INTRODUCTION

Resident endoplasmic reticulum (ER)¹ membrane proteins are integrated into the membrane cotranslationally or post-translationally. A typical example of the former case is microsomal cytochrome P450 (P450), which is synthesized on the rough ER (1), and inserted into and anchored to the ER membrane by its amino-terminal hydrophobic sequence (2). Thus, the hydrophobic sequence functions not only as an uncleavable signal sequence for signal recognition particle-dependent insertion into the ER membrane but also as a stop-transfer signal (3, 4). Recent studies have shown that the hydrophobic domain of P450 spans the ER membrane and that the amino terminus is luminally oriented (5, 6).

On the other hand, microsomal aldehyde dehydrogenase (msALDH) has no amino-terminal signal/anchor sequence (7) and is synthesized on free ribosomes (8). This protein has a hydrophobic sequence at its carboxyl terminus (7) (see Fig. 1) which functions as a membrane anchor (9). Recently, we showed that two hydrophilic sequences, one on both sides of the membrane binding domain at the carboxyl terminus (see Fig. 1), play important roles in ER targeting (9). Thus, msALDH is post-translationally inserted into the ER membrane, and most of the molecular portion is exposed to the cytoplasm (8, 9). However, it remains to be determined whether the carboxyl terminus of msALDH is located in the luminal side of the ER (a transmembrane structure) or is oriented toward the cytoplasm (a hairpin loop structure). Elucidation of the membrane topology of msALDH is important for understanding the mechanism of post-translational insertion of tail-anchored proteins.

Another unresolved question is how msALDH is retained in the ER. Recent studies have shown that a Lys-Asp-Glu-Leu (KDEL) sequence in soluble luminal ER proteins and a carboxyl-terminal sequence of Lys-Lys-X-X (KKXX) or Lys-X-Lys-X-X (KXKXX) in integral ER membrane proteins function as retrieval signals for the transport of these proteins back to the ER from the intermediate compartment (10, 11, 12). In contrast, P450 seems to be retained in the ER without recycling through the intermediate compartment (5, 13, 14). However, nothing is known about the mechanism of ER retention of msALDH.

In this study, we have examined the membrane topology and the ER retention of msALDH by adding potential glycosylation sites to the carboxyl terminus of this protein. The resulting chimeric proteins were glycosylated in transfected COS cells, supporting the transmembrane structure of the carboxyl terminus of msALDH. In addition, the sensitivity of the carbohydrate chain to endoglycosidases and glycosidase in addition to immunofluorescence microscopic data suggest that this protein is retained in the ER without recycling between the ER-Golgi compartments.

EXPERIMENTAL PROCEDURES

Fetal bovine serum and Dulbecco's minimal essential medium were purchased from Filtron Co. (Brooklyn, Australia) and Nissui Pharmaceutical Co., Ltd. (Tokyo, Japan), respectively. Brefeldin A and antibiotics were obtained from Epicenter Technologies (Madison, WI) and Life Technologies, Inc., respectively. [³⁵S]Methionine-cysteine was from DuPont NEN.

Goat anti-rabbit IgG conjugated with peroxidase was purchased from Tago, Inc. (Burlingame, CA). Goat anti-rabbit IgG conjugated with rhodamine and anti-mouse IgG conjugated with fluorescein were obtained from Protos Immunoresearch (San Francisco) and American Qualex Antibodies & Immunochemicals Co. (La Mirada, CA), respectively. Mouse monoclonal antibodies against human protein disulfide isomerase were from Fuji Yakuhin Kogyo Co., Ltd. (Toyama, Japan). Rabbit antibodies against rat liver msALDH were obtained and characterized as described (7).

Protein A-Sepharose 4B was from Pharmacia LKB Biotechnology Inc. (Uppsala, Sweden). Endoglycosidase H (endo H), endoglycosidase D (endo D), and N-glycosidase F were purchased from Boehringer Mannheim. Restriction enzymes and DNA-modifying enzymes were obtained from Nippon Gene (Toyama, Japan) and Takara Co., Ltd. (Kyoto, Japan). DNA sequencing kits were from U. S. Biochemical Corp. Oligonucleotide primers were synthesized with an Applied Biosystem model 381A DNA synthesizer. All other chemicals were of the highest purity commercially available.

The cDNA encoding bovine opsin in the pSP vector and a 50% suspension of Staphylococcus aureus were kindly provided by Dr. Takashi Morimoto (New York University) and Dr. Shigeru Taketani (Kansai Medical University, Osaka, Japan), respectively.

All constructions were verified by the dideoxy chain termination method (15) and restriction enzyme digestion. The full-length cDNA for rat msALDH (7) was inserted into the HindIII-EcoRV sites of the mammalian expression vector, pMIW (16), to construct pMIWALDH. Chimeric cDNAs between msALDH and bovine opsin were constructed by the gapped duplex method of oligonucleotide-directed mutagenesis (17), followed by either annealing of oligonucleotides or the polymerase chain reaction. Synthetic oligonucleotide 1 (5-TCAAGGATCAGCTGTGATTGAGCCTCTACTGAAG-3), with the mutated nucleotides underlined, was used for the generation of XbaI and AccI sites at the carboxyl terminus of msALDH. The mutated cDNA was inserted into pMIW digested with HindIII and EcoRV to construct pMIWALDHXA, which encodes msALDH with an additional Ser-Arg-Val-Asp (SRVD) sequence (amino acids 485-488) at its carboxyl terminus.

For construction of pMIWALDH/OP1, oligonucleotides 2 (5-CTGTCTAGATCCAACAAGACGGTCGACTGATGAA-3) and 3 (5-TTCATCAGTCGACCGTCTTGTTGGATCTAGACAG-3) were annealed, digested with XbaI and AccI, and then ligated into the XbaI-AccI sites of pMIWALDHXA. Similarly, oligonucleotides 4 (5-CTGTCTAGATCCAACAAGAAGGTCGACTGATGAA-3) and 5 (5-TTCATCAGTCGACCTTCTTGTTGGATCTAGACAG-3) were used to create pMIWALDH/OP1TK. pMIWALDH/OP2 was produced from oligonucleotides 6 (5-CTGTCTAGATTCTACGTGCCTTTCTCCAACAAGACGGTCGACTGA-3) and 7 (5-TCAGTCGACCGTCTTGTTGGAGAAAGGCACGTAGAATCTAGACAG-3). Polymerase chain reactions involving the following primers and templates were performed to amplify DNA fragments termed OP2TK, OP3, and OP3TK: OP2TK, oligonucleotides 8 (5-AACAAACTCAGGTACCCT-3) and 9 (5-TCAGTCGACCTCTTGTTGGA-3), with the mutated nucleotide underlined, and pMIWALDH/OP2 as a template; OP3, oligonucleotides 10 (5-CTGTCTAGAATGAACGGGACCGAGGG-3) and 11 (5-TTCACTAGTCGACCGTCTTGTTGGAGAAAG-3), and pSPOpsin as a template; and OP3TK, oligonucleotides 10 and 12 (5-TCAGTCGACCTCTTGTTGGA-3), and pMIWALDH/OP3 as a template. The XbaI-AccI fragments of polymerase chain reaction products were ligated into pMIWALDHXA digested with XbaI and AccI. The resultant plasmids were designated as pMIWALDH/OP2TK, pMIWALDH/OP3, and pMIWALDH/OP3TK, respectively.

The transfection of COS cells, subcellular fractionation, membrane extractions, and indirect immunofluorescence microscopy were performed as described previously (9). Samples were analyzed by sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) (18), followed by immunoblotting as described (9).

Forty-four hours after transfection, the cells were preincubated at 37 °C with or without brefeldin A (10 µg/ml) in Dulbecco's minimal essential medium devoid of methionine and fetal bovine serum and then pulse labeled for 30 min in the same medium containing 200 µCi/ml [³⁵S]methionine-cysteine. After labeling, the cells were chased with or without brefeldin A in complete medium containing 10% fetal bovine serum for 3 h, washed three times with cold phosphate-buffered saline, and then lysed in 10 mM Tris-HCl, pH 7.5, 0.15 M NaCl, 1% Nonidet P-40, 5 mM EDTA, 1% Trasyol, 1 mM phenylmethylsulfonyl fluoride, and 1 mM dithiothreitol for 15 min on ice. The lysates were centrifuged for 30 min in a microcentrifuge, and the resulting supernatants were preincubated with the S. aureus suspension for 30 min on a rotating device at 4 °C. Then ALDH/OP3 was immunoprecipitated from the precleared medium by the addition of rabbit anti-msALDH antibodies followed by incubation with protein A-Sepharose 4B for 2 h. Immunoprecipitates were washed four times with RIPA buffer (50 mM Tris-HCl, pH 7.5, 0.15 M NaCl, 0.3% CHAPS, and 2 mM EDTA). The proteins were eluted by heating at 100 °C for 2 min in 1% SDS and 1% -mercaptoethanol.

For endo H treatment, the eluates were adjusted to 50 mM sodium citrate, pH 5.5, 1% Nonidet P-40, and 1% -mercaptoethanol and then incubated with or without 1 milliunit of endo H overnight at 37 °C. Digestions with endo D (2 milliunits of enzyme) were carried out overnight at 37 °C in 100 mM sodium citrate, pH 6.5, 10 mM EDTA, and 2% Triton X-100. Digestions with N-glycosidase F (0.4 unit of enzyme) were carried out overnight at 37 °C in 50 mM sodium phosphate, pH 7.2, and 1% Nonidet P-40. Samples were separated by SDS-PAGE followed by fluorography.

RESULTS

To examine the membrane topology and the intracellular localization of msALDH, we fused the amino-terminal region of bovine opsin (19), which contains two N-glycosylation sites at Asn-2 and Asn-15 (Fig. 1), to the carboxyl terminus of msALDH. Three chimeras, designated as ALDH/OP1-3, have extensions of different sizes at their carboxyl termini (Fig. 1). Asn-488 of ALDH/OP1 and Asn-493 of ALDH/OP2 correspond to Asn-15 of bovine opsin. They are located at distances of 8 and 13 amino acids from the membrane anchor, respectively. ALDH/OP3 has the longest extension, and Asn-488 and Asn-501, which are present at distances of 8 and 21 amino acids from the membrane anchor, correspond to Asn-2 and Asn-15 of bovine opsin, respectively.

These mutants as well as wild-type msALDH were expressed transiently in COS cells under the control of the -actin promoter and Rous sarcoma enhancer in the pMIW expression vector (16). We first determined the intracellular localization of the expressed proteins by double indirect immunofluorescence microscopy. Cells were fixed and permeabilized 44 h after transfection, and then the expressed proteins were detected by incubation with anti-msALDH antibodies. Endogenous protein disulfide isomerase was stained with the corresponding monoclonal antibodies. When wild-type msALDH was expressed in COS cells, large granular structures containing both msALDH and protein disulfide isomerase appeared in addition to reticular structures (Fig. 2, A and B), as reported previously (9). Recently, we revealed that the structures are composed of regularly arranged crystalloid smooth ER and that overexpression of msALDH in COS cells causes the formation of crystalloid ER (20). A similar pattern of staining was observed when COS cells were transfected with cDNA for either ALDH/OP1 (not shown), ALDH/OP2 (not shown), or ALDH/OP3 (Fig. 2, C and D). These immunofluorescence microscopic data indicate that the attachment of the short extension to the carboxyl terminus of msALDH does not influence either the ER localization of the chimeras or the formation of crystalloid ER in COS cells.

We next analyzed the expressed proteins by immunoblotting. Forty-four hours after transfection, the cells were harvested and subjected to subcellular fractionation. In this case, the postnuclear supernatant was centrifuged at 88,000 × g for 80 min to separate the membrane fraction from the cytosol fraction. As expected, the three chimeric proteins as well as wild-type msALDH were recovered exclusively in the membrane fraction (Fig. 3A). In addition, these mutants in the membrane fraction were resistant to alkali extraction (Fig. 3B), showing that the three chimeras as well as msALDH are integral membrane proteins. These results, in agreement with indirect immunofluorescence microscopic data, also indicate that the carboxyl-terminal extensions do not interfere with integration of the chimeric proteins into the ER membrane. As shown in Fig. 3A (lanes 1 and 3), ALDH/OP1 exhibited a mobility almost identical to that of msALDH (54 kDa). Two bands were observed for ALDH/OP2 (lane 5): one with a mobility corresponding to that of the protein moiety (55 kDa) of the chimera and a slower migrating band with a mobility shift of about 3 kDa. The mobility of ALDH/OP3 was also shifted about 3 kDa compared with that of its protein moiety (56 kDa) (lane 7). These results suggest that Asn-493 of ALDH/OP2 and Asn-501 of ALDH/OP3 are glycosylated, whereas Asn-488 of ALDH/OP1 and Asn-488 of ALDH/OP3 are not.

To confirm the N-glycosylation of ALDH/OP2 and ALDH/OP3, we constructed an additional series of ALDH/OP chimeras (TK mutants, ALDH/OP1TK-3TK) (Fig. 4A), in which each threonine residue (Thr-490 of ALDH/OP1, Thr-495 of ALDH/OP2, or Thr-503 of ALDH/OP3) in the N-glycosylation signal (consensus sequence Asn-X-Thr), except for Thr-490 of ALDH/OP3, is replaced by a lysine residue. They were expressed in COS cells, and their intracellular localization was determined by indirect immunofluorescence microscopy. As expected, the staining patterns of all TK mutants were similar to that of msALDH, indicating the ER localization of the mutants (not shown). They were then analyzed by immunoblotting. ALDH/OP1TK exhibited a mobility identical to that of ALDH/OP1 (Fig. 4B, lanes 2 and 3), and ALDH/OP2TK showed the same mobility as that of the lower band of ALDH/OP2 (lanes 4 and 5). In contrast, ALDH/OP3TK migrated faster than ALDH/OP3 and exhibited a mobility corresponding to that of its protein moiety (lanes 6 and 7). Taken together, these biochemical results support the transmembrane topology model, in which the carboxyl terminus is translocated across the ER membrane to be glycosylated in the luminal side. In addition, these results suggest that the efficiency of N-glycosylation in ALDH/OP chimeras depends on the position of an Asn residue from the membrane anchor. Therefore, it seems that Asn-488 of ALDH/OP1 or ALDH/OP3, which is located at a distance of 8 amino acids from the membrane anchor, is too close to the membrane to be glycosylated by oligosaccharyltransferase.

We analyzed the carbohydrate structure of ALDH/OP3 to examine its intracellular localization. Forty-four hours after transfection, COS cells were pulse labeled for 30 min with [³⁵S]methionine-cysteine and chased for 3 h, and then the radiolabeled chimera was immunoprecipitated with anti-msALDH antibodies. As shown in Fig. 5 (lane 1), two major products were immunoprecipitated: one with the molecular mass of the unglycosylated ALDH/OP3 (56 kDa), and a slower migrating one with the size expected for its glycosylated form. Upon endo H treatment of the immunoprecipitated proteins, the upper species was shifted to the position of the lower one (lane 2), demonstrating that the upper product is indeed glycosylated and that the carbohydrate structure of the chimera is of a high mannose type. These results suggest that ALDH/OP3 does not reach the medial Golgi compartment, where the modification of glycoproteins to endo H-resistant forms occurs (21). To rule out the possibility that the chimera reaches the cis Golgi compartment and then is recycled back to the ER, immunoprecipitates were digested with endo D. As shown in Fig. 5 (lane 3), the glycosylated ALDH/OP3 was resistant to endo D. Since glycoproteins become sensitive to endo D after processing by -mannosidase 1A, which is supposed to be located in the cis Golgi (22), these results suggest that the chimera does not reach the cis Golgi compartment.

To confirm further that ALDH/OP3 is retained in the ER, metabolic labeling was performed in the presence of brefeldin A, an inhibitor of ER-Golgi transport (23). It is well known that glycoproteins retained in the ER are processed rapidly by redistributed cis/medial Golgi enzymes in brefeldin A-treated cells (24). Therefore, this reagent is a good indicator for proving the ER retention of ALDH/OP3. Upon treatment of transfected COS cells with brefeldin A, the chimera took on a form with a heterogeneous mobility insensitive to endo H (Fig. 5, lanes 4 and 5). The smear disappeared on digestion of the immunoprecipitates with N-glycosidase F (lane 6), an enzyme known to remove all N-linked carbohydrate chains, indicating that ALDH/OP3 was indeed processed by cis/medial Golgi enzymes redistributed to the ER as a result of the brefeldin A treatment. These biochemical results suggest that ALDH/OP3 is not transported to the cis Golgi compartment but is retained in the ER through a mechanism different from the retrieval or recycling model (10, 11, 12).

DISCUSSION

Native or artificially introduced N-glycosylation has been used to determine the membrane topology of proteins, especially those with multiple membrane-spanning domains (25). By taking advantage of this strategy, we have shown here that the carboxyl-terminal membrane binding domain of msALDH spans the phospholipid bilayer and that the carboxyl terminus is located in the luminal side of the ER. We have also demonstrated that the efficiency of glycosylation depends on the distance between the N-glycosylation site and the membrane anchor. Asn-488, located at a distance of 8 amino acids from the membrane anchor, was not glycosylated, whereas asparagine located at a distance of 13 or 21 was glycosylated. These data suggest that Asn-488 is too close to the membrane to be glycosylated by the oligosaccharyltransferase, which is composed of ribophorins I and II and a third 48-kDa protein (26). This result is in agreement with that obtained by Nilsson and von Heijne (27), who showed that the N-glycosylation site should be at least 12-14 amino acids away from the membrane anchor by in vitro transcription/translation of model proteins in the presence of rough microsomes. Since msALDH is post-translationally targeted by the ER-targeting sequences located on either side of the carboxyl-terminal membrane binding domain and inserted into the ER (9), it is noteworthy that N-glycosylation takes place not only cotranslationally during the translocation of nascent polypeptides but also post-translationally after insertion of the carboxyl-terminal hydrophobic domain of ALDH/OP chimeras into the ER membrane.

Recent studies have shown that the carboxyl termini of two tail-anchored proteins, microsomal cytochrome b₅ (28) and synaptobrevin (29), exhibit a luminal orientation. After insertion into the ER membrane in a signal recognition particle- and Sec61-independent fashion, synaptobrevin is transported to synaptic-like vesicles through the Golgi apparatus in PC12 cells (29). Additionally, we have shown that HPC-1/syntaxin 1A (30) is transported to the plasma membrane after post-translational insertion into the ER through the carboxyl-terminal hydrophobic domain and that this protein has a similar membrane topology.² Although the transmembrane topology of only the four tail-anchored proteins has been clarified at present, it appears that most tail-anchored proteins have a similar topology. Fig. 6 shows the alignment of the carboxyl-terminal sequences of six tail-anchored proteins, including the four proteins described above. Since the hydrophobic domains of these proteins differ in length (16-23 amino acids) and the composition of hydrophobic amino acids, it is less likely that these domains are important determinants for the transmembrane topology. Probably, they function as targeting or retention signals to localize to their final destinations after insertion into the ER membrane. Similarly, the carboxyl termini of these proteins exhibit no common feature. Cytochrome b₅ (31), msALDH (7), and SSO1 (32) have charged residues, whereas the other three proteins do not. On the contrary, positively charged residues are commonly located on the amino-terminal side of the hydrophobic domain with the exception in cytochrome b₅, suggesting the important role of positively charged residues in the translocation of the hydrophobic domains across the ER membrane. This agrees with the positive inside rule (33), which states that positively charged residues are preferentially found on the cytoplasmic sides of membrane proteins. However, the detail mechanism for the targeting of tail-anchored proteins to the ER and for integration of the carboxyl-terminal hydrophobic domains into the ER membrane remains to be resolved.

It is widely accepted that there are two mechanisms for the localization of ER resident proteins; one is the retrieval mechanism from the intermediate compartment, and the other is the retention mechanism, that is, the blockading of exit from the ER. The carboxyl-terminal KDEL sequence of luminal ER proteins functions as a retrieval signal from the intermediate compartment to the ER (10, 11). In addition, the di-lysine motif (KKXX or KXKXX) on the cytoplasmic side of some ER membrane proteins also serves as a retrieval signal (12). On the other hand, the ER retention signal of P450 has been defined. P450 is inserted into the ER membrane by its amino-terminal signal/anchor sequence (2, 3, 4), which also functions as an ER retention signal (14, 34). Additionally, recent studies have shown that the cytoplasmic and amino-terminal transmembrane domains of P450 contain independent redundant signals for retention in the ER (35) and that P450 is retained in the ER without recycling through the intermediate or the cis Golgi compartment (5, 13, 14). In this study, N-glycosylation of ALDH/OP chimeras provided a powerful tool for determining the mechanism for their ER localization. To the best of our knowledge, this is the first report on the ER retention of a tail-anchored protein. The carbohydrate chain of ALDH/OP3 was sensitive to endo H but insensitive to endo D, suggesting strongly that this chimera does not reach the cis Golgi. The ER retention of the chimera was checked further by treatment of transfected COS cells with brefeldin A, which effectively blocks membrane transport out of but not back to the ER (24). We have shown here that the carbohydrate chain of ALDH/OP3 is processed to an endo H-resistant form by cis/medial Golgi enzymes redistributed in the ER as a result of brefeldin A treatment. These biochemical results are consistent with morphological observations. Recently, we found that overexpression of msALDH in COS cells induced the formation of crystalloid ER (20), an aggregate of the smooth ER in a regular array. In addition, we showed that the smooth ER proliferated from the rough ER was transformed to the crystalloid ER and that the bulky cytoplasmic domain of msALDH was necessary for the formation of the crystalloid ER. By indirect immunofluorescence microscopy, we have shown here that ALDH/OP chimeras also form the crystalloid ER. Taken together, it appears that msALDH is retained in the ER through the blockading of the exit from the ER and that the smooth ER is assembled into the crystalloid ER through head-to-head association between the cytoplasmic domains of msALDH overexpressed on opposing membranes.

In summary, we have found that msALDH has a transmembrane topology and suggest that this protein is retained in the ER without recycling between the ER and the intermediate and cis Golgi compartments. Further investigation is required to elucidate the mechanisms for post-translational integration into the ER and for retention of msALDH in the ER.