Iron Export through the Transporter Ferroportin 1 Is Modulated by the Iron Chaperone PCBP2*

Ferroportin 1 (FPN1) is an iron export protein found in mammals. FPN1 is important for the export of iron across the basolateral membrane of absorptive enterocytes and across the plasma membrane of macrophages. The expression of FPN1 is regulated by hepcidin, which binds to FPN1 and then induces its degradation. Previously, we demonstrated that divalent metal transporter 1 (DMT1) interacts with the intracellular iron chaperone protein poly(rC)-binding protein 2 (PCBP2). Subsequently, PCBP2 receives iron from DMT1 and then disengages from the transporter. In this study, we investigated the function of PCBP2 in iron export. Mammalian genomes encode four PCBPs (i.e. PCBP1–4). Here, for the first time, we demonstrated using both yeast and mammalian cells that PCBP2, but not PCBP1, PCBP3, or PCBP4, binds with FPN1. Importantly, iron-loaded, but not iron-depleted, PCBP2 interacts with FPN1. The PCBP2-binding domain of FPN1 was identified in its C-terminal cytoplasmic region. The silencing of PCBP2 expression suppressed FPN1-dependent iron export from cells. These results suggest that FPN1 exports iron received from the iron chaperone PCBP2. Therefore, it was found that PCBP2 modulates cellular iron export, which is an important physiological process.

siderable challenge to cells and organisms, which have evolved sophisticated mechanisms to satisfy their metabolic needs and concomitantly minimize the risk of toxicity.
It has long been known that there is very little, yet potentially toxic, low-molecular-weight iron in rapidly metabolizing cells (6). However, the molecular mechanisms involved in intracellular iron transport have remained elusive (7). Recently, it was revealed that one of the mechanisms of iron transport and metabolism involves intracellular chaperone proteins (8). It was reported that poly(rC)-binding protein 1 (PCBP1), 3 also referred to as ␣-CP1 or hnRNP E1, is a cytosolic iron chaperone protein that delivers iron to ferritin (9). PCBP1 binds to iron with micromolar affinity at a molar ratio of 3:1 Fe:PCBP1. Notably, PCBP1 has been shown previously to function as an RNAand DNA-binding protein and has been identified as a member of a family of four homologous proteins containing three heterogeneous nuclear ribonucleoprotein K-homology (KH) domains (10).
Mammalian genomes encode four PCBPs, and every PCBP has been found to exhibit iron chaperone activity (11). Interestingly, PCBP1 is an intronless gene (12) that likely arose from the retrotransposition of a splice variant of PCBP2 mRNA. This became fixed in the genome, probably because the associated gene product performed a unique function not shared by the other PCBPs (13). The other members of the PCBP gene family undergo alternative splicing that results in greater diversity in the expression of PCBPs in mammals (13). PCBP1 is specific to mammals, whereas PCBP2, PCBP3, and PCBP4 are present throughout vertebrates, flies, and worms (8).
Ferroportin 1 (FPN1) is an important cellular iron exporter, which is found on the plasma membranes of certain types of cells, especially duodenal enterocytes and macrophages (20 -22). One of the DMT1 isoforms is expressed specifically on the apical membrane of duodenal enterocytes (18), whereas FPN1 is expressed on the basolateral membrane of duodenal enterocytes (22). Hepcidin is a 25-amino acid peptide synthesized by hepatocytes (23)(24)(25). The major molecular target of hepcidin is FPN1, which transfers cellular iron to the plasma (26). Hepcidin binds to the extracellular loop of FPN1, and this binding induces the endocytosis and proteolysis of FPN1, leading to a decrease in the export of iron to the plasma (27). This hepcidin-FPN1 binding interaction is an essential mechanism for the regulation of body iron stores; it is not yet clear how FPN1 receives cytoplasmic iron to export it from cells. However, the process of iron release is crucial to the regulation of systemic iron metabolism (28,29).
In the present study, we identified a novel function of PCBP2 as an iron chaperone. We demonstrated that the C-terminal cytoplasmic region of FPN1 interacts specifically with PCBP2 but not with PCBP1, PCBP3, or PCBP4. We also determined that iron-depleted FPN1 interacts with iron-loaded PCBP2 but not with iron-depleted PCBP2. Silencing PCBP2 expression suppressed the iron export activity of FPN1. In addition, we showed that DMT1 binds to PCBP2 but not PCBP1, PCBP3, or PCBP4. This observation indicates that the interaction of DMT1 with these candidate iron chaperone proteins is highly specific to PCBP2. All four PCBPs exhibit iron binding activity, and their target molecules could potentially be different. The results of our study suggest that PCBP2 has important functions in both iron import and export and serves as a gateway keeper for the safe transport of iron.

FPN1 Interacts with PCBP2 but Not with PCBP1, PCBP3, or PCBP4 in Yeast Cells
Our previous study showed that PCBP2 was identified as a binding partner of DMT1 using the N-terminal cytoplasmic region of DMT1 as bait in yeast two-hybrid screening (18). FPN1 contains three major cytoplasmic regions (Fig. 1A): an N-terminal region (FPN1-N; amino acids , an intracellular loop between the sixth and seventh transmembrane regions (FPN1-M; amino acids 220 -307) and a C-terminal region (FPN1-C; amino acids 534 -572) (30). We hypothesized that one of these regions might function as a PCBP-binding domain. To assess this possibility, yeast two-hybrid assays were performed using these three regions of FPN1 as bait (Fig. 1).
In this study, the full-length form of PCBP1 (amino acids 1-356, I-F) or PCBP2 (amino acids 1-366, II-F) was cloned into the prey vector and used for the yeast two-hybrid assay (Fig. 1A). We also cloned the full-length form of PCBP3 (amino acids 1-371, III-F) and PCBP4 (amino acids 1-403, IV-F) and made the recombinant yeast prey vectors in Escherichia coli (Fig. 1A). However, when we transformed the yeast with the prey vectors containing full-length PCBP3 or PCBP4, recombinant yeast clones could not be grown on selection plates. We repeatedly tried to isolate the recombinant yeast clones but were not successful. Hence, it was presumed that PCBP3 and PCBP4 in the prey vector were toxic for yeast cells.
To examine the expression of recombinant proteins and assess whether each of them was expressed at their expected molecular weight in yeast cells, whole yeast protein was extracted and analyzed by immunoblotting (Fig. 1B). The expected molecular sizes of the bait fusion proteins were obtained (Fig. 1B, left panel): vector (v), 23 kDa; FPN1-N, 25 kDa; FPN1-M, 31 kDa; and FPN1-C, 26 kDa. Of note, when bait proteins were examined using the anti-Myc mAb, an extra band in the bait vector alone was evident (Fig. 1B, left panel). When the same bait vector and yeast strain were used in a previous report by others, a similar extra band was detected by the anti-Myc mAb (32).
When prey proteins were examined by anti-HA mAb, multiple bands on the blots were evident, even using the prey vector alone (Fig. 1B, middle and right panels). Considering this finding, two other sources of anti-HA mAbs (from Wako Chemicals (Osaka Japan) and Covance Laboratories (Madison, WI)) were utilized for detection, and again, a similar pattern of multiple bands was detected (data not shown). Of interest, when the same prey vector and yeast strain were used in previous reports by others, similar multiple bands were detected by the anti-HA mAb (32,33).
In summary, the band in each lane indicated by an arrowhead (Fig. 1B) was the expected molecular size of the fusion proteins. Hence, we demonstrated that all bait and prey constructs worked appropriately in yeast cells generating recombinant proteins of the correct size. These were then used for the studies below in Fig. 1C.
When the N-terminal cytoplasmic region (FPN1-N) or the intracellular loop (FPN1-M) was expressed in yeast cells, these regions of FPN1 did not interact with PCBP1/PCBP2 or with any of the KH domains of PCBP1, PCBP2, PCBP3, and PCBP4 on SD-Leu-Trp-His-Ade-selective plates (Fig. 1C). In contrast, when the C-terminal cytoplasmic region of FPN1 (FPN1-C) was used as bait, we detected an interaction between FPN1 and the full-length form (II-F) or the KH2 domain (II-2) of PCBP2 (Fig. 1C). Importantly, the only other observed interaction under these conditions was the P53-SV40-T positive control. In summary, these findings suggest that the C-terminal cytoplas- were cloned into a vector containing the GAL4 DNA-activating domain (prey vector). B, confirmation of bait and prey expression. Yeast cells were harvested and extracted for whole protein using the method described under "Experimental Procedures." Bait protein expression was detected by using the anti-Myc mAb, and prey was identified by anti-HA mAb. The arrowheads indicate the appropriate molecular weight of each recombinant protein. Results are typical experiments of at least three performed. C, analysis of the interaction between FPN1 and PCBPs. After mating yeast expressing the bait protein with yeast expressing the prey protein, yeast cells were grown on both SD/-Trp/-Leu and SD/-Trp/-Leu/-His/-Ade selection media and incubated for 48 at 30°C. The vector (v) served as a negative control, and the well characterized interaction between p53 and SV40-T served as a positive control, as suggested by the manufacturer. Similar results were obtained from at least three independent experiments. mic region of FPN1 interacts with PCBP2 and specifically the KH2 domain of PCBP2.

PCBP2 KH2 Domain Is a Crucial Region to Bind Not Only FPN1 but Also DMT1 in Yeast Cells
As shown in Fig. 1, the C-terminal cytoplasmic region of FPN1 as bait interacted with PCBP2 in yeast cells. Next, we investigated the interactions between DMT1 and the PCBP family (Fig. 2) and further confirmed the interaction between FPN1 and PCBP2 by means of exchanging the bait and prey inserts. We used the same region of the KH domains in prey (Fig. 1A) and bait ( Fig. 2A) constructs. Using cDNA segments corresponding to the KH1, KH2, and KH3 domains of PCBP1, PCBP2, PCBP3, and PCBP4 or full-length forms of PCBP1, PCBP2, PCBP3, and PCBP4 as bait ( Fig. 2A) and DMT1 or FPN1 as prey, we performed yeast two-hybrid assays (Fig. 2). In this case, the full-length forms of PCBP3 and PCBP4 cloned into bait vectors could be isolated in yeast cells.
The band in each lane indicated by an arrowhead (Fig. 2B, bottom panel) was the expected molecular size of the prey fusion proteins, that is: vector (v), 23 kDa; FPN1-C, 26 kDa; and DMT1-N, 40 kDa. When FPN1-C or DMT1-N was expressed in yeast cells and detected by anti-HA mAb, multiple bands were again detected as shown in Fig. 1B. However, the band in each lane indicated by an arrowhead was the expected molecular size (Fig. 2B, bottom panel).
Hence, we determined that all bait and prey constructs worked appropriately in yeast cells generating the correct size fusion proteins. These were then used for the further studies described below.
Notably, using SD-Leu-Trp-His-Ade-selective plates (Fig.  2C), the interactions of the KH2 domain and the full-length form of PCBP2 with the N-terminal cytoplasmic region of DMT1 and the C-terminal cytoplasmic region of FPN1 were detected in yeast cells along with the positive control (i.e. P53-SV40-T). However, other members of the PCBP family, or any of the KH domains of PCBP1, PCBP2, (except PCBP2 KH2), PCBP3, and PCBP4, were not demonstrated to interact with DMT1 or FPN1 (Fig. 2C).

FPN1/DMT1 Interacts with PCBP2 but Not with PCBP1, PCBP3, or PCBP4 in Mammalian Cells
Considering the results in yeast cells described above, we then examined the interactions of FPN1/DMT1 with all members of the PCBP family in mammalian cells (Fig. 2D). Expression of all four PCBPs was clearly detected in both the total protein extraction and the membrane fraction using immuno-blotting (Fig. 2D). It should be noted that multiple PCBP2 bands on the immunoblot were observed throughout the present study. When the expression of PCBP2 was silenced using siRNA, the intensities of these multiple bands were reduced (as shown later in Figs. 6A and 7D). Previous investigations have reported that the multiple PCBP2 bands on the immunoblots may represent alternative PCBP2 gene splice forms (34 -36).

The N-terminal Region of DMT1 and the C-terminal Region of FPN1 Are Crucial for Their Interactions with PCBP2 in Mammalian Cells
To investigate the importance of the cytoplasmic domains of both DMT1 and FPN1 in their interactions with PCBP2 in mammalian cells, two deletion mutants were constructed and expressed in HEp-2 cells (Fig. 3). These mutants were DMT1, lacking its N-terminal cytoplasmic region (DMT1⌬N), and FPN1, lacking its C-terminal cytoplasmic region (FPN1⌬C) (Fig.  3A). Mutants were prepared as described previously (18,37).
We then analyzed the subcellular localization of the deletion mutants, as this could affect their function (Fig. 3B). In these studies, confocal microscopy was used to examine the localization in HEp-2 cells of full-length (wild type (WT)) or deletion mutants of DMT1 or FPN1. The GFP-tagged deletion mutant forms of DMT1 or FPN1 (green) were transiently transfected with the mCherry-tagged full-length (WT) forms of DMT1, FPN1, or NADPH cytochrome P450 reductase (CPR) (red), and co-localization (yellow) in the electronic merge was then assessed (Fig. 3B). Cells transfected with the GFP-tagged deletion mutant forms of DMT1 or FPN1 (green) were also stained with the anti-protein disulfide isomerase (PDI) mAb (red) or anti-LAMP2 mAb (red) to examine co-localization in the endoplasmic reticulum (ER) lumen and lysosome, respectively (Fig. 3B).
The full-length forms of DMT1A-I and FPN1 were localized to the plasma membrane as described previously (18,27). Importantly, both DMT1⌬N and FPN1⌬C were also expressed at the cell surface and intracellularly (Fig. 3B). Furthermore, they partially co-localized with the full-length forms of these molecules at the plasma membrane (Fig. 3B).
PDI is an ER protein that is localized to the ER lumen (38); it is notable that neither DMT1⌬N nor FPN1⌬C was strongly co-localized with PDI (Fig. 3B). In contrast, DMT1⌬N or FPN1⌬C demonstrated pronounced co-localization with CPR, which is known to be markedly expressed in the ER membrane (18). Neither DMT1⌬N nor FPN1⌬C was co-localized with LAMP2 ( Fig. 3B), which is a marker of late endosomes/lysosomes (39).
DMT1 is known to exist in several isoforms (18,40,41). In fact, DMT1A-I has 98 amino acids in its N-terminal cytosolic region, and DMT1B-I lacks N-terminal amino acids 1-29 of DMT1A-I (18, 40, 41). DMT1A-I is localized to the plasma Vectors containing the GAL4 DNA-activating domain fused to the N-terminal cytoplasmic region of DMT1 or the C-terminal cytoplasmic region of FPN1 were used as prey. B, confirmation of bait and prey expression. Yeast cells were harvested, and whole protein was prepared using the method described under "Experimental Procedures." Bait protein expression was detected by using anti-Myc mAb, and prey was detected by anti-HA mAb. The arrowheads indicate the appropriate molecular weight of each recombinant protein. Results are typical experiments of at least three performed. v, vector. C, analysis of the interactions of DMT1 or FPN1 with PCBPs. After mating yeast expressing the bait protein with yeast expressing the prey protein, yeast cells were grown on both SD/-Trp/-Leu and SD/-Trp/-Leu/-His/-Ade selection media and incubated for 48 h at 30°C. The vector alone served as a negative control, and the well characterized p53 and SV40-T interaction served as a positive control, as suggested by the manufacturer. Similar results were obtained from at least three independent experiments. D, analysis of the interactions of DMT1 or FPN1 with PCBP1-4 using a co-immunoprecipitation assay. HEp-2 cells were transfected with FPN1-GFP or DMT1A-I-GFP and incubated for 48 h at 37°C. After incubation, total protein was extracted in TNE buffer and assessed by immunoblotting using the appropriate anti-PCBP1, -2, -3, or -4 or GFP Ab (left lane). The membrane fraction was prepared using the method described under "Experimental Procedures." Using either the total protein extract or the membrane fraction, co-immunoprecipitation analyses (co-IP) were performed. The precipitates were analyzed by immunoblotting with anti-PCBP1 pAb, anti-PCBP2 mAb, anti-PCBP3 pAb, anti-PCBP4 pAb, and anti-GFP pAb (right lane). Similar results were obtained from at least three independent experiments. FIGURE 3. The N-terminal cytoplasmic region of DMT1 and the C-terminal cytoplasmic region of FPN1 are essential for their binding to PCBP2 in mammalian cells. A, the schematics show the full-length or deletion mutants of DMT1 or FPN1. The N-terminal cytoplasmic region of full-length DMT1A-I consists of 98 amino acids. DMT1⌬N is a deletion mutant of DMT1 that lacks amino acids 2-98. Amino acids 534 -572 correspond to the C-terminal cytoplasmic region of FPN1. FPN1⌬C is a deletion mutant of FPN1 that lacks amino acids 535-572. B, the cellular localization of full-length or the deletion mutants of DMT1 or FPN1. The GFP-tagged deletion mutant forms of DMT1 and FPN1 were transiently transfected with mCherry-tagged full-length forms of DMT1, FPN1, and CPR. Cells were stained with anti-PDI mAb or anti-LAMP2 mAb. The arrowhead shows co-localization of full-length and deletion-mutant forms of DMT1 or FPN1 at the plasma membrane. Similar results were obtained from at least three experiments. C, GFP-tagged full-length and deletion-mutant forms of DMT1 and FPN1 were transiently transfected into HEp-2 cells. The cells were lysed in TNE buffer and analyzed by immunoblotting using ␣-PCBP2 Ab. Co-immunoprecipitation (IP) was also performed with anti-GFP pAb-conjugated protein A beads. These samples were analyzed by immunoblotting (WB) using an anti-PCBP2 mAb. Similar results were obtained from at least three independent experiments. D, iron transport activity of full-length and deletion-mutant forms of DMT1 or FPN1. Full-length or deletion-mutant forms of DMT1 were transfected into HEp-2 cells. Cells were loaded with 100 M FAC for 6 h at 37°C. The cells were lysed in 50 mM NaOH for the ferrozine-based iron assay. HEp-2 cells stably expressing DMT1A-I were cultured in serum-free DMEM and preloaded with 100 M FAC for 6 h at 37°C. The cells were washed with PBS, transfected with the full-length or deletion mutant of FPN1-GFP expression plasmid, and then incubated in serum-free RPMI 1640 medium for 18 h at 37°C. The cells were then lysed in 50 mM NaOH for the ferrozine-based iron assay. The results shown on the histogram are presented as the means Ϯ S.E. of three independent experiments. Significance was determined by Student's t test: ‡, p Ͻ 0.01; §, p Ͻ 0.001. membrane, whereas DMT1B-I is localized to late endosomes/ lysosomes (16,18). As demonstrated from our studies examining the co-localization with LAMP2, the localization of DMT1⌬N was obviously different from late endosomes/lysosomes (shown in Fig. 3B). Indeed, DMT1⌬N and FPN1⌬C were localized to the plasma membrane and the ER membrane.
Co-immunoprecipitation was then performed to investigate the interaction of PCBP2 with DMT1 and FPN1 (Fig. 3C). In these studies, ␣-GFP was used to immunoprecipitate FPN1 and DMT1 from cells transfected with DMT1A-I-GFP, FPN1-GFP, and their deletion mutants. Immunoblotting was then performed using anti-PCBP2 (␣-PCBP2). When the full-length form of DMT1 or FPN1 was expressed in HEp-2 cells, PCBP2 was co-immunoprecipitated with DMT1 or FPN1 (Fig. 3C).
In contrast, when DMT1⌬N or FPN1⌬C was expressed in HEp-2 cells, PCBP2 was not co-immunoprecipitated with DMT1⌬N or FPN1⌬C at all (Fig. 3C). This occurred even though deletion mutants were co-localized with the full-length forms of these proteins at the plasma membrane (Fig. 3B). Therefore, neither DMT1⌬N nor FPN1⌬C interacted with PCBP2 (Fig. 3). These results confirmed that the N-terminal cytoplasmic region of DMT1 is a domain that binds to PCBP2, as shown in our previous report (14). Moreover, these studies demonstrated that the C-terminal cytoplasmic region of FPN1 is responsible for its binding to PCBP2 in mammalian cells.

DMT1⌬N and FPN1⌬C Lose Iron Transport Activity
The cells expressing the full-length form of DMT1 incorporated iron very efficiently (Fig. 3D). When DMT1⌬N was expressed in cells, the cells did not incorporate iron from ferric ammonium citrate (FAC) (100 M), and thus DMT1⌬N lost its iron import activity (p Ͻ 0.001). The cells expressing the fulllength form of FPN1 exported iron efficiently (Fig. 3D). On the other hand, when FPN1⌬C was expressed, the cells did not export iron efficiently, indicating that FPN1⌬C lost its iron export activity (p Ͻ 0.01). Thus, these deletion mutants lost their iron transport activity in HEp-2 cells (Fig. 3D).
Taken together, the results shown in Figs. 1-3 confirmed the interactions between DMT1/FPN1 and PCBP2 in yeast and mammalian cells. In contrast, no interactions between DMT1/ FPN1 and other members of the PCBP family were detected (Fig. 3). Therefore, we next investigated the function of PCBP2 in iron export.

FPN1 Interacts with Iron-loaded PCBP2
Previously, we described the iron-dependent DMT1-PCBP2 interaction (14). Iron-loaded, but not iron-depleted, DMT1 interacts with PCBP2 (14). However, on the other hand, the association of PCBP2 with iron does not alter its DMT1 binding activity (14). Therefore, in this study, we investigated the iron dependence of the FPN1-PCBP2 interaction and then examined whether FPN1 or PCBP2 needed to be in an iron-bound state to enable the FPN1-PCBP2 interaction (Fig. 4).
It has been reported that PCBP2 is localized primarily to the nucleus and the cytoplasm (42,43), and we discovered that some part of PCBP2 also resides within the cell membrane (14). Fig. 4A, using immunoblotting, demonstrates that FPN1 was not depleted from this fraction via Na 2 CO 3 treatment, whereas FIGURE 4. Iron-depleted FPN1 can bind to iron-loaded PCBP2 in mammalian cells. A, preparation of membrane and m⌬P2 fractions. Membrane fractionation was performed on HEp-2 cells expressing FPN1-GFP as described under "Experimental Procedures." The membrane fraction was pelleted via centrifugation, resuspended in homogenization buffer containing 0.1 M Na 2 CO 3 (pH 11), and centrifuged again at 100,000 ϫ g. The resultant pellet was collected as the PCBP2-depleted membrane (m⌬P2) fraction and analyzed by immunoblotting using the appropriate FPN1, PCBP2, and TfR1 antibodies. B, analysis of the iron dependence of the FPN1-PCBP2 interaction. The m⌬P2 fraction was incubated with iron (freshly prepared FeCl 2 ) or the iron chelator CDTA for 4 h at 4°C. The soluble fraction was preincubated with iron or CDTA for 4 h at 4°C. Before immunoprecipitation, the m⌬P2 fraction or soluble fraction was pretreated with FeCl 2 or CDTA and analyzed by immunoblotting (WB) with anti-GFP pAb and anti-PCBP2 mAb. Similar results were obtained from at least three independent experiments. C, analysis of the iron dependence of the FPN1-PCBP2 interaction. The m⌬P2 fraction was incubated with iron (freshly prepared FeCl 2 ) or the iron chelator CDTA for 4 h at 4°C and then immunoprecipitated (IP) with an anti-GFP pAb. The soluble fraction was preincubated with iron or CDTA for 4 h at 4°C. The immunoprecipitated m⌬P2 fraction and the preincubated soluble fraction were mixed for 2 h at 4°C and then analyzed by immunoblotting with an anti-PCBP2 mAb. The supernatant (Sup) after mixing the m⌬P2 fraction and the soluble fraction was collected and analyzed by immunoblotting with anti-PCBP2 mAb. The results are typical of at least three independent experiments. membrane-bound PCBP2 was removed. This fraction is referred to as m⌬P2 (Fig. 4A). Importantly, Na 2 CO 3 treatment efficiently removes certain membrane-bound proteins that are not integrally associated with the membrane (14,16,44) Indeed, TfR1 is a transmembrane protein that resides in the membrane fraction (7), and its levels were not affected by Na 2 CO 3 treatment (Fig. 4A). Thus, TfR1 acted as an appropriate control, which demonstrated that Na 2 CO 3 did not lead to the extraction of integral membrane proteins.
To assess whether FPN1 can interact with iron-loaded PCBP2, studies were then performed to implement immunoprecipitation using ␣-GFP followed by immunoblotting using ␣-PCBP2 (Fig. 4B). These experiments utilized the m⌬P2 fraction pretreated with either iron (i.e. FeCl 2 ) or an iron chelator (i.e. 1,2-cyclohexylenedinitrilotetraacetic acid (CDTA) from Wako Chemicals) to deplete this fraction of iron (Fig. 4B). This step was performed to load or deplete the m⌬P2 fraction, particularly FPN1, with iron. Iron or chelator pretreatment did not significantly (p Ͼ 0.05) alter the amount of FPN1 recovered in the m⌬P2 fraction (Fig. 4B, upper panel). The free iron or chelator was then subsequently removed from the membrane by extensive washing with PBS as described previously (14).
The soluble cytosolic fraction containing PCBP2 was also premixed with FeCl 2 or the CDTA iron chelator to load or deplete it of the metal ion, respectively (Fig. 4B, lower panel). Iron or chelator pretreatment did not alter the amount of PCBP2 in the soluble fraction. Then, the m⌬P2 fraction and the soluble cytosolic fraction were mixed, immunoprecipitated with anti-GFP pAb-conjugated protein A beads and immunoblotted using an anti-PCBP2 mAb (Fig. 4C, top panel). The amount of FPN1 recovered by immunoprecipitation with anti-GFP pAb-conjugated protein A beads was not considerably changed among these conditions (Fig. 4C, bottom panel).
When the soluble cytosolic fraction was premixed with iron (PCBP2ϩFe) and then mixed with the chelator-pretreated m⌬P2 fraction (m⌬P2 ϩ CDTA), a substantial amount of PCBP2 was recovered on the immunoblot (Fig. 4C, top panel). However, after the soluble cytosolic fraction was pretreated with iron chelator (PCBP2 ϩ CDTA) and mixed with the chelator-pretreated m⌬P2 fraction (m⌬P2 ϩ CDTA), there was no appreciable co-immunoprecipitation of PCBP2 with FPN1 (Fig.  4C, top panel). This observation suggested that PCBP2 loaded with iron interacted preferably with immunoprecipitable FPN1-GFP in the m⌬P2 membrane fraction relative to PCBP2 depleted of iron.
A markedly lower amount of PCBP2 was detected when the iron-premixed m⌬P2 fraction (m⌬P2 ϩ Fe) was added to the iron-premixed soluble cytosolic fraction (PCBP2 ϩ Fe; Fig. 4C, top panel). Hence, this observation suggested that potentially FPN1 with iron bound (or at least the m⌬P2 ϩ Fe fraction) inhibited the interaction with iron-loaded PCBP2 relative to when m⌬P2 was depleted of iron (Fig. 4C, top panel).
When PCBP2 was depleted of iron using CDTA (PCBP2 ϩ CDTA) and then mixed with the m⌬P2 fraction loaded with iron (m⌬P2 ϩ Fe), no interaction was detected (Fig. 4C, top panel). Hence, unlike PCBP2 loaded with iron, these experiments suggested that when the PCBP2 fraction was depleted of iron, it could not interact at all with the iron-loaded immunoprecipitable FPN1 (m⌬P2 ϩ Fe).
Supernatants were analyzed after mixing the m⌬P2 and the soluble fraction by immunoblotting using an anti-PCBP2 mAb (Fig. 4C, middle panel). When the soluble cytosolic fraction was premixed with iron (PCBP2 ϩ Fe) and then added to the chelator-pretreated m⌬P2 fraction (m⌬P2 ϩ CDTA), less PCBP2 was recovered in the supernatant fraction (Fig. 4C, middle  panel). This observation suggests that iron-loaded PCBP2 in the soluble cytosolic fraction efficiently binds to iron-depleted FPN1, which is then immunoprecipitated with membranebound FPN1. Thus, the recovered PCBP2 in the supernatant fraction is reduced (Fig. 4C, middle panel).
Considering the results described above, and particularly the studies examining the iron loading of FPN1 and the decreased interaction with PCBP2 (i.e. m⌬P2ϩFe 4 PCBP2 ϩ Fe relative to m⌬P2 ϩ CDTA 4 PCBP2 ϩ Fe) (Fig. 4C, top panel), it is suggested that it is difficult to achieve saturation of the ironbinding site of FPN1 by applying iron under these conditions. This may result in some part of FPN1, which remains as an iron-lacking form, binding to iron-loaded PCBP2 (Fig. 4C, top  panel). Such an interaction could explain the limited binding we observed. However, these data are consistent with the hypothesis that when FPN1 binds iron, there is a conformational change or alteration that prevents or diminishes the binding of iron-bound PCBP2.
Collectively, the results shown in Fig. 4 suggest that irondepleted FPN1 interacts with iron-loaded PCBP2 but not with iron-depleted PCBP2. The depletion of iron from the PCBP2 fraction markedly decreased its FPN1 binding activity (Fig. 4C,  top panel). Conversely, the association of iron with the PCBP2 fraction markedly increased its FPN1 binding activity. These observations agree with the hypothesis that the FPN1-PCBP2 interaction is iron-dependent and the transfer of iron from PCBP2 to FPN1 may occur via direct binding of FPN1 to PCBP2.

PCBP2 Directly Transfers Iron to FPN1
NTBI Iron-loading Studies-In general, there are two major pathways of iron uptake by mammalian cells: (i) the uptake of NTBI by DMT1A-I, which is localized to the plasma membrane; and (ii) the uptake of TBI via Tf-TfR1 endocytosis and subsequent transmembrane transport of Tf-derived iron by DMT1A-II/DMT1B-II, which are localized to recycling endosomes (7,15,16). In our previous report, we demonstrated that PCBP2 is indispensable for iron import through DMT1 via both pathways in HEp-2 cells (14). Therefore, we attempted to determine whether PCBP2 is involved in iron export through FPN1.
To investigate the function of PCBP2 in conjunction with FPN1, the ferritin and TfR1 expression levels and intracellular iron levels were analyzed (Figs. 5 and 6). It is well known that high levels of iron lead to increased ferritin and decreased TfR1 expression, whereas the opposite occurs under cellular iron depletion (7). Many cell types, including HEp-2 cells, internalize NTBI (e.g. FAC or ferric nitrilotriacetate (Fe-NTA)) resulting in iron loading (7,44). In fact, their efficiency in incorporating NTBI is enhanced by the expression of DMT1A-I on the plasma membrane (45). It is notable that the expression level of endogenous FPN1 in HEp-2 cells was below the limit of detection of immunoblotting (Fig. 5A).
Incubation of HEp-2 cells stably expressing DMT1A-I with NTBI (i.e. FAC or Fe-NTA) for 6 h/37°C significantly increased the intracellular iron level (Fig. 5). This was demonstrated indirectly by the finding that TfR1 expression was significantly (p Ͻ 0.05) decreased and that ferritin expression was markedly and significantly (p Ͻ 0.001) increased under these conditions (Fig.  5, A and B). Furthermore, it was also assessed directly by quantitating the cellular iron levels, which were markedly and significantly (p Ͻ 0.001) increased after the loading period of 6 h at 37°C (Fig. 5C).
Upon incubation of cells for an additional 18 h at 37°C after NTBI had been withdrawn, TfR1 expression remained low and ferritin expression remained high (Fig. 5, A and B). In agreement with this observation, the intracellular iron level was not significantly (p Ͼ 0.05) decreased under these conditions (Fig.  5C). However, transfection of exogenous FPN1 into the NTBI-preloaded cells (confirmed by immunoblotting using an anti-FPN1 pAb (Fig. 5A)) markedly increased TfR1 (p Ͻ 0.05) and decreased ferritin (p Ͻ 0.001) (Fig. 5, A and B). When GFP as a control was transfected into the NTBI-preloaded cells (confirmed by immunoblotting using an anti-GFP pAb (Fig. 5A)), TfR1 expression remained low and ferritin expression remained high (Fig. 5, A and B). In agreement with these observations, the intracellular iron level remained high after GFP transfection (Fig. 5C). These changes in expression were consistent with the known ability of FPN1 to export iron out of cells to induce iron deficiency (29). This hypothesis was strongly supported by the significant (p Ͻ 0.05) decrease in the intracellular iron level after transfection of exogenous FPN1 (Fig. 5C). These results indicated that exogenously expressed FPN1 functioned properly as an iron exporter.
Considering these observations, to investigate the role of PCBP2 in iron export through FPN1, we then silenced PCBP2 expression using a specific siRNA (Fig. 6). The FPN1 expression plasmid and PCBP2 siRNA were transfected into NTBIpreloaded cells, and after incubation for 18 h at 37°C in culture medium lacking iron, protein expression levels and intracellular iron levels were analyzed (Fig. 6). In the non-siRNA(Ϫ)/FPN1 or negative control-siRNA (N)/FPN1transfected cells, the intracellular iron level was notably decreased, TfR1 expression was increased, and ferritin expression was decreased compared with the iron-replete, NTBI-preloaded cells incubated for 0 h (Fig. 6). These observations were considered to result from iron depletion in the cells by exogenously expressed FPN1.
When PCBP2 expression was silenced by specific PCBP2 siRNA (P2) in the FPN1-transfected cells, its levels were decreased by 80%, TfR1 was decreased by 70%, and ferritin was increased by 5-fold compared with their respective expression levels in non-siRNA(Ϫ)/FPN1 or negative control-siRNA (N)/ FPN1-transfected cells (Fig. 6, A and B). The return of TfR1 and ferritin expression to almost the same levels as those observed in cells incubated for 0 h was attributed to PCBP2 silencing, which probably prevented iron export via the PCBP2-FPN1 interaction (Fig. 6, A and B). Importantly, the expression level of recombinant FPN1 was not altered by PCBP2 silencing (Fig. 6,  A and B).
When GFP as a control was transfected into the NTBIpreloaded cells, TfR1 expression remained low and ferritin expression remained high in three conditions, namely non-siRNA(Ϫ)/GFP, negative control-siRNA(N)/GFP, and PCBP2-siRNA(P2)/GFP (Fig. 6, A and B). In agreement with these observations, the intracellular iron level remained high after GFP transfection (Fig. 6C).
The intracellular iron level in the PCBP2-siRNA/FPN1transfected cells was 4-fold higher than that of the negative control-siRNA(N)/FPN1-transfected cells (Fig. 6C). Hence, the iron export activity of FPN1 was markedly reduced by PCBP2 silencing. These results suggest that PCBP2 functions to provide the iron supplied to the cells as NTBI to be exported out via FPN1.
Holo-Tf Iron-loading Studies-Next, we investigated the function of PCBP2 in conjunction with FPN1 when holo-Tf was used to load cells with iron (Fig. 7). Using a protocol similar to that implemented with NTBI loading (Figs. 5 and 6), very sim-ilar results were obtained utilizing holo-Tf as an iron donor (Fig. 7, A-C). That is, when cells were incubated without holo-Tf for 12 h at 37°C, they expressed TfR1, whereas ferritin was not detected, suggesting a relative cellular iron deficit (Fig.  7A). In contrast, upon loading the cells with iron for 12 h at 37°C with holo-Tf, there was a decrease in TfR1 expression and an increase in ferritin expression, denoting iron loading (Fig.  7A). This conclusion was confirmed by the direct determination of cellular iron levels (Fig. 7C). Then, upon an additional incubation for 18 h at 37°C, this iron-loading phenotype was maintained in the absence of FPN1 transfection, as shown by TfR1 and ferritin expression (Fig. 7, A  and B). Again, the maintenance of iron loading was confirmed by the direct estimation of cellular iron levels (Fig. 7C). In contrast, when the cells were transfected with FPN1 and then incubated for 18 h at 37°C, TfR1 increased (p Ͻ 0.05) and ferritin decreased (p Ͻ 0.001), suggesting cellular iron depletion via FPN1 (Fig. 7A). The decrease in the cellular iron levels was confirmed by direct estimation, showing a significant (p Ͻ 0.01) decrease in iron levels upon transfection with FPN1 (Fig. 7C).
When GFP as a control was transfected into the holo-Tfpreloaded cells, TfR1 expression remained low and ferritin expression remained high (Fig. 7, A and B). In agreement with this observation, the intracellular iron level remained high after GFP transfection (Fig. 7C). Collectively, these studies showed that holo-Tf loaded iron to the cells and that their transfection with FPN1 led to an iron deficiency.
Considering the results shown in Fig. 7, A-C, our studies then assessed the effect of silencing PCBP2 on cellular iron levels using this holo-Tf iron-loading system (Fig. 7, D-F). When PCBP2 expression was silenced using specific PCBP2 siRNA (P2) in FPN1-transfected cells, TfR1 was significantly (p Ͻ 0.05) decreased by 60%, whereas ferritin was significantly (p Ͻ 0.001) increased by 8-fold compared with their respective expression in non-siRNA(Ϫ)/FPN1 or negative control-siRNA(N)/FPN1-transfected cells (Fig. 7, C and D). In fact, both TfR1 and ferritin expression levels returned to nearly the same levels as those observed in the cells incubated for 0 h (Fig. 7D). Furthermore, the intracellular iron level in PCBP2-siRNA(P2)/ FPN1-transfected cells was 3-4 times higher than that in non-siRNA(Ϫ)/FPN1 or negative control-siRNA(N)/FPN1transfected cells (Fig. 7F). Hence, silencing PCBP2 resulted in the iron export activity of FPN1 being markedly reduced.
When GFP as a control was transfected into the holo-Tfpreloaded cells, TfR1 expression remained low and ferritin expression remained high under all three conditions, namely non-siRNA(Ϫ)/GFP, negative control-siRNA(N)/GFP, and PCBP2-siRNA(P2)/GFP (Fig. 7, D and E). In agreement with these observations, the intracellular iron level remained high after transfection with GFP (Fig. 7F).
Collectively, the results shown in Figs. 5-7 suggest that PCBP2 functions to provide iron to FPN1 prior to FPN1-dependent iron export. This occurs irrespective of whether cellular iron is initially supplied in the form of NTBI (Figs. 5 and 6) or holo-Tf (Fig. 7). Therefore, PCBP2 transfers iron to FPN1 to enable transmembrane export of iron.

Discussion
In this study, we investigated the interactions between FPN1 and the PCBP iron chaperones and found a specific interaction between FPN1 and PCBP2. It was reported that the KH domain of the PCBPs could interact with RNAs, DNAs, or proteins (12,13). To analyze which domain in the PCBPs was important for the interaction between FPN1 and PCBPs, we focused on the three types of KH domains in these proteins using yeast twohybrid assays. Bait and prey expression in yeast cells were confirmed by immunoblotting. The results obtained from the yeast two-hybrid assays showed that the C-terminal cytoplasmic region of FPN1 could bind to PCBP2, especially the KH2 domain of PCBP2 (Fig. 1).
Significantly, we did not detect any specific binding of FPN1 to members of the PCBP family aside from PCBP2. Hence, the specific membrane transporter binding activity of PCBP2 was revealed in this investigation. This specificity relative to the other members of the PCBP family could be due to differences in the amino acid sequences of the KH2 domains of the PCBPs (13). It is also possible that each member of the PCBP family may interact with its own specific target molecule in specific subcellular locations and/or cell types to perform specific functions.
To examine whether the N-terminal cytoplasmic region of DMT1 or the C-terminal cytoplasmic region of FPN1 was important for PCBP2 binding, we made deletion mutants (DMT1⌬N and FPN1⌬C) and then examined their intracellular distribution in mammalian cells. DMT1⌬N, lacking the N-terminal amino acids 1-98 of DMT1A-I, was localized to the plasma membrane and ER. DMT1B-I, lacking the N-terminal amino acids 1-29 of DMT1A-I, was localized to late endosomes/lysosomes (18,40,41). In a previous study, it was revealed that mutations in N-terminal cytoplasmic region of DMT1 caused mislocalization in the cells (37). Furthermore, FPN1⌬C was localized to the plasma membrane and also the ER. Therefore mislocalization of these proteins must be due to mutations in their cytoplasmic regions, which might induce mis-sorting of the molecules.
Notably, PCBP2 exists abundantly in the cytosol, and all four DMT1 isoforms can bind to PCBP2 regardless of the different distribution of the transporters (14). Therefore, the PCBP2-binding domains in the cytoplasmic regions could bind to PCBP2 even though they were localized to various intracellular membrane structures such as endosomes, lysosomes, and the ER.
In mammalian cells, both DMT1⌬N and FPN1⌬C completely lost their PCBP2 binding activity (Fig. 3C). Importantly, these deletion mutants remained localized to the plasma membrane (Fig. 3B). Irrespective of their plasma membrane location, and partial co-localization with the respective WT transporters (Fig. 3B), they still lost their PCBP2 binding activity (Fig.  3C) and iron transport activity (Fig. 3D). These data indicate that the reason for losing the iron transport activity in mammalian cells was the loss of the DMT1-PCBP2 or FPN1-PCBP2 interaction and not that these transporters became totally mislocalized.
We reported previously that mutants in the cytoplasmic regions of DMT1 still function as iron transporters in yeast cells (37), and the transmembrane-domain structure must be important for their iron transport activity itself (46). Thus, collectively, the specific PCBP2 binding activity of the cytoplasmic regions of each transmembrane iron transporter has been confirmed in both yeast and mammalian cells.
Using the UniProt Knowledgebase, we did not detect any amino acid sequence homology between the N-terminal cytoplasmic region of DMT1 and the C-terminal cytoplasmic region of FPN1. It remains unclear how PCBP2 binds to these specific regions of DMT1 or FPN1. We investigated whether FPN1 or PCBP2 requires iron for the FPN1-PCBP2 interaction. After iron was removed from a PCBP2-containing fraction, neither iron-loaded FPN1 nor iron-depleted FPN1 could bind to PCBP2 (Fig. 4). Intriguingly, iron-depleted FPN1 strongly bound to iron-loaded PCBP2 (Fig. 4). In the case of DMT1, iron-depleted DMT1 did not bind to PCBP2 at all (14), but iron-loaded DMT1 did bind to PCBP2. Therefore, it is apparent that PCBP2 can bind to both DMT1 and FPN1 but that the effects and properties of each interaction are very different. The lack of similarity between the PCBP2binding regions of these two transporters suggests a difference in their affinity for PCBP2. Hence, further structural analyses of iron-loaded and -depleted PCBP2 and these transporters are necessary but were beyond the scope of the current study.
We have shown that PCBP2 is essential for iron uptake via DMT1 (14). When PCBP2 expression was silenced, cells could not import iron. The process by which PCBP2 receives iron from DMT1 suggests a model in which PCBP2 disengages from DMT1 and subsequently transports iron to another acceptor. This could occur potentially through a subsequent interaction with a downstream acceptor such as ferritin or FPN1. Moreover, the direct delivery of iron from DMT1 to PCBP2 probably occurs while they are bound to each other. This integrated process could avoid the generation of hazardous free iron in the cytoplasm. Accordingly, we hypothesized that certain iron chaperones deliver iron directly to FPN1.
There are two forms of iron that can be imported by cells: NTBI and TBI (7). The NTBI is predominately relevant for iron-loading diseases such as hemochromatosis and ␤-thalassemia, whereas TBI is the physiologically important iron donor to tissues (7). The iron export activity of FPN1 was inhibited by PCBP2 silencing regardless of the pathway by which iron was supplied, suggesting that the pathways of iron uptake from NTBI or TBI merge intracellularly. As free iron is extremely hazardous to life because of its ability to induce reactive oxygen species production, it cannot be liberated freely into the cytoplasm. Thus, the task of each iron chaperone molecule is probably crucial to direct iron appropriately irrespective of whether it is supplied as NTBI or TBI. It can be further speculated that the role of each iron chaperone is likely to be somewhat unique in iron metabolism, as found for the metabolism of copper (47).
Considering this latter suggestion, emerging evidence indicates that PCBP1 and PCBP2 deliver iron not only to the iron storage protein, ferritin, but also directly to some enzymes that contain mono-and dinuclear iron centers (48). For instance, the prolyl and asparaginyl hydroxylases that modify the important transcriptional regulator hypoxia-inducible factor 1␣ (HIF-1␣) accept iron from PCBPs (49). Similarly, deoxyhypusine hydroxylase, which is required for the modification of eukaryotic translation initiation factor 5A (eIF5A), is also a downstream target of these iron chaperones (48). These functions, together with our findings demonstrating that PCBP2 acts as a gateway keeper for cellular iron uptake (14) and also for cellular iron efflux via FPN1, should now focus attention on other possible molecular interactions of PCBPs. For instance, it is conceivable that PCBPs could interact with other iron-transporting transmembrane proteins such as ZIP14 (50) and mito-ferrin (51). Hence, the participation of different PCBP family members in such processes needs to be clarified.
Another challenge in understanding the biology of the PCBPs will be to explain how functions in iron transport overlap with their described nuclear and cytoplasmic control over gene expression (52). For example, Pcbp2 Ϫ/Ϫ mice demonstrate that PCBP2 has a non-redundant role in vivo during hematopoiesis, which is attributed to its post-transcriptional control over ␣-globin gene expression (53). However, another possibility regarding this latter finding emerges from our current and previous studies regarding PCBP2 function (14). That is, the potential ability of PCBP2 to control Tf iron uptake by DMT1 and also iron release by FPN1 in erythroid precursors may possibly affect erythropoiesis, and these results could explain the non-redundant role of PCBP2 in hematopoiesis.
In conclusion, PCBP2 acts as a recipient of iron from DMT1 and as a donor of iron to FPN1; both transmembrane transporters function as intracellular receptors for the iron chaperone PCBP2 but not PCBP1, PCBP3, or PCBP4. The role of PCBP2 as a crucial molecular partner in the release of iron from cells via FPN1 provides another link to understanding the complex protein network involved in cellular iron metabolism.

Experimental Procedures
Antibodies and Reagents-A mouse anti-human ␣-tubulin monoclonal antibody (mAb) was purchased from BD Transduction Laboratories (Franklin Lakes, NJ); a mouse anti-human PCBP2 mAb was purchased from Abnova (Taipei, Taiwan, catalog No. H00005094-M07); a goat anti-human PCBP1 polyclonal antibody (pAb) was purchased from Acris (Herford, Germany, catalog No. AP22434PU-N); a rabbit anti-human PCBP3 pAb was purchased from Atlas Antibodies (Stockholm, Sweden, catalog No. HPA030247); a rabbit anti-human PCBP4 pAb was purchased from Novus Biologicals (Minneapolis, MN, catalog No. NBP1-76729); a rabbit anti-human ferritin pAb was purchased from Santa Cruz Biotechnology (Santa Cruz, CA, catalog No. sc-25617); a rabbit anti-human PDI mAb was purchased from Cell Signaling Technology (Boston MA, catalog No. 3501); an anti-human LAMP2 mAb (H4B4, developed by Drs. J. E. K. Hildreth and J. T. August) was obtained from the Developmental Studies Hybridoma Bank, University of Iowa; a mouse anti-human c-Myc mAb was purchased from Cell Signaling Technology (catalog No. 2276); a mouse anti-HA mAb was purchased from Cell Signaling Technology (catalog No. 2367). We previously raised a rabbit polyclonal antibody against GFP (16,17). A mouse anti-TfR1 mAb (N-2) was prepared as described previously (54).
A rabbit anti-human FPN1 pAb was raised against amino acids 220 -307 of FPN1 for the present study. Briefly, the affinity-purified anti-FPN1 pAb was purified from antiserum on a HiTrap TM N-hydroxysuccinimide-activated column (GE Healthcare Life Sciences) coupled with maltose-binding protein-tagged FPN1-(220 -307), which was produced in E. coli strain JM109 and purified on a column of amylose resin. All animal experiments were approved by the Animal Ethics Committee of Kawasaki Medical School.
Cell Culture-HEp-2 cells were maintained in DMEM supplemented with 10% FBS. Cells were transfected with GFPtagged full-length and deletion mutant forms of DMT1 and FPN1 or the mCherry-tagged full-length form of CPR by Lipofectamine 3000 (Thermo Fisher Scientific).
Protein Extraction and Immunoblot Analysis-Whole cell lysates were extracted using TNE buffer (10 mM Tris-HCl (pH 7.4), 150 mM NaCl, and 2% (w/v) Nonidet P-40) and immunoblotting performed by standard methods as described previously (55). Briefly, equal amounts of protein (15 g), loaded and separated on 10 or 12.5% SDS-PAGE, were then transferred to PVDF membranes. The membranes were incubated with primary antibodies (see "Antibodies and Reagents" above) at 4°C overnight and then incubated with HRP-conjugated secondary antibodies (as detailed above) at room temperature for 1 h.
Immunoprecipitation Protocol-Whole cell lysates or membrane fractions solubilized in TNE buffer were incubated with anti-GFP pAb conjugated to protein A-Sepharose beads (GE Healthcare Life Sciences) and incubated at 4°C overnight. The beads were recovered by centrifugation and washed three times with TNE buffer. Proteins were extracted from the beads and then analyzed by immunoblotting (14).
Vectors and Plasmid Constructs-To express human DMT1A-I or FPN1, each cDNA was cloned into a pEGFP-N1 vector. The deletion mutant forms of DMT1 and FPN1 were generated via PCR-mediated mutagenesis using KOD-Plus DNA polymerase (Toyobo, Osaka, Japan). To express human CPR, the cDNA was cloned into the pRSET-B-mCherry vector (kindly provided by Dr. Roger Y. Tsien, University of California, San Diego). The amino acid coding regions of human PCBP1, PCBP2, and PCBP4 were amplified using total RNA from human HEp-2 cells, and the amino acid coding region of PCBP3 was amplified using total RNA from human placenta.
Yeast Two-hybrid Assay-The yeast two-hybrid assay was performed as described by the manufacturer (Clontech, Mountain View, CA). For bait plasmids, the N-or C-terminal cytoplasmic region or the intracellular loop between the sixth and seventh transmembrane domains of FPN1 (Fig. 1) (Fig. 1A).
Briefly, yeast cells were cultured in yeast nitrogen base with 2% glucose and the appropriate amino acids at 30°C until A 600 ϭ 1.0 -3.0. Yeast cells were harvested and treated with 1.0 M lithium acetate for 10 min at room temperature and then treated for 5 min with 0.4 M NaOH on ice. These pretreated yeast cells were boiled in SDS-sample buffer for 3 min. The supernatant was collected and analyzed by immunoblotting.
Confocal Microscopy-Cells grown on glass coverslips were fixed with 4% paraformaldehyde in PBS for 15 min at room temperature and permeabilized with 0.1% Triton X-100 in PBS for 15 min. The coverslips were washed and blocked in 0.1% fish skin gelatin in PBS. Cells were incubated with primary antibodies for 60 min at room temperature. Coverslips were washed with 0.1% fish skin gelatin in PBS. Secondary antibody coupled to Alexa 594 was incubated with cells for 60 min at room temperature. Coverslips were washed and mounted on slides with Vectashield (catalog No. H-1000, Vector Laboratories, Burlingame, CA). The images were obtained by using a Zeiss LSM 700 confocal laser-scanning microscope system (Jena, Germany) (18).
Membrane Fractionation-HEp-2 cells were grown to confluence in 100-mm dishes. Cells were washed twice with PBS, removed from the plate in homogenization buffer (0.25 M sucrose, 10 mM Tris-HCl (pH 7.4), and protease inhibitor mixture), and homogenized in a Teflon-coated glass Potter homogenizer rotating at 1000 rpm for 15 strokes. The homogenate was then centrifuged at 1,000 ϫ g for 10 min at 4°C to obtain the postnuclear supernatant, which was centrifuged at 10,000 ϫ g for 20 min/4°C to obtain the postmitochondrial fraction. Then, the postmitochondrial fraction was centrifuged at 100,000 ϫ g for 30 min at 4°C to obtain the membrane fraction as described previously (14,17). The membrane fraction was resolved in 250 l of TNE buffer.
To remove PCBP2 from the membrane fraction, this fraction was resuspended in homogenization buffer containing 0.1 M Na 2 CO 3 (pH 11.0). The suspension was incubated on ice for 30 min and then centrifuged at 100,000 ϫ g for 30 min at 4°C. The resulting precipitate, consisting of PCBP2-depleted cell membranes, is termed the "m⌬P2 fraction" (14,16). Protein concentrations were determined using Lowry's method (57).
In Vitro FPN1 and PCBP2-binding Assay: Loading or Depleting the m⌬P2 Fraction of Iron-The m⌬P2 fraction was solubilized in TNE buffer and mixed with the anti-GFP pAb conjugated to protein A-Sepharose beads as described previously (14). Then, 100 M FeCl 2 (Wako Chemicals; freshly prepared at time of use) or the chelator CDTA (100 M, Wako Chemicals) was added to the mixture to load or deplete the fraction of iron, respectively. The beads were washed three times with PBS and then mixed with the soluble cytosolic fraction containing 100 M FeCl 2 or 100 M CDTA for 2 h at 4°C. The samples were washed again and analyzed by immunoblotting.
Stealth RNAi siRNA-Stealth RNAi siRNA targeting a 25-nucleotide sequence of PCBP2 was purchased from Thermo Fisher Scientific. HEp-2 cells were transfected with 10 nM Stealth RNAi using Lipofectamine RNAiMAX (Thermo Fisher Scientific) and incubated for 18 h at 37°C. The Lipofectamine RNAiMAX:siRNA ratio was 1.0 l:7.5 pmol. The sequence of PCBP2 Stealth RNAi was: 5Ј-CCGGUGUGAUU-GAAGGUGGAUUAAA-3Ј. A Stealth RNAi with similar guanine-cytosine (GC) content (Stealth RNAi TM siRNA negative control medium-GC duplex No. 2) was purchased from Thermo Fisher Scientific (catalog No. 12935-112) and used as a control.
Iron Export Assay-Two procedures were used to load cellular iron; one method used NTBI and the other TBI. To load NTBI into cells, HEp-2 cells stably expressing DMT1A-I-GFP were cultured in serum-free DMEM supplemented with 100 M FAC or 100 M Fe-NTA (Fe:NTA ratio, 1:4) for 6 h at 37°C.
To load TBI into cells, HEp-2 cells were cultured in serumfree DMEM supplemented with 0.3 mg/ml (3.75 M) holo-Tf (Calbiochem) for 12 h at 37°C. Cells were washed with PBS and then transfected with FPN1-GFP using Lipofectamine 3000 and with PCBP2 siRNA using Lipofectamine RNAiMAX (Thermo Fisher Scientific). Cells were incubated in serum-free RPMI 1640 medium (Thermo Fisher Scientific) for 18 h at 37°C. Then, the cells were washed with PBS containing 1 mM EDTA and lysed in TNE buffer for immunoblotting analysis or in 50 mM NaOH for the measurement of iron concentration.
Ferrozine-based Iron Assay-To directly quantify the intracellular iron concentration, we performed a ferrozine-based iron assay. The detailed method was described previously (58). Briefly, cells were lysed in 50 mM NaOH and then treated with acidic KMnO 4 to release iron from proteins. The samples were then incubated in an iron detection agent (6.5 mM ferrozine, 6.5 mM neocuproine, 2.5 M ammonium acetate, and 1 M ascorbic acid dissolved in water). A colorimetric measurement was performed by calculating the absorbance of each sample relative to the absorbance of FeCl 3 standards at a range of concentrations.
Statistics-The density of the specific bands from immunoblotting was measured with a computer-assisted imaging analysis system (NIH ImageJ) and normalization was to total protein loaded. All statistical analysis was performed using Student's t test and considered significant when p Ͻ 0.05. All results are typical of at least three experiments.
Author Contributions-I. Y. conducted most of the experiments, analyzed the results, and wrote the paper. D. R. R. suggested experiments and also was involved in analyzing results and writing the manuscript. K. I. conducted experiments for the binding assay in yeast. F. K. conceived the study and wrote the paper in association with I. Y.