An Internal Signal Sequence Mediates the Targeting and Retention of the Human UDP-Glucuronosyltransferase 1A6 to the Endoplasmic Reticulum*

The human UDP-glucuronosyltransferase isoform UGT1A6 is predicted to be a type I transmembrane protein anchored in the endoplasmic reticulum by a single C-terminal transmembrane domain, followed by a short cytoplasmic tail. This topology is thought to be established through the sequential action of a cleavable N-terminal signal peptide and of a C-terminal stop transfer/anchor sequence. We found that the deletion of the signal peptide did not prevent membrane targeting and insertion of this protein expressed in an in vitro transcription/translation system or in yeast Pichia pastoris. Interestingly, the same results were obtained when the protein was depleted of both the signal peptide and the C-terminal transmembrane domain/cytoplasmic tail sequences, suggesting the presence of an internal topogenic element able to translocate and retain UGT1A6 in the endoplasmic reticulum membrane in vitro and in yeast cells. To identify such a sequence, the insertion of several N-terminal deletion mutants of UGT1A6 into microsomal membranes was investigated in vitro. The data clearly showed that the deletion of the N-terminal end did not affect endoplasmic reticulum targeting and retention until residues 140–240 were deleted. The signal-like activity of the 140–240 region was demonstrated by the ability of this segment to confer endoplasmic reticulum residency to the cytosolic green fluorescent protein expressed in mammalian cells. Finally, we show that this novel topogenic sequence can posttranslationally mediate the translocation of UGT1A6. This study provides the first evidence that the membrane assembly of the human UGT1A6 involves an internal signal retention sequence.

Targeting and insertion of integral membrane proteins involve characteristic topogenic sequences that act to initiate (signal peptide sequence) and terminate (stop transfer sequence) translocation. In mammalian cells, the N-terminal signal peptide generally targets the protein to the endoplasmic reticulum (ER) 1 via a cotranslational pathway by binding to the signal recognition particle, which then interacts with its receptor and releases the signal sequence (1). A second signal recognition event within the ER involving the Sec61 complex, the main constituent of a protein conducting channel, is then required for the subsequent translocation of the nascent polypeptide across the lipid bilayer (2). The elongating polypeptide is then extruded through the membrane until a stop transfer sequence terminates translocation and integrates laterally into the membrane. A number of reports describing overlapping properties of signal sequences, signal anchors, and stop transfer sequences question the nature and specificity of these topogenic sequences. The stop transfer sequence of the IgM heavy chain (3) and the seventh transmembrane segment of band 3 of the erythrocyte anion exchanger (4) are able to initiate the translocation process. Signal anchors can also act as signal sequences as shown for cytochrome b 5 (5) and for the yeast UBC6 transmembrane protein (6).
Single spanning membrane proteins that present a cleavable signal peptide sequence usually have a type I topology with the N-terminal part of the transmembrane domain exposed to the exterior of the membrane and the C-terminal portion exposed to the cytoplasmic side (N-exo/C-cyto). Examples of such bitopic membrane proteins include the low density lipoprotein receptor (7), influenza virus hemagglutinin (8), and glycophorin A (9). In this study, we have investigated the topogenic determinants that drive the membrane assembly of the human UDPglucuronosyltransferase (UGT) isoform UGT1A6, an integral ER glycoprotein with a type I orientation. The UGT superfamily consists of numerous isoforms catalyzing the addition of a glycosidic residue, glucuronic acid, on to a wide variety of hydrophobic acceptor substrates, i.e. small endogenous and exogenous molecules, as well as complex glycolipids and glycoproteins (10,11). UGT1A6 is a major human UGT isoform playing an essential role in the elimination of planar phenols, carcinogenic environmental compounds, and therapeutic agents such as naftazone and paracetamol (12)(13)(14).
Sequence alignment of more than 50 mammalian UGT cDNAs indicated the presence of a hydrophobic N-terminal signal peptide sequence that is cleaved during membrane insertion (15). Examination of the sequences also suggested the presence of a single TMD of 17 amino acid residues near the C-terminal end, followed by about 26 cytoplasmic residues containing a di-lysine motif that could function as a ER retrieval/ retention signal (16,17). Additional analysis of the membrane topology of native UGTs in rat and human liver microsomes using proteases and antibodies (18 -20) or photoaffinity probes (21) as well as studies performed on recombinant rat enzymes (22,23) suggested the picture of a transferase slightly exposed to the cytoplasmic surface of the ER, joined by a single transmembrane region to the bulk of the protein located inside the ER. However, the membrane organization and the topological determinants of UGTs remain hypothetical since no structural determination by x-ray crystallography or nuclear magnetic resonance is yet available. This study represents the beginning of a systematic effort to understand the structural implications of the membrane-associated regions of UGTs. We previously provided evidence that UGT1A6 with its signal peptide sequence deleted remained integrally associated with the inner membrane of Escherichia coli, when heterologously expressed in this host cell (24).
To further address the mechanism by which UGT1A6 is targeted and inserted into the ER membrane, we generated a series of truncations deleting individually or simultaneously the N-and C-terminal ends of the protein. Using an efficient expression system in the yeast Pichia pastoris together with in vitro expression experiments in a reticulocyte lysate, we demonstrate the existence of a previously unidentified internal topogenic sequence mediating the translocation and retention of polypeptides across the ER membrane. Moreover, we provide evidence for a posttranslational mechanism of translocation of the full-length and truncated UGT1A6 polypeptides.

EXPERIMENTAL PROCEDURES
Materials-[ 35 S]Methionine was from Amersham Pharmacia Biotech (Les Ulis, France). Bacterial and yeast culture media were from Difco (Detroit, MI). Mammalian (Dulbecco's modified Eagle's medium) cell culture medium was from Life Technologies, Inc. (Cergy-Pontoise, France). Protein assay reagent was obtained from Bio-Rad (Yvry sur Seine, France). Restriction enzymes, T4 DNA ligase, RNAsin TM , pGEM-3Z, rabbit reticulocyte lysate, dog pancreas microsomes, and competent E. coli JM109 cells were purchased from Promega (Charbonnières, France). Prepro-␣-factor mRNA used as control in in vitro experiments was also provided by Promega. The plasmid pEGFP-N1 for expression of protein fused to a red shift variant of the green fluorescent protein of jellyfish was from CLONTECH (Palo Alto, CA). The P. pastoris yeast expression system was from Invitrogen (Groningen, The Netherlands). Vent DNA polymerase and endoglycosidase H were provided by New England Biolabs (Hitchin, United Kingdom). Other reagents including cycloheximide, proteinase K, methanol, paraformaldehyde, fluorescein isothiocyanate (FITC)-conjugated anti-sheep immunoglobulins, rhodamine-conjugated anti-mouse immunoglobulins, and anti-goat alkaline phosphatase-conjugated immunoglobulins were purchased from Sigma (Saint Quentin Fallavier, France). The glycosylation acceptor tripeptides N-benzoyl-asparagine-asparagine-threonine-N-methylamide and N-benzoyl-asparagine-alanine-serine-N-methylamide were synthesized by Dr. G. Bloomberg (Recognition Center Peptide Synthesis Facility, Bristol, United Kingdom) and used to inhibit N-linked glycosylation, as described previously (25).
Plasmid Constructions-Cloning and sequencing of full-length UGT1A6 cDNA (GenBank accession no. M84130) have been reported elsewhere (26). For in vitro expression of UGT1A6, the corresponding cDNA was isolated from the eukaryotic expression vector pcDNA1-UGT1A6 (27) and subcloned into the EcoRI-XbaI sites of pGEM-3Z. UGT1A6 lacking the signal peptide sequence was obtained by PCR amplification using a sense primer comprising an EcoRI site, a Kozak sequence, and nucleotides 79 -102 corresponding to UGT1A6 coding region together with an antisense primer comprising a XbaI site, a stop codon, and nucleotides 1593-1576. The PCR fragment was subcloned into the EcoRI-XbaI sites of pGEM-3Z to generate UGT1A6⌬SP. Constructs encoding polypeptides lacking the 26, 43, 68, 140, or 240 N-terminal amino acid residues as well as 43 C-terminal residues (corresponding to the TMD/CT sequence) were generated by PCR amplification using sense primers containing an EcoRI site, a Kozak sequence, and nucleotides 79 -102, 130 -153, 205-228, 421-444, 721-744 corresponding to UGT1A6 coding region, respectively, and an antisense primer comprising a XbaI site, a stop codon, and nucleotides 1515-1498. The PCR products were individually subcloned into the EcoRI-XbaI sites of pGEM-3Z to generate UGT1A6⌬SP/TMD/CT, UGT1A6⌬-N43/TMD/CT, UGT1A6⌬-N68/TMD/CT, UGT1A6⌬-N140/TMD/CT, and UGT1A6⌬-N240/TMD/CT, respectively. The recombinant plasmids were then used for in vitro transcription-translation in a rabbit reticulocyte lysate using the T7 RNA polymerase.
To generate the UGT141-240/EGFP expression plasmid encoding a chimeric protein between the 141-240 region of UGT1A6 and EGFP, a sense primer comprising a HindIII site, a Kozak sequence, and nucleotides 421-441 corresponding to UGT1A6 coding region was used together with an antisense primer comprising a BamHI site and nucleotides 720 -702. The resulting amplified fragment was subcloned into the HindIII-BamHI sites of the pEGFP-N1 vector in frame with the EGFP coding sequence. The recombinant vector was used for transient expression of the chimeric UGT141-240/EGFP protein in HeLa 229 cells (European Collection of Cell Cultures, ECACC).
For expression in P. pastoris of wild-type and truncated UGT1A6 mutants, the sequences coding for UGT1A6, UGT1A6⌬SP, and UGT1A6⌬SP/TMD/CT were isolated from the corresponding pGEM-3Z constructs and subcloned into the EcoRI-XbaI sites of pPICZB. The corresponding expression vectors were named pPICZ-UGT1A6, pPICZ-UGT1A6⌬SP, and pPICZ-UGT1A6⌬SP/TMD/CT. All mutant clones were screened for Taq-introduced errors by dideoxy sequencing (28).
In Vitro Protein Synthesis-The in vitro transcription-translation procedures were performed with T7 RNA polymerase in the presence of rabbit reticulocyte lysate and [ 35 S]methionine, at 30°C for 60 min using the TNT TM coupled reticulocyte lysate kit from Promega. For membrane translocation tests, dog pancreas microsomal membranes (Promega) were added to the reaction mixture according to the suggestions of the supplier. N-Glycosylation was prevented by acceptor tripeptides (0.2 mM) added to the incubation mixture before the microsomal membranes. For alkaline extraction, the incubation mixture containing the microsomes was diluted two times in buffer A (50 mM KOAc, 200 mM Na 2 CO 3 , 20 mM triethanolamine, 1 mM Mg(OAc) 2 , pH 11.5) and kept in ice for 20 min to fracture microsomes and remove non-integral proteins. The mixture was then layered over a 250-l sucrose cushion (buffer B: 0.25 M sucrose, 50 mM triethanolamine, 140 mM KOAc, 2.5 mM Mg(OAc) 2 ) containing 100 mM Na 2 CO 3 , pH 11.5, and the membranes were pelleted at 100,000 ϫ g in a Sorvall RC M120 micro-ultracentrifuge for 45 min. The pellet was washed with 25 l of buffer B and suspended in sample loading buffer. Supernatant proteins were precipitated with trichloroacetic acid (10% (v/v)), and the samples were electrophoresed on 10% (w/v) SDS-polyacrylamide gels (29). Protease sensitivity and topology of the expressed polypeptide were analyzed using proteinase K as described (30). The transcription-translation products were treated with 0.2 mg/ml proteinase K in the presence or absence of 0.1% (v/v) Triton X-100 for 30 min on ice, and the reaction was stopped by addition of 1 mM phenylmethylsulfonyl fluoride. After 10 min, the samples were heated at 100°C in Laemmli sample buffer prior to SDS-PAGE analysis. The following calibrated prestained SDS-PAGE standards (Bio-Rad) were used: phosphorylase B, 107 kDa; bovine serum albumin, 76 kDa; ovalbumin, 52 kDa; carbonic anhydrase, 36.8 kDa; soybean trypsin inhibitor, 27.2 kDa; and lysozyme, 19 kDa. After electrophoresis, gels were dried and exposed directly to Kodak Biomax films for visualization.
Posttranslational Membrane Processing-Following linearization by XbaI, recombinant pGEM-3Z plasmids encoding the full-length and truncated UGT polypeptides served as templates for in vitro transcription by T7 RNA polymerase using mCAP RNA capping kit (Stratagene, La Jolla, CA). The RNAs produced were translated in vitro using the Flexi TM rabbit reticulocyte lysate expression system (Promega). After translation for 30 min, 2 mM cycloheximide was added for 10 min to inhibit further elongation. Translocation reactions were then initiated by addition of dog pancreas microsomes and the incubation was continued for 30 min at 30°C. The same procedure was carried out using the yeast prepro-␣-factor mRNA as a control.
Subcellular Fractionation and Protein Analysis of Recombinant Yeast Cells-Cells were harvested after 48 h of induction, washed once, and suspended in cold breaking buffer (50 mM sodium phosphate, pH 7.4, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and 5% glycerol). The cells were then broken by vortexing with glass beads. The resulting homogenate was centrifuged at 5,000 ϫ g for 15 min, and the supernatant was centrifuged at 12,000 ϫ g for 20 min. Membranes were then pelleted from the supernatant for 1 h at 100,000 ϫ g at 4°C. The pellet fraction was resuspended by Dounce homogenization in sucrose-Hepes buffer (0.25 M sucrose, 5 mM Hepes, pH 7.4). Alkaline extraction was carried out as described above. For high salt extraction, membranes were incubated in buffer (50 mM triethanolamine, 140 mM KOAc, 2.5 mM Mg(OAc) 2 , 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 M NaCl, pH 7.4) at 4°C for 20 min, then layered over a 250-l sucrose cushion (0.25 M sucrose, 50 mM triethanolamine, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 M NaCl, pH 7.4) and fractionated by centrifugation at 100,000 ϫ g to yield pellet and soluble fractions. For endoglycosidase H digestion, microsomal membranes were denatured in 0.5% SDS, 1% ␤-mercaptoethanol at 100°C for 10 min and then incubated with endoglycosidase H in 50 mM sodium citrate, pH 5.5, buffer containing Complete Mini TM protease inhibitors (Roche Molecular Biochemicals, Meylan, France) at 37°C for 1 h, according to the recommendations of the supplier (New England Biolabs).
Protein concentration was evaluated by the method of Bradford (31). SDS-PAGE (29) and immunoblot analysis were performed using anti-UGT1A6 antibodies and alkaline phosphatase-conjugated secondary antibodies as described previously (27).
Immunofluorescence and Fluorescence Microscopy-HeLa cells were replated in 30-mm diameter wells on coverslips 1 day before transfection with the pUGT141-240/EGFP plasmid using Exgen TM as transfecting reagent (Euromedex, Souffelweyersheim, France) and fixed after 24 h of culture with 3% paraformaldehyde (w/v) in PBS. Cells were permeabilized with 0.1% (w/v) Triton X-100/PBS, followed by blocking with 0.2% gelatin/PBS (w/v). Permeabilized cells were then incubated for 1 h with monoclonal anti-calnexin antibodies (Affinity Bioreagents, Golden, CO), washed in PBS, and incubated for 20 min with rhodamineconjugated secondary antibodies (32). Finally, cells were washed in PBS and mounted on microscope slides using mounting medium (Sigma).
Microscopic observations were performed using a Zeiss Axiophot microscope with a 100ϫ objective. Kodak Ektachrome 400 film was used for photography.

The N-terminal Signal Peptide Sequence Is Not Required for
Membrane Targeting and Integration of UGT1A6 -To study the process of UGT1A6 targeting and insertion into the microsomal membranes, we employed an in vitro transcriptiontranslation expression system using in the first instance two constructs. One of these directed the synthesis of the fulllength UGT1A6 precursor, and the other was an engineered mutant, UGT1A6⌬SP, lacking 26 N-terminal amino acids (the entire signal peptide sequence). A schematic representation of the constructs appears in Fig. 1A. The insertion of the polypeptides in dog pancreas microsomes was first characterized by their glycosylation pattern and resistance to alkaline treatment. The in vitro expression of UGT1A6 cDNA yielded a polypeptide of about 54 kDa (Fig. 1B, lane a) corresponding to the expected apparent molecular mass of the precursor, as evaluated by comparison with calibrated molecular masses of prestained SDS-PAGE standards. In the presence of microsomal membranes, polypeptide bands exhibiting a slower mobility on gel electrophoresis (55-56 kDa, Fig. 1B, lane c) were observed. This increase of apparent molecular mass was consistent with the cleavage of the signal peptide followed by N-asparagine-linked glycosylation, previously described for this UGT isoform (26). In the presence of acceptor tripeptides that are competitive inhibitors of asparagine-linked glycosylation, a single polypeptide band of 52 kDa was generated, which corresponds to the mature unglycosylated form of UGT1A6 resulting from the cleavage of the signal peptide sequence (Fig.  1B, compare lane e to lanes a and c). In the case of UGT1A6⌬SP, a polypeptide of about 52 kDa was produced in the absence of microsomal membranes (Fig. 1B, lane b), which was, as expected, about 2 kDa shorter than the product of the wild-type cDNA expressed in the same conditions. Interestingly, addition of microsomes to the transcription-translation incubation mixture resulted in the appearance of higher mo-lecular mass species (Fig. 1B, lane d) that was prevented by addition of glycosylation inhibitory tripeptides (Fig. 1B, lane f).
These results indicate that the deletion mutant, like the precursor form, was translocated and N-glycosylated in the lumen of the ER. A high concentration of a strong base is classically used to test the integration of proteins into the membrane. Such treatment extracts only peripheral membrane and secretory proteins but not integral membrane-bound proteins. Both the wild-type UGT1A6 and the form lacking the signal peptide sequence were found to be tightly integrated into the lipid bilayer of microsomal membranes with the majority of the polypeptides recovered in the pellet fraction (Fig. 1B, lanes i  and j). Fig. 1B (lanes g and h) shows that, indeed, only small amounts of protein could be detected in the supernatant. As a positive control, the yeast prepro-␣-factor expressed in the presence of microsomes was processed and glycosylated to produce a 30-kDa polypeptide. This secreted protein was effectively removed from the membrane pellet, when subjected to alkaline treatment (Fig. 1B, lanes k and l). The results of this set of experiments (glycosylation of newly synthesized chains and resistance to alkaline extraction) showed that UGT1A6, despite the lack of its N-terminal cleavable signal peptide, was translocated and integrated into the ER membrane as the wild-type protein.
The in vitro targeting and translocation of UGT1A6 deprived of its natural signal peptide prompted us to further investigate ER targeting in vivo. A yeast P. pastoris expression system was developed to analyze the topogenic elements mediating membrane insertion of UGT1A6 in vivo. After subcloning both fulllength and truncated UGT1A6 cDNA into the expression vector pPICZB, the resultant recombinant plasmids were individually transformed in the methyltrophic yeast strain P. pastoris. Upon methanol induction, expression of both wild-type UGT1A6 and mutant lacking the signal peptide sequence was successfully achieved. Western blot analysis of whole cell extracts showed that the recombinant wild-type UGT1A6 migrated on SDS-polyacrylamide gel as a polypeptide band of approximately 55-56 kDa (Fig. 1C, lane a). Subfractionation experiments showed that, as expected, the protein was associated with the microsomal membranes and was resistant to alkaline (Fig. 1C, lanes b and c) and to high salt extraction (Fig.  1C, lanes d and e). Treatment of the microsomal fraction with endoglycosidase H under denaturating conditions resulted in the appearance of a lower molecular mass band (Fig. 1C, compare lane g to lane f), indicating that UGT1A6 was translocated and retained in the ER of recombinant yeast cells. Interestingly, Western blot analysis performed on recombinant yeast cell extracts expressing UGT1A6⌬SP (Fig. 1D) showed that this polypeptide exhibited the same apparent molecular mass as the wild-type (55-56 kDa), suggesting that despite the absence of signal sequence, the protein was also targeted and translocated across the ER (Fig. 1D, lanes a and b). Upon sodium carbonate (Fig. 1D, lanes c and d) or high salt (Fig. 1D, lanes e and f) extraction, the polypeptide remained associated with the microsomal fraction, indicating that the protein was integrated into the membranes. Moreover, the protein was sensitive to endoglycosidase H digestion (decrease of apparent molecular mass of 2 kDa, Fig. 1D, lane g) similarly to the wild-type protein suggesting a compartmentalization restricted to the ER. To further assess the intracellular localization of wild-type and mutant UGT1A6 expressed in yeast cells, immunofluorescence microscopy studies were carried out. The wild-type UGT1A6 exhibited a typical ER staining pattern (data not shown). The same ER staining pattern was observed in P. pastoris expressing UGT1A6⌬SP, confirming that the deletion of the UGT1A6 signal peptide did not prevent targeting and ER residency of the protein (data not shown). Based both on in vitro and in vivo experiments, these results provide evidence that the absence of the signal peptide does not preclude ER targeting, translocation, and retention of UGT1A6.
Evidence for an Internal Signal-like Sequence-Members of the UGT family present a hydrophobic stretch of 17 amino acids at the C terminus followed by a positively charged tail of about 26 amino acids, thought to be important for membrane integration and ER retention (16,17). Therefore, we analyzed the consequences of the deletion of this region by creating a mutant lacking 43 C-terminal amino acid residues as well as the N-terminal leader peptide (UGT1A6⌬SP/TMD/CT represented in Fig. 2A). In vitro transcription-translation showed that in the absence of microsomal membranes, the UGT1A6⌬SP/TMD/CT construct yielded a polypeptide of the expected size (about 48 -49 kDa) (Fig. 2B, lane b). Expression of UGT1A6 full-length cDNA in the same conditions is shown for comparison (Fig. 2B, lane a). When microsomal membranes were added, polypeptide bands of higher apparent molecular mass (about 49 -51 kDa) were produced (Fig. 2B, lane c). In the presence of competitive glycosylation inhibitor tripeptides, the expressed polypeptide comigrated with the unglycosylated product formed when the membranes were omitted (Fig. 2B,  compare lane d to lane b), indicating that UGT1A6 lacking both signal peptide and TMD/CT sequences was translocated and glycosylated in the lumen of the ER. Alkaline treatment (also shown in Fig. 2B) demonstrates that the majority of the The above construct was also successfully expressed in the yeast P. pastoris, allowing the in vivo examination of the subcellular targeting of the N-and C-terminal truncated UGT1A6 polypeptide (Fig. 2C). In whole yeast cell extracts, an immunoreactive protein about 3-4 kDa smaller than UGT1A6 and UGT1A6⌬SP was detected (Fig. 2C, lane c compared with lanes  a and b). In addition, similarly to the latter polypeptides, strong base (Fig. 2C, lanes d and e) and high salt extraction (Fig. 2C, lanes f and g) did not release UGT1A6⌬SP/TMD/CT protein from the microsomal membranes, as the majority of the protein was recovered in the pellet. This result indicates that, despite the absence of the predicted transmembrane domain, UGT1A6 remains an integral membrane-bound protein. Sensitivity to endoglycosidase H, as evidenced by a 2-kDa decrease of apparent molecular mass, indicated that UGT1A6⌬SP/ TMD/CT was N-glycosylated and retained in the ER of yeast cells (Fig. 2C, lane h).
Finally, to analyze the membrane orientation of UGT1A6 expressed as wild-type or truncated polypeptides, we employed a "protease protection assay" as described by Connolly et al. (30). The cDNAs encoding the UGT1A6 precursor, UGT1A6 lacking the N-terminal signal peptide only or lacking both the signal peptide and the C-terminal TMD/CT sequences (represented in Fig. 3A), were expressed in vitro. Fig. 3B shows the effect of proteinase K on the transcription-translation products. It is clear that UGT1A6, UGT1A6⌬SP, and UGT1A6⌬SP/ TMD/CT polypeptides were protease-protected, when synthesized in the presence of membranes (Fig. 3B, lanes a, b, and c,  respectively), whereas, in the absence of microsomes, the polypeptides were completely degraded (Fig. 3B, lanes d, e, and f, respectively) as well as in the presence of detergent (data not shown). These results suggested that the products of the truncated constructs have probably acquired a lumenal orientation similarly to that of the wild-type protein. This set of experiments provided evidence that the absence of the N-terminal signal peptide and of the C-terminal TMD/CT sequence did not prevent ER targeting, translocation, and membrane integra-tion of UGT1A6 either in vitro or in yeast cells. The most likely explanation of these observations is the presence of internal topogenic information capable of mediating the membrane targeting and retention process.
Identification of the Region Containing the Internal Signal Peptide Sequence-To find out the position of the putative signal-like region, serial N-terminal deletions of UGT1A6 were generated to successively remove computer-predicted hydrophobic stretches (33). Mutants lacking the coding sequence for 43, 68, 140, and 240 N-terminal amino acid residues and the C-terminal TMD/CT region were generated (Fig. 4A) and used for in vitro transcription-translation in the presence or absence of dog pancreas microsomes.
The deletion of 43 and 68 N-terminal amino acid residues and of the C-terminal TMD/CT region of UGT1A6 did not modify the glycosylation pattern nor the resistance to alkaline treatment of the protein (data not shown). The consequence of further deletion, i.e. 140 N-terminal amino acid residues after expression in a cell-free system, is illustrated in Fig. 4B. The in vitro expression of the mutant UGT1A6⌬-N140TMD/CT produced a polypeptide of about 35 kDa (Fig. 4B, lane a). When the expression system was supplemented with microsomes, higher molecular mass bands of 35-37 kDa were produced (Fig. 4B, lane e). This gel migration shift was reverted by the presence of glycosylation inhibitor tripeptides (Fig. 4B, lane d), suggesting that the UGT1A6⌬-N140/TMD/CT polypeptide was translocated across the ER membrane. Resistance to alkaline extraction further demonstrated that it was integrated into the lipid bilayer (Fig. 4B, lanes b and c). By contrast, the in vitro expression of the UGT1A6⌬-N240/TMD/CT mutant resulted in the production of polypeptide bands of an identical size when transcribed and translated in the presence or in the absence of membranes (Fig. 4B, lanes f and j), indicating that glycosylation did not take place and providing evidence that the removal  d and i,  respectively). Following expression of UGT1A6⌬-N140/TMD/CT or UGT1A6⌬-N240/TMD/CT, microsomal membranes were subjected to alkaline extraction and fractionated by centrifugation on a sucrose cushion to yield membrane-bound (P, lanes b and g, respectively) and supernatant (S, lanes c and h, respectively) fractions.
of 100 more amino acid residues (residue 141-240 region of UGT1A6) destroyed the capacity of UGT1A6 to cross the ER membrane. This was supported by the absence of an effect of acceptor tripeptides on the electrophoretic mobility of the expressed polypeptide (Fig. 4B, lane i). As expected, the majority of the expressed unglycosylated polypeptide was found in the supernatant following sodium carbonate extraction and 100,000 ϫ g centrifugation (Fig. 4B, compare lanes h and g). These findings strongly support the presence of an internal topogenic element located in the 141-240 region of UGT1A6 able to target and to retain UGT1A6 to the ER membrane.
UGT1A6 Inserts into Microsomal Membranes at the Co-and Posttranslational Levels-We next examined a possible posttranslational translocation mechanism of UGT1A6. For this purpose, translation of mRNAs encoding full-length and truncated UGT1A6 mutants was completed in a rabbit reticulocyte lysate. Further protein elongation was blocked by cycloheximide prior to the addition of microsomal membranes. The results clearly show that UGT1A6 and UGT1A6⌬SP as well as UGT1A6⌬SP/TMD/CT and UGT1A6⌬-N140/TMD/CT were efficiently translocated and glycosylated when expressed posttranslationally, as evidenced by the presence of slower migrating bands compared with their counterparts expressed in the absence of membranes (Fig. 5A, compare lanes a and b, c and d,  e and f, and g and h). As control, yeast prepro-␣-factor synthesized in identical conditions to UGT polypeptides (Fig. 5B, lane i) exhibited no posttranslational translocation (Fig. 5B, lane j), but was efficiently translocated and glycosylated when the membranes were present cotranslationally (Fig. 5B, lane k), in agreement with previous reports (34). Altogether, these data indicate that the UGT1A6 preprotein and truncation mutants can be targeted and inserted into membranes not only at cobut also at posttranslational level.
The 141-240 UGT1A6 Domain Addresses the EGFP Reporter Protein to the ER of Mammalian Cells-From the in vitro studies reported above, it was likely that the 141-240 region of UGT1A6 contains an internal signal-like sequence capable of membrane targeting and translocation. To assess the signal activity of this region in vivo, we tested its ability to target the fluorescent reporter EGFP protein to the ER. The chimeric and the native EGFP cDNAs were transiently expressed in HeLa cells. We examined the intracellular localization of the ex-pressed proteins by direct fluorescence microscopy 24 h after transfection. The native EGFP expression resulted in an homogeneous staining characteristic of a cytoplasmic location (Fig. 6A). Interestingly, cells expressing the UGT141-240/ EGFP chimera presented a dramatically different fluorescent staining pattern consistent with an ER localization (Fig. 6B). This staining pattern overlapped with that of calnexin, a typical ER marker (Fig. 6C). These observations provide strong evidence that the 141-240 region of UGT1A6 contains a targeting and retention signal able to confer ER residency to the cytoplasmic EGFP reporter protein in mammalian cells.

DISCUSSION
It has become clear over the past decade that selective trafficking and compartmentalization of proteins throughout the cell requires the recognition of sorting determinants present in their primary structure (1,35,36). A significant finding from the present study is that internal topogenic information contained within the N-terminal half of UGT1A6 functions as an ER targeting and retention signal, in addition to the predicted N-terminal cleavable signal peptide sequence and C-terminal transmembrane anchor domain.
For eukaryotic proteins with a type I topology such as UGT1A6, the best characterized insertion mechanism is the signal recognition particle-dependent pathway in which membrane targeting is initiated cotranslationally by a signal peptide at the N terminus of the nascent peptide. For these proteins, integration occurs via a complex multistep process ending with the release of the polypeptide into the lipid bilayer coincident with the completion of protein synthesis. A first important result of this study is that the deletion of the Nterminal cleavable leader sequence of UGT1A6 does not prevent targeting and ER insertion. The addition of carbohydrate as an endogenous marker for peptides exposed to the lumen of the ER, resistance to alkaline extraction and protease protection showed that UGT1A6 lacking the signal peptide expressed in vitro is an integral membrane-bound protein and that the N-glycosylation sites have achieved a lumenal orientation. We further investigated the role of the signal peptide in vivo using, for the first time, the yeast P. pastoris expression system. High levels of expression enabled us to analyze the fate of the mutant lacking the signal peptide compared with the wild-type protein. Based on the same criteria as for in vitro analysis, we found that ER targeting of UGT1A6 in yeast cells did not require the N-terminal cleavable signal peptide. Both wild-type and mutant proteins were integral membrane-bound polypeptides, as evidenced by resistance to alkaline and high salt extraction. Moreover, the protein was sensitive to endoglycosidase H, indicating that despite the lack of the N-terminal cleavable signal peptide, the predicted intralumenal domain of UGT1A6 was translocated across the yeast ER membrane. Immunofluorescence observation confirmed that UGT1A6⌬SP was localized in the ER of yeast cells, as the wild-type protein.
Altogether, within the resolution of the in vitro and in vivo assays performed in this study, ER targeting, translocation, and retention of UGT1A6 were not prevented by the absence of the N-terminal signal peptide sequence. The ability of proteins deprived of their natural signal peptide to be targeted and translocated across the ER is unusual but not unprecedented. It was reported that the cleavable signal peptide of the yeast secreted carboxypeptidase Y (37) and acid phosphatase (38) could be entirely removed without abolishing the ability of these proteins to be translocated and glycosylated in the ER. A related observation was also recently described for the human sodium-calcium exchanger, an integral plasma membrane polytopic protein containing a N-terminal cleavable leader sequence (39). The authors showed that, despite the lack of signal peptide sequence, the Na ϩ -Ca 2ϩ exchanger was addressed to its normal subcellular localization.
It has been proposed that, when the wild-type N-terminal signal peptide is deleted or made dysfunctional, hydrophobic internal segments can promote the translocation of the protein (40). This prompted us to search for some structural features other than the typical N-terminal cleavable signal peptide involved in ER targeting and assembly of UGT1A6. UGT1A6 is predicted by hydropathy analysis to feature a C-terminal membrane spanning domain, which may act as a targeting retention signal in the absence of the N-terminal signal peptide. The finding that a UGT1A6 polypeptide lacking both the N-terminal leader peptide and the C-terminal TMD/CT domain was targeted into the ER when expressed either in vitro or in yeast cells was a crucial argument for the presence of an internal signal-like sequence. Furthermore, the observation that this Nand C-terminal truncated protein was tightly associated with the microsomal membranes showed that some region of UGT1A6 not only acts as a signal sequence but also as an ER retention determinant. Moreover, the sensitivity of UGT1A6⌬SP/TMD/CT to endoglycosidase H deglycosylation clearly shows that translocation of the lumenally exposed domain of UGT1A6 was achieved. These data also suggest that the truncated polypeptide has not moved past an ER/pre-Golgi compartment. Since computer-based secondary structure prediction identifies only the N-terminal cleavable signal peptide and C-terminal TMD/CT region as topogenic elements, the membrane targeting and association of UGT1A6⌬SP/TMD/CT, in both expression systems used, was unexpected. Interestingly, computer based analysis predicted some buried ␣-helices in UGT1A6 as well as in other members of the UGT1 family (41). Although these domains are not, in theory, long enough to span the membrane, they could interact with the lipid bilayer via hydrophobic interactions. Resistance to protease treatment carried out in this study and by others (18,20) does not favor the hypothesis of the presence of other membrane-spanning segments than the C-terminal TMD. The observation that the interactions of UGT1A6⌬SP/TMD/CT with the membrane cannot be disrupted by standard extraction techniques utilizing high pH and high salt concentration confirms that it is an integral membrane-bound polypeptide. In agreement with our results, Meech and Mackenzie (42) found that the rat liver UGT2B1 isoform remained resident in the ER of recombinant mammalian cells in the absence of a cytosolic tail and transmembrane domain. It is noteworthy that the hydrophobic transmembrane helix is not the only possible membrane binding motif. Determination of the crystal structure of the prostaglandin H2 synthase has led to a model of membrane insertion on the lumenal face of the ER mediated by a motif of short ␣-helices, which do not span the ER membrane (43). Based on crystal structure analysis, similar membrane-binding characteristics have also been reported for a squalene cyclase from Alicyclobacillus acidocaldarius (44).
N-terminal deletion analysis performed in an in vitro expression system allowed us to localize the targeting-retention signal within region 141-240 of UGT1A6. Indeed, addition of N-linked carbohydrates and resistance to alkaline extraction were abolished when 240 N-terminal residues were deleted. The ability of the 141-240 domain to confer ER residency to the soluble EGFP reporter protein unequivocally established the signal and retention function of this sequence. The highly conserved secondary structural features of this region supports the notion that it has a specific role in the biogenesis of UGTs. It has been suggested that the need for an internal signal sequence arises from the entire protein being too large to be translocated across the membrane and hence the requirement to re-engage the translocation channel (4). Such an internal signal may be critical in the case of UGTs to allow the localization of the majority of the polypeptide on the lumenal face of the ER.
It must be emphasized that aglycone substrates of UGTs are liphophilic molecules. It is accepted that they reach their binding site localized in the N-terminal half of the proteins by passive diffusion through the lipid bilayer (45). An attractive hypothesis is that the membrane interaction conferred by the 141-240 region may provide a hydrophobic path from the membrane interior to the catalytic site. This mode of membrane interaction may also explain why lipids are important in the functional integrity of these phospholipid-dependent enzymes.
Translocation of type I transmembrane proteins with a Ncleavable signal peptide such as UGTs across the ER membrane involves the cotranslational delivery of the nascent chain. Surprisingly, we found that the UGT1A6 precursor as well as the N-and C-terminal mutants can insert into the microsomal membranes posttranslationally. The existence of a posttranslational mode of translocation has been clearly demonstrated in yeast (46,47) but is far less documented in higher eukaryotic species (48). It has been suggested in some cases such as type IV or C-anchored transmembrane proteins, UCB6 (6), and Bcl-2 proto-oncogenic protein (49). These proteins are characterized by hydrophobic segments close to or at their C termini, precluding cotranslational membrane insertion (6).
From the results of this study, we propose that UGT1A6 membrane assembly is mediated by an internal ER targeting and retention signal, presumably acting through a posttranslational pathway, in addition to the known predicted topogenic determinants, i.e. the N-terminal cleavable signal sequence and the C-terminal TMD/CT domain. It is noteworthy that the involvement of a combination of topogenic elements for the achievement of correct subcellular localization is described for a growing number of membrane proteins. A polytopic membrane protein, the cystic fibrosis transmembrane conductance regulator, has been convincingly demonstrated to acquire its N terminus topology through the arrangement of the cotranslational action of the first transmembrane domain and the posttranslational action of the second transmembrane domain (50). In the same manner, the membrane integration of the gap junctional protein connexin 26 was shown to occur posttranslationally possibly involving an internal signal sequence (51).
We bring here convincing evidence that the membrane assembly of the human UGT1A6 in the ER membrane involves a previously unidentified internal signal and retention sequence in addition to the N-terminal cleavable signal peptide and stop-transfer domain. Our results support the idea that this topogenic element not only mediates targeting and translocation of UGT1A6 but also features a characteristic membrane association domain likely to be of major importance in the structure and function of this protein.