An Iron Delivery Pathway Mediated by a Lipocalin

Molecular Cell, Vol. 10, 1045–1056, November, 2002, Copyright 2002 by Cell Press An Iron Delivery Pathway Mediated by a Lipocalin

Jun Yang,1 David Goetz,2,5 Jau-Yi Li,1,5 Wenge Wang,1 iron to many tissues, including epithelia. Hypotransferri- Kiyoshi Mori,1 Daria Setlik,1 Tonggong Du,1 nemic micehpx/hpx (Huggenvik et al., 1989; Trenor et al., Hediye Erdjument-Bromage,3 Paul Tempst,3 2000) and atransferrinemic humans (Hamill et al., 1991; Roland Strong,2 and Jonathan Barasch1,4 Hayashi et al., 1993) have severe defects in hematopoie- 1College of Physicians and Surgeons sis and central nervous system development, but most of Columbia University epithelial organs are normal. Likewise, mice lacking the New York, New York 10032 transferrin receptor1 (Levy et al., 1999) initiate organo- 2 The Fred Hutchinson Cancer Institute genesis but succumb to the effects of anemia. Given Seattle, Washington 98109 that iron is necessary for all cells, there must be other 3 The Memorial Sloan-Kettering Cancer Center pathways for iron acquisition in nonhematopoietic cells. New York, New York 10021 These additional pathways might involve the local expression of iron-trafficking proteins, such as lactoferrin (Ward et al., 1999b), putative low molecular weight sider- Summary ophores (Fernandez-Pol, 1978; Jones et al., 1980), and cell surface iron transporters like DMT1. Despite the critical need for iron in many cellular reac- Developing organs may be served by a variety of iron tions, deletion of the transferrin pathway does not delivery proteins. Transferrin is expressed in the primor- block organogenesis, suggesting the presence of al- dial liver (Cassia et al., 1997), but circulation of this ternative methods to deliver iron. We show that a protein is uncertain before E12, by which time organo- member of the lipocalin superfamily (24p3/Ngal) deliv- genesis is advanced (Gustine and Zimmerman, 1973). ers iron to the cytoplasm where it activates or re- Transferrin is probably most critical for later stages of presses iron-responsive genes. Iron unloading de- hematopoiesis and neural tube development because pends on the cycling of 24p3/Ngal through acidic these cells undergo apoptosis as late as E13 in trans- endosomes, but its pH sensitivity and its subcellular ferrin receptor1-deleted mice (Levy et al., 1999). Like- targeting differed from transferrin. Indeed, during the wise, the final steps of kidney development are best conversion of mesenchyme into epithelia (where we supported by transferrin, but initial steps do not require discovered the protein), 24p3/Ngal and transferrin this protein (Ekblom et al., 1983; Landschulz et al., 1984). were endocytosed by different cells that characterize In these examples, iron delivery by transferrin is thought different stages of development, and they triggered to regulate proliferation of determined cells. Hence, unique responses. These studies identify an iron deliv- early stages of a lineage likely acquire iron by different ery pathway active in development and cell physiology. mechanisms and perhaps for different purposes. How- ever, specific iron delivery pathways are unknown. Introduction The cardinal feature of kidney development is the formation of epithelial tubules from nonepithelial mesen- The delivery of iron to cells is critical for cell growth and chymal cells. The conversion of cell type is controlled development. Iron is located at the active site of a large by factors secreted from the ureteric bud (Barasch et number of proteins including regulators of intermediary al., 1999; Plisov et al., 2001), and we have now purified metabolism and DNA synthesis (Cooper and Porter, and sequenced an epithelial inducer from ureteric bud 1997; Nyholm et al., 1993), and it stabilizes three-dimen- cell lines (UB cells; Barasch et al., 1996). The protein is sional protein structure (e.g., Anderson et al., 1990). In called 24p3 (the mouse ortholog of human neutrophil- addition, iron is a unique regulator of gene expression, gelatinase-associated lipocalin, Ngal), and it is a mem- activating both transcriptional (Yamaguchi-Iwai et al., ber of the lipocalin superfamily. The lipocalin family is 1996) and posttranscriptional mechanisms (Rouault and thought to regulate growth and development by binding Klausner, 1997). For these reasons, the acquisition of cell surface receptors (Wojnar et al., 2001) and activating iron is central to cell survival, growth, and maturation, downstream targets (Devireddy et al., 2001) or, alterna- and its metabolism is tightly regulated by many proteins. tively, by delivering low molecular weight ligands (Flower Most cells acquire iron by capturing iron-loaded trans- et al., 2000). By X-ray crystallography, atomic absorp- ferrin. After binding to receptors, transferrin enters an tion, and X-ray fluorescence analyses, we found that endocytic pathway (van Renswoude et al., 1982; Ya- 24p3/Ngal contains iron, and we now report that 24p3 mashiro et al., 1984). Within endosomes, iron dissoci- and Ngal are iron-trafficking proteins. Remarkably, in ates from transferrin and is transferred across the vesi- tissue culture cells, the uptake, subcellular trafficking, cle membrane into the cell cytoplasm. In epithelia, this and final destination of 24p3/Ngal differed from those pathway is uniquely located in the basal domain of the taken by transferrin. In the primordial kidney, 24p3/ cell (Fuller and Simons, 1986). However, despite its ubiq- Ngal targeted early epithelial progenitors and stroma, uity and quantitative importance in adult physiology, the whereas transferrin targeted later staged epithelial pro- transferrin pathway is not essential for the delivery of genitors. These data show that, during organogenesis, mechanisms for the acquisition of iron are stage and 4 Correspondence: [email protected] cell-type specific and, consequently, iron is likely to 5 These authors contributed equally to the work. activate different programs upon delivery. Molecular Cell 1046

Figure 1. Induction of Epithelia and Tubules by the Ureteric Bud Protein 24p3 (A) A purified UB protein epithelializes rat metanephric mesenchyme. The epithelia form chains and kidney tubules called S-shaped bodies (inset) that contain many types of epithelia including visceral (v) and parietal (p) cells of the glomerulus (inset). Bars, 100 ␮m. (B) Silver stain of the active fraction reveals a single 25 kDa protein, identified as 24p3. (C) Whole mount in situ hybridization of E15 rat kidney (overnight culture) shows 24p3 is expressed predominantly by the ureteric bud (arrows and inset). S-shaped (s) nephrons also express 24p3, but metanephric mesenchyme is negative. (D) Frozen section of E13 rat kidney stained with anti-24p3. Staining is found in the ureteric bud (UB) and its basement membrane but not in mesenchyme. Wolffian duct (w). (E) Immunoblots confirm expression of 24p3 in ureteric bud. 20 ␮g/lane. (D and E) Preimmune were negative.

Results To confirm the identification of an inducer of mesenchymal to epithelial conversion, we used a recombinant A Member of the Lipocalin Superfamily Induces glutathione S-transferase (gst)-24p3 fusion protein and Mesenchymal to Epithelial Conversion its derivative, pure 24p3. These proteins (50–100 ␮g/ml; We previously showed that factors produced by mouse 2–4 ␮M) induced metanephric mesenchyme to convert UB cells (Barasch et al., 1996) convert metanephric mes- into epithelia (n Ͼ 100). Human Ngal exists as a monomer enchyme into complex epithelia. Using an in vitro assay, or a disulfide-linked dimer (Kjeldsen et al., 1993), and we and others have already purified one of these factors, recombinant versions of both forms induced cell conver- the IL-6 cytokine, LIF (Barasch et al., 1999; Plisov et al., sion (n Ͼ 100). Likewise, the Ngal fraction of a commer- 2001). One additional activity bound to heparin-Sepha- cial preparation of disulfide-linked Ngal-gelatinaseB in- rose beads, and we proceeded to purify it to homogene- duced cellular conversion (n ϭ 2), but the gelatinase ity by cation exchange, chromatofocusing, and hydro- fraction was negative. These assays confirm that mouse phobic interaction chromatography. We found that the 24p3 and its human ortholog Ngal are epithelial inducers inducing activity (Figure 1A) was a 25 kDa protein (Figure and show that the activity is conserved across species. 1B). The protein was identified by matrix-assisted laser- 24p3 behaved like a classical inducer in that it was desorption/ionization reflectron time-of-flight (MALDI- largely expressed by the ureteric bud and not by mesen- reTOF) and by electrospray ionization (ESI) MS/MS of chymal cells. Figures 1C and 1D show expression by in tryptic peptides. The protein was mouse 24p3, the mouse situ hybridization and by immunocytochemistry in the ortholog of human neutrophil-gelatinase-associated- rat ureteric bud. Immunoblots confirmed this expression lipocalin (Ngal) and a member of the lipocalin superfam- pattern (Figure 1E) as did gene chip analysis. These data ily of transport proteins. demonstrate that the ureteric bud expresses 24p3 in vivo. Iron Traffic in Embryonic Kidney 1047

Figure 2. Iron-24p3 (A) UB cells were incubated with 59Fe or 35S-methionine for 3 days and 24p3 immunoprecipitated. Immune sera were specific for a single 35S- labeled 25 kDa protein. The immunoprecipitate contained four times more 59Fe than the preimmune precipitate. (B) 59Fe-labeled conditioned media were fractionated by heparin-Sepharose chromatography. Proteins were eluted by a gradient of NaCl and measured (280 nm light, red line), 59Fe was detected by liquid scintillation (black line), and 24p3 was detected by immunoblot. 59Fe and 24p3 coeluted. Inspection of the sample showed a copurifying low molecular weight substance that was reddish-yellow.

24p3/Ngal Binds Iron Using a Cofactor observation that 24p3-containing fractions of ureteric The structure of Ngal was solved by X-ray crystallogra- bud proteins contained a low molecular weight reddish- phy, and a combination of X-ray crystallography, atomic yellow substance (Figure 2B, top), similar in color to absorption, and X-ray spectroscopy showed that bacte- enterobactin. Moreover, the molecular mechanism by rially cloned human Ngal contains stoichiometric quanti- which Ngal recognizes enterobactin is likely applicable ties of iron (Goetz et al., 2002 [this issue of Molecular to other types of siderophores (oxazaline or thiazoline Cell]). In addition, the cloned protein contained a small type) in addition to the tris-catecholate (Goetz et al., ringed structure called enterobactin or enterochelin, a 2002). tris catecholate siderophore (Neilands, 1995) that mediates the association of iron with Ngal. To determine Iron Trafficking by 24p3/Ngal whether native mouse 24p3 can also associate with iron, To test whether cloned 24p3/Ngal can deliver iron, we we metabolically labeled UB cells with 59Fe and then incubated UB cells (mouse embryonic kidney collecting purified 24p3 using immunoprecipitation or column duct cells; Barasch et al., 1996), human Wilms tumor chromatography. The immunoprecipitate contained a kidney cells (WT), Madin Darby Canine Kidney (MDCK; single protein of 25 kDa (detected in a parallel immuno- dog kidney collecting duct cells), 7.1 (rat embryonic kid- precipitation from 35S-methionine-labeled cells) and ney epithelia; Herzlinger et al., 1993), and BF-2/Foxd1ϩ 18,462 Ϯ 963 cpm of 59Fe. In contrast, preimmune sera cells (mouse kidney stroma; Hatini et al., 1996) with 59Fe- failed to precipitate a specific protein, and it contained labeled human Ngal or mouse 24p3. All cells accumu- on average only 3938 Ϯ 370 cpm of 59Fe (Figure 2A). In lated 59Fe in a time-dependent manner (Figure 3A). Iron addition, we partially purified 24p3 from the radiolabeled uptake was blocked at 4ЊC or by the inclusion of a 20- media using a heparin-Sepharose column and found fold molar excess of unlabeled ligands, indicating that that 24p3 (Figure 2B, top) and 59Fe copurified (Figure the uptake was receptor mediated. As would be ex- 2B, bottom). Last, we repurified 24p3 from UB cells, pected from numerous studies, these cells also accumu- depleted it of endogenous iron by prolonged dialysis lated 59Fe-transferrin (both species-matched and human with the chelator deferoxamine, followed by ultrapure transferrin) in a time- and temperature-dependent fash- water, and then added increasing concentrations of 59Fe. ion. Interestingly, it appeared that 59Fe delivery by 24p3/ We found that pure 24p3 trapped 59Fe, that the associa- Ngal reached an isotope equilibrium distribution that tion was competed by 100-fold excess nonradioactive was lower than that produced by transferrin despite iron, and that the stoichiometry of iron to protein was the addition of identical isotope concentrations to the less than 1:1. These results show that mammalian ex- medium. These results suggest that the two proteins pressed 24p3 contains iron. The nature of the iron-pro- deliver iron to different compartments (see below) and tein association is not clear, however, but it may involve that iron trafficking by 24p3/Ngal could mediate a sub- a mammalian homolog of enterobactin, whose existence stantial rate of iron delivery. was suggested by prior studies (Fernandez-Pol, 1978; Because acquisition of iron from an iron-loaded pro- Jones et al., 1980). This hypothesis is supported by the tein like transferrin requires endocytosis of the carrier Molecular Cell 1048

Figure 3. Subcellular Trafficking of 24p3/Nga1 (A) Iron delivery by Ngal. MDCK cells were incubated with 59Fe-Ngal or 59Fe-transferrin at 37Њor 4ЊC and time- and temperature-dependent uptake of iron was found (n ϭ 4). Addition (arrows) of the vacuolar HϩATPase inhibitor, bafilomycin (0.5 ␮M) blocked 59Fe accumulation. Similar data were obtained from other renal cells and with human or dog transferrin. (B) Endocytosis of Alexa568-24p3 in UB cells. (C and D) Uptake was blocked by a 20-fold molar excess of unlabeled 24p3(C) but not by a 20-fold molar excess of unlabeled transferrin (D). (E) Conversely, unlabeled 24p3 did not block endocytosis of Alexa488-transferrin. Note the different intracellular distributions of 24p3 and transferrin in UB cells (compare [B] and [E]) and Ngal and transferrin in WT renal epithelia (compare [F] and [G]). In addition, some cells show an exclusive uptake of one of these proteins (compare [F] and [G]). (F–H) At steady state, Ngal organelles were not occupied by transferrin (green or red arrows), except in a single perinuclear compartment in a minority of cells ([H], yellow arrows). Confocal 3 ␮m projection. Bar, 15 ␮m. protein (Dautry-Varsat et al., 1983), we examined whether beled by different fluorescent tags. We found that all 24p3/Ngal also entered cells. Cells were incubated with of the cell lines took up fluorescent and gst-labeled purified gst-24p3 fusion protein or with 24p3/Ngal la- lipocalins and that the internalized protein was localized Iron Traffic in Embryonic Kidney 1049

to vesicles scattered throughout the cytoplasm (Figures than transferrin for maximal iron release (Figure 4F), and 3B and 3F). Uptake was inhibited at 4ЊC and by a 20- only 50% of the exchangeable iron was released by our fold molar excess of nonfluorescent ligand (Figure 3C), acidified solutions. To determine whether 24p3/Ngal- confirming that it was receptor dependent. Next we ex- loaded organelles had an acid pH that could mediate amined whether 24p3/Ngal shared the same endocytic iron release, we labeled renal epithelia and stroma with pathway as transferrin. First, we found that transferrin fluorescently labeled Ngal and the pH probe DAMP (An- did not block the entry of labeled 24p3 into the test cell derson et al., 1984). DAMP accumulated in vesicles con- lines (Figure 3D) and conversely 24p3 did not block taining 24p3, indicating that it was located in acidic the entry of labeled transferrin (Figure 3E). These data vesicles (Figure 4G). The accumulation of DAMP re- demonstrate that these two iron transport proteins enter flected the transvesicular pH difference because it was cells by independent receptor-mediated mechanisms. dissipated when ammonium chloride (30 mM) was To study the polarized distribution of the two recep- added just prior to terminating the experiment (Figure tors, we grew MDCK and UB cells on porous filters. We 4H). To determine whether the low pH of the endocytic found that both 24p3/Ngal and transferrin were captured pathway was required for the delivery of iron from 24p3/ from the basal surface of MDCK, but in UB cells 24p3 Ngal, we fed epithelia and stroma 59Fe-Ngal and bafilo- was captured at both apical and basal surfaces, while mycin (0.5 ␮M), an inhibitor of the vacuolar HϩATPase transferrin entered only from the basal surface. Similar (Bowman et al., 1988). We found that bafilomycin studies were obtained using 59Fe-24p3/Ngal or 59Fe- blocked the incorporation of 59Fe from Ngal and from transferrin, showing that the initial phase of trafficking of 59Fe-transferrin (Figure 3A), whereas endocytosis of li- these two molecules is independent. Filter-grown MDCK gands continues under these conditions (Clague et al., and UB cells formed tight monolayers that were imper- 1994). These data show that 24p3/Ngal traffics in acid meable to gst-24p3 and to biotin-transferrin, added to vesicles and that low pH is essential for iron release. either apical or basal media. Hence, no significant trans- After the release of iron from its carrier protein in epithelial flux of either molecule appeared over 12 hr of cytoplasmic vesicles, iron is transferred to the cytosol. incubation (measured by immunoblot), ruling out signifi- Transferrin, for example, delivers iron to vesicles that cant transcytosis or leakage (data not shown). contain the membrane protein NRAMP2/DMT1(divalent To examine the steady-state distributions of 24p3/ metal transporter1) which is an Fe:Hϩ synporter (Gun- Ngal and transferrin, we examined a variety of cells after shin et al., 1997). To determine whether 24p3/Ngal deliv- prolonged incubation with the fluorescent conjugates ers iron to vesicles that contain this transporter, we and found that 24p3/Ngal and transferrin occupied dis- localized DMT1 in a number of cell lines with specific tinct intracellular compartments (Figure 3). For example, antibodies. We found that DMT1 partially colocalized in human WT cells (Figures 3F–3H), human transferrin with fluorescent 24p3 (Figure 4I). These data show that occupied a small perinuclear compartment. In contrast, vesicles containing 24p3/Ngal are potential sites of iron human Ngal occupied larger vesicles located at the pe- acquisition. riphery of the transferrin compartment, as well as To assess whether 24p3/Ngal is significantly de- throughout the cytoplasm. There was almost no overlap graded by intracellular trafficking, we measured the loss of Ngal and transferrin, but in some cells a small perinu- of 35S-methionine-labeled 24p3 and 125I-Ngal. For com- clear compartment contained both proteins (Figure 3H, parison, we used 125I-transferrin, which is not degraded yellow arrows). To further characterize the distribution after endocytosis (van Renswoude et al., 1982) and 125I- of 24p3/Ngal in WT and stromal cells, we used markers ␣2-macroglobulin which traffics to lysosomes and is of different organelles and found that Ngal did not oc- degraded (Tycko and Maxfield, 1982; Yamashiro et al., cupy early endosomes (EEA1 and rab5; Figures 4A and 1984). We found that transferrin and 24p3/Ngal were not 4B). Likewise, the majority of Ngal did not occupy late digested and remained insoluble (Figure 4J), while 125I- recycling endosomes (rab11; Figure 4C) or late endo- ␣2 macroglobulin was converted into a TCA soluble somes/prelysosomes (mannose-6-phosphate receptor; form. We confirmed these data by recovering full-length Figure 4D), although a colabeled population was readily gst-24p3 (measured by immunoblot) and TCA insoluble detectable (yellow arrows). Similarly, most 24p3 was 125I-Ngal after contact with a variety of cells for 12 hr, distinct from dextran-labeled lysosomes (Yamashiro et whereas 125I-␣2 macroglobulin became TCA soluble. In al., 1984), but again there was a small population with addition, we quantitatively recovered 24p3 from meta- colabeling (Figure 4E). These data show that 24p3/Ngal nephric mesenchyme after 7 days of coculture. These can reach a late endosomal compartment and probably data show that, like transferrin, 24p3/Ngal enters an crosses the transferrin pathway in the perinuclear recy- endocytic pathway and the protein is not significantly cling endosome. However, at steady state the bulk of degraded. Hence, it is plausible that 24p3/Ngal might 24p3/Ngal and transferrin is distinct. be reutilized after releasing iron. This was further sug- That the endocytic pathway of 24p3/Ngal and trans- gested by 59Fe binding to acidified and then realkalinized ferrin could be distinguished raised the question of 24p3/Ngal. However, further analysis of recycling must whether iron delivery by these proteins required different include structural and functional studies of the recycled mechanisms. Dissociation of iron from transferrin is me- species, since the siderophore may be labile (Goetz et diated in part by the acid pH of the endosome (Dautry- al., 2002). Varsat et al., 1983). To determine whether 59Fe-24p3/ The data show that 24p3/Ngal is incorporated by a Ngal was also pH sensitive, we incubated 59Fe-Ngal in saturable mechanism into acidic vacuoles with trans- solutions of varying pH and recovered the protein-bound membrane iron transporters. These vesicles are distinct 59Fe on a filter. Like transferrin, Ngal released iron at low from transferrin vesicles, identifying a separate path for pH (n ϭ 5); however, Ngal required a much lower pH iron delivery. Like transferrin, 24p3/Ngal may recycle. Molecular Cell 1050

Figure 4. Distribution of 24p3/Ngal in Acidic Vesicles WT (A–D) and stromal cells (E, G–I). Ngal-Alexa568 (red) is compared to EEA (A), Rab5-GFP (B), Rab11-HA (C), and mannose 6 phosphate receptor (D), and 24p3-Alexa568 (red) is compared to dextran-marked lysosomes (E). 24p3/Ngal showed minor colocalization with late endosomal markers rab11 and mannose-6-phosphate as well as with lysosomes (single labeling, red or green arrows; colocalization, yellow arrows). (F) Dissociation of 59Fe from Ngal or transferrin at acid pH. Protein-bound 59Fe was captured on a microcon filter. Ngal shows an acid-shifted pH sensitivity (n ϭ 5). (G) 24p3-Alexa488 (green) localizes to acid vesicles that trap DAMP, a pH probe (red) that is released by alkalinization (NH4; H). 24p3 also colocalized with the divalent metal transporter, DMT1 ([I]; colocalization, yellow arrows). (J) 24p3/Ngal is not degraded during intracellualr transport. Renal stroma (squares) and 7.1 cells (circles) were loaded with125I-24p3 or Ngal and washed, and the media were collected after 1 hr. There was no change in the fraction of degraded, TCA-soluble 24p3/Ngal or transferrin, but 125I-␣2 macroglobulin converted into a TCA-soluble form in a temperature-sensitive manner. Similar data were obtained from 35S-24p3, gst-24p3, and unlabeled Ngal. Confocal 3 ␮m projection. Bars, 25 ␮m.

24p3/Ngal Regulates Iron-Responsive Genes test this, we added iron-loaded Ngal to a number of cell Iron regulates the expression of a number of genes by lines in the absence of transferrin. We found that the enhancing translation of message or conversely by pro- expression of ferritin was enhanced, while transferrin moting degradation of the mRNA. As a result, the iron receptor1 was suppressed. Hence, iron delivery by Ngal storage protein ferritin and the transferrin receptor1 are was sufficient to regulate iron-responsive genes. Similar reciprocally regulated by cytoplasmic iron. Treating cells data were obtained in the cultured metanephric mesen- with iron-loaded transferrin enhances the translation of chyme, which expressed a higher basal level of ferritin ferritin while reducing the expression of the transferrin compared to the cell lines. receptor1 (Figure 5). Conversely, chelating cellular iron with deferroxamine (Figure 5) inhibits the translation of 24p3/Ngal and Transferrin Have Nonredundant ferritin, but enhances the expression of transferrin re- Functions during Organogenesis ceptor1. Hence, if 24/p3Ngal can deliver physiological The distinct trafficking pathways of 24p3/Ngal and quantities of iron to cells, then ferritin expression should transferrin might be utilized to supply iron to different increase and transferrin receptor1 should decrease. To cells in a developing organ. To test this idea, we added Iron Traffic in Embryonic Kidney 1051

in tandem: 24p3/Ngal induces primordial cells, while transferrin acts on their progeny to complete the developmental program.

Discussion

Lipocalins are structurally related proteins that transport a variety of ligands (Flower et al., 2000). By delivering these cargo, the lipocalins are believed to act as survival factors (e.g., purpurin, Schubert and LaCorbiere, 1985) and developmental regulators (e.g., lazarillo, Sanchez et al., 1995; retinol binding protein, Newcomer and Ong, 2000). For example, glycodelin (Kamarainen et al., 1997) promotes epithelialization after endocytosis, suggesting that it delivers a small molecule that stimulates differentiation (Zhou et al., 2000). However, the specific actions of the lipocalins have not been established because their ligands and receptors are unknown. Hence, it has not been clear why 24p3/Ngal is highly expressed at sites Figure 5. 24p3/Ngal Regulate Iron-Responsive Genes of rapid epithelial turnover such as malignant or inflamed 7.1 cells were cultured with 24p3/Ngal (100 ␮g/ml), transferrin (200 ␮g/ml), additional free iron (50 ␮M ferric ammonium citrate or ferric epithelia (Nielsen et al., 1996; Stoesz et al., 1998). chloride), or the iron chelator deferroxamine (10 ␮M) for 48 hr. Trans- We now report that 24p3 and its human ortholog Ngal ferrin receptor1 and ferritin were measured by immunoblot (20 ␮g/ are iron-trafficking proteins. The evidence for iron asso- lane). 24p3/Ngal suppressed transferrin receptor1 but enhanced the ciation includes the purification of 59Fe-24p3 from UB expression of ferritin, just like iron loading with transferrin or free cells that were metabolically labeled with 59Fe, the bind- iron. Deferoxamine had the opposite effect. Ferritin expression un- ing of iron to pure preparations of 24p3, and the struc- der basal conditions could be seen after prolonged exposure of the immunoblot. tural studies of Goetz et al. (2002). Evidence for 24p3/ Ngal mediated iron delivery includes time- and temperature-dependent radiolabeling of cells by 59Fe-24p3/Ngal in a saturable fashion, traffic of these proteins through fluorescently labeled proteins to kidneys at different acidified endosomes with divalent metal transporters, stages of development. In all cases, transferrin was con- and the subsequent modulation of iron-responsive centrated by mesenchymal cells adjacent to the tips of genes. Trafficking by 24p3/Ngal was distinct from trans- the invading ureteric bud (Figure 6A, green). These cells ferrin at the subcellular level, and their pathways were are late-staged epithelial progenitors as judged by their even more divergent in embryonic kidney, where they location in the developing kidney as well as by their targeted different cells. 24p3/Ngal is expressed by the intense expression of transcription factors (e.g., Pax-2, ureteric bud in the embryonic kidney, suggesting that Figure 6B; Rothenpieler and Dressler, 1993). Their prog- it regulates local transfer of iron between renal cells, eny, the renal tubular epithelia, also prominently incor- while transferrin transfers iron from the circulation. porated transferrin (Figure 6A). These cells highly ex- These data show that 24p3/Ngal is an iron-trafficking press the transferrin receptor1 (J.B., unpublished data). protein, with functions distinct from those of transferrin. 24p3, in contrast, had the reverse distribution. It predominantly targeted peripheral cells (Figures 6A and Subcellular Trafficking of 24p3/Ngal 6C), including epithelial progenitors at an early stage of 24p3/Ngal loaded organelles were larger than transferrin the lineage (Pax-2ϩ;Pϩ nuclei in Figure 6C) and stroma vesicles, suggesting that they were lysosomes. How- (PϪ nuclei; data not shown). To confirm the distinct tar- ever, this was ruled out by the recovery of intact 24p3/ geting of transferrin and 24p3/Ngal, we explanted meta- Ngal from cell lines and metanephric mesenchyme. In nephric mesenchyme (rat E13) and again found that trans- addition, most internalized 24p3/Ngal failed to colocal- ferrin targeted late epithelial progenitors (Yang et al., ize with dextran, which marks lysosomes. Alternatively, 2002) while 24p3 targeted peripheral cells (Figure 6D). the 24p3/Ngal vesicles are late endosomes. Evidence Distinct trafficking in the metanephric mesenchyme includes the lack of association with early endosomal implies that 24p3/Ngal and transferrin have different markers and a minor, but discernable, overlap with late functions. 24p3/Ngal (50 ␮g/ml; 48 hr) stimulated mod- markers. In addition, 24p3/Ngal colocalized with DMT1 est growth of early progenitors (Figure 6E) but failed to iron transporters, which are found in late endosomes produce morphologically evident epithelia, unless trans- (Tabuchi et al., 2000). These compartments are more ferrin was added after 24p3/Ngal (data not shown). On acidic than the early recycling vesicles, perhaps more the other hand, treatment of the metanephric mesen- closely matching the pH sensitivity of 24p3/Ngal. Traf- chyme with transferrin (100 ␮g/ml) failed to stimulate ficking in acidic endosomes is very important for the growth (Figure 6F) or epithelial conversion, and the sub- removal of iron from catecholate siderophores because sequent addition of 24p3/Ngal was ineffectual. Hence, it converts ferric-catecholate to ferric-salicylate, which these proteins provide stage specificity to the iron re- is a substrate for reduction by a NADH or NADPH cou- quirement by targeting different cells in the epithelial pled ferrireductase. Reduction results in the discharge lineage. This results in complementary activities that act of iron, because Fe2ϩdoes not bind the acidified sidero- Molecular Cell 1052

Figure 6. 24p3 and Transferrin Target Different Rat Kidney Cells (A) Confocal shows that mesenchymal cells and their epithelial progeny near the ureteric bud (UB) take up increasing amounts of Alexa488- transferrin (green). (B) These cells express high levels of Pax-2 (red, nuclei), typical of late preepithelial development. (C) In contrast, Alexa568-24p3 (red) is taken up by cells throughout the kidney, particularly at the cortical margin and between branches of the ureteric bud. These cells include Pax-2ϩ epithelial precursors (Pϩ, green nuclei) and Pax-2Ϫ (PϪ) stroma (data not shown). (D) Cultured metanephric mesenchyme contains clusters of late-staged Pax-2ϩ cells (Yang et al., 2002) that incorporate transferrin (green), whereas 24p3 (red) is taken up by peripheral cells. (E) 24p3 without transferrin stimulates 5Ј-bromodeoxyuridine incorporation by peripheral cells, whereas transferrin without 24p3 had no effect on cellular proliferation. (F) Confocal 20 ␮m projection. (A–C) Bars, 150 ␮m. (D–F) Bars, 100 ␮m. phore (Ward et al., 1999a). In fact, ferrireductase activi- labeled at steady state by this protein will further define ties have been found in late endosomes (Scheiber and the mechanism of iron delivery. It is notable that lactofer- Goldenberg, 1993). In addition, siderophores such as rin, a second iron transport protein with a very acidic iron-hydroxamates are also sensitive to reduction (Ward pH sensitivity (Nicholson et al., 1997), also traffics to et al., 1999a). Hence, it is plausible that 24p3/Ngal traf- late endosomes (Birgens et al., 1988). fics to late endosomes where acidification and reduction unload iron. Structural analysis of acidified and reduced Stage Specificity in Iron Capture 24p3/Ngal will clarify this mechanism. Further, a kinetic Developing and mature organs have multiple cellular description of the endocytic pathway of 24p3/Ngal, as compartments. These include progenitors, transit ampli- well as the isolation and identification of the organelles fying cells, and mature progeny. The least developed Iron Traffic in Embryonic Kidney 1053

epithelial progenitors and associated stroma incorpo- version by regulating the cytoplasmic pool of iron in rated 24p3/Ngal but not transferrin, while later staged early progenitors. This idea is plausible because 24p3/ epithelial progenitors (akin to transit amplifying cells) Ngal delivers sufficient iron to regulate iron-responsive took up transferrin but less 24p3/Ngal. This demon- genes (Figure 5). In fact, epithelial induction by 24p3/ strates that iron capture in an embryonic organ is stage Ngal was markedly reduced by substituting galliumϩ3 specific. This is likely to be a general principle because for ironϩ3, whereas induction by LIF (Barasch et al., 1999; iron transporters and receptors appear to be expressed Plisov et al., 2001) was not blocked by an equal quantity in a stage-specific manner at many sites. Transferrin of gallium (data not shown). Hence, iron delivery may receptor1 is expressed in intestinal progenitors, while be an essential component of 24p3/Ngal-mediated in- DMT1 and ferroportin are expressed by their progeny duction. Conversely, Fe-ionophores (enterobactin and at the duodenal villus tip (reviewed in Roy and Enns, Fe-PIH; Landschulz et al., 1984) could partially replace 2000). Likewise, transferrin receptor1 is transiently ex- 24p3/Ngal and induce small clusters of epithelia (data pressed in hematopoietic lineages (Chitambar et al., not shown). Because no such activity was found for 1983; Vaisman et al., 2000; Kanayasu-Toyoda et al., transferrin in our induction assay (even at 200 ␮g/ml), 1999), and the transporter lactoferrin is transiently ex- the ionophores must have targeted the same cells as pressed in the epithelia of the upper gut and airway 24p3/Ngal. These data suggest that iron itself is a signal (Ward et al., 1999b). for epithelial development in early progenitors. The targeting of primordial renal cells by 24p3/Ngal In sum, the distinct function of transferrin and 24p3/ and their progeny by transferrin explains the sequential Ngal in our model argues that iron delivery is not only actions of these proteins. 24p3/Ngal maintained viable mediated by different proteins but also has different mesenchymal cells in the absence of transferrin but functions at different cellular stages. These ideas are could not convert them to epithelia without the subse- supported by the diverse actions of iron in cells that are quent addition of transferrin. Conversely, transferrin had lineally related including crypt and villous cells in the no obvious action on the metanephric mesenchyme duodenum (Roy and Enns, 2000), hematopoietic stem without prior stimulation by 24p3/Ngal. These findings cells and their different progeny (Kanayasu-Toyoda et are consistent with those of Ekblom et al. (Ekblom et al., 1999), fibroblasts and adipocytes (Festa, et al., 2000), al., 1983; Landschultz et al., 1984) who demonstrated and glial restricted precursors and differentiated oligo- that while transferrin was not required for early stages dendrocytes (Morath and Mayer-Proschel, 2001). In all of induction or cell conversion (Ekblom et al., 1983), of these examples, the earliest stages of cell determina- transferrin was required after the mesenchyme was ex- tion were sensitive to iron acquisition, whereas later posed to an inducing factor. Similarly, hematopoietic stages were not sensitive. Finally, the surprising speci- precursors (Sposi et al., 2000; Kanayasu-Toyoda et al., ficity of the transferrin receptor1 knockout supports the 1999; Wu et al., 1999) must be stimulated by inducing contention that both the delivery of iron and its functions factors to respond to transferrin. Hence, the exclusive are cell type and cell stage dependent. expression of transferrin receptor1 in late progenitors (J.B., unpublished data) could result from antecedent Experimental Procedures stimulation by 24p3/Ngal. Thus, transferrin stimulates determined cells to complete their developmental pro- Purification, Identification, and Cloning of Lipocalin (24p3 and Ngal) gram, while 24p3/Ngal acts on earlier populations. E13 rat metanephric mesenchymes were separated from the ureteric bud and cultured in serum-free basal media that included FGF-2 Functions of Iron Delivery (50 ng/ml), TGF␣ (10 ng/ml; Karavanova et al., 1996), and transferrin The stage-specific capture of different iron transporters (5 ␮g/ml). Basal media failed to convert the cells. is likely to reflect changing cellular requirements, and, Forty liters of media conditioned by UB cells (Barasch et al., 1999) were desalted and fractionated by heparin-Sepharose (buffer A, 10 conversely, iron delivery might stimulate different re- mM NaPO4 [pH 7.0]; buffer B, buffer A ϩ 2 M NaCl; activity eluted sponses at different stages. Transferrin-iron stimulates at 0.25 M NaCl) followed by cation exchange (buffer A:50mM HEPES the proliferation of many different cell types (Trowbridge [pH 7.6]; buffer B: buffer A ϩ 0.5 M NaCl; activity eluted at 0.1 M and Shackelford, 1986), and transferrin-gallium (an iron NaCl), chromatofocusing (buffer A, 25 mM Triethanolamine [pH antagonist) results in cell cycle arrest (Chitambar et al., 10.5]; buffer B: Pharmalyte 1:45 [pH 9.0]; activity eluted at pH 9.5– 1983; Brodie et al., 1993) and suppression of cyclins, 9.7), and hydrophobic interaction chromatography (buffer A, 1.7 M NH4SO4, 50 mM Tris base [pH 8.5]; buffer B, 50mM Tris base [pH bcl-2, and myc families (Alcantara, et al., 2001). The 8.5]; activity eluting at 50% buffer B) to yield a single protein (Biorad inductive activity of 24p3/Ngal, however, cannot be ex- silver stain) that induced epithelia when added to the basal media. plained by an effect on proliferation alone. Instead, one Trypsin peptides of this protein were fractionated by Poros50 R2 possibility is that iron induces early cells of the mesen- RP micro-tips (Erdjument-Bromage et al., 1998) and analyzed by chymal lineage to epithelialize. Iron regulates many matrix-assisted laser-desorption/ionization reflectron time-of-flight genes including those not directly involved in iron me- (MALDI-reTOF) MS using a Reflex III instrument (Bru¨ ker Daltonics, Bremen, Germany). Peptides were also analyzed by electrospray tabolism by both posttranscriptional (iron response ele- ionization (ESI) MS/MS on an API 300 triple quadrupole instrument ments; Rouault and Klausner, 1997; Lieu et al., 2001) (Applied Biosystems/MDS Sciex, Thornhill, Canada) modified with and transcriptional mechanisms (iron-dependent tran- an ultrafine ionization source (Geromanos et al., 2000). Selected scription factors; Yamaguchi-Iwai et al., 1996). Some of mass values from MALDI-TOF experiments were taken to search these proteins are critical for development, including the protein nonredundant database (NR; NCBI, Bethesda, MD) using PKC-␤, p21 (Alcantara et al., 1994; Gazitt et al., 2001) PeptideSearch (Mann et al., 1993). MS/MS spectra were inspected for y″ ion series and compared with the computer-generated series and eya, which produces the brachio-oto-renal malfor- from predicted tryptic peptides. mation when deleted (Ye and Connor, 1999). These ex- Ngal-MMP9 heterodimers (Roche) were dissociated by gel filtra- amples suggest that 24p3/Ngal could induce cell con- tion with DTT (5 mM). Mouse gst-24p3 was cloned, expressed (Bund- Molecular Cell 1054

gaard et al., 1994), and 24p3 derived by thrombin cleavage. Proteins by the NIH (DK 55388, DK 58872). P.T. is supported by NCI Core were gel filtered and then authenticated by total amino acid analysis. Grant P30 CA08748. Dimeric and monomeric Ngal were produced by C87S mutagenesis (50 ␮g/ml; Goetz et al., 2002). Each protein was tested on mesen- Received: April 9, 2002 chyme (50 ␮g/ml). Revised: September 30, 2002 Rabbits were immunized with recombinant gel-filtered 24p3. Sera were incubated with gst-Sepharose4B beads to remove contami- References nants. Frozen sections (2% paraformaldehyde fixed) and immunoblots (3% milk/Tris-buffered saline/0.05% Tween20) were stained Alcantara, O., Obeid, L., Hannun, Y., Ponka, P., and Boldt, D. (1994). at 1:100 and 1:500, respectively. Antisense and sense digoxigenin- Regulation of protein kinase C (PKC) expression by iron: effect of labeled riboprobes were synthesized (Promega) from 24p3-pCR- different iron compounds on PKC-beta and PKC-alpha gene expres- TOPOBLUNT (In Vitrogen). sion and role of the 5Ј-flanking region of the PKC-beta gene in the response to ferric transferrin. Blood 84, 3510–3517. Alcantara, O., Kalidas, M., Baltathakis, I., and Boldt, D. (2001). Ex- Iron and Protein Trafficking pression of multiple genes regulating cell cycle and apoptosis in Cells were grown in MEM, 10% FCS to 70% confluence on glass differentiating hematopoietic cells is dependent on iron. Exp. Hema- or for 7 days after reaching 100% confluence on transwell filters. tol. 29, 1060–1069. BF-2/Foxd1ϩ cells were grown in low glucose DMEM, 10% FCS. UB cells were radiolabeled (3 days) with 59Fe (50 ␮Ci) or 35S-methio- Anderson, R., Falck, J., Goldstein, J., and Brown, M. (1984). Visual- nine (100 ␮Ci). Media were clarified and protein A beads carrying ization of acidic organelles in intact cells by electron microscopy. anti-24p3 or preimmune sera were added (2 hr, 4ЊC). Beads were Proc. Natl. Acad. Sci. USA 81, 4838–4842. then collected without centrifugation and washed (PBS X3; Anderson, B.F., Baker, H.M., Norris, G.E., Rumball, S.V., and Baker, PBSϩ0.3% Triton X-100 X3). 59Fe-conditioned media was also frac- E.N. (1990). Apo-lactoferrin structure demonstrates ligand-induced tionated by heparin-Sepharose (see above) and compared with free conformational change in transferrins. Nature 344, 784–787. 59Fe. Purified UB 24p3 was dialyzed with deferoxamine (10 ␮M) then Barasch, J., Pressler, L., Connor, J., and Malik, A. (1996). A ureteric water (Specialty Media), and incubated with 59Fe Ϯ cold iron (NaCl bud cell line induces nephrogenesis in two steps by two distinct 59 10 mM, HEPES 5 mM, NaHCO3 2 mM [pH 7.2]). Fe-protein was signals. Am. J. Physiol. 271, F50–F61. detected after rapid filtration on PD-10 columns. Only the first half Barasch, J., Yang, J., Ware, C., Taga, T., Yoshida, K., Erdjument- 59 of the desalted protein was collected (free- Fe negative). Bromage, H., Tempst, P., Parravicini, E., Malach, S., Aranoff, T., 59 ␮ 59 ␮ 59 Fe-Ngal (105,000 cpm/ g), Fe-24p3 (98,000 cpm/ g), or Fe- and Oliver, J. (1999). Mesenchymal to epithelial conversion in rat ␮ transferrin (91,000 cpm/ g) was purified by PD-10 columns followed metanephros is induced by LIF. Cell 99, 377–386. by overnight dialysis and added to cells for 0.3–12 hr in serum-free Birgens, H., Kristensen, L., Borregaard, N., Karle, H., and Hansen, MEM (37ЊCor4ЊC). Canine or human transferrin was used for MDCK N. (1988). Lactoferrin-mediated transfer of iron to intracellular ferritin and mouse or human transferrin (Sigma) was used for other cells. in human monocytes. Eur. J. Haematol. 41, 52–57. Cells were washed in five cycles (pH 5.0, pH 8.0) and 0.5 M NaCl once. Bowman, E., Siebers, A., and Altendorf, K. (1988). Bafilomycins: 125I-transferrin (NEN), 125I- Ngal, and 125I-␣2macroglobulin (Sigma) a class of inhibitors of membrane ATPases from microorganisms, were prepared with IodoBeads (Pierce). 35S-24p3 used 35S-methio- animal cells, and plant cells. Proc. Natl. Acad. Sci. USA 85, 7972– nine (1.5 mCi) for 2 hr prior to IPTG induction. Cell lines were incu- 7976. bated continuously with probes or loaded (2 hr), washed, and release Brodie, C., Siriwardana, G., Lucas, J., Schleicher, R., Terada, N., detected 1 hr later. Media was treated with TCA (7%) and DOC Szepesi, A., Gelfand, E., and Seligman, P. (1993). Neuroblastoma (0.125%). Transepithelial flux was tested by adding gst-24p3 (200 sensitivity to growth inhibition by deferrioxamine: evidence for a ␮g/ml) to apical or basal media of confluent cells followed by anti- block in G1 phase of the cell cycle. Cancer Res. 53, 3968–3975. gst immunoblot (1:1000). Bundgaard, J., Sengelov, H., Borregaard, N., and Kjeldsen, L. (1994). Cells were labeled with 5–20 ␮g/ml of 24p3- or Ngal-Alexa 568 Molecular cloning and expression of a cDNA encoding Ngal: a lipo- or 488 (Molecular Probes) for 3 hr (37ЊC, 4ЊC) Ϯ 20-fold molar excess calin expressed in human neutrophils. Biochem. Biophys. Res. Com- of unlabeled protein. Alternatively, gst-24p3 (20 ␮g/ml) was used. mun. 202, 1468–1475. Early endosomes were detected with anti-EEA (1:500; BD) or tran- Cassia, R., Besnard, L., Fiette, L., Espinosa de los Monteros, A., sient transfection with Rab-5-GFP (gift from S. Greenberg) by lipo- Ave, P., Py, M., Huerre, M., de Vellis, J., Zakin, M., and Guillou, F. fectamine-plus. Late endosomes were detected by transfection with (1997). Transferrin is an early marker of hepatic differentiation, and Rab-11-HA and anti-HA (1:250; Babco; gift of S. Greenberg). Prely- its expression correlates with the postnatal development of oligo- sosomes were detected with anti-mannose 6 phosphate receptor dendrocytes in mice. J. Neurosci. Res. 50, 421–432. (1:500 in 0.1% NP-40; Calbiochem), and lysosomes were detected with fluorescent-dextran (3 hr) followed by overnight washout. Cells Chitambar, C., Massey, E., and Seligman, P. (1983). Regulation of were also characterized with anti-gst (1:100; Pharmacia), anti- transferrin receptor expression on human leukemic cells during pro- DMT1 (1:100; Alpha Diagnostic), anti-Pax2 (1:200; Zymed), anti- liferation and induction of differentiation. Effects of gallium and di- ␤-galactosidase (1:2000; Biogenesis), or anti-bromodeoxyuridine methylsulfoxide. J. Clin. Invest. 72, 1314–1325. (manufacturer’s instructions; Amersham) after 5Ј bromodeoxyuri- Clague, M., Urbe, S., Aniento, F., and Gruenberg, J. (1994). Vacuolar dine uptake (10 ␮M; 1 hr), or anti-dinitrophenol (1:50; Oxford Bio- ATPase activity is required for endosomal carrier vesicle formation.

chemical) after DAMP uptake (50 ␮M, 1 hr Ϯ 30 mM NH4Cl). Many J. Biol. Chem. 269, 21–24. hundreds of cells were examined by confocal microscopy (Zeiss, Cooper, C., and Porter, J. (1997). Ribonucleotide reductase, lipoxy- LSM software) and each section was examined for costaining. Pro- genase and the intracellular low-molecular-weight iron pool. Bio- jections (3–4 ␮m) from 0.5 ␮m sections of cells or projections (20 chem. Soc. Trans. 25, 75–80. ␮ ␮ m) from 2 m intervals of tissues are illustrated to show staining Dautry-Varsat, A., Ciechanover, A., and Lodish, H. (1983). pH and in its entirety. the recycling of transferrin during receptor-mediated endocytosis. Immunoblot for transferrin receptor1 (1:500; Zymed) and ferritin Proc. Natl. Acad. Sci. USA 80, 2258–2262. (1:200; Biogenesis) used 5% milk, 0.2% Tween20. Devireddy, L., Teodoro, J., Richard, F., and Green, M. (2001). Induc- tion of apoptosis by a secreted lipocalin that is transcriptionally Acknowledgments regulated by IL-3 deprivation. Science 293, 829–834. Ekblom, P., Thesleff, I., Saxen, L., Miettinen, A., and Timpl, R. (1983). We thank Q. Al-Awqati and G.M. Brittenham for review and the Transferrin as a fetal growth factor: acquisition of responsiveness referees for critical insights. J.B. is supported by a Research Grant related to embryonic induction. Proc. Natl. Acad. Sci. USA 80, 2651– (FY00165) from the March of Dimes Birth Defects Foundation and 2655. Iron Traffic in Embryonic Kidney 1055

Erdjument-Bromage, H., Lui, M., Lacomis, L., Grewal, A., Annan, R., Effect of granulocyte colony stimulating factor on differentiation and MacNulty, D., Carr, S.A., and Tempst, P. (1998). Micro-tip reversed- proliferation. J. Biol. Chem. 274, 25471–25480. phase liquid chromatographic extraction of peptide pools for mass Karavanova, I., Dove, L., Resau, J., and Perantoni, A. (1996). Condi- spectrometric analysis. J. Chromatogr. 826, 167–181. tioned medium from a rat ureteric bud cell line in combination with Fernandez-Pol, J.A. (1978). Isolation and characterization of a sider- bFGF induces complete differentiation of isolated metanephric mes- ophore-like growth factor from mutants of SV40-transformed cells enchyme. Development 122, 4159–4167. adapted to picolinic acid. Cell 14, 489–499. Kjeldsen, L., Johnsen, A., Sengelov, H., and Borregaard, N. (1993). Festa, M., Ricciardelli, G., Mele, G., Pietropaolo, C., Ruffo, A., and Isolation and primary structure of Ngal, a novel protein associated Colonna, A. (2000). Overexpression of H ferritin and upregulation of with human neutrophil gelatinase. J. Biol. Chem. 268, 10425–10432. iron regulatory protein genes during differentiation of 3T3-L1 pre- Landschulz, W., Thesleff, I., and Ekblom, P. (1984). A lipophilic iron adipocytes. J. Biol. Chem. 275, 36708–36712. chelator can replace transferrin as a stimulator of cell proliferation Flower, D., North, A., and Sansom, C. (2000). The lipocalin protein and differentiation. J. Cell Biol. 98, 596–601. family: structural and sequence overview. Biochim. Biophys. Acta Mann, M., Højrup, P., and Roepstorff, P. (1993). Use of mass spectro- 1482, 9–24. metric molecular weight information to identify proteins in data- Fuller, S., and Simons, K. (1986). Transferrin receptor polarity and bases. Biol. Mass Spectrom. 22, 338–345. recycling accuracy in “tight” and “leaky” strains of Madin-Darby Morath, D., and Mayer-Proschel, M. (2001). Iron modulates the differ- canine kidney cells. J. Cell Biol. 103, 1767–1779. entiation of a distinct population of glial precursor cells into oligo- Gazitt, Y., Reddy, S., Alcantara, O., Yang, J., and Boldt, D. (2001). dendrocytes. Dev. Biol. 237, 232–243. A new molecular role for iron in regulation of cell cycling and differen- Neilands, J. (1995). Siderophores: structure and function of microbial tiation of HL-60 human leukemia cells: iron is required for transcrip- iron transport compounds. J. Biol. Chem. 270, 26723–26726. tion of p21(WAF1/CIP1) in cells induced by phorbol myristate ace- tate. J. Cell. Physiol. 187, 124–135. Newcomer, M., and Ong, D. (2000). Plasma retinol binding protein: structure and function of the prototypic lipocalin. Biochim. Biophys. Geromanos, S., Freckleton, G., and Tempst, P. (2000). Tuning of an Acta 1482, 57–64. electrospray ionization source for maximum peptide-ion transmis- sion into a mass spectrometer. Anal. Chem. 72, 777–790. Nicholson, H., Anderson, B., Bland, T., Shewry, S., Tweedie, J., and Baker, E. (1997). Mutagenesis of the histidine ligand in human Goetz, D.H., Holmes, M.A., Borregaard, N., Bluhm, M.E., Raymond, lactoferrin: iron binding properties and crystal structure of the histi- K.N., and Strong, R.K. (2002). The neutrophil lipocalin NGAL is a dine-253 methionine mutant. Biochemistry 36, 341–346. bacteriostatic agent that interferes with siderophore-mediated iron acquisition. Mol. Cell 10, this issue, 1033–1043. Nielsen, B., Borregaard, N., Bundgaard, J., Timshel, S., Sehested, M., and Kjeldsen, L. (1996). Induction of Ngal synthesis in epithelial Gunshin, H., Mackenzie, B., Berger, U., Gunshin, Y., Romero, M., cells of human colorectal neoplasia and inflammatory bowel dis- Boron, W., Nussberger, S., Gollan, J., and Hediger, M. (1997). Clon- eases. Gut 38, 414–420. ing and characterization of a mammalian proton-coupled metal-ion transporter. Nature 388, 482–488. Nyholm, S., Mann, G., Johansson, A., Bergeron, R., Graslund, A., and Thelander, L. (1993). Role of ribonucleotide reductase in inhibition of Gustine, D., and Zimmerman, E. (1973). Developmental changes in mammalian cell growth by potent iron chelators. J. Biol. Chem. 268, microheterogeneity of foetal plasma glycoproteins of mice. Bio- 26200–26205. chem. J. 132, 541–551. Plisov, S., Yoshino, K., Dove, L., Higinbotham, K., Rubin, J., and Hamill, R., Woods, J., and Cook, B. (1991). Congenital atransferri- Perantoni, A. (2001). TGF beta 2, LIF and FGF2 cooperate to induce nemia. A case report and review of the literature. Am. J. Clin. Pathol. nephrogenesis. Development 128, 1045–1057. 96, 215–218. Hatini, V., Huh, S., Herzlinger, D., Soares, V., and Lai, E. (1996). Rothenpieler, U., and Dressler, G. (1993). Pax-2 is required for mes- Essential role of stromal mesenchyme in kidney morphogenesis enchyme-to-epithelium conversion during kidney development. De- revealed by targeted disruption of Winged Helix transcription factor velopment 119, 711–720. BF-2. Genes Dev. 10, 1467–1478. Rouault, T., and Klausner, R. (1997). Regulation of iron metabolism Hayashi, A., Wada, Y., Suzuki, T., and Shimizu, A. (1993). Studies on in eukaryotes. Curr. Top. Cell. Regul. 35, 1–19. familial hypotransferrinemia: unique clinical course and molecular Roy, C., and Enns, C. (2000). Iron homeostasis: new tales from the pathology. Am. J. Hum. Genet. 53, 201–213. crypt. Blood 96, 4020–4027. Herzlinger, D., Abramson, R., and Cohen, D. (1993). Phenotypic con- Sanchez, D., Ganfornina, M., and Bastiani, M. (1995). Developmental versions in renal development. J. Cell Sci. Suppl. 17, 61–64. expression of the lipocalin Lazarillo and its role in axonal pathfinding Huggenvik, J., Craven, C., Idzerda, R., Bernstein, S., Kaplan, J., and in the grasshopper embryo. Development 121, 135–147. McKnight, G. (1989). A splicing defect in the mouse transferrin gene Scheiber, B., and Goldenberg, H. (1993). NAD(P)H:ferric iron reduc- leads to congenital atransferrinemia. Blood 74, 482–486. tase in endosomal membranes from rat liver. Arch. Biochem. Bio- Landschulz, W., Thesleff, I., and Ekblom, P. (1984). A lipophilic iron phys. 305, 225–230. chelator can replace transferrin as a stimulator of cell proliferation Schubert, D., and LaCorbiere, M. (1985). Isolation of an adhesion- and differentiation. J. Cell Biol. 98, 596–601. mediating protein from chick neural retina adherons. J. Cell Biol. Levy, J., Jin, O., Fujiwara, Y., Kuo, F., and Andrews, N. (1999). Trans- 101, 1071–1077. ferrin receptor is necessary for development of erythrocytes and Sposi, N., Cianetti, L., Tritarelli, E., Pelosi, E., Militi, S., Barberi, T., the nervous system. Nat. Genet. 21, 396–399. Gabbianelli, M., Saulle, E., Kuhn, L., Peschle, C., and Testa, U. (2000). Lieu, P.T., Heiskala, M., Peterson, P.A., and Yang, Y. (2001). The Mechanisms of differential transferrin receptor expression in normal roles of iron in health and disease. Mol. Aspects Med. 22, 1–87. hematopoiesis. Eur. J. Biochem. 267, 6762–6774. Jones, R., Peterson, C., Grady, R., and Cerami, A. (1980). Low molec- Stoesz, S., Friedl, A., Haag, J., Lindstrom, M., Clark, G., Gould, M. ular weight iron-binding factor from mammalian tissue that potenti- (1998). Heterogeneous expression of the lipocalin Ngal in primary ates bacterial growth. J. Exp. Med. 151, 418–428. breast cancers. Int. J. Cancer. 79, 565–572. Kamarainen, M., Seppala, M., Virtanen, I., and Andersson, L. (1997). Tabuchi, M., Yoshimori, T., Yamaguchi, K., Yoshida, T., and Kishi, Expression of glycodelin in MCF-7 breast cancer cells induces dif- F. (2000). Human NRAMP2/DMT1, which mediates iron transport ferentiation into organized acinar epithelium. Lab. Invest. 77, across endosomal membranes, is localized to late endosomes and 565–573. lysosomes in HEP-2 cells. J. Biol. Chem. 275, 22220–22228. Kanayasu-Toyoda, T., Yamaguchi, T., Uchida, E., and Hayakawa, T. Trenor, C., Campagna, D., Sellers, V., Andrews, N., and Fleming, M. (1999). Commitment of neutrophilic differentiation and proliferation (2000). The molecular defect in hypotransferrinemic mice. Blood 96, of HL-60 cells coincides with expression of transferrin receptor. 1113–1118. Molecular Cell 1056

Trowbridge, I., and Shackelford, D. (1986). Structure and function of transferrin receptors and their relationship to cell growth. Bio- chem. Soc. Symp. 51, 117–129. Tycko, B., and Maxfield, F.R. (1982). Rapid acidification of endocytic vesicles containing ␣2-macroglobulin. Cell 28, 643–651. Vaisman, B., Meyron-Holtz, E., Fibach, E., Krichevsky, A., and Konijn, A. (2000). Ferritin expression in maturing normal human erythroid precursors. Br. J. Haematol. 110, 394–401. van Renswoude, J., Bridges, K., Harford, J., and Klausner, R. (1982). Receptor-mediated endocytosis of transferrin and the uptake of Fe in K562 cells: identification of a non-lysosomal acidic compartment. Proc. Natl. Acad. Sci. USA 79, 6186–6190. Ward, T., Lutz, A., Parel, S., Ensling, J., Gutlich, P., Buglyo, P., and Orvig, C. (1999a). An iron-based molecular redox switch as a model for iron release from enterobactin via the salicylate binding mode. Inorg. Chem. 38, 5007–5017. Ward, P., Mendoza-Meneses, M., Mulac-Jericevic, B., Cunningham, G., Saucedo-Cardenas, O., Teng, C., and Conneely, O. (1999b). Re- stricted spatiotemporal expression of lactoferrin during murine embryonic development. Endocrinology 140, 1852–1860. Wojnar, P., Lechner, M., Merschak, P., and Redl, B. (2001). Molecular cloning of a novel lipocalin-1 interacting human cell membrane receptor using phage display. J. Biol. Chem. 276, 20206–20212. Wu, K., Polack, A., and Dalla-Favera, R. (1999). Coordinated regulation of iron-controlling genes, H-ferritin and IRP2, by c-MYC. Sci- ence 283, 676–679. Yamaguchi-Iwai, Y., Stearman, R., Dancis, A., and Klausner, R. (1996). Iron-regulated DNA binding by the AFT1 protein controls the iron regulon in yeast. EMBO J. 15, 3377–3384. Yamashiro, D., Tycko, B., Fluss, S., and Maxfield, F. (1984). Segrega- tion of transferrin to a mildly acidic (pH 6.5) para-Golgi compartment in the recycling pathway. Cell 37, 789–800. Yang, J., Blum, A., Novak, T., Levinson, R., and Lai, E., and Barasch, J. (2002). An epithelial precursor is regulated by the ureteric bud and by the stroma. Dev. Biol. 246, 296–310. Ye, Z., and Connor, J. (1999). Screening of transcriptionally regulated genes following iron chelation in human astrocytoma cells. Biochem. Biophys. Res. Commun. 264, 709–713. Zhou, H., Ramachandran, S., Kim, J., Raynor, D., Rock, J., and Parthasarathy, S. (2000). Implications in the management of preg- nancy: II. Low levels of gene expression but enhanced uptake and accumulation of umbilical cord glycodelin. Fertil. Steril. 73, 843–847.