Endoplasmic Reticulum Localization of Gaa1 and PIG-T, Subunits of the Glycosylphosphatidylinositol Transamidase Complex*

After integration into the endoplasmic reticulum (ER) membrane, ER-resident membrane proteins must be segregated from proteins that are exported to post-ER compartments. Here we analyze how human Gaa1 and PIG-T, two of the five subunits of the ER-localized glycosylphosphatidylinositol transamidase complex, are retained in the ER. Neither protein contains a known ER localization signal. Gaa1 is a polytopic membrane glycoprotein with a cytoplasmic N terminus and a large luminal loop between its first two transmembrane spans; PIG-T is a type I membrane glycoprotein. To simplify our analyses, we studied Gaa1 and PIG-T constructs that could not interact with other subunits of the transamidase. We now show that Gaa1282, a truncated protein consisting of the first TM domain and luminal loop of Gaa1, is correctly oriented, N-glycosylated, and ER-localized. Removal of a potential ER localization signal in the form of a triple arginine cluster near the N terminus of Gaa1 or Gaa1282 had no effect on ER localization. Fusion proteins consisting of different elements of Gaa1282 appended to α2,6-sialyltransferase or transferrin receptor could exit the ER, indicating that Gaa1282, and by implication Gaa1, does not contain any dominant ER-sorting determinants. The data suggest that Gaa1 is passively retained in the ER by a signalless mechanism. In contrast, similar analyses of PIG-T revealed that it is ER-localized because of information in its transmembrane span; fusion of the PIG-T transmembrane span to Tac antigen, a plasma membrane-localized protein, caused the fusion protein to remain in the ER. These data are discussed in the context of models that have been proposed to account for retention of ER membrane proteins.

The mechanism(s) by which ER 1 -resident proteins are sorted from those destined for post-ER compartments is a subject of ongoing debate (1)(2)(3)(4)(5). Possible scenarios to explain ER retention of proteins include trapping of ER residents within a matrix or oligomeric structure, signal-mediated retention, lack of an export signal, and signal-mediated retrieval from post-ER compartments. These scenarios must be viewed in the context of bulk anterograde membrane flow from the ER that has been hypothesized to nonselectively carry membrane and fluid content to post-ER compartments (6,7). The two best characterized ER localization signals, the KDEL and dilysine motifs, are based on retrieval mechanisms that involve direct or indirect binding of the motif to the COPI coats of vesicles mediating retrograde transport from the Golgi to the ER. The KDEL motif is found on soluble proteins that reside within the ER lumen (8 -11); the cytoplasmically exposed dilysine motif serves a similar function for ER-resident membrane proteins (12,13). However, only a small fraction of any given type of molecule bearing a KDEL or dilysine motif appears to exit the ER and needs to be retrieved; the majority appear not to leave the ER at all. The question remains. How do ER resident proteins remain resident in the ER?
It is possible that many ER resident proteins are components of hetero-or homo-oligomeric complexes that in turn interact with other complexes to yield an aggregate that is too large to export efficiently. Consistent with this are observations of the slow lateral diffusion of oligosaccharyltransferase complexes that associate with the protein translocon (14). Alternatively, some ER residents, like ER chaperones, may be continuously engaged with protein traffic entering the ER such that their own exit is impeded. These interactions or the numerous quasi-interactions that are probably promoted by the high protein density in the ER may be sufficient to retain proteins if the rate of bulk flow of membrane and fluid departing the ER is low and if the proteins lack explicit export signals. We classify these possibilities under the heading of "signalless" retention mechanisms.
In contrast, it is possible to consider retention resulting from an explicit signal within the retained protein. For the particular case of ER membrane proteins, it is conceivable that sequence motifs in one or more of the topological domains of the protein (cytoplasmic, transmembrane (TM), or luminal) interact specifically with effectors to block exit of the protein from the ER. However, many ER membrane proteins appear to enjoy relatively unencumbered lateral diffusion in the membrane, suggesting that their presumed exclusion from ER-derived transport vesicles is not due to immobility in the membrane resulting from, for example, cytoskeletal association or large complex formation (15,16). The few experimentally characterized ER retention motifs, such as cytoplasmic N-terminal and internal diarginine sequences, cytoplasmic C-terminal tyrosine-based sequences, or transmembrane domain sequences, are not universal and/or share no sequence similarity between other proteins (1,(17)(18)(19)(20)(21)(22)(23). The mechanism by which these motifs promote retention is unknown.
Many ER-translocated proteins are covalently modified with a glycosylphosphatidylinositol (GPI) anchor (24). GPI anchor attachment is catalyzed by GPI transamidase (GPIT), a membrane-bound protein complex composed of at least five sub-units: Gpi8, Gaa1, PIG-S, PIG-T, and PIG-U (25-30). GPIT removes a C-terminal GPI-directing signal peptide from the translocated protein and attaches a preassembled GPI anchor via an amide linkage to the newly exposed C-terminal amino acid residue of the protein (31,32). Gpi8 is presumed to form the catalytic center of GPIT (33,34), but all subunits are essential for function. Recent data suggest that Gaa1 and possibly PIG-U may play a role in recruiting GPI to GPIT (28,35), and it has been proposed that a ␤-propeller structure in the luminal domain of PIG-T gates access to the active site of Gpi8 (36). The role of PIG-S is unknown. Gpi8 and PIG-T are type I transmembrane proteins; the other three subunits have multiple transmembrane spans. Recent studies show that human Gaa1 has a cytoplasmically oriented N terminus and spans the ER membrane seven times (37); the membrane topology of PIG-S and PIG-U remains to be experimentally determined.
We are interested in understanding how the GPIT complex is assembled into a functional unit that is localized to the ER membrane. The individual subunits are presumably independently synthesized and integrated into the ER membrane and must remain in the ER long enough to encounter the other subunits in order to form the GPIT complex. The complex itself, after assembly, must remain in the ER in order to execute its catalytic function. Thus, individual subunits as well as the GPIT complex as a whole must be able to avoid entering the secretory pathway and exiting the ER.
To identify the mechanism(s) by which the subunits of GPIT are localized to the ER, we chose initially to analyze Gaa1 and PIG-T. We used fluorescence microscopy and glycosidase digestion to analyze the subcellular localization of epitope-tagged variants of these two proteins as well as that of fusion proteins prepared using elements of Gaa1 or PIG-T combined with sequences from the Golgi protein ␣2,6-sialyltransferase or the plasma membrane-localized proteins transferrin receptor and Tac antigen. This "cut-and-paste" approach revealed that Gaa1 does not contain any dominant ER localization information, whereas PIG-T is localized to the ER by virtue of a signal contained within its TM domain. These data are discussed in the context of other recent examples of signal-mediated and signalless retention of ER-resident membrane proteins.

MATERIALS AND METHODS
Antibodies-Anti-Gpi8 and anti-Gaa1 rabbit polyclonal antibodies were generated as described previously (37). Affinity-purified rabbit anti-PIG-T antibody was a gift from Dr. Taroh Kinoshita (Osaka University). Mouse monoclonal anti-FLAG antibody M2 was purchased from Sigma; mouse monoclonal anti-V5 antibody was purchased from Invitrogen. Rabbit polyclonal anti-V5 antibody was a gift from Dr. Karen Colley (University of Illinois at Chicago). Horseradish peroxidase-conjugated anti-rabbit and anti-mouse IgGs were from Promega (Madison, WI).
To replace the N-terminal cytoplasmic part and TM domain of rat ␣2,6-sialyltransferase (ST; the construct we used encoded ST with tyrosine, rather than cysteine, at amino acid 123) with the aminoterminal cytoplasmic part and TM of Gaa1, two PCR fragments were generated: (i) from pEF/D1V5 vector using T7 and GN3 (5Ј-tgggaattcgaggcgcagcactagg-3Ј) primers and (ii) from pcDNA3.1/ST TYR vector (a gift from Dr. Karen Colley, University of Illinois at Chicago) using primers STLUM (5Ј-gggcagaattcgagcgactatgaggccct-3Ј) and BGH (5Јcaactagaaggcacagtcgagg-3Ј). The obtained PCR products were digested with BamHI/EcoRI and EcoRI/XbaI, respectively, and ligated into BamHI/XbaI pEF6/V-5 His vector. The resulting Gaa1 55 -ST 30 -402 construct includes N-terminal sequence of the Gaa1 (aa 1-55), followed by Glu and Phe and then residues 30 -402 of human ST, fused with V5 epitope and His 6 sequence. The same strategy was used to generate other chimeric proteins. The Gaa1 282 -ST 30 -402 construct contains Nterminal sequence of the Gaa1 (aa 1-282) fused to ectodomain of the ST (aa 30 -402).
To obtain human transferrin receptor fused at its C terminus with V5 and His 6 epitope tags, sense T7 promoter and antisense TFR3 (5Јgccctctagaaactcattgtcaatgtccca-3Ј) primers were used for PCR amplification from plasmid pSVT7, (bearing human transferrin receptor cDNA) as template DNA (a gift from Dr. Paul Gleeson, University of Melbourne, Australia). The resulting PCR product was digested with EcoRI/XbaI and ligated into EcoRI/XbaI pEF6/V-5 His vector.
To replace the N-terminal part of human transferrin receptor (TfR) with the N-terminal cytoplasmic part of Gaa1, two PCR fragments were generated: (i) from pEF/D1V5 vector using T7 and GN3 (5Ј-tgggaattcgaggcgcagcactagg-3Ј) primers and (ii) from pSVT7/TfR vector using primers TFR62 (5Ј-tgggaattctgtagtggaagtatctgctatgg-3Ј) and TFR3. Amplified DNA products were digested with BamHI/EcoRI and EcoRI/ XbaI, respectively, and ligated into BamHI/XbaI pEF6/V-5 His vector. The resulting Gaa1 19 -TfR 63-760 construct includes the N-terminal 19 residues of hGaa1, followed by Glu and Phe followed by residues 63-760 of the human transferrin receptor, fused with V5 epitope and His 6 sequence. The same strategy was used to generate other chimeric proteins. The Gaa1 55 -TfR 87-760 construct contains N-terminal sequence of the Gaa1 (aa 1-55) fused to the ectodomain of TfR (aa 87-760).
Cell Culture and Transfection-HeLa cells were cultured at 37°C in a humidified 5% CO 2 atmosphere in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal bovine serum. Cells were transfected with cDNA constructs by electroporation as described previously (37).
Immunoprecipitations and Immunoblots-Transfected HeLa cells (1-2 ϫ 10 7 cells) were harvested by scraping 48 h post-transfection, washed once with PBS, resuspended in 1 ml of MSB buffer (20 mM Hepes-KOH, pH 7.6, 200 mM NaCl, 1% digitonin or Nonidet P-40, and 1ϫ protease inhibitor mixture (Calbiochem)), and solubilized on ice for 30 min. The resulting cell lysate was clarified by centrifugation at 10,000 ϫ g for 20 min at 4°C. The 10,000 ϫ g supernatant was incubated with 30 l of anti-FLAG M2, or anti-V5 agarose (Sigma) slurry was added and incubated at 4°C for 4 h with gentle agitation. The beads were pelleted by 15-s centrifugation at 10,000 ϫ g. The samples were washed four times for 5 min each in 1.5 ml of buffer MSB. Bound antigen was released from the beads by incubation with FLAG or V5 peptide (200 g/ml) in buffer MSB. Immunoprecipitated fractions were analyzed by SDS-PAGE, followed by immunoblotting using chemiluminescence reagents (Pierce).
Pulse-Chase Labeling-Transfected HeLa cells were analyzed 48 h post-transfection. The cells were pulse-labeled for 10 min with EXPRE 35 S 35 S Protein Labeling Mix (1 mCi/ml; PerkinElmer Life Sciences) in methionine/cysteine-free complete medium and then chased for 2-8 h at 37°C. Cell lysates were prepared, and immunoprecipitations were carried out as described above. Immunoprecipitated fractions were resolved by SDS-PAGE. After electrophoresis, the gels were dried and radiolabeled proteins were visualized and quantitated using a PhosphorImager (Amersham Biosciences) and ImageQuant software (Amersham Biosciences).
Fluorescence Microscopy-Transfected HeLa cells were plated on glass coverslips, cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum, fixed with 4% paraformaldehyde, permeabilized with 0.3% (w/v) Triton X-100, and incubated with primary and secondary antibodies as described previously (37). The labeled cells were visualized using a Bio-Rad confocal microscope (type MRC 1000).

Endoglycosidase H (Endo H) and
Peptide:N-Glycosidase F (PNGase F) Treatment-Endo H and PNGase F from New England Biolabs was used for carbohydrate digestion. The treatment was done as described previously (35).

A Triple Arginine Motif in the Cytoplasmically Oriented Nterminal Sequence of Gaa1 Does Not Control ER Localization of
the Protein-We previously suggested that the cytoplasmic N terminus of Gaa1 might function as an ER sorting determinant (37). Inspection of the primary sequence of human Gaa1 identified an internal triple arginine cluster (Arg 9 -Arg 10 -Arg 11 ) near the N terminus (Fig. 1A). This arginine cluster resembles the Arg-Xaa-Arg ER localization motifs recently identified in subunits of multimeric protein complexes displayed at the cell sur-face (18,39,40). Correct assembly of these protein complexes and their transport to the cell surface appears to be regulated through control of the trafficking of individual subunits, such that incompletely assembled complexes are retained in the ER. The secretory transport machinery is able to differentiate between unassembled units and assembled complexes because the Arg-Xaa-Arg motif in individual subunits can be masked after assembly of the protein into the complex or because the ability of the motif to signal ER retention/retrieval can be suppressed in some cases by phosphorylation of adjacent residues.
We hypothesized that the Arg 9 -Arg 10 -Arg 11 motif in Gaa1 acts as an ER localization signal. We tested this hypothesis by preparing a Gaa1 variant (denoted Gaa1(R3 A)) in which the three arginines were replaced with alanine residues. HeLa cells were transfected with plasmids encoding wild type, V5epitope-tagged Gaa1 (Fig. 1A), or similarly tagged Gaa1(R3 A). The subcellular distribution of the transiently expressed proteins was determined by indirect immunofluorescence microscopy using anti-V5 antibodies. Fig. 1B shows that both wild type Gaa1 and Gaa1(R3 A) display a reticular staining pattern consistent with ER localization. Gaa1 and Gaa1(R3 A) were extracted from cells in detergent-containing buffer, treated with glycosidases, and analyzed by SDS-PAGE and immunoblotting. Fig. 1C shows that both proteins are modified with a single, endoglycosidase H-susceptible N-glycan, suggesting that the wild-type Gaa1 and Gaa1(R3 A) are not exposed to Golgi N-glycan modification enzymes.
We considered the possibility that, without its arginine cluster, Gaa1(R3 A) could potentially be exported from the ER but is unable to leave the compartment as a result of improper folding. Misfolding would mitigate the ability of the protein to pass ER quality control systems for secretion and potentially accelerate its rate of turnover. Thus, compared with wild-type Gaa1, Gaa1(R3 A) may be turned over more rapidly by ERassociated degradation pathways. To test whether this was the case, we used [ 35 S]methionine/cysteine to pulse-label HeLa cells that were transiently expressing wild-type Gaa1 or Gaa1(R3 A) and then chased the cells for various periods of time (up to 8 h), extracted proteins in detergent-containing buffer, and used immunoprecipitation, SDS-PAGE, and Phos-phorImager analysis to determine the amount of radiolabeled Gaa1 or Gaa1(R3 A) remaining in the cells. The SDS-PAGE/ PhosphorImager data for one experiment are shown in Fig. 1D, whereas the quantitation of the rate of decay taken from three independent experiments is shown in Fig. 1E. The data show that Gaa1(R3 A) turns over with a half-time of ϳ3.2 h, similar to the rate of turnover of wild-type Gaa1 (a single, monoexponential decay curve is plotted for both proteins in Fig. 1E). This result suggests that it is unlikely that folding/degradation problems underlie the ER retention of Gaa1(R3 A). Together with the immunofluorescence localization of Gaa1(R3 A) to the ER and the Endo H-susceptibility of its N-glycan, these data suggest that the triple arginine motif (Arg 9 -Arg 10 -Arg 11 ) is not necessary for ER localization of Gaa1.
Gaa1 282 , a Construct Consisting of the N-terminal Cytoplasmic Tail, First TM Domain, and Part of the N-Glycosylated Luminal Loop of Gaa1, Is ER-localized-Co-immunoprecipitation experiments indicated that Gaa1(R3 A), like Gaa1, interacts with Gpi8, PIG-T, and PIG-S (data not shown). It is therefore conceivable that Gaa1(R3 A) is retained in the ER through its interactions with other subunits of the GPI transamidase and that these interactions render potential localization information in the triple arginine motif redundant. Alternatively, it is possible that full-length Gaa1 contains more than one ER localization signal, rendering individual signals redundant.
To clarify this issue, we investigated the subcellular local- ization of Gaa1 282 , a truncated molecule consisting of the first 282 amino acids of Gaa1 capped with a C-terminal V5 epitope tag ( Fig. 2A). Our choice of truncation site was based on previous results that indicated that a similar construct containing a larger stretch of the luminal loop (up to amino acid 367, right before the start of the second TM domain) was proteolytically cleaved in situ to yield a membrane-anchored fragment analogous to Gaa1 282 ; this fragment was ER-localized but not able to bind to Gpi8, PIG-T, and PIG-S (37). In addition to Gaa1 282 , we generated a construct, Gaa1 282 (R3 A), in which the triple arginine cluster (Arg 9 -Arg 10 -Arg 11 ) was altered to Ala 9 -Ala 10 -Ala 11 . Neither Gaa1 282 nor Gaa1 282 (R3 A) is expected to interact with other transamidase subunits. We confirmed this via co-immunoprecipitation experiments that showed that the Gaa1 282 constructs, unlike full-length Gaa1, could not co-immunoprecipitate endogenous Gpi8 or PIG-T (Fig. 2B). Co-immunoprecipitation experiments done on lysates prepared from cells that co-expressed the Gaa1 282 constructs with green fluorescent protein-tagged PIG-T similarly demonstrated that the truncated proteins could not co-immunoprecipitate PIG-Tgreen fluorescent protein (data not shown). Indirect immunofluorescence microscopy showed that both Gaa1 282 and Gaa1 282 (R3 A) were ER-localized (Fig. 2C), and glycosidase digestion showed that both possessed an Endo H-susceptible N-glycan (Fig. 2D). These results reflect the data obtained with full-length Gaa1 constructs (Fig. 1) and confirm that the arginine cluster in the N-terminal cytoplasmic tail of Gaa1 is not required for correct topological insertion of the protein or for its ER localization.

Gaa1 282 Does Not Contain a Dominant ER Retention
Signal-Our results thus far eliminate a role for the triple arginine cluster (Arg 9 -Arg 10 -Arg 11 ) in ER localization of Gaa1 and Gaa1 282 but leave open the question of how these proteins are localized to the ER. To simplify our analyses, we focused on Gaa1 282 . The correct topological insertion and N-glycosylation of Gaa1 282 (Fig. 2D), together with its inability to interact with other GPIT subunits (Fig. 2B), suggested to us that this model protein would provide a good starting point from which to identify sequence elements in Gaa1 that contribute to its ER localization. We considered the possibilities that Gaa1 is ERlocalized, because it contains an ER localization signal located within the Gaa1 282 fragment or because both it and Gaa1 282 lack a signal for efficient export from the ER (3). To distinguish these possibilities, we examined the subcellular localization of chimeric constructs consisting of sequence elements of Gaa1 282 fused with membrane proteins destined for post-ER compartments. Since the first TM span of Gaa1 crosses the ER membrane in a type II orientation, we chose two well characterized type II membrane proteins, trans-Golgi network-localized ST (we used the STtyr isoform for our studies (41)) and plasma membrane-localized TfR, to prepare fusion proteins for these experiments.
We first generated a series of Gaa1-ST fusion proteins bearing a C-terminal V5 epitope tag (Fig. 3A). Indirect immunofluorescence microscopy showed V5-tagged ST to be primarily Golgi-localized (Fig. 3B), and glycosidase digestion showed that a portion of the ST population possessed Endo H-resistant N-glycans, consistent with its post-ER localization (Fig. 3C,  lanes 1-3; the Endo H-resistant band is indicated by an asterisk in lane 2). ST is glycosylated at Asn 146 and Asn 158 (41), but, despite the trans-Golgi network localization of the protein, only a fraction of the ST population is known to bear glycans that are Endo H-resistant (42). This is consistent with our results.
A fusion construct, Gaa1 19 -ST 7-402 , consisting of the first 19 amino acids of Gaa1 appended to ST 7-402 (ST lacking the first 6 amino acids) (Fig. 3A), showed intense Golgi fluorescence and a background of reticular staining by indirect immunofluorescence microscopy when transiently expressed in HeLa cells (fluorescence microscopy data for this construct were described in a previous publication (37) but are reiterated here to provide an experimental control). Similar to the result obtained with ST, a fraction of the Gaa1 19 -ST 7-402 molecules displayed an Endo H-resistant glycan (Fig. 3C, lanes 4 -6; the Endo H-resistant fraction is denoted by an asterisk in lane 5). A second fusion construct (Gaa1 55 -ST 30 -402 (Fig. 3A)), consisting of the N-terminal sequence and first TM domain of Gaa1 (amino acids 1-55) fused to the ectodomain of ST (amino acids 30 -402), behaved similarly to Gaa1 19 -ST 7-402 . When transiently expressed in HeLa cells, the protein acquired an Endo H-resistant form (asterisk in Fig. 3C, lanes 7-9) and was detected both in the Golgi and in the ER (Fig. 3B). Like ST, Gaa1 55 -ST 30 -402 could be detected at the cell surface by immunofluorescence microscopy; however, unlike ST (41), Gaa1 55 -ST 30 -402 was not cleaved efficiently to release its ectodomain into the extracellular medium (data not shown). The export of a significant fraction of Gaa1 19 -ST 7-402 and Gaa1 55 -ST 30 -402 to the Golgi indicates that the cytoplasmically disposed, 19-amino acid N-terminal sequence of Gaa1, alone or in concert with the first TM domain, does not contain a dominant ER localization signal. Gaa1 19 -ST 7-402 contains a short remnant of the N-terminal cytoplasmic tail of ST that includes a cluster of lysine residues (Lys 7 -Lys 8 -Lys 9 ) located proximal to the TM domain (Fig. 3A). It has recently been suggested that dibasic (R/K)X(R/K) motifs in the membrane-proximal region of the cytoplasmic tail of Golgiresident glycosyltransferases act as ER exit signals by promoting interactions between the glycosyltransferase tail sequence and the COPII vesicle coat (43). It is therefore possible that the export of Gaa1 19 -ST 7-402 from the ER is dictated by this motif. However, the motif is absent in Gaa1 55 -ST 30 -402 , a fusion protein that is exported with similar efficiency. This suggests that the luminal part of ST may also contain an ER exit determinant(s).
To complete our analysis of the role of sequence elements of Gaa1 282 that could play a role in ER-localization of the protein, we next prepared Gaa1 282 -ST 30 -402 , a chimeric protein in which the luminal domain of ST (amino acids 30 -402) was appended to the C terminus of Gaa1 282 . As with the other fusion proteins described above, a significant portion of Gaa1 282 -ST 30 -402 was detected in the Golgi and in an Endo H-resistant form (Fig. 3, B and C). These results clearly show that the N-terminal part of Gaa1 (amino acids 1-282) is unable to retain the ST ectodomain in the ER. This result indicates that Gaa1 282 either does not contain an explicit ER localization signal or, if it does, the signal is dominated by export information in the ST ectodomain. On a separate note, our data mitigate concerns about folding of the unnatural chimeras that we describe, since, in each case (Gaa1 19 -ST 7-402 , Gaa1 55 -ST 30 -402 , and Gaa1 282 -ST 30 -402 ), at least a fraction of the synthesized chimeras is export-competent and, by implication, correctly folded. Problems stemming from potential saturation of export or retention machinery by overexpression of protein cargo are also not of major concern, since the expression levels of all our constructs are similar; this suggests that retention of a particular construct (e.g. Gaa1 282 ) is not due to saturation of the export machinery, since a related construct (e.g. Gaa1 282 -ST 30 -402 ), expressed at a similar level, can be exported.
Similar data were obtained when sequence domains of Gaa1 282 were appended to the human TfR (Fig. 4A). Cell surface display of transiently expressed Gaa1 19 -TfR 63-760 and Gaa1 55 -TfR 87-760 and Endo H resistance of the chimeric proteins indicate that they exit the ER (Fig. 4, B and C; Endo H-resistant bands are indicated by an asterisk). We conclude that neither the cytoplasmically disposed N-terminal sequence of Gaa1 nor its first TM domain contains an ER localization signal capable of retaining Gaa1-TfR fusion proteins in the ER. Our overall conclusion from these studies is that the ER localization of Gaa1, as embodied by our analyses of Gaa1 282 , is not signal-mediated and may instead reflect the lack of an export signal.
ER Localization of PIG-T-We next turned our attention to the ER localization of PIG-T, a GPIT subunit with a simpler membrane topology than Gaa1. The 34-amino acid C-terminal cytoplasmic tail of human PIG-T ( 545 NLLTRTFHIEEPRTG-GLAKRLANLIRRARGVPPL 578 ; the charged amino acids are shown in boldface type) has 10 charged residues with an overall positive charge. As in the case of cytoplasmically disposed  arginine or lysine clusters that have been shown to act as ER localization signals for the membrane proteins (44), we hypothesized that the charged residues in the PIG-T cytoplasmic tail may be involved in its ER retention. To test this, we made a T 545 construct (Fig. 5A, middle; compare with the image for intact, epitope-tagged PIG-T in the top panel) in which the C-terminal 33 amino acids were replaced with a V5 epitope tag. (We previously showed that the membrane-proximal cytoplasmic display of the V5 tag does not cause ER localization of proteins destined for post-ER compartments (38); see also be-low.) When transiently overexpressed in HeLa cells, the recombinant protein was found located only in the ER, suggesting that the C-terminal sequence is not required for ER localization (Fig. 5A, middle). However, like PIG-T, T 545 interacts with other GPIT subunits as detected by co-immunoprecipitation of endogenous Gpi8 (Fig. 5B, lanes 1-4) and Gaa1 (data not shown); these interactions may be a contributing factor in its ER localization.
We prepared a second truncated construct, T 514 , consisting of only the luminal portion of PIG-T terminating in a C-terminal FLAG epitope tag. We tested whether this mutant was localized to the ER and able to bind other transamidase subunits. T 514 did not co-precipitate Gpi8 (Fig. 5B, lanes 5 and 6) or Gaa1 (data not shown), suggesting that without its transmembrane domain PIG-T does not interact efficiently with other GPIT subunits. Although fluorescence microscopy showed an ER-Golgi staining pattern for T 514 (Fig. 5A, bottom), the protein was secreted into the medium (Fig. 5C, lane 4) in the form of two higher molecular weight forms suggestive of N-glycan processing and, for the upper band, possibly O-glycan addition. Both secreted bands were Endo H-resistant but PNGase Fsensitive, consistent with this proposal. These results suggest that PIG-T without its membrane anchor and lacking the ability to complex with other GPIT subunits can be exported from the ER. We conclude that the TM domain of PIG-T may be required directly or indirectly (by facilitating interaction between PIG-T and other GPIT subunits) for localizing PIG-T in the ER.
To test whether the TM domain plays a direct role in ER retention of PIG-T, we prepared a chimeric construct, Tac e -T 516 -578 consisting of the 237-amino acid ectodomain of human interleukin-2 receptor (Tac) appended to the TM domain and C-terminal cytoplasmic tail of PIG-T (amino acids 516 -578) (Fig. 6A). When HeLa cells expressing Tac e -T 516 -578 were assessed by indirect immunofluorescence microscopy, the fusion protein was found located in the ER (Fig. 6B, lower panels), whereas Tac was displayed, as expected, at the cell surface (Fig. 6B, top panels). Unlike Tac, Tac e -T 516 -578 appears not to acquire high molecular weight O-glycans (Fig. 6C, lanes 1 and  3), and its N-glycans are Endo H-sensitive (Fig. 6D). These data are consistent with the conclusion that Tac e -T 516 -578 is ERlocalized. Co-precipitation experiments show that Gpi8 does not interact with Tac e -T 516 -578 (Fig. 6C, lanes 3 and 4). Immunoblotting of the Tac e -T 516 -578 IP fraction with anti-Gaa1 antibodies revealed that the chimeric protein does not interact with Gaa1 either (data not shown), indicating that ER retention of the fusion protein is unlikely to be due to its complexation with GPIT subunits. We conclude that the TM domain and cytoplasmic tail of PIG-T can confer ER localization to a type I membrane protein destined for the cell surface. DISCUSSION In this paper, we explore the mechanisms by which the human GPI transamidase subunits Gaa1 and PIG-T are retained in the ER. ER localization of these two membrane proteins as well as that of the three other known GPIT subunits (Gpi8, PIG-S, and PIG-U) must occur at two levels; the first involves ER retention of the individual proteins prior to assembly into the GPIT complex, and the second involves retention of the complex itself. Our studies focus on the first issue. We show that there is no primary sequence element in Gaa1 that can act as a dominant ER localization signal. In contrast, we show that the single TM domain in PIG-T is sufficient to ER-localize the ectodomain of Tac, a protein that is normally exported to the plasma membrane. Thus, PIG-T contains ER localization information, whereas Gaa1 does not.
In order to identify ER localization information within Gaa1 A, schematic and localization of C-terminally V5 or FLAG epitope-tagged human PIG-T, T 545 , and T 514 constructs. HeLa cells were fixed 24 h after transfection, permeabilized or not, and stained with mouse monoclonal anti-FLAG or anti-V5 antibodies and Alexa Fluor 568-conjugated anti-mouse Ig. Scale bar, 10 m. No fluorescence was detected in cells that had not been permeabilized. B, co-immunoprecipitation of Gpi8 with PIG-T constructs. HeLa cells were transiently transfected with expression vectors encoding PIG-T, T 545 , and T 514 . Digitonin extracts of transfected cells were subjected to immunoprecipitation with anti-V5 or anti-FLAG M2 affinity beads. 10% of the immunoprecipitated samples (IP) and 2% of the nonbound material (postimmunoprecipitation; Post-IP) were resolved by SDS-PAGE and immunoblotted with anti-V5, anti-FLAG, or anti-Gpi8 antibodies. C, analysis of T 545 and T 514 constructs expressed in HeLa cells. The proteins were immunoprecipitated from cell lysates and medium fractions, and immunoprecipitates were analyzed by SDS-PAGE and immunoblotting. The immunoprecipitates corresponding to ϳ10 and ϳ4% (lane 2) or 5% (lane 4) of the immunoprecipitated material from the cell lysates and from the medium, respectively, were analyzed by SDS-PAGE-immunoblotting. Immunoblotting was carried out with anti-V5 (lanes 1 and 2) or anti-FLAG (lanes 3 and 4) antibodies. and PIG-T, it was necessary to generate and study variants of these proteins that do not interact with other GPIT subunits. Thus, our information on Gaa1 relied on analyses of Gaa1 282 , an ER-localized truncated construct that displays the same membrane topology and N-glycosylation as Gaa1 but does not interact with other GPIT subunits. Similarly, our results concerning the ER retention of PIG-T rely on T 514 and Tac e -T 516 -578 , PIG-Tderived proteins that do not interact with other GPIT subunits. It is important to note that our conclusions concerning the absence (for Gaa1) or presence (for PIG-T) of ER localization signals rely ultimately on the ability to observe protein export from the ER. We view export of a fusion protein (e.g. Gaa1 55 -ST 30 -402 (Fig. 3)) or of a tagged, albeit truncated, construct (e.g. T 514 (Fig. 5)) as evidence that the protein is correctly folded and that the ER retention of simpler variants of these constructs (e.g. Gaa1 282 or T 545 ) is due to one of the retention mechanisms described in the Introduction rather than a result of the protein being targeted for degradation. In the case of the intact Gaa1 molecules described in Fig. 1, we explicitly assess and rule out the latter possibility by measuring protein turnover. The approaches discussed here con-trast with those used to assess sorting signals for proteins located in post-ER compartments; in these situations, protein folding is less of an issue, since the protein reporters being tested have typically already negotiated ER quality control mechanisms en route to their final destination.
Our results contribute to the ongoing debate about the nature of signal-mediated protein transport in the secretory pathway. The controversy concerning whether protein transport in the secretory pathway occurs by a passive ("bulk flow") mechanism, active (receptor-mediated) mechanism, or both raises similar questions about retention mechanisms of ER-resident proteins. The vectorial flow of proteins from the ER to the plasma membrane would suggest that active mechanisms are required to retain ER-resident proteins, whereas selective sorting of proteins for export would imply that the absence of positive ER export signals and an inefficient bulk flow pathway would be sufficient to retain ER residents (45).
Recent studies addressing the importance of sorting signals for COPII-dependent export of proteins from the ER provide support for the idea of passive ER retention of membrane proteins that lack functional ER exit signals. For example, the exit of wild-type cystic fibrosis transmembrane conductance regulator from the ER is blocked by mutation of a diacidic ER exit motif, which is required for protein binding to COPII vesicle coats (46). Similar results were obtained when a diacidic motif in Gap1p, the general amino acid permease, was mutated to other amino acids (45). Also, mutations of a novel export motif, FNX 2 LLX 3 L, in the C terminus of the human pituitary vasopressin V1b/V3 receptor abolished export of the protein (47). Thus, the lack of positive exit codes or retention signals appears to be one way to retain membrane proteins in the ER. Our data on the ER localization of Gaa1 are consistent with this idea.
The molecular basis for the TM-mediated ER retention of PIG-T is unclear. Like other ER-retained membrane proteins (15,16), PIG-T diffuses freely in the plane of the ER membrane (green fluorescent protein-tagged PIG-T has a lateral diffusion coefficient (D) of ϳ1.5 ϫ 10 Ϫ9 cm 2 ⅐s Ϫ1 and a mobile fraction of ϳ77%), 2 suggesting that it is not a stable component of a large matrix or supercomplex. The length and hydrophobicity of TM segments has been proposed to provide sorting information for some tail-anchored and integral membrane proteins (20, 22, 48 -52), and it is conceivable that something similar holds true for PIG-T. Protein sequence alignment reveals that the Nterminal portion of the PIG-T TM domain ( 518 FSMPYNVI-CLTCTVVAVCYGSFY 544 ; conserved amino acids are shown in boldface type) is highly conserved within the family of PIG-T orthologs; however, since the sequence does not possess significant homology with other membrane proteins in the data bases, it is difficult to make generalizations. Further investigations and other approaches will be needed to determine the mechanism involved in the TM-mediated ER retention of PIG-T.