Degradation of AP2 During Reticulocyte Maturation Enhances Binding of Hsc70 and Alix to a Common Site on TfR for Sorting into Exosomes

Authors


Corresponding author: Michel Vidal, mvidal@@univ-montp2.fr

Abstract

Reticulocytes release small membrane vesicles termed exosomes during their maturation in erythrocytes. The transferrin receptor (TfR) is completely lost from the red cell surface by its segregation in the secreted vesicles where it interacts with the heat shock cognate 70 kDa protein (hsc70). We have now determined a region of the TfR that can potentially interact with hsc70. The peptide P1 (YTRFSLARQV) from the TfR cytosolic domain: (i) binds to hsc70 (ii) with an increased affinity in oxidative conditions, (iii) competes for binding of an unfolded protein to hsc70, and (iv) inhibits the interaction of hsc70 with a recombinant protein corresponding to the cytosolic domain of the receptor. This peptide encompasses the internalization motif (YTRF) of the receptor, and accordingly an affinity column made with the immobilized peptide retains hsc70 and also the AP2 adaptor complex. On the other hand, we show that AP2 is degraded by the proteasome system during reticulocyte maturation and that the presence of the proteasome inhibitor during in vitro red cell maturation inhibits AP2 degradation and specifically decreases TfR secretion via exosomes. Finally, coimmunoprecipitation of Alix with the exosomal TfR, and binding of P1 peptide to the Alix homolog PalA suggest that Alix also interacts with the YTRF motif and contributes to exosomal TfR sorting.

A reticulocyte remodels its membrane surface during its maturation in the bloodstream by losing different proteins, and mainly the transferrin receptor (TfR). These membrane proteins are endocytosed and sorted in portions of the endosomal membrane that bud into the lumen of the endocytic compartment, leading to the formation of multivesicular intracellular structures. Fusion of these multivesicular bodies with the plasma membrane leads to the release of internal vesicles (then called exosomes) into the medium (1). Exosomes are also secreted by other cell types, mainly of hematopoietic origin (2). The molecular basis of protein sorting during exosome formation is not fully understood. In reticulocyte exosomes, the heat shock cognate 70 kDa protein (hsc70) has been shown to interact with the cytosolic domain of TfR and it was suggested that this interaction induces sorting of the receptors into exosomes (3). We recently demonstrated that hsc70 binds to exosomal TfR with the characteristics expected of a chaperone/peptide interaction (4). On the other hand, machinery involved in protein sorting during multivesicular body (MVB) formation has recently been identified. At least three protein complexes, named endosomal sorting complex, responsible for transport (ESCRT-I, II and III), were shown to be subsequently involved in the sorting process, including ubiquitination and deubiquitination of selected cargo proteins (5,6). This ubiquitin-dependent endosomal sorting machinery could be relevant for the biogenesis of exosomes since different proteins involved in MVB sorting, such as the E3-ubiquitin ligase c-Cbl (7) and components of the class E vacuolar protein sorting (VPS) machinery, including Tsg101 (8) and Alix (9), have been found in exosomes (10–12).

During the final erythroid differentiation stage, the reticulocyte also gets rid of all of its intracellular compartments to take on the appearance of a typical erythrocytic ‘hemoglobin bag’. Mitochondria are the last organelles lost by reticulocytes since they are still needed at this stage for ATP production and heme synthesis. They are the main physiologic substrates of proteasomes during reticulocyte maturation, although cytosolic proteins are also degraded. Protein turnover via the ATP ubiquitin-dependent pathway proceeds through degradative intermediates generated by the ATP-coupled covalent conjugation of ubiquitin to susceptible target proteins (13). During erythroid differentiation, the ubiquitin pathway does not seem to be activated in a general manner, but rather through the specific induction of certain conjugating enzymes. Indeed, genes encoding the ubiquitin-conjugating enzymes E2–20K and E2–230K are induced in erythroid cells (14). Moreover, the high levels of both proteins noted in reticulocytes decrease in mature red cells, suggesting a function in the differentiation process. Induction of these enzymes could be a way to selectively degrade some proteins during reticulocyte maturation.

Moreover, clathrin-coated pits, i.e. the structures in charge of surface receptor internalization, are never found on the erythrocyte surface, suggesting that not only cargo molecules (e.g. TfR), but also constituents of the endocytic machinery are lost during maturation. The cytosolic adaptor complex AP2 binds to clathrin by its β2 subunit, which can drive clathrin cage polymerization in vitro (15). AP2 is also a critical component required for membrane receptor segregation in clathrin-coated pits, which occurs by interaction of its μ2 subunit with internalization motifs, e.g. YXXΦ (where Φ is a bulky hydrophobic amino acid and X represents any amino acid), present on the cytosolic domain of receptors (16).

Here, we present data showing that AP2 is downregulated during reticulocyte maturation by the proteasome system. Consequently, this likely favors binding of hsc70 to TfR on a region that encompasses the tyrosine signal (YTRF), and this binding is also favored by the oxidative conditions encountered in the cytosol of maturing reticulocytes. Interestingly, inhibition of AP2 degradation during in vitro reticulocyte maturation decreases TfR secretion via exosomes. Coimmunoprecipitation of Alix with the exosomal TfR and data showing that an Alix homolog binds to a peptide from the TfR cytosolic domain suggest that AP2 degradation allows Alix binding to the YXXΦ motif, which in turn may induce TfR sorting in exosomes.

Results

We previously used immunoprecipitation of exosomal TfR to characterize the interaction of this receptor with hsc70 (4). As this interaction occurs with the cytosolic part of the receptor, we expressed a recombinant protein corresponding to the TfR cytosolic domain fused with a His tag. As shown in Figure 1A, the protein of about 10 kDa, recognized by an antibody directed against the N-terminal region of the TfR, was expressed only after induction of bacteria and could be purified using the His tag (arrow). However, a significant fraction of the protein was systematically eluted with the flow-through, likely due to a propensity of the protein to aggregate. We used this recombinant protein, which we called cytTfR, immobilized on the surface of agarose beads through the His tag, in a pull-down assay in which purified hsc70 from rat brain (4), or reticulocyte cytosol were incubated with beads bearing the cytTfR or not. Hsc70, either from the purified batch or present in the reticulocyte cytosol, was co-pelleted only when cytTfR was immobilized on the bead surface (Figure 1B). This interaction occurred with nucleotide characteristics similar to that described (4) with the immunoprecipitated exosomal TfR (not shown).

Figure 1.

Figure 1.

    Interaction of cytTfR with hsc70.A, Lysates from IPTG-induced or noninduced bacteria transformed to express the recombinant cytTfR (see Materials and Methods) were loaded on an Ni-NTA-agarose column. Fractions recovered in the flow-through (FT) or eluted with increasing concentrations of imidazole were analyzed by Coomassie blue staining (lower panel) or by Western blot using an antibody directed against the cytosolic domain of the TfR (upper panel). B, Ni-NTA-agarose beads (20 μL of packed gel) with or without bound cytTfR were obtained as in A, and incubated with rat brain-purified hsc70 (9 μg) or with reticulocyte cytosol (800 μg). Hsc70 pelleted with the beads was assessed by Western blot after SDS-PAGE separation and membrane transfer.

    Hsp70 molecular chaperones interact with newly synthesized or unfolded polypeptides that generally expose hydrophobic stretches. Previous studies found that peptides with high aromatic and hydrophobic amino acid contents are preferential binding substrates for the endoplasmic reticulum hsp70, BiP (17). Moreover, the presence of basic amino acids within these peptides was reported to favor hsc70 binding whereas negatively charged amino acids diminished the interaction (18). The amino acid sequence 20YTRFSLARQV29 present on the TfR cytosolic domain could be in line with such a ‘consensus’ recognition sequence, and we hypothesized that it might constitute a binding domain of hsc70 on exosomal TfR. We thus synthesized a peptide (P1) corresponding to this sequence in order to test this hypothesis. Since TfR can be phosphorylated on serine 24 by protein kinase C (19), we also synthesized a peptide (P2) in which the serine was replaced by a glutamic acid to mimic phosphorylation.

    First, we determined whether the peptides compete with reduced carboxymethylated lactalbumin (Rcmla), a well-known unfolded protein (20), for hsc70. In this assay, incubation of hsc70 and Rcmla leads to the formation of a complex of the two proteins that can be revealed by Coomassie blue staining after electrophoresis in nondenaturing conditions (Figure 2A). Note that the complex formed (asterisks) migrates faster than hsc70, as previously described (18). We verified that the complex contained both proteins, as demonstrated by hsc70 Western blotting (Figure 2B) and using radiolabeled Rcmla (Figure 2C). Note that doublet complex migration may be due to Rcmla that migrates as two major bands (arrowheads in Figure 2C). When P1 was added during incubation, complex formation decreased, as indicated by Coomassie blue staining, hsc70 Western blot and the 125I-Rcmla content (Figure 2A,B,C). The P2 peptide, in which the serine was replaced by a glutamic acid, did not present any significant differences in inhibitory potential as compared to P1. The P10K peptide (PLSRTLSVAAKK), demonstrated to be one of the most powerful peptides to induce competition for the formation of a complex between Rcmla and hsc70 (18), was taken to estimate the binding affinity of P1 and P2. Interestingly, this P10K peptide was less effective in inhibiting the hsc70–Rcmla interaction compared to P1 (Figure 2A,B). Western blot (Figure 2B) quantification indicated that about 24% and 18% of the original hsc70 content was found in the complex when P1 and P2 were present, respectively, compared to 55% when P10K was added during incubation. We then tested the P1 peptide in our pull-down assay using cytTfR immobilized on the surface of beads. As shown in Figure 2D, the presence of P1 in the assay completely abolished hsc70 pelleting by cytTfR immobilized on beads, indicating competition of cytTfR and P1 for the substrate binding domain of hsc70, while the irrelevant peptide Act10 did not affect hsc70 binding (not shown).

    Figure 2.

    Figure 2.

      P1 peptide competition for hsc70 binding. Recom- binant hsc70 was incubated with reduced carboxymethylated lactal- bumin in the absence or presence of a 300-fold excess molar ratio of the indicated peptides, as described in Materials and Methods. After separation by electrophoresis in nondenaturing conditions, gels were analyzed by Coomassie blue staining (A)and hsc70 Western blotting (B), or by autoradiography using 125I-Rcmla (C).Asterisks indicate bands corresponding to Rcmla/hsc70 complexes. D, the hsc70 pull-down assay was carried out as described in Figure 1B, with or without the presence of a large excess P1 peptide. Hsc70 was loaded separately as a migration control.

      Further evidence of peptide binding by hsc70 was assessed after biotinylation of P1 and P2 peptides, and detection using streptavidin-peroxidase after migration in nondenaturing conditions (Figure 3A). In these conditions, we demonstrated a direct interaction between the peptides and hsc70, but not with bovine serum albumin and transferrin, two proteins used as controls. Here again, no difference between P1 and P2 was detected. While these experiments were underway, Ménoret and colleagues published their work describing changes of affinity between hsc70 and substrate peptides as a function of oxidative conditions (21). Moreover, they showed that short radiolabeled peptides remained bound to hsc70 even after electrophoresis migration in denaturing conditions. We thus set up a similar protocol to study interactions between hsc70 and P1 using biotinylated peptide (P1-biot) and streptavidin-peroxidase overlay after SDS-PAGE and membrane transfer. In these conditions, hsc70/P1-biot complex formation could be decreased by addition of unlabeled P1, while an irrelevant peptide (Act10) and a peptide (HbS) with intermediate affinity for binding to hsc70 (18) were not as efficient (Figure 3B). As shown in Figure 3(C), oxidative conditions (1 mm H2O2) markedly increased the amount of peptide bound to hsc70, whereas reducing conditions (1 mm DTT) diminished the interaction, in agreement with published data (21). Quantification of the blot indicated that peptide binding was 3-fold increased in oxidative conditions compared to the control, and amounted to 39% of control binding in reducing conditions. Moreover, interactions between hsc70 and P1 in oxidative conditions occurred with nucleotide characteristics of the chaperone (Figure 3D). Note that, unexpectedly, although similar amounts of protein were used in both conditions (ADP and ATPγS) and loaded on the gel, the amount of hsc70 detected by Western blot at its normal molecular weight was lower when loaded with ADP (lower panel), meanwhile the amount of peptide associated with the chaperone was higher (upper panel).

      Figure 3.

      Figure 3.

        Interaction of biotinylated peptide with hsc70.A, Biotinylated peptides (P1-biot and P2-biot) were incubated with recombinant bovine hsc70, bovine serum albumin (BSA) or human transferrin (Tf), and the protein mixture was subjected to non-denaturing gel electrophoresis, transferred (without SDS) onto membrane and the protein–peptide complexes were revealed by streptavidin-peroxidase overlay and the ECL method. Note the faint bands corresponding to migration of BSA and Tf in nondenaturing conditions. B, GST-hsc70 (2 μg per assay, 1 μm final concentration) was preincubated with 500 μm peptides Act10, P1, HbS or without peptide (Ctrl) for 30 min at 30 °C before addition of 500 μm biotinylated P1 peptide (P1-biot) for 30 min at 30 °C. Chaperone/P1-biot complex was revealed by peroxidase-conjugated streptavidin overlay after separation on SDS-PAGE (10%) and transfer onto PVDF membrane. C, Biotinylated P1 peptide was incubated with GST-hsc70 (2 μg) in the indicated redox conditions. The protein–peptide complexes were analyzed after SDS-PAGE and membrane transfer, by streptavidin-peroxidase overlay (upper panel) and hsc70 Western blot (lower panel). D, Protein–peptide complex formation was carried out in duplicate by incubation of GST-hsc70 and biotinylated peptide in buffer containing 1 mm H2O2 supplemented with ADP (5 mm) or ATPγS (5 mm), and revealed as described in C.

        Very interestingly, the TfR internalization motif (YTRF) is part of the sequence that was found to interact with hsc70. This TfR region is known to associate with the AP2 complex through binding of the tyrosine motif to the μ2 subunit of AP2 (22). Hence, in order to determine whether the P1 peptide could also bind AP2, we coupled the peptide on diaminodipropylamine agarose beads through its carboxy terminus. Recombinant GST-hsc70 and AP2 complexes enriched from rat brain were loaded on the P1-affinity column. As shown in Figure 4A, in both cases the proteins were retained on the column and then eluted with an acid wash, whereas transferrin did not bind to the affinity column. The low molecular weight bands detected with the anti α-adaptin antibody may correspond to the N-terminal trunk, resulting from proteolytic cleavage of the α-subunit (22). Moreover, when reticulocyte cytosol was loaded on the column, endogenous hsc70 was retained on the gel (Figure 4B), and the chaperone eluted in the flow-through was likely due to the large excess of hsc70 present in the reticulocyte cytosol. Note that we did not observe any AP2 binding using cytosol, due to its low AP2 adaptor complex content compared to hsc70.

        Figure 4.

        Figure 4.

          Immobilization of hsc70 and AP2 complex on a P1 affinity column.A, P1 peptide was cross-linked to agarose beads as described in Materials and Methods. The peptide column was then loaded with recombinant GST-hsc70, AP2 complex or transferrin, washed with PBS, and proteins bound to the column were released by washing using a glycine buffer (100 mm, pH 2.3). Proteins contained in the fractions were separated by SDS-PAGE, transferred to membrane and revealed by Western blot using the corresponding antibodies. B, Reticulocyte cytosol (400 μg protein) was loaded on the column and proteins were collected as described in A. Fractions were analyzed for their hsc70 content by Western blot.

          Hsc70 and the AP2 complex interacted with the same TfR region. Hsc70 was found to be highly concentrated in exosomes whereas AP2 was absent from the vesicles (not shown), suggesting that the chaperone displaces the adaptor complex during reticulocyte maturation. We thus assessed the presence of the adaptor complex during the final stage of erythropoiesis. Percoll density gradient allows the separation of cells, yielding essentially pure populations of age-synchronized reticulocytes (23). As shown in Figure 5A, the amount of α-adaptin detected in reticulocytes markedly decreased as a function of their maturation state, contrary to flotillin-1. Moreover, in vitro maturation of the youngest reticulocyte population led to a decrease in α-adaptin associated with cells (Figure 5B). Since AP2 was not lost by reticulocytes via the exosomal pathway, we assessed its intracellular degradation by the proteasomal system. The proteasome inhibitor lactacystin was added at different concentrations during in vitro maturation of young reticulocytes. As shown in Figure 5C, increasing concentrations of lactacystin inhibited α-adaptin degradation, whereas similar amounts of hsc70 were found to be cell-associated. Western blot quantification indicated that without lactacystin, about 43% of the initial amount of adaptor was cell-associated after 48 h maturation, whereas in the presence of the highest inhibitor concentration, about 86% of the initial amount of α-adaptin was detected.

          Figure 5.

          Figure 5.

            Degradation of α-adaptin during reticulocyte maturation.A, Different populations of age-synchronized reticulocytes were obtained by Percoll density gradient centrifugation using blood of anemic rats, as described in Materials and Methods. Analysis of the α-adaptin and flotillin-1 content of the reticulocyte subpopulations (2 μL packed cell volume) was carried out by Western blot after SDS-PAGE and transfer onto membrane. F1–F4 correspond to the Percoll fractions from lower to higher density, i.e. from younger to older reticulocyte stages. B, Fraction F1 (i.e. young reticulocytes) was matured in vitro as described in Materials and Methods. After the indicated periods of time (days 1–6), cells were analyzed for their α-adaptin content as above. C, Young reticulocytes were matured in vitro for 48 h in the absence or presence of the indicated concentrations of lactacystin. Cells were then analyzed for their α-adaptin content and hsc70 by Western blot.

            We then assessed the in situ effects of α-adaptin degradation. First, we carried out an immunofluorescence analysis to localize the AP2 complex in reticulocytes. As shown in Figure 6, a characteristic punctated pattern (panel B) compared to the control (without primary antibody, panel D) was obtained with young reticulocytes before in vitro maturation. This typical punctated pattern likely corresponded to assembled AP2 complexes with clathrin-coated pits and/or coated vesicles. After 48 h in vitro maturation, α-adaptin immunolabeling was much weaker (panel E), but was partly conserved when the proteasome inhibitor MG132 was present during maturation (panel F). The corresponding Western blots of α-adaptin associated with cells (panel G) perfectly agreed with the immunolabeling data. To assess the influence of AP2 degradation on membrane traffic during reticulocyte maturation, we quantified the cell endocytic capacity before and after maturation in the presence or absence of proteasome inhibitor. Reticulocytes were thus briefly pulsed with transferrin-rhodamine (Tf-Rh) and analyzed by cytofluorimetry. Classically, as controlled by fluorescence microscopy, Tf-Rh internalization was temperature dependent, and the punctated pattern obtained was typical of vesicular labeling (Figure 7A, upper panel). Cytofluorimetry allowed quantification of the amount of Tf-Rh internalized by freshly prepared reticulocytes (t0) and after 48 h maturation. As shown, cell labeling substantially decreased after maturation, but to a lesser extent when MG132 was present during subculture. Since more Tf-Rh remained cell-associated in the presence of proteasome inhibitor, it could be possible that less TfR was lost by reticulocytes via exosomes. Indeed, Western blot analysis of exosomes secreted in these conditions demonstrated that the presence of MG132 during in vitro reticulocyte maturation inhibited TfR secretion via exosomes (Figure 7B). Moreover, this was likely due to a specific impairment in TfR sorting, since the exosomal flotillin-1 content was much less affected by the presence of the proteasome inhibitor, as indicated by the Western blot quantification results.

            Figure 6.

            Figure 6.

              Disappearance of the AP2 complex during reticulocyte maturation. Freshly isolated reticulocytes were adsorbed on glass cover slips and processed for immunofluorescence as described in the Materials and Methods. Nomarski (A, C) and corresponding fluorescence imaging (B, D) was carried out on cells incubated in the presence (A, B) or absence (C, D) of antirat β-adaptin antibody, followed by incubation with FITC-conjugated antimouse antibody. Young reticulocytes were matured in vitro for 48 h in the absence (E) or presence of MG132 (F), and processed for AP2 immunofluorescence as in (B). Aliquots (2 μL packed cell volume) of cell suspensions corresponding to (B, E, F) were processed by Western blot for detection of α-adaptin (G).

              Figure 7.

              Figure 7.

                MG132 inhibits the loss of TfR during maturation.A, Young reticulocytes were observed by fluorescence microscopy after a short pulse (30 min) with Tf-Rh (20 μg/mL) either at 4 °C or 37 °C (upper panel). Cytofluorimetry analysis was carried out on freshly prepared reticulocytes (t0) or matured cells (48 h) in the presence or absence of MG132 (4 μm), after Tf-Rh labeling in the presence (bold trace) or absence (thin trace) of unlabeled Tf (80-fold excess). B, Exosomes collected from the supernatant after reticulocyte in vitro maturation (48 h) in the presence or absence of MG132 (4 μm) were analyzed for TfR and flotillin-1 by Western blot. The band (+ MG132) quantification results are given on the right of the blot as percentages of the corresponding bands (– MG132) set as 100%.

                AP2 degradation was correlated with TfR secretion in exosomes, indicating that exposure of the AP2 domain was a critical event for TfR sorting in MVB. Since we demonstrated that hsc70 binding decreases targeting of TfR molecules towards the exosome pathway (4), we hypothesized that binding of another protein to the same domain may be involved in TfR exosomal sorting. Alix, a protein first identified as a protein involved in apoptosis (24), was recently demonstrated to bind the YPXL/I protein interaction motif (25) and identified as a component of the viral budding machinery linking the YPXL-type (26) l-domain of various Gag proteins to the ESCRT machinery (26,27). Interestingly, Alix has been identified by mass spectroscopy in exosomes secreted by dendritic cells (10). As shown in Figure 8A, Alix was also detected by Western blot in reticulocyte exosomes and, more importantly, Alix was coimmunoprecipitated with the TfR. Moreover, we demonstrated that PalA, the Aspergillus homolog of Alix, was retained in the P1-affinity column (Figure 8B). Conversely, PalA adsorbed on plastic wells bound to biotinylated P1 peptide (Figure 8C).

                Figure 8.

                Figure 8.

                  Alix binds in vivo and in vitro to TfR.A, Rat reticulocyte exosomes were lysed in RIPA buffer and subjected to TfR immunoprecipitation as previously described (4). Mouse monoclonal anti-TfR (CL071) was used for immunoprecipitation and a nonrelevant antibody from the same species was used as control. Lysate: 20 μL of lysate was loaded on the gel; Bound: protein A-Sepharose beads were mixed with Laemmli buffer and loaded on the gel. Proteins were transferred to PVDF membrane and immunoblotted for TfR and Alix. B, GST-PalA was expressed in bacteria, purified on GSH-Sepharose and treated by thrombin. 50 μg of recovered PalA was loaded on the P1 affinity column and analyzed as described in Figure 4, using the anti-Alix for Western blot detection. C, Microplate wells were coated with GST or GST-PalA (100 μL at 100 μg/mL) and blocked by gelatin, as described in Materials and Methods. After washing, biotinylated P1 peptide (200 μm) binding was quantified as described in Materials and Methods. Peroxidase activity of biotinylated peptide incubated in gelatin-coated plates is presented (CTRL). Data are presented as mean ± SD (triplicate).

                  Discussion

                  We have previously shown that hsc70 binding to the TfR cytosolic domain in reticulocyte exosomes occurs with specific chaperone characteristics (4). Moreover, inhibition of hsc70 binding to TfR during in vitro reticulocyte maturation increased the amount of TfR secreted via exosomes, indicating that the chaperone binding was probably not the direct cause of receptor sorting, as previously suggested (3). Rather, in agreement with our previous data (28), we hypothesized that receptor aggregation was recognized as a sorting signal. However, the fact that TfR was specifically enriched in reticulocyte exosomes and not in vesicles secreted by other cell types, together with the finding that hsc70 binds to TfR in exosomes and not at the plasma membrane, indicates that hsc70 binding is a key event of TfR sorting in reticulocyte exosomes. We thus decided to conduct an in-depth study of this interaction between hsc70 and TfR by looking for its cause, determining first the binding site on the TfR. As previously described, members of the hsp70 family generally bind to hydrophobic stretches presented by unfolded proteins either during their translation or upon modification. A ‘consensus’ sequence of at least seven residues with high aromatic and hydrophobic amino acid contents has been described for interactions with proteins of the hsp70 family (17), with the presence of basic amino acids favoring interaction with the hsc70 cognate form (18). Such an amino acid sequence can be found in the N-terminal region of the TfR cytosolic domain, whereas residues near the transmembrane domain are more acidic and less hydrophobic. We thus hypothesized that the 20YTRFSLARQV29 sequence, possessing characteristics of a potential hsp70 interacting peptide, could be the TfR binding site of the chaperone. The presence in this sequence of a serine that can be phosphorylated by protein kinase C (19) suggested that the hsc70 interaction could be regulated depending on the phosphorylation state of the receptor. We indeed found that a synthetic peptide (P1) corresponding to this sequence could bind to the chaperone, as shown by competition with reduced carboxymethylated lactalbumin for hsc70 binding in a well-defined assay (18). Moreover, comparison with a peptide (P10K) already tested in this assay and rated as a high affinity peptide for hsc70 (18), indicated that the TfR synthetic peptide has a high binding affinity for the chaperone. Surprisingly, replacing the serine by a glutamic acid to mimic phosphorylation in the other synthetic peptide (P2) did not lead to a significant difference in binding affinity for hsc70 in the in vitro assay. Accordingly, both peptides were able to stimulate hsc70 ATPase activity (not shown), as already shown for the entire exosomal TfR (4). These data thus suggest that TfR phosphorylation by protein kinase C is not involved in hsc70 binding to exosomal TfR. The YTRFSLARQV sequence was confirmed to be a potential binding site for hsc70 through a pull-down assay that we set up using the TfR cytosolic domain expressed as a recombinant protein for interaction with the chaperone. The assay also showed that the P1 peptide completely abolished cytTfR binding to hsc70. Direct evidence of specific peptide binding to hsc70 was obtained using biotinylated peptides. The overall results of these experiments indicated that the YTRFSLARQV sequence could be the binding site for hsc70 on the TfR cytosolic domain. If this is the case in vivo, hsc70 binding to exosomal TfR could impair further phosphorylation of the serine located in the binding pocket of the chaperone. This would be in agreement with data showing that exosomal TfR, contrary to cell-associated TfR, could not be phosphorylated in vitro by exogenously added protein kinase C, and suggesting that the phosphorylation site was no longer accessible (29).

                  Our next goal was to determine the signal that triggers hsc70 binding. As already mentioned, chaperone binding can be associated with the stabilization of unfolded polypeptides. It is well known that mitochondrial degeneration is a crucial event during reticulocyte maturation (30). We assessed, by flow cytometric experiments (not shown), the formation of reactive oxygen species (ROS) and loss of mitochondrial membrane potential (Δψm), as described (31). ROS likely contribute to mild oxidation of proteins during reticulocyte maturation, as we detected (not shown) by the formation of carbonyl groups (32). Mild oxidation of proteins was shown to target them to proteasome degradation (33). Moreover, oxidative conditions have been demonstrated to influence the binding of immunogenic peptides to hsp70 and hsc70 (21). As in the case of radiolabeled immunogenic peptides, biotinylated P1 remained associated with hsc70 after SDS-PAGE, and allowed quantification of the interaction between hsc70 and the peptide in various redox conditions. As reported for other peptides (21), P1 binding to hsc70 was improved under an oxidative environment and decreased in reducing conditions, in agreement with the conformational change of hsc70 in oxidative conditions, which is more favorable for peptide association (21). The fact that the stress-inducible form (hsp70) can no longer be synthesized by reticulocytes during this oxidative stress period likely contributes to mobilization of the cognate form (hsc70) due to its chaperoning function, including interaction with the TfR cytosolic domain.

                  A crucial point is that this sequence of the TfR cytosolic domain encompasses the internalization motif (YTRF) of the receptor. This motif has been described to bind to the μ2 subunit of the AP2 adaptor complex. We thus hypothesized that the TfR sequence, which is engaged with partners of the endocytic machinery such as AP2 and consequently inaccessible to hsc70, could be unmasked after degradation of the adaptor complex. In this scenario, the TfR cytosolic domain would not be modified to interact with hsc70. Rather, hsc70 chaperones the receptor when the internalization motif is not interacting with AP2. In this respect, it is very important to note that a physiologic role of hsc70 in facilitating adaptor complex release from clathrin-coated vesicles has already been described (34). We first verified that an affinity column involving P1 peptide cross-linking on Sepharose beads could retain both the AP2 complex and hsc70. A similar TfR peptide (15GEPLSYTRFSLARQVDG31) was already tested for AP2 binding and revealed a somewhat higher affinity than other tyrosine-based signals (Dr Stefan Höning, personal communication). Moreover, we demonstrated that the AP2 complex disappeared during reticulocyte maturation and that at least the α-subunit was degraded during this period. Note that α-adaptin degradation was relatively slow when tested in in vitro maturation conditions, but its disappearance was clear when subpopulations of different maturation states were isolated from blood, likely due to in vitro maturation limitations. This degradation is carried out by the proteasomal system, as demonstrated by using two inhibitors (MG132 and lactacystin). Moreover, proteasome inhibition during in vitro reticulocyte maturation diminished the loss of endocytic capacities, as shown by transferrin uptake, and this was concomitant with a decrease in TfR secretion via exosomes. Most interestingly, flotillin-1, another protein released in exosomes (12), was not significantly affected by proteasome inhibitors, indicating a specific effect on the TfR sorting event. Taken together, these data strongly suggest a causal relationship between AP2 disappearance and TfR sorting in exosomes. Similarly, it is possible that proteasomes could also degrade other components of the adaptor sorting machinery such as the adaptor-associated kinase AAK1 (35), which would decrease μ2 affinity for the tyrosine motif (36) and favor hsc70 binding to TfR, even in the presence of a significant amount of AP2 left in the cells. However, in agreement with our previous study which demonstrated that blocking hsc70 binding to TfR increases receptor secretion via exosomes (4), the present work suggests that another protein, namely Alix, may be the binding partner that induces TfR exosomal sorting. Indeed, different groups have recently demonstrated that Alix is directly involved in virus budding by linking the YPXL-type l-domain of Gag proteins to the ESCRT machinery (26,27). Remarkably, the equine infectious anemia virus (EIAV) p9 Gag protein containing a YPXL motif has been shown to colocalize with AP2 complexes (37), probably reflecting an interaction between the different binding partners (i.e. AP2/Alix/YXXΦ) depending on the protein expression balance. In the case of maturing reticulocytes, AP2 degradation would increase Alix binding to the YXXΦ motif and induce TfR sorting in MVB by the ESCRT machinery. Indeed, Alix can also simultaneously bridge ESCRT-I to ESCRT-III through binding to Tsg101 and CHMP4 proteins, respectively (9,26,27). In accordance with this, we found Tsg101 in reticulocyte exosomes (12). Moreover, sorting through direct binding of Alix to the TfR could also overcome the need for TfR ubiquitination, as reported for p9 Gag during EIAV budding (38). Indeed, although four lysine residues are present near the transmembrane region, our trials to obtain evidence of TfR ubiquitination in vitro have been unsuccessful so far.

                  In summary, hsc70 binding to TfR was found to occur in a region encompassing the internalization motif recognized by the AP2 adaptor complex. This could reflect a physiologic function of the chaperone, facilitating AP2 complex recycling by interacting with the same TfR domain. In maturing reticulocytes, chaperone binding would be favored by AP2 degradation by the proteasomal system, and by a change in redox status due to mitotosis. Alix, could also bind to the same region and link the TfR to the ESCRT sorting machinery. The fact that these events are very specific to the final red cell differentiation stage could explain the specific enrichment of TfR in exosomes secreted by reticulocytes as compared to other cell types.

                  Materials and Methods

                  Reagents

                  Protein A-Sepharose and GSH-Sepharose gels, peroxidase-conjugated streptavidin (RPN 1231 V) and ECL reagents were obtained from Amersham Biosciences Europe GmbH (Orsay, France). ATP-agarose and Ni-NTA-agarose beads were from Sigma (St. Louis, MO) and Qiagen S.A. (Courtaboeuf, France), respectively. The proteasome inhibitors lactacystin (clasto-lactacystin β-lactone) and MG132 were from Calbiochem (Schwalbach, Germany). Tetramethylrhodamine-conjugated transferrin (T-2872), MitoTracker Red CMXRos (M-7512) and dihydroethidium (D-1168) were from Molecular Probes (Eugene, OR). Purified recombinant bovine hsc70 was from StressGen (Victoria, Canada). Peptides P1 (YTRFSLARQV) and P2 (YTRFELARQV) were synthesized by Synt:em (Nîmes, France) and biotinylated using N-(4-aminobenzoyl)biocytin (A-1604) from Molecular Probes. O-phenylenediamine (OPD), human transferrin, reduced carboxymethylated lactalbumin (Rcmla) and the peptides P10K (PLSRTLSVAAKK; catalog no. P5307) and HbS (VHLTPVEK; catalog no. V5130) were obtained from Sigma. Peptide Act10 (DDEETALVCD) was a gift of C. Roustan (UMR 5539, Montpellier).

                  DNA constructs cDNA corresponding to GST fused with human hsc70 was a generous gift from F. Dice (Tufts University School of Medicine, Boston, MA). The plasmid encoding for GST-PalA (25) was the kind gift of O. Vincent (Department Microbiología Molecular, CSIC, Madrid, Spain). cDNA encoding the full length TfR was kindly provided by C. Enns (Oregon Health & Science University, Portland, OR). A polymerase chain reaction (PCR) product encoding the cytoplasmic tail of TfR was generated using two oligos (Sigma-Genosys, Cambridge, UK) introducing 3' Xho (TTCCACATCACCTTCATAGACGGAGCTCGTCGCA) and 5' NdeI (GGAATTCCATATGATGGATCAAGCTAGATCAGCAT) restriction sites. Fusion of a 6His tag at the C-terminus of the cytosolic domain of TfR was obtained by ligation of Xho/NdeI digested PCR products to Xho/NdeI digested pET-23a-d(+) vector (Novagen, Madison, WI). The construction was then checked by sequence analysis (Genome express, Meylan, France) before use.

                  Expression and purification of recombinant proteins

                  Recombinant GST-hsc70 was expressed in JM109 cells induced with 0.15 mm IPTG (isopropyl-1-thio-β-d-galactopyranoside) for 2 h, GST-PalA was expressed in BL21 cells induced with 1 mm IPTG for 2 h, and the proteins were purified on GSH-Sepharose. When indicated, the protein of interest was cleaved from GST by thrombin treatment according to the manufacturer's instructions (Amersham Biosciences). The purity and size of the proteins were verified by SDS-PAGE (10% gel) and Western blotting. The proteins were then dialyzed overnight against 50 mm NaCl, 20 mm Tris-HCl, pH 7.

                  The DNA construct corresponding to cytTfR was introduced in BL21 cells. Protein expression was induced with 1 mm IPTG for 2 h at 30 °C. After centrifugation, the pellet was resuspended with binding buffer (5 mm imidazole, 0.5 m NaCl, 20 mm Tris, 0.1% Triton X100, 100 μg/mL lysozyme and protease inhibitors) and the cells lyzed by sonication. After 15 min centrifugation at 15 000 × g, the supernatant was incubated for 2 h with Ni-NTA-agarose beads. Beads were washed three times and retained proteins were eluted by an imidazole gradient (20–100 mm). The protein purity and size were verified on 15% SDS-PAGE gel and Western blots.

                  Antibodies

                  Rat monoclonal antibody raised against hsc70 (SPA-815) was from StressGen. Mouse monoclonal antihuman transferrin receptor raised against the cytoplasmic tail of the receptor was from Zymed Laboratories Inc. (South San Francisco, CA) and was used for Western blot. Mouse monoclonal antirat TfR antibody (CL071) was obtained from Cedarlane (Hornby, Canada) and was used for immunoprecipitation. Mouse monoclonal antiflotillin-1 antibody (clone-18) and mouse monoclonal antibody (clone-8) raised against the α-subunit (α-adaptin) of the AP2 complex, used for Western blotting, were from BD Transduction Laboratories (BD Biosciences, Heidelberg, Germany). Anti-human transferrin developed in goat was from Sigma Immunochemicals. Peroxidase-conjugated goat antirat IgG, peroxidase-conjugated donkey antimouse IgG and donkey antigoat IgG were obtained from Jackson Immunoresearch Laboratories, Inc. (West Grove, PA). Affinity-purified mouse monoclonal antibody (10 A) against rat β1 and β2-adaptin subunits, used for immunofluorescence, was the gift of T. Kirchhausen (CBR/HMS, Boston, MA). Peroxidase-conjugated donkey antirabbit IgG and FITC-conjugated donkey antimouse IgG were from Rockland (Gilbertsville, PA). Polyclonal rabbit anti-Alix was kindly donated by R. Sadoul (INSERM-UJF, Grenoble, France). This antibody cross-reacts with PalA (not shown).

                  Reticulocyte and exosome preparation

                  Reticulocyte-enriched blood was obtained from phenylhydrazine-treated Sprague-Dawley white rats (28). Exosomes were collected by sequential centrifugation from the supernatant of reticulocyte subcultures, as previously described (39). Briefly, blood samples were centrifuged at 1000 × g for 10 min at 4 °C to remove plasma and the buffy coat. After three washes with Hanks' buffer, the cells (3% final packed cell volume) were cultured in RPMI 1640 supplemented with 5 mm adenosine, 10 mm inosine, 5 mm glutamine, 3% (v/v) fetal calf serum. To assess the effect of MG132 and lactacystin on exosome formation, the compounds were added during in vitro reticulocyte maturation. After the indicated periods of time at 37 °C, the cells were pelleted and the supernatants centrifuged for 20 min at 20 000 × g to remove cellular fragments. Supernatants were then ultracentrifuged for 2 h at 100 000 × g and the vesicle pellets were resuspended in phosphate-buffered saline (PBS).

                  Purification of hsc70 and AP2 complex from rat brain

                  Hsc70 was purified as previously described (4). Briefly, rat brains were homogenized with a Dounce homogenizer and loaded onto a Q-Sepharose column. After column washing, bound proteins were eluted with a KCl gradient (50 to 600 mm). Fractions containing hsc70 (detected by Western blotting) were pooled, dialyzed and loaded onto an ATP-agarose column. The column was then extensively washed, and bound hsc70 was eluted with ATP-containing buffer. Hsc70-containing fractions were then pooled and extensively dialyzed.

                  Clathrin-coated vesicles (CCVs) were isolated as described (34) with minor modifications. Briefly, rat brains were harvested immediately after blood recollection by cardiac puncture, and frozen at − 80 °C for later use. Frozen brains (7 g) were thawed in MES buffer (100 mm MES, pH 6.5, 0.5 mm MgCl2, 1 mm EGTA, 1 mm PMSF, 0.8 mm DTT, 0.02% NaN3) supplemented with a protease inhibitor cocktail (Roche Diagnostics GmbH, Mannheim, Germany) and homogenized using a Dounce homogenizer. The homogenate was centrifuged at 15000 × g for 30 min at 4 °C, and the supernatant was collected and centrifuged at 100 000 ×g for 1 h at 4 °C in a Ti50 rotor (Beckman Coulter, Roissy, France). The CCV pellet was resuspended in 700 μL of MES buffer with protease inhibitors and stored at 4 °C for no longer than 1 week. AP2 complex was released from CCVs by incubation with 0.5 m Tris/HCl, pH 7, 0.2 mm DTT for 30 min at 4 °C. The supernatant obtained after 15 min centrifugation at 120 000 × g in a TLA-100.3 rotor (Beckman Coulter) was then dialyzed overnight against 20 mm Hepes/NaOH, pH 7, 0.15 m NaCl, 10 mm KCl, 2 mm MgCl2, 2 mm DTT.

                  Western blot and overlay analysis

                  Proteins were separated by SDS-PAGE according to Laemmli (40) using 10% polyacrylamide gels, and were electrophoretically transferred (41) to PVDF membrane (Immobilon-P, Millipore, Bedford, MA). Proteins separated by native polyacrylamide gel electrophoresis, as previously described (18), were blotted on membrane using transfer buffer without SDS. Membranes were blocked and incubated with primary antibodies and peroxidase-conjugated secondary antibodies. The corresponding bands were detected using an enhanced chemiluminescence detection kit. Interactions between hsc70 and biotinylated peptides were revealed by overlay with peroxidase-conjugated streptavidin. Blots were scanned and quantified using ImageQuaNT software.

                  Competition of hsc70-Rcmla complex formation by peptides

                  The competition binding assay developed by Fourie et al. (18), involving the formation of a complex between hsc70 and an unfolded form of lactalbumin (Rcmla), was used to assess peptide binding affinity to hsc70. Briefly, 2.4 μg of recombinant hsc70 was incubated with 4.8 μg Rcmla (Rcmla/hsc70 molar ratio = 10), in the presence or absence of the indicated competing peptide (peptide/hsc70 molar ratio = 3000) for 2 h at 37 °C. Free hsc70 (or similarly migrating hsc70-peptide complexes) and free Rcmla were separated from hsc70-Rcmla complexes by native polyacrylamide gel electrophoresis and visualized by subsequent Coomassie blue staining and hsc70 Western blot. In some experiments, Rcmla was radiolabeled as previously described (42), added in the binding assay and revealed by autoradiography.

                  Interaction of proteins in a P1 affinity column

                  P1 peptide was cross-linked through its carboxyl group to immobilized diaminodipropylamine gel using 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide as recommended by the manufacturer (Pierce, Rockford, IL). Purified AP2 complex (50 μg), GST-hsc70 (5 μg), PalA (50 μg), human transferrin (5 μg), or reticulocyte cytosol (400 μg), obtained as described (43), were added to the column (1 mL of packed gel) and incubated for 1 h at room temperature. After extensive washing of the column, affinity-bound proteins were eluted by a 100 mm glycine buffer (pH 2.3). Collected fractions were analyzed for the presence of the indicated proteins by Western blot.

                  P1 peptide microplate binding assay

                  GST-PalA or GST were coated in 96-well microtiter plates (polysorp NUMC Life Technologies) by incubation of 100 μL protein solution (100 μg/mL) in Na carbonate/Na bicarbonate 0.1 m (pH 9.5) overnight at 4 °C. Biotinylated P1 peptide alone was incubated in wells as control. After washing the wells three times with PBS (pH 7.3), nonspecific plastic sites were blocked by 1 h incubation at room temperature with 100 μL of PBS 0.5% gelatin. After three washes with PBS, biotinylated P1 peptide (200 μm) was then incubated overnight at 4 °C. Unbound peptide was carefully washed away using PBS containing 0.1% Tween 20 (PBST). Peroxidase-conjugated streptavidin was distributed (100 μL of the stock solution diluted 1/1000) and after five washes with PBST, peroxidase activity was assayed by adding 100 μL of OPD (4 mm) and H2O2 (0.003%) in sodium acetate (50 mm) buffer (pH 5). The reaction was terminated by adding 50 μL of sulfuric acid (2 n). Plates were read on a plate reader (Labsystems multiscan EX, Helsinki, Finland) at OD450nm. Each point represents the result of triplicate trials.

                  Immunofluorescence and cytofluorimetry

                  Reticulocytes were allowed to adhere to Alcian blue-coated glass coverslips for 30 min at room temperature, washed with PBS and cells were fixed and permeabilized by dipping coverslips in cold methanol (5 min) and then acetone (2 min). After rehydration, 15 min in PBS, coverslips were stained by incubation with anti β-adaptin antibody (1 h, 4 °C) and FITC-conjugated antimouse secondary antibody (30 min, RT) diluted in 1% PBS-FCS. Cells mounted in Mowiol were observed on a Reichert Polyvar fluorescence microscope equipped with Nomarski optics and specific rhodamine and fluorescein excitation and emission filters. Images were captured from a 12-bit CCD camera (Princeton Instruments Inc., Trenton, NJ). Tf-Rh containing reticulocytes were observed directly after labeling (20 μg/mL Tf-Rh) without any processing.

                  Freshly isolated or in vitro matured reticulocytes were labeled (30 min, 37 °C) with Tf-Rh (20 μg/mL) in the presence or absence of a 80-fold excess of unlabeled Tf, and analyzed on a FACSCalibur flow cytometer (BD Biosciences, Pont de Claix, France), with a 488 nm argon laser and standard band pass filters for FL-2 (585/42 nm).

                  Acknowledgments

                  We thank Olivier Vincent, Caroline Enns and Fred Dice for the gift of plasmids, and Rémi Sadoul and Tom Kirchhausen for providing antibodies. We acknowledge Bernard Bayard for helping with cytTfR plasmid construction, and Rose Johnstone for helpful discussions.

                  Supported by grants from the CNRS, the Ministère de la Recherche and the Université Montpellier II. C.G. was supported by fellowships from the Association pour la Recherche sur le Cancer and the Fondation pour la Recherche Médicale.

                  Ancillary