Release of the Soluble Transferrin Receptor Is Directly Regulated by Binding of Its Ligand Ferritransferrin*

The human transferrin receptor (TfR) is shed by an integral metalloprotease releasing a soluble form (sTfR) into serum. The sTfR reflects the iron demand of the body and is postulated as a regulator of iron homeostasis via binding to the hereditary hemochromatosis protein HFE. To study the role of transferrin in this process, we investigated TfR shedding in HL60 cells and TfR-deficient Chinese hamster ovary cells transfected with human TfR. Independent of TfR expression, sTfR release decreases with increasing ferritransferrin concentrations, whereas apo-transferrin exhibits no inhibitory effect. To investigate the underlying mechanism, we generated several TfR mutants with different binding affinities for transferrin. Shedding of TfR mutants in transfected cells correlates exactly with their binding affinity, implying that the effect of ferritransferrin on TfR shedding is mediated by a direct molecular interaction. Analysis of sTfR release from purified microsomal membranes revealed that the regulation is independent from intracellular trafficking or cellular signaling events. Our results clearly demonstrated that sTfR does not only reflect the iron demand of the cells but also the iron availability in the bloodstream, mirrored by iron saturation of transferrin, corroborating the important potential function of sTfR as a regulator of iron homeostasis.

The human transferrin receptor (TfR) is shed by an integral metalloprotease releasing a soluble form (sTfR) into serum. The sTfR reflects the iron demand of the body and is postulated as a regulator of iron homeostasis via binding to the hereditary hemochromatosis protein HFE. To study the role of transferrin in this process, we investigated TfR shedding in HL60 cells and TfR-deficient Chinese hamster ovary cells transfected with human TfR. Independent of TfR expression, sTfR release decreases with increasing ferritransferrin concentrations, whereas apo-transferrin exhibits no inhibitory effect. To investigate the underlying mechanism, we generated several TfR mutants with different binding affinities for transferrin. Shedding of TfR mutants in transfected cells correlates exactly with their binding affinity, implying that the effect of ferritransferrin on TfR shedding is mediated by a direct molecular interaction. Analysis of sTfR release from purified microsomal membranes revealed that the regulation is independent from intracellular trafficking or cellular signaling events. Our results clearly demonstrated that sTfR does not only reflect the iron demand of the cells but also the iron availability in the bloodstream, mirrored by iron saturation of transferrin, corroborating the important potential function of sTfR as a regulator of iron homeostasis.
The transferrin receptor (TfR) 3 is a homodimeric type II membrane protein that mediates iron uptake into the cell. Each subunit possesses one binding site for the iron carrier protein transferrin. The complex of ferritransferrin (ferri-Tf) and TfR is transported into the cell by clathrin-mediated endocytosis. Inside the acidic environment of the endosome, iron dissociates, and the complex of apo-transferrin and TfR recycles back to the plasma membrane, where apo-transferrin is released, and TfR is presented for a new uptake cycle.
The TfR consists of a small cytoplasmic domain, a single-pass transmembrane region for each subunit, and a large butterfly-shaped ectodomain, which is kept by a stalk at a distance of 2.9 nm from the plasma membrane (1). The crystal structure of the extracellular domain revealed that each homodimer has three structurally distinct domains: a protease-like domain following the stalk, a helical domain mediating dimer contact, and an apical domain oriented toward the outside (2).
Early approaches to identify the transferrin binding site on the TfR were based on chimeras of human TfR and the chicken counterpart that does not bind human transferrin. Mapping of the receptor-ligand interaction revealed the helical domain of the TfR as binding site (3). Another mutational study with soluble chimeras of TfR and alkaline phosphatase identified a critical RGD sequence for transferrin binding within the helical domain (4). This observation is consistent with a newer extensive mutational analysis wherein further important residues for transferrin binding were identified (5). The combined data from structural analyses with x-ray hydroxyl radical footprinting (6) and a subnanometer resolution by cryo-electron microscopy (7), as well as from the crystal structures of ferri-Tf (8,9) and the TfR ectodomain (2), demonstrated that the C-lobe of transferrin abuts against the helical domain, whereas the N-lobe is sandwiched between the membrane and the TfR ectodomain (7), indicating a binding to the stalk region of the TfR.
Besides the membrane-associated cellular TfR, a soluble form (sTfR) exists in human serum. It is released by proteolytic cleavage of the TfR C-terminal of Arg-100 within the stalk (1,10). The shedding process is primarily mediated by an integral membrane metalloprotease sensitive to the inhibitor TAPI-2 (11). The protease probably belongs to the disintegrin and metalloprotease family (ADAM) (11). Nevertheless, other proteases are also involved in cleavage at alternative sites (12,13).
In mammalian organisms, iron is required for the function of various proteins in fundamental biochemical reactions. Since iron also harbors a toxic potential, uptake into the body must be strictly controlled to preserve iron homeostasis. The divalent metal transporter DMT-1 mediates apical dietary iron absorption in the duodenum via villus cells. On the basolateral surface, iron is exported by the iron transporter ferroportin (also named iron-regulated protein IREG-1) and then bound to transferrin. The mechanism of how the enterocytes sense the iron demand of the body, however, is unknown. Three soluble proteins or peptides are believed to be involved in the basolateral regulation of iron export from duodenal enterocytes to the serum, namely sTfR, transferrin, and hepcidin. The concentrations of these molecules in serum change depending on the iron status of the body. It remains unclear how they transfer the signal to the enterocyte. Recently, it has been shown that hepcidin binds to ferroportin and induces its internalization (14).
The sTfR mirrors the availability of functional iron independent of the iron stores in the body (15)(16)(17). In thalassemia major, the patients exhibit high sTfR levels (18) as well as high duodenal iron uptake, although the iron stores of the body are filled (19). Unlike other iron markers, sTfR is not affected by chronic inflammation or infection. The average sTfR level in normal subjects is 5 mg/liter and can vary from 1 ⁄ 8 to 20-fold of average if iron demand is changed (17). Since erythroid precursor cells contain 80% of total body TfR mass, the sTfR is directly correlated with the total mass of erythroid precursors (18). Nevertheless, not only erythroid precursors but all cells of the body are involved in sTfR production, in particular rapidly dividing cells and cells of the liver and placenta. Thus, the sTfR is more likely a marker for tissue iron deficiency and not only a determinant of erythroid precursor mass. Iron deficiency leads to an sTfR increase proportional to the severity of iron deficit (20).
New insights into the regulation of iron metabolism are gained from a common disease connected to iron homeostasis, the hereditary hemochromatosis type I. A mutation in the hemochromatosis protein (HFE) results in severe body iron overload and progressive organ failure. HFE has a direct influence on the iron concentration in reticuloendothelial cells (21). Furthermore, studies using the macrophage cell line THP-1 and the intestinal cell line HT-29 reveal an effect of HFE on iron export rather than iron uptake (22,23), suggesting an inhibitory effect of HFE on ferroportin function. Other authors propose a function of HFE in the regulation of hepcidin expression in the liver, which in turn could regulate iron uptake from the gut (reviewed in Ref. 24).
It has been observed that sTfR as well as TfR bind to HFE in tissues and cell lines (25)(26)(27), suggesting that the inhibitory activity of HFE on iron export is regulated by sTfR. Since sTfR reflects the iron demand of the body and has the ability to bind HFE, it is discussed in recent years as the erythroid regulator of iron homeostasis (24,28). A regulation of TfR shedding by extracellular compounds is, however, unknown to date. Since iron saturation of transferrin reflects iron availability, the release of sTfR from the cell surface may be controlled by transferrin. Therefore, in the present study, we analyzed the effect of transferrin on TfR shedding.

EXPERIMENTAL PROCEDURES
Reagents and Antibodies-Mouse monoclonal antibody OKT9 directed against the extracellular domain of TfR was prepared from a hybridoma cell line as described previously (29). Polyclonal rabbit antibody pAB063 was generated by immunization with purified human placental TfR (in cooperation with R. Gessner, Charité, Campus Virchow-Klinikum, Berlin, Germany). Horseradish peroxidase-labeled anti-mouse and anti-rabbit antibodies were obtained from Dako A/S (Glostrup, Denmark), FITC-labeled goat anti-mouse antibody was from Alexis (Grünberg, Germany), ferri-Tf was from Sigma, blasticidin S and Dulbecco's PBS were from Invitrogen, FuGENE 6 was from Roche Applied Science, NHS-Sepharose was from Amersham Biosciences, and other materials were purchased from Sigma.

Construction of TfR Mutants-
The open reading frame of the TfR-cDNA was cloned from pGEM1-TR (kindly provided by Marino Zerial, Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany) into AgeI-deficient pcDNA6/V5-His B (Invitrogen). The mutations within the RGD sequence of the TfR were generated by exchange of the bases cgt (nucleotides ϩ1936 to ϩ1938) to aaa (amino acid R646K), g (nucleotide ϩ1940) to c (amino acid G647A), and c (nucleotide ϩ1944) to g (amino acid D648E) using the QuikChange site-directed mutagenesis kit (Stratagene Europe, Amsterdam, Netherlands) and the following primer pairs (reverse primers were reverse complementary to forward primers): R646K (5Ј-CAG TGG CTG TAT TCT GCT AAA GGA GAC TTC TTC CGT GCT AC-3Ј), G647A (5Ј-GCT GTA TTC TGC TCG TGC AGA CTT CTT CCG TGC-3Ј), and D648E (5Ј-GTA TTC TGC TCG TGG AGA GTT CTT CCG TGC TAC TTC-3Ј).
Cell Culture and Transfection of CHO-TRVb Cells-CHO-TRVb cells, designated hereafter as TRVb, do not express functional TfR on their cell surface (kindly provided by Timothy E. McGraw, Cornell Uni-versity, Ithaca, NY). TRVb and human myeloid leukemia HL60 cells (the German Resource Centre for Biological Material) were cultured in a 5% CO 2 atmosphere and 37°C in RPMI (HL60) or Dulbecco's modified Eagle's medium (TRVb) with Glutamax supplemented with 100 units/ml penicillin, 100 g/ml streptomycin, and 10% fetal calf serum. Transfection of TRVb cells was performed with FuGENE 6 according to the manufacturer's instructions, and cells were selected in the presence of 10 g/ml blasticidin S.
Detection of sTfR from Cell Culture Medium-HL60 cells (10 6 /ml and 10 7 /ml) or TRVb cells (10 6 /ml) transfected with TfR were seeded in 35-mm dishes and incubated for 12 h (i) in the absence or presence of 200 M desferal in 1.5 ml of medium containing 200 g/ml ferri-Tf or (ii) in the presence of ferri-Tf (0 -1000 g/ml) in medium without fetal calf serum. After incubation, the cells were solubilized in 750 l of PBS, 1% Triton X-100. For kinetic studies of sTfR release, 5 ϫ 10 4 cells/ml were seeded and cultured for 1-4 days (final cell number 1.5 ϫ 10 6 /ml) in ferri-Tf-containing medium. The sTfR in the cell culture supernatant and TfR in the cell lysates were quantified in a TfR-specific ELISA as described previously (11). Briefly, 96-well microplates were coated with 200 ng of OKT9 in PBS for 90 min, blocked with 10% fetal calf serum and 3% bovine serum albumin in PBS for 30 min, and incubated for 2.5 h with cell supernatants or cell lysates (dilution 1:10 to 1:100), which were centrifuged before (20,800 ϫ g, 4°C, 20 min) to remove cellular debris. Bound TfR (or sTfR) was quantified in a colorimetric reaction using tetramethylbenzidine after detection with anti-TfR polyclonal antibody pAB063 (1:1000 in PBS, 0.05% Tween 20) and peroxidase-labeled swine anti-rabbit IgG (1:2000 in PBS, 0.05% Tween 20). The purification of placental TfR, which served as internal standard for the quantitation of sTfR, and colorimetric reaction were performed as described (30).
Detection of sTfR Release from Membranes-After removal of the medium, transfected TRVb cells were placed on ice, washed twice with 10 ml of PBS, and then incubated for 15 min with 5 ml of 1:10 diluted PBS. The cells were homogenized by using a Dounce homogenizer 30 times and differentially centrifuged at 500 ϫ g for 15 min followed by 2600 ϫ g for 15 min and finally 100,000 ϫ g for 30 min. The pellet from the last centrifugation step containing microsomal membranes was washed once in 1 ml of PBS and centrifuged at 20,000 ϫ g for 20 min. The membrane pellet was resuspended in PBS to a final concentration of 1 mg/ml total protein, and 35-l aliquots were incubated at 37°C for 12 h in the presence or absence of 10 g of ferri-Tf. The samples were then centrifuged at 20,000 ϫ g for 20 min at 4°C and sTfR in the supernatants analyzed in the TfR-specific ELISA as described above.
Depletion of sTfR from Ferri-Tf Solution-OKT9 antibody (650 g) was covalently bound to NHS-Sepharose as described by the manufacturer. Ferri-Tf solution (4 mg/ml containing 46 ng/ml sTfR as measured in TfR-ELISA) was incubated overnight with OKT9-NHS-Sepharose. After removal of the Sepharose, neither OKT9 (tested by ELISA with rabbit anti-mouse antibody) nor sTfR (tested by TfR-ELISA) was detectable (detection limit for both 0.08 ng/mg of ferri-Tf).
Detection of TfR Located at the Cell Surface of Transfected TRVb Cells-After washing once with PBS, transfected TRVb cells were detached from the plate by PBS, 0.1% EDTA. The suspended cells (3 ϫ 10 6 /sample) were blocked with PBS, 1% goat serum and incubated with 1 g/ml IgG 1 (isotype control) or 1 g/ml OKT9. After washing three times, the cells were stained with FITC-labeled goat anti-mouse IgG (1:50 in blocking solution) and washed again. Each incubation step was performed for 30 min at 4°C. Finally, cells were resuspended in 150 l of PBS and TfR on the cell surface determined by flow cytometry (FACScalibur, BD Biosciences) and CellQuest software.
Binding of Ferri-Tf to Transfected TRVb Cells-After washing, detaching, and blocking as described above, 3 ϫ 10 6 transfected TRVb cells were incubated with 4 g/ml ferri-Tf-FITC (generated by labeling ferri-Tf using the FluoReporter FITC protein labeling kit from Molecular Probes, Leiden, Netherlands) for 30 min at 4°C. Untransfected TRVb cells served as control. After washing three times, cells were resuspended, and bound ferri-Tf was quantitated by flow cytometry as described above.
Determination of the Equilibrium Dissociation Constants of the Ferri-Tf-TfR Complex-Equilibrium dissociation constants for the complex of TfR mutants and ferri-Tf were determined using Liliom plot analyses (31) based on data from a receptor-ligand ELISA (32). Briefly, 96-well microplates were coated with 500 ng of ferri-Tf in PBS for 90 min, blocked with PBS, 10% fetal calf serum, 3% bovine serum albumin for 30 min, and incubated for 2.5 h with lysates of TfR-transfected TRVb cells in concentrations between 55 pM and 55 nM (determined using the TfR-specific ELISA). Bound TfR was detected as described for the TfRspecific ELISA. The absorbance (recorded at 450 nm) at maximal binding was derived from a plot of the TfR concentration versus absorbance by curve fitting. K D values were determined as the reciprocal value of the slope in a Liliom plot (c/i versus 1/(1 Ϫ i), where c is the TfR concentration and i the quotient of the absorbance at a given TfR concentration and the absorbance at maximal binding) (31).

Influence of Ferri-Tf on TfR
Shedding-Since the sTfR concentration in human serum mirrors the iron demand of the body, it can be expected that TfR shedding is regulated by proteins that reflect the actual iron status. The concentration of iron-loaded transferrin (ferri-Tf) is high when available body iron is abundant. Since the sTfR concentration is reduced in this case, we raised the hypothesis that ferri-Tf has an inhibitory effect on TfR shedding.
To test this hypothesis, we measured sTfR release into serum-free culture medium from HL60 cells, a human promyelocytic cell line generally used to study TfR shedding (11,33,34). Since we found that commercially available ferri-Tf (Sigma) is contaminated with sTfR (12 ng/mg of ferri-Tf), which interferes with the measurement, we first completely removed the sTfR from ferri-Tf by treatment with OKT9-NHS-Sepharose. Using an ELISA developed in our laboratory (11), we specifically determined both the amount of released sTfR and the amount of cellular TfR. To depict alterations in sTfR release independent of possible changes in cellular TfR during ferri-Tf incubation, we calculated the amount of released sTfR as a percentage of cellular TfR. Our results demonstrate that with increasing sTfR-free ferri-Tf concentrations, TfR shedding significantly (p values less than 0.05, in most cases less than 0.01) decreases gradually, independent of the cellular TfR expression (Fig. 1A). Moreover, within the time interval of the experiment, the expression of cellular TfR is virtually unchanged. Only a slight scattering of the values from a minimum of 88 to a maximum of 113% TfR expression (average 103% Ϯ 8) as compared with the control was observed, which was clearly independent from the ferri-Tf concentration (correlation coefficient of 0 -0.05). Thus, ferri-Tf has the ability to directly or indirectly inhibit TfR shedding; maximal inhibition (36 -48%) is attained at 200 g/ml ferri-Tf on 10 6 cells/ml. Since the ferri-Tf level sufficient for maximal inhibition is expected to be dependent on the total amount of cellular TfR, we performed the assay additionally with a 10-fold higher concentration of HL60 cells, which is in the range of the concentration of sTfR-producing cells in the blood. For a direct comparison, we normalized TfR shedding to the untreated control and plotted the sTfR release observed for 10 6 and 10 7 cells (Fig. 1B). Maximal inhibition of sTfR release is reached at 200 -400 g/ml for 10 6 and above 1000 g/ml for 10 7 cells. Thus, the concentration for maximal inhibition shifted, as expected, to higher concentrations. Since significantly higher cell concentrations than 10 7 cells/ml cannot be applied in the chosen experimental design, an exact titration of cell number versus concentration cannot be performed. Nevertheless, the experiment demonstrated that the concentration interval, in which inhibition is subject to significant changes, is in the physiological range of ferri-Tf (150 -500 g/ml). The results are corroborated by the observation that treatment with the Fe 3ϩ chelator desferal counters the effect of ferri-Tf on TfR shedding (Fig. 1D), indicating that iron-free apo-transferrin, which has a low affinity to TfR, cannot inhibit TfR release.
For further studies, we used transfected TRVb cells, which enabled us to investigate TfR mutants. TRVb cells release high amounts of sTfR (13.3 fg/ml/cell as compared with 5.8 fg/ml/cell in HL60). Furthermore, they exhibit the same protease inhibition pattern on TfR shedding as human HL60 cells (13). Cells transfected with human TfR were analyzed by SDS-PAGE of the lysates and Western blotting with the anti-TfR antibodies H68.4 and OKT9 directed against the intracellular and extracellular domain of TfR, respectively, and shown to be clearly expressed with the expected molecular weight. The correct localization was proven by immunofluorescence studies. Permeabilized and non-permeabilized cells were stained with OKT9 and rhodamine-conjugated goat anti-mouse antibody. The receptor was exclusively localized on the cell surface and inside endosomes as expected (not shown).
To exclude a feedback effect on intracellular TfR translation during variation of extracellular iron supply, we transfected the TRVb cells with human TfR lacking all the iron-responsive elements 3Ј to the open read-  FEBRUARY 10, 2006 • VOLUME 281 • NUMBER 6 ing frame. Analogous to HL60 cells, TfR shedding in transfected TRVb cells significantly (p values less than 0.005) decreases with increasing ferri-Tf concentrations (Fig. 1C) and increases with decreased transferrin saturation generated by adding desferal (Fig. 1D). The observed effect is clearly independent of the cellular TfR expression, which exhibits only an unsteady variation of maximal 6% at the different ferri-Tf concentrations. As in HL60 cells (Fig. 1A), maximal inhibition (25-31%) is attained at ferri-Tf concentrations of 200 g/ml. The lower shedding rate per TfR (Fig. 1, compare A with C) is a result of 11 times higher TfR expression in transfected TRVb cells and not of a decrease in the amount of released sTfR.

Transferrin Inhibits Transferrin Receptor Shedding
Mammalian cells usually shed a number of different cell surface proteins. To show that the effect mediated by ferri-Tf is specific for the TfR, we additionally quantified the shedding of L-selectin in HL60 cells by an analogous ELISA. The amount of released soluble L-selectin unsteadily varied in a narrow range of 0.97-1.21 ng/ml for the different ferri-Tf concentrations with virtually no correlation (r ϭ 0.27), demonstrating that ferri-Tf does not affect shedding processes in general.
TfR Mutants with Decreased Affinity to Transferrin-To solve the question of whether the effect of ferri-Tf on TfR shedding is mediated by direct binding to the TfR or indirectly by binding to another component that mediates protease inhibition by intracellular signal transduction events or adapter proteins, we generated TfR mutants with different binding affinities for transferrin. Results from other authors show that mutations within the RGD sequence in its extracellular domain lead to altered dissociation constants for the ferri-Tf-TfR complex (see "Discussion"). Therefore, we introduced the conservative mutations R646K, G647A, and D648E into full-length TfR.
Expression and localization of mutant TfR was analyzed by immunofluorescence as described above for wild-type (WT) TfR and found to be identical (not shown). To quantify the cell surface expression of WT-TfR and TfR mutants on individual cells, TfR was labeled with OKT9 and FITC-conjugated goat anti-mouse antibody and analyzed by flow cytometry (Fig. 2, left panel). Dependent on the mutant, the percentage of cells that express TfR varied from 56% for WT to 73% for R646K (Fig.  2, left panel, right peaks). Cells not expressing TfR do not interfere since TRVb cells do not express endogenous TfR and are therefore unable to bind transferrin nor to release sTfR. The cells expressing WT-TfR, G647A, and D648E exhibit similar levels of expression on the cell surface; the R646K mutant possesses fewer TfRs per cell (Fig. 2B, left panel); however, this mutant has the highest number of cells that express TfR. Thus, the mutants behave like WT-TfR with respect to localization and distribution and are therefore suitable for the further studies. Differences in the total amount of TfR expression were taken into account by calculating the ratio of released sTfR and cellular TfR.
To demonstrate that the mutants indeed exhibit different binding affinities for transferrin, the cells were incubated with ferri-Tf-FITC and analyzed by flow cytometry. No ferri-Tf binding is detectable in untransfected cells, whereas cells expressing WT-TfR bind substantial amounts of ferri-Tf ( Fig. 2A, right panel, dashed and continuous line,  respectively). As expected, a part of the transfected cells do not bind ferri-Tf ( Fig. 2A, right panel, left peak of the continuous line), corresponding to the cells that do not express WT-TfR ( Fig. 2A, left panel, left  peak). To normalize the binding of ferri-Tf for the different mutants, we calculated the ratio between TfR expression and ferri-Tf binding. The D648E mutant behaves akin to WT-TfR (ratios of 3.9 and 4.1, respectively), whereas the binding of ferri-Tf to cells expressing R646K (ratio of 4.8) is slightly reduced and almost not detectable when transfected with G647A (ratio of 584).
After measuring the binding of ferri-Tf to transfected TRVb cells, we determined the affinity constants of the binding partners in vitro to confirm the results and to quantitate the binding, permitting a clear comparison with the amount of sTfR release. The dissociation constants were determined by a Liliom plot using binding data derived from a receptor-ligand ELISA. This analysis yields K D values of 7 and 26 nM for WT-TfR and the D648E mutant, respectively. The K D value was slightly increased to 93 nM for the R646K mutant and strongly elevated to 3160 nM for the G647A mutant. These results explicitly corroborated the binding data observed in living cells by fluorescence-activated cell sorter (compare K D values with ferri-Tf binding ratios in fluorescenceactivated cell sorter experiments). Due to their gradual reduction in ferri-Tf binding, the TfR mutants are optimal tools to examine the influence of transferrin binding on TfR shedding.
sTfR Release from Transfected Cells-Transfected TRVb cells were cultured in ferri-Tf-containing medium. TfR shedding from these cells was determined using the TfR-specific ELISA. The amount of sTfR in the culture supernatant increases in a time-dependent manner for all cell lines, but the extent of the increase differs (Fig. 3A). Nevertheless, the cell proliferation for all cell lines is nearly identical (Fig. 3B). Independent of TfR expression and time point, significantly higher amounts of sTfR were released by cells expressing the mutants R646K (2-fold) and G647A (3-fold) (Fig. 3, A and C). It is striking that the increase in sTfR release exactly correlates with the pK D of the ferri-Tf-TfR complex for all mutants (Fig. 3C, compare also with Fig. 2, right panel). At 12 h of incubation of 200 g/ml ferri-Tf, the shedding of the mutant with very low ferri-Tf binding affinity G647A is in the same range as the shedding of WT-TfR in the absence of ferri-Tf (compare Fig. 1C and 3C). This demonstrates that the effect on TfR shedding mediated by ferri-Tf is a consequence of a direct interaction of ferri-Tf with the TfR.
Nevertheless, it remains unclear whether the inhibitory effect of ferri-Tf on TfR shedding is mediated (i) by signal transduction events initiated by ferri-Tf binding, (ii) by preventing the receptor from colocalizing with the shedding protease due to impeding intracellular trafficking, or (iii) by directly preventing cleavage either by competition for the binding site or by inducing conformational changes that obstruct protease binding or at least cleavage after binding. To clarify the regulation, we analyzed TfR release in a membrane assay instead of from intact cells. As proved recently, this in vitro experiment with microsomal membranes is a suitable tool to examine TfR shedding with the advantage of excluding trafficking or signal transduction events (11). Purified membrane fractions of TfR expressing TRVb cells were incubated in the presence or absence of ferri-Tf. Both TfR and the shedding protease were located in the purified membranes and were still active under the assay conditions. Released sTfR was quantitated in the supernatant of the membranes by ELISA. The addition of ferri-Tf clearly reduces the amount of released WT-sTfR up to 46% (Fig. 4A). Thus, the inhibitory effect of ferri-Tf cannot be attributed to signal transduction in the cytosol or relocalization of the TfR in vicinity to the shedding protease but rather results from a direct inhibition. If the direct effect is a consequence of an interaction of ferri-Tf with the shedding protease, the inhibition of TfR shedding should not be affected in the TfR mutants. In contrast, if ferri-Tf competes for the binding site or induces allosteric inhibition, the blocking of TfR shedding by ferri-Tf should be decreased in the low affinity mutants. In Fig. 4B, the higher percentage represents a stronger inhibitory effect by ferri-Tf. Inhibition of TfR shedding in microsomal membranes is slightly reduced from 46% for WT-TfR to 38% for D648E and 34% for R646K, whereas shedding of the low affinity mutant G647A is clearly reduced to 23%. These values clearly mirror the corresponding binding affinities for ferri-Tf (inhibition is proportional to the pK A ). Thus, our results demonstrated that binding of ferri-Tf to the TfR down-regulates TfR shedding by blocking the binding of the shedding protease or by preventing cleavage due to induced structure alterations. In conclusion, an increased intracellular iron demand was manifested in elevated sTfR release due to constitutive shedding of an augmented amount of TfR, whereas extracellular iron availability mirrored by ferri-Tf antagonized TfR shedding.

DISCUSSION
In the present study, we investigated the influence of transferrin on TfR shedding. Due to its altered concentration in diseases of iron metabolism and its ability to bind HFE, sTfR is, together with hepcidin, one of the most discussed proteins involved in systemic iron regulation (24,28).
Transferrin saturation represents a reliable indicator for instantaneous iron availability of the organism; however, as known to date, it does not influence iron absorbance from the gut (22). We demonstrated that increasing ferri-Tf concentrations results in decreased TfR shedding rates in both HL60 and TRVb cells transfected with WT-TfR, whereas the iron chelator desferal has the opposite effect. The concentration of  . Effect of ferri-Tf on sTfR release from purified membranes. Purified membranes of transfected TRVb cells were incubated for 12 h at 37°C in the presence or absence of ferri-Tf. The release of sTfR was measured in three independent experiments in a TfR-specific ELISA. A, influence of ferri-Tf on sTfR release from purified microsomal membranes derived from TRVb cells transfected with WT-TfR. B, dark gray bars, the relative inhibitory effect of ferri-Tf on sTfR release from purified microsomal membranes containing WT-TfR (corresponding to the relative difference of the bars in panel A) and the different TfR mutants. Light gray bars, association constants (K A ) for the ferri-Tf-TfR complex (derived from Fig. 3C for better comparison). FEBRUARY 10, 2006 • VOLUME 281 • NUMBER 6 JOURNAL OF BIOLOGICAL CHEMISTRY 3301 ferri-Tf that blocks half of the sTfR release was in a range corresponding to a typical transferrin saturation in the transition from iron deficit to normal iron status. Our in vitro data correspond very well with in vivo analyses newly published by Brandao et al. (35), who find an sTfR decrease in human serum with increasing transferrin saturation as long as the saturation is below 25%. Maes et al. (36) showed that sTfR and iron inversely correlate to each other in humans. The inhibitory effect of ferri-Tf on TfR shedding is based on a mechanism independent of TfR transcription and translation. Under the applied conditions, varying ferri-Tf concentrations had, if any, a marginal effect on TfR expression in HL60 cells and, as expected, no effect in transfected TRVb cells due to the lack of iron-responsive elements. These observations are in agreement with earlier studies of Baynes (37), who showed that TfR expression first changes after 24 h of incubation, whereas no difference was detectable after 6 h. In the same study, the author showed that an increased ferri-Tf level decreases the ratio of sTfR release and TfR expression, which is consistent with our data. In HL60 cells, the author determined a decrease in sTfR release up to 59% after a 6-h incubation with 1320 pmol/liter iron as diferric transferrin (53 g/ml ferri-Tf). In contrast, Kohgo et al. (38) and Chitambar et al. (39) assumed an activating effect of ferri-Tf on TfR shedding or no influence, respectively. As proposed by Baynes (37), these findings could be a result of sTfR contamination of the ferri-Tf preparation. Our data supported this assumption as we observed an apparent high sTfR release exerted by contaminated ferri-Tf. We therefore depleted the sTfR from commercially available ferri-Tf by antibody purification.

Transferrin Inhibits Transferrin Receptor Shedding
To address the question of how transferrin mediates this effect, we generated TfR mutants with different affinities for ferri-Tf. Earlier studies showed that the exchange of glycine 647 to alanine (G647A) within the RGD motif results in a prominently decreased binding affinity of ferri-Tf for recombinant soluble forms of TfR expressed in insect cells (5,40). In contrast to the glycine residue, the influence of arginine and aspartate was only analyzed in chimeras in which the extracellular domain of the TfR was fused to alkaline phosphatase, resulting in a soluble reporter protein (4). In these chimeras, an R646K mutation led to reduced transferrin binding, whereas a D648E mutation had only a weak influence. In agreement with these studies, our results likewise showed that mutation of the amino acids Arg and Gly within the RGD sequence results in a moderate and strong decrease in affinity for ferri-Tf, respectively, whereas the influence of the Asp residue was low but significant. The dissociation constant of 3160 nM for the mutant G647A is in correlation with other investigations (K D ϭ 2345 nM (5)). In contrast to previous studies by other authors, we determined dissociation constants for full-length TfR that was expressed in mammalian cells. Furthermore, we determined for the first time concrete dissociation constants for Arg and Asp mutations within the RGD sequence of TfR.
By use of transfected cells expressing these different TfR mutants, we confirmed the data obtained by varying the ferri-Tf concentration. The shedding activity of these cells showed a strong inverse correlation with the binding affinity of the TfR mutants for ferri-Tf. Thus, inhibition of TfR shedding by ferri-Tf depended on direct binding of ferri-Tf to its receptor. Since our experiments on purified microsomal membranes reveal that the influence of ferri-Tf on TfR shedding is the same as in living cells, the inhibitory effect of ferri-Tf is not a result of either signal transduction events or enhanced endocytosis of TfR, as supposed by Baynes (37). Moreover, other causes, such as changes in the cellular distribution of the receptor in relation to the shedding protease, are also excluded, indicating instead a direct effect on proteolytic cleavage, most probably by blocking the cleavage site of the TfR. This is corroborated by new structural analyses of the Tf-TfR-complex, showing that the N-lobe of ferri-Tf is lodged between the ectodomain of the TfR and the membrane (7), representing the stalk region. This allows ferri-Tf to act as a competitive inhibitor for the binding of the protease to the TfR, although it cannot be finally excluded that the binding of ferri-Tf only changes the structure of the receptor, which in turn prevents binding of the protease or, if not, at least cleavage. To clarify this, purification and final identification of the protease would be helpful; however, it has already been shown that at least two proteases are involved in TfR shedding and that the proteases lose their activity when removed from the membrane (11,13). Moreover, the membrane-mediated lateral positioning of protease and substrate is important for recognizing the correct cleavage site (41).
In the last decade, it has become obvious that shedding of membrane proteins is an important process including such different proteins as membrane-anchored growth factors and precursors of cytokines, receptors, ectoenzymes, cell adhesion molecules, and the Alzheimer precursor protein. Most of the proteases identified to be involved in shedding processes are metalloproteases, but proteases of other classes are also described (reviewed in Refs. [42][43][44][45]. Ectodomain shedding is a widely branched network of competitive processes, i.e. one type of protein can be cleaved by different proteases, or vice versa, one protease can cleave several different proteins. To date, it is unclear how shedding is regulated. The size and composition of the stalk as well as the activation High iron demand leads to up-regulation of TfR expression and thus to up-regulation of sTfR release due to constitutive shedding. This effect is antagonized by ferri-Tf-mediated inhibition of the shedding process, leading to an sTfR concentration that reflects the balance between iron demand and availability. The middle panel shows that the effect of sTfR may be buffered in the bloodstream by ferri-Tf to generate a threshold effect. In the lower panel, the concentration of free sTfR is calculated dependent on transferrin saturation for two different sTfR concentrations (5 mg/liter and 25 mg/liter) and assuming a K D of 7 nM. The small case letters each indicate the same iron status in all panels.
of the protease and its localization with respect to the substrate are discussed as static and dynamic conditions (13, 46 -51). Our finding that ferri-Tf inhibits shedding of the TfR demonstrated that the binding of an extracellular molecule may be a general mechanism to regulate ectodomain shedding by environmental parameters.
In addition to general mechanisms, the elucidation of ferri-Tf-dependent regulation of TfR shedding sheds new light on iron homeostasis. To date, it is unclear how iron homeostasis is ensured. The sTfR level is closely related to erythroid TfR turnover, and the prime determinants of the sTfR concentration are cellular iron demand and erythroid proliferation rate (18,20). Furthermore, our studies demonstrate that sTfR does not only reflect the sum of cellular iron demand but simultaneously reflects the instantaneous iron availability of the whole body mirrored by the iron saturation of transferrin in the bloodstream, as depicted in the model in Fig. 5 (upper panel). Thus, the organism has the ability to act at an early stage to an imbalance between iron demand and iron availability. Decreasing iron saturation of transferrin (like in iron deficiency) and/or increasing cellular iron demand cause enhanced TfR shedding. In the bloodstream, released sTfR can bind to circulating ferri-Tf (Fig. 5, middle panel). Indeed, in the case of normal transferrin saturation, 95 Ϯ 3% of sTfR can be coprecipitated with antibodies against transferrin (18). An analysis of the equilibrium between free sTfR and sTfR complexed with ferri-Tf demonstrates, however, that the concentration of free sTfR increases drastically when transferrin saturation drops below 10% (Fig. 5, lower panel). Assuming a K D of 7 nM for the dissociation of sTfR from ferri-Tf, a 5-fold elevated sTfR concentration of 25 mg/liter due to iron deficit and low transferrin saturation (10%) results in a concentration of 1000 pM free sTfR in the bloodstream, a value 20-fold greater than under normal conditions (50 pM free sTfR in the case of 35% transferrin saturation and 5 mg/liter sTfR). The absolute free sTfR concentration and the dynamics of its variation are in the range for typical hormones like insulin (36 -600 pM) and aldosterone (180 -790 pM) (52).
Our data are consistent with the hypotheses formulated by Townsend and Drakessmith (28) and Cazzola et al. (53) and that free sTfR levels in the blood may activate iron export from reticuloendothelial cells or duodenal enterocytes, probably by an interaction with HFE in these cells. Our experiments give a hint about how sTfR can represent both the cellular iron demand and systemic iron availability. Indeed, Cook et al. (54) found a slight correlation between sTfR concentration and iron absorption in human. The correlation was abrogated if persons with iron deficiency (ferritin Ͻ15 g/ml) were excluded; however, this in turn means that in particular under iron deficiency conditions, sTfR may play an important role supporting our observations. With the described properties derived from our data, the sTfR completely fulfills the requirements for an erythroid regulator as formulated by Finch (55). It is effective when the total cellular iron demand of the body is greater than the compensative capability of the storage regulator to release iron from stores, in the case of the study of Cook et al. (54) under iron deficiency conditions. The model of sTfR as erythroid regulator is in line with increased iron resorption observed in ␤-thalassemia. In this disease, the number of reticulocytes is elevated (56), and thus, sTfR release is increased. This model is also supported by the observation that a lack of transferrin expression in humans and mice causes uncontrolled iron absorption from the gut, whereas repeated injections of transferrin in hypotransferrinemic animals results in reduced intestinal iron resorption (57). This can be explained by a drastic increase in free sTfR and by complexing free sTfR, respectively. Interestingly, plasma transfused to healthy mice does not affect iron absorption (58), indicating that the erythroid regulator must possess a short half-life or be neutralized as proposed by Hentze et al. (59). Our model suggests that after plasma transfusion to normal mice, ferri-Tf complexes the transfused sTfR and thus reduces the effective free form of sTfR, which is an effect analogous to the transferrin injections in hypotransferrinemic animals.
Since the binding of sTfR to HFE has been already shown, the hypothesis proposed by Townsend and Drakessmith (28) and Cazzola et al. (53) may be probable, but other roles should also be taken into consideration, e.g. sTfR may influence the expression of the iron storage regulator hepcidin in the liver. It is supposed that the expression of this small peptide is regulated by HFE and a second transferrin receptor (TfR2) (24). In this case, the cross-talk between the cellular iron demand and the iron storage situation may be mediated by sTfR. The demonstration that sTfR also functions as a regulator of iron export from reticuloendothelial cells and duodenal enterocytes is currently under investigation.