Endosomal recycling of the mammalian iron transporter DMT1 is assumed to be important for efficient and rapid uptake of iron across the endosomal membrane in the transferrin cycle. Here, we show that the retromer, a complex that mediates retrograde transport of transmembrane cargoes from endosomes to the trans-Golgi network, is required for endosomal recycling of DMT1-II, an alternative splicing isoform of DMT1. Bacterially expressed Vps26-Vsp29-Vsp35 trimer, a retromer cargo recognition complex, specifically binds to the cytoplasmic tail domain of DMT1-II in vitro. In particular, this binding is dependent on a specific hydrophobic motif of DMT1-II, which is required for its endosomal recycling. DMT1-II colocalizes with the Vps35 subunit of the retromer in TfR-positive endosomes. Depletion of the retromer by siRNA against Vps35 leads to mis-sorting of DMT1-II to LAMP2-positive structures, and expression of siRNA-resistant Vps35 can rescue this effect. These findings demonstrate that the retromer recognizes the recycling signal of DMT1-II and ensures its proper endosomal recycling.

The steady-state localization of transmembrane (TM) proteins in the endocytic system is a consequence of many sorting events that occur at various points throughout the endosomal pathway. The sorting decisions are governed by a complex system of sorting signals in the itinerant proteins and molecular machineries that recognize those signals and deliver the proteins to their intended destinations (Bonifacino and Traub, 2003; Seaman, 2008). In general, sorting signals are located in the cytoplasmic domain of TM proteins, and some TM proteins have two or more sorting signals in their individual cytoplasmic domains, which allow them a more complex and unique itinerary in the endosomal pathway. For instance, the cation-independent mannose-6-phosphate receptor (CI-MPR), which sorts lysosomal hydrolase precursors to lysosomes, shuttles between the trans-Golgi network (TGN) and endosomes by the actions of several different sorting mechanisms (Ghosh et al., 2003; Seaman, 2005; Bonifacino and Hurley, 2008). At the TGN, CI-MPR binds lysosomal acid hydrolase and is then recognized by the Golgi-associated, γ-adaptin-ear-containing ARF-binding (GGA) proteins, via the acidic cluster dileucine signal located in its cytoplasmic domain, and sorted into clathrin-coated vesicles for delivery to the endosomes (Puertollano et al., 2001; Takatsu et al., 2001; Zhu et al., 2001). After delivery to the endosomes, CI-MPR dissociates from its ligand and is recognized by the retromer complex, via the aromatic motif located in its cytoplasmic domain, and is then retrieved to the TGN for further rounds of lysosomal hydrolase sorting (Arighi et al., 2004; Carlton et al., 2004; Seaman, 2004; Seaman, 2007). In addition to the sorting signals for GGAs and the retromer, the cytoplasmic domain of CI-MPR has several other sorting signals for substances such as PACS1, TIP47 and the clathrin adaptor complexes AP1 and AP2 (Ghosh et al., 2003). Thus, intracellular traffic of CI-MPR is regulated by several different sorting signals coupled with each sorting mechanism, which ensures its proper recycling between the TGN and the endosomes. Unlike single-pass TM proteins such as CI-MPR, the multipass TM proteins, such as nutrient transporters, have several cytoplasmic domains that make it more difficult to understand the molecular mechanisms for their sorting events in the endosomal pathway.

Divalent metal transporter 1 (DMT1 or SLC11A2, formerly called NRAMP2/DCT1) is an integral membrane protein consisting of 12 predicted TM domains with two potential N-glycosylation sites; both of its N- and C-terminal tail domains face the cytosol (Gunshin et al., 1997). DMT1 has been shown to transport a number of divalent metals including Fe2+, Mn2+, Co2+, Cu2+, Ni2+, Pb2+, Zn2+ and Cd2+ by a proton cotransport mechanism (Gunshin et al., 1997). A mutation in Dmt1 (G185R) has been identified in microcytic anemia (mk) mice and Belgrade (b) rats (Fleming et al., 1997; Fleming et al., 1998), which have significant defects in intestinal iron absorption and assimilation of iron by erythroid precursor cells, indicating that the protein has lost the iron transport function in those tissues of the mutant animals. Indeed, DMT1 localizes at the brush border of duodenal enterocytes (Canonne-Hergaux et al., 1999), where it is responsible for dietary iron absorption across the apical plasma membrane, and in the endocytic compartments of cells of peripheral tissues (Su et al., 1998; Gruenheid et al., 1999; Tabuchi et al., 2000; Tabuchi et al., 2002), where it is responsible for the transport across endosomal membranes of iron internalized via the transferrin (Tf)-transferrin receptor (TfR) complex.

Dmt1 produces at least two distinct mRNAs by alternative splicing of two 3′ exons encoding different 3′ untranslated regions (UTRs) and protein products with distinct C-termini (Lee et al., 1998). The DMT1 isoform I (DMT1-I, +IRE) contains an iron-responsive element (IRE) in its 3′ UTR and is expressed at the brush border of duodenal enterocytes, where its expression is induced by dietary iron deprivation (Canonne-Hergaux et al., 1999). DMT1 isoform II (DMT1-II, —IRE) lacks the IRE and encodes a protein that has a different C-terminal 25 amino acid segment instead of the 18 amino acid segment of DMT1-I (see Fig. 1A). DMT1-II is expressed preferentially in nonepithelial cells and is very abundant in reticulocytes. Its expression is induced by the hematopoietic hormone erythropoietin (Canonne-Hergaux et al., 2001). Recently, additional isoforms of Dmt1 mRNA have been identified based on alternative promoter usage at exon 1 (exon1A vs 1B) (Hubert and Hentze, 2002). This alternative promoter usage is assumed to produce a DMT1 protein bearing an additional 29 amino acids (exon1A) upstream of the previously identified start codon of DMT1-I and II (exon1B) (Hubert and Hentze, 2002). The role of these additional residues in the function and targeting of DMT1 has not yet been studied. However, recently, it has been shown that DMT1-I expression from the exon1A promoter is activated by hypoxia-inducible factor signaling induced in the duodenum following an acute iron deficiency (Shah et al., 2009). This indicates that DMT1-I, with an N-terminus derived from exon1A, specifically functions in dietary iron absorption from the apical plasma membrane in the duodenum. These observations suggest that the cytoplasmic tail domains of the N- and C-termini of the DMT1 molecule are substituted by the alternative promoter usage and the alternative splicing, respectively, in a tissue-specific manner, and that these substitutions adapt the DMT1 localization to the tissue-specific iron acquisition (i.e. the dietary iron absorption from the apical plasma membrane in the duodenum and endosomal iron acquisition in the Tf cycle).

Previously, we demonstrated that isoforms DMT1-I and DMT1-II, with their N-termini derived from exon1B, are targeted to distinct endosomal compartments. We also identified critical amino acids in the cytoplasmic tail domain of DMT1-II as a determinant for the isoform-specific localization (Tabuchi et al., 2002). It has recently been shown that this signal is required for endosomal recycling of DMT1-II, and sorting of DMT1-II to the recycling pathway occurs at the endosomes (Touret et al., 2003; Lam-Yuk-Tseung et al., 2005). In contrast to DMT1-II, DMT1-I is not efficiently recycled but is subsequently targeted to LAMP2-positive structures (Tabuchi et al., 2000; Tabuchi et al., 2002; Lam-Yuk-Tseung and Gros, 2006). These data indicate that a putative sorting receptor specifically recognizes the recycling signal of DMT1-II and ensures its endosomal recycling. However, the molecular basis for the sorting of signal-dependent endosomal recycling of DMT1-II has yet to be elucidated.

In this study, we report a molecular mechanism underlying the sorting signal-dependent endosomal recycling of DMT1-II. We found that the retromer is required for proper endosomal recycling of DMT1-II. We discuss the molecular mechanism for retromer-mediated sorting of DMT1-II to the recycling pathway and its role on iron acquisition in the Tf cycle.

Structural requirements for endosomal recycling of DMT1-II

To identify the structural requirements for the recycling signal of DMT1-II, we performed a detailed mutational analysis of the cytoplasmic tail sequence. We expected that this approach might reveal a conserved sequence motif by comparison with known sorting signal sequences and then allow us to identify the molecule(s) required for endosomal recycling of DMT1-II. We constructed various GFP-tagged mutants of DMT1-II, which had deletions or amino acid substitutions within the 25-amino acid sequence specific for DMT1-II, to narrow the region required for endosomal recycling of DMT1-II, and to identify the critical amino acids and the structural requirements (Fig. 1A and listed in Fig. 1G). These constructs were expressed in HEp-2 cells that were immunostained with antibodies against GFP and LAMP2, a marker for organelles of the late endocytic pathway. Their localizations were analyzed by confocal microscopy. To measure the colocalization of DMT1-II mutants with LAMP2, pixel-by-pixel analysis using Pearson's correlation coefficient (Pc) was used and Pc with LAMP2 (PcLAMP2) was calculated. A Pc value of 1 indicates perfect colocalization. During the course of this analysis, we noticed that the localization of these mutants could be divided into three classes based on the comparison between their images and PcLAMP2 counts (typical patterns of their localization and PcLAMP2 counts are shown in Fig. 1B,C): ‘class A’ mutants, similarly to DMT1-I, displayed a predominant localization in the perinuclear region of transfected cells and significantly colocalized with LAMP2 (PcLAMP2≥0.8), indicating a severe defect of endosomal recycling of DMT1-II, as typified by the DMT1-II L557A mutant. Mutants classified as ‘class B’ displayed a dual localization in both the perinuclear region and in punctate structures diffusely scattered throughout the cytoplasm, and a partial colocalization of GFP signals with LAMP2 was observed exclusively in the perinuclear region (0.5<PcLAMP2<0.8), indicating a partial defect of endosomal recycling of DMT1-II, as typified by the DMT1-II Y555A mutant. ‘Class C’ mutants displayed a predominant localization in punctate structures scattered throughout the cytoplasm, and almost no colocalization of GFP signals with LAMP2 (PcLAMP≤0.5) was observed, indicating normal endosomal recycling of DMT1-II, as typified by the DMT1-II wild-type. Thirty cells for each sample were examined and their localizations were scored by the above criteria (Fig. 1D-F).

Deletion analysis of the cytoplasmic tail domain of DMT1-II mapped the most important element for endosomal recycling within the eight amino acid stretch 551-558 of DMT1-II (Fig. 1D,G). Alanine-scanning analysis in this region confirmed that Tyr555 and Leu557 are the crucial amino acids for endosomal recycling of DMT1-II, which is in good agreement with our previous report (Tabuchi et al., 2002). Notably, the L557A mutant displayed a severe mis-sorting of DMT1-II to LAMP2-positive structures, whereas the Y555A mutant displayed a modest mis-sorting (PcLAMP2 L557A vs Y555A=0.86±0.024; n=30 vs 0.71±0.126; n=30, P<0.0001, Fig. 1E), indicating that Leu557 is a more important residue for proper endosomal recycling of DMT1-II than Tyr555. Deletion of the Y555LL sequence led to a complete mis-sorting of DMT1-II to LAMP2-positive structures (PcLAMP2=0.91±0.018; n=30, Fig. 1E), indicating that this mutation drastically reduces the ability of endosomal recycling of DMT1-II. Although each single amino acid substitution to alanine within the Q551PEL sequence did not affect the proper localization of DMT1-II, deletion of this region or substitution of all amino acids to alanine in this region led to a mis-sorting of DMT1-II to LAMP2-positive structures (PcLAMP2= 0.63±0.107; n=30, Fig. 1E). Importantly, deletion analysis indicated that the distance from the TM domain is also important for this signal (Fig. 1E,G). The crucial amino acid Tyr555 must be more than 15 amino acids away from the TM domain (Fig. 1G).

We next examined the structural requirements for the specific amino acid residues present at positions 555 and 557. Substitution of Tyr555 with Ser led to a significant mis-sorting of DMT1-II to LAMP2-positive structures (PcLAMP2=0.86±0.043; n=30, Fig. 1F), whereas substitution by Phe, Trp or Leu did not affect their localizations (Fig. 1F,G), suggesting that any hydrophobic amino acid at position 555, either aromatic or aliphatic, is sufficient for endosomal recycling of DMT1-II. Substitution of Leu557 by any hydrophobic amino acid except for Met led to a striking mis-sorting of DMT1-II to LAMP2-positive structures (Fig. 1F,G). On the basis of the detailed mutational analysis of the cytoplasmic tail domain of DMT1-II in this study, we conclude that a Øx(Leu/Met)x motif (where Ø can be any hydrophobic amino acid and x can be any amino acid) is the recycling signal responsible for proper endosomal recycling of DMT1-II.

Fig. 1.

Mutational analysis of the cytoplasmic tail domain of DMT1-II reveals the structural requirements for endosomal recycling of DMT1-II. (A) Predicted secondary structure of DMT1. DMT1 is an integral membrane protein, which consists of 12 putative transmembrane domains and two potential glycosylation sites. Two isoforms of DMT1 mRNA result from alternative splicing of a single gene product in mammalian cells. The two polypeptides (DMT1-I and DMT-II) share 543 residues on the N-terminal portion but differ primarily in the last 18 or 25 C-terminal residues. (B) Isoform-specific localization of DMT1. DMT1-I localizes in LAMP2-positive structures (class A), whereas DMT1-II localizes in maturing endosomes or recycling endosomes (TfR-positive endosomes) (class C). Some DMT1-II mutants show a dual localization of both TfR-positive endosomes and LAMP2-positive structures (class B). (C) The localization of GFP-tagged DMT1-II was split into three classes. HEp-2 cells expressing GFP-DMT1-I (class A), GFP-DMT1-II Y555A (class B) or GFP-DMT1-II wild-type (class C) were fixed and immunostained with anti-GFP (green) and LAMP2 (red) antibodies. Scale bar: 20 μm. Pixel-by-pixel colocalization analysis yielded a Pearson's correlation coefficient (Pc) with LAMP2 (PcLAMP2). The PcLAMP2 counts in each selected region are indicated. Pc=1 indicates perfect colocalization. (D-F) Quantification of the localization of various mutants of DMT1 in HEp-2 cell. Samples were inspected by confocal microscopy and the PcLAMP2 counts were analyzed (n=30 cells per sample). Error bars represent s.d. The cut-off criteria for each class of localization were determined by the comparison of their images and PcLAMP2 counts. Class A, PcLAMP2≥0.8; Class B, 0.5<PcLAMP2<0.8; Class C, PcLAMP2≤0.5. (G) Summary of the mutational analysis of the cytoplasmic tail domain of DMT1-II. Amino acid sequences of the cytoplasmic tail domain of wild-type DMT1-I and DMT1-II and their various mutants, together with their class of localization. Based on the results from the mutational analysis of DMT1-II, we concluded that the sequence Øx(Leu/Met)x is a structural requirement for a recycling signal of DMT1-II (x indicates any amino acid; Ø indicates any hydrophobic amino acid).

Fig. 1.

Mutational analysis of the cytoplasmic tail domain of DMT1-II reveals the structural requirements for endosomal recycling of DMT1-II. (A) Predicted secondary structure of DMT1. DMT1 is an integral membrane protein, which consists of 12 putative transmembrane domains and two potential glycosylation sites. Two isoforms of DMT1 mRNA result from alternative splicing of a single gene product in mammalian cells. The two polypeptides (DMT1-I and DMT-II) share 543 residues on the N-terminal portion but differ primarily in the last 18 or 25 C-terminal residues. (B) Isoform-specific localization of DMT1. DMT1-I localizes in LAMP2-positive structures (class A), whereas DMT1-II localizes in maturing endosomes or recycling endosomes (TfR-positive endosomes) (class C). Some DMT1-II mutants show a dual localization of both TfR-positive endosomes and LAMP2-positive structures (class B). (C) The localization of GFP-tagged DMT1-II was split into three classes. HEp-2 cells expressing GFP-DMT1-I (class A), GFP-DMT1-II Y555A (class B) or GFP-DMT1-II wild-type (class C) were fixed and immunostained with anti-GFP (green) and LAMP2 (red) antibodies. Scale bar: 20 μm. Pixel-by-pixel colocalization analysis yielded a Pearson's correlation coefficient (Pc) with LAMP2 (PcLAMP2). The PcLAMP2 counts in each selected region are indicated. Pc=1 indicates perfect colocalization. (D-F) Quantification of the localization of various mutants of DMT1 in HEp-2 cell. Samples were inspected by confocal microscopy and the PcLAMP2 counts were analyzed (n=30 cells per sample). Error bars represent s.d. The cut-off criteria for each class of localization were determined by the comparison of their images and PcLAMP2 counts. Class A, PcLAMP2≥0.8; Class B, 0.5<PcLAMP2<0.8; Class C, PcLAMP2≤0.5. (G) Summary of the mutational analysis of the cytoplasmic tail domain of DMT1-II. Amino acid sequences of the cytoplasmic tail domain of wild-type DMT1-I and DMT1-II and their various mutants, together with their class of localization. Based on the results from the mutational analysis of DMT1-II, we concluded that the sequence Øx(Leu/Met)x is a structural requirement for a recycling signal of DMT1-II (x indicates any amino acid; Ø indicates any hydrophobic amino acid).

Structural similarity between the recycling signal of DMT1-II and the retromer recognition sequence of CI-MPR

To find sequence motif(s) that are similar to the recycling signal of DMT1-II, we investigated the known sorting motifs and found a structurally similar sequence motif WLM, which has been recently identified in the CI-MPR tail as a sorting motif required for retromer-mediated endosome-to-TGN retrieval (Seaman, 2007). The retromer was originally identified as an essential protein complex in yeast for the proper retrieval of the vacuolar hydrolase receptor Vps10, a functional analog of mammalian CI-MPR, from endosomes back to the TGN (Seaman et al., 1998), and the mammalian retromer has also been shown to function in endosome-to-TGN retrieval of CI-MPR (Arighi et al., 2004; Carlton et al., 2004; Seaman, 2004). The mammalian retromer comprises a sorting nexin (SNX) dimer composed of four possible combinations of SNX1-SNX5, SNX1-SNX6, SNX2-SNX5 and SNX2-SNX6 (Wassmer et al., 2009), and a cargo-recognition complex composed of Vps26, Vps29 and Vps35 (Bonifacino and Hurley, 2008). Intriguingly, recent reports showed that the retromer also functions in the endosomal recycling of several TM proteins including the yeast iron transporter Fet3-Ftr1 complex (Strochlic et al., 2007) and Wntless, a membrane protein dedicated to the secretion of the morphogen Wnt proteins (Eaton, 2008). The sequence similarity and recent findings of the retromer function in membrane protein recycling prompted us to examine the potential involvement of the retromer in endosomal recycling of DMT1-II. We first swapped the YLL motif in the cytoplasmic tail domain of DMT1-II for WLM and analyzed the localization of this mutant by immunofluorescence confocal microscopy. This substitution did not affect the proper localization of DMT1-II, indicating that the WLM motif is sufficient for endosomal recycling, even in the case of DMT1-II (Fig. 1F,G). Seaman (Seaman, 2007) has also reported the FLV motif in sortilin as a retromer recognition sequence. However, the FLV motif is not sufficient for endosomal recycling in the case of DMT1-II (Fig. 1F,G).

Retromer directly interacts with the recycling signal of DMT1-II in vitro

We next examined whether the recycling signal of DMT1-II could directly bind to the retromer complex. Bacterially expressed glutathione S-transferase (GST)-DMT1-II tail domain (Fig. 2A) was used to pull down a bacterially expressed FLAG-tagged retromer cargo-recognition complex, which consists of 3×FLAG-Vps26-Vps29-Vps35-His6 (Fig. 2B). Since a previous study showed that the cytoplasmic tail domain of CI-MPR interacts with Vps35 using a yeast two-hybrid assay (Arighi et al., 2004), we used the GST-CI-MPR tail (Fig. 2A) as a positive control. Immunoblot analysis showed that both GST-CI-MPR and GST-DMT1-II tails, but not GST alone, efficiently pulled down the recombinant retromer in a dose-dependent manner (Fig. 2C), showing that the cytoplasmic tail domain of DMT1-II can directly bind to the retromer. Strikingly, this interaction was almost completely abrogated by deletion of the YLL sequence from the DMT1-II tail domain (Fig. 2C), which leads to a complete mis-sorting of DMT1-II to LAMP2-positive structures in HEp-2 cells (Fig. 1E,G).

Using the results of the in vivo mutagenesis experiments (Fig. 1G), we next expressed GST-DMT1-II tail fusion proteins with nonfunctional mutations Y555A or L557A, and GST-DMT1-II tail fusion proteins with the functional mutation L557M. These proteins were used for pull-down assays to investigate the binding specificity of the DMT1-II-cytoplasmic tail domain to the retromer. As shown in Fig. 2D, both the GST-DMT1-II tail Y555A and L557A mutant pulled down the retromer less efficiently than the GST-DMT1-II tail wild type (reduced to <65% compared with that of the wild type). Notably, the GST-DMT1-II tail L557M mutant pulled down the retromer more efficiently (increased to more than 125% compared with the wild type), indicating that this L557M mutant has higher affinity to the retromer than the wild type. The binding specificity of the DMT1-II tail domain to the retromer in vitro was well correlated with the structural requirements for the recycling signal of DMT1-II in cultured cells.

DMT1-II colocalization with the retromer in TfR-positive endosomes

We next determined the subcellular localization of DMT1-II and the retromer. HEp-2 cells stably expressing 3×HA-tagged DMT1-II were immunostained with antibodies against HA epitope Vps35 and various organellar markers. A triple-staining experiment showed that DMT1-II colocalized with Vps35 in relatively large punctae, where TfR, a marker for organelles of the early endocytic and recycling pathways, also colocalized with Vps35 (Fig. 3Ae-h, arrows). To clarify the localization of DMT1-II and Vps35 double-positive punctate-structures, antibody against EEA1, a marker for organelles of the early endocytic pathway, was used for triple staining instead of TfR (Fig. 3B). Only the small patches of the DMT1-II and Vps35 double-positive punctate structures overlapped with EEA1, whereas the majority of the double-positive punctate structures were juxtaposed to EEA1 and not colocalized with EEA1 (Fig. 3Be-h, arrows). Furthermore, these double-positive punctate structures colocalized particularly with CI-MPR in the perinuclear region (supplementary material Fig. S1Ae-h, arrows) but not with TGN marker p230 or recycling endosome marker Rab11 (supplementary material Fig. S1B,C). These data showed that DMT1-II colocalizes with the retromer in TfR-positive endosomes, whereas EEA1 is mostly excluded from this configuration. These results support the hypothesis that the retromer might regulate endosomal recycling of DMT1-II at maturing endosomes (see Discussion).

Retromer depletion by RNAi leads to mis-sorting of DMT1-II to LAMP2-positive structures

To examine directly whether the retromer is required for endosomal recycling of DMT1-II, we knocked down VPS35 mRNA using small interfering RNA (siRNA). HEp-2 cells stably expressing 3×HA-tagged DMT1-II were transfected with two types of siRNA oligonucleotide pairs against VPS35: VPS35 siRNA#1 and #2. After two rounds of transfection at 72 hour intervals with each of these siRNAs, the protein expression of Vps35 decreased to undetectable levels in both samples of siRNA-treated cells, which was confirmed by immunoblot analysis using anti-Vps35 antibody (Fig. 4A). The protein expression of DMT1-II was not significantly affected by depletion of Vps35 (Fig. 4A). Notably, an additional band at approximately 40 kDa was observed in Vps35-depleted cells but not in control siRNA-treated cells (Fig. 4A, indicated with a single asterisk).

Fig. 2.

Retromer complex binds to the cytoplasmic tail domain of DMT1-II. (A) Schematic representation of the structure of GST-fusion proteins used for GST pull-down assay. (B) Schematic representation for the bacterial expression and purification of 3×FLAG-tagged retromer complex. (C) Purified recombinant FLAG-tagged retromer complex (consisting of proteins, 3×FLAG-Vps26-Vps29-Vps35-His6) was incubated with purified GST or GST fusion proteins bearing the wild-type tails of CI-MPR or DMT1-II, or the tail with deletion of the YLL recycling motif of DMT1-II, as shown in Fig. 1G (GST-DMT1-IIΔaYLL tail), and then the proteins were pulled down with glutathione-Sepharose 4B beads. Serial dilutions of proteins bound to the beads were resolved by SDS-PAGE. Bound FLAG-tagged retromer components were detected by immunoblotting with antibody to the FLAG-tag, and proteins were visualized by Coomassie blue staining. (D) Recombinant retromer complex was tested for interaction with GST or GST fusion proteins bearing the wild-type tails or mutant tails of DMT1-II, as above. Interaction of the retromer complex with Y555A or L557A mutant tails of DMT1-II, whose full-length proteins are mis-sorted to LAMP2-positive strucures in HEp-2 cells shown in Fig. 1E, was significantly decreased compared with the wild-type tail of DMT-II, whereas interaction of the retromer complex with the L557M mutant tail of DMT1-II, whose full-length protein is normally sorted to TfR-positive endosomes, was notably increased compared with the wild-type tail of DMT1-II. We observed less 3×FLAG-Vps35-His6 protein than 3×FLAG-Vps26, or 3×FLAG-Vps29 proteins, presumably because of an inefficient transfer of the protein to the PVDF membrane or a partial proteolytical digestion of the N-terminal region during protein purification.

Fig. 2.

Retromer complex binds to the cytoplasmic tail domain of DMT1-II. (A) Schematic representation of the structure of GST-fusion proteins used for GST pull-down assay. (B) Schematic representation for the bacterial expression and purification of 3×FLAG-tagged retromer complex. (C) Purified recombinant FLAG-tagged retromer complex (consisting of proteins, 3×FLAG-Vps26-Vps29-Vps35-His6) was incubated with purified GST or GST fusion proteins bearing the wild-type tails of CI-MPR or DMT1-II, or the tail with deletion of the YLL recycling motif of DMT1-II, as shown in Fig. 1G (GST-DMT1-IIΔaYLL tail), and then the proteins were pulled down with glutathione-Sepharose 4B beads. Serial dilutions of proteins bound to the beads were resolved by SDS-PAGE. Bound FLAG-tagged retromer components were detected by immunoblotting with antibody to the FLAG-tag, and proteins were visualized by Coomassie blue staining. (D) Recombinant retromer complex was tested for interaction with GST or GST fusion proteins bearing the wild-type tails or mutant tails of DMT1-II, as above. Interaction of the retromer complex with Y555A or L557A mutant tails of DMT1-II, whose full-length proteins are mis-sorted to LAMP2-positive strucures in HEp-2 cells shown in Fig. 1E, was significantly decreased compared with the wild-type tail of DMT-II, whereas interaction of the retromer complex with the L557M mutant tail of DMT1-II, whose full-length protein is normally sorted to TfR-positive endosomes, was notably increased compared with the wild-type tail of DMT1-II. We observed less 3×FLAG-Vps35-His6 protein than 3×FLAG-Vps26, or 3×FLAG-Vps29 proteins, presumably because of an inefficient transfer of the protein to the PVDF membrane or a partial proteolytical digestion of the N-terminal region during protein purification.

We next examined the localization of DMT1-II in Vps35-depleted cells. We observed that DMT1-II colocalized with TfR but not with LAMP2 in control siRNA-treated cells (Fig. 4Ba-d and i-l, respectively). However, upon depletion of Vps35 by treatment with VPS35 siRNA#1, localization of DMT1-II changed dramatically. We observed almost no colocalization of DMT1-II with TfR in Vps35-depleted cells (Fig. 4Be-h). Instead, DMT1-II colocalized well with LAMP2 (Fig. 4Bm-p). To quantify this effect, we performed pixel-by-pixel colocalization analysis and Pc was calculated. Depletion of Vps35 caused a significant reduction of a PcTfR (Pc with TfR) (control siRNA vs VPS35 siRNA=0.61±0.09; n=50 vs 0.28±0.14; n=50, P<0.0001; Fig. 4C), and conversely a significant increase of PcLAMP2 (control siRNA vs VPS35 siRNA=0.12±0.10; n=50 vs 0.77±0.06; n=50, P<0.0001; Fig. 4D). This analysis statistically confirmed that DMT1-II was mis-sorted to LAMP2-positive structures in Vps35-depleted cells, indicating that endosomal recycling of DMT1-II was perturbed by the loss of retromer function.

Phenotypic rescue of Vps35-depleted cells by expression of siRNA-resistant Vps35

To further confirm the retromer function in endosomal recycling of DMT1-II, we monitored the rescue of this phenotype by transfecting Vps35-depleted cells with GFP alone, or GFP-tagged siRNA-resistant Vps35 wild-type or GFP-tagged nonfunctional mutant Vps35 R107A (Gokool et al., 2007). The expression of the siRNA-resistant Vps35 wild-type in Vps35-depleted cells well restored the mis-sorting of DMT1-II. GFP-Vps35-positive cells in Vps35-depleted cells displayed almost no colocalization of DMT1-II with LAMP2 (Fig. 5Aa′-e′), but significant colocalization of DMT1-II with TfR (supplementary material Fig. S2Aa′-e′). To measure the colocalization of DMT1-II with LAMP2 in both GFP-Vps35-positive and -negative cells, pixel-by-pixel analysis was used and Pc with LAMP2 was calculated. This analysis revealed almost no colocalization of DMT1-II with LAMP2 in GFP-Vps35-positive cells (PcLAMP2=0.24±0.16; n=50; Fig. 5Aa′-e′), but did show significant colocalization in GFP-Vps35-negative cells (PcLAMP2= 0.62±0.10; n=50; Fig. 5Aa″-e″). These results confirmed that expression of GFP-Vps35 rescued the defect of endosomal recycling of DMT1-II in endogenous Vps35-depleted cells.

Fig. 3.

DMT1-II and Vps35 colocalizes in TfR-positive endosomes in HEp-2 cells. HEp-2 cells stably expressing a 3×HA-tagged DMT1-II were fixed, permeabilized, and immunostained with rabbit anti-HA pAb (Aa,Ba), mouse anti-TfR mAb (Ab) or anti-EEA1 mAb (Bb) and goat anti-human Vps35 pAb (Ac,Bc), followed by Alexa Fluor 555-conjugated donkey anti-rabbit IgG, Alexa Fluor 488-conjugated donkey anti-mouse IgG, and Alexa Fluor 633-conjugated donkey anti-goat IgG. Cells were examined by confocal microscopy. Merged images show colocalization of 3×HA-DMT1-II with Vps35 (Ae,Be), 3×HA-DMT1 with TfR (Af) and Vps35 with TfR (Ag) in relatively large puncta (arrows), but almost no colocalization of 3×HA-DMT1-II with EEA1 (Bf) and Vps35 with EEA1 (Bg). Scale bars: 5 μm.

Fig. 3.

DMT1-II and Vps35 colocalizes in TfR-positive endosomes in HEp-2 cells. HEp-2 cells stably expressing a 3×HA-tagged DMT1-II were fixed, permeabilized, and immunostained with rabbit anti-HA pAb (Aa,Ba), mouse anti-TfR mAb (Ab) or anti-EEA1 mAb (Bb) and goat anti-human Vps35 pAb (Ac,Bc), followed by Alexa Fluor 555-conjugated donkey anti-rabbit IgG, Alexa Fluor 488-conjugated donkey anti-mouse IgG, and Alexa Fluor 633-conjugated donkey anti-goat IgG. Cells were examined by confocal microscopy. Merged images show colocalization of 3×HA-DMT1-II with Vps35 (Ae,Be), 3×HA-DMT1 with TfR (Af) and Vps35 with TfR (Ag) in relatively large puncta (arrows), but almost no colocalization of 3×HA-DMT1-II with EEA1 (Bf) and Vps35 with EEA1 (Bg). Scale bars: 5 μm.

In contrast to the wild-type GFP-Vps35, neither GFP alone nor GFP-tagged nonfunctional mutant Vps35 R107A could rescue this phenotype (supplementary material Fig. S2B,C). More than 90% of cells displayed colocalization of DMT1-II with LAMP2 in cells positive for GFP alone or GFP-Vps35 R107A, whereas less than 3% of cells displayed colocalization of DMT1-II with LAMP2 in cells positive for wild-type GFP-Vps35 (Fig. 5B; supplementary material Fig. S2C,D), confirming that the rescue of DMT1-II localization by ectopic expression of siRNA-resistant Vps35 in Vps35-depleted cells is due to the restoration of normal retromer function. Thus, these results demonstrate that the defect of endosomal recycling of DMT1-II in Vps35-depleted cells is a direct consequence of loss of retromer function.

In conclusion, these results strongly suggest that DMT1-II is recognized by a retromer cargo-recognition complex via its recycling signal at maturing endosomes, thereby becoming incorporated into tubular vesicles formed by the retromer and sorting DMT1-II to the recycling pathway.

In this study, we report a molecular mechanism underlying the sorting-signal-dependent endosomal recycling of DMT1-II. First, we identified the structural requirements for the recycling signal of DMT1-II and found the retromer to be a candidate molecule required for endosomal recycling of DMT1-II. Second, we demonstrated that the retromer cargo-recognition complex could directly bind to the cytoplasmic tail domain of DMT1-II in vitro in a recycling-signal-dependent manner. Third, we showed that DMT1-II colocalized with a Vps35 subunit of the retromer in TfR-positive endosomes. Finally, we showed that depletion of Vps35 by RNAi led to a complete mis-sorting of DMT1-II to LAMP2-positive structures, and ectopic expression of siRNA-resistant Vps35 rescued this effect.

Fig. 4.

Mis-sorting of DMT1-II to LAMP2-positive structures upon retromer depletion. (A) Depletion of Vps35 by RNAi is assessed by immunoblotting of each siRNA-treated cell lysate. Tubulin is assumed to be unaffected by the siRNA treatment and can serve as a loading control in this experiment. A single asterisk indicates the protein bands found in VPS35-depleted cells, which presumably correspond to some degradation product of 3×HA-DMT1-II protein. A double asterisk indicates the immature protein bands of 3×HA-DMT1-II (Tabuchi et al., 2002). (B) HEp-2 cells stably expressing a 3×HA-DMT1-II were transfected twice at 72 hour intervals with control siRNA or siRNAs directed against human VPS35. At 24 hours after the second transfection, cells were seeded on coverslips and cultured for an additional 48 hours. Cells were fixed at 72 hours after the second transfection, double-labeled with rat anti-HA mAb (b,f,j,n) and mouse anti-TfR mAb (c,g) or mouse anti-LAMP2 mAb (k,o), followed by Alexa Fluor 555-conjugated goat anti-rat IgG and Alexa Fluor 633-conjugated goat anti-mouse IgG. Cells were examined by confocal microscopy. Scale bar: 10 μm. (C,D) Pixel-by-pixel colocalization analysis yielded a Pearson's correlation coefficient with TfR (PcTfR) or LAMP2 (PcLAMP2). A total of 50 cells were analyzed for each sample. Depletion of Vps35 significantly reduced PcTfR, and inversely increased PcLAMP2. Error bars indicate s.d. *P<0.0001.

Fig. 4.

Mis-sorting of DMT1-II to LAMP2-positive structures upon retromer depletion. (A) Depletion of Vps35 by RNAi is assessed by immunoblotting of each siRNA-treated cell lysate. Tubulin is assumed to be unaffected by the siRNA treatment and can serve as a loading control in this experiment. A single asterisk indicates the protein bands found in VPS35-depleted cells, which presumably correspond to some degradation product of 3×HA-DMT1-II protein. A double asterisk indicates the immature protein bands of 3×HA-DMT1-II (Tabuchi et al., 2002). (B) HEp-2 cells stably expressing a 3×HA-DMT1-II were transfected twice at 72 hour intervals with control siRNA or siRNAs directed against human VPS35. At 24 hours after the second transfection, cells were seeded on coverslips and cultured for an additional 48 hours. Cells were fixed at 72 hours after the second transfection, double-labeled with rat anti-HA mAb (b,f,j,n) and mouse anti-TfR mAb (c,g) or mouse anti-LAMP2 mAb (k,o), followed by Alexa Fluor 555-conjugated goat anti-rat IgG and Alexa Fluor 633-conjugated goat anti-mouse IgG. Cells were examined by confocal microscopy. Scale bar: 10 μm. (C,D) Pixel-by-pixel colocalization analysis yielded a Pearson's correlation coefficient with TfR (PcTfR) or LAMP2 (PcLAMP2). A total of 50 cells were analyzed for each sample. Depletion of Vps35 significantly reduced PcTfR, and inversely increased PcLAMP2. Error bars indicate s.d. *P<0.0001.

Cumulatively, these data argue that the retromer can directly bind and sort the DMT1-II molecule at maturing endosomes (Fig. 6B), thereby ensuring proper endosomal recycling of this molecule.

Retromer cargo-recognition mechanism

Detailed mutational analysis of the cytoplasmic tail domain of DMT1-II revealed the structural requirements for the recycling signal of DMT1-II and identified the Øx(Leu/Met)x motif as its recycling signal. This motif is also required for the binding of the DMT1-II-cytoplasmic tail domain to the retromer in vitro. These data suggest that this motif represents a novel retromer-binding motif. So far, no strong consensus retromer-binding motifs have been identified, and none have been characterized at a structural level (Bonifacino and Hurley, 2008; Collins, 2008). The WLM motif of CI-MPR (Seaman, 2007) fits the structural requirement for the recycling signal of DMT1-II and the swapping of the YLL motif in DMT1-II for the WLM motif did not affect its proper localization in HEp-2 cells, showing that the WLM motif is sufficient for endosomal recycling of DMT1-II. Other known retromer-recognition sequences, such as FLV in sortilin (Seaman, 2007) or FxFxD in yeast DPAP-A (Nothwehr et al., 2000), are structurally similar, but the residues Val and Phe at amino acid position 2 in these sequences do not fit the structural requirement for the recycling signal of DMT1-II (Fig. 6A). Futhermore, the FLV motif is not sufficient for endosomal recycling in the case of DMT1-II. We found that the upstream and downstream sequences of the crucial amino acids Y555LL were also important for the proper recycling of DMT1-II (Fig. 1D,E,G). The upstream and downstream sequences of the FLV motif in sortilin might also be important for the efficient binding of this motif to the retromer.

Fig. 5.

Rescue of DMT1-II localization by expression of siRNA-resistant Vps35. (A) Cells transfected with VPS35 siRNA#1 followed by transfection of plasmid encoding VPS35 siRNA#1-resistant GFP-tagged VPS35 wild type were fixed, immunostained with rabbit anti-GFP pAb, rat anti-HA mAb and mouse anti-LAMP2 mAb followed by Alexa Fluor 488-conjugated goat anti-rabbit IgG, Alexa Fluor 555-conjugated goat anti-rat IgG and Alexa Fluor 633-conjugated goat anti-mouse IgG. Cells were examined by confocal microscopy. Scale bars: 10 μm. PcLAMP2 in GFP-positive cells=0.24±0.16 (n=50), PcLAMP2 in GFP-negative cells=0.62±0.10 (n=50). (B) Quantification of the rescue of DMT1-II mislocalization to LAMP2-positive structures in Vps35-depleted cells after transfection with GFP alone, or siRNA-resistant Vps35 wild type or R107A mutant. Samples were inspected by confocal microscopy and scored for colocalization of DMT1-II with LAMP2. The average counts were calculated from three independent experiments, each analyzing at least 45 cells per condition. Errors bars indicate s.d. *P<0.0001.

Fig. 5.

Rescue of DMT1-II localization by expression of siRNA-resistant Vps35. (A) Cells transfected with VPS35 siRNA#1 followed by transfection of plasmid encoding VPS35 siRNA#1-resistant GFP-tagged VPS35 wild type were fixed, immunostained with rabbit anti-GFP pAb, rat anti-HA mAb and mouse anti-LAMP2 mAb followed by Alexa Fluor 488-conjugated goat anti-rabbit IgG, Alexa Fluor 555-conjugated goat anti-rat IgG and Alexa Fluor 633-conjugated goat anti-mouse IgG. Cells were examined by confocal microscopy. Scale bars: 10 μm. PcLAMP2 in GFP-positive cells=0.24±0.16 (n=50), PcLAMP2 in GFP-negative cells=0.62±0.10 (n=50). (B) Quantification of the rescue of DMT1-II mislocalization to LAMP2-positive structures in Vps35-depleted cells after transfection with GFP alone, or siRNA-resistant Vps35 wild type or R107A mutant. Samples were inspected by confocal microscopy and scored for colocalization of DMT1-II with LAMP2. The average counts were calculated from three independent experiments, each analyzing at least 45 cells per condition. Errors bars indicate s.d. *P<0.0001.

In addition to the known retromer-binding sequences, we found structurally similar sequences F746LLQ and Y532KLT in the cytoplasmic tail domains of human polymeric immunoglobulin receptor (pIgR) and human Wntless, respectively, both of which are known to be retromer cargo proteins (Verges et al., 2004; Yang et al., 2008). Mostov and colleagues have shown that the retromer directly binds to the cytoplasmic tail domain of pIgR and regulates transcytosis of pIgR from the basolateral to the apical surface (Verges et al., 2004). Interestingly, they showed that a pIgR mutant lacking the 30 C-terminal residues of its cytoplasmic domain, which completely lacks the F746LLQ sequence, impaired its retromer-binding ability. Furthermore, the same group recently found that the F746LLQ sequence of pIgR is sufficient to transcytose the pIgR from the basolateral to the apical surface (Luton et al., 2009). These data suggest that the F746LLQ sequence functions as a retromer-binding sequence, which is in good agreement with our prediction. The relative contribution of the Y532KLT to retromer-dependent recycling of human Wntless remains to be determined. Further experiments are needed to define the consensus retromer-binding motif.

Endosomal sorting pathways for DMT1 isoforms

Endosomes are highly dynamic organelles that comprise distinct subdomains occupied by different Rab proteins within the same continuous membrane. For example, during endocytosis of degradative cargos, such as low-density lipoproteins to lysosomes, endosomes containing the cargos gradually mature from early to late stages with a continuous conversion of Rab5 to Rab7 (Rink et al., 2005). It has recently been reported that the retromer cargo-recognition complex is recruited onto the endosomal membrane by the action of GTP-bound Rab7 (Rojas et al., 2008; Seaman et al., 2009). Based on our present results and these recent findings, we assume that the endosomal recycling pathway for DMT1-II is as follows: Rab5-positive endosomes containing DMT1-II gradually mature by recruiting the GTP-bound Rab7, which further recruits the retromer cargo-recognition complex onto its endosomal membrane (Fig. 6B). DMT1-II is subsequently recognized by the retromer cargo-recognition complex via its recycling signal and incorporated into the tubular vesicle formed by the retromer cargo-recognition complex and the BAR-domain-containing proteins SNXs, which stimulate membrane-curvature formation (McMahon and Gallop, 2005). DMT1-II is then transported to the TGN before endosomes are completely matured. At the TGN, DMT1-II is transported via the recycling endosome en route to the plasma membrane. Since DMT1-I cannot be recognized by the retromer, this isoform cannot escape from maturing endosomes and is consequently transported to LAMP2-positive structures.

Although depletion of the retromer by treatment with VPS35 siRNA caused a complete mis-sorting of DMT1-II to LAMP2-positive structures, we did not observe any significant reduction of the DMT1-II protein level in Vps35-depleted cells. Nevertheless, we did observe a faint band at ~40 kDa, which presumably corresponds to a degradation product of DMT1-II (Fig. 3A). In addition, Vps35-depleted cells treated with cycloheximide for 12 hours displayed a DMT1-II protein level that was comparable with that in control cells treated with ethanol for 12 hours (supplementary material Fig. S3), showing that the turnover of DMT1-II protein is very slow at LAMP2-positive structures. These data indicate that a small amount of DMT1-II is degraded in lysosomes, presumably through sorting into a multivesicular body pathway (Katzmann et al., 2002), but the majority of DMT1-II is probably transported to the limiting membrane of LAMP2-positive structures in Vps35-depleted cells. Likewise, DMT1-I must be transported to the limiting membrane of LAMP2-positive structures, where it could function as an iron transporter (Fig. 6B).

Fig. 6.

A model for retromer-dependent endosomal recycling. (A) Representative retromer-binding motifs. (B) Schematic diagram of the role of the retromer for endosomal recycling of DMT1-II in the Tf cycle. DMT1-II and TfR are functionally coupled to operate pH-dependent iron uptake across the endosomal membrane, and both molecules are recycled between the plasma membrane and endosomes. TfR is directly transported to the recycling endosome, whereas DMT1-II is recognized by the retromer cargo-recognition complex and transported through the TGN en route to the recycling endosome; subsequently, both molecules are recycled back to the plasma membrane for another round of the Tf cycle.

Fig. 6.

A model for retromer-dependent endosomal recycling. (A) Representative retromer-binding motifs. (B) Schematic diagram of the role of the retromer for endosomal recycling of DMT1-II in the Tf cycle. DMT1-II and TfR are functionally coupled to operate pH-dependent iron uptake across the endosomal membrane, and both molecules are recycled between the plasma membrane and endosomes. TfR is directly transported to the recycling endosome, whereas DMT1-II is recognized by the retromer cargo-recognition complex and transported through the TGN en route to the recycling endosome; subsequently, both molecules are recycled back to the plasma membrane for another round of the Tf cycle.

Role for retromer-mediated endosomal recycling of DMT1-II in the Tf cycle

In the Tf cycle, DMT1-II and TfR are functionally coupled to operate pH-dependent iron uptake across the endosomal membrane, and both molecules are recycled between the plasma membrane and endosomes (Andrews, 1999; Touret et al., 2003; Andrews, 2008). However, the molecular mechanisms underlying their recycling from endosomes to the plasma membrane are apparently distinct. DMT1-II is directly recognized by the retromer cargo-recognition complex and transported through the TGN en route to the recycling endosome, whereas TfR is directly transported to the recycling endosome by a retromer-independent mechanism (Traer et al., 2007), and subsequently both molecules are recycled back to the plasma membrane for another round of the Tf cycle (Fig. 6B). Since endosomes are highly dynamic organelles forming a continuum of maturing organelles from early to late stages (Bonifacino and Rojas, 2006), the retromer-mediated dynamic recycling system must be important for retention of the iron transporter at TfR-positive endosomes to facilitate the endosomal iron uptake in the Tf cycle, which is particularly important for the production of hemoglobin in mitochondria in erythroid precursors (Fig. 6B). Further studies are required to gain a better understanding of the relative contribution of retromer-mediated endosomal recycling of DMT1-II on endosomal iron uptake in the Tf cycle.

Antibodies and reagents

We used mouse or rat monoclonal antibodies (mAbs) against the following proteins or epitope tags: EEA1, p230 and Rab11 (BD Biosciences); CI-MPR (2G11, Affinity BioReagents™); LAMP2 (H4B4, Developmental Studies Hybridoma Bank); α-tubulin (GE Healthcare); HA tag (mouse mAb HA.11, Covance; rat mAb 3F10, Roche Applied Science); FLAG tag (M2, Sigma); TfR [N2, provided by Tamotsu Yoshimori, Osaka University, Osaka, Japan (Yoshimori et al., 1988)]. A rabbit polyclonal antibody (pAb) to the FLAG tag was obtained from Sigma, and one to the HA tag was obtained from MBL, Japan. A goat pAb to Vps35 was obtained from Imgenex. We raised a rabbit pAb against the GFP. A rabbit was immunized with a maltose-binding protein-GFP fusion protein, and the anti-GFP antiserum was affinity purified using a GST-GFP fusion-protein-immobilized column. Goat anti-rabbit IgG or donkey anti-mouse IgG conjugated with Alexa Fluor 488, goat anti-rat IgG or donkey anti-rabbit IgG conjugated with Alexa Fluor 555, and goat anti-mouse IgG or donkey anti-goat IgG conjugated with Alexa Fluor 633 were purchased from Invitrogen. Horseradish-peroxidase-conjugated anti-mouse and anti-rabbit IgG were purchased from Cell signaling technology. Horseradish-peroxidase-conjugated anti-goat IgG was purchased from R&D systems. Most of the other general reagents were from Wako chemicals, Nacalai Tesque or Sigma.

Recombinant DNA procedures

To introduce the mutations into the cytoplasmic domain of DMT1-II, the pUC13-DMT1-II plasmid (Tabuchi et al., 2002) was used for site-directed mutagenesis to create the deletion or alanine-scanning mutants. Site-directed mutagenesis was performed by the inverse PCR method (Imai et al., 1991). The plasmids carrying mutant constructs were prepared from the Escherichia coli strain SCS110 (Stratagene) and digested with BclI. The BclI fragments containing the mutated constructs of DMT1-II were ligated into the BamHI site of pIRES neo-GFP plasmid (Clontech) to generate the mammalian expression plasmids carrying GFP-tagged DMT1-II mutants.

To generate plasmids encoding GFP or FLAG-tagged VPS26, VPS29 and VPS35, DNA encoding full-length human Vps26, Vps29 and Vps35 was amplified by PCR using KOD plus DNA polymerase (Toyobo). Each amplified fragment was cloned into pDONR221 by BP reaction using Gateway™ technology (Invitrogen) to generate entry clones for VPS26, VPS29, and VPS35. Each entry clone was transferred into pIRES neo-GFP or 3×FLAG-based Gateway vectors by LR reaction to generate pIRES neo-GFP or 3×FLAG-VPS26, -VPS29 and -VPS35 plasmids.

To generate a plasmid encoding VPS35 siRNA#1-resistant wild type of VPS35, pDONR221-siRNAR-VPS35WT, inverse PCR was performed with primers (the forward primer: 5′-GTA GTT AGC TTA CAA GTC TCT-3′; the reverse primer: 5′-GTC TTC ACT AGG CAT GTC TTG-3′; substituted nucleotides without an amino acid change are underlined) corresponding to the sequence of VPS35 siRNA#1 using the pDONR221-VPS35 plasmid as a template, and the PCR product was self-ligated with T4 DNA ligase and T4 polynucleotide kinase (see supplementary material Fig. S4 for further details). The resulting clone was transferred into a pIRES neo-GFP-based Gateway vector by LR reaction to generate a pIRES neo-GFP-siRNAR-VPS35WT plasmid.

To generate a plasmid expressing the GST-CI-MPR tail fusion protein in E. coli, the fragment containing the region encompassing residues 2328-2491 of the human CI-MPR was amplified by PCR from a human skin fibroblast cDNA library. The amplified fragment was digested with BamHI and XhoI, and the BamHI/XhoI fragment was cloned into the BamHI-XhoI sites of pGEX-5X-1 (GE Healthcare) to generate a pGEX-5X-1-CI-MPR tail plasmid.

To generate plasmids expressing fusion proteins of GST-DMT1-II tail wild-type or various mutants in E. coli, the fragment containing the region encompassing residues 533-568 of human DMT1-II was amplified by PCR from pUC13-DMT1-II plasmids containing a wild-type or mutant version of DMT1-II, as described above. The amplified fragments were digested with EcoRI and XhoI, and the EcoRI-XhoI fragments were cloned into the EcoRI-XhoI sites of pGEX 5X-1 to generate plasmids of pGEX-5X-1-DMT1-II tail wild type or various mutants.

To generate a plasmid expressing the 3×FLAG-tagged retromer complex in E. coli, DNA fragments encoding 3×FLAG-tagged full-length molecules of each Vps26, Vps29 and Vps35 with a Shine-Dalgarno sequence at 5′-upstream of each ORF, were amplified by PCR from pIRES neo-3×FLAG-VPS26, -VPS29 and -VPS35 plasmids, respectively. The amplified fragments were sequentially cloned into pET23d plasmid in the order of 3×FLAG-VPS26, 3×FLAG-VPS29 and then 3×FLAG-VPS35 to generate a pET23d-3×FLAG-Retromer plasmid. 3×FLAG-Vps35 was fused to a C-terminal His6 tag.

The cloned regions of all plasmids were completely sequenced to ensure that unplanned mutations were not introduced during the cloning procedures.

Protein expression and purification

For purification of the GST-fusion protein, Rosseta-gamiB (DE3) cells (EMD) transformed with protein expression plasmids were grown in Luria broth with 100 μg/ml ampicillin at 37°C to an OD600 of 0.8, and then protein expression was induced by the addition of isopropyl-1-thio-β-D-galactopyranoside (IPTG) to 0.3 mM for 3 hours at 37°C. Cell pellets were resuspended in buffer A (0.5% Triton X-100 in PBS) containing 1 mM phenylmethylsulfonyl fluoride (PMSF) and 1 mg/ml lysozyme, and disrupted by sonication. Lysates were cleared by centrifugation at 20,000 g for 30 minutes at 4°C and GST-tagged proteins were purified from the supernatant using glutathione-Sepharose 4B beads (GE Healthcare). After incubation for 2 hours with the beads, the beads were washed five times with buffer A, and then proteins were eluted with G buffer (10 mM reduced glutathione, 50 mM Tris-HCl, pH 9.0). Proteins were subsequently dialyzed in PBS and protein concentrations were determined using a micro BCA protein assay kit (Pierce).

For the purification of the FLAG-tagged retromer complex, Rosseta-gami2 (DE3) pLys cells (EMD) transformed with pET23d-3×FLAG-Retromer plasmid were grown in Terrific broth (Invitrogen) with 34 μg/ml chloramphenicol, 50 μg/ml streptomycin, 12.5 μg/ml tetracycline, and 100 μg/ml ampicillin at 37°C to an OD600 of 0.8. The bacteria were then moved to 16°C for 30 minutes and protein expression was induced by the addition of IPTG to 0.3 mM for 19 hours. The cell pellet was resuspended in buffer B (50 mM NaH2PO4, 300 mM NaCl, 5 mM imidazole, pH 8.0) containing complete EDTA-free protease inhibitor cocktail (Roche Molecular Biochemicals), 1 mM PMSF and 1 mg/ml lysozyme, and disrupted by sonication. The lysate was cleared by centrifugation at 20,000 g for 30 minutes at 4°C and the cleared lysate was applied to the Ni-NTA agarose column (Qiagen) equilibrated with buffer B. The column was then washed with 10 column volumes of wash buffer (50 mM NaH2PO4, 300 mM NaCl, 10 mM imidazole, pH 8.0) and eluted with elution buffer (50 mM NaH2PO4, 300 mM NaCl, 250 mM imidazole, pH 8.0). Fractions containing the recombinant FLAG-retromer complex were collected and subsequently dialyzed into PBS, and protein concentrations were determined using a micro BCA protein assay kit (Pierce).

GST pull-down assay

GST or GST fusion proteins (0.1 mg/ml) were incubated with recombinant FLAG-tagged retromer cargo recognition complex (3×FLAG-Vps26, -Vps29 and -Vps35-His6) in 0.5 ml GST pull-down buffer [50 mM HEPES-KOH, pH 7.4, 50 mM KCl, 1 mM MgCl2, 10% (v/v) glycerol, 2 mM dithiothreitol, and 0.6 mg/ml bovine serum albumin] for 2 hours at 4°C. Glutathione-Sepharose 4B beads (50 μl) were added and the incubation was continued for an additional 30 minutes. The beads were washed with 1 ml GST pull-down buffer five times and the bound proteins were eluted with 50 μl of G buffer (10 mM glutathione, 50 mM Tris-HCl, pH 9.0). Samples were analyzed by immunoblot analysis. Immunoblot analysis was performed using mouse anti-FLAG M2 mAb and horseradish peroxidase-conjugated anti-mouse IgG. Signals were detected using Pierce SuperSignal West Pico chemiluminescent detection reagent (Pierce).

Cell transfection and immunofluorescence microscopy

Human HEp-2 epithelial cells were cultured on 12-well dishes at 37°C in DME high-glucose medium (Sigma) supplemented with 10% (v/v) FBS. For mutational analysis of the DMT1-II-recycling signal, plasmids encoding various GFP-tagged mutants of DMT1-II were transfected using FuGENE 6 transfection reagent (Roche Molecular Biochemicals) into HEp-2 cells according to the manufacturer's instructions. At 24 hours after transfection, the cells were stripped with trypsin-EDTA, and seeded onto glass coverslips, and cultured for an additional 48 hours. Cells were fixed, permeabilized, and immunostained with rabbit anti-GFP pAb and mouse anti-LAMP2 mAb, followed by Alexa Fluor 488-conjugated goat anti-rabbit IgG and Alexa Fluor 594-conjugated goat anti-Mouse IgG. Cell images were obtained using a confocal microscope (TSC SP-2; Leica) equipped with 488-nm Ar-Kr, 543/594 nm He-Ne, and 633 nm He-Ne lasers. Images were acquired in a sequential mode using a 63× Plan Apochromat NA/1.4 oil objective and the appropriate filter combination. All images were saved as TIFF files, their contrast was adjusted with Photoshop (version CS2; Adobe), and the images were imported into Illustrator (version CS2; Adobe). Pixel-by-pixel colocalization analysis was performed using ImageJ (plug-in JACoP) to determine the Pc with marker proteins (Bolte and Cordelieres, 2006).

To generate HEp-2 cells stably expressing 3×HA-DMT1-II, HEp-2 cells were grown at 50-60% confluency in a 10 cm dish and transfected with 5 μg pIRES neo-3×HA-DMT1-II plasmid using FuGENE 6 transfection reagent. At 24 hours after the transfection, cells were stripped with trypsin-EDTA and cultured in 96-well dishes in DMEM supplemented with 10% FBS containing 0.75 mg/ml G418 (Nacalai Tesque). The medium was changed every 3 days until colonies were formed. The expression of the G418-resistant clones was tested by immunostaining and immunoblot analysis, and clones with normal localization of DMT1-II and normal cell morphology were selected.

For triple staining of 3×HA-DMT1-II, Vps35 and various organellar markers, HEp-2 cells stably expressing 3×HA-DMT1-II were grown on glass coverslips for 72 hours and fixed, permeabilized, and immunostained with mouse anti-TfR, EEA1, CI-MPR, p230, or Rab11 mAb, rabbit anti-HA pAb, and goat anti-Vps35 pAb followed by donkey Alexa Fluor 555-conjugated anti-rabbit IgG, donkey Alexa Fluor 488-conjugated anti-mouse IgG, and donkey Alexa Fluor 633-conjugated anti-goat IgG.

RNA interference

The siRNA knockdown oligonucleotides were obtained from Invitrogen (Stealth™/siRNA duplex oligoribonucleotides) and resuspended in water according to the manufacturer's instructions. HEp-2 cells stably expressing 3×HA-DMT1-II were grown to 30% confluency and transfected twice at 72-hours intervals with 33 nM siRNA oligonucleotides using Lipofectamine™ RNAiMAX (Invitrogen). For most experiments, cells were analyzed 72 hours after the second round of transfection.

In rescue experiments, cDNA encoding GFP-tagged siRNA-resistant Vps35 wild-type or R107A mutant, nonfunctional mutant Vps35 (Gokool et al., 2007), or GFP alone were transfected using FuGENE 6 transfection reagent (Roche Molecular Biochemicals) into Vps35-depleted cells 24 hours after the second siRNA treatment. Twelve hours after the plasmid transfection, cells were stripped with trypsin-EDTA and seeded onto glass coverslips, and cultured for an additional 48 hours. Cells were fixed, permeabilized, and immunostained with rabbit anti-GFP pAb, rat anti-HA mAb, and mouse anti-LAMP2 or TfR mAb, followed by Alexa Fluor 488-conjugated anti-rabbit IgG, Alexa Fluor 555-conjugated anti-rat IgG, and Alexa Fluor 633-conjugated anti-mouse IgG.

The following oligonucleotides were used to ablate expression of Vps35 (VSP35 siRNA#1, 5′-G CCU UCA GAG GAU GUU GUA UCU UUA-3′; VPS35 siRNA#2, 5′-GC AUG AGU UGU UAU GUG CUU AGU AA-3′).

We thank Scott D. Emr and Kazuhisa Nakayama for helpful discussions and their critical reading of the manuscript. This work was supported in part by the Asahi Glass Foundation, and also partly supported by the Research Project Grants from Kawasaki Medical School and the Okayama Medical Foundation.

Andrews
N. C.
(
1999
).
Disorders of iron metabolism
.
N. Engl. J. Med.
341
,
1986
-
1995
.
Andrews
N. C.
(
2008
).
Forging a field: the golden age of iron biology
.
Blood
112
,
219
-
230
.
Arighi
C. N.
,
Hartnell
L. M.
,
Aguilar
R. C.
,
Haft
C. R.
,
Bonifacino
J. S.
(
2004
).
Role of the mammalian retromer in sorting of the cation-independent mannose 6-phosphate receptor
.
J. Cell Biol.
165
,
123
-
133
.
Bolte
S.
,
Cordelieres
F. P.
(
2006
).
A guided tour into subcellular colocalization analysis in light microscopy
.
J. Microsc.
224
,
213
-
232
.
Bonifacino
J. S.
,
Traub
L. M.
(
2003
).
Signals for sorting of transmembrane proteins to endosomes and lysosomes
.
Annu. Rev. Biochem.
72
,
395
-
447
.
Bonifacino
J. S.
,
Rojas
R.
(
2006
).
Retrograde transport from endosomes to the trans-Golgi network
.
Nat. Rev. Mol. Cell. Biol.
7
,
568
-
579
.
Bonifacino
J. S.
,
Hurley
J. H.
(
2008
).
Retromer
.
Curr. Opin. Cell Biol.
20
,
427
-
436
.
Canonne-Hergaux
F.
,
Gruenheid
S.
,
Ponka
P.
,
Gros
P.
(
1999
).
Cellular and subcellular localization of the Nramp2 iron transporter in the intestinal brush border and regulation by dietary iron
.
Blood
93
,
4406
-
4417
.
Canonne-Hergaux
F.
,
Zhang
A. S.
,
Ponka
P.
,
Gros
P.
(
2001
).
Characterization of the iron transporter DMT1 (NRAMP2/DCT1) in red blood cells of normal and anemic mk/mk mice
.
Blood
98
,
3823
-
3830
.
Carlton
J.
,
Bujny
M.
,
Peter
B. J.
,
Oorschot
V. M.
,
Rutherford
A.
,
Mellor
H.
,
Klumperman
J.
,
McMahon
H. T.
,
Cullen
P. J.
(
2004
).
Sorting nexin-1 mediates tubular endosome-to-TGN transport through coincidence sensing of high-curvature membranes and 3-phosphoinositides
.
Curr. Biol.
14
,
1791
-
1800
.
Collins
B. M.
(
2008
).
The structure and function of the retromer protein complex
.
Traffic
9
,
1811
-
1822
.
Eaton
S.
(
2008
).
Retromer retrieves wntless
.
Dev. Cell
14
,
4
-
6
.
Fleming
M. D.
,
Trenor
C. C.
3rd
,
Su
M. A.
,
Foernzler
D.
,
Beier
D. R.
,
Dietrich
W. F.
,
Andrews
N. C.
(
1997
).
Microcytic anaemia mice have a mutation in Nramp2, a candidate iron transporter gene
.
Nat. Genet.
16
,
383
-
386
.
Fleming
M. D.
,
Romano
M. A.
,
Su
M. A.
,
Garrick
L. M.
,
Garrick
M. D.
,
Andrews
N. C.
(
1998
).
Nramp2 is mutated in the anemic Belgrade (b) rat: evidence of a role for Nramp2 in endosomal iron transport
.
Proc. Natl. Acad. Sci. USA
95
,
1148
-
1153
.
Ghosh
P.
,
Dahms
N. M.
,
Kornfeld
S.
(
2003
).
Mannose 6-phosphate receptors: new twists in the tale
.
Nat. Rev. Mol. Cell. Biol.
4
,
202
-
212
.
Gokool
S.
,
Tattersall
D.
,
Reddy
J. V.
,
Seaman
M. N.
(
2007
).
Identification of a conserved motif required for Vps35p/Vps26p interaction and assembly of the retromer complex
.
Biochem. J.
408
,
287
-
295
.
Gruenheid
S.
,
Canonne-Hergaux
F.
,
Gauthier
S.
,
Hackam
D. J.
,
Grinstein
S.
,
Gros
P.
(
1999
).
The iron transport protein NRAMP2 is an integral membrane glycoprotein that colocalizes with transferrin in recycling endosomes
.
J. Exp. Med.
189
,
831
-
841
.
Gunshin
H.
,
Mackenzie
B.
,
Berger
U. V.
,
Gunshin
Y.
,
Romero
M. F.
,
Boron
W. F.
,
Nussberger
S.
,
Gollan
J. L.
,
Hediger
M. A.
(
1997
).
Cloning and characterization of a mammalian proton-coupled metal-ion transporter
.
Nature
388
,
482
-
488
.
Hubert
N.
,
Hentze
M. W.
(
2002
).
Previously uncharacterized isoforms of divalent metal transporter (DMT)-1: implications for regulation and cellular function
.
Proc. Natl. Acad. Sci. USA
99
,
12345
-
12350
.
Imai
Y.
,
Matsushima
Y.
,
Sugimura
T.
,
Terada
M.
(
1991
).
A simple and rapid method for generating a deletion by PCR
.
Nucleic Acids Res.
19
,
2785
.
Katzmann
D. J.
,
Odorizzi
G.
,
Emr
S. D.
(
2002
).
Receptor downregulation and multivesicular-body sorting
.
Nat. Rev. Mol. Cell. Biol.
3
,
893
-
905
.
Lam-Yuk-Tseung
S.
,
Gros
P.
(
2006
).
Distinct targeting and recycling properties of two isoforms of the iron transporter DMT1 (NRAMP2, Slc11A2)
.
Biochemistry
45
,
2294
-
2301
.
Lam-Yuk-Tseung
S.
,
Touret
N.
,
Grinstein
S.
,
Gros
P.
(
2005
).
Carboxyl-terminus determinants of the iron transporter DMT1/SLC11A2 isoform II (-IRE/1B) mediate internalization from the plasma membrane into recycling endosomes
.
Biochemistry
44
,
12149
-
12159
.
Lee
P. L.
,
Gelbart
T.
,
West
C.
,
Halloran
C.
,
Beutler
E.
(
1998
).
The human Nramp2 gene: characterization of the gene structure, alternative splicing, promoter region and polymorphisms
.
Blood Cells Mol. Dis.
24
,
199
-
215
.
Luton
F.
,
Hexham
M. J.
,
Zhang
M.
,
Mostov
K. E.
(
2009
).
Identification of a cytoplasmic signal for apical transcytosis
.
Traffic
10
,
1128
-
1142
.
McMahon
H. T.
,
Gallop
J. L.
(
2005
).
Membrane curvature and mechanisms of dynamic cell membrane remodelling
.
Nature
438
,
590
-
596
.
Nothwehr
S. F.
,
Ha
S. A.
,
Bruinsma
P.
(
2000
).
Sorting of yeast membrane proteins into an endosome-to-Golgi pathway involves direct interaction of their cytosolic domains with Vps35p
.
J. Cell Biol.
151
,
297
-
310
.
Puertollano
R.
,
Aguilar
R. C.
,
Gorshkova
I.
,
Crouch
R. J.
,
Bonifacino
J. S.
(
2001
).
Sorting of mannose 6-phosphate receptors mediated by the GGAs
.
Science
292
,
1712
-
1716
.
Rink
J.
,
Ghigo
E.
,
Kalaidzidis
Y.
,
Zerial
M.
(
2005
).
Rab conversion as a mechanism of progression from early to late endosomes
.
Cell
122
,
735
-
749
.
Rojas
R.
,
van Vlijmen
T.
,
Mardones
G. A.
,
Prabhu
Y.
,
Rojas
A. L.
,
Mohammed
S.
,
Heck
A. J.
,
Raposo
G.
,
van der Sluijs
P.
,
Bonifacino
J. S.
(
2008
).
Regulation of retromer recruitment to endosomes by sequential action of Rab5 and Rab7
.
J. Cell Biol.
183
,
513
-
526
.
Seaman
M. N.
(
2004
).
Cargo-selective endosomal sorting for retrieval to the Golgi requires retromer
.
J. Cell Biol.
165
,
111
-
122
.
Seaman
M. N.
(
2005
).
Recycle your receptors with retromer
.
Trends Cell Biol.
15
,
68
-
75
.
Seaman
M. N.
(
2007
).
Identification of a novel conserved sorting motif required for retromer-mediated endosome-to-TGN retrieval
.
J. Cell Sci.
120
,
2378
-
2389
.
Seaman
M. N.
(
2008
).
Endosome protein sorting: motifs and machinery
.
Cell Mol. Life Sci.
65
,
2842
-
2858
.
Seaman
M. N.
,
McCaffery
J. M.
,
Emr
S. D.
(
1998
).
A membrane coat complex essential for endosome-to-Golgi retrograde transport in yeast
.
J. Cell Biol.
142
,
665
-
681
.
Seaman
M. N.
,
Harbour
M. E.
,
Tattersall
D.
,
Read
E.
,
Bright
N.
(
2009
).
Membrane recruitment of the cargo-selective retromer subcomplex is catalysed by the small GTPase Rab7 and inhibited by the Rab-GAP TBC1D5
.
J. Cell Sci.
122
,
2371
-
2382
.
Shah
Y. M.
,
Matsubara
T.
,
Ito
S.
,
Yim
S. H.
,
Gonzalez
F. J.
(
2009
).
Intestinal hypoxia-inducible transcription factors are essential for iron absorption following iron deficiency
.
Cell Metab.
9
,
152
-
164
.
Strochlic
T. I.
,
Setty
T. G.
,
Sitaram
A.
,
Burd
C. G.
(
2007
).
Grd19/Snx3p functions as a cargo-specific adapter for retromer-dependent endocytic recycling
.
J. Cell Biol.
177
,
115
-
125
.
Su
M. A.
,
Trenor
C. C.
,
Fleming
J. C.
,
Fleming
M. D.
,
Andrews
N. C.
(
1998
).
The G185R mutation disrupts function of the iron transporter Nramp2
.
Blood
92
,
2157
-
2163
.
Tabuchi
M.
,
Yoshimori
T.
,
Yamaguchi
K.
,
Yoshida
T.
,
Kishi
F.
(
2000
).
Human NRAMP2/DMT1, which mediates iron transport across endosomal membranes, is localized to late endosomes and lysosomes in HEp-2 cells
.
J. Biol. Chem.
275
,
22220
-
22228
.
Tabuchi
M.
,
Tanaka
N.
,
Nishida-Kitayama
J.
,
Ohno
H.
,
Kishi
F.
(
2002
).
Alternative splicing regulates the subcellular localization of divalent metal transporter 1 isoforms
.
Mol. Biol. Cell
13
,
4371
-
4387
.
Takatsu
H.
,
Katoh
Y.
,
Shiba
Y.
,
Nakayama
K.
(
2001
).
Golgi-localizing, gamma-adaptin ear homology domain, ADP-ribosylation factor-binding (GGA) proteins interact with acidic dileucine sequences within the cytoplasmic domains of sorting receptors through their Vps27p/Hrs/STAM (VHS) domains
.
J. Biol. Chem.
276
,
28541
-
28545
.
Touret
N.
,
Furuya
W.
,
Forbes
J.
,
Gros
P.
,
Grinstein
S.
(
2003
).
Dynamic traffic through the recycling compartment couples the metal transporter Nramp2 (DMT1) with the transferrin receptor
.
J. Biol. Chem.
278
,
25548
-
25557
.
Traer
C. J.
,
Rutherford
A. C.
,
Palmer
K. J.
,
Wassmer
T.
,
Oakley
J.
,
Attar
N.
,
Carlton
J. G.
,
Kremerskothen
J.
,
Stephens
D. J.
,
Cullen
P. J.
(
2007
).
SNX4 coordinates endosomal sorting of TfnR with dynein-mediated transport into the endocytic recycling compartment
.
Nat. Cell Biol.
9
,
1370
-
1380
.
Verges
M.
,
Luton
F.
,
Gruber
C.
,
Tiemann
F.
,
Reinders
L. G.
,
Huang
L.
,
Burlingame
A. L.
,
Haft
C. R.
,
Mostov
K. E.
(
2004
).
The mammalian retromer regulates transcytosis of the polymeric immunoglobulin receptor
.
Nat. Cell Biol.
6
,
763
-
769
.
Wassmer
T.
,
Attar
N.
,
Harterink
M.
,
van Weering
J. R.
,
Traer
C. J.
,
Oakley
J.
,
Goud
B.
,
Stephens
D. J.
,
Verkade
P.
,
Korswagen
H. C.
, et al. 
. (
2009
).
The retromer coat complex coordinates endosomal sorting and dynein-mediated transport, with carrier recognition by the trans-Golgi network
.
Dev. Cell
17
,
110
-
122
.
Yang
P. T.
,
Lorenowicz
M. J.
,
Silhankova
M.
,
Coudreuse
D. Y.
,
Betist
M. C.
,
Korswagen
H. C.
(
2008
).
Wnt signaling requires retromer-dependent recycling of MIG-14/Wntless in Wnt-producing cells
.
Dev. Cell
14
,
140
-
147
.
Yoshimori
T.
,
Shimonishi
Y.
,
Uchida
T.
(
1988
).
Binding properties of monoclonal antibody to the cytoplasmic domain of transferrin receptor
.
Cell Struct. Funct.
13
,
311
-
324
.
Zhu
Y.
,
Doray
B.
,
Poussu
A.
,
Lehto
V. P.
,
Kornfeld
S.
(
2001
).
Binding of GGA2 to the lysosomal enzyme sorting motif of the mannose 6-phosphate receptor
.
Science
292
,
1716
-
1718
.

Supplementary information