Receptor-mediated endocytosis of transferrin in K562 cells.

Human diferric transferrin binds to the surface of K562 cells, a human leukemic cell line. There are about 1.6 X 10(5) binding sites per cell surface, exhibiting a KD of about 10(-9) M. Upon warming cells to 37 degrees C there is a rapid increase in uptake to a steady state level of twice that obtained at 0 degree C. This is accounted for by internalization of the ligand as shown by the development of resistance to either acid wash or protease treatment of the ligand-cell association. After a minimum residency time of 4-5 min, undegraded transferrin is released from the cell. Internalization is rapid but is dependent upon cell surface occupancy; at occupancies of 20% or greater the rate coefficient is maximal at about 0.1-0.2 min-1. In the absence of externally added ligand only 50% of the internalized transferrin completes the cycle and is released to the medium with a rate coefficient of 0.05 min-1. The remaining transferrin can be released from the cell only by the addition of ligand, suggesting a tight coupling between cell surface binding, internalization, and release of internalized ligand. There is a loss of cell surface-binding capacity that accompanies transferrin internalization. At low (less than 50%) occupancy this loss is monotonic with the extent of internalization. Even at saturating levels of transferrin, the loss of surface receptors upon internalization never exceeds 60-70% of the initial binding capacity. This suggests that receptors enter the cell with ligand but are replaced so as to maintain a constant, albeit reduced, receptor number on the cell surface. In the absence of ligand, the cell surface receptor number returns at 37 degrees C. Neither sodium azide nor NH4Cl blocks internalization of ligand. However, they both prevent the release of transferrin from the cell thus halting the transferrin cycle. Excess ligand can overcome the block due to NH4Cl but not azide although the cycle is markedly slower. Iron is delivered to these cells by transferrin at 37 degrees C with a rate coefficient of 0.15 to 0.2 min-1. The iron is released from the transferrin and the majority is found in intracellular ferritin. There is a large internal receptor pool comprising 70 to 80% of the total cell receptors and this may be involved in maintaining the steady state iron uptake.

Human diferric transferrin binds to the surface of K562 cells, a human leukemic cell h e . There are about 1.6 x lo6 binding sites per cell surface, exhibiting a KO of about lo-' M. Upon warming cells to 37 O C there is a rapid increase in uptake to a steady state level of twice that obtained at 0 "C. This is accounted for by internalization of the ligand as shown by the development of resistance to either acid wash or protease treatment of the ligand-cell association. After a minimum residency time of 4-5 min, undegraded transferrin is released from the cell. Internalization is rapid but is dependent upon cell surface occupancy; at occupancies of 20% or greater the rate coefficient is maximal at about 0.1-0.2 min-l. In the absence of externally added ligand only 60% of the internalized transferrin completes the cycle and is released to the medium with a rate coefficient of 0.05 min-'. The remaining transferrin can be released from the cell only by the addition of ligand, suggesting a tight coupling between cell surface binding, internalization, and release of internalized ligand. There is a loss of cell surface-binding capacity that accompanies transferrin internalization. At low (~5 0 % ) occupancy this loss is monotonic with the extent of internalization.
Even at saturating levels of transferrin, the loss of surface receptors upon internalization never exceeds 60-70% of the initial binding capacity. This suggests that receptors enter the cell with ligand but are replaced so as to maintain a constant, albeit reduced, receptor number on the cell surface. In the absence of ligand, the cell surface receptor number returns at 37 "C. Neither sodium azide nor m C 1 blocks internalization of ligand. However, they both prevent the release of transferrin from the cell thus halting the transferrin cycle. Excess ligand can overcome the block due to m C 1 but not azide although the cycle is markedly slower. Iron is delivered to these cells by transferrin at 37 "C with a rate coefficient of 0.15 to 0.2 min-'.
The iron is released from the transferrin and the majority is found in intracellular ferritin. There is a large internal receptor pool comprising 70 to 80% of the total cell receptors and this may be involved in maintaining the steady state iron uptake.
The mechanism by which iron is taken up by cells is still unclear. The delivery of iron to cells is mediated primarily, if not uniquely, by transferrin, a serum glycoprotein capable of * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
f Supported by stipends from the Niels-Stensen Stichting, The Netherlands, and from the Netherlands Organizations for the Advancement of Pure Scientific Research (Z.W.O.), The Netherlands. binding two ferric ions (reviewed in Ref. 1). A specific membrane receptor for transferrin appears to be the first step in the iron uptake process (2). This receptor is found on many cells and has recently been purified and shown to be a glycoprotein (3, 4). AU cells require iron as a constituent of respiratory enzymes and other heme-containing proteins. This probably underlies the requirement for transferrin cell culture systems (5). Of particular interest has been the iron uptake of the erythroid cell lines because of the requirement for relatively large amounts of iron for hemoglobin production. Most of these studies have focused on short term cultures of reticulocytes. Although these cells are active in iron uptake, they are constantly changing as they differentiate terminally to become erythrocytes.
All of the studies on transferrin and iron uptake by cells have led to some disagreement about the events that occur after binding of transferrin to the surface receptor. Two major models have been presented. In one the iron enters the cell together with the transferrin in a process of receptor-mediated endocytosis (6). The study of Karin and Mintz using a mouse teratocarcinoma cell line strongly suggests the existence of such a process (7). An alternative model envisages the iron to be removed from the transferrin at the membrane and internalized by, as yet, undefined cellular processes (8). It is possible that different cells employ different mechanisms of iron uptake.
We have chosen to study the uptake of iron from transferrin using the human erythroleukemia cell line K562 (9). This cell line has a number of features that make it an excellent system for such studies. It is a stable cell line that appears morphologically similar to a blast cell. These cells produce a small amount of hemoglobin and this can be greatly enhanced by the presence of hemin in the growth medium without the occurrence of terminal differentiation (10).
The studies reported here strongly support the initial model of Karin and Mintz. We find that there are a large number of high affinity transferrin receptors on these cells which mediate the internalization of diferric transferrin. After internalization, both iron molecules are removed from the protein and apotransferrin is released from the cell in an energy-dependent process.

MATERIALS AND METHODS
Transferrin-Human transferrin (Calbiochem) was passed over a s-200 (Pharmacia) column to remove any aggregated protein. This material showed a single peak on high pressure liquid chromatography. In all experiments diferric transferrin was used. Saturation of transferrin with iron was performed as follows. Six mg of Tf' was dissolved in 1 ml of 0.25 M Tris-C1, pH 8.0, 10 p NaHC03. To this solution 20 pl of 100 n m disodium nitrilotriacetate/l2.5 mM FeCb were added. The sample was incubated at 37 "C for 30 min and then passed over a PD-10 column (Sephadex G-25, Pharmacia) previously equilibrated with 0.15 M NaC1/0.02 M Tris-C1, pH 7.4. The amount of iron bound by the transferrin was estimated from the A465 ,,,,,/Azm ,,,,, ratio, which was routinely found to be 0.046, consistent with full saturation.
59FeC13 (1 mCi/mk 79 pg of Fe/ml) was obtained from Amersham Corp. Human transferrin was obtained from Pentex. Disodium ascorbate and disodium nitrilotriacetate were purchased from Sigma. Rabbit anti-human transferrin antibody and rabbit anti-human ferritin antibody were obtained from Boehringer Mannheim. Desferrioxamine was a generous gift from Dr. Arthur Nienhuis, National Institutes of Health. Pansorbin was obtained from Calbiochem-Behring. PD-10 columns (Sephadex G-25) were purchased from Pharmacia. Beckman Company was the source of the tabletop microfuge and the series 5000 y counter. ACA-34 (Ultrogel) chromatography gel was obtained from LKB Instruments, Inc.
Iodination of Tf-Diferric Tf in 0.15 M NaC1/0.02 M Tris-C1, pH 7.4, was combined with 50 pl of a suspension of lactoperoxidase/ glucose oxidase immobilized on agarose beads (Enzymobeads; Bio-Rad). Ten ,d of Na? (100 mCi/ml, Amersham Corp.) or 40 pl of NaI3lI (400 mCi/ml, Amersham Corp.) followed by 50 pl of 1% p-Dglucose were added to this mixture. The sample was incubated at room temperature for 30 min and then passed over a PD-10 column equilibrated with 0.15 M NaC1/0.02 M Tris-C1, pH 7.4. The specific activities of lZ5Iand 13LI-labeled Tf were similar and both varied between preparations from 600-1200 cpm/ng.
Preparation of 59Fe-Tf-Apotransferrin was prepared as previously described (11) except that the PD-10 column was equilibrated with 10 pM NaHCOd0.25 M Tris-C1, pH 8.0. 100 pl of 59FeC13 (0.1 mCi) were added to 500 pl of a 100 mM disodium nitrilotriacetate solution. This solution was combined with the apotransfemin and incubated at room temperature for 60 min. The 59Fe-Tfsolution was then passed over a PD-10 column and the peak tubes were pooled.
Based upon the specific activity of the 59Fe solution, the molar ratio of Fe/Tf for the 59Fe-Tf was calculated to be 1.6.
Cells and Binding Assays"K562 cells were grown in suspension in RPMI with 10 mM 4-(2-hydroxyethy1)-I-piperazineethanesulfonic acid, 100 I " glutamine, and 10% fetal bovine serum (heat inactivated) (Gibco Laboratories, Grand Island, NY). For these studies, cells were grown to a density of 5 X IO6 cells/ml and then divided one to two. The following day they were harvested at 4-6 X 105/ml. Cell viability, as judged by the exclusion of 0.5% trypan blue, was always >95%. Cells to be studied were washed twice with either RPMI alone or with RPMI containing 1% serum by centrifugation at 1000 rpm in a JEC centrifuge at 4 "C and subsequently suspended to a final concentration of 10 X IO6 cells/ml. with radiolabeled Tf and 0.5-ml aliquots were added to 1 ml of RPMI Two types of binding assays were used. In one, cells were incubated with 1% BSA (w/v) in a 1.5-ml Eppendorf centrifuge tube. This was centrifuged for 0.5 min at 8000 rpm, and the pellet was resuspended and washed twice in RPMI. The second technique which was adopted for these studies was to layer 200 pl of the cell sample over 150 p1 of dibutylphthalate oil in a 400-pl Eppendorf centrifuge tube. After 10-20 s at 8000 rpm, the bottom of the tube containing the cell pellet was cut off and counted. The fist technique allowed us to quantitate the cell recovery; the second technique, which was much faster, gave identical results and lower background levels of nonspecific binding as measured in the presence of excess cold Tf.
During these studies we observed a gradual loss of cell viability in cells kept at 37 "C for more than 45 min. The loss of cell viability under those conditions could be completely prevented by carrying out the incubations in the presence of 1% (w/v) fetal calf serum. At this concentration the serum inhibits the binding of radiolabeled Tf to the cells by less than 5% for the concentrations of Tf used. Fetal bovine Tf, the presumed responsible inhibitor, has a much lower affinity than human Tf for the Tf receptor on the human K562 cells. Binding isotherms at 0 "C were carried out in the absence of fetal calf serum.
To distinguish cell surface-bound from internalized ligand, we examined the release of 1z51-Tf from cells by acid (12). Two hundred pl of labeled cells were added to 200 pl of 0.25 M acetic acid/0.5 M NaCl (pH 2.3). After 5 s 100 pl of 1 M sodium acetate were added, which returned the pH to 6.0, and this was immediately spun through dibutylphthalate. Cell recovery was unaffected by the exposure to acid. Incubation with the acid for longer times did not alter the amount of ligand released. To compare acid sensitivity to protease sensitivity, cells were incubated with lz5I-Tf at 0 "C or 37 "C for varying periods of time and cooled to 0 "C. Pronase B (Calbiochem) was added to give a final concentration of 100 pg/ml. Aliquots of the cells were removed at different times and cell-associated counts were determined by centrifugation through oil. Cell-associated counts decreased for 15 to 20 min and then remained constant. Pronaseresistant counts were, therefore, determined after 20 min.
Uptake of 59Fe by K562 Cells-Cells were centrifuged at 300 x g for 5 min and resuspended in RPMI 1640 medium containing 1% fetal bovine serum at a cell concentration of 1 X 107/ml. 59Fe-Tf was added to the cell suspension to a final concentration of 100 pg/ml. Cellassociated counts were determined by spinning 200 pl of a cell suspension through an oil layer in a microfuge and counting the pellet. Acid lability of cell-associated 59Fe was determined as described above.
I n Vitro Assay for Soluble Transferrin Receptors-In general, freshly harvested K562 cells were solubilized by suspending 0.5-1.0 X lo7 cells per ml of Tris-buffered saline, pH 7.2, containing 0.1% Triton X-100. To 0.05-0.20 ml of the solubilized cells, 0.02 ml of Iz5Itransferrin (1 pg, 5-10 X IO5 cpm) was added and the volume brought to 0.50 ml with the above buffer-detergent solution (Buffer A). In all cases, nonspecific binding was corrected by subtracting the values obtained with incubations wherein an excess (50 to 100-fold) of nonradioactive transferrin had been added. The incubation was carried out for 10 min at room temperature and terminated by the addition of an equal volume (0.5 m l ) of 60% saturated ammonium sulfate (room temperature) which had been adjusted to pH 7.4. The pH of the latter solution should be checked shortly prior to each assay on a daily basis. After standing at 0 "C for 5 min, the precipitate was removed by filtration on GF/C Whatman filters which had been previously soaked in 30% saturated ammonium sulfate containing 0.8% BSA at pH 7.4 (Buffer B). The residue on the filters was washed 3 times with 1 ml each of Buffer B and the filters were counted in a Beckman autogamma spectrometer. In the range indicated, the assay was proportional to the number of cells added.

Binding of 1251-Tf to K562
Cells-Binding of Iz51-Tf to K562 cells at 0 "C was extremely rapid but, as expected, the time required to reach maximum binding increased with decreasing ligand concentration. Preincubation of the cells with excess (100-fold) cold Tf for 5 min resulted in a dramatic inhibition of subsequent '251-Tf binding. Mixing labeled and unlabeled Tf prior to binding showed the two preparations to bind identically. A 1000-fold excess of bovine serum albumin or asialo-orosomucoid (a glycoprotein) had no effect on Tf binding. The binding isotherm at 0 "C, plotted as a Scatchard analysis, yields an apparent Kd of about 0.9 X lo-' M and a receptor number of about 1.6 X 105/cell. The dissociation of Tf at 0 "C was obtained by washing cells that had been loaded with Tf at 0 "C and measuring the loss of cell-associated counts as a function of time. The off rate at 0 "C is slow with a half-time of approximately 70 min; it is not affected by extraneous proteins such as BSA but can be specifically increased to approximately 48 min by the addition of excess cold Tf.

The amount of Tf taken up by these cells increases signif-
icantly when the incubation is performed at 37 "C. This rise occurs rapidly and plateaus by 10 to 15 min at a level two times that attained at 0 "C. At 25 "C, the same plateau is reached much more slowly, requiring 30 to 35 min. The binding at 37 "C is specifically inhibited in the presence of excess cold Tf. Plasma membranes isolated from K562 cells specifically bind Tf. In contrast to whole cells, however, the membranes bind Tf equally well at 0 and 37 "C, thereby suggesting that the increased uptake of Tf by whole cells at 37 "C might reflect an active metabolic process. We examined the release of cell-bound Tf at 37 "C ( Fig.  L4). Cells were loaded with Tf for 30 min at 37 "C, chilled to 0 "C, and then washed free of unbound ligand. These cells were rapidly warmed to 37 "C, and the loss of cell-associated counts was monitored. In contrast to the results at 0 "C there was a rapid loss of cell-associated Tf at 37 "C. When, at any point during this release, the cells were chilled to 0 "C, the loss of cellular ' "I stopped. The possibility that degradation at 37 "C was responsible for this release could be eliminated from consideration since the released counts were completely precipitated by trichloroacetic acid. Furthermore, when the released material was run on an sodium dodecyl sulfate-polyacrylamide gel all of the counts comigrated with the initial ligand and no breakdown products were observed. An initially puzzling aspect of the release of Tf at 37 "C was that it reached a maximum between 40-60% and then remained constant (Fig. 1A). Subsequently, we found that the addition of increasing amounts of cold Tfled to a progressively greater amount of labeled Tf recoverable in the medium (Fig.  1B). The addition of BSA or asialo-orosomucoid had no effect on this release. This observation raised the question as to wl.ether the released transferrin was able to rebind to the cells and whether the effect of excess cold Tf was simply to block rebinding. This seemed unlikely since the amount of protein released represented a 1000-fold dilution compared to the original incubation. Furthermore, altering the incubatipn volume during release had little effect on the extent of Tf release. To examine the rate and extent of release of Iz5I-Tf at 37 "C as a function of the occupancy of the surface receptors, the cells were incubated with varying amounts of ligand at 0 "C, washed thoroughly to remove unbound ligand, and rapidly warmed to 37 "C. Although the surface occupancy varied from less than 10% to greater than 95%, the pattern of release was similar, plateauing at 58% of the initial value (Fig.   M). Curve fitting revealed the rate constant for release to be about 0.05 min" and the release to be overwhelmingly due to a single exponential process. As further evidence against a rebinding phenomenon, we made use of the observation that the addition of anti-human Tf antibodies completely inhibited the binding of labeled Tf. Irrelevant antibody was ineffective. Establishing this, we asked whether a large excess of antibody would alter the extent or rate of Tf release from the cells. As shown in Fig. 2B, the antibody had no effect on the extent of Tf released from the cells. However, the initial rate of release was increased. Furthermore, the ability of excess cold l'f to enhance the release could be inhibited by excess anti-Tf antibody (data not shown).
Internalization of Transferrin by K562 Cells-To test for the internalization of Tf, we used the approaches employed by Karin and Mintz (7): 1 ) the ability to release cell surface bound ligand by a rapid acid wash; 2) the accessibility of such ligands to added proteases. The latter assay, which was slower and more cumbersome, was used only to confirm the data obtained by the acid wash method. Upon loading cells at 0 "C, 90 to 96% of the cell-associated Tf was released by the acid treatment. In contrast, only 15% of the cell-associated counts could be stripped from cells loaded at 37 "C. Comparable results were obtained with protease digestion. When Tf was bound to isolated plasma membranes from K562 cells at 0 or 37 "C, all of the specifically bound Tf was released with the acid wash. During the course of 60 min at 0 "C about 15% of the cell-associated counts became acid resistant. Entry into the acid-resistant compartment at 37 "C was quite rapid. We found, however, that this was dependent upon receptor occupancy (Fig. 3A). Cells, loaded at 0 "C for 5 min with varying concentrations of Iz5I-Tf, were washed and rapidly warmed to 37 "C in the absence of ligand in the medium. At 95% occupancy, the ligand had maximally entered the acid-resistant compartment by 4-6 min. At 5% occupancy, it took about 20 min before the ligand became maximally acid resistant. This change was not linear with occupancy; the maximum rate of internalization was seen at about 20% occupancy.
We compared the rate of internalization, ie. development of acid resistance to the rate of exit from the cell. Cells were loaded with Tf at 0 "C for 10 min (Fig. 3B), washed thoroughly, and warmed to 37 "C. Internalization proceeded rapidly with only a minimal lag. In contrast there was a 4-min delay before the Tf left the cell. It is also clear that most, if not all, ligand that leaves the cell does so via the acid-resistant surprising (Fig. 4). The amount on the cell surface peaked compartment.
between 2-5 min at a level close to that seen upon binding at The amount of total cell-associated Tf at 37 "C plateaus at 0 "C and then rapidly fell to a small fraction of the initial about twice the amount found at 0 "C for any given amount binding. These results strongly suggest that the Tf-binding of ligand. We determined how much of the cell-associated capacity of the cell surface is decreased upon uptake of ligand ligand was on the surface of the cell by acid resistance at each at 37 "C. time point during uptake at 37 "C. The results were quite To examine this apparent loss of cell surface-binding activ- and 5% (0) occupancy. Both samples reached about 85% acid resistance. This was taken to be a fractional internalization of 1.0, and the data shown was normalized with respect to this. B, internalization uersus release of Tf at 37 "C. Cells (1 X 107/ml) were incubated with 10 pg of "'I-Tf for 10 min at 0 "C, washed free of unbound ligand, and resuspended in media and placed in a 37 "C water bath. Aliquots were removed at timed intervals and assayed for total cellassociated counts (+) and acid-sensitive counts (A). ity at 37 "C, cells at 107/ml were incubated with various concentrations of lZ5I-Tf for 30 min at either 0 or 37 "C. The initial cell surface binding was determined from the 0 "C cells (Fig. 5). The amount of cell-associated Tf at 37 "C at each concentration was approximately twice that seen for each corresponding 0 "C incubation (data not shown). At the end of the 30-min incubation, the 37 "C cells were chilled to 0 "C, washed free of surface and unbound ligand, and resuspended to the initial cell density. To each sample 100 p g / d of 13'I-Tf was added and allowed to bind for 20 min at 0 "C. The surface binding of the I3lI-Tf was determined. The depletion of cell surface binding capacity is clearly illustrated by the binding of 1311-Tf at 0 "C. When cells were preincubated at 37 "C in the absence of lZ5I-Tf, about 50 ng of 13'I-Tf were bound to the surface per 2 X lo6 cells at 0 "C. This number decreased progressively as the amount of lZ5I-Tf in the preincubation was increased. Furthermore, it can be concluded that there is a close to one to one correlation between the number of surface-bound molecules of lz5I-Tf internalized and the decrease in subsequent surface-binding activity. This monotonic down regulation exists for receptor occupancy levels from zero to 60-708. At higher levels, the subsequent surface binding remained constant with the maximum down regulation attainable at about 60-70%. No amount of ligand present in the 37 "C preincubation is capable of inhibiting the subsequent binding of I3lI-Tf to below about 15 ng per 2 X lo6 cells.
After the ligand-induced loss of surface receptors at 37 "C, there was no recovery of binding if the cells were kept at 0 "C. On the other hand, surface-binding activity recovered during a 30-min incubation at 37 "C in the absence of external ligand as is shown in Fig. 6. Here, cells were loaded at 37 "C with '"I-Tf. Acid-sensitive cell-associated ligand was measured.
After 45 mi n, the cells were cooled and washed free of ligand. Surface binding (at 0 "C) was then determined after different times at 37 "C by the addition of a saturating amount of lZ5I-Tf. The initial 37 "C incubation exhibited a maximum surface binding of 14,000 cpm/l X lo6 cells which corresponded to about 25 ng of Tf/106 cells. This number was in good agreement with a parallel experiment in which these cells were incubated with the same amount of ligand at 0 "C. The acidsensitive ligand then decreased rapidly and stabilized at about 3000 cpm/106 cells. After 30 m i n , the cell surface binding rose to about 70% of the initial value and further incubations resulted in minimal additional recovery. In other experiments, the addition of cycloheximide (50 p i , which inhibited I4Cleucine incorporation into proteins by greater than 90%) to the cells had no influence on the recovery of binding capacity or on any aspects of the data reported in this paper.  Effect of drugs on release of 1251-Tf from cells. A, cells were allowed to take up Iz5I-Tf at 37 "C for 30 min. At this time the cells were divided into two aliquots one of which received azide (final concentration 1%) and 2-deoxyglucose (50 mM) and the incubation continued for 10 min. Cells were then chilled, washed thoroughly free of unbound ligand, and resuspended in media at 37 "C containing no additions (A), 100 pg/ml of cold Tf (+), azide plus deoxyglucose (0) or azide, deoxyglucose plus 100 p g / d of cold Tf (A). The release of '"I-Tf at 37 "C was followed for each sample. B, cells were incubated with 20 mM ammonium chloride at 37 "C for 15 min. Cells were chilled to 0 "C and lZ5I-Tf was added for 20 min in the presence of ammonium chloride and washed to remove unbound '"1-Tf. Cells were rewarmed to 37 "C either in the absence (0) or presence (m) of 100 pg/ml of unlabeled Tf. Control cells (no NHXI) were likewise rewarmed in the presence (A) or absence ( X ) of excess cold Tf.
Effect of Selectiue Inhibitors upon the Uptake of Tf by Cells-Two types of inhibitors were tested for their effects upon the processes described above. These included (a) metabolic inhibitors such as sodium azide and 2-deoxyglucose that act by depletion of cellular ATP and (b) ammonium chloride which presumably acts as a lipid-soluble weak base to neutralize acidic cellular compartments. Neither of these regimes affected the binding of Tf to cells at 0 "C. This was true even when the cells were treated with the drugs at 37 "C for as long as 30 min before cooling. At 37 "C, the uptake of Tf by cells treated with azide and deoxyglucose was much lower than either control cells or cells treated with NH&1 and was comparable to the binding at 0 "C. Binding to isolated plasma membranes was not affected by treatment with either of these drugs at any temperature.
Significantly, both the rate and extent of internalization of surface-bound Tf was unaffected by azide and deoxyglucose.
In contrast the release of Tf from the cells was profoundly affected (Fig. 7 A ) . After the cells had taken up Tf at 37 "C, exposure to azide and 2-deoxyglucose led to nearly complete inhibition of release. More strikingly, the addition of a large excess of cold Tf failed to enhance the release of the Tf (Fig.  7 A ) . N&Cl yielded similar results when the spontaneous release of ligand was monitored. However, in contrast to azide, the addition of excess cold Tf overcame the NH&l block with resultant release of the cell-associated ligand although the rate of release was slowed significantly (Fig. 7 B ) . Uptake and Fate of Iron-At 4 "C there is rapid binding of 59Fe-Tf to K562 cells with the steady state being reached in less than 5 min. In contrast, at 37 "C cell-associated 59Fe continued to rise over a period of 60 min, and in the presence of sufficient ligand, uptake was nearly linear. At both temperatures, 59Fe uptake was inhibited by excess unlabeled transferrin but not by comparable amounts of proteins such as bovine serum albumin or orosomucoid. 59Fe taken up at 37 "C, (e) (100 p g / d ) or lz5I-Tf (0) (100 pg/ml) was added to cells at 37 "C and aliquots were removed at the indicated times and spun through oil to determine the total cell-associated counts. After 45 min, the cells were chilled to 4 "C and washed twice with cold incubation medium. They were returned to 37 "C, and cell-associated radioactivity was determined at the indicated times after rewarming.
but not at 4 "C, was resistant to displacement by 0.25 M NaCI/ 0.25 M acetic acid, suggesting that it was internalized. Fig. 8 demonstrates the marked difference in behavior of the alternatively labeled ligand in cells incubated with 59Fe-Tf and lz5I-Tf at 37 "C. In the latter case, lz5I-Tf rapidly reached a steady state level approximately twice that seen at 4 "C. In contrast, cell-associated 59Fe continued to rise, implying that 59Fe had been separated from the transferrin. By 45 min, the amount of iron delivered to the cells represented about 8 to 10 times the amount expected from the steady state level of transferrin (assuming 2 atoms of Fe per mol of transferrin). At this time, cells washed free from ligand and resuspended at 37 "C revealed a progressive 50% loss of lZ5I-associated counts over a period of 60 min with no diminution in the radioactive iron content, thereby providing additional evidence for dissociation of iron from the transferrin molecule. Based on the specific activity of the 59Fe-Tf used in these studies, a rate coefficient for the uptake of transferrin was calculated to be 0.13 min". This value correlates well with the corresponding figure of 0.15-0.2 min" for lZ5I-Tf and indicates that the uptake of iron and transferrin are coupled. Desferrioxamine, a powerful iron chelator which binds free Fe3+ in the medium but which does not remove iron from transferrin over short periods of time, had no effect on the uptake of 59Fe from transferrin (data not shown). This is a further indication that 59Fe is not released from transferrin into the medium prior to its uptake by the cells.
The prolonged linear uptake of 59Fe to levels stoichiometrically in excess of those achieved by lz5I-Tf at steady state indicates that the cycling of transferrin described above provides for a continuous delivery of iron to the cells. Having established that the iron is released from the transferrin, we sought to determine the locus of iron binding within the cell. To approach this problem, cells were incubated with 59Fe-Tf or lZ5I-Tf at both 0 and 37 "C. At the end of each incubation, the cells were washed to remove unbound ligand solubilized in a Triton X-100-containing buffer, and the resulting extract was applied to an ACA-34 sizing column. As shown in Fig. 9A, authentic lZ5I-Tf emerged as a single peak within the included volume of the gel and was clearly separable from the peak observed after lZ5I-Tf had been incubated with cells at 4 "C. The latter peak which was similarly recovered following incubation of 59Fe-Tf at 4 "C is presumed to represent the initial transferrin-receptor complex. In contrast, all of the radioactive iron taken up during 1 h at 37 "C emerged as a higher molecular weight component which co-eluted with radiolabeled horse spleen ferritin.
To establish ferritin as the site to which 59Fe was transferred, immunoprecipitation using anti-human ferritin and anti-human transferrin antibody was performed. The majority of the 59Fe in the cells was immunoprecipitated by anti-human ferritin antibody. In contrast, the amount of 59Fe immunoprecipicated by the anti-human transferrin antibody was only slightly greater than the control (Fig. 9B).
Internal Receptor Pool-The linear rate of iron uptake occurs with a rate coefficient of 0.15-0.2 min". This is quite similar to that found for the rate of Tf uptake by the cells. This presents a problem in that the release of Tf from the cell occurs with a rate coefficient of about 0.05 min". Thus, if this represented the rate of return of unoccupied receptors to the surface, this would be rate limiting for iron uptake in the steady state. A simple resolution of this problem would be if there existed an "internal" receptor pool from which unoccupied receptors could be recruited, replacing the rapidly internalized surface receptors. Internal receptor pools are seen in '"I-transferrin; (X) 1251-transferrin standard which had not been exposed to cells. All curves were normalized by assigning a value of 1 to the peak fraction. B, Pansorbin immunoprecipitation of 59Fe from solubilized cells. Cells were incubated with 59Fe-Tf at 37 "C for 1 h and then chilled to 4 "C and washed twice with ice-cold phosphate-buffered saline. After the second wash, the cell pellet was solubilized with 0.1% Triton X-100, 0.15 M NaCl, 0.02 M Tris-C1, pH 7.4. Centrifugation removed insoluble material (10% of the radioactivity) and varying amounts of anti-human ferritin antibody, anti-human transferrin antibody, or buffer were added to aliquots of the solubilized material. After a 1-h incubation at 4 "C 100 pl of Pansorbin cells previously washed in the above buffer were added. The mixture was incubated at 4 "C for 1 additional hour. Then the Pansorbin cells were washed twice with cold buffer and the pellet counted. 1,2, 3, 10  Initial experiments were conducted wherein K562 cells were incubated at 0 "C with an excess of nonradioactive transferrin, washed free from unbound material, and solubilized in Buffer A (see under "Materials and Methods"). Control cells were similarly incubated in the absence of added transferrin. After solubilization, both sets of cells were incubated with a saturating amount of '251-transferrin for 15 min at 0 "C prior to passage over a column of Sephacryl-200 (1 X 36 cm). The receptor-ligand complex appeared in the void volume in contrast to the unbound ligand which was recovered in the included volume. Comparison of the amount of radioactivity recovered in the void volume, from cells previously exposed to cold transferrin, to saturate external receptors revealed that only 20% of the total bound radioactivity could be accounted for as being bound to the plasma membranes of the cell.
In order to provide a more quantitative estimate of the ratio of external to internal receptors, 1 X lo7 K562 cells were  Table I, which reveals that the ratio of external to internal receptors is 1:3.5, a value in accord with the Sephacryl-200 column chromatographic estimate of 1:4 (80% internal). These data were further substantiated by a Scatchard analysis of the total binding capacity of solubilized K562 cells. A value of 500,000 receptor sites per cell was obtained with a dissociation constant of 1.8 X M. As indicated earlier, only 160,000 binding sites were detected on the cell surface of intact cells. Consequently, it could be estimated that the ratio of internal to external binding sites was 1 to 3.3. From all of the above, it seems safe to conclude that 3 to 4 binding sites are available within the cell for each site expressed on the cell surface.

DISCUSSION
The human erythroleukemia cell line K562 binds human transferrin. The binding is rapid and slowly reversible at 0 "C. At this temperature the half-time for the release of Tf from the surface of these cells is about 70 min. Binding can be blocked completely by unlabeled Tf but not by other serum proteins. Scatchard analysis of the 0 "C binding isotherm revealed a linear curve consistent with a single type of binding site with an apparent affinity of lo9 M". Such an analysis leads to an estimate of 1.6 X lo5 receptors per cell. This is in the range for the number of Tf receptors calculated on a variety of cells (14-16). Isolated plasma membrane fractions of K562 cell homogenates also displayed specific binding sites.
A striking alteration in the cellular uptake of transferrin was obtained when the temperature was raised from 0 to 37 "C. Under these conditions, the amount of cell-associated Tf increased about 2-fold. That no such increase in binding to isolated membranes was observed suggested a cellular process was involved in the increase. The acid wash experiments provided further support for this idea. Several workers have used this technique to distinguish surface receptor-bound ligand from internalized ligand (12). Karin and Mintz showed that with mouse teratocarcinoma cells, acid resistance correlated with resistance to extracellular proteases and concluded that this procedure could be used to determine the internalization of transferrin (7). Our findings support and extend theirs. Furthermore, the absence of acid resistance at any temperature when isolated membranes are used is consistent with this interpretation.
At 37 "C, there is significant release of cell-associated Tf which ceases when the cells are chilled to 0 "C. Evidence that the released material is intact transferrin was provided by the demonstration that it comigrates with transferrin on sodium dodecyl sulfate-polyacrylamide gels. This finding is in marked contrast to other endocytic systems in which the ligands are degraded after internalization (17, 18). Strikingly, only about 50% of the ligand taken up by the cells is released irrespective of the actual level of Tf internalized. We examined the question of whether the failure to release all of the ligand was due to rebinding of released Rf. However, all manipulations designed to alter rebinding were without effect upon the extent to release. A large excess of anti-Tf antibody which was capable of blocking the binding of Tf to cells did alter the rate but not the extent of release.
To elucidate further the release at 37 "C it is helpful to define which cellular pool of Tf is being lost from the cell. We have defined two pools of Tf in this study, acid sensitive and acid resistant. When cells are loaded with Tf at 37 "C about 85% of the Tf is in the acid-resistant pool after 30-45 min of incubation. Thus, it can be concluded that the vast majority of Tf released from the cell at 37 "C in the presence or absence of excess ligand emerges from the acid-resistant compartment. This is further supported by the results shown in Fig. 3B. When cells are loaded with Tf at 0 "C, all the protein is in the acid-sensitive pool. Upon warming the cells, the ligand rapidly enters the acid-resistant compartment. However, there is a lag time of about 4 min before any ligand is released from the cell. This suggests an obligate cellular pathway for Tf prior to release.
We believe that our results are best interpreted as follows.
Tf binds to surface receptors. It is subsequently internalized and, after completing an undefined cellular journey, it is released from the cell. The process of release is apparently coupled to internalkation. Hence, without continued occupancy of the surface receptors, the cycle stops and the release ceases. Only when Tf binds to unoccupied surface receptors is the cycle initiated, leading to the continued release from the "internal" pool. This predicts that during incubation at 37 "C, in the continuous presence of ligand, this cycle continues to turn. Why only 50% of the ligand that enters the internal compartment is able to leave the cell in the absence of added Tf is unclear. However, the lack of dependence of this percentage upon the total amount of cellular T f suggests a process linked to a possible bimolecular interaction of Tf, receptors, or ligand-receptor complexes. The 50% value suggests that this interaction involves either two Tf molecules or Tf-recep-tors. At this point in our studies the exact mechanism behind this phenomenon remains to be elucidated. That the receptor accompanies the ligand into the sequestered pool is suggested by the experiments shown in Figs.

4-6.
It is striking that the amount of surface or acid-sensitive ligand drops rapidly even in the presence of a large excess of available Tf. Experiments employing 1311 and '251-labeled Tf helped to clarify this. Binding capacity on the surface of the cell is lost upon internalization of ligand. In addition, the greater the amount of original surface receptors occupied the greater the loss of subsequent surface-binding capacity. The total number of surface receptors measured as the number originally present and occupied plus the amount of subsequent binding remains relatively constant at 040% occupancy. This suggests a close to one to one loss of surface receptor with internalized ligand over this range. However, at higher levels of initial occupancy, subsequent binding capacity is maintained, albeit at lower levels.
The function of transferrin is to deliver iron to cells. In this cellular system, as in others, the iron is released as the transferrin cycles through the cell. A linear rate of iron uptake is observed. In order for this to be maintained, the rate of internalization of transferrin receptor must be equal to the appearance of new surface receptors. However, our data show that the rate of internalization of transferrin is about four times greater than the rate of transferrin release from the cell. If the receptor canies the ligand through its entire cellular journey the entrance and exit of ligand w i l l reflect the rate of depletion and repletion of a single cycling receptor. This would result in a rate of iron uptake that would be limited by the release rate. One way around this would be if the internalized receptors are rapidly replaced by internal unoccupied receptors. Our soluble receptor assay supports the existence of a cryptic receptor pool. This pool size is three to four times the number of surface receptors. Interestingly, the ratio of the internalization to release rate coefficients is about four and thus this size internal pool would be sufficient to account for the maintenance of a rapid and linear uptake of iron.
Both azide and NH&l altered the transferrin cycle. Neither of these agents inhibited the internalization of surface-bound ligand. Both agents, however, blocked the release of Tf from the cells. The block with azide could not be overcome with excess ligand while the N&C1 block could. If there is a receptor cycle that accompanies the ligand cycle then we would predict that the continued cycling of this system will be blocked due to the failure to return internal receptors to the surface. In fact both of these agents inhibit the continued iron uptake by the cells (7, 19). The ability to overcome the NH&l exit block with excess ligand again demonstrates the coupling of receptor occupancy and internalization to the completion of the transferrin cycle.
Although the transferrin cycle is now well established, it remains intuitively peculiar that the endocytosis of transferrin should occur at all, since its function is simply the delivery of iron to the cell. It would appear to be simpler for the transferrin to remain at the cell surface, deposit its iron, and depart for another load. A possibility, mentioned earlier, is that an acid environment is important for release of iron from transferrin. Since this can be achieved only in a closed vesicular compartment, endocytosis of the transferrin would be obligatory. Alternatively, uptake may be part of a more general process in which polypeptide ligands are internalized after binding to their receptors, irrespective of their function. Such receptor-mediated endocytosis of polypeptides comprises a growing list of participants as the process is studied more closely (18).
A common finding with the ligands studied so far, such as asialoglycoproteins and low density lipoproteins, is that after endocytosis they are shuttled to the lysosomes and degraded. Although transferrin also enters the cell by receptor-mediated endocytosis, it is not degraded and emerges intact in the cell medium (Fig. 10). Recently, Regoeczi and co-workers have reported that asialotransferrin can be endocytosed into hepatocytes via the asialoglycoprotein receptor and subsequently released into the medium without being degraded (20). Since transferrin is not resistant to proteases as is a-2-macroglobulin, for instance, its lack of degradation during its sojourn through the cell means it must either bypass the lysosome or have a very fleeting stay. In a series of studies recently completed in our laboratory using Percoll gradients to achieve fractionation of homogenized cells, we have obtained evidence that the internalized transferrin enters an acidic compartment distinct from the lysosomes and at no point resides within lysosomes (21). Continued studies of the receptor-mediated endocytosis of transferrin may lead to a further understanding of this fascinating problem of intracellular trafficking of ligands.