Molecular Characterization of the Signal Responsible for the Targeting of the Interleukin 2 Receptor β Chain toward Intracellular Degradation*

During receptor-mediated endocytosis, most growth factor receptors are transported to late endocytic compartments and degraded. This process is important to control their expression on the cell surface and requires sorting in early endocytic compartments. Little is known about the mechanisms and the signals involved. We have studied the signal involved in targeting the interleukin 2 receptor β chain (IL2Rβ), a member of the cytokine receptor superfamily, toward degradation after internalization. We show that a motif of 8 amino acids in the cytosolic tail of IL2Rβ is sufficient to target a normally recycling receptor toward degradation. Deletion of this signal strongly impairs IL2Rβ degradation. Further molecular characterization of the motif shows that it does not resemble the well documented tyrosine and dileucine families of trafficking signals.

During receptor-mediated endocytosis, most growth factor receptors are transported to late endocytic compartments and degraded. This process is important to control their expression on the cell surface and requires sorting in early endocytic compartments. Little is known about the mechanisms and the signals involved. We have studied the signal involved in targeting the interleukin 2 receptor ␤ chain (IL2R␤), a member of the cytokine receptor superfamily, toward degradation after internalization. We show that a motif of 8 amino acids in the cytosolic tail of IL2R␤ is sufficient to target a normally recycling receptor toward degradation. Deletion of this signal strongly impairs IL2R␤ degradation. Further molecular characterization of the motif shows that it does not resemble the well documented tyrosine and dileucine families of trafficking signals.
Receptor-mediated endocytosis is a process used by cells for rapid and specific uptake of extracellular macromolecules. After their internalization from the plasma membrane, receptors travel from one organelle to another. They may recycle back to the cell surface, thus following the default pathway, or reach other compartments such as late endosomes and lysosomes, the trans-Golgi network, or other specialized compartments. Sorting may take place at several steps during this journey, to ensure that each receptor reaches its proper destination (1).
Sorting mechanisms described to date involve short amino acid sequences carried by the cytosolic region of the membrane protein to be routed. A large majority of these motifs, which target receptors to various intracellular compartments, resemble the well described internalization signals, which allow receptors to interact with clathrin-coated pits (2). They have been classified in two groups, a tyrosine-based motif and a dileucinebased motif (3).
Many internalization signals have now been described, but much less is known about the characteristics of the motifs involved in targeting to late endocytic compartments and to lysosomes. Most of our knowledge comes from studies on lysosomal membrane glycoproteins. After synthesis, these proteins either are directly targeted from the trans-Golgi network to endosomes and lysosomes, or first appear on the cell surface before reaching their final destination. Sequences necessary for targeting lysosomal membrane glycoproteins to lysosomes have been described by mutation analysis (reviewed in Ref. 4). They belong to the tyrosine or dileucine families of signals described for clathrin-coated pit endocytosis. Usually, when these signals are mutated, both the information for endocytosis and for targeting to lysosomes are lost.
Sequences necessary for internalization and degradation of a few nonlysosomal membrane proteins have been reported and are also similar to these internalization motifs, i.e. CD4, CD3 ␥ and ␦, HLA-DM, and the major histocompatibility complex (MHC) 1 associated invariant chain (5)(6)(7)(8). An important category of membrane proteins that are also targeted to late endocytic compartments after endocytosis are the growth factor receptors. In this case, degradation of the receptors plays an important role in controlling their expression and consequently the cellular response to growth factors. Little is known about the motifs involved in their sorting.
The cytokine interleukin 2 (IL2) is produced by activated helper T lymphocytes and stimulates proliferation and effector functions of a variety of cells of the immune system. High affinity IL2 receptors (K d Ϸ 10 -100 pM) are composed of three distinct components, the ␣, ␤, and ␥ chains, which are associated in a noncovalent manner (9). Both the ␤ and ␥ chains, but not the ␣ chain, belong to the cytokine receptor superfamily (10). This hematopoietic cytokine receptor family includes receptors for several cytokines, for erythropoietin, the granulocyte colony-stimulating factor, the granulocyte-macrophage colony-stimulating factor, the leukemia inhibitory factor, growth hormone, prolactin, and ciliary neurotrophic factor. Many receptor subfamily members share at least one component; thus, the receptors for IL2, -4, -7, -9, and -15 have a common ␥ chain, and the receptors for IL2 and IL15 share the ␤ chain (reviewed in Ref. 11).
One of the early events following IL2 binding to high affinity receptors on the cell surface is the internalization of IL2 receptor complexes (12,13). After endocytosis, the components of this multimolecular receptor have different intracellular fates: one of the chains, ␣, recycles to the plasma membrane, while the others, ␤ and ␥, are routed to late endocytic compartments (14).
We have previously shown that the interleukin 2 receptor ␤ chain (IL2R␤) is constitutively endocytosed and degraded when expressed in cells lacking the ␣ or the ␥ chains (15). A truncated form of IL2R␤, ␤ 27 , composed of the extracellular and transmembrane domains followed by the first 27 cytosolic amino acids out of 286, is also efficiently internalized and degraded (16). We have reported that the ␤ 27 construct carries several weak internalization signals within the transmembrane and cytosolic domains, acting in an additive fashion (17). Furthermore, its cytosolic tail contains a signal that is sufficient to target a recycling receptor toward degradation (16). Here we show that this signal is unique since, when it is deleted, IL2R␤ is no longer degraded. We have further characterized this sequence by mutational analysis. It does not have the characteristics of the tyrosine or dileucine families of signals.

EXPERIMENTAL PROCEDURES
Cells and Monoclonal Antibodies-K562, a human erythroleukemia cell line, was grown in suspension in RPMI 1640, 10% decomplemented fetal calf serum, supplemented with 2 mM L-glutamine. Stably transfected K562 cells were grown in the same medium supplemented with 1.5 mg/ml G418.
Plasmids-All ␤ 27 mutants and ␣ Y ␤ nm chimeras were generated by polymerase chain reaction (PCR) as described (16). The truncated form of the ␤ chain lacking cytosolic residues 22-27 (␤ ⌬22-27 ) was constructed by PCR amplification of the ␤ chain region, encompassing the extracellular and transmembrane domains plus the first 21 cytosolic amino acids. A NotI restriction site was introduced in the upstream primer. The downstream primer ATTGAGCTCGAGGGGTCTGGGGTGTTAC contained a SacI restriction site overlapping the junction between residue 21 and residues 28 -29. The PCR product was cloned in the NT expression vector using the NotI/SacI restriction sites. The fragment coding for amino acid 28 to the last amino acid (286) of the ␤ chain was then added using the unique SacI site in IL2R␤ cDNA.
Cell Transfection-All kinetics and half-life analyses described here have been performed in stably transfected K562 cells. To generate stable transfectants, 7 ϫ 10 6 K562 cells were washed once in Dulbecco's modified Eagle's medium, 4.5 g/liter glucose and resuspended in 800 l of the same medium, with 20 g of the plasmid of interest. Electroporation was performed using the Easyject electroporator (Eurogentec) with a single pulse, 240 V, 1500 microfarads. Selection with 1.5 mg/ml G418 (Geneticin, Life Technologies, Inc.) was initiated 2 days after transfection, and the cells were cloned in 96-well dishes. G418-resistant clones were assayed for expression by flow cytometry using anti-␣ (7G7B6) or anti-␤ (341) antibodies. The expression levels of recombinant proteins in all clones tested were the same or less than their normal level in activated lymphocytes.
Internalization Assays-Internalization of the different ␤ constructs was measured using radiolabeled anti-␤ antibody. Antibody 561 was radiolabeled with 125 I by the chloramine T method to a specific activity of 2-10 Ci/g. Cells (2 ϫ 10 6 ) were incubated in 100 l of RPMI-Hepes, pH 7.2, 1 mg/ml bovine serum albumin, at 37°C, and 1-5 nM 125 Ilabeled anti-␤ antibody was added. After incubation at 37°C for the indicated times, the cells were rapidly cooled to 4°C and washed twice. Cell surface-associated radioactive ligand was then removed by two successive acid pH treatments (2 min at pH 2.0) at 4°C as described previously (20). Nonspecific binding, measured for each ligand by adding a 100-fold excess of the same unlabeled ligand, was less than 10% and was subtracted. The efficiency of removal of cell surface-associated ligands by acid pH washes was measured for each ligand and was more than 90%. The data presented here thus show specific binding and internalization.
In all figures, the ratio of intracellular to cell surface-associated 125 I-labeled antibody is represented. All experiments were done at least three times, with different clones expressing the same construct. The means Ϯ S.E. are shown.
Internalization of 2A3A1H antibodies directed against the ␣ Y ␤ nm chimeras were quantitated by flow cytometry as described previously (21). Briefly, cells were incubated at 4°C for 60 min with 2A3A1H (1/200 ascites fluid). The cells were washed in phosphate-buffered saline at 4°C, and after incubation at 37°C for the indicated times, rapidly cooled to 4°C and washed twice in cold phosphate-buffered saline, 2% fetal calf serum. The cells were incubated at 4°C for 1 h with phycoerythrin-conjugated goat F(ab)Ј 2 anti-murine IgG and washed once at 4°C. Internalization was monitored by the decrease in mean fluorescence as a function of time at 37°C. The background fluorescence intensity, determined with cells incubated only with the secondary antibody, was less than 10% and was subtracted.
Cell Surface Half-life Measurements-To measure the half-life on the cell surface of the different ␤ 27 mutants or ␣ Y ␤ nm chimeras, cells were incubated with cycloheximide to prevent the synthesis of new receptors. After different times of incubation at 37°C in culture medium with 50 M cycloheximide (Sigma), the cells were cooled to 4°C, and cell surface expression of the constructs was assayed by flow cytometry as described (22). Time 0 on the graph corresponds to a 30-min incubation in cycloheximide, which is the time required for a newly synthesized IL2 receptor to reach the cell surface (20). All experiments were done at least four times, with different clones expressing the same construct. The means Ϯ S.E. are shown.

The IL2R␤ Cytosolic Domain Contains a Unique Sorting
Signal to Target the Receptor toward Degradation after Endocytosis-After internalization, IL2R␤ is targeted to late endocytic compartments and is degraded. We have previously described a 10-amino acid sequence, in its cytosolic domain, which functions as a sorting signal for the receptor (16). IL2R␤ cytosolic domain is 286 amino acids long and could contain more than one sorting signal. We constructed a mutated receptor in which an essential part of the signal we have characterized (amino acids [22][23][24][25][26][27] was deleted (␤ ⌬22-27 ). To analyze the fate of this receptor during endocytosis, we measured its halflife on the cell surface when protein synthesis was inhibited. We have previously shown that, under these conditions, the loss of IL2R␤ from the cell surface can be equated with degradation as measured in a pulse-chase labeling experiment (16). The half-life of ␤ ⌬22-27 on the cell surface was very long, more than 8 h, compared with that of IL2R␤, 3 h (Fig. 1A). This difference was not due to a defect in internalization since both proteins were internalized with similar kinetics (Fig. 1B). This result indicates that ␤ ⌬22-27 is no longer sorted to late endocytic compartments, but is recycled back to the cell surface after internalization. Thus, IL2R␤ cytosolic domain contains a unique sorting signal that targets the receptor toward degradation after endocytosis.
The Rate of Endocytosis of ␤ 27 Is Not the Limiting Factor for Its Degradation-We have shown that IL2R␤ sorting signal functions in ␤ 27 , a truncated form with only the first 27 cytosolic amino acids of IL2R␤. We decided to better define IL2R␤ sorting signal by inserting point mutations in ␤ 27 and measuring the half-life on the cell surface of these mutant proteins. When protein synthesis is inhibited, disappearance of a membrane protein from the cell surface is the net result of two events: internalization from the cell surface and sorting out of the general recycling pathway in endosomes. The limiting step(s) in these processes are not known. If the endocytic rate is limiting, a decrease or an increase in the endocytic rate should increase or decrease the half-life of the protein on the cell surface, respectively. Analysis of different mutants showed that it was not the case. For example, ␤ 27 ⌬S25 or ␤ 27 S25T (in which Ser-25 was deleted or replaced by a similar amino acid, threonine, respectively), were internalized about 2-fold slower than ␤ 27 ( Fig. 2A), but their half-life was the same as that of ␤ 27 (Fig. 2B). A further decrease in endocytosis might affect the half-time on the cell surface, since the half-life of mutants that were internalized about 3-fold slower than ␤ 27 , such as ␤ 27 F24A and ␤ 27 F24P, was increased by about 50%. We concluded that the rate of internalization was not the rate-limiting step for the net disappearance of ␤ 27 from the cell surface, but internalization might become limiting if reduced more than 2-fold. In any case, to characterize the intracellular sorting step, we will compare the half-life of mutants that are internalized with similar rates. This way, differences in half-lives between two mutants reflect a difference in sorting after endocytosis.
IL2R␤ Sorting Signal Does Not Have the Characteristics of Tyrosine-based Internalization Motifs-Most sorting signals described to date resemble internalization motifs, and indeed also promote endocytosis. Although the IL2R␤ sorting signal is not by itself an internalization motif (16), it contains two phenylalanines, which can sometimes replace tyrosine in tyrosinebased internalization signals. In the previously described cases, Tyr usually forms a better internalization signal than Phe, and other mutations at this position abolish the signal (23)(24)(25).
Of the two phenylalanines, only Phe-24 is part of a sequence that resembles tyrosine-based internalization motifs (26,27). Mutation of Phe-24 to a tyrosine resulted in an increase in endocytosis, while mutation to an alanine or a proline had the opposite effect, suggesting that Phe-24SQL could function as a weak internalization motif related to the tyrosine-based signal family ( Fig. 3A and Ref. 17). However, ␤ 27 F24Y half-life on the cell surface was not accelerated compared with that of ␤ 27 (Fig.  3B). The half-lives of ␤ 27 F24A and ␤ 27 F24P were increased by about 50%, a difference that might be entirely due to a defect in endocytosis and not in sorting, since the internalization rates for both mutants were more than 3-fold slower than that of ␤ 27 (Fig. 2).
Thus, mutations known to produce stronger or weaker inter-nalization signals, and which indeed had the expected effect on ␤ 27 internalization, did not affect ␤ 27 sorting in a corresponding way. We concluded that IL2R␤ sorting signal does not have the same requirements as tyrosine-based internalization motifs. The Sorting Signal of IL2R␤ Is Not a Strictly Defined Sequence, but Certain Amino Acids Tolerate Only Limited Replacement by Another Residue-In our previous study, we had shown that the IL2R␤ sorting signal was comprised within amino acids 18 -27 of its cytosolic domain. To better define this signal, we measured the half-life on the cell surface of ␤ 27 mutants in which amino acids within this sequence had been deleted or replaced by other residues. The internalization rates and half-lives of the mutants are reported in Fig. 3.
Residues 18 and 19 do not seem to play a critical role in the sorting signal (for further details, see below and Fig. 4). We had previously reported that Pro-20 could be replaced by a leucine, without affecting sorting after endocytosis (16). Deletion of Ser-21 or Lys-22 did not modify ␤ 27 half-life. These results suggest that amino acids 20 -23 are not essential for the IL2R␤ sorting signal function.
Mutation of Phe-23 to a tyrosine or an alanine increased the half-life of ␤ 27 to 165 and 250 min, respectively. This was not only due to a decrease in internalization compared with ␤ 27 since ␤ 27 ⌬S25, which was internalized with similar kinetics as ␤ 27 F23Y and ␤ 27 F23A, had a 125-min half-life. This result shows that Phe-23 is an important amino acid in IL2R␤ sorting signal. Its replacement by another aromatic amino acid impairs the signal only slightly, but replacement by an alanine increased the half-life of ␤ 27 2-fold.
As reported above, mutation of Phe-24 to an alanine or a proline increased ␤ 27 half-life to 180 and 200 min, respectively, a delay that could be entirely due to a defect in internalization, since ␤ 27 F24A and ␤ 27 F24P are internalized about 3-fold slower than ␤ 27 . The half-life of ␤ 27 F24Y is slightly longer than that of ␤ 27 , despite an accelerated rate of entry.
Deletion or substitution of Ser-25 by a glycine had no effect on ␤ 27 half-life. Thus, Ser-25 is not an essential amino acid in IL2R␤ sorting signal. On the contrary, deletion or mutation of Gln-26 of ␤ 27 had a dramatic effect; the half-life of both constructs increased to about 400 min, while their internalization was faster than that of ␤ 27 ⌬S25. These results show that Gln-26 is important for the IL2R␤ sorting signal to function.
Finally, we investigated the role of Leu-27 in this signal. The dramatic effect of its deletion (␤ 26 half-life was more than 12 h) suggested that this residue played an essential role in IL2R␤ sorting signal (16). Replacement of Leu-27 by an isoleucine slightly increased ␤ 27 half-life, while replacement by a proline and a threonine increased it about 2-fold. The internalization rates of ␤ 27 L27P and ␤ 27 L27T were lower than that of ␤ 27 , but similar to that of ␤ 27 ⌬S25, and therefore do not account for the half-life increase. These results suggest that Leu-27 plays an essential role in IL2R␤ sorting signal.
Analysis of IL2R␤ Sorting Signal Using Chimeric Proteins-In our previous study, we had shown that a 10-amino acid sequence in IL2R␤ was sufficient to mediate sorting of a recycling receptor, ␣ Y , toward degradation (16). The ␣ Y chimera was constructed by inserting the transferrin receptor internalization motif at the carboxyl-end of the IL2 receptor ␣ chain cytosolic domain. We have shown that this chimera was efficiently internalized and recycled back to the cell surface. In contrast, ␣ Y ␤ 18 -27 , in which amino acids 18 -27 from IL2R␤ had been added to the cytosolic domain of ␣ Y , was sorted toward degradation after internalization. We constructed a shorter chimera, ␣ Y ␤ 20 -27 . Its half-life on the cell surface was the same as that of ␣ Y ␤ 18 -27 , 200 min, showing that IL2R␤ sorting signal can be limited to 8 amino acids (Fig. 4). Studies on the ␤ 27 mutants presented here suggested that, within these 8 residues, amino acids between Phe-23 and Leu-27 were important for sorting. We prepared a chimera between ␣ Y and these 5 amino acids, ␣ Y ␤ 23-27 . While ␣ Y ␤ 23-27 was as efficiently internalized as the ␣ Y ␤ 20 -27 chimera, its half-life on the cell surface was 350 min, indicating that its sorting was impaired (Fig. 4). As shown in Fig. 4, only 4 amino acids separate 23 FFSQL 27 from the YTRF internalization signal in ␣ Y ␤ 23-27 , while in ␣ Y ␤ 20 -27 there is a 7-amino acid spacing between the two signals. A minimal spacing between these two motifs might be necessary for the sorting signal to function. To test this hypothesis, we studied the trafficking of a chimera with a three-alanine spacer between the internalization and lysosomal sorting signals, ␣ Y AAA␤ [23][24][25][26][27] . This chimera was internalized with the same kinetics as ␣ Y ␤ 20 -27 but was poorly degraded, showing that the difference in spacing between the two signals is not sufficient to account for the prolonged half-life of the ␣ Y ␤ 23-27 chimera.

DISCUSSION
The IL2R␤ chain is constitutively internalized and targeted to late endocytic compartments, where it is degraded (14,16). We had previously shown that amino acids 18 -27 of the IL2R␤ cytosolic domain are sufficient to target a recycling receptor toward degradation. Here, we show that this signal is unique in the IL2R␤, and that it is comprised within amino acids 20 -27. We further characterize this sorting signal by mutational analysis.
Several sorting signals have been described since the identification of the low density lipoprotein receptor internalization motif 12 years ago (28). Most of them fall into two categories: the tyrosine-and the dileucine-based motifs. These families have been first described and best characterized in the case of the internalization signals, but are also encountered as signals for sorting to the trans-Golgi network, to late endocytic compartments, to MHC class-II compartments, or for polarized sorting (for reviews, see Refs. 3, 4, and 26). An important characteristic of these signals is that, in addition to a specific sorting step, they also promote internalization from the plasma membrane. In some cases, several signals are found in the same receptor and can have an additive effect. Another important feature is that these signals do not consist of a fixed amino acid sequence, but are relatively flexible. Some positions in the signal, such as the tyrosine itself, or one leucine, tolerate only very limited change, while others are less restricted. Our study allows to draw a comparison between the IL2R␤ sorting signal and the other well documented trafficking signals.
The IL2R␤ Sorting Signal Functions as Other Trafficking Signals-On one hand, the IL2R␤ sorting signal shares many properties with other trafficking motifs. It is also a short amino acid sequence, which functions in different protein surroundings (here the IL2R␤ or ␣ y ), and which tolerates limited change in its composition. Leu-27 is clearly an essential residue in this signal, since even a conservative substitution, by an isoleucine, slightly increased the half-life of ␤ 27 L27I, a defect that was further enhanced by substitution by a proline or a threonine, and since deletion of Leu-27 totally impaired sorting (16). Deletion or mutation of Gln-26 increased the half-life of ␤ 27 more than 3-fold, showing that Gln-26 is important for the signal to function. Serine 25, which could be deleted or replaced by a glycine without any consequence on the half-life, is not a crucial residue. The same seems to be true for Ser-21 and Lys-22, whose deletion did not affect sorting, and for Pro-20, which could be replaced by a leucine (16). Phe-23 and Phe-24 could each be replaced by another aromatic residue, a tyrosine, with only a small increase in the half-life. Replacement by an alanine further impaired sorting, especially in ␤ 27 F23A, whose half-life was twice as long as that of ␤ 27 , suggesting that these aromatic residues play a role in the signal function. In summary, our results show the importance of the FFSQL motif for lysosomal targeting of IL2R␤, residues Phe-23, Gln-26, and Leu-27 being critical within this motif.
Another feature of IL2R␤ sorting signal common to other trafficking motifs is that many of these signals are not extremely efficient, but perform their function in an iterative manner. We measured that half of the cell surface IL2R␤ gets internalized in about 20 min, while the half-life of the receptor on the cell surface is about 180 min. Cycloheximide, used in the half-life measurement, does not slow receptor entry (data not shown). These results imply that the IL2R␤ recycles several times between the cell surface and endosomes before it gets sorted to late endocytic compartments. The same is true for the lysosomal proteins that are delivered to the plasma membrane before being sorted to lysosomes, and for the epidermal growth factor (EGF) receptor (29,30).
We had previously shown that IL2R␤ sorting signal is part of an ␣-helix (16). Again, several other trafficking motifs have also been located to structured domains of receptors, either in ␣-helix or in ␤-turns (8,(31)(32)(33)(34)(35). Another point in common is that in none of these examples is it known whether the structure is important for the signal to function. In vitro studies, on the recognition of internalization motifs placed in different protein environments, suggest that the surrounding amino acids, which will determine the overall structure, are not important for the signal recognition (27). Here we found that replacement of Ser-25 by a glycine did not affect sorting. Secondary structure predictions using AGADIR program (36), allowing to detect the ␣-helix tendencies of peptides, showed that the ␣-helix is very likely disrupted in the ␤ 27 S25G mutant. This result therefore suggests that IL2R␤ sorting signal structure is not important for its function, as long as it is accessible to the cell machinery that recognizes it. In this respect, it is interesting to note that IL2R␤ sorting signal is more efficient at the COOHterminal end of ␤ 27 (half-life ϳ 130 min) than in IL2R␤ cytosolic tail (half-life ϳ 180 min). A carboxyl-terminal location is also found in most, if not all, lysosomal proteins, and it has been suggested that this location promotes signal recognition and subsequently enzyme transport to the lysosomes (37).
The IL2R␤ Sorting Signal Is Different from Other Trafficking Motifs-On the other hand, the IL2R␤ sorting signal differs in several ways from the trafficking motifs described so far. First, although it includes a F 24 SQL sequence, reminiscent of tyrosine-based sequences, we show here that IL2R␤ sorting signal does not share the requirements of this category of signals. Mutation of Phe-24 to a tyrosine increased the endocytic rate of the mutant, without decreasing its half-life. Interestingly, the replacement of the tyrosine in the LAMP1 motif by alanine was shown to impair targeting to lysosomes (38). In our case, mutation to an alanine increased the half-life only slightly, a difference that could be entirely due to the 3-fold decrease on the endocytic rates of this mutant.
Second, the IL2R␤ sorting signal is not by itself an internalization motif (16). The observation that ␤ ⌬22-27 , in which it is deleted, is internalized with similar kinetics as IL2R␤ also shows that this sequence is not critical for the receptor internalization. However, we have shown that ␤ 27 internalization is promoted by several sequences in its cytosolic and transmembrane domains, and this sorting signal, together with other sequences, may also be involved in ␤ wt internalization (17).
Third, transmembrane proteins often contain several signals, which seem to have redundant roles in targeting (39 -41). This redundancy has also been illustrated in the sorting toward degradation of the EGF receptor and of the ␥ and ␦ chains of the T cell receptor (6,30,42). Although IL2R␤ cytosolic domain is quite large, it contains a unique signal for sorting to degradation after endocytosis, since the receptor is no longer degraded when this signal is deleted in ␤ ⌬22-27 . This last characteristic seems to be shared by some other cytokine receptors, the ␥ chain of the IL2R, the gp130 chain of the IL6 receptor (43,44).
A novel lysosomal targeting signal has been recently characterized within the cytosolic tail of P-selectin (45,46). It shares some features with the IL2R␤ sorting signal; it does not belong to the tyrosine-or dileucine-based families, it does not function as an internalization motif, and it appears to be the unique degradation signal. This sorting signal (KCPL) shares no obvious sequence homology with that of IL2R␤ (FFSQL).
What Might Recognize the IL2R␤ Sorting Signal?-Very little is known about the cellular machinery responsible for targeting membrane proteins to late endocytic compartments and lysosomes. Lysosomal proteins might be sorted to lysosomes directly from the trans-Golgi network by interaction of their sorting motif with the 1 chain of AP-1 adaptors (47,48). The EGF receptor has been the principal model to study plasma membrane receptor degradation after endocytosis. Identification of a protein that interacts with its lysosomal targeting code has been reported, but it might be specific for this receptor (49). COPI proteins, known for their role in intra-Golgi transport, might also be involved in endosome function, but the protein(s) that interact specifically with sorting signal remain to be identified (50 -52).
Where Does Signal Recognition Take Place?-The location where the signals for sorting to late endocytic compartments are recognized is also unknown. It is generally accepted that membrane proteins, even internalized by different pathways, reach a common endosome (53). In this compartment, sorting physically occurs; membrane proteins carrying trafficking signals are targeted according to this information toward different compartments, while signal-free proteins are trafficked to the recycling endosome by a default mechanism, and then back to the cell surface (54). Sorting signals might be recognized in the sorting endosome, but not necessarily. Recognition could also take place earlier in the endocytic pathway, even before internalization proceeds. This model is supported by our observation that the endocytic rate is not, to some extent, a limiting factor for degradation, in agreement with another report (55). For example, ␤ 27 , which was internalized 2-fold faster than ␤ 27 ⌬S25, was not degraded faster. If the signal recognition took place after internalization, one could expect that a faster rate of entry, which results in a higher intracellular distribution at equilibrium, would favor signal recognition and therefore degradation. However, if sorting signals were recognized at the plasma membrane, this step would be favored in slowly internalized receptor, a positive effect that could compensate for the negative effect of slow internalization.
The identification and characterization of the molecular signals that target membrane proteins toward degradation compartments is only at its beginning. Our detailed study on the IL2R␤ chain degradation signal may help to identify similar signals in other receptors. It may also open the way to the identification of the cellular components that interact with this signal.