HFE Modulates Transferrin Receptor 2 Levels in Hepatoma Cells via Interactions That Differ from Transferrin Receptor 1-HFE Interactions*

Mutations in the transmembrane glycoproteins transferrin receptor 2 (TfR2) and HFE are associated with hereditary hemochromatosis. Interactions between HFE and transferrin receptor 1 (TfR1) have been mapped to the α1- and α2-helices in HFE and to the helical domain of TfR1. Recently, TfR2 was also reported to interact with HFE in transfected mammalian cells. To test whether similar HFE residues are important for both TfR1 and TfR2 binding, a mutant form of HFE (W81AHFE) that has a ∼5,000-fold lower affinity for TfR1 than HFE was employed. As expected, W81AHFE does not interact with TfR1. However, we found that the same mutation in HFE does not affect the TfR2/HFE interaction. This finding indicates that the TfR2/HFE and TfR1/HFE interactions are distinct. We further observed that, unlike TfR1/HFE, Tf does not compete with HFE for binding to TfR2 and that binding is independent of pH (pH 6–7.5). TfR2-TfR1 and HFE-HLA-B7 chimeras were generated to map the domains of the TfR2/HFE interaction. TfR1 and HLA-B7 were chosen because of their similar overall structures with TfR2 and HFE, respectively. We mapped the interacting domains to the putative stalk and protease-like domains of TfR2 located between residues 104 and 250 and to the α3 domain of HFE, both of which differ from the TfR1/HFE interacting domains. Furthermore, we found that HFE increases TfR2 levels in hepatic cells independent of holo-Tf.

Hereditary hemochromatosis (HH) 3 is a prevalent inherited iron metabolism disorder characterized by excessive iron dep-osition in the liver, heart, pancreas, and parathyroid and pituitary glands. The excess iron deposited in these organs is toxic, which leads to multi-organ dysfunction (1,2). HH type 1, an autosomal recessive form of the disease, is caused by mutations in the HFE gene (3,4), which encodes an atypical MHC class I protein. The majority of HH type 1 patients are homozygous for C282Y 4 substitution, which has a prevalence of ϳ1 per 400 individuals of northern European descent (5). Interestingly, patients with this same mutation present very different phenotypes, ranging from mild to severe (6). The different presentations of the same mutation suggest that HFE function depends on the presence of modifiers (7)(8)(9)(10), acting directly or indirectly with HFE.
The first modifier was reported to be transferrin receptor 1 (TfR1). In vitro (11) and in vivo (12) studies showed that, at slightly alkaline pH 7.5, HFE interacts with the ecto-domain of TfR1. Interestingly, both HFE and diferric-transferrin (holo-Tf) recognize overlapping regions on TfR1 (13,14), which results in competition between HFE and holo-Tf for binding to TfR1 (15,16). At concentrations of holo-Tf (100 nM) well below the physiological level of holo-Tf in the blood (ϳ10 M), no binding of HFE to TfR1 can be detected (17). Expression of wild type HFE decreases intracellular iron levels in both the cells expressing endogenous TfR1 (16,18,19) and the cells lacking endogenous TfR1 (20). Moreover, a mutated form of HFE (W81AHFE), which has ϳ5,000-fold lower affinity to TfR1 than the unmutated HFE (14), also down-regulates Tf-mediated iron uptake to the same extent as wild-type HFE in HeLa cells (17). These studies point out that HFE can lower intracellular iron levels independently of TfR1, and lead to the conclusion that TfR1 is not the only modulator of HFE function in regulation of iron homeostasis at the cellular level (17). Thus HFE likely regulates iron homeostasis by interacting with proteins other than TfR1.
The predominant site of HFE expression in the liver is hepatocytes (21). Transferrin receptor 2 (TfR2), a homolog of TfR1, is also primarily expressed in hepatocytes (22). TfR2 and TfR1 show 45% sequence identity in their ecto-and transmembrane domains. They both bind to Tf but with different affinities (23,24). Unlike TfR1, which is inversely regulated at the level of mRNA stability by an intracellular iron pool (25), TfR2 is regulated at the level of protein stability by a novel mechanism, involving the stabilization of receptor upon ligand binding (26 -28). Moreover, while TfR1 knock-out mice die in utero with defective erythropoiesis and neurological abnormalities (29), mutations in TfR2 result in an HH type 3, a rare form of iron overload (30). These differences between TfR2 and TfR1 indicate that the two proteins play different roles in iron metabolism.
HH type 1 and type 3 patients show very similar patterns of iron overload. Both groups have increased serum transferrin saturation, serum ferritin levels, and iron absorption at the duodenal level (30,31). This increased iron absorption leads to parenchymal iron overload. The similarities in patients with HH type 1 and 3 suggest that HFE and TfR2 function in the same or parallel pathways to regulate iron homeostasis. In line with these observations, a recent study identified that HFE and TfR2 interact in mammalian cells, the interaction is TfR1-independent, and neither HH type 1-nor HH type 3-associated mutations abrogate the interaction (32).
The binding sites of TfR1 and HFE were mapped to the helical domain of TfR1 and to ␣1 and ␣2 helices of HFE (13). However, the binding sites between TfR2 and HFE are still unknown, raising questions about the nature of the TfR2-HFE interaction. To test whether TfR2 utilizes similar domain(s) as TfR1 to interact with HFE, a set of TfR2-TfR1 chimeras was generated to measure their interaction with HFE and W81AHFE, the form of HFE that does not detectably co-immunoprecipitate with TfR1 (17). In complementary experiments, chimeras of HFE and another MHC class I molecule HLA-B7, which has a similar overall structure to HFE but no apparent function in iron homeostasis, were generated to map the domains of HFE that interact with TfR2. The presented data show that the interactions between HFE and TfR2 differ from those between HFE and TfR1. Moreover, further experiments reveal that HFE increases TfR2 levels in hepatic cells independent of holo-Tf.

EXPERIMENTAL PROCEDURES
Plasmid Construction-The pUHD10-3/TfR2 plasmid, which was used to generate a stable HeLa/tTA-TfR2 cell line, was constructed by partial digest of the pcDNA3.1/TfR2 plasmid with EcoRI and XbaI. The EcoRI/XbaI fragment containing the entire TfR2 cDNA, as well as parts of 5Ј-and 3Ј-untranslated regions, was then subcloned into pUHD10-3. To obtain a plasmid for stable expression of human HFE-FLAG in Hela/ tTA-TfR2 cells, the HFE-FLAG cDNA was excised as a BamHI/ NotI fragment from the pcDNA3.1/HFE-FLAG plasmid (33) conferring the Geneticin (G418, Calbiochem, San Diego, CA) resistance marker, and subcloned into pcDNA3-hyg-A (gift from Dr. Maurer, OHSU, Portland, OR) conferring the hygromycin resistance marker.
The HLA-B7-␣1␣2/HFE-␣3-tm-cyto and HLA-B7 plasmids were generated by bridge PCR as described (34). To add a FLAG epitope tag at the C terminus of each protein, cDNAs were amplified by PCR with a combination of primers and templates listed in Supplemental Table S1. The pcDNA3.1/HLA-B7-␣1␣2/HFE-␣3/HLA-B7-tm-cyto-FLAG plasmid was constructed by overlapping PCR using primers and templates listed in Supplemental Table S1. In all cases, the gel-purified PCR products were first inserted into the pGEM-T (Promega, Madison, WI), and then subcloned as XbaI fragments into the pcDNA3.1 vector. All constructs were verified by sequencing ( Fig. 1).
Generation of Stable Cell Lines-HeLa/tTA cells (15,16) were co-transfected with the TfR2/pUHD10-3 plasmid and the pBSpac plasmid containing the puromycin resistance gene as described (39) using Lipofectamine (Invitrogen) to generate HeLa/tTA-TfR2 cells. HeLa/tTA-TfR2 cells were then used to create a cell line stably co-expressing TfR2 and HFE-FLAG (HeLa/tTA-TfR2ϩHFE-FLAG). Here, the cells were seeded at 1 ϫ 10 6 cells per 60-mm dish, grown for 20 h (until 70% confluent), and then co-transfected with 3.6 g of the pcDNA3.1/HFE-FLAG plasmid containing hygromycin resistant gene and 0.4 g of pBSpac using Lipofectamine (Invitrogen) in the ratio 1 g of DNA: 3.75 l of Lipofectamine. Stable clones were selected 10 days after transfection and screened for the co-expression of HFE-FLAG and TfR2 via immunoblotting. HepG2 cells were stably transfected with pcDNA3.1/HFE-FLAG and selected with MEM medium containing 600 g/ml G418 to create HepG2/HFE-FLAG cell line.
Cell Culture and Transient Transfection-Hela/tTA-HFE-FLAG (15,16), Hela/tTA-W81A HFE-FLAG (17) and HeLa/ tTA-TfR2 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (FBS), 400 g/ml G418 and 300 ng/ml puromycin. HeLa/tTA-TfR2ϩHFE-FLAG cells were maintained in the medium described above plus 800 g/ml hygromycin B (Roche Applied Sciences GmbH, Mannheim, Germany). Hep3B/TfR2 cells and Hep3B/TfR2CD cells (28) were maintained in Minimum Essential Medium Eagle (MEM) (Invitrogen, Bethesda, MD) supplemented with 1.0 mM sodium pyruvate, 0.1 mM nonessential amino acids (Invitrogen), 10% FBS, and 400 g/ml G418. Hep3B and HepG2 cells were maintained as described above without G418. In all transient transfection experiments, the cells were seeded at 3 ϫ 10 4 cells/cm 2 16 h prior the experiment. The DNA was used at 0.2 g of DNA/cm 2 , and a Lipofectamine: DNA ratio was kept in the range of 3-3.75 l of Lipofectamine (Invitrogen) per 1 g of DNA in all experiments. The transfection medium was replaced with the normal medium 6 h later, and cells were grown for another 42-66 h prior to lysis on ice in NET/T buffer (150 mM NaCl, 5 mM EDTA, 10 mM Tris, pH 7.4, 1% Triton X-100) containing 1ϫ Complete Mini Protease Inhibitor Mixture (Roche Applied Science, Indianapolis, IN).
The effect of holo-Tf on the interaction between TfR2 and HFE was studied as described above with the following additions. Precleared cell lysates (50 g) were treated with PBS or holo-Tf (0.01, 0.1, or 10 M) for 30 min at 4°C while rotating prior addition of rabbit anti-FLAG antibody.
The pH dependence of TfR2 and HFE association was done as described above with the following modifications. The cells were scraped into PBS (pH 7.4), washed once more, and divided to four aliquots. Each aliquot was washed once in cold PBS titrated to pH 6, 6.5, 7, or 7.5 just before the experiment, and the lysis was initiated by addition of PBS buffers titrated as above, each containing 1% Triton X-100 and protease inhibitors. The lysates were precleared with beads equilibrated with the PBS 1% Triton X-100 buffer at the indicated pH. After overnight immunoprecipitation, the beads were washed with PBS/1% Triton X-100 of the indicated pH. These conditions assured that the different pH was maintained during the entire course of immunoprecipitation experiment.
Immunofluorescence-Hela/tTA-HFE-FLAG cells growing on coverslips were transiently transfected with pcDNA3/ TfR1-cyto-tm/TfR2 104 -250 -HA for 2 days, washed twice with PBS, and fixed for 15 min with PBS/4% (v/v) paraformaldehyde. After washing three times with PBS and blocking with PBS/10% FBS for 60 min at room temperature, the cells were incubated in 1.25 g/ml mouse anti-HA antibody diluted in PBS/5% FBS for 60 min, washed 3 ϫ 5 min with PBS, incubated with goat antimouse IgG AlexaFluor 546 (1:500) diluted in PBS for 60 min, and washed 3 ϫ 5 min with PBS. Coverslips were then mounted FIGURE 1. Schematic representation of the chimera constructs used in this study. A, TfR2-TfR1 chimeras. TfR1 is shown in white, TfR2 is shown in gray. Numbers under diagrams for each full-length protein designate the first amino acid in the cytoplasmic domain (cd), transmembrane domain (tm), and ecto-domain (ecto). FLAG epitope tag was added to the C terminus of individual proteins, HA epitope tag was added either to the N terminus or the C terminus of the chimeras. B, HFE-HLA-B7 chimeras. HFE is shown in gray, HLA-B7 is shown in white. Numbers under diagrams for each full-length protein designate the first amino acid in the signal sequence (ss), ␣1-␣2 domain (␣1-␣2), ␣3 domain (␣3), transmembrane domain (tm), and cytoplasmic domain (cd). FLAG epitope tag was added to the C terminus of each protein. In both cases (panels A and B), chimeras were constructed by replacing the appropriate segments between the parent proteins. The first coding methionine in all four parent proteins is labeled as number 1. We refer to individual recombinant proteins used throughout the study by names given above the illustration of the construct.
in ProLong Gold anti-fade reagent (Molecular Probes/Invitrogen, Eugene, OR). Images were acquired by laser-scanning confocal microscopy using a Plan-Apochromat 63ϫ objective on a Zeiss LSM 5 Pascal confocal inverted microscope.

Effects of HFE on TfR2 and the Chimeras of TfR2-cd-FLAG and TfR2 104 -250 -HA Levels in Hep3B
Cells-To determine if HFE has any effect on TfR2 and TfR2-cd levels, Hep3B/TfR2 and Hep3B/TfR2CD (28) cells were seeded at 3.6 ϫ 10 4 cell/ cm 2 , and 16 -20 h later, they were transfected with pcDNA3.1, pcDNA3.1/HFE-FLAG, or pcDNA3.1/W81A HFE-FLAG using Lipofectamine (Invitrogen) as described (41). The cells were split into two dishes 24-h post-transfection to ensure equal transfection efficiency in the analyzed sets. 24 h later, the cells were treated with PBS or 25 M holo-Tf for 24 h and then lysed as described above. To test whether HFE also affects TfR2 104 -250 -HA levels, Hep3B cells seeded at 3.6 ϫ 10 4 cell/ cm 2 were first transiently transfected with pcDNA3 or pcDNA3/TfR2 104 -250 -HA for 30 h and then split into two dishes to ensure equal transfection efficiency in the following analyzed sets. After 16 h, the cells transfected with pcDNA3 were re-transfected with pcDNA3, and the cells transfected with TfR2 104 -250 -HA were re-transfected with pcDNA3.1 or pcDNA3.1/HFE-FLAG for 24 h. Subsequently, each dish was split into two and incubated with or without 25 M holo-Tf for 24 h. Cell lysates were made as described above, and similar amounts of protein per lane were separated on SDS-PAGE as indicated in the figure legends. The levels of TfR2, TfR2-cd-FLAG, TfR2 104 -250 -HA and actin were quantified using fluorescent immunoblotting as described previously (41).
Isolation of the TfR2/HFE Complex with Bovine Tf-Bovine serum albumin (BSA) and bovine Tf (bTf) were coupled to Affi-Gel-15 (Ag) following the manufacturer's instructions (Bio-Rad), at the concentration of 2 mg of protein/ml Ag. BSA-Ag and Hela/tTA-HFE-FLAG, which do not express TfR2 were used as negative controls to confirm the binding specificity of bTf-Ag with TfR2. Lysates (200 g) of HepG2, HepG2/HFE-FLAG, and Hela/tTA-HFE-FLAG cells were incubated with 200 l of bTf-Ag or BSA-Ag at 4°C for 2 h, the samples were centrifuged at 10,000 rpm for 2 min, and the pellets were washed with PBS. The proteins were eluted with 2ϫ Laemmli buffer and heated at 95°C for 5 min and subjected to 12% SDS-PAGE and immunoblot analysis.

RESULTS
TfR2 Binds W81AHFE-The primary contact of TfR1 with HFE is through two ␣-helices in the helical domain of TfR1, which partially overlaps with the Tf binding site (13). An interaction between mouse TfR2 and HFE expressed in HEK293T, CHO-TVRb-0, and AML12 cells has also been reported recently (32). Because of the high sequence identity between the TfR1 and TfR2 ecto-and transmembrane domains (45%), we sought to determine whether human HFE binds to the ectodomain of human TfR2, as it does with human TfR1. Previous studies indicated that mutation of tryptophan 81 to alanine (W81A) in the ␣1 domain helix of HFE reduces the affinity of HFE for TfR1 by ϳ5,000-fold (14), and TfR1 no longer co-immunoprecipitates with HFE (17). HeLa/tTA-W81AHFE-FLAG cells were thus used in the next experiment. After transient transfection with TfR2 for 48 h, cells were lysed, and W81AHFE-FLAG was immunoprecipitated with rabbit anti-FLAG antibody. Co-immunoprecipitated proteins were then detected with mouse anti-TfR2 and sheep anti-TfR1 antibodies. TfR2 co-immunoprecipitated with W81AHFE-FLAG (Fig.  2). In contrast, TfR1 did not co-immunoprecipitate with W81AHFE-FLAG, which is consistent with previous data (17). Therefore, HFE interacts differently with TfR2 than with TfR1.
TfR2 Binds HFE in the Presence of Holo-Tf and Independent of pH-The interaction between TfR1 and HFE is disrupted by the presence of holo-Tf and cannot be detected at holo-Tf concentrations Ն100 nM (20). The ability of TfR2 to co-immunoprecipitate with W81AHFE-FLAG (Fig. 2) suggests that TfR2 binds to HFE and Tf via two different regions. To explore this possibility in more detail, the effect of holo-Tf on the TfR2/HFE interaction was examined in HeLa/tTA-HFE-FLAG cells. After transient transfection with TfR2 for 48 h, the cells were treated for 30 min at 4°C with increasing amounts of holo-Tf either prior to 5 or after the lysis with NET/T. HFE-FLAG was immunoprecipitated with rabbit anti-FLAG antibody, and interacting proteins were detected by immunoblotting. As reported previously (17) and shown in Fig. 3A, TfR1 binding to HFE-FLAG was reduced in the presence of 0.01 M holo-Tf and completely eliminated in the presence of 0.1 M holo-Tf (Fig.  3A). In contrast, TfR2 binding to HFE-FLAG was similar in the absence and the presence of either 0.01 M or 0.1 M holo-Tf. Even 10 M holo-Tf did not affect the interaction between TfR2 and HFE-FLAG. 5 Moreover, similar amounts of TfR2 interacted with W81AHFE-FLAG regardless of the presence of 25 M holo-Tf (Fig. 3B). As expected, no interaction of TfR1 with W81AHFE-FLAG could be detected (Fig. 3B). Thus, unlike TfR1, the binding site of Tf on TfR2 does not overlap with that of HFE.
Previous structural and biochemical studies of HFE showed that HFE associates with TfR1 in a pH-dependent manner. At 5 M. Chloupkova, personal observation. Pre-absorb lanes correspond to the lysates treated identically to immunoprecipitated samples but without anti-FLAG antibody. Inclusion of this control assured that the observed interaction between TfR2 and HFE is not due to nonspecific binding of either protein to Sepharose 4B/protein A beads. This control is not shown in the remaining figures for simplicity. These results were repeated twice with similar results. DECEMBER 21, 2007 • VOLUME 282 • NUMBER 51 slightly alkaline pH 7.5, TfR1 and HFE bind with a very high affinity (K d ϳ0.6 nM) while at pH 6.0 no binding is detected (11).

Domain-Domain Interaction between HFE and TfR2
To determine whether the interaction of TfR2 and HFE is also pH-dependent, cells stably co-expressing both TfR2 and HFE-FLAG (HeLa/tTA-TfR2ϩHFE-FLAG) were lysed in PBS 1% Triton X-100 buffers titrated to different pH values between 6.0 and 7.5. Proteins were immunoprecipitated at the indicated pH with the rabbit anti-FLAG antibody and analyzed by immunoblotting. TfR2 remained associated with HFE throughout the entire range of pH tested (Fig. 3C).
In contrast, the interaction of TfR1 with HFE could not be detected below pH 6.5 (Fig. 3C) as previously shown (11). These results substantiate the difference between the interactions of TfR1 and TfR2 with HFE.
TfR2 104 -250 Interacts with HFE-The above data shows that the TfR2/HFE and TfR1/HFE interactions differ significantly in nature. Therefore, we began a detailed analysis to determine the portion of TfR2 responsible for HFE binding. For this purpose, a set of TfR2-TfR1 chimeras was constructed (Fig. 1). For simplicity, we refer to TfR2-TfR1 chimeras by names given in Fig. 1. First, TfR2-tm-ecto chimera or full-length TfR2 were transiently transfected into HeLa/tTA-HFE-FLAG cells. Cell lysates were generated 48-h post-transfection, and HFE-FLAG was immunoprecipitated with rabbit anti-FLAG antibody. The immunoblotting results show that both TfR2-tm-ecto and fulllength TfR2 interact with HFE-FLAG (Fig. 4A) as well as with . HFE-FLAG was immunoprecipitated with rabbit anti-FLAG antibody as described under "Experimental Procedures." Interacting proteins were detected by immunoblotting using mouse anti-TfR2, sheep anti-TfR1 and M2 anti-FLAG antibodies. The input lanes correspond to 1:10 of the material used for immunoprecipitation. B, TfR2 binds W81AHFE in the presence of holo-Tf. HeLa/tTA-W81AHFE-FLAG cells were transiently transfected with TfR2 for 24 h, split to two wells, and grown for additional 24 h in the absence (0 M) or the presence (25 M) of holo-Tf. Equal amounts of whole cell lysates prepared in NET/1% Triton X-100 were used to immunoprecipitate W81AHFE-FLAG with rabbit anti-FLAG antibody as described under "Experimental Procedures." Co-immunoprecipitated proteins were detected with antibodies as above (A), actin was detected with a mouse anti-actin antibody. C, pH dependence of the TfR2/HFE interaction. HeLa/ tTA-TfR2ϩHFE-FLAG cells were lysed in PBS/1% Triton X-100 buffers titrated to different pH between 6.0 and 7.5. HFE-FLAG was immunoprecipitated at indicated pH with the rabbit anti-FLAG antibody as described under "Experimental Procedures." Co-immunoprecipitated proteins were detected by immunoblotting with antibodies as above (A). Negative control (pre-absorb, see legend in Fig. 2) was omitted from the figure because it showed no signal for either protein. These results were repeated at least once with similar results.

FIGURE 4. Region between amino acids 104 and 250 in TfR2 interacts with HFE.
A, TfR2-tm-ecto chimera interacts with HFE. TfR2-tm-ecto chimera or full-length TfR2 was transiently transfected into HeLa/tTA-HFE-FLAG cells. Whole cell lysates were prepared in NET/T buffer 48 h post-transfection, and HFE-FLAG was immunoprecipitated with rabbit anti-FLAG antibody. Co-immunoprecipitated proteins were detected by immunoblotting with mouse anti-TfR2 and M2 anti-FLAG antibodies. B, cytoplasmic and transmembrane domains of TfR2 do not interact with W81AHFE. HA-TfR2-cd-FLAG or HA-TfR2tm-FLAG was transiently transfected into HeLa/tTA-W81AHFE-FLAG cells and immunoprecipitated with mouse anti-HA antibody. Co-immunoprecipitated proteins were analyzed by immunoblotting with rabbit anti-FLAG and mouse anti-HA antibodies. C, TfR2 104 -250 expresses on the cell surface. HeLa/tTA-HFE-FLAG cells were seeded on coverslips, and transiently transfected with TfR2 104 -250 -HA for 48 h. Fixed, but not permeabilized, cells were used to localize the expressed chimera. Here, the chimera was first labeled with mouse anti-HA antibody and then detected with goat anti-mouse Alex-aFluor 546 (see "Experimental Procedures"). Scale bar represents 10 m. D, HeLa/tTA-HFE-FLAG cells were transiently transfected with TfR2 104 -250-HA for 48 h. The chimera was immunoprecipitated with mouse anti-HA antibody, and co-immunoprecipitated proteins were detected with rabbit anti-FLAG and mouse anti-HA antibodies. These results were repeated twice with similar results. W81AHFE-FLAG. 6 To verify that the cytoplasmic domain is not required and to establish whether the transmembrane domain is sufficient for the TfR2/HFE interaction, double tagged chimeras containing only the cytoplasmic domain (HA-TfR2-cd-FLAG) or transmembrane domain (HA-TfR2-tm-FLAG) of TfR2 were transiently transfected into HeLa/tTA-W81AHFE-FLAG cells and immunoprecipitated with mouse anti-HA antibody. HeLa/tTA-W81AHFE-FLAG cells were chosen for these studies because they exhibit no detectable interaction between TfR1 and W81AHFE by immunoprecipitation (17), a property that would interfere with the interpretation of results. We found that neither the TfR2 cytoplasmic domain nor the TfR2 transmembrane domain chimeras interacted with W81AHFE-FLAG (Fig. 4B). These experiments eliminate both the cytoplasmic and transmembrane domains of TfR2 as possible sites for HFE interaction. Therefore, the ectodomain of TfR2 is responsible for the interaction with HFE.
A previous study identified that mouse TfR2Y 245X , which corresponds to the human Y250XTfR2 mutant characterized in HH type 3 patients, interacts with HFE (32). Based on this information and on our data (Fig. 4B), we predicted that the binding site of TfR2 for HFE would be located in the putative stalk and protease-like domains of TfR2 within the first 146 amino acids of the ecto-domain. To address this, another chimera (TfR2 104 -250 -HA) was constructed and transiently transfected into HeLa/tTA-HFE-FLAG cells. Although TfR2 104 -250 -HA contains only about a fifth of the TfR2 ecto-domain, immunofluorescence analysis using mouse anti-HA antibody showed that the protein traffics to the cell surface (Fig.  4C). Immunoprecipitation of the chimera with mouse anti-HA antibody, followed by immunoblotting with rabbit anti-FLAG antibody, revealed that TfR2 104 -250 -HA interacts with HFE (Fig. 4D). Thus, the putative stalk and proteaselike domains of TfR2 between amino acid residues 104 and 250 are sufficient for binding to HFE.
HFE-␣3 Domain Interacts with TfR2-Crystallographic studies of the TfR1/HFE complex show that the ␣1 and ␣2 helices of HFE contact TfR1 (13). To map which domain(s) of HFE interacts with TfR2, chimeras composed of various portions of HFE and HLA-B7, an unrelated MHC class I molecule (5), were constructed. Initially, HFE-␣3-tm-cd-FLAG chimera, HFE-FLAG, or HLA-B7-FLAG (Fig. 1) were transiently transfected into Hela/tTA-TfR2 cells for 48 h, and proteins with FLAG epitope tag were immunoprecipitated with rabbit anti-FLAG antibody. The immunoblot probed with mouse anti-TfR2 antibody showed that TfR2 interacts with both HFE-␣3-tm-cyto-FLAG and HFE-FLAG but not with HLA-B7-FLAG (Fig. 5). To determine whether the HFE-␣3 domain alone is sufficient for binding to TfR2, a chimera containing only the ␣3 domain of HFE (HFE-␣3-FLAG) was constructed and analyzed for the TfR2/ HFE interaction via immunoprecipitation using rabbit anti-FLAG antibody. We found that HFE-␣3 domain in the context of the HLA-B7 was sufficient to interact with TfR2 (Fig. 5).
HFE Increases TfR2 and TfR2-cd Chimera Levels in Hep3B Cells-TfR2 is stabilized by holo-Tf in a dose-and concentration-dependent manner in hepatoma cell lines (26,27). Hav-ing mapped the TfR2/HFE interaction, we first confirmed that TfR2 and HFE also interacted in Hep3B/TfR2 cells. 6 We then asked if HFE had any effect on TfR2 protein levels in hepatoma cells stably expressing TfR2 (Hep3B/TfR2). HFE-FLAG or W81AHFE-FLAG were transiently transfected into Hep3B/ TfR2 cells, which were split 24 h later and then treated with or without 25 M holo-Tf for an additional 24 h prior to lysis. The cell lysates were subjected to 8% SDS-PAGE for quantitative immunoblotting of TfR2 (Fig. 6, A and B). In the absence of holo-Tf, both HFE-FLAG and W81AHFE-FLAG led to ϳ2-fold increase in TfR2 levels, reaching those measured in the mocktransfected Hep3B/TfR2 cells treated with holo-Tf. Interestingly, holo-Tf treatment caused a further stabilization of TfR2 in the presence of HFE protein. The levels of TfR2 in holo-Tftreated cells increased by an additional ϳ1.5-2-fold in both HFE-FLAG-and W81AHFE-FLAG-expressing cells compared with the levels of TfR2 in untreated HFE-FLAG-and W81AHFE-FLAG-expressing cells.
In these experiments, we could not completely dissociate the effects of Tf and HFE on TfR2 levels. The same set of experiments was thus repeated using a recently described TfR2-cd-FLAG chimera (28), which consists of the TfR2 cytoplasmic domain followed by the TfR1 transmembrane and ecto-domains (Fig. 1). As such, the chimera interacts with HFE but not with W81AHFE-FLAG (17). Moreover, this chimera is stabilized by holo-Tf (28), making it ideally suited for studying how HFE and holo-Tf affect TfR2 levels independently. The levels of TfR2-cd-FLAG increased by ϳ4-fold in the presence of HFE-FLAG (Fig. 6, C and D). In contrast to full-length TfR2 (Fig. 6B), however, the chimera levels did not further increase upon holo-Tf treatment in cells expressing HFE-FLAG (Fig. 6D). The existing competition between holo-Tf and HFE for binding to TfR1 ecto-domain, which dissociates HFE from the chimera, explains this observation. In the presence of W81AHFE-FLAG, the levels of TfR2-cd-FLAG mimicked the response observed in mock-transfected cells, e.g. only holo-Tf increased levels of the chimera (Fig. 6D).
We sought to investigate whether HFE also stabilizes TfR2 104 -250 -HA chimera, since TfR2 104 -250 -HA chimera is the smallest domain that was identified to interact with HFE. In addition, TfR2 104 -250 -HA chimera does not bind Tf, which would allow us to test whether Tf treatment of cells had a direct or indirect effect on the stability of the construct. Hep3B cells  DECEMBER 21, 2007 • VOLUME 282 • NUMBER 51 transiently transfected with pcDNA3/TfR2 104 -250 -HA for 30 h were split into two dishes and re-transfected with pcDNA3 or pcDNA3/TfR2 104 -250 -HA for 24 h, and then split into two dishes in presence or absence of 25 M holo-Tf for 24 h. The levels of TfR2 104 -250 -HA increased by ϳ2-fold in the presence of HFE-FLAG (Fig. 6, E and F). In contrast to full-length TfR2 (Fig. 6B), which is stabilized when cells are treated with holo-Tf, the amount of TfR2 104 -250 -HA does not change because TfR2 104 -250 -HA does not bind holo-Tf. These combined experiments with TfR2, TfR2-cd-FLAG and TfR2 104 -250 -HA indicated that HFE modulates TfR2 levels independent of holo-Tf.

Domain-Domain Interaction between HFE and TfR2
HFE Interacts with TfR2 and Tf Independent of TfR1-The interaction of HFE with TfR2 in the presence of physiological concentrations of Tf showed that TfR2 behaves very differently than TfR1 where HFE and Tf compete for binding to TfR1. We assayed the TfR2/HFE complex to determine if Tf bound to TfR2 in the presence of HFE using two different approaches. In the first set of experiments HeLa cells expressing W81AHFE-FLAG were transiently transfected with the TfR2 plasmid and treated with or without 25 M holo-Tf prior to lysis. W81AHFE-FLAG was then immunoprecipitated with M2 anti-FLAG antibody. The immunoblot probed with rabbit anti-TfR2 and goat anti-Tf antibodies showed the presence of both pro-teins in the purified immunocomplexes (Fig. 7A). Similar results were obtained when either HFE-FLAG or TfR2 was immunoprecipitated (data not shown). No interaction between W81AHFE-FLAG and TfR1 was detected in co-precipitation studies. Tf competes with HFE for binding to TfR1 (Fig. 2) (17). In the second set of experiments we tested the ability of bovine Tf (bTf) to co-isolate the TfR2/HFE complex. bTf binds readily to human TfR2 (42) but has a 2,000-fold lower affinity for human TfR1 (43). Both HFE and TfR2 were isolated by bTf-Affi-Gel (bTf-Ag) whereas no TfR1 could be detected (Fig. 7B). These results indicate that both HFE and Tf bind to TfR2 simultaneously. The interaction of the HFE/TfR2 complex is very different than that of TfR1 where no TfR1/HFE complex can be detected in the presence of 100 nM holo-Tf.

DISCUSSION
Hereditary hemochromatosis protein HFE is related to the class 1 major histocompatibility complex (MHC-1) proteins in both structure and sequence. One of the interesting characteristics of MHC-1-related and MHC-1 proteins is that despite striking similarities among their overall structures, individual proteins interact with their ligands by very distinct domains. For instance, classical MHC-1 molecules use the ␣1-␣2 domain helices in combination with bound peptides to interact with 0 FIGURE 6. HFE increases TfR2 levels in Hep3B/TfR2 cells. A and B, full-length TfR2 levels increase in the presence of HFE and W81AHFE and respond to holo-Tf. Hep3B cells stably expressing TfR2 (Hep3B/TfR2) were transiently transfected with HFE-FLAG or W81AHFE-FLAG. 24 h later, the cells were split, and then treated with or without 25 M holo-Tf for additional 24 h. After lysis in NET/T buffer, equal amounts of the whole cell lysates were subjected to 8% SDS-PAGE followed by immunoblotting using rabbit anti-TfR2, rabbit anti-HFE, and mouse anti-actin antibodies. A representative immunoblot from three independent experiments is shown in panel A. Relative levels of TfR2 determined with the use of fluorescence-labeled secondary antibodies are shown in panel B. Here, intensities of individual bands were quantified using the Odyssey software (41). Relative levels of TfR2 were first normalized to actin and then expressed as relative % of control levels. Control represents mock-transfected cells that were treated with PBS (0 M holo-Tf). Quantitative data represent the average values obtained from three independent experiments; the bars correspond to standard errors. The Student's t test revealed that no statistical differences in TfR2 levels were measured between the HFE-and W81AHFE-transfected cells (p Ͼ 0.05). C and D, levels of TfR2-cd-FLAG chimera increase in the presence of HFE but not in the presence of W81AHFE. Hep3B cells stably expressing TfR2-cd-FLAG (Hep3B/TfR2CD) (28) were transiently transfected with HFE-FLAG or W81AHFE-FLAG and treated as described in A. Equal amounts of proteins were subjected to 8% SDS-PAGE and analyzed by immunoblotting with rabbit anti-FLAG and mouse anti-actin antibodies. A representative immunoblot from three independent experiments is shown in panel C. All the studied proteins contain FLAG epitope tag, which was used for immunodetection. TfR2-cd-FLAG migrates as ϳ98 kDa, but both HFE-FLAG and W81AHFE-FLAG migrate as ϳ45 kDa species on SDS-PAGE. These differences in size prevented interference from overlapping migration patterns of individual proteins. Intensities of individual bands were quantified and normalized as described in the legend for panel B. Relative levels of TfR2-cd-FLAG chimera graphed in panel D represent the average values obtained from three independent experiments. The bars correspond to standard errors. The Student's t test revealed no statistical differences in TfR2-cd-FLAG levels between the mock and W81AHFE-transfected cells, as well as between PBS-treated and holo-Tf-treated cells transfected with HFE (p Ͼ 0.05). These results were repeated twice with similar results. E and F, TfR2 104 -250 -HA levels increase in the presence of HFE and do not respond to holo-Tf. Hep3B cells seeded at 3.6 ϫ 10 4 cell/cm 2 were transiently transfected with pcDNA3 (mock) or pcDNA3/TfR2 104 -250 -HA for 30 h and then split into two dishes. After 16 h, the cells transfected with pcDNA3 were re-transfected with pcDNA3.1, and the cells transfected with TfR2 104 -250 -HA were re-transfected with pcDNA3.1 (mock) or pcDNA3.1/HFE-FLAG for 24 h. The cells were then split into two dishes in presence or absence of 25 M holo-Tf for 24 h, washed with PBS and lysed in NET/T buffer. Equal amounts of the whole cell lysates were subjected to 12% SDS-PAGE followed by immunoblotting using mouse anti-HA, rabbit anti-FLAG, and mouse anti-actin antibodies. A representative immunoblot from two independent experiments is shown in panel E. Relative levels of TfR2 104 -250 -HA chimera (F) were first normalized to actin and then expressed as relative % of control levels. Here, Control represents TfR2 104 -250 -HA/pcDNA3.1-transfected cells that were treated with PBS (0 M holo-Tf). The Student's t test revealed no statistical differences in TfR2 104 -250 -HA levels between PBS-treated and holo-Tf-treated cells transfected with empty vector or HFE (p Ͼ 0.05). These results were repeated once with similar results.
In the case of HFE, the ␣1-␣2 domain helices interact with TfR1 (13), connecting HFE to iron homeostasis. Effects of HFE on iron homeostasis, however, do not depend solely on the interaction with TfR1 and are more complex than simple competition between HFE and Tf for binding to TfR1 (17,20). Thus, given the nature of MHC-1 and MHC-1-related molecules to accommodate more than one protein ligand, the function of HFE in iron metabolism might depend on ligands in addition to TfR1. In agreement with this idea, a recent study showed that mouse HFE and TfR2 interact in transfected mammalian cells. In that study, the authors demonstrated that none of the HHrelated mutations are capable of abrogating the interaction between TfR2 and HFE (32). The finding that the truncated TfR2 from the TfR2 Y245X mouse, which lacks 553 amino acids in its ecto-domain, interacts with HFE was surprising because the C-terminal region of TfR1 is crucial for interaction with HFE (48). Consistent with that observation, the authors also demonstrated that a mutation in the conserved RGD motif, located within a putative holo-Tf binding pocket of TfR2, does not affect the TfR2/HFE interaction (32). In contrast, the same mutation in TfR1 abrogates its interaction with HFE (49). These two findings inspired us to map the interactions of TfR2 with HFE and to determine how HFE affects TfR2 function.
We first observed that TfR2 interacts with W81AHFE, a mutant form of HFE that does not co-immunoprecipitate with TfR1 (17). It suggested that the HFE domain(s) binding to TfR2 differs from that binding to TfR1. All our subsequent findings confirmed the different nature of TfR2/HFE interaction. The interaction between TfR2 and HFE is not competed by holo-Tf (15,16) and TfR2 remains associated with HFE even when the pH is lowered to 6.0 (11). Using TfR2-TfR1 chimeras, the region of TfR2 interacting with HFE was mapped to the putative stalk and protease-like domains of TfR2 located between residues 104 and 250, which is consistent with the ability of mouse TfR2 Y245X to interact with mouse HFE (32). Using HFE-HLA-B7 chimeras, the ␣3 domain of HFE is sufficient for the interaction between TfR2 and HFE. This finding indicates that the docking site in HFE for TfR2 differs from that for TfR1, in which ␣1 and ␣2 domains of HFE associate with TfR1 (13). The data also explain why an earlier study did not detect an interaction between purified HFE ecto-domain lacking residues 276 -285 and TfR2 ecto-domain lacking residues 104 -132 in vitro (24). Taken together, the results presented here confirm our hypothesis that TfR2 and HFE interact via different regions than TfR1 and HFE.
The possible physiological significance of the identified TfR2/HFE interaction was further investigated. Hepatoma cells, in which the levels of TfR2 are regulated by holo-Tf (26,27) were employed for this study. Our experiments show that TfR2 levels increase in response to transient expression of either HFE-FLAG or W81AHFE-FLAG in Hep3B/TfR2 cells. Moreover, TfR2 still responds to holo-Tf treatment in the same cells, which is not surprising since holo-Tf does not abrogate the binding of HFE to TfR2. Further experiments of TfR2-cd-FLAG and TfR2 104 -250 -HA chimeras in response to holo-Tf treatment and HFE expression demonstrate that HFE modulates TfR2 levels independent of Tf. These results indicate that not only holo-Tf but also HFE increases the amount of TfR2 in hepatoma cell lines. Thus holo-Tf affects TfR2 levels in cells expressing HFE in two ways. It binds to TfR2 and stabilizes the complex by increasing the recycling of TfR2 and thus decreasing its trafficking to the lysosome (28,41). Holo-Tf also competes with HFE for binding to TfR1 and releases HFE to bind TfR2, which increases TfR2 levels. Because Tf saturation in the blood correlates with TfR2 levels and hepcidin transcription, the Tf/TfR2/HFE complex may be part of the sensing complex leading to hepcidin induction. HFE would increase the levels of TfR2 and hence the transcription of hepcidin. Thus HFE would serves as a set point for the hepcidin transcription. The degree of iron saturation would be the sensor for body iron levels. How this Tf/TfR2/HFE complex signals to increase hepcidin transcription remains to be determined. A, HeLa/tTA-W81AHFE-FLAG cells were transiently transfected with TfR2 for 48 h. Prior to lysis in NET/T buffer, the cells were treated with or without 25 M holo-Tf for 1 h at 37°C. Cell lysate (10 g) was loaded (Input) directly onto gels, and 100 g was immunoprecipitated with M2 anti-FLAG antibody as described under "Experimental Procedures." Proteins were detected with rabbit anti-TfR2, goat anti-Tf, and rabbit anti-FLAG antibodies. Negative control (Pre-absorb, see legend in Fig. 2) was omitted from the figure because it showed no signal for either protein. These results were repeated once with similar results. B, lysates (200 g) of HepG2, HepG2/HFE-FLAG, and Hela/tTA-HFE-FLAG cells were either loaded directly onto the gel (Input) or incubated with 200 l of bTf-Ag or BSA-Ag and rotated at 4°C for 2 h. The samples were centrifuged at 10,000 rpm for 2 min, washed with PBS, the pellets were eluted with 2ϫ Laemmli buffer and denatured by heating at 95°C for 5 min. Samples were subjected to 12% SDS-PAGE, and transferred to nitrocellulose. The bands were detected as described under "Experimental Procedures." Controls included: HeLa/HFE cells lacking TfR2 incubated with bTf-Ag; bTf-Ag extracted with loading buffer (NC); BSA-Ag extracted with loading buffer (NC); and HepG2/HFE cell extracts incubated with BSA-Ag (to determine specificity of binding to the bTf-Ag beads). Both BSA and Tf have approximately the same charge (pI). The bands shown in B were cropped from a single exposure of the immunoblot. These experiments were repeated twice with similar results.