The α-Helical Region in p24γ2 Subunit of p24 Protein Cargo Receptor Is Pivotal for the Recognition and Transport of Glycosylphosphatidylinositol-anchored Proteins*

Background: Glycosylphosphatidylinositol-anchored proteins (GPI-APs) depend on p24 cargo receptors for intracellular trafficking. Results: The luminal α-helical but not GOLD domain of p24γ2 was required for efficient GPI-AP transport. Conclusion: A p24 complex containing p24γ2 recognizes GPI-AP cargo using the α-helical region of p24γ2. Significance: The α-helical regions of p24 proteins are involved in complex formation and cargo recognition. Glycosylphosphatidylinositol-anchored proteins (GPI-APs) are group of proteins that depend on p24 cargo receptors for their transport from the endoplasmic reticulum to the Golgi apparatus. The GPI anchor is expected to act as a sorting and transport signal, but so far little is known about the recognition mechanism. In the present study we investigate the GPI-AP transport in cell knockdown of p24γ, the most diverse p24 subfamily. Knockdown of p24γ2 but not of other p24γ family members impaired the transport of a reporter GPI-AP. Restoration of the knockdown-induced phenotype using chimeric constructs between p24γ2 and the related p24γ1 further implied a role of the α-helical region of p24γ2 but not its GOLD domain in the specific binding of GPI-APs. We conclude that motifs in the membrane-adjacent α-helical region of p24γ2 are involved in recognition of GPI-APs and are consequently responsible for the incorporation of these proteins into coat protein complex II-coated transport vesicles.

Specific protein trafficking is essential to maintain the unique functions of different cellular compartments in eukaryotic cells. Coat protein complex I (COPI) 4 -and COPII-coated vesicles are implicated in the bidirectional transport between the endoplasmic reticulum (ER) and the Golgi apparatus. COPII assembles at the ER exit sites, and generated vesicles are transported from the ER toward the ER-Golgi intermediate compartment (ERGIC) (1,2). The role of COPI vesicles includes the retrieval of proteins from the Golgi to the ER (3). Sec24 proteins, components of the multisubunit COPII complex, harbor several motifs for cargo binding. In contrast to transmembrane proteins, which are capable of interacting directly with the Sec24 subunits, luminal proteins depend on cargo receptors for their specific sorting into COPII vesicles.
A distinct set of cargo receptors, recognizing either peptide and/or carbohydrate signals, have been described (4). Among them is a family of type I transmembrane proteins of approximately 24 kDa, termed p24 proteins, which can be divided into four subfamilies (p24␣, ␤, ␥, and ␦) (5). Because of the expansion of two of the subfamilies (p24␣ and p24␥) containing three and five members, respectively, 10 p24 proteins are present in most vertebrates. However, tissue expression studies in mice revealed restricted expression patterns for p24␣ 1 and p24␥ 5 , and due to an in-frame stop codon, p24␣ 1 is a pseudogene in humans (6). Furthermore, an alignment of p24␣ 2 and p24␣ 3 excluding the signal sequences reveals 96% similarity of the two proteins. The largest variety is hence provided by the p24␥ subfamily containing four different and widely expressed proteins, whereas the other subfamilies exhibit no or very limited diversity.
All p24 proteins share a similar domain organization, consisting of a GOLD domain and an ␣-helical region, followed by a transmembrane region and a short C-terminal tail. Whereas Jenne et al. (7), propose the occurrence primarily of monomers and dimers, several studies describe the building of heterotetrameric or larger complexes involving one member from each of four subfamilies (8 -11).
Glycosylphosphatidylinositol (GPI) anchoring is a posttranslational modification, exerted to a wide variety of proteins in eukaryotic cells. Early steps of the GPI anchor assembly occur on the cytosolic side of the ER, and following flipping of an intermediate product, the synthesis is completed in the ER lumen. After this multistage assembly process, the GPI anchor is transferred en bloc to targeted proteins and after further remodeling reactions, GPI-anchored proteins (GPI-APs) are transported via the Golgi to the cell surface. Several studies revealed a role of p24 proteins in the sorting of GPI-APs into COPII vesicles in yeast (12)(13)(14)(15). The impaired transport of GPIanchored, but not other cargo molecules upon knockdown of p24␦ 1 or p24␤ 1 in mammalian cells, supported these findings in yeast (16,17). Fujita et al. reported that p24␣ 2, p24␤ 1, p24␥ 2 , and p24␦ 1 are associated with GPI-APs in the ER, supporting a model of heterotetrameric or larger complex of p24 cargo receptors (10). They further showed that two GPI anchor remodeling reactions in the ER, occurring after the transfer to proteins, are crucial for the interaction with these p24 proteins and efficient sorting into the ER exit sites. Hence, the GPI anchor is expected to act as a sorting and transport signal in the ER although little is known so far about the recognition mechanism.
Due to the largest variability, it is likely that the respective p24␥ subunit determines the cargo specificity in the receptor complexes. Here, we demonstrate that knockdown of p24␥ 2 but not knockdown of other p24␥ subfamily members, results in delayed GPI-AP transport. Using chimeric and mutant constructs, we define the region required for GPI anchor recognition and further confirm the results by a binding assay.

EXPERIMENTAL PROCEDURES
Cells-FCAT5 is a cell line obtained as a result of three separate stable transfections of CHO-K1 cells. In a first step 3B2A cells were established by stably transfecting CHO-K1 cells with pME-NEO plasmid expressing DAF and CD59, human GPI-APs, under the control of an SR␣ promoter, and selecting by cell sorting a clone expressing DAF and CD59 at high levels (18). 3B2A cells were stably transfected with pTRE2-puro-VSVG ex -FF-mEGFP-GPI in conjunction with pUHrT62-1, an expression plasmid for reverse tetracycline-controlled transactivators (16,19) to obtain FF8 cells. Finally, for use in a retrovirus system, FF8 cells were stably transfected with a plasmid, expressing mouse CAT1, a receptor for ecotropic retroviruses to generate FCAT5 cells. FCAT5 cells stably expressing p24␥ 2 shRNA or p24␥ 2 shRNA in combination with various restoration constructs were established by infection with a retrovirus produced in PLAT-E packaging cells (a gift from T. Kitamura, University of Tokyo, Tokyo, Japan), followed by selection with 7 g/ml blasticidin (BSD). FCAT5 cells and their derivatives were maintained in Ham's F-12 medium (Sigma-Aldrich) supplemented with 10% FCS, 600 g/ml G418, 800 g/ml hygromycin, 6 g/ml puromycin, and if necessary 7 g/ml BSD.
Reagents and Antibodies-Lipofectamine 2000 and Lipofectamine RNAiMAX were purchased from Invitrogen. Rabbit anti-p24␥ 2 antibody was provided by H. Hauri and H. Farhan (University of Basel, Basel, Switzerland). Rabbit anti-p24␣ 2 , rabbit anti-p24␤ 1 , rabbit anti-p24␥ 1 , rabbit anti-p24␥ 3 , and guinea pig anti-p24␥ 4 antibodies were generous gifts from F. Wieland and A. Herrmann (Heidelberg University, Heidelberg, Germany). Anti-p24␦ 1 antibody was obtained by immunizing rabbit with the peptide LRRFFKAKKLIE followed by affinity purification on a peptide column. These are antibodies against the C-terminal tails of human p24 proteins, but they recognized mouse p24 proteins as well. Mouse anti-FLAG (clone M2), rabbit anti-ERGIC-53 and mouse anti-␣ tubulin (clone DM1A) were purchased from Sigma-Aldrich and mouse anti-GFP from Roche Applied Bioscience. Horseradish peroxidase (HRP)-conjugated sheep anti-mouse IgG and donkey anti-rabbit IgG were from GE Healthcare. HRP-conjugated goat anti-guinea pig IgG H&L was purchased from Abcam and phycoerythrin (PE)-conjugated goat anti-mouse IgG from BD.
For generating full-length restoration constructs mouse p24␥ 2 cDNA was cloned into the EcoRI site of pLIB2. The introduction of an MluI site and mutation of the EcoRI site in the forward primer allowed us to retain the EcoRI as a single cutting site downstream of the insert. hU6 promoter, knockdown (307), and BSD resistance gene sequences were subsequently integrated between the EcoRI and XhoI sites. pLIB2-p24␥ 1 -307-BSD was cloned in a similar way to pLIB2-p24␥ 2 -307-BSD after amplifying full-length p24␥ 1 from a mouse testis cDNA library. To ensure the recognition of p24␥ 1 by rabbit anti-p24␥ 2 antibody, the C-terminal part of the protein was subsequently changed from RFFHDKRPVPT (p24␥ 1 ) to SLFEDKRKSRT (p24␥ 2 ) to have an epitope recognized by the anti-p24␥ 2 antibody. Chimeric constructs were prepared by using pLIB2-p24␥ 2 -307-BSD and pLIB2-p24␥ 1 -307-BSD as PCR templates. See Table 1 for a list of primers used.
Cells were transfected two times (48-h interval) according to the manufacturer's instructions using Lipofectamine RNAiMAX. Knockdown efficiency was confirmed by quantitative PCR and immunoblotting.
The quantitative PCR using SYBR Premix Ex TaqII (Tli RNaseH Plus), Bulk (TaKaRa) was performed by Thermal Cycle Dice Real Time System according to the manufacturer's instructions. The RNA expression level was normalized to HPRT and the relative expression calculated using Ϫ⌬⌬C T .
a Indicated reverse primers were used to clone m-p24␥ 1 , chimera 1, and chimera 2 containing the C terminus of p24␥ 1 . To allow detection of all chimeric constructs, the Cterminal part was subsequently changed to p24␥ 2 using the primer pair listed under "Epitope." Transport Assay of the Reporter GPI-AP-FCAT5 cells harboring Tet-inducible VSVGts-FLAG-GFP-GPI (VFG-GPI) were incubated at 40°C in the presence of 1 g/ml doxycycline. Due to misfolding of the VSVGts domain at 40°C, the reporter is retained in the ER. After 24 h of incubation, the cultures were harvested with trypsin-EDTA solution (Sigma-Aldrich), and transferred to 32°C to allow proper folding and transport. Samples collected after different incubation times were stained with anti-FLAG antibody, followed by PE-conjugated goat anti-mouse secondary antibody. The surface amount of VFG-GPI reporter protein was analyzed using FACSCanto II (BD Biosciences) (19).
GPI-AP Immunoprecipitation and Immunoblotting-FCAT5 cells stably transfected with p24␥ 2 siRNA in combination with different restoration constructs were cultured for 24 h at 40°C in the presence of 1 g/ml doxycycline (4 ϫ 10 7 cells). The cultures were harvested with trypsin-EDTA solution and transferred to 32°C for 20 min. The cells were washed once in PBS and lysed for 1 h at 4°C in buffer containing 20 mM MES/HEPES (pH 7.4), 100 mM NaCl, 1% digitonin, 1ϫ protease inhibitor mixture  (Roche Applied Science). After centrifugation (20,000 ϫ g, 15 min, 4°C) the soluble lysate was incubated with anti-FLAG M2-agarose beads (Sigma-Aldrich) for 2 h at 4°C, followed by four washing steps (20 mM MES/HEPES (pH 7.4), 100 mM NaCl, 0.5% digitonin). Finally, the beads were boiled in SDS-PAGE sample buffer and analyzed by immunoblotting.

Transient Knockdown of p24␥ 2 but Not of Other p24␥s
Impairs GPI-AP Transport-To investigate the role of p24␥ proteins in the transport of GPI-AP, we used FCAT5 cells, which express a Tet-inducible, temperature-sensitive GPI reporter protein, VFG-GPI. After expression and accumulation of VFG-GPI in the ER at 40°C, cultures were transferred to 32°C to allow protein folding and transport. The kinetics of GPI-AP transport to the cell surface can be determined by measuring the amount of surface-associated VFG-GPI at different time points by flow cytometry using anti-FLAG antibody.
We transiently knocked down the p24␥ subfamily members in FCAT5 cells using two different siRNAs. After transfecting the cells two times at a 48-h interval, the mRNA levels of p24␥ 1ϳ4 were decreased to 3-16% of those in control cells (Fig.  1A). Substantial decreases in protein levels were also detected (Fig. 1B). Impaired GPI-AP transport could be observed only upon knockdown of p24␥ 2 (Fig. 1, C and D).
The basic mRNA expression level of p24␥ 5 was substantially lower compared with the other p24␥ subfamily members. This might explain why in p24␥ knockdown cells a relative decrease to only 36 and 45%, respectively, could be observed (Fig. 1A). Under the conditions of the reduction of the mRNA to Ͻ50%, we could not observe any transport delay in our assay (Fig. 1D). Although a role of p24␥ 5 in GPI-AP transport cannot certainly be excluded under these circumstances, it is rather unlikely. The restricted tissue expression of p24␥ 5 (6) further indicates a very specific function of this protein and hence questions a role in the ubiquitous GPI-AP transport. These data provide evidence that p24␥ 2 may be a key subunit in the cargo receptor for the specific recognition of GPI in transport of GPI-APs. This finding is also supported by our previous result, where in a mass spectrometry analysis p24␥ 2 was the only member of the p24␥ family found to be associated with VFG-GPI (10).
Localization of Regions in p24␥ 2 for Recognition of GPI Anchor-Among the p24␥ proteins, p24␥ 1 is most closely related to p24␥ 2 , revealing 53.6% sequence identity (Fig. 3). Despite this similarity, only the expression of p24␥ 2 but not p24␥ 1 was able to restore the transport phenotype in stable p24␥ 2 knockdown cells (Fig. 4). This finding further points toward specific roles of p24␥ 1 and ␥ 2 proteins with no or only limited redundancy.
Because the p24␥ 1 did not influence GPI-AP transport kinetics (Figs. 1C and 4), we next tested p24␥ 1 /p24␥ 2 chimeric constructs for their ability to restore the transport phenotype in stable p24␥ 2 knockdown cell line to determine which region in p24␥ 2 is functionally important (Fig. 5A). To assure the detection by the p24␥ 2 antibody that recognizes the cytoplasmic domain (C-term), each construct was designed to contain the C-term of p24␥ 2 . All chimeric constructs could be detected by immunoblotting (Fig. 5B).
Using chimeras 1-4, we first studied the role of the GOLD domain, the ␣-helical region, and the transmembrane domain in GPI-AP transport. Chimera 1 consisted of a p24␥ 2 GOLD domain and ␣-helical region, but a p24␥ 1 transmembrane region. Chimera 2 contained a p24␥ 2 GOLD domain but the p24␥ 1 ␣-helical region and transmembrane domain. Chimeras 3 and 4 were the opposite of chimera 2 and chimera 1, respectively (Fig. 5A). Chimeras 1 and 3 were able to restore the transport delay. At time point 40 min, their level of VFG-GPI on the surface reached 79.1% and 97.0% of vector-transfected FCAT5 cells, respectively (Fig. 5C). However, expression of chimeras 2 and 4 was not sufficient to restore the transport delay (56.5% and 50.1%, respectively). This implies that the ␣-helical region rather than the GOLD domain is involved in GPI-AP recognition and transport.
The ␣-helical regions of p24␥ 1 and p24␥ 2 have only 27.6% sequence identity in amino acids 127-153, whereas the remaining ␣-helical region (amino acids 154 -196) shows a sequence identity of 65.1% (Fig. 5D). Four chimeric constructs (chimeras 5-8) were designed to test the role of these two parts of the ␣-helical region in GPI-AP transport (Fig. 5A). Chimeras 5 and 8 were able to restore the transport phenotype (92.2% and 83.1% respectively), whereas chimeras 6 and 7 were not (42.3% and 41.6%, respectively) (Fig. 5C). The result indicates that the critical region for GPI-AP transport is not localized in amino acids 127-153 but in amino acids 154 -196. The ␣-helical region spanning amino acids 154 -196 was further divided into two parts (154 -174 and 175-196), and their influence was tested with the expression of chimeras 9 -12 (Fig. 5A). All cell lines showed an intermediate phenotype (67.4%, 66.5%, 70.2%, and 61.1%, respectively), neither reaching the level of p24␥ 2 knockdown cells restored by wild-type p24␥ 2 nor showing a reduction comparable with nonrestored p24␥ 2 knockdown cells. This suggests that motifs from both regions are required to restore the transport delay fully. When the p24␥ 2 part spanning amino acids 175-196 was extended 2 or 4 more amino acids to amino acids 173-196 (chimera 13) or 171-196 (chimera 14), transport was fully recovered (158.5% and 118.3%, respectively) (Fig. 5C). We therefore conclude that the IQ motif at positions 173/174 together with a motif present within amino acids 175-196 are involved in GPI-AP binding. Although this combination was shown to be very effective, chimera 15, full-length p24␥ 2 with the GHIQ sequence at positions 171-174, changed to the respective amino acid in p24␥ 1 (IQML) (Fig. 5, A and D) was still able to restore the knockdown-induced phenotype (135.5%) (Fig. 5C). Hence, although the IQ motif at 173/174 together with amino acid 175-196 of p24␥ 2 is sufficient to allow normal GPI-AP transport, the motif seems to be negligible in other sequence combinations.
Transport Phenotype of Chimeric Constructs Correlates with GPI-AP Binding-To correlate the result of the transport assay with binding ability of various cargo receptor complex, we next studied the interaction of chimeric constructs 7, 8, 12, and 13 with the reporter GPI-AP by immunoprecipitation. Chimeric constructs 8 and 13, which were very efficient in restoring the knockdown-induced phenotype, were tested in combination with their counterpart constructs (chimeras 7 and 12), which showed delayed transport. . The ␣-helical region of p24␥ 2 is relevant for GPI-AP transport. A, schematic overview shows used p24␥ 2 /p24␥ 1 chimeric constructs. B, stable p24␥ 2 knockdown FCAT5 cells expressing chimeric constructs were lysed and the protein level analyzed by immunoblotting. Note that a 25-kDa band seen in vector lane (left end) is endogenous hamster p24␥ 2 , which was significantly reduced with p24␦ 1 shRNA and was more greatly reduced with p24␥ 2 shRNA. C, stable p24␥ 2 knockdown FCAT5 cells expressing chimeric constructs were prepared as described under "Experimental Procedures." The expression of the GPI reporter (VFG-GPI) was induced by the addition of 1 g/ml doxycycline, and the cultures were incubated at 40°C. After 24 h the cells were harvested and transferred to 32°C to allow transport. Samples were collected after 0 and 40 min and stained with anti-FLAG antibody, followed by PE-conjugated goat anti-mouse secondary antibody. The surface amount of VFG-GPI reporter protein was analyzed by flow cytometry. The graph represents the quantified data. Error bars, S.E. D, ␣-helical region of p24␥ 1 and p24␥ 2 were aligned using the BLAST tool of NCBI. Similar amino acids are indicated by ϩ. Matching amino acids are indicated by the amino acid letter code. Amino acid positions mentioned in the text are indicated. They correspond to the position in mouse p24␥ 2 (NP_083152). The IQ motif of p24␥ 2 (chimera 13) as well as the IQML motif of p24␥ 1 (chimera 15) are framed.
After accumulating VFP-GPI in the ER at 40°C, the temperature was shifted to 32°C for 20 min to allow protein folding, receptor binding, and transport initiation, and then lysed in buffer containing 1% digitonin. VFG-GPI was collected with anti-FLAG beads and the amount of bound chimeric constructs analyzed by immunoblotting. Chimeras 8 and 13 were specifically precipitated whereas chimeras 7 and 12 showed only mild binding to the VFP-GPI (Fig. 6). It should be noted that the remaining level of endogenous hamster p24␥ 2 was efficiently precipitated with VFG-GPI in all cell lines. The immunoprecipitation result hence supports the data observed in the transport assay.

DISCUSSION
Ten different p24 proteins, involved in the specific transport between the ER and the Golgi, are expressed in vertebrates. They can be divided into four subfamilies, p24␣, ␤, ␥, and ␦. Whether p24 proteins exist mainly as monomers, dimeric, tetrameric, or larger complexes is still debated. Jenne et al. showed that the state of oligomerization and concentration of p24 proteins vary, depending on the cellular localization (7). Several studies confirmed the occurrence of heterotetrameric complexes involving one member of each subfamily (8 -11). In contrast to the limited diversity of the p24␣, ␤, and ␦ subfamilies, five different p24␥ proteins are expressed in vertebrates. It is therefore likely that in heterotetrameric receptor complexes, the respective ␥ subunit determines the cargo specificity. In this study we also showed that levels of p24 are interdependent. The inverse correlation among the levels of p24␥ proteins supports the competition for the common p24␣, ␤, and ␦ interaction partners. Consistent with that, we additionally detected elevated kinetics of GPI-AP transport upon knockdown of p24␥ 1 , ␥ 3 , ␥ 4 , and ␥ 5 , which further strengthens the competition among these proteins (Fig. 1D). Redundancy among p24␣ and p24␥ proteins had been described recently in yeast (11). Because the GPI-AP transport was impaired in p24␥ 2 knockdown cells and could not be compensated by the slight overexpression of other p24␥ proteins (Fig. 1D), p24␥ proteins seem, however, to have a specific function in the GPI-AP transport in mammalian cells.
Until now, impaired transport of GPI-APs had only been reported upon knockdown of p24␤ 1 and p24␦ 1 (12,16,17). Because these knockdowns lead to the instability and reduction of all p24 proteins, no conclusions about the specificity could be drawn. In the present study we could, however, show that knockdown of p24␥ 2 specifically affects the GPI-AP transport without impairment of basic p24 complex formation.
A major conclusion of this study is that the ␣-helical region of p24␥ 2 is involved in the binding of GPI. The ␣-helical region had previously been implicated in oligomerization and stabilization of p24 receptor complexes and the more prominent GOLD domain was suggested as a cargo binding site. The GOLD domains are not only implicated in protein-protein interactions (20), but are also often observed in sugar-and lipid-binding proteins (21). It therefore seemed feasible that the GOLD domain is capable of interacting with a variety of cargos including sugar-lipid motifs like GPI. We however found that the GOLD domains of p24␥ 1 and p24␥ 2 are interchangeable, as chimera 3, although containing the GOLD domain of p24␥ 1 , was still able to restore the p24␥ 2 knockdown-induced GPI-AP transport delay (Fig. 5A). The ␣-helical regions of p24 proteins in complexes interact and form coiled-coil structures. It is therefore possible that in this region protein-overlapping binding motifs arise. Changes in amino acids directly involved in cargo binding as well as changes in residues important for proper p24 protein interaction could thereby affect efficient cargo binding and transport. As all of our chimeric constructs were stably expressed, we assume that the basic complex assembly was efficient. Small conformational changes, however, might have led to masking or unfavorable positioning of the binding motif.
We were able to narrow down the binding site to the IQ motif at positions 173/174 together with a motif present within amino acids 175-196. We did not, however, conclude that the IQ motif is involved directly in interaction with GPI-AP. Chimera 15, full-length p24␥ 2 with the GHIQ motif at positions 171-174 changed to the respective amino acid in p24␥ 1 (IQML), was still able to restore the knockdown-induced phenotype. We therefore propose a model that amino acids 175-196 of p24␥ 2 are indispensable for proper GPI-AP transport and presumably involved directly in the binding process. The IQ motif at 173/174, however, does not seem to be involved directly in the interaction but rather has a role in increasing the affinity of binding or favoring proper folding or binding motif accessibility. Also, amino acids 154 -170 seem to be able to undertake such a supportive function, and it appears that either amino acids 154 -170 or IQ motif at 173/174 in combination with amino acids 175-196 is sufficient to allow efficient GPI-AP transport.
Two remodeling reactions in GPI occur in the ER after its attachment to proteins. First, the acyl chain linked to inositol is removed by PGAP1 followed by the removal of a side chain ethanolamine phosphate attached to the second mannose by PGAP5 (22,23). We showed previously that these reactions are crucial for association with p24 proteins and the recruitment of GPI-APs into ER exit sites (23). As in our chimeric constructs the transmembrane regions of p24␥ 1 and p24␥ 2 were interchangeable (Fig. 5), we conclude that the lipid part of GPI is not FIGURE 6. Binding of chimeric constructs to the GPI reporter. Stable p24␥ 2 knockdown FCAT5 cells expressing chimeric constructs were prepared as described under "Experimental Procedures." The expression of the GPI reporter (VFG-GPI) was induced by the addition of 1 g/ml doxycycline, and the cultures were incubated at 40°C. After 24 h the cells were harvested and transferred to 32°C to allow transport. Samples were collected after 20 min and lysed. VFG-GPI was collected with anti-FLAG beads and the amount of bound chimeric constructs analyzed by immunoblotting. Blots are representative of at least three experiments.
involved directly in p24 binding. The removal of the acyl chain from inositol ring is nevertheless a crucial step for the interaction with p24 proteins. We expect that cleavage by PGAP1 changes the orientation of inositol to the membrane and structure of the glycan plus inositol and could consequently contribute to the formation of the correct binding site for the p24 complex. The same might be the case for the removal of the ethanolamine phosphate. It is also likely that a combination of both remodeling reactions is required to generate the final binding site. Our findings provide evidence that p24␥ proteins show only a limited redundancy and that p24␥ 2 is specifically important for GPI-AP transport in mammalian cells.