Our laboratory was interested in characterizing the molecular composition of non-caveolar lipid rafts. Thus, we generated monoclonal antibodies to lipid raft proteins of human myelomonocytic cells. Two of these proteins, KE04p and C8orf2, were found to be highly enriched in the detergent-insoluble, buoyant fraction of sucrose gradients in a cholesterol-dependent manner. They contain an evolutionarily conserved domain placing them in the prohibitin family of proteins. In contrast to other family members, these two proteins localized to the ER. Furthermore, the extreme N-termini of KE04p and C8orf2 were found to be sufficient for heterologous targeting of GFP to the ER in the absence of classical ER retrieval motifs. We also demonstrate that all prohibitin family members rely on sequences in their extreme N-termini for their distinctive subcellular distributions including the mitochondria, plasma membrane and Golgi vesicles. Owing to their subcellular localization and their presence in lipid rafts, we have named KE04p and C8orf2, ER lipid raft protein (erlin)-1 and erlin-2, respectively. Interestingly, the ER contains relatively low levels of cholesterol and sphingolipids compared with other organelles. Thus, our data support the existence of lipid-raft-like domains within the membranes of the ER.
It has long been appreciated that the plasma membrane (PM) is not a random distribution of lipids and proteins. Rather, it contains discrete domains responsible for mediating specific cellular functions (Harder and Engelhardt, 2004). Lipid rafts, an example of a class of such domains, are PM microdomains enriched in glycosphingolipids and cholesterol (Brown and London, 1998). Owing to their distinct morphology, the first lipid raft domains to be identified were caveolae (Palade, 1953; Yamada, 1955). Caveolae are 50-100 nm flask-shaped invaginations of the plasma membrane containing the cholesterol-binding protein caveolin-1 (Holthuis et al., 2001). Biochemically, caveolae are characterized by their insolubility in non-ionic detergents at low temperatures (such as solutions of Triton X-100 at 4°C) and their buoyancy in sucrose gradients (Brown and Rose, 1992). It is by these biochemical criteria that non-caveolin-containing lipid rafts were isolated from and identified in various hematopoietic cells (Fra et al., 1994; Robbins et al., 1995).
As a result of their lack of visibility as morphological structures, non-caveolar lipid rafts are notoriously difficult to characterize. The most commonly used method for studying these structures has been biochemical isolation of the buoyant, insoluble material from cells lysed in non-ionic detergent and subjected to sucrose gradient centrifugation, a feature that has earned these structures the biochemically descriptive name `detergent-resistant membranes' or DRMs (Brown and Rose, 1992). Thus, protein markers that co-fractionate with these sphingolipid-rich domains have been instrumental in the characterization of the size, composition, and functions of these membrane microdomains, through indirect means in both biochemical and microscopy studies (Rietveld and Simons, 1998; Subczynski and Kusumi, 2003).
To characterize non-caveolar lipid rafts in cells of hematopoietic origin, our laboratory has undertaken a combined immunologically based, proteomics approach. To this end, low-density, Triton X-100-insoluble lipid raft preparations were isolated from differentiated U937 and HL60 cells and inoculated into mice to produce a panel of monoclonal antibodies. Herein, we focus on the identification and characterization of two novel proteins enriched in lipid raft fractions from myelomonocytic cells: KE04p and its highly related family member, chromosome 8 open reading frame 2 (C8orf2). Both proteins fall within the growing family of prohibitin domain-containing (PHB) proteins, which includes the prohibitins, the stomatins and the flotillins (also known as reggies). Like other PHB family members, both KE04p and C8orf2 are found in the detergent-insoluble, low-density fractions of cell lysates of many different cell types. Neither of these proteins localized to the PM; instead, they were observed to localize to the endoplasmic reticulum (ER). This was somewhat surprising considering that the ER contains extremely low levels of sphingolipids and cholesterol (Holthuis et al., 2001; Prinz, 2002; van Meer and Lisman, 2002). Collectively, our data provide evidence for the existence of lipid-raft-like domains within the membranes of the ER. Because of their subcellular localization and their enrichment in lipid rafts we propose the names erlin-1 and -2 (for endoplasmic reticulum lipid raft protein) for KE04p and C8orf2, respectively.
Identification of known and novel lipid-raft-resident proteins
We were interested in determining the molecular composition of non-caveolar lipid rafts in cells of hematopoeitic origin. To this end we used an immunologically based, proteomics approach. Briefly, U937 and HL60 cells were differentiated along the macrophage-like, and granulocytic lineage, using tetradecanoyl phorbyl myristate acetate (TPA) and dimethylsulphoxide (DMSO), respectively. Detergent resistant membrane (DRM) isolates were prepared from these cells and used to immunize mice for the generation of a panel of monoclonal antibodies.
Out of 2880 original hybridomas cultured, approximately one-third (>900) of the clones survived. After screening by western blotting, we identified 32 antibodies that recognized antigens within DRM fractions of these cells. A number of hybridoma clones reacted with antigens of multiple molecular weights (data not shown), indicating a possible recognition of common epitopes or modifications commonly found on a diverse range of proteins, such as phosphorylation or ubiquitylation. For these antibodies we have not yet identified the specific antigens that they recognize. Most of the hybridoma antibodies, however, recognized antigens represented by a discrete band, indicating a single protein species. Of these, we were able to identify the specific proteins recognized by 26 of these monoclonals using mass spectrometry of both 1D and 2D gel samples as well as cDNA expression cloning. Confirmation of the identified antigens was accomplished by cloning the candidate proteins into a mammalian expression system along with epitope tags and western blotting with the hybridoma supernatants. Through this approach we were able to identify both known and predicted protein residents of DRMs (Table 1).
The plasma membrane protein CD14 represented the antigen recognized by the largest number of hybridoma clones (Table 1). This was to be expected based on its abundance in macrophages and its mode of membrane attachment through a glycophosphatidylinositol (GPI) anchor, a moiety known to be sufficient for DRM association in a number of different cell types (Brown et al., 1989; Dai et al., 2005; Lisanti et al., 1989). Other antigens identified include the plasma membrane protein EMMPRIN/CD147 and the dually acylated Src family kinase member, Lyn. Consistent with the cell lines used in this study, both EMMPRIN and Lyn have been reported to be upregulated upon monocyte differentiation (Major et al., 2002; Yamanashi et al., 1989). DRM association has also been reported previously for both of these proteins (Tang and Hemler, 2004; Young et al., 2003).
Other proteins identified were not restricted to the PM, but are known to populate the membranes of intracellular organelles, including both isoforms of the mitochondrial protein mitofilin and the ER protein, BAP31 (Breckenridge et al., 2003; Groenendyk and Michalak, 2005; Simmen et al., 2005). Although mitofilin was only recently reported to reside in DRMs (Mielenz et al., 2005), to our knowledge this is the first report describing a raft localization for BAP31.
Interestingly, two independent hybridoma clones, 10E6 and 7D3, identified the same novel antigen: the predicted protein product of the chromosome 10 open reading frame 69 gene (C10orf69), known as KE04p or stomatin-prohibitin-flotillin-HflC/K (SPFH) domain protein 1 (herein renamed erlin-1). While attempting to identify the antigen for the 10E6 and 7D3 antibodies from bands of silver-stained acrylamide gels, another highly related protein was identified by mass spectrometry from U937 DRM preparations, known as the protein product of chromosome 8 open reading frame 2 gene (C8orf2), or the SPFH domain protein 2 (renamed erlin-2). Both proteins have been identified in DRM fractions by mass spectrometry in previous reports (Blonder et al., 2005; Blonder et al., 2004b; Blonder et al., 2004c; Brown and London, 1998; Ledesma et al., 2003; Li and Prinz, 2004; Sprenger et al., 2004). We decided to focus on the molecular characterization of erlin-1 and erlin-2 since they had not been previously characterized in the literature.
Erlin-1 and erlin-2 are highly related members of the prohibitin family of proteins
The most striking feature of erlin-1 and -2 is their high degree of relatedness, sharing 83% sequence identity and 89% similarity at the amino acid level (Fig. 1A) (Ikegawa et al., 1999; Li et al., 2000). They belong to the prohibitin family of proteins by virtue of a conserved prohibitin-homology domain (PHB) of ∼160 amino acids (Fig. 1B). In fact, they are more closely related to prohibitin, the protein for which this family is named, than any of the other family members (Fig. 1C). Common properties shared by PHB family members, such as association with DRMs, the tendency to form oligomers, and their diverse subcellular distribution, guided our studies of erlin-1 and -2.
Erlin-1 can be distinguished from erlin-2 using monoclonal antibodies
Since erlin-1 and erlin-2 are highly related proteins, we wanted to ensure the specificity of the hybridoma clones for erlin-1 over erlin-2. To this end, we expressed full-length erlin-1 and hemagglutinin-tagged erlin-2 proteins in NIH 3T3 fibroblasts, which do not endogenously express the antigen(s) for either 10E6 or 7D3. Western blotting with either 10E6 or 7D3 revealed a single, ∼40 kDa band in DRM fractions from both the native U937 cell line as well as NIH 3T3 cells that ectopically express a full-length erlin-1 cDNA (3T3-erlin-1), (Fig. 2A,B). This observed molecular mass corresponds very well with the predicted mass of 38.925 kDa for the protein (Li et al., 2000). No immunoreactivity was observed with vector control 3T3 cells.
Western blotting with the monoclonal anti-hemagglutinin (HA) antibody (12CA5) revealed high levels of expression of erlin-2HA in retrovirally infected NIH 3T3 cells (3T3-erlin-2HA), (Fig. 2C). In addition to a band at ∼43 kDa, higher molecular mass bands were readily observed for 3T3-erlin-2HA extracts on western blots (Fig. 2C, arrow). We inferred that these bands represented higher-order oligomers of the protein. The propensity to form oligomers is a common feature shared by PHB family members (Morrow and Parton, 2005).
Since erlin-1 and erlin-2 were highly related, we wanted to determine the specific epitopes on erlin-1 recognized by the 10E6 and 7D3 antibodies. To this end, we generated GST fusions to various truncation mutants of the erlin-1 protein (Fig. 3A). Western blotting with the 10E6 and 7D3 antibodies revealed bands in extracts of bacteria expressing GST fusions containing residues 305-330 of erlin-1 (Fig. 3A-C). Appropriate expression of all constructs was confirmed by western blotting with anti-GST antibody (Fig. 3D). Specific recognition of this C-terminal region of erlin-1 by these antibodies is consistent with the lower degree of sequence identity with erlin-2 at this site compared with the rest of the protein (Fig. 2A). This accounts for the inability of these antibodies to detect erlin-2.
Erlin-1 and erlin-2 are enriched in the buoyant, detergent-insoluble fractions of sucrose gradients
Although both erlin-1 and erlin-2 were identified from the Triton-X-100-insoluble, low-density fraction of U937 and HL60 cells using mass spectrometry, some proteins identified from DRM fractions by mass spectrometry were not enriched as determined by western blotting such as the transferrin receptor and calnexin (Fig. 4B, supplementary material Fig. S1). Therefore we wanted to ensure that these proteins were indeed enriched in DRMs. First, to confirm that erlin-1 and erlin-2 were membrane proteins, we subjected U937, as well as NIH 3T3 cells stably expressing either erlin-1 (3T3-erlin-1) or HA-tagged erlin-2 (3T3-erlin-2HA) to hypotonic lysis. Equivalent proportions of the membrane (pellet) and cytosolic (supernatant) fractions were analyzed by western blotting. Both erlin-1 and erlin-2 were highly enriched in the membrane fractions of these cell lines (Fig. 4A). By contrast, minimal immunoreactivity was observed in lanes representing the cytosolic fractions of cells. The relative purity of membrane and soluble fractions was determined by western blotting for Src and pyruvate kinase, respectively. As expected, Src was enriched in cellular membranes (Kaplan et al., 1990) whereas the ubiquitous, cytosolic protein, pyruvate kinase, was highly enriched in cytosolic fractions (Fig. 4A).
Having demonstrated their presence in membranes, we wanted to confirm that these proteins were indeed enriched in DRMs. Schuck et al. (Schuck et al., 2003) show that DRM isolation using the non-ionic detergent Triton X-100 results in the isolation of the least number of proteins in the insoluble fractions compared with other detergents tested. Therefore, we based our criteria for DRM-resident proteins on this stringent method: insolubility in Triton X-100 at 4°C and buoyancy in sucrose gradients (Brown, 1992; Schuck et al., 2003). Western blotting of equal proportions (by volume) of low-density, insoluble and soluble fractions using 10E6 revealed that erlin-1 was highly enriched in the DRMs of both the U937 and 3T3-erlin-1 cell lines (Fig. 4B).
In accordance with the results for erlin-1, erlin-2 was also highly enriched in DRM fractions as assessed by western blotting (Fig. 4B). Based on the consistent behaviour of erlin-1 in both the native U937 and the ectopically expressing 3T3-erlin-1 cell line, it was concluded that the DRM enrichment observed for erlin-2 in 3T3-erlin-2HA cells represents its natural cellular distribution. Western blotting for the ubiquitous DRM marker protein, flotillin-1, and the transferrin receptor (CD71), a non-raft transmembrane protein indicated that our isolation procedure was specific for DRMs and not generic cellular membranes (Fig. 4B). Similar to other PHB family members, we were able to demonstrate the enrichment of both erlin-1 and erlin-2 in DRMs.
Erlin-1 and erlin-2 are refractory to plasma membrane cholesterol depletion
In addition to defining DRMs merely by detergent resistance, Schuck et al. (Schuck et al., 2003) recommend procedures aimed at disrupting these structures, such as cholesterol or sphingomyelin depletion, to confirm the presence of proteins in such microdomains. To assess the requirement of cholesterol for the ability of erlin-1 and -2 to associate with DRMs we treated live cells with the cholesterol sequestration agent, methyl-β-cyclodextrin (M-β-C) or mock-treated cells with PBS alone. Unexpectedly, M-β-C treatment had little effect on the solubility of either protein or on their buoyancy in sucrose gradients (Fig. 5). Similar to mock-treated cells, both erlin-1 and -2 remained predominantly in the insoluble fractions in both the U937 and the 3T3 cells lines (Fig. 5). Concordantly, there was no increase in the presence of either protein in the soluble fraction. Consistent with other reports, flotillin-1 was also observed to be relatively unaffected by M-β-C treatment (Fig. 5) (Gkantiragas et al., 2001).
We envisioned two possible scenarios for explaining these results: (1) Erlin-1 and -2 may occupy detergent-resistant, buoyant complexes that do not rely on cholesterol for their integrity or, (2) since M-β-C is only able to extract cholesterol from the exoplasmic leaflet of the plasma membrane, erlin-1 and -2 may not be at the cell surface but instead populate intracellular detergent-resistant membranes (Neufeld et al., 1996). To test these hypotheses we depleted cholesterol from internal membranes by subjecting cells to hypotonic lysis before treatment of pelleted membranes with M-β-C in order to disrupt intracellular DRMs.
Following resuspension of M-β-C-treated membranes in Triton X-100 and fractionation of continuous sucrose gradients we observed a dramatic shift in both buoyancy and solubility of erlin-1 in both the U937 and 3T3-erlin-1 cell lines from the low-density insoluble fraction to the high-density soluble fraction (Fig. 5A,B). Similar results were observed for the control DRM marker protein flotillin-1, which localizes to the cytoplasmic leaflet of the PM as well as intracellular compartments including the Golgi and endocytic system (Morrow and Parton, 2005). Once again this is consistent with a report that flotillin-1 was only susceptible to cholesterol sequestration once internal Golgi membranes were exposed to M-β-C (Gkantiragas et al., 2001). Mobility in the sucrose gradient as well as increased solubility in Triton X-100 was also observed for erlin-2 following M-β-C-treatment of exposed intracellular membranes (Fig. 5C). The consistency between the native U937 and the 3T3-erlin-1 cell line, in terms of response to treatment, suggested that the 3T3-erlin-2HA line was behaving in a manner indicative of its native state. Based on these results, we conclude that like flotillin-1, erlin-1 and erlin-2 populate intracellular DRM domains.
Erlin-1 and Erlin-2 are localized to the ER
Since we deduced that erlin-1 and erlin-2 were present in intracellular DRMs, we further characterized their specific subcellular localization using microscopy. Although erlin-1 was first identified in hematopoietic cells, western blotting revealed that it was expressed in a wide array of human cell lines (data not shown). Therefore, we used the human HCT116, MCF-7 and HeLa cell lines for staining of endogenous erlin-1 with the 10E6 antibody. All cells displayed a perinuclear staining pattern consistent with nuclear envelope localization, a substructure of the ER (Fig. 6A) (Estrada de Martin et al., 2005). This staining pattern was reminiscent of those achieved with the ER marker antibodies anti-calnexin and anti-calreticulin (compare with Fig. 7). 10E6 staining was also observed in places where involution of cell nuclei was apparent, consistent with nuclear envelope staining (Fig. 6A, arrowheads). Thus, staining for endogenous erlin-1 revealed a subcellular localization consistent with the ER.
Since we did not generate a monoclonal antibody to erlin-2 we assessed its subcellular distribution by immunostaining of NIH 3T3 cells expressing epitope-tagged erlin-2 using the anti-HA antibody, 12CA5. Using this approach we observed a reticular cytoplasmic pattern upon staining with 12CA5, reminiscent of ER morphology (Fig. 6B). Similar to the results achieved with erlin-1, staining of the nuclear envelope was also observed for erlin-2HA, illustrated by examination of a single focal plane through the cell (Fig. 6B, NE).
To extend our studies of the subcellular localization of these molecules we generated EGFP fusions with full length erlin-1 and -2 and transiently transfected these constructs into HeLa cells. Both erlin-1 and erlin-2 demonstrated a reticular pattern of fluorescence, as well as nuclear envelope staining, consistent with ER localization (Fig. 6). A high degree of co-localization was observed when full-length erlin-1-EGFP- and erlin-2-EGFP-expressing cells were counterstained with the ER membrane protein marker antibodies anti-calnexin and anti-calreticulin (Fig. 7).
Other PHB family members do not localize to the ER
As an additional control for non-specific accumulation of overexpressed membrane proteins in the ER, we took advantage of the reported distinct subcellular localizations of endogenous PHB family proteins as controls for correct subcellular localization of the erlins. To this end, EGFP-tagged versions of prohibitin-1, found in the mitochondria, flotillin-1, known to localize to the PM and Golgi as well as endocytic compartments, and stomatin 1 isoform a, known to localize to the PM were generated (Gkantiragas et al., 2001; Morrow and Parton, 2005). None of the overexpressed, EGFP-tagged, PHB family members were observed to localize to the ER as assessed by the lack of correspondence of EGFP fluorescence with anti-calnexin staining (Fig. 8). EGFP-tagged prohibitin-1 displayed mitochondrial localization, whereas both stomatin-1a and flotillin-1 were observed at the plasma membrane as well as at intracellular vesicular structures, probably representing Golgi and/or endocytic compartments (Fig. 8, arrowheads). Consistent with other reports of endogenous protein, all ectopically expressed, EGFP-tagged PHB family members tested were observed at their previously reported subcellular locales (Morrow and Parton, 2005) (Fig. 8).
The extreme N-terminus of PHB proteins is sufficient for targeting to their distinctive subcellular localizations
Since the prohibitin family of proteins exhibits diversity in terms of subcellular localization, we characterized the sequences required for the appropriate targeting of several PHB family members. To this end we generated EGFP fusions with the extreme N-termini (including the putative transmembrane domains) of erlin-1, erlin-2, and the PHB family members prohibitin-1 and stomatin-1a. Schematic representations of the various TM-EGFP fusion constructs are outlined in Fig. 9A.
While the extreme N-terminal domains of erlin-1 and -2 were sufficient for targeting of EGFP to the ER, the analogous domains of prohibitin efficiently targeted EGFP to the mitochondria (Fig. 9B,C). The longer N-terminal sequence of stomatin targeted EGFP to the PM as well as to cytoplasmic vesicular structures probably representing Golgi or endocytic membranes (Fig. 9D, arrowhead). These experiments demonstrate that the extreme N-termini of PHB family members are sufficient for their distinctive subcellular targeting.
In the past few years lipid rafts have garnered considerable attention based on the diverse functional roles that have been attributed to them, including transcytosis (Simionescu et al., 1983), potocytosis (Anderson, 1992), an alternate route for endocytosis and phagocytosis (Nichols, 2003; Parton and Richards, 2003; Smart et al., 1999), internalization of various pathogens (Fivaz et al., 1999; Kurzchalia, 2003; Manes et al., 2003; Parton et al., 1994; Shin et al., 2000), cholesterol transport (Oram and Yokoyama, 1996; Smart et al., 1996), calcium homeostasis (Isshiki and Anderson, 1999), protein sorting (Simons and Ikonen, 1997) and signal transduction (Anderson, 1993; Simons and Toomre, 2000; Smart et al., 1999; Zajchowski and Robbins, 2002). There is now compelling evidence to suggest that there are distinct populations of lipid rafts on different cell types as well as within the same cell. These data include distinct populations of lipid rafts on polarized epithelial cells (Roper et al., 2000; Scheiffele et al., 1998) as well as on polarized migrating T cells (Gomez-Mouton et al., 2001). In addition it is clear that different GPI-anchored proteins are localized to distinct membrane microdomains (Madore et al., 1999).
Based on the lack of knowledge of the molecular composition of lipid rafts we embarked on an immunologically based approach to generate unique tools to investigate the molecular heterogeneity and composition of lipid rafts. We chose to focus our initial studies on human cell lines of myelomonocytic origin because they lack caveolae (Fra et al., 1994) and thus would allow us to focus on the composition of non-caveolin containing lipid rafts. Using this approach we were able to produce antibodies to: (1) known lipid-raft-resident proteins, (2) known proteins not previously reported to reside in lipid rafts as well as (3) identify and characterize proteins whose function or existence was previously unknown or unconfirmed (Table 1). In the present study we have focused our attention on two proteins, erlin-1 and erlin-2 (also known as KE04p and C8orf2, respectively) because of their high degree of enrichment within lipid raft fractions and because they were two previously uncharacterized proteins.
Our studies complement a growing number of large-scale proteomics studies aimed at characterizing the molecular composition of lipid rafts (Bae et al., 2004; Bini et al., 2003; Blonder et al., 2005; Blonder et al., 2004a; Chakraborty et al., 2005; Foster et al., 2003; Karsan et al., 2005; Kim et al., 2004; Li et al., 2003; Li et al., 2004; Mielenz et al., 2005; Nebl et al., 2002; Tu et al., 2004; Wollscheid et al., 2004). Some of these studies have identified erlin-1 and erlin-2 within lipid raft fractions of a number of cell types but very little if any information regarding their biochemical and molecular characterization is available (Blonder et al., 2005; Blonder et al., 2004b; Blonder et al., 2004c; Ledesma et al., 2003; Li et al., 2004; Sprenger et al., 2004). Although these large-scale proteomic studies provide valuable insights into the global molecular composition of DRMs, they are not able to distinguish between compositionally distinct rafts within the same cell, nor are they able to further characterize the molecules that they identify.
By contrast, our approach results in the generation of tools for the further characterization of identified molecules as well as the specific microdomains that they inhabit. Using one of the antibodies that we generated (10E6) we were able to confirm the enrichment of erlin-1 in lipid raft fractions of both endogenously and ectopically expressing cells as well as reveal its intracellular localization. In addition, using an epitope-tagged allele of erlin-2 we were able to show that it had similar biochemical properties to erlin-1 and was highly enriched in lipid raft fractions. Using immunofluorescent staining of endogenous erlin-1 and expression of EGFP-tagged versions of both proteins we were able to demonstrate that these two molecules are localized within the endoplasmic reticulum. Based on their biochemical properties and unique subcellular localization we have named these proteins ER lipid raft protein-1 and -2 (erlin-1 and erlin-2) for KE04p/SPFH domain protein 1 and C8orf2/SPFH domain protein 2, respectively.
Erlin-1 and erlin-2 represent the most recently discovered members of the growing prohibitin family of proteins. In addition to sharing a PHB domain, members of the prohibitin family of proteins share many common characteristics including: detergent insolubility, buoyancy in sucrose gradients, association with cellular membranes, and a propensity to form higher-order oligomers (Mairhofer et al., 2002; Neumann-Giesen et al., 2004; Nijtmans et al., 2002; Salzer and Prohaska, 2001; Snyers et al., 1998). However, they differ in their subcellular distributions, which include the plasma membrane, phagosomes, lipid droplets, Golgi vesicles, mitochondria, the nucleus and now the ER (Dermine et al., 2000; Garin et al., 2001; Gkantiragas et al., 2001; Green et al., 2004; Morrow et al., 2002; Salzer and Prohaska, 2001; Tatsuta et al., 2005; Umlauf et al., 2004). Proteins with a PHB domain are highly conserved throughout evolution and are found within diverse kingdoms including those of plants, bacteria and animals (Morrow and Parton, 2005; Tavernarakis et al., 1999).
In this study, we established that the most recently discovered members of the PHB family, erlin-1 and -2, like other family members, are associated with cellular membranes and are enriched in the detergent-insoluble, buoyant fractions of sucrose gradients (Mairhofer et al., 2002; Mielenz et al., 2005; Morrow et al., 2002; Salzer and Prohaska, 2001; Snyers et al., 1999). We confirmed their association with lipid rafts, as opposed to other detergent-insoluble, buoyant complexes, based on their sensitivity to the cholesterol sequestering agent M-β-C following exposure of intracellular membranes. Another common feature of PHB family members shared by these proteins is that erlin-1 and especially erlin-2 appear to form oligomeric complexes that are not fully dissociated by heating in protein sample buffer containing SDS and β-mercaptoethanol. This oligomerization can be greatly enhanced if the reducing reagent is omitted from the sample buffer (data not shown). We also discovered another common property of PHB family members: they are targeted to their distinctive subcellular locations by their extreme N-termini. Within the extreme N-terminus of erlin-1 and erlin-2 is a predicted transmembrane domain that is sufficient for the targeting of EGFP to the ER. Similarly, we determined that other PHB family members rely on putative transmembrane sequences in their N-termini for their distinctive subcellular targeting. Our data extend the observations made in previous reports that the N-terminal sequences of flotillin-1 and the yeast prohibitins (Phb1 and Phb2) are sufficient for localization to the PM and mitochondria, respectively (Liu et al., 2005; Morrow et al., 2002; Tatsuta et al., 2005). Furthermore, our data suggest that the transmembrane domains of erlin-1 and -2 contain ER-targeting information independent of known ER retrieval signals such as di-lysine or di-arginine motifs.
DRMs have been reported to reside within multiple intracellular membranes including the endosomal system, mitochondria, the Golgi, phagosomes and lysosomes (Bae et al., 2004; Dermine et al., 2001; Parton and Richards, 2003). The members of the prohibitin family of proteins represent excellent examples of protein markers that populate diverse subsets of detergent-resistant membranes within the cell (Morrow and Parton, 2005). Adding to the diversity of the PHB family in terms of subcellular localization, we demonstrate that both erlin-1 and -2 reside in the ER.
It has long been established that DRMs (characterized by their insolubility in non-ionic detergent and their buoyancy on sucrose gradients) are not exclusive to the PM. Early studies pioneering the biochemical isolation of lipid rafts demonstrated that association of proteins with DRMs occurs as early as the Golgi in the secretory pathway (Brown and Rose, 1992; Fra et al., 1994; Paladino et al., 2004). Interestingly, both the early Golgi and the ER are involved in the de novo synthesis of sphingolipids and cholesterol (Prinz, 2002; van Meer and Lisman, 2002). However, although the Golgi contains a substantial proportion of these components, the ER is known as the organelle with the lowest sphingolipid and cholesterol content within the secretory pathway (Holthuis et al., 2001; Prinz, 2002; van Meer and Lisman, 2002). For this reason the ER has been viewed as an unlikely place for the existence of lipid raft domains.
Nonetheless, there are reports that DRMs exist at the ER in mammalian cells. For example, proteins involved in the synthesis and transfer of the GPI anchors to GPI-anchored proteins (GPI-APs) have been observed at DRMs within the ER (Pielsticker et al., 2005; Sevlever et al., 1999). Additionally, the GPI-linked, prion protein PrP has been demonstrated to associate with DRMs as early as the ER within the secretory pathway (Paladino et al., 2004; Sarnataro et al., 2004). Finally, DRMs were reported to exist within a substructure of the ER, ER-lipid droplets (Hayashi and Su, 2003). Hayashi and Su (Hayashi and Su, 2003) demonstrated that σ-1 receptors of the brain are associated with DRMs of ER-lipid droplets; however, they exhibit lower buoyancy on sucrose gradients than plasma membrane lipid rafts and do not co-stain with calnexin. By contrast, we report that the novel transmembrane proteins erlin-1 and erlin-2 are enriched in buoyant DRMs and are residents of the ER, supporting the notion that lipid raft-like domains are present in the ER. This contention is based on the isolation of lipid rafts using the stringent detergent, Triton X-100, as well as by using methods aimed at disrupting these cholesterol-dependent structures (Schuck et al., 2003). Collectively, our data, as well as the aforementioned reports provide strong evidence for the existence of lipid raft-like domains in the ER.
There is certainly increasing data to suggest that there are subdomains of the ER such as specific regions that are continuous with a peroxisomal reticulum from which a major source of peroxisomes can originate (Bascom et al., 2003; Hoepfner et al., 2005), as well as ER subdomains that give rise to lipid droplets (Ozeki et al., 2005; Robenek et al., 2004). In addition there is now increasing evidence to suggest that there are regions of the ER that are in direct contact with the plasma membrane as well as in intimate association with mitochondria (Levine and Rabouille, 2005). Specific regions of the ER interacting with mitochondria regulate important biological functions including apoptosis and calcium release (Breckenridge et al., 2003; Chandra et al., 2004). It is interesting to note that we identified the pro-apoptotic protein BAP31 in our lipid raft screen, a protein known to reside in the ER that mediates the pro-apoptotic calcium signal that passes from the ER to the mitochondria after BAP31 is cleaved by a caspase-mediated event (Breckenridge et al., 2003; Chandra et al., 2004).
Despite the large body of literature on the PHB family of proteins, very little is known about their exact biological functions. The best-defined members of the PHB family are the prohibitins, which are also the closest relatives of the erlins within this family (Fig. 1C). They are reported to function as chaperones for mitochondrial membrane proteins (Nijtmans et al., 2000). Specifically, they protect proteins from degradation by the mitochondrial m-AAA protease (Steglich et al., 1999). This function is conserved in the bacterial kingdom where the prohibitin homologues HflK and HflC protect E. coli membrane proteins from the AAA-protease, FtsH (Kihara and Ito, 1998). A number of mammalian membrane proteins including the cystic fibrosis transmembrane conductance regulator (CFTR), (Jensen et al., 1995), HMG CoA reductase (Moriyama et al., 1998) and stearoyl-CoA desaturase (Heinemann and Ozols, 1998), are degraded by ER-resident proteases as a complementary pathway to the classical ER-associated degradation (ERAD) cytosolic degradation route. Given their high degree of similarity to prohibitin and their presence in the ER, it is possible that the erlins regulate the stability of proteins by protecting them from degradation by ER-resident proteases.
Based on these reports as well as the structure and subcellular location of erlin-1 and erlin-2, these newly discovered PHB family members in particular, and detergent-resistant ER microdomains in general, may represent specialized regions of the ER for mediating diverse activities such as GPI-anchor biosynthesis and attachment, quality control, protein degradation, proteolytic signal peptide processing, chaperone functions and/or antigen processing and presentation. More work is required to fully understand the roles that the erlins in particular and ER lipid-raft-like domains in general are playing in various aspects of cellular function.
Materials and Methods
Multiple amino acid sequence alignment was accomplished using CLUSTALW analysis. PHYLIP rooted dendrogram was produced based on the foregoing analysis. Putative transmembrane segments were predicted using the TMAP algorithm, all available at SDSC Biology Workbench (www.workbench.sdsc.edu). PHB domains were determined based on the conserved domain database with protein-protein BLAST (BLASTp) available at NCBI.
Antibodies and chemicals
The following antibodies were used: anti-Src MAb327 (J. Brugge, Harvard Medical School, Boston, MA); anti-flotillin-1 MAb (Transduction Labs); rabbit anti-calnexin (Stressgen); goat anti-pyruvate kinase (Chemicon); rabbit anti-CD71/transferrin receptor (Santa Cruz); Fugene 6, 12CA5 (Roche); goat anti-mouse Alexa Fluor 488 and 568, goat anti-rabbit Alexa Fluor 568, SP-DiIC18 and Mitotracker CMX-Ros (Molecular Probes); goat anti-mouse HRP conjugate (BD Pharmingen); goat anti-rabbit HRP (Santa Cruz); rabbit anti-goat HRP (Pierce). Puromycin, paraformaldehyde, tetradecanoyl phorbol myristate acetate (TPA), DMSO, polybrene and methyl-β-cyclodextrin were all from Sigma. RIBI adjuvant was from ImmunoChem Research and pGEX-2T from Amersham.
Generation of monoclonal antibodies
Triton X-100 detergent-insoluble fractions were purified from 1.5×108 U937 cells differentiated with 20 ng/ml TPA for 48 hours and from 1.5×108 HL-60 cells differentiated for 4 days with 1.3% DMSO as per our normal protocol (Robbins et al., 1995). The isolated DRMs from the two cell types were pooled and suspended in RIBI adjuvant before injecting subcutaneously into two Balb/C mice with one-sixth of the total DRM fraction. Boosters were given to each mouse on days 21 and 35 and the spleens isolated from the mice on day 42. Spleen cells were fused with the Sp2 myeloma cell line that expresses the murine IL-6 gene according to standard protocols (Fuller et al., 1998). After fusion, cells were plated into 96-well plates and selected in HAT medium. Supernatants from individual hybridomas were screened by western blotting of DRMs made from the differentiated U937 cells and HL60 cells as described above. Of the approximately 900 hybridomas that grew, 32 different antibodies were selected for further study based on their strong immunoreactivity with the DRM fractions.
Mass spectrometry (peptide mass fingerprinting/MALDI-TOF MS/MS)
Protein gel slices were sent to Dr Liang Li (ACB Proteomics Mass Spectrometry Facility, Department of Chemistry, University of Alberta, Canada) and analyzed on a Bruker REFLEX III time-of-flight mass spectrometer using MALDI positive ion mode. Selected peptides for each sample were fragmented using MALDI MS/MS on a PE Sciex API-QSTAR pulsar to obtain sequence information to identify the proteins.
Erlin-1 (KE04; Acc. No. AF064093), erlin-2HA (C8orf2HA; Acc. No. NM_007175) and prohibitin (Acc. No. NM_002634) were cloned by RT-PCR from U937 cDNA. Flotillin-1 (NM_005803) and stomatin transcript variant 1 (Acc. No. NM_004099) were cloned from a human RNA library (Stratagene). Except where indicated, PCR inserts were first cloned into the pGEM-T-Easy vector (Promega), and subcloned into the indicated mammalian expression vectors. Primer pairs for erlin-1 were XhoI-KE04-F, 5′-CCGCTCGAGAATGAATATGACTCAAG-3′ and ClaI-KE04p-R, 5′-CCATCGATCAACCTGTGCTCTCTTTG-3′; for erlin-2HA were XhoI-C8orf2-F, 5′-CCGCTCGAGCCGCCATGGCTCAGTTGGGA-3′ and ClaI-C8orf2HA-R, 5′-CCATCGATTCAAGCGTAATCTGGAACATCGTATGGGTATCCTCCATTCTCCTTAGTGGCCGTCTCCAAG-3′. Erlin-1 and erlin-2HA were cloned into pBabepuro3 using XhoI and ClaI restriction sites. For erlin-1 GST fusions the primer pairs used were GST-N-terminus: BamHI-KE04-F, 5′-GGATCCGAGAATGAATATGACTCAAGCCC-3′ and EcoRI-KE04-191-R, 5′-GAATTCTCATGTCTTCTCAGCCTCCATTAACTC-3′; GST-C-terminus: BamHI-KE04-184-F, 5′-GGATCCGAGTTAATGGAGGCTGAGAAGAC-3′ and EcoRI-KE04-end-R, 5′-GAATTCTCAACCTGTGCTCTCTTTGTTTTG-3′; GST-305-346: BamHI-KE04-305-F, 5′-GGATCCTTCGTGGACTCCTCATGTGCTTTG-3′ and EcoRI-KE04-end-R. GST-330-346 was made by hybridizing the following oligonucleotides and cloning directly into pGEX-2T: KE04-330-F, 5′-GATCCGCTCTTGAACCCTCTGGAGAGAACGTCATCCAAAACAAAGAGAGCACAGGTTGAG-3′ and KE04-330-R, 5′-AATTCTCAACCTGTGCTCTCTTTGTTTTGGATGACGTTCTCTCCAGAGGGTTCAAGAGCG-3′. GST fusions with erlin-1 fragments were generated by cloning into pGEX-2T vector with BamHI and EcoRI restriction sites. For GFP constructs, the primer pairs are as follows: erlin-1-FL-GFP: XhoI-KE04-F and AgeI-KE04-FL-GFP-R, 5′-ACCGGTCCACCTGTGCTCTCTTTGTTTTGGATG-3′; erlin-2-FL-GFP: XhoI-C8orf2-F and AgeI-C8orf2-FL-GFP-R, 5′-ACCGGTCCATTCTCCTTAGTGGCCGTCTCCAA-3′; prohibitin-FL-GFP: XhoI-prohibitin-F, 5′-CTCGAGCGGCCGCATGGCTGCCAAAGTGTTTGAGTC-3′ and AgeI-prohibitin-FL-GFP-R, 5′-ACCGGTCCCTGGGGCAGCTGGAGGAGCA-3′; stomatin-FL-GFP: XhoI-stomatin-F, 5′-CTCGAGGGCAGCATGGCCGAGAAGCG-3′ and AgeI-stomatin-FL-GFP-R, 5′-ACCGGTCCCTCTTTTATAATCTTTATGCACATCC-3′; flotillin-1-FL-GFP: XhoI-flotillin-F, 5′-CTCGAGTGAACCATGTTTTTCACTTGTGGCC-3′ and AgeI-flotillin-FL-GFP-R, 5′-ACCGGTCCGGCTGTTCTCAAAGGCTTGTGA-3′; erlin-1-N29-GFP: XhoI-KE04-F and AgeI-KE04-TM29-R, 5′-ACCGGTCCGCCCTCCTCAATCTTGTGGATGG-3′; erlin-2-N25-GFP: XhoI-C8orf2-F and AgeI-C8orf2-TM25-R, 5′-ACCGGTCCCTTGTGCACAGCTGAGAAGAGAG-3′; prohibitin-N38-GFP: XhoI-prohibitin-F and AgeI-prohibitin-TM38-R, 5′-ACCGGTCCGATGACAGCTCTGTGCCC AGCATC-3′; stomatin-N59-GFP: XhoI-stomatin-F and AgeI-stomatin-TM59-R, 5′-ACCGGTCCATCTTTATGCACATCCATATTGAGATTG-3′. Erlin-1, erlin-2, prohibitin-1, stomatin-1 and flotillin-1 full-length and truncation mutants were cloned into pEGFP-N1 using XhoI and AgeI restriction sites.
Cell lines and cell culture
U937 and HL60 cells were maintained in RPMI complete medium. HeLa and MCF-7 cells were maintained in DMEM 10% FBS, 1% penicillin-streptomycin (P/S). NIH 3T3 cells were maintained in DMEM, 5% FBS, 1% P/S. Erlin-1 and erlin-2HA stably expressing 3T3 cell lines were created using previously described methods and maintained in the presence of 2 μg/ml puromycin (Davy et al., 1999; Robbins et al., 1995).
Treatments with methyl-β-cyclodextrin
Whole cells were incubated in PBS or 20 mM methyl-β-cyclodextrin in PBS for 30 minutes at 37°C. For treatment of cellular membranes, membrane isolates were prepared using previously described methods in hypotonic medium (HM; 10 mM HEPES-KOH, pH 7.5, 10 mM KCl, 1 mM MgCl2, 1 mM PMSF, 1 μg/ml aprotinin, 10 μg/ml leupeptin) (Robbins et al., 1995). Briefly, cells were subjected to one cycle of freeze-thaw and incubated in HM for 30 minutes on ice. Next, cells were passed through a 26.5 gauge needle for 20 strokes and membranes pelleted by centrifugation at 16,000 g for 10 minutes. Membrane pellets were then incubated in methyl-β-cyclodextrin using the same conditions as whole cells described above. Following treatment, cells or membrane isolates were pelleted by centrifugation and subjected to sucrose gradient fractionation outlined below.
DRM preparation and cell fractionation
DRM isolates were prepared as described previously (Robbins et al., 1995). The insoluble (DRM) pellet was resuspended in 1× protein sample buffer [1× PSB: 62.5 mM Tris-HCl, pH 6.8, 10% glycerol (w/v), 5% β-mercaptoethanol (v/v), 2.3% SDS (w/v), 0.01% Bromophenol Blue (w/v)]. The soluble, high-density fraction was obtained by resuspending the bottom 2 ml of the gradient (40% sucrose) in 10 ml of 1× MBS and subjected to centrifugation for a further hour. The supernatant was removed and proteins precipitated with trichloroacetic acid (TCA), (20% w/v final) on ice for 10 minutes. Precipitate was centrifuged at 11,000 rpm in a Beckman JA-17 rotor. Pellet was resuspended in a 1:1 mixture of 1 M Tris-HCl pH 11 and 2× PSB. For gradient fractionation 2 ml fractions were collected from the top of the gradient (1) to the bottom of the gradient (6). These fractions were further subfractionated into soluble and insoluble subfractions by dilution in 10 ml of 1× MBS and ultracentrifugation for a further hour. Soluble (supernatant) and insoluble (pellet) subfractions were retained. Soluble fractions were obtained by TCA precipitation, centrifugation and resuspension in 1:1, 1 M Tris-HCl (pH 11): 2× PSB. Insoluble (raft) pellets were resuspended in 1× PSB.
For immunofluorescent staining with the 10E6 antibody, serum-free hybridoma supernatant was concentrated by ammonium sulphate precipitation. MCF-7 and HeLa cells were prepared by fixation with 4% paraformaldehyde/PBS for 20 minutes at RT. HCT116 cells were fixed with methanol:ethanol at 1:1 for 5 minutes at -20°C. Cells were then permeabilized by incubation in 0.2% Triton X-100-PBS for 10 minutes on ice. Next, cells were incubated in undiluted, concentrated 10E6 for 1 hour at RT, washed with PBS and incubated in goat-anti-mouse-Alexa Fluor 488 (Molecular Probes) at 20 μg/ml (1:100 dilution) in PBS for 1 hour at RT. Finally, stained cells were washed with PBS, counterstained with 500 nM DAPI and mounted on slides with PPDA/glycerol. NIH 3T3 cells expressing erlin-2HA were prepared by fixation in paraformaldehyde and permeabilization as outlined above. Staining was performed with ammonium sulphate-precipitated 12CA5 at a dilution of 1:500 in PBS, and goat-anti-mouse-Alexa Fluor 568 (20 μg/ml; 1:100 dilution) secondary with PBS washes as described above. For EGFP experiments, HeLa cells were plated onto acid-washed coverslips in 24-well plates on the day before transfection. Cell transfection was carried out using 0.2 μg of plasmid DNA and 0.6 μl of Fugene 6 according to manufacturer's instructions. For mitochondrial staining 100 nM Mitotracker was added 30 minutes before fixation according to the manufacturer's instructions. Cells were fixed with 4% paraformaldehyde, and permeabilized as above, 24-48 hours following transfection and washed with PBS. Cells were incubated in rabbit-anti-calnexin at a dilution of 1:1000 or rabbit-anti-calreticulin at a dilution of 1:200 in PBS for 1 hour at RT. Cells were washed with PBS and incubated in goat-anti-rabbit-Alexa Fluor 568 at 10 μg/ml (a dilution of 1:200) in PBS, washed again and mounted on slides as outlined above. Cells were visualized using a PlanApo 1.4/60× oil immersion objective on an Olympus IX70 microscope with mercury arc lamp excitation. Band pass filters for visualizing EGFP fluorescence or Alexa Fluor 488 and Alexa Fluor 568 fluorescence were excitation/emission 490±10 nm/528±19 nm, and 555±14 nm/617±37 nm, respectively. Acquisition of z-stacks with 0.2 μm intervals was accomplished using a cooled Photometrics CH350 camera (-40°C) and the DeltaVision RT restoration imaging system. Images were subjected to full iterative digital deconvolution on a Silicon Graphics Octane 2 workstation running IRIX software and checked for convergence. Images represent an individual focal plane except in Fig. 6B where the micrograph with inset magnification represents a projected stack of deconvolved z-planes. To increase fluorescence detection, images representing full-length prohibitin-GFP and calnexin were binned at 2×2 (Fig. 7A). Images were coloured and merged using Adobe Photoshop 7.0.
We would like to thank members of the Robbins' laboratory for input during the course of this work. We would like to thank Nancy Quintrell for her help with the initial generation and screening of thelipid raft monoclonal antibodies. We would also like to thank Dr Liang Li in the Department of Chemistry at the University of Alberta for identification of proteins by mass spectrometry. We would also like to express our gratitude to Dr Pina Colarusso and Betty Pollock of the Live Cell Imaging Core Facility associated with the Canadian Institutes of Health Research (CIHR) Group in Inflammatory Disease for guidance and technical assistance with the microscopy studies. This work was supported by an operating grant from the CIHR to S.M.R. S.M.R. currently holds a Canada Research Chair in Cancer Biology and is an Alberta Heritage Foundation for Medical Research Scientist.
Supplementary material available online at http://jcs.biologists.org/cgi/content/full/119/15/3149/DC1
- Accepted May 10, 2006.
- © The Company of Biologists Limited 2006