200,000 g for 16 hours; 1 ml fractions were collected. NADPH-cytochrome C reductase and lactate dehydrogenase assays were performed on all fractions to determine ER and cytosolic activity, respectively (Graham, 1993).

To demonstrate the distribution of IRPs in SW1088 microsomal fractions, 6% SDS-PAGE was used; 5 μg of total protein was loaded for IRP1 analysis and 2.5 μg of total protein was loaded for IRP2 analysis. The proteins were transferred to PVDF membranes and incubated with 5% powdered milk in Tris-buffered saline-Tween 20 overnight at 4°C. The primary antibody was diluted 1:1000 and incubated with the membrane for 1 hour. Following washes in Tris-buffered saline-Tween 20 (TBS-Tween), the membrane was incubated in secondary antibody that had been diluted 1:5000. The IRP1 primary antibody was the same as used for immunohistochemistry. The IRP2 primary antibody (clone 4G11) was a gift from Wolff Kirsch (Loma Linda University, Loma Linda, CA, USA). The secondary antibody used for IRP1 was a rabbit anti-chicken IgG conjugated to horseradish peroxidase. The secondary antibody used for IRP2 was rabbit anti-mouse IgG conjugated to horseradish peroxidase. The distribution of protein on the membrane was visualized using a KPL chemiluminescence kit.

Sprague-Dawley rats were fed either control or iron-deficient diets prepared as described previously (Pinero et al., 2000) for 6 weeks, after which the animals were sacrificed, and the livers were removed, weighed and minced. Homogenization medium (0.25 M sucrose, 1 mM EDTA, 10 mM Hepes-NaOH, pH 7.4) was added and the livers were Dounce homogenized. The homogenates were centrifuged at 4000 g for 5 minutes. The microsomal pellets were obtained by centrifugation at 140,000 g for 1 hour. The microsomal pellets were resuspended in 1 ml of homogenization medium and each was top-loaded onto a 2.5%-30% pre-formed discontinuous Optiprep gradient. The gradients were centrifuged at 200,000 g for 2.5 hours and 1 ml fractions were isolated by tube puncture.

To demonstrate the distribution of IRPs in rat liver microsomal fractions, 4%-20% gradient SDS-PAGE was used; 10 μg of total protein was loaded for IRP1 analysis and 5 μg of total protein was loaded for ER and Golgi analysis. The proteins were transferred to PVDF membrane and incubated with 5% powdered milk in TBS-Tween overnight at 4°C. The following primary antibodies were used (anti-IRP1, Alpha Diagnostics, International; San Antonio, TX; and anti-calnexin and anti-Golgi 58K protein, Abcam). Both anti-IRP1 and anti-calnexin antibodies were diluted 1:1000 and anti-Golgi 58K protein was diluted 1:5000 and incubated with the membrane for 1 hour. Following washes in TBS-Tween, the membrane was incubated in secondary antibody that had been diluted 1:5000. The secondary antibody used for IRP1 and calnexin was a goat anti-rabbit IgG conjugated to horseradish peroxidase and the secondary antibody used for Golgi 58K protein was rabbit anti-mouse IgG conjugated to horseradish peroxidase. The distribution of protein on the membrane was visualized using a KPL chemiluminescence kit. All analyses were quantitated using the Fuji LAS-3000 system.

The following sets of experiments were performed to evaluate the intracellular distribution of IRP binding activity after activation of IRPs. Human astrocytoma cells (SW1088) and human embryonic kidney cells (HEK293) were purchased from ATCC. The cells were grown in a 5% CO₂ atmosphere at 37°C in DMEM supplemented with 10% fetal bovine serum (Biocell, Rancho Dominguez, CA, USA), 4 mmol/l glutamine (Sigma, St Louis, MO, USA), 1000 U/ml penicillin G and 100 μg/ml streptomycin sulfate and 250 ng/ml amphotericin B. All the experiments were performed when the cells reached 70-80% confluence. The cells were treated with 0.2 μM phorbol 12-myristate 13-acetate (PMA; Biomol, Plymouth Meeting, PA, USA) (for 16 hours), a phorbol ester shown to induce IRP activity in HL-60 cells (Schalinske and Eisenstein, 1996), interleukin-1 (Sigma) (for 16 hours), hydrogen peroxide (for 30 minutes), chelerythrine (Biomol; 2.0 μM, for 16 hours) an inhibitor of protein kinase C, as well as the intracellular calcium chelator BAPTA-AM (1,2-bis-(o-aminophenoxy)-ethane-N,N,-N′N′-tetraacetic acid tetraacetoxy-methyl ester; Biomol) (for 30 minutes). Untreated cells, as well as cells treated with DMSO at the same concentration used for the cells treated with PMA or BAPTA-AM, served as controls. Additionally, cells were also treated for 16 hours with either ferric ammonium citrate (FAC; 100 mg/l; Sigma) or the iron chelator deferoxamine (DFO, 100 μmol/l; Sigma) to demonstrate their responsiveness to iron addition or chelation. When the cells were treated with BAPTA-AM, the medium was changed after 30 minutes and replaced with either control medium or FAC- or DFO-containing medium for the next 16 hours. The results were consistent for all cell types used.

After completion of the experiments, the cells were washed with Hank's solution (Sigma), treated with trypsin until detachment, and collected. To separate membrane-bound organelles from cytosol, the cells were washed and homogenized in lysis buffer (10 mm/l Hepes pH 7.4, 40 mmol/l KCl, 5% glycerol, 3 mmol/l MgCl₂, 0.1 mmol/l EDTA) containing a protease inhibitor cocktail (Sigma) and Prime RNase inhibitor (Eppendorf Scientific, Westbury, NY, USA). Homogenization was carried out using a sonicator with a microtip (Branson Sonifier 250, Branson Inc., Danbury, CT, USA), on pulse mode. The homogenates were centrifuged at 4,000 g for 5 minutes at 4°C to sediment the nuclei. The resulting supernatant was centrifuged at 145,000 g for 1 hour at 4°C and the cytosolic fraction was collected. The pellet was washed twice in 1 ml lysis buffer, resuspended in 100 μl of lysis buffer and either processed or stored at -70°C. 50 μl of this solution were treated with Triton X-100 (final concentration 0.5%) at 4°C for 1 hour, centrifuged at 20,000 g for 5 minutes at 4°C, and the supernatant collected. This supernatant will be referred to here as membrane fraction.

A synthetic RNA transcript containing the IRE was generated from the oligonucleotide template 5′-ttatgctgagtgatatccctctcctaggacgaagttgtcacgaacctgcctagg-3′ with T7 polymerase in the presence of [³²P]CTP. Binding reactions were carried out as described previously (Leibold and Munro, 1988). Briefly, equal amounts of protein for the cytoplasmic and membrane fractions were incubated with the synthetic radiolabeled IRE probe. To measure total IRP binding activity, the samples were first treated with 2% β-mercaptoethanol (BME). The RNA-protein complexes were separated on a 6% native polyacrylamide gel and visualized by autoradiography. The autoradiograms were scanned and the resulting digital image subjected to densitometric band analysis using the Labworks Analysis Software (UVP Inc., Upland, CA, USA). The results of the RNA band shift assays were expressed as active IRP binding from each treatment as a ratio of control.

Neither IRP1 nor IRP2 was detected in the nucleus and both were predominantly located in the perikaryal cytoplasm. IRP2 extended from the perikaryal cytoplasm into cellular processes more frequently than IRP1. IRP2 immunoreaction product also had a more punctate appearance than IRP1. A similar observation was made with the c-myc tagged IRP proteins (Fig. 1C,D).

In order to further explore the perikaryal location of IRP1, confocal microscopy was used to analyze SW1088 cells that had been stained for IRP1, ER and Golgi distribution. Fig. 2 also demonstrates a perikaryal cytoplasmic distribution of IRP1 as well as the co-localization with anti-calnexin antibody. Fig. 3 demonstrates the co-localization of IRP1 with the Golgi antibody anti-58K Golgi protein. Fig. 4 demonstrates that localization of IRP1 to the ER in microscopic studies is unaffected by changes in iron status.

On the basis of the microscopy results, a crude microsomal SW1088 cell fraction was obtained and sub-fractionated using density centrifugation to further assess the location of IRPs. IRP1, but not IRP2, was detected in fractions containing ER membrane (based on NADPH cytochrome C reductase assay), however, it was apparent that the distribution of IRP1 was biphasic (Fig. 5). Cytosolic contamination of the ER fraction was determined by lactate dehydrogenase assay to be less than 2%.

To further study the biphasic distribution of IRP1 that was observed in the SW1088 fractionation experiments in a model system that would provide greater sample size we isolated liver membrane fractions by density gradient analysis from rats fed control and iron-deficient diets. As shown in Fig. 6 immunoblot analysis revealed that IRP1 did not fractionate with ER in control rat livers, however, a relatively small amount of IRP1 was detected in the Golgi-containing fractions. The largest amount of IRP1 was found in those fractions that were less dense than the Golgi. These less dense fractions were determined by LDH activity to have higher cytosolic content, indicating that under control conditions, the majority of IRP1 remains in the cytosolic fraction. As shown in Fig. 7, immunoblot analysis revealed more IRP1 in fractions containing the Golgi marker in iron-depleted rat livers than in livers from rats on control diets. Additionally, a small fraction of IRP1 was found in the ER-containing fractions. IRP2 was absent in the membrane fractions from both the control and iron-deficient rat livers, but was present in the cytosolic fractions of both (data not shown).

Two cell types were chosen for this study, HEK293 cells and SW1088 to demonstrate that the membrane-associated IRPs were not unique to a single cell type. Fig. 8 shows the changes in IRE-binding activity for both the cytosolic fraction and the membrane fractions after treatment with ferric ammonium citrate (FAC), DFO, H₂O₂ or IL-1β in HEK293 cells. Treatment with FAC produced a decrease in the IRE binding activity, as did treatment with IL-1β. DFO and H₂O₂ treatment resulted in an increase in IRE-binding activity. The treatment with iron or H₂O₂ produced a larger response in the membrane than in the cytosolic fraction, while the response to treatment with IL-1β was greater in the cytosolic fraction.

A similar IRE binding profile to that observed in HEK293 cells was seen in SW1088 cells. However, following iron manipulation, the responses were amplified in SW1088 cells (Figs 8 and 10). Therefore SW1088 cells were chosen for further investigation into the possible mechanisms involved in the differential localization of IRPs in the two subcellular fractions.

Because IRPs reportedly have sites for phosphorylation (Schalinske et al., 1997), SW1088 cells were treated with PMA (an activator of protein kinase C) or chelerythrine (an inhibitor of protein kinase C) to determine the role of phosphorylation on IRP activity in the membrane fraction. Treatment with PMA resulted in a decrease in IRE binding while treatment with chelerythrine resulted in an increase in IRE binding (Fig. 9). Because protein kinase C is a calcium-dependent protein, we examined the role of calcium as a regulator of IRP activity in the membrane fraction using the intracellular calcium chelator BAPTA-AM. When the SW1088 cells were treated with BAPTA-AM, there was an increase in the IRE-binding activity in both the cytosolic and the membrane fraction similar to the one produced by the treatment with DFO. However, when the cells were treated with iron or DFO after treatment with BAPTA-AM, the IRE binding activity was similar to the ones that received iron or DFO without BAPTA-AM (Fig. 10).

Our laboratory previously reported that IRE-binding activity in NIH3T3 cells and brain was present in both membrane (microsomal) and cytosolic fractions. Additionally, an in vitro analysis suggested that iron chelation resulted in an increase in the IRP binding activity in both the cytosolic and microsomal fractions and an increase in the amount of IRP1 in the microsomal fraction (Pinero et al., 2001). The purpose of the current study was to more specifically identify the microsomal fraction and begin to elucidate the intracellular events associated with the putative redistribution of IRP1 between the cytosolic and membrane compartments. In this study we used multiple microscopic and biochemical approaches to demonstrate the subcellular compartmentalization of IRPs. The microscopic studies reveal that both IRP1 and IRP2 are localized in the cytosol with IRP1 having a more perinuclear distribution than IRP2. Immunocytochemistry was performed on cells in culture as an initial attempt to determine the subcellular localization of IRP1. Under control conditions, IRP1 was co-localized with both the ER and Golgi. Iron loading or iron depletion of the cultured cells did not result in observable changes in the amount of IRP1 that co-localized with the ER or Golgi markers. In our previous study in which we examined the subcellular distribution of IRP in both rat brains and various cell lines, under varying iron conditions, we determined that about 10% of the IRP/IRE binding activity is membrane associated (Pinero et al., 2001). Changes in this relatively small population of IRP1 in the membrane compartment may be below the detectable level microscopically.

To further explore this apparent association between IRP1, ER and Golgi, a biochemical approach was utilized. The microsomal membrane fraction from the SW1088 cells was isolated and further separated into ER-containing fractions, which revealed that IRP1 co-localized with the ER-activity marker NADPH-cytochrome C reductase. This finding is consistent with the microscopical analysis and confirmed a novel subcellular location for IRP1. The IRP1 distribution appeared biphasic, which suggested that IRP1 might be associated with other membrane-bound organelles in addition to the ER, supporting the microscopical evidence that IRP1 is also associated with the Golgi. The presence of IRP2 in the ER fraction in SW1088 cells could not be detected with either immunocytochemical or biochemical methods.

To obtain sufficient material to perform additional subcellular fractionation analyses and to demonstrate that the subcompartmentalization of IRP1 is not unique to cells in culture we examined livers from rats fed control or iron-deficient diets (Han et al., 2003). This model system also enabled us to determine if iron status affected the intracellular distribution of IRP1, and allowed us to postulate that the IRP1 distribution may affect mRNA stabilization. In the livers of the rats fed a control diet, IRP1 is predominantly associated with the cytosolic fraction, as is well established (Pinero et al., 2001), but is also clearly present in the Golgi apparatus. This novel finding is consistent with the microscopical analysis of SW1088 cells. Unlike the microscopical analysis of IRP1 distribution in the cell culture studies, no IRP1 was found in the ER fraction of the control rat livers. However, the ER fraction from iron-deficient rats did contain a relatively small amount of IRP1 in addition to IRP1 in the Golgi membrane fraction.

The association of IRP1 with intracellular membranes is a novel finding. The function of IRP1 is well established. We are not proposing that there is a new function for the IRPs in these subcellular compartments, but rather a novel concept in their processing and how they function. For example, the Golgi apparatus could be a site where IRP1 obtains an iron-sulfur cluster. The iron-sulfur cluster assembly proteins, IscU1 and IscS, are reportedly localized in the cytoplasm suggesting that the machinery for Fe-S cluster assembly may also be present in the cytosol (Tong and Rouault, 2000). Roy et al. have identified CFD1, a highly conserved P-loop ATPase that is critical for Fe-S cluster biogenesis and have reported that the subcellular distribution of CFD1 is 90% cytosolic, and 10% membrane (Roy et al., 2003). The subcellular localization of CFD1 mirrors the IRP1 subcellular localization that was reported here. It has been proposed that there is an intracellular iron trafficking system that is comparable to the copper trafficking system (Tong and Rouault, 2000). Expounding on this proposition, IRP1 may obtain a metal cluster in the Golgi apparatus similar to the process identified for ceruloplasmin. Copper is transported to the Golgi by the Cu-binding chaperone protein, HAH1 and incorporated into ceruloplasmin via a copper-transporting ATPase (Cox and Moore, 2002).

Because of the novel location of the IRPs in the membrane fraction, we performed gel-shift assays to determine if the activity of the IRPs in this fraction were activated similarly to those in the cytosol. We performed IRE-binding assays following iron chelation, iron supplementation, treatment with interleukin-1β and treatment with hydrogen peroxide, and each of the IRP/IRE interactions were similar to those reported for other cell types in the cytosol (Pantopoulos and Hentze, 1995). HEK293 cells and SW1088 cells, however, have a different level of response to manipulation of iron status. Therefore most of the cell culture experiments were performed in the SW1088 cell line.

Eisenstein and co-workers have previously reported that phosphorylation may play a role in IRP/IRE interaction (Eisenstein et al., 1993; Schalinske and Eisenstein, 1996). Activation of protein kinase C resulted in increased IRP/IRE interaction in HL-60 cells (Schalinske and Eisenstein, 1996). In the SW1088 cells, protein kinase C activation decreased IRP/IRE interaction in the cytosolic and membrane fractions. Inhibition of protein kinase C with chelethyrine resulted in an increase in IRP/IRE binding in the cytosolic fraction. These results demonstrate consistency within the two fractions of the SW1088 cells and suggest that the role of protein kinase C in IRP/IRE interaction could be cell-specific.

IRPs in HL-60 cells are activated by phosphorylation via protein kinase C, which is Ca²⁺-dependent (Schalinske and Eisenstein, 1996). We examined the potential role of calcium in IRP activation in the membrane fraction using the calcium chelator, BAPTA-AM. SW1088 cells that were treated with BAPTA-AM alone had an increase in IRP/IRE interaction in both the cytosolic and membrane fractions, and an increase in the amount of IRP binding in the microsomal fraction. The increase in IRP/IRE interaction was similar to that observed in SW1088 cells treated with DFO alone. The finding that concomitant addition of BAPTA-AM with DFO to SW1088 cells did not result in an increase in IRP/IRE interaction beyond that observed for either treatment alone suggests that both chelators individually elicit a maximum effect on IRPs available to bind IREs. FAC treatment of SW1088 cells concomitant with BAPTA-AM provided results similar to those observed with FAC treatment alone, suggesting that FAC treatment is able to overcome the influence of calcium chelation in IRP/IRE interaction.

In a number of IRP binding assays, we noted that there was a difference in the relative amounts of binding activity in the cytosol or membrane fraction following treatment. This difference in amount of total IRP binding activity following treatment was first observed in our earlier study on NIH3T3 cells in the membrane fraction following treatment with an iron chelator (Pinero et al., 2001) and was notable in this study following treatment with CET, hydrogen peroxide, DFO and BAPTA. These data raise the possibility that IRPs may translocate between the cytosolic and membrane fractions unbound to IREs. The evidence that the IRPs in the ER and Golgi membranes may not be simply a co-migration with IREs on the mRNAs that were bound in the cytosol is that IRP1 activity could be detected with the addition of exogenous probe indicating that unbound but available IRPs are present in the membrane fraction. Treatment of the homogenates with 2% BME increases the amount of IRP activity, further indicating a resident population of IRP in the membrane fraction.

The aforementioned findings do not preclude the possibility that some IRP1 could co-migrate from the cytosol to the membrane fraction while bound to IREs. Indeed, as shown in Fig. 11, we propose that IRP1 initially binds to the Tf receptor mRNA in the cytosol to stabilize the mRNA, and then under iron-deficient conditions, the additional IREs are bound when the mRNA reaches the ER, offering additional stabilization once the mRNA reaches its site of translation. Tf receptor mRNA was chosen for this model because this mRNA is synthesized on the rough ER. Furthermore, the mechanism of additional stabilization for the Tf receptor mRNA from IRP1 in the ER membrane is particularly appealing because only two of the five IREs in the Tf receptor mRNA must be bound to promote mRNA stabilization (Erlitzki et al., 2002). Once the translation of the IRE-containing mRNA has occurred, we propose that the IRP1 remains attached to the ER membrane and cycles to the cis-Golgi membrane where it is released into the cytosol. Insertion of an iron-sulfur cluster may not be a requirement for release of IRP1 from the Golgi membrane into the cytosol. It is also possible that the IRP1 in the Golgi membrane could be retrieved back to the ER membrane particularly in situations when intracellular iron status is low.

A major question in our proposal is the mechanism by which IRP1 would become integrally associated with the ER and Golgi membrane. An analysis of the protein sequence of IRP1 using the TMHMM (Tied Mixture Hidden Markov Model) Server 2.0 revealed a putative hydrophobic tail in human IRP1 that is not present in human IRP2, which may facilitate insertion of IRP1 into a membrane (Krogh et al., 2001). This observation not only supports our hypothesis that IRP1 is found in association with intracellular membranes but also supports our inability to detect IRP2 in these fractions. Tail-anchored proteins are a group of integral membrane proteins that are retained in a phospholipid bilayer via a single C-terminal hydrophobic domain (Borgese et al., 2003). These proteins are unlike other integral membrane proteins because they localize to intracellular membranes post-translationally. The protein sequences of tail-anchored proteins do not contain a signal recognition site, which would direct the partially synthesized protein to the ER. The N-terminus of tail-anchored proteins, such as cytochrome b(5) and heme oxygenase, are synthesized and folded on free polysomes in the cytosol prior to the insertion of the hydrophobic domain into a membrane (D'Arrigo et al., 1993). Because IRP1 is a cytosolic protein it is also synthesized on free polysomes. Wattenberg and Lithgow (Wattenberg and Lithgow, 2001) propose that tail anchoring of proteins serve the purpose of localizing certain enzyme activities close to the surface of various intracellular membranes, which assists in the enzymes' catalytic activity. This is consistent with the novel concept that we propose, namely, that IRP1 can insert in the intracellular membrane to facilitate stabilization of IRE containing mRNAs. It is noteworthy for the new concepts for IRP1 in this paper that cytochrome b(5) and heme oxygenase require insertion of a metal complex for activity similar to IRP1 and all have a hydrophobic tail that allows transient insertion into the Golgi membrane.

In summary, we have shown that there are distinct subcellular populations of IRPs. These subcellular populations include, but are not limited to, the cytosol, the ER and the Golgi membrane. Additionally, we have determined that IRP1, but not IRP2, can be localized to Golgi apparatus and ER membrane containing fractions. This latter observation has resulted in the presentation of a novel concept for the activity of IRP1 in cells that further distinguishes IRP1 from IRP2.

This research was supported by a grant from the National Institutes of Health: National Institute of Diabetes and Digestive and Kidney Diseases (1 P01 DK53430-04) and Tobacco Settlement Funds from the State of Pennsylvania.