Variants of the Carboxyl-terminal KDEL Sequence Direct Intracellular Retention*

Soluble proteins which reside in the lumen of the endoplasmic reticulum share a common carboxyl-terminal tetrapeptide Lys-Asp-Glu-Leu (KDEL). Addition of the tetrapeptide to a normally secreted protein is both necessary and sufficient to cause retention in the endoplasmic reticulum. In order to characterize the critical residues in the KDEL signal, cDNAs encoding proneuropeptide Y (pro-NPY) with the 4-amino acid carboxyl-terminal extension KDEL or a series of KDEL variants were expressed in the AtT-20 cell line. AtT-20 cells, a mouse anterior pituitary corticotrope cell line, synthesize, process, and secrete the pro-ACTH/endorphin precursor. Since post-translational processing in AtT-20 cells has been extensively characterized, it provides a model system in which the processing of a foreign peptide precursor (pro-NPY) and the endogenous precursor (pro-ACTH/endorphin) can be compared. Altered cDNAs encoding pro-NPY with KDEL, DKEL, RDEL, KNEL, KDQL, or KDEA at the COOH terminus were used to generate stable AtT-20 cell lines. The processing of pro-NPY to neuropeptide Y and the carboxyl-terminal peptide was studied using sodium dodecyl sulfate-polyacrylamide gel electrophoresis, tryptic peptide mapping, and radiosequencing. Addition of the tetrapeptides KDEL, DKEL, RDEL, or KNEL to the COOH terminus of the neuropeptide Y precursor, a peptide hormone normally processed and secreted from neuronal cells, caused complete intracellular retention of the unprocessed prohormone in AtT-20 cells. However, KDQL and KDEA-extended pro-NPY molecules were processed and secreted like wild-type pro-NPY when expressed in AtT-20 cells. The secretion of proNPY-derived peptides in these cell lines paralleled secretion of endogenous pro-ACTH/endorphin-derived products under both basal and stimulated conditions. These mutagenesis studies demonstrate that variants of the KDEL retention signal can direct intracellular retention.

level. Correct folding and oligomerization appear to be required for transport to the Golgi apparatus (1, 2). Improperly folded molecules or incompletely assembled protein complexes not competent for Golgi transport are retained and eventually degraded in the endoplasmic reticulum (3). Another mechanism must regulate the retention of correctly folded resident proteins (4,5). The tetrapeptide Lys-Asp-Glu-Leu (KDEL) is found at the carboxyl terminus of a set of soluble endoplasmic reticulum proteins and is essential for their retention (6, 7). Pelham et al. (8) has proposed that a "KDEL receptor" in an early Golgi compartment recognizes and selectively recycles proteins bearing the KDEL tetrapeptide. We have employed a model system to characterize the specificity of the KDEL retention signal. This system utilizes the expression of the neuropeptide Y precursor (pro-NPY)' under the control of the mouse metallothionein I promoter (9) in AtT-20 cells.
Neuropeptide Y is a 36-amino acid neuropeptide which is synthesized as a larger precursor, prepro-NPY (10). Prepro-NPY consists of a 28-amino acid signal sequence followed by 69 residues (Fig. 1). The mature hormone is located at. the amino terminus of the propeptide and is cleaved proteolytitally during biosynthesis at a pair of basic amino acids (Lys3'-Arg3"). The precursor contains neither disulfide bonds nor asparagine-linked glycosylation sites, making it an attractive model system to examine the effects of placing KDEL or variants of this sequence at the COOH terminus of pro-NPY. When AtT-20 cells are stably transfected with a plasmid expressing prepro-NPY, the biosynthesis, processing, storage, and secretion of the "foreign" precursor protein proceeds as in neuronal cells which synthesize the peptide from the endogenous gene product (11-13).

Construction of Expression
Vector-Cloning of the cDNA for human pro-NPY has been described (14 When Mt.NPYla cells (11) which produce wild-type pro-NPY were labeled for 2 h and the immunoprecipitable material was analyzed by SDS-PAGE, the medium contained a single NPY-related peptide which comigrated with iz51-labeled NPY (data not shown). After a 6-h chase, essentially all of the pro-NPY precursor had been processed to NPY within the cell (Fig. 2) sites and poly(A) addition signal as well as the gene for neomycin phosphotransferase.
Each of the plasmids was introduced into AtT-20 cells, and stable cell lines were selected in the presence of geneticin. and secretion of the endogenous pro-ACTH/endorphin precursor was unaltered in the Mt.KDEL cell line (Fig. 2). The high level of expression attained in Mt.KDEL cells allowed analysis of the biosynthetically labeled pro-NPY-KDEL peptide. Microsequencing of the [3H]tyrosine-labeled precursor following isolation by immunoprecipitation verified that attachment of the COOH-terminal KDEL sequence and prolonged retention in the cell did not alter the site of signal peptide cleavage between residues 28 and 29 of the preprohormone and demonstrated that the NPY sequence was located at the amino terminus of the propeptide.
[3H]Tyrosine (cycles 1, 20, 21, and 27) was observed only at those amino acid residues predicted from the cDNA sequence (data not shown). In addition, [3H]leucine-labeledpro-NPY-KDEL was digested with trypsin, and the peptides were separated by HPLC. A radioactive peptide with a retention time identical to an authentic sample of DEL was observed, suggesting that the pro-NPY-KDEL protein was unaltered in the AtT-20 cells (data not shown). Immunoelectron microscopy studies by Schnabel et al. (27) on proopiomelanocortin processing in AtT-20 cells indicate that endoproteolytic processing and a-amidation begin in the trans-Golgi cisterna and continue in the secretory granules. The experiments described in Fig. 2 suggest that the lack of dibasic processing and proteolysis of pro-NPY-KDEL is due to selective retention prior to the trans-Golgi. and P-endorphin antibody (10% of total, bottompanel). Immunoprecipitates were subjected to SDS-PAGE in tube gels, sliced, eluted, and scintillation-counted.
"'I-Labeled NPY eluted in slice 26 in a parallel gel. Similar results were seen in extracts from eight other incubations.
retention, we felt that pro-NPY expressed in AtT-20 cells provided a model system with which to characterize the retention signal. To accomplish this goal, a series of KDEL variants was produced. The secretory fate of pro-NPY variants with the indicated 4-amino acid carboxyl-terminal extensions is shown in Fig. 3.
The pro-NPY peptides containing COOH-terminal KDQL and KDEA extensions were processed in the same way as wild-type pro-NPY when expressed in AtT-20 cells. Pro-NPY-KDQL and pro-NPY-KDEA were efficiently processed to NPY and the corresponding COOH peptide (Fig. 4). The reduction of radiolabeled precursor protein in the cells was paralleled by the accumulation of labeled processed peptides in the medium. We were concerned that the addition of the COOH-terminal KDEL tetrapeptide might alter pro-NPY folding and cause retention in the endoplasmic reticulum. The processing and secretion of these mutants with minimal changes in the KDEL sequence imply that pro-NPY retention is not due to misfolding or inappropriate aggregation but is a consequence of the KDEL sequence. When the mutant cell lines Mt.DKEL, Mt.RDEL, and Mt.KNEL were labeled, the propeptides were efficiently retained in an unprocessed state within the cell (Fig. 4). No labeled immunoprecipitable material was detected in the medium after chase periods of 6-10 h. Fig. 4  cellular retention signals. Although the subcellular location of the retained KDEL mutants has not been directly determined, the removal of the signal peptide indicates transport into the endoplasmic reticulum, and the lack of proteolysis and dibasic processing strongly suggests that the mutants are retained in a compartment prior to the trans-Golgi. These results are in agreement with the recycling model proposed by Pelham (7). Evidence for such a model was obtained by addition of the KDEL retention signal to the lysosomal protein cathepsin D (8). The addition of mannose g-phosphate to lysosomal enzymes is one of the earliest postendoplasmic reticulum-processing events. KDEL-extended cathepsin D was retained in the endoplasmic reticulum but was still a substrate for the first of the enzymes involved in mannose 6phosphate addition.
No further enzymatic processing was observed, providing strong evidence for retrieval from an early Endoplasmic Reticulum Retention Signal Mutants 5955 postendoplasmic reticulum compartment. machinery involved in KDEL retention. The regulation of "foreign peptide" secretion was also investigated in each AtT-20 cell line. Both the endogenous pro-ACTH/endorphin and pro-NPY polypeptide should be targeted to the regulated secretory pathway, the products of which are released in response to secretagogues (28, 29). Secretagogue (5 mM 8-bromo-CAMP) treatment increased the rate of endogenous pro-ACTH/endorphin-and pro-NPY-derived peptide secretion 2-2.5fold in Mt.KDQL, Mt.KDEA, and Mt.NPYla cells (data not shown). The molar ratio of pro-NPY-derived peptides to pro-ACTH/endorphin-derived peptides in these cell lines was constant, showing parallel regulation of secretion and implying a common secretory pathway. Secretagogue treatment increased endogenous but not pro-NPY-derived peptide secretion in Mt.KDEL, Mt.KNEL, Mt.RDEL, and Mt.DKEL cells. The sorting of regulated proteins appears to take place in the trans-Golgi (30), suggesting that intracellularly retained pro-NPY molecules never reach this intracellular compartment.
The results of the amino acid substitution experiments allow some predictions regarding the nature of the retention signal. For the purpose of discussion K has been assigned as residue 1 with L as residue 4 (within the KDEL sequence). The ability of Lys to be functionally substituted by Arg and Asp at the first position, and Asp to be replaced by Gln and Lys in position 2 implies that the structural requirements for the initial two positions of the retention signal are not rigid. Interestingly, Pelham et al. (31) have shown that the COOHterminal sequence HDEL is not an efficient retention signal in COS cells so the initial two positions of the tetrapeptide are not completely elastic. In contrast, the negatively charged Glu at position 3 cannot be substituted with Gln, a residue with a similar size but no charge, suggesting that an acidic residue might be important at this position for signal recognition. Also, the inability of Ala to functionally replace Leu implies that leucine or a branched nonpolar side chain might be needed in position 4. Further mutations will be needed before a consensus is reached on the minimal retention signal. However, the results of our model studies employing pro-NPY-KDEL mutants emphasize the role of the terminal two amino acids in the functioning of the endoplasmic reticulum retention signal.