Characterization of the Signal for Rapid Internalization of the Bovine Mannose 6-Phosphate / Insulin-like Growth Factor 41 Receptor *

The signal for rapid internalization of the mannose 6-phosphate/insulin-like growth factor I1 receptor has been localized to the sequence Tyr-Lys-Tyr-Ser-LysVal in positions 24-29 of its 163-residue cytoplasmic tail. Most of the activity of this signal is mediated by the carboxyl 4 amino acids, especially Tyr” and Val2@ (Canfield, W. M., Johnson, K. F., Ye, R. D., Gregory, W. and Kornfeld, S . (1991) J. Biol. Chem. 266,66825688). In this study, we have tested the effect of a series of mutations on the internalization rate of a mutant receptor that contains a 29-amino acid cytoplasmic tail terminating with the 4-amino acid internalization sequence Tyr-Ser-Lys-Val. Replacement of Tyr” with Phe or Trp gave rise to mutant receptors that were internalized at 10% the wild-type rate, while receptors with Ala, Leu, Ile, Val, or Asn at this position were totally inactive. Val2@ could be replaced by other large hydrophobic residues (Phe, Leu, Ile, or Met) with no loss of activity, but the presence of Ala, Gly, Arg, Gln, or Tyr in this po ition inactivated the signal. Ser“ could be effectively replaced by many different amino acids, but not by Pro or Gly. However, Gly” could be tolerated if the residues at positions 28 and 29 were also changed. A change in the 2-residue spacing between Tyr2‘ and Val2@ destroyed the signal. These data show that the essential elements of this signal are an aromatic residue, especially a Tyr in the first position, separated from a large hydrophobic residue in the last position by 2 amino acids. The residues in positions 2 and 3 of the signal may have a modulating effect on its activity. The Tyr-Ser-Lys-Val signal could be moved to a more proximal region of the cytoplasmic tail with only a modest loss of activity. In addition, the signal could be effectively replaced by the putative 4-residue signals of seven other receptors and membrane proteins known to undergo rapid endocytosis, including the Tyr-Thr-Arg-Phe sequence of the transferrin receptor, a Type I1 membrane protein. These results are compatible with the 4-residue signals of this type being

interchangeable, even among Type I and Type I1 membrane proteins.
The 275-kDa cation-independent mannose 6-phosphate/ insulin-like growth factor-I1 receptor (MGP/IGF-II receptor)' is a Type I transmembrane protein that plays a key role in the biogenesis of lysosomes (1,2). This receptor functions in the shuttling of newly synthesized lysosomal hydrolases from the Golgi to prelysosomal compartments and in the retrieval of lysosomal enzymes from the extracellular medium. The receptor also serves to internalize extracellular IGF-I1 (1,2). In the endocytic pathway, the receptor with its bound ligand is initially concentrated in clathrin-coated pits at the plasma membrane and is then internalized via clathrin-coated vesicles, the first organelle involved in the vesicular transport of the ligand to its final destination, the lysosome (3). The ability of the MGP/IGF-II receptor to be efficiently concentrated in the clathrin-coated pits depends on the presence of an internalization signal in its cytoplasmic tail which is presumed to interact with the adaptin proteins of the coated pits (4,5). In previous studies we localized this signal to the sequence TyrZ4-Lys-Tyr-Ser-Lys-Val" of the 163-residue cytoplasmic tail (6,7). Alanine scanning mutagenesis identified TyrZ6 and Valz9 as the most important residues for rapid receptor internalization. TyrZ4 and LysZs also contributed to the signal, while the other amino acids were not critical.
The requirement for a tyrosine residue as a component of the internalization signal has been demonstrated for a number of receptors and membrane proteins (8)(9)(10)(11)(12)(13)(14)(15)(16), and in a few of these cases it has been shown that the tyrosine must be in the proper context relative to the surrounding amino acids to be functional (8,(10)(11)(12)17).
When we compared the YKYSKV sequence with the sequences neighboring the critical tyrosines in the cytoplasmic tails of other proteins known to undergo rapid internalization, we noted that the crucial elements of the internalization sequence were present in a number of these proteins (7). In particular, the common features were a 4-or 6-amino acid motif with an aromatic residue, usually a Tyr, in the first position and a large hydrophobic residue (Val, Phe, or Ile) in the last position. Taken together, these data were consistent with the notion that the requirement for an effective internalization signal is a general motif rather than a specific amino acid sequence. Collawn et al. (10,18) and Ktistakis e t al. (17) studying the internalization signals of the transferrin receptor and the influenza virus hemagglutinin, respectively, came to a similar conclusion.
In this study we have analyzed the effect of a series of mutations on the internalization rate of a mutant MGP/IGF-I1 receptor that contains a 29-amino acid cytoplasmic tail terminating with the 4-amino acid internalization sequence YZ6SKVz9. The goal was to define the structural requirements at each position of the 4-amino acid internalization motif and t o determine if the putative internalization motifs of other recycling receptors and membrane proteins would function when transplanted onto the MGP/IGF-I1 receptor. Our results indicate that the internalization signal must contain an aromatic residue (preferably a Tyr) in the first position and a large hydrophobic residue in the last position in order to be functional. In addition, the spacing between these two residues is critical. We also demonstrate that the YSKV sequence still mediates rapid endocytosis when it is moved to position 8-11 of the cytoplasmic tail. Furthermore, the putative 4amino acid internalization signals of other proteins are functional when transplanted onto the MGP/IGF-I1 receptor cytoplasmic tail in place of the YSKV sequence.

EXPERIMENTAL PROCEDURES
Materials-Enzymes used in molecular cloning were obtained from New England Biolabs, Bethesda Research Laboratories, or Promega. Lactoperoxidase (bovine milk), HzO2, Sephadex G-150 (120 mesh), and bovine serum albumin were obtained from Sigma. Lipofectin and G418 (neomycin sulfate) were purchased from GIBCO-BRL. Na'*'I and [36S]dATP were obtained from Amersham Corp. Other reagent grade chemicals were obtained from standard suppliers.
Cassette Mutagenesis of MGPIZGF-ZI Receptor cDNA-Oligonucleotides were synthesized with an Applied Biosystems model 380A solid phase synthesizer by the Protein Chemistry Facility of Washington University, St. Louis, MO. In most cases the 5' dimethoxy-trityl derivative was prepared and the oligonucleotide was isolated and deblocked using the oligonucleotide purification cartridge procedure as recommended by the manufacturer (Applied Biosystems). When multiple substitutions at a single position were desired, oligonucleotides containing random nucleotides at specific positions were utilized.
Plasmids encoding mutant MGP/IGF-II receptors were constructed by a cassette mutagenesis strategy (Fig. 1 termini of the -13,300-bp fragments following SfiI digestion are illustrated. The 5' termini of these fragments are all the same and contain a 3' overhang with the sequence GGA which is complementary to the CCT overhang found on the 1302-hp fragment. The 5' terminus of the 1302-bp fragment provides the stop codon. For example, the mutant oligonucleotides used to prepare construct 344 are illustrated. Two stop codons are used to decrease the possibility of read-through. 15,094 base pairs in length and lacked useful unique restriction sites, we inserted sites for the restriction enzyme SfiI. The recognition sequence of SfiI is GGCCNNNN/NGGCC, defining specificity for 8 nucleotides while allowing a half restriction site of 4 nucleotides to be placed without disrupting the coding sequence. Because the overhang sequence from different SfiI sites may be unique, multiple SfiI sites may be utilized in a single ligation reaction while retaining the advantages of directional subcloning. cleotide with the sequence GACGGAGCGCGAACGTGTCGGCCAA-A plasmid (pBC 222) was constructed by insertion of an oligonu-GTAGGCCATCGATGGCCATAGC between the PstI site at nucleotide 7053 and a Not1 site at nucleotide 7652 in the expression plasmid by the strategy previously described (7). The oligonucleotide retains the MGP/IGF-II receptor coding sequence up to amino acid 26 of the cytoplasmic domain, where the first SfiI site is located. The second SfiI site provides the stop codon and adjoins the non-coding sequence in the MGP/IGF-II receptor expression plasmid. The two SfiI sites have distinct overhang sequences to permit mutant-specific oligonucleotides to be inserted by directional cloning. The vector also contains a third SfiI site with a CCT overhang sequence. SfiI digestion of pBC 222 generates three fragments of 13356, 1302, and 19 base pairs. The two larger fragments were isolated by agarose gel electrophoresis (19) and binding and elution from SiOz (20). Mutant-specific oligonucleotides were synthesized, annealed, and 20 fmol were ligated to the 13,356-and 1302-base pair fragments (8 fmol each) to generate plasmids (pSFFV-neo-MGP/IGF-II receptor), which encode mutant MGP/IGF-II receptors. This strategy allowed the generation of receptors containing mutations beyond amino acid 26 of the cytoplasmic domain (the 163-amino acid cytoplasmic domain begins with lysine 2337). The remaining receptor sequence is wild-type, except for an alanine that has been substituted for tyrosine 24 of the cytoplasmic domain. Two additional plasmids were constructed to allow mutagenesis in other regions of the cytoplasmic or transmembrane domains by an analogous strategy. In pBC 223, an SfiI site (GGCCAACGTGGCC) is inserted at nucleotide 7065 which allows pBC 224 an SfiI site (GGCCTGTCTGGCC) is inserted at nucleotide mutagenesis beyond amino acid 22 of the cytoplasmic domain. In 6981 which allows mutagenesis beyond amino acid -7 of the transmembrane domain. The 13-kilobase SfiI fragment from each of these plasmids was ligated with the 1302-bp fragment from pBC 222 and appropriate oligonucleotides to generate plasmids encoding mutant receptors.
All sequences derived from synthetic oligonucleotides were confirmed by sequencing. Dideoxy sequencing (21) was performed with [Y3]dATP and T7 DNA polymerase utilizing the 7-deaza-dGTP protocol for double-stranded plasmid templates as recommended by the manufacturer (Pharmacia LKB Biotechnology, Inc.).
The Cczand Dd, cell lines were isolated previously (6). and secretes human p-glucuronidase (parental cell 1ine:mouse L(rec-)) was maintained in growth medium consisting of Dulbecco's modified Eagle medium supplemented with 0.9 mM sodium pyruvate, 26 mM sodium bicarbonate, 1.2 mM glutamine, 5% dialyzed fetal bovine serum, 3.2 p~ methotrexate, 100 units/ml penicillin G, and 100 pg/ml streptomycin sulfate at 37 "C in 5% COz. For collection of @-glucuronidase, the cells were grown in Waymouth MB752/1 medium containing 0.5 mg/ml human serum albumin and 1 X insulintransferrin-selenium mix (Collaborative Research). Pooled growth and collection media were precipitated by the addition of ammonium sulfate to 55% saturation, and the precipitated proteins collected by centrifugation, dissolved in a minimal volume of PBS, and dialyzed extensively at 4 "C versus PBS. The dialyzed enzyme was purified by chromatography on a MGP/IGF-II receptor affinity column as previously described for pooled lysosomal enzymes (21). Following elution with 10 mM mannose &phosphate, the 0-glucuronidase was concentrated and washed in a PM-10 Centricon (Amicon), and stored at -20 "C. 0-Glucuronidase (30 pg) was iodinated with 1 mCi of NalZ5I using soluble lactoperoxidase as described (22). The iodination reaction was gel filtered on a column (1.0 X 110 cm) of G-150 Sephadex equilibrated with 0.05 M NaP04, pH 7.6, 0.15 M NaCl, 1 mg/ml bovine serum albumin. The column was eluted at a flow rate of 0.2 ml/min, and the material inkhe first peak (VJV, = 0.40) was pooled, avoiding the leading edge of aggregated material. The specific radioactivity of the pooled p-glucuronidase was 8-16 pCi/pg assuming complete recovery of the &glucuronidase.
Short Internalization Assay-Cells were seeded into 22-mm wells of a 12-well tissue culture plate and grown in MEM-a, 10% fetal calf serum for 36-48 h until confluent. The cells were rinsed 2 times with the same media, and then 0.5 ml of MEM-a/lO% fetal calf serum containing approximately 0.1 pg (between 1 and 2 X lo6 cpm) of lZ5I-@glucuronidase was added. After a 30-min incubation on ice, the cells were rapidly washed 3 times with ice-cold PBS, 1% bovine serum albumin and 3 times with ice-cold PBS. The plate was then quickly transferred to a 37 'C water bath, where it was floated without its lid. A 0.5-ml portion of a mixture of 0.035% trypsin, 0.013% EDTA, 20 mM Man-6-P in 15 mM citrate-phosphate buffer, pH 5, was immediately added to each of two wells that were used for the measurement of the total surface binding of ligand (see below), while the other wells received an addition of 0.5 ml of 37 "C MEM-a. After an incubation period that usually ranged from 0.25 to 3 min, the medium was collected and 0.5 ml of the pH 5 trypsin-Man-6-P mixture was added to each well. At the end of an additional 3-min incubation at 37 "C, 0.8 ml of MEM-a, 10% fetal calf serum was added to each well.
The cells were then harvested in a 1.5-ml microcentrifuge tube and sedimented in an Eppendorf centrifuge 3200 for 1 min. After aspirating the supernatant with a pulled out Pasteur pipette, the radioactivity in the cell pellet was measured in a y-counter. The cells from the two wells used to determine the total surface binding were treated similarly except that the radioactivity of the harvesting media was measured directly, prior to centrifugation. The initial rate of internalization was measured by calculating the maximal slope of the uptake curve following the short lag of 15-30 s. In general, two determinations were made with two different clones of cells expressing each mutant receptor. The "internalization ratio" was calculated by taking the ratio of the initial rate of internalization of the mutant construct relative to the initial rate of internalization of construct 344 (Tyr" + Ala"; STOP3') in the same experiment.
Long Internalization Assay-The long internalization assays were performed by a slight modification of the previously described procedure (7). Cells were seeded into the wells of a 12-well tissue culture plate in triplicate and maintained for 36-48 h until confluent. To initiate uptake, the cells were rinsed 3 times with medium (MEM-a, 10% fetal calf serum), and then 1 ml of medium containing 1-2 X IO6 cpm of '251-~-glucuronidase was added to each well. The cells were incubated at 37 'C for 4 h in a 5% CO, atmosphere. The medium was then collected, and aliquots were precipitated with 10% trichloroacetic acid. The cells were rapidly washed 5 times with cold PBSbovine serum albumin and the surface-bound ligand was eluted with two washes of cold 12 mM citrate phosphate buffer, pH 5.0, 0.15 M NaCl. The cells were then solubilized in two 1-ml aliquots of 0.1 N NaOH. The radioactivity in the acid-eluted fraction, the NaOHsolubilized cells, and the trichloroacetic acid-soluble media were measured in a y-counter to determine surface-bound, internalized, and degraded 1251-~-glucuronidase, respectively. For each measurement the values obtained with wells containing L(rec-) cells were used to correct for nonspecific binding and uptake. The internalization index was calculated according to Goldstein et al. (23).

RESULTS
Assay for Rapid Receptor Internalization-In our previous studies, we assayed the rate of receptor internalization by incubating the cells with radioiodinated lysosomal enzymes for 4 h at 37 "C followed by a 1-h incubation at 4 "C . At the end of this time we measured the surface-bound, internalized and degraded lysosomal enzymes which allowed the calculation of an internalization index (23). This index is an indication of the number of times that the receptors cycle between the cell surface and t h e cell interior, and thereby serves as an indirect measurement of t h e rate of receptor internalization. T h e availability of a large amount of @-glucuronidase that binds with high affinity to the MGP/IGF-II receptor enabled us to design a more direct internalization assay (Fig. 2). The cells to be tested are first incubated with the lz51-@-glucuronidase for 30 min on ice to allow binding to the cell surface receptors. The cells are then washed free of unbound ligand and quickly brought to 37 "C to initiate internalization of the surface receptors with their bound ligands. After incubation for 0.25 to 3 min, the media is collected for determination of released ligand molecules and the internalization process is then stopped by the addition of a mixture of trypsin and Man-6-P i n pH 5.0 buffer. This cocktail rapidly releases any noninternalized ligand from the cell surface. The amount of ligand that has been internalized is then determined by harvesting the cells and measuring the radioactivity associated with the cell pellet. Fig. 2 shows t h e results of assays performed with the Cc2 and D& cell lines, which express the wild-type receptor and a receptor containing only 7 residues of the cytoplasmic tail, respectively. It is apparent that the wild-type receptor ligand complex is internalized very rapidly after a short lag period (Fig. 2 A ) . The t1/2 of the internalization process is less than 1 min. In contrast, the tail-less receptor enters the cell extremely slowly (Fig. 2B). These data also show that a significant proportion of the pre-bound @-glucuronidase is released into the media during the 3-min warming period. About onethird of t h e pre-bound ligand is released from the wild-type receptor, whereas more than 50% of the ligand is released from the tail-less receptor. This observation explains why the maximal internalization of ligand by the wild-type receptor is 50-60% of the surface-bound material.
T h e release of the surface-bound @-glucuronidase into the media is a temperature-dependent process. At 37 "C 80% of t h e pre-bound ligand is released from the tail-less receptor after 4 min, whereas the cells retain 85% of the ligand when kept on ice. At 21 "C the ligand is released at an intermediate rate.
Analysis of a Truncated Receptor Containing a T y P + Ala24 Mutation-Since the goal of this study was to analyze the 4amino acid internalization motif YZ6SKV2' in detail, we designed a mutant receptor in which Tyr" was substituted with an Ala residue and the cytoplasmic tail was truncated from 163 amino acids to 29 amino acids (construct 344: TyrZ4 + Alaz4; STOP3'). By using this mutant receptor as the reference construct for subsequent mutagenesis, we hoped to avoid secondary effects of mutations on other regions of t h e cyto- plasmic tail. When this construct was analyzed in the internalization assay, it was found to be rapidly endocytosed with an initial rate of uptake similar to that observed with the wild-type receptor (Fig. 3). The only difference between the two receptors was that the mutant receptor internalized a smaller fraction of the pre-bound ligand (45% versus 62% for the wild-type receptor). This suggests that the off-rate for the bound ligand is somewhat greater for the mutant receptor than the wild-type receptor. Nevertheless, the rapid internalization of this mutant receptor is consistent with YZ6SKV2' containing the key element of the internalization signal.
Since construct 344 also contains lysine 25, which is part of the YKYSKV 6-residue signal, we prepared a construct in which this residue was changed to alanine (construct 360 TyrZ4Lysz5 + Alaz4* 25* , STOP3'). As shown in Fig. 4, this mutant receptor was internalized at essentially the same rate as construct 344, confirming that the critical element of the internalization signal is the YSKV sequence.

Analysis of Mutant Receptors with VaP9 Substitutwns-
Previously we reported that a Valzg + Alaz9 substitution in the cytoplasmic tail of the MGP/IGF-II receptor results in a drastic reduction in the rate of receptor internalization (7). In addition, when we compared the cytoplasmic tail sequences of a number of membrane proteins known to undergo rapid endocytosis, using the essential tyrosine as the basis of the alignment, we noted six other instances where a bulky hydrophobic residue (either a valine, phenylalanine, or isoleucine) was located 3 amino acids to the carboxyl side of the tyrosine. Based on these findings we reasoned that the presence of a bulky hydrophobic residue in this position may be an essential component of the internalization signal. To test this prediction, a series of constructs were generated in which valine 29 was replaced by residues with or without this characteristic. These cDNAs were transfected into receptor-negative mouse L cells and stable clones expressing the mutant receptors were selected and analyzed for their ability to bind and internalize @-glucuronidase using the rapid internalization assay. The results of typical experiments are shown in Fig. 5, and a summary of all the determinations is given in Table I (section  A). These values are expressed as the ratio of the initial rate of internalization of the mutant receptor compared with the initial rate obtained with the YZ6SKVz9-containing receptor in the same experiment. When valine 29 was substituted with either phenylalanine, leucine, isoleucine, or mcthionine, the  resultant mutant receptors were internalized as rapidly as, or even faster than, the receptor with valine in position 29 (Fig.   5, A and B ) . The average internalization ratios were 2.22, 1.76, 1.38, and 1.32 for the phenylalanine, leucine, isoleucine, and methionine-containing mutants, respectively, compared to 1.00 for valine (Table I). By contrast, mutant receptors containing a glutamine, glycine, arginine (Fig. 5 A ) , or a tyrosine (Fig. 5 B ) at position 29 were internalized very slowly (internalization ratios of 0.05, 0.02, 0.03, and 0.12, respectively). These results demonstrate that bulky, hydrophobic residues (Phe, Leu, Ile, and Met) can substitute for Valz9 without any impairment of the signaling function of the YZ6SKVz9 sequence, whereas small (Gly) or polar (Tyr, Gln, Arg) residues cannot.
Analysis of Mutant Receptors with TY?~ Substitutions-In our previous study we showed that a TyrZ6 + Alaz6 substitution results in a striking decrease in the rate of receptor internalization (7). To further examine the amino acid re-quirement at this position of the signal, we replaced tyrosine 26 with 6 different amino acids and determined the effect on the rate of receptor internalization. The results are shown in Fig. 6 and summarized in Table I (section B). Replacement of tyrosine 26 with either a leucine, isoleucine, valine or asparagine gave rise to receptors that were internalized at the basal rate (internalization ratios of 0.00 to 0.01). Mutant receptors with either a phenylalanine or a tryptophan residue at position 26 were internalized more rapidly than the negative control, but much slower than construct 344 (internalization ratios of 0.06 and 0.13, respectively). These results indicate a specific requirement for tyrosine in this position although other aromatic residues can substitute to some extent.
The slow rate of internalization of the mutant receptor with the Tyf' + Phe" mutation was surprising since our previous experiments had shown that a mutant receptor with a Tyr24*26 -PheZ4,"j substitution in the context of a full-length cyto- plasmic tail was rapidly internalized (7). To test the possibility that a phenylalanine could substitute for tyrosine 26 provided that a phenylalanine was also located at position 24, we constructed a mutant receptor containing a FZ4KFSKV2' sequence in the truncated 29-amino acid cytoplasmic tail (construct 364). This mutant receptor had an internalization ratio of 0.07, similar to the value of 0.06 obtained with the Tyr26+ Phez6 mutant (Fig. 6B, Table I). This demonstrates that phenylalanines will function in place of tyrosines at positions 24 and 26 only in the context of a full-length cytoplasmic tail.
Analysis of Mutant Receptors with Substitutions Involving S e P and Lys2'-The consequence of replacing serine 27 with a variety of amino acids is shown in Fig. 7 and summarized in Table I (section C). Serine 27 can be substituted with isoleucine, alanine, phenylalanine, valine, threonine, or methionine with little or no effect on the rate of internalization of the mutant receptor. The result with the SeS7 Alaz7 mutant confirms our previous finding (7). However, not all residues are allowed in this position since the replacement of serine 27 with a glycine or a proline results in a drastic impairment in endocytosis (Fig. 7B). Interestingly, the effect of a glycine at position 27 varies depending on the residues at positions 28 and 29. Thus a mutant receptor with the sequence YZ6GVF2' is internalized relatively well, with an average internalization ratio of 0.40 (Fig. 7B). These data demonstrate that the residue at position 27 can modulate or even abolish the functioning of the internalization signal. We also tested the effect of reversing the SerZ7Lysz8 sequence in the Y26SKV29 signal. As shown in Fig. 7B, the resulting mutant receptor (construct 345) containing a YZ6KSVz9 sequence was internalized about one-third as well as the control receptor (internalization ratio of 0.35).
Importance of the Spacing between Residues 26 and 29 of the Cytophmic Tail-To determine if the spacing between tyrosine 26 and valine 29 is critical for the functioning of the internalization signal, four mutant receptors were constructed in which the spacing between these two residues was altered. In construct 346, serine 27 was removed to generate the sequence YZ6Kz7VB, whereas in constructs 347-349 1, 2, or 3 alanines were introduced between serine 27 and the lysine initially present at position 28. As summarized in Table I (section D), none of the mutant receptors were internalized over the basal rate. Clearly the correct spacing between tyrosine 26 and valine 29 is critical for the functioning of the internalization signal.
The YSKV Sequence Remins Functional When Moved to a Different Location in the Cytophmic Tail-Construct 420 was prepared to determine whether the YSKV internalization signal would still function when placed closer to the transmembrane segment and adjacent to a different set of amino acids. This construct contains an 11-amino acid cytoplasmic tail with the sequence K'KERRAKYSKV". The first 5 amino acids represent the initial residues of the MGP/IGF-II receptor cytoplasmic tail. Surprisingly we found that the clones expressing this mutant receptor bound the "61-j3-gl~~~ronidase very poorly at 4 "C, making it impossible to perform the standard rapid internalization assay.' The basis for the poor binding at 4 "C is not clear. However, the mutant receptor did bind the ligand quite well at 37 "C, and therefore it was possible to evaluate its rate of endocytosis using the long internalization assay described previously (7). Table I1 shows that construct 420 was internalized about 40-60% as well as the wild-type receptor with a full-length cytoplasmic tail and 4-5 times better than a receptor (D&) with a 7-amino acid cytoplasmic tail. These data demonstrate that the YSKV sequence remains functional in endocytosis, albeit at a somewhat reduced rate, when transplanted to a different region of the cytoplasmic tail.

Transplantation of Putative Internalization Signals of Other
Proteins onto the MGpIIGF-II Receptor-As mentioned previously, the alignment of the amino acid sequences of the cytoplasmic tails of proteins known to undergo rapid internalization revealed six instances (in addition to the M6P/ IGF-I1 receptor) of putative tetrapeptide internalization signals (Table VI of Canfield et ul.) (7). To determine if these putative internalization signals would function in the context of the MGP/IGF-II receptor, we constructed six hybrid receptor mutants in which the putative internalization sequences were substituted for the YZ6SKVm sequence in the truncated 29-amino acid MGP/IGF-II receptor cytoplasmic tail. The transplanted sequences were YQTI and YEQF from the lysosomal membrane glycoproteins LAMP-1 and lgp 110, respectively (13,24), YRGV from the cation-dependent M6P receptor (9), YRHV from human acidphosphatase (16), YSAF from the rabbit poly (Ig) receptor (15), and YTRF from the human transferrin receptor (10). Each of these hybrid receptors was efficiently internalized, demonstrating that the internalization signal of one protein can function when transplanted onto another protein (Fig. 8, A and B ) . In fact, two of the hybrid receptors were internalized somewhat faster than the control receptor with the YZ6SKVz9 sequence. The average internalization ratios for the receptors containing the YSAF and YTRF sequences were 1.54 and 1.66, respectively (Table I,

DISCUSSION
In our previous study we presented evidence that the signal for the rapid internalization of the MGP/IGF-I1 receptor is contained within the YZ4KYSKVz9 sequence of the 163-amino acid cytoplasmic tail (7). The most important elements of this signal were shown to be TyP' and Val" with LysZ8 and TyP4 having lesser roles. Since the YZ6SKVz9 sequence was almost as potent as the YKYSKV sequence in promoting rapid endocytosis, this internalization signal could be considered as either a 6-or a 4-amino acid motif. When we compared the YKYSKV sequence with the sequences neighboring the critical tyrosines in the cytoplasmic tails of eight other proteins known to undergo rapid internalization, we noted six instances where the tyrosine was separated from a valine, phenylalanine, or isoleucine by 2 intervening amino acids. Based on this information, we suggested that a common internalization motif may be a sequence of 4 amino acids with an aromatic residue, especially a tyrosine, in the first position separated from a bulky hydrophobic residue in the fourth position.
The current experiments provide further evidence to support this proposal. Using a construct with a 29-amino acid cytoplasmic tail that terminates with the sequence AKYSKV, we have systematically tested the effect of numerous amino acid substitutions at the first, second, and fourth position of the YSKV internalization sequence. Our results clearly show a requirement for an aromatic residue in the first position of the signal, particularly a tyrosine. Tryptophan and phenylalanine were about 10% as effective as tyrosine in promoting rapid endocytosis while valine, isoleucine, leucine, and asparagine were totally inactive. The finding that a w6 + Phe" substitution resulted in a marked decrease in the rate of receptor internalization was unexpected since a TyrZ4sz6 4 PheZ4*'' replacement in the context of a full-length cytoplasmic tail did not impair receptor endocytosis (7). The addition of a second phenylalanine at position 24 in the mutant receptor with a truncated cytoplasmic tail (construct 364) did not enhance the ability of the receptor to be rapidly endocytosed. The reason for this discrepancy is unclear, but we presume that the presence of the full-length cytoplasmic tail allows the phenylalanines to function better than they do in the context of a truncated cytoplasmic tail. One possibility is that elements in the full-length cytoplasmic tail enable the FKFSKV sequence to achieve a structure that is required for interaction with the adaptin proteins of the coated pits. Alternatively, there could be additional elements in the fulllength cytoplasmic tail that interact directly with the adaptin proteins and thereby complement the weak signal generated by the FKFSKV sequence. In this regard, it is of interest that phenylalanine will substitute for tyrosine in some internalization signals (12,25,26), but not in others (16,24). The analysis of the substitutions at the fourth position of the signal shows that valine 29 can be replaced with leucine, isoleucine, methionine, or phenylalanine with no impairment in receptor internalization, whereas mutant receptors with glycine, arginine, glutamine, or tyrosine in this position are endocytosed extremely slowly. These results are consistent with a requirement for a large hydrophobic residue in this position rather than a particular amino acid. Collawn et al. (18) have made a similar finding with the transferrin receptor. They reported that the carhoxyl-terminal phenylalanine in the tetrapeptide internalization sequence YTRF could be replaced by methionine, isoleucine, or tryptophan without impairment in the rate of receptor internalization. However, when the phenylalanine was replaced with an alanine or a glycine, rapid endocytosis was lost.
One function of the residues at positions 2 and 3 of the internalization signal is to provide the correct spacing between the tyrosine in position one and the hydrophobic residue in position 4. This is most clearly shown by construct 346, which has a deletion of serine 25 to produce a 3-amino acid signal, and construct 347, which contains an alanine inserted between the serine and lysine to produce a 5amino acid signal. Neither of these mutant receptors was endocytosed above the base-line value, showing an absolute requirement for the correct spacing of the 4-amino acid signal.
While many amino acid substitutions were well tolerated in position 2 of the YSKV signal, proline and glycine were not. In addition, it is apparent that the nature of the residues in positions 2 and 3 may modulate the activity of the signal. For example, reversing the serine and lysine residues to produce a signal with a YKSV sequence results in a 65% decrease in the rate of internalization.
An even more striking finding was made with mutant receptors containing glycine substitutions in position 2. Construct 352 with a YGKV sequence was internalized at the baseline value, whereas construct 422 with the sequence YGVF was internalized reasonably well (internalization ratio of 0.35). Interestingly the YGVF sequence is present in the cytoplasmic tail of the P-selectin molecule which is known to be translocated to the cell surface of platelets and endothelial cells upon activation and then cleared from the plasma membrane over a lo-20-min period (27). In these examples, the consequence of having a glycine in position 2 of the signal is dependent, in part, on the nature of the residues in positions 3 and 4. Since the same residue at a particular position of the signal may have a different effect on the activity depending on the nature of the other residues, it may be difficult to predict whether a putative internalization signal will be functional or not. However, this will only be a problem when certain amino acids, such as glycine and proline, are present in the signal.
The finding that the YSKV sequence can be moved to position 8-11 of the cytoplasmic tail and still mediate relatively rapid endocytosis shows that this 4-amino acid motif can function when placed in a different context. This result is consistent with our previous alanine scanning data which showed that the 6 residues preceding the YSKV signal in the cytoplasmic tail were not essential for rapid endocytosis (7).
One prediction from these studies is that the putative of these studies involved transplanting putative internalization signals onto receptors that normally undergo constitutive recycling. Thus it is possible that these receptors have other properties that allow them to be rapidly endocytosed when they contain the internalization signals. For this reason it will be of interest to determine if these signals function in the same way when transplanted onto the cytoplasmic tails of membrane proteins that do not normally undergo rapid endocytosis. Several recent studies have suggested that the internalization sequences have a propensity to form tight turn structures. Collawn et al. (10,18) have presented indirect evidence that the 6-residue internalization signals of the MGP/IGF-II and the LDL receptors as well as the 4-residue signal of the transferrin receptor adopt tight turn structures based on their analysis of similar sequences in known protein crystal structures. In the case of the MGP/IGF-II receptor, they identified four structural analogs of the YKYSKV sequence. The four carboxyl-terminal residues of these analogs were noted to be in tight turns and the four functionally important side chains at positions -2, 1, 3, and 4 were simultaneously accessible from one side of the turn. (The numbering refers to the positions relative to the turn region.) In addition, the two aromatic rings in positions -2 and 1 were oriented in a roughly parallel manner, providing an indication of how one of the aromatic rings could compensate for the loss of the other. Collawn et al. (18) suggested that the similarity in threedimensional placement of the critical residues in the 6-and 4-residue signals would allow both types of signals to interact with the same recognition structure in coated pits. Ktistakis et al. (17) have postulated that the tyrosine recognition signals may form a small surface loop structure, but this structure differs from the one proposed by Collawn et al. in terms of the positioning of the tyrosine in the loop. Bansal and Gierasch (29) have obtained more direct evidence that the NPVY sequence of the LDL receptor forms a @-turn structure by performing NMR analysis of nonapeptides containing this sequence. They showed that peptides containing the NPVY sequence assume a reverse-turn conformation with the Asn in position 1 and the Tyr in position 4 of the turn. Substitution of either the Asn, Pro, or Tyr with residues known to be inactive in endocytosis resulted in a disruption of the turn conformation. Eberle et al. (30) used a similar approach to obtain evidence that the PPGY sequence of the acid phosphatase cytoplasmic tail forms a type 1 @ turn with the Tyr in position 4 of the turn.
All of these studies indicate that the critical tyrosine of the internalization signal is presented to the adaptin proteins in the context of a tight turn motif. However, the LDL receptor internalization motif differs from the signal of the MGP/IGF-I1 receptor and the transferrin receptor in that its essential tyrosine is in the carboxyl-terminal position of the signal, whereas the critical tyrosine of the other two receptors is in the amino-terminal position of the signal. In fact, our data show that a mutant receptor with a Valz9 -* Tyrz9 substitution (construct 419) is internalized very poorly.
The finding that the P6PGY8 sequence of the acid phosphatase cytoplasmic tail achieves a tight turn conformation must be reconciled with the results of our transplantation experiments showing that the Y8RHV" sequence of this cytoplasmic domain functions to mediate rapid endocytosis in the context of the MGP/IGF-II receptor (construct 375). One possibility is that the internalization signal of acid phosphatase comprises a more extended structure, involving residues 5-11, with the RHV sequence providing additional interactions with the adaptin proteins over those achieved by the PPGY turn structure. If this is the case, then the YRHV sequence may function well enough in the content of the MGP/IGF-II receptor to allow rapid endocytosis. The possibility that some internalization signals may have more extended structures has been suggested by several investigators (11, 12, 17, 30). Ktistakis et al. (17) have proposed a "tyrosine internalization signal" that extends from -6 to +2 relative to the essential tyrosine with polar or basic residues preferred at certain positions on both sides of the tyrosine. However, there are several exceptions to these rules, including the MGP/IGF-II receptor.
In summary, we have presented evidence that a tetrapeptide sequence with an aromatic residue, especially a tyrosine, in the first position and a large hydrophobic residue in the fourth position, functions as an effective signal for the rapid internalization of the MGP/IGF-II receptor and possibly many other receptors and membrane proteins as well. The relationship of this 4-residue signal to other internalization motifs needs further exploration, particularly at the structural level.