Direct transport across the plasma membrane of mammalian cells of Leishmania HASPB as revealed by a CHO export mutant

Transport of most secretory proteins to the extracellular space is mediated by the ER/Golgi-dependent secretory pathway (Palade, 1975; Rothman, 1994; Rothman and Wieland, 1996; Nickel et al., 2002). In the case of soluble factors, the principal targeting motifs are N-terminal signal peptides that direct classical secretory proteins to the translocation machinery of the ER (Keenan et al., 2001). However, for a number of soluble factors with defined extracellular functions, it has been demonstrated that ER/Golgi-independent routes of protein secretion exist (Muesch et al., 1990; Rubartelli and Sitia, 1991; Cleves, 1997; Hughes, 1999; Nickel, 2003; Prudovsky et al., 2003). Among these, the most prominent examples are the angiogenic growth factors FGF-1 (Jackson et al., 1992; Jackson et al., 1995; Shin et al., 1996; LaVallee et al., 1998; Tarantini et al., 1998; Landriscina et al., 2001a; Landriscina et al., 2001b; Prudovsky et al., 2002; Mandinova et al., 2003) and FGF-2 (Mignatti and Rifkin, 1991; Mignatti et al., 1992; Florkiewicz et al., 1995; Trudel et al., 2000; Engling et al., 2002), cytokines like interleukin 1β (IL1β) (Rubartelli et al., 1990; Andrei et al., 1999; Andrei et al., 2004) and migration inhibitory factor (MIF) (Flieger et al., 2003) as well as the galectin protein family, lectins of the extracellular matrix (Cooper and Barondes, 1990; Cleves et al., 1996; Lutomski et al., 1997). These and other unconventional secretory proteins are characterized by the lack of a signal peptide, by not being glycosylated (despite bearing multiple consensus sites for this post-translational modification) and, most importantly, by an export mechanism that is fully functional in the presence of brefeldin A (Cleves, 1997; Hughes, 1999; Nickel, 2003), a drug that blocks ER/Golgi-dependent protein transport (Misumi et al., 1986; Lippincott-Schwartz et al., 1989; Orci et al., 1991).

Leishmania HASPB is another interesting kind of unconventional secretory protein that is a component of the surface coat of Leishmania parasites (Flinn et al., 1994; McKean et al., 1997; Alce et al., 1999). HASPB is exclusively expressed in infective parasites in both extracellular metacyclics and intracellular amastigotes of L. major and L. donovani (Rangarajan et al., 1995; McKean et al., 1997; Alce et al., 1999) suggesting a role in parasite virulence, although gene deletion mutants retain their viability in the host (McKean et al., 2001). Intriguingly, following heterologous expression, HASPB is also externalized by mammalian cells (Denny et al., 2000) suggesting that endogenous factors exist in higher eukaryotes that are exported in a mechanistically similar manner.

HASPB contains an N-terminal SH4 domain that becomes dually acylated by myristoylation of glycine 2 and palmitoylation of cysteine 5 (Denny et al., 2000). These post-translational modifications are essential for HASPB targeting to the cell surface of Leishmania parasites (Pimenta et al., 1994; Denny et al., 2000). Interestingly, the extreme N-terminus of HASPB (HASPB-N18) is sufficient for targeting of HASPB-N18-GFP to the cell surface (Denny et al., 2000). A myristoylated HASPB mutant that cannot be palmitoylated (HASPB^C5S-GFP) has been found to localize to Golgi membranes (Denny et al., 2000) suggesting that HASPB associates at least transiently with membranes of the classical secretory pathway. Therefore, even though palmitoylacyltransferase activities have not only been localized to the Golgi but also to the ER and plasma membranes (reviewed by Bijlmakers and Marsh, 2003), it is likely that transfer of palmitate to the SH4 domain of HASPB occurs at the Golgi. Following palmitoylation, HASPB is transferred to the plasma membrane but the mechanism of this transport process is unknown. Interestingly, HASPB surface expression in mammalian cells is not affected by brefeldin A (Denny et al., 2000), making it unlikely that dually acylated HASPB reaches the plasma membrane bound to the cytoplasmic leaflet of secretory vesicles. Based on these considerations, HASPB trafficking appears to be distinct from all three known classes of plasma membrane-targeted proteins carrying dual acylation motifs at their N-termini (Bijlmakers and Marsh, 2003). The first of these, the Src kinase Lck, is palmitoylated at the Golgi followed by brefeldin A-sensitive transport to the plasma membrane (Bijlmakers and Marsh, 1999; Bijlmakers and Marsh, 2003). Secondly, Gα_ζ subunits of plasma membrane-resident trimeric G proteins are transiently associated with intracellular membranes like Lck, but these factors get palmitoylated at the plasma membrane and transport to this site is fully operational in the presence of brefeldin A (Fishburn et al., 1999; Bijlmakers and Marsh, 2003). Finally, a third mechanism of plasma membrane targeting of N-terminally acylated proteins is exemplified by the Src kinase Fyn that does not appear to contact intracellular membranes but rather is directly targeted from the cytoplasm to the plasma membrane. Consistently, brefeldin A does not affect Fyn targeting (van't Hof and Resh, 1997; Bijlmakers and Marsh, 2003). As indicated above, current knowledge about HASPB targeting to the plasma membrane does not fit with any of these examples as HASPB is transiently associated with, and most likely palmitoylated at, the Golgi followed by plasma membrane targeting in a brefeldin A-insensitive manner (Denny et al., 2000).

Based on these considerations, many questions about the molecular mechanism of HASPB targeting to the plasma membrane remain unresolved. There is also a complete lack of knowledge about the subcellular site of membrane translocation of HASPB, a process that eventually allows HASPB exposure on the cell surface of eukaryotic cells. In the current study, we introduce a novel experimental system that permits the precise quantification of HASPB export from mammalian cells based on flow cytometry. Using this assay, we screened for somatic CHO mutants that are incapable of exporting HASPB-GFP fusion protein employing retroviral insertion mutagenesis. Based on a detailed biochemical and morphological characterization of a mutant with this phenotype, we conclude that the site of HASPB membrane translocation is the plasma membrane. These data suggest that in the CHO mutant described, a component of a plasma membrane-resident machinery has been disrupted causing the fully acylated HASPB-GFP reporter molecule to accumulate in the inner leaflet of the plasma membrane. These results support a two-step process of HASPB biogenesis, in which dual acylation of the HASPB SH4 domain is required for delivery to the inner leaflet of the plasma membrane and subsequent recognition by a putative translocation machinery. Intriguingly, the component of the HASPB export apparatus that has been disrupted in the mutant cell line appears to be specific for this export pathway, as the angiogenic growth factor FGF-2, another protein secreted by unconventional means, is found to be exported normally from the HASPB export mutant cell line.

Affinity-purified anti-GFP antibodies were generated as described earlier (Engling et al., 2002; Seelenmeyer et al., 2003). Antibodies directed against the transferrin receptor were from Zymed, antibodies directed against GM130 were purchased from BD Transduction Laboratories. Monoclonal anti-CD4 antibodies were obtained from cell culture supernatants of the hybridoma cell line Okt4 (ATCC CRL-8002). APC-conjugated anti-rabbit and anti-mouse IgG antibodies used for FACS analyses were from Molecular Probes.

All HASPB fusion proteins used in this study are based on enhanced GFP (Clontech). The 18 N-terminal amino acids of HASPB (MGSSCTKDSAKEPQKRAD) were fused to the N-terminus of eGFP through a short linker sequence (GPVAT). In case of HASPB-N18-GFP Δmyr/palm and HASPB-N18-GFP Δpalm, two single amino acid changes were introduced to prevent myristoylation and palmitoylation (G2A), and palmitoylation (C5S), respectively. All constructs were cloned into the retroviral expression vector pREV-TRE2 that contains a doxicycline-dependent promoter. Generation of CHO cell lines expressing the various HASPB fusion proteins employing retroviral transduction was performed as described previously (Engling et al., 2002).

CHO cells expressing the various kinds of HASPB-GFP fusion proteins were cultured for 2 days at 37°C in the presence of doxicycline (1 μg/ml). Following detachment using PBS/EDTA, cells were collected by centrifugation (500 g, 5 minutes) and resuspended in PBS/sucrose (10% w/w) homogenization buffer. Cells were broken up by sonication and then ultracentrifuged for 60 minutes at 100,000 g. The resulting supernatant was defined as the cytosolic fraction, the sediment was resuspended in homogenization buffer and defined as the membrane fraction. For carbonate extraction experiments, membranes were again collected by ultracentrifugation. Following resuspension in Na₂CO₃ (0.1 M, pH 11.5), the samples were incubated for 30 minutes at 4°C. Membranes were re-isolated by ultracentrifugation and separated into supernatant (peripheral membrane proteins) and membrane sediment (integral and tightly associated proteins).

CHO cells expressing HASPB-GFP fusion proteins were grown on six-well plates for 48 hours at 37°C in the presence of doxicycline (1 μg/ml) to about 80% confluency. Following 2 hours of incubation in FCS-free medium, cells were incubated in FCS-free medium containing either 250 μCi [³H]palmitic acid or 100 μCi [³H]myristic acid. After 3 hours of incubation in labeling medium at 37°C, cells were washed and lysed in buffer containing detergent by sonication. Following removal of insoluble material by centrifugation (14,000 g, 10 minutes), the lysates were subjected to immunoprecipitation of HASPB-GFP fusion proteins employing affinity-purified anti-GFP antibodies. Antibody-bound protein was eluted with SDS sample buffer and subjected to SDS-PAGE. For this purpose, the samples were split into two fractions in order to analyze the eluted proteins by both fluorography and silver staining. For fluorography, SDS gels were treated with NAMP100 amplifier solution (Amersham) and then dried for exposure using Hyperfilm™ (Amersham). Typically, SDS gels containing [³H]myristic acid-labeled proteins were exposed for about 2 weeks whereas SDS gels containing [³H]palmitic acid-labeled proteins were exposed for up to 6 weeks.

CHO cells expressing HASPB-GFP fusion proteins were grown on glass cover slips for 48 hours at 37°C in the presence of 1 μg/ml doxicyclin. The cells were then processed including paraformaldehyde fixation (3% w/v) for 20 minutes at 4°C. The specimens were mounted in Fluoromount G (Southern Biotechnology Associates) and GFP-derived fluorescence was viewed with a Zeiss LSM 510 confocal microscope.

CHO cells expressing HASPB-GFP fusion proteins were grown under the conditions indicated in the corresponding figure legends. In order to detach the cells from the culture plates without using protease-based protocols, cell dissociation buffer (Life Technologies) was used to generate a cell suspension devoid of cell aggregates. Where indicated, cells were treated with affinity-purified anti-GFP antibodies for 1 hour at 4°C on a rotating wheel. Wash procedures were carried out by sedimenting the cells at 200 g for 3 minutes at 4°C. Primary antibodies were detected with goat anti-rabbit secondary antibodies coupled to allophycocyanin (APC; Molecular Probes). Prior to FACS analysis, propidium iodide (1 μg/ml) was added in order to detect damaged cells. GFP- and APC-derived fluorescence was analyzed using a Becton Dickinson FACSCalibur flow cytometer. Autofluorescence was determined by measuring non-induced cells that were treated with primary and APC-coupled secondary antibodies. FACS-mediated isolation of cell populations and clones was performed using a Becton Dickinson FACSVantage cell sorter.

CHO cells expressing HASPB-GFP fusion proteins were grown under the conditions indicated in the corresponding figure legends. Following removal of the medium, the cells were washed twice with PBS containing Ca²⁺ and Mg²⁺ (2 mM each). A membrane-impermeable biotinylation reagent (EZ-Link Sulfo-NHS-SS-biotin; Pierce) dissolved in 10 mM triethanolamine, 150 mM NaCl and 2 mM CaCl₂ (pH 9) was added at a final concentration of 1 mg/ml. Following incubation for 30 minutes at 4°C quenching of excess amounts of biotinylation reagent was achieved by adding 100 mM PBS/glycine. The cells were then washed twice with PBS containing 2 mM Ca²⁺ and Mg²⁺. Cell lysates were prepared by adding PBS containing NP-40. Cell lysates were cleared by centrifugation and biotinylated and non-biotinylated proteins were separated by streptavidin affinity chromatography. The two fractions were probed for the presence of HASPB-GFP fusion proteins using SDS-PAGE and western blotting with affinity-purified anti-GFP antibodies.

To randomly mutate CHO cells, we prepared retroviral particles encoding the cell surface protein CD4 (pBI-CD4) (Liu et al., 2000) using the Viraport system (Stratagene). CHO cells expressing HASPB-N18-GFP were transduced according to standard procedures. Following 3 days of incubation at 37°C CD4-positive cells were selected by FACS sorting. In order to isolate HASPB export mutants CD4-positive cells were subjected to multiple rounds of FACS sorting, selecting for cells with high GFP fluorescence and low cell surface staining. In the third round of FACS sorting, individual cells were collected and further propagated in 96-well plates at 37°C. The corresponding clonal cell lines were then characterized as described.

In order to study the export mechanism of HASPB on a quantitative basis in mammalian cells, we generated CHO cells that express various kinds of HASPB-GFP fusion proteins in a doxicycline-dependent manner. CHO_MCAT-TAM2 cells (Engling et al., 2002) were transduced with retroviruses carrying HASPB-N18-GFP, HASPB-N18-GFP-Δpalm (C5S), HASPB-N18-GFP-Δmyr/palm (G2A) or GFP, respectively. Clonal cell lines were obtained by several rounds of FACS sorting in the absence and presence of doxicycline, respectively, followed by the selection of single cells characterized by GFP-derived fluorescence only after the addition of doxicycline. As shown in Fig. 1A, expression of the various reporter molecules could only be detected following incubation of the cells with doxicycline as analyzed by western blotting employing anti-GFP antibodies.

We then tested the overall membrane association of the various HASPB fusion proteins by subcellular fractionation into a soluble pool (referred to as `cytosol'; Fig. 1B, lane 1) and a membrane fraction (Fig. 1B, lane 2). Additionally, the membrane fraction was subjected to carbonate extraction (Fujiki et al., 1982) to discriminate loosely attached material (Fig. 1B, lane 3) from protein tightly associated with membranes (Fig. 1B, lane 4). Typically roughly equal amounts of HASPB-N18-GFP were found in the cytosolic and the membrane fractions. About two thirds of the membrane-bound material was found to be resistant to carbonate treatment. In the case of HASPB-N18-GFP-Δpalm (C5S), as expected, the population found in the cytosolic fraction was significantly larger than that present in the membrane fraction (Fig. 1B). About 50% of membrane-associated HASPB-N18-GFP-Δpalm was found to be resistant to carbonate treatment. The vast majority of HASPB-N18-GFP-Δmyr/palm (G2A), a mutant that lacks both myristoylation and palmitoylation, was found in the cytosolic fraction which is also the case for GFP lacking the HASPB-N18 sequence (Fig. 1B). Thus, subcellular fractionation combined with carbonate extraction allows analysis of the acylation status of HASPB fusion proteins.

We next analyzed the subcellular distribution of the HASPB-GFP fusion proteins employing confocal microscopy (Fig. 2). HASPB-N18-GFP was found to localize mainly to the plasma membrane as well as to droplet-like structures that were typically located in close proximity to the plasma membrane (Fig. 2A). By contrast, HASPB-N18-GFP-Δpalm was not at all associated with the plasma membrane but was rather localized on intracellular membranes in a perinuclear position (Fig. 2B). As expected, HASPB-N18-GFP-Δmyr/palm was found exclusively in the cytoplasm (Fig. 2C) with a staining pattern indistinguishable from that of GFP (Fig. 2D). Thus, membrane association of HASPB-GFP is completely dependent on dual acylation of the HASPB N-terminus. These results are consistent with previous studies analyzing similar constructs in Leishmania parasites (Denny et al., 2000).

Based on previous observations (Denny et al., 2000), an export assay was developed that simultaneously measures both the expression level of a given HASPB-GFP fusion protein and its export to the outer leaflet of the plasma membrane on a quantitative basis. Following membrane translocation, HASPB-GFP fusion proteins remain membrane-anchored through their acylated N-termini, allowing decoration of the extracellular population with anti-GFP antibodies. This subpopulation is then specifically detected using APC-conjugated secondary antibodies. Using flow cytometry, GFP-derived fluorescence (overall expression level) and APC-derived fluorescence (exported population) is measured simultaneously. All three HASPB-GFP fusion proteins as well as GFP lacking an N-terminal HASPB tag are expressed at similar levels provided that the cells have been treated with doxicycline (Fig. 3A). Only the expression of HASPB-N18-GFP gives rise to surface staining whereas HASPB-N18-GFP-Δpalm, HASPB-N18-GFP-Δpalm/myr and GFP are negative with regard to cell surface staining (Fig. 3B). Again, these data are consistent with previous observations made in parasites (Denny et al., 2000) in that export to the extracellular leaflet of the plasma membrane of mammalian cells is critically dependent on dual acylation of the SH4 N-terminus of HASPB. Thus, the in vivo assay introduced in Fig. 3 allows precise quantification of HASPB export under normalized conditions with respect to the expression level of a given HASPB-GFP fusion protein.

CHO cells In order to demonstrate an external population of HASPB-N18-GFP associated with the outer leaflet of the plasma membrane using an independent method, biotinylation experiments were carried out employing a membrane-impermeable biotinylation reagent. Cells were incubated in the presence of doxicycline, serum-containing medium was removed followed by treatment with a biotinylation reagent. After quenching and removing excess amounts of the biotinylation reagent, cells were converted into a detergent lysate that was then subjected to streptavidin affinity chromatography to separate biotinylated (cell surface) from non-biotinylated (intracellular) proteins. HASPB-N18-GFP can be detected in the fraction eluted from streptavidin beads (Fig. 4, lane 3) whereas the various controls (HASPB-N18-GFP-Δpalm, HASPB-N18-GFP-Δmyr/palm and GFP) are absent from this fraction. These results demonstrate that the biotinylation reagent does not traverse the plasma membrane and, therefore, the positive signal for HASPB-N18-GFP in the fraction of biotinylated proteins represents a population located on the surface of these cells. Under steady-state conditions, the amount of HASPB-N18-GFP on the surface of CHO cells was calculated by densitometry to vary between experiments within a range of ∼3-6% of the overall population (data not shown).

The principal aim of this study was to identify the subcellular site of HASPB membrane translocation in CHO cells. To address this point, we generated clonal CHO mutants by retroviral insertion mutagenesis, which are characterized by a negative HASPB export phenotype. In order to make sure that such mutants can be identified in the presence of wild-type cells, we first conducted experiments designed to detect intercellular spreading of HASPB-N18-GFP following translocation onto the cell surface. For this purpose, HASPB-N18-GFP-expressing and HASPB-N18-GFP-non-expressing cells (CHO_MCAT-TAM2) were mixed in culture, treated with doxicycline and subjected to a FACS analysis measuring GFP and APC-derived cell surface fluorescence. Two populations of cells from a mixed culture of CHO_MCAT-TAM2 cells (Fig. 5A, green curve) (negative for GFP fluorescence, data not shown) and HASPB-N18-GFP-expressing cells (blue curve) (positive for GFP fluorescence, data not shown) can be easily distinguished by anti-GFP cell surface staining, demonstrating that HASPB-N18-GFP-exporting cells retain the secreted material on their surface. However, cell surface staining of HASPB-N18-GFP-non-expressing cells is in fact significantly higher than the autofluorescence background defined by antibody-treated CHO_MCAT-TAM2 cells (Fig. 5A, red curve), suggesting that a certain amount of HASPB-N18-GFP was transferred from HASPB-N18-GFP-expressing cells to non-expressing cells. This observation emphasizes the conclusions drawn from the experiments shown in Figs 3 and 4, in that an extracellular population of HASPB-N18-GFP can be defined by its appearance on the surface of cells that are incapable of expressing the reporter molecule when mixed with HASPB-N18-GFP-expressing cells. With regard to the mutagenesis strategy, however, these experiments demonstrate that HASPB export mutants are distinguishable from wild-type cells on the basis of FACS sorting and can therefore be identified and isolated by this method, even in the presence of a large excess of wild-type cells.

To randomly generate somatic CHO mutants, we chose retroviral insertion mutagenesis. CHO cells carrying the HASPB-N18-GFP reporter gene were transduced with retroviral particles encoding the open reading frame of the integral plasma membrane protein CD4 (Fig. 5B) (Liu et al., 2000). In this context, CD4 was used as a marker for mutated cells that were subsequently enriched by FACS. The pool of CD4-positive mutated cells (Fig. 5B, blue curve) was then subjected to a selection of CHO mutants expressing the HASPB-N18-GFP reporter molecule at a normal level (as compared to wild-type cells) but without localisation to the cell surface. For this purpose, cells were viewed in the FACS set-up in dot blot mode and a sorting window was defined to select for the phenotype described (Fig. 5C). Cells isolated from this window in the first round of sorting (about 0.5% of the total population) were propagated and subjected to a second round of cell sorting in which 3.7% of the selected population displayed the desired phenotype. In a third round of cell sorting, single cells were selected in the sorting window (defined in Fig. 5C).

In the current study, we provide a detailed characterization of a clonal CHO cell line (from here on referred to as `K3') derived from the selection procedure of the experiments described in Fig. 5. In order to make sure that the reporter molecule itself does not contain a mutation in the CHO K3 mutant cell line, we isolated genomic DNA and amplified the HASPB-N18-GFP open reading frame. As demonstrated by sequencing of the corresponding PCR product, the sequence of the N-terminal SH4 domain of the reporter molecule was confirmed to be unchanged compared to the parental CHO wild-type cells (data not shown).

Based on GFP-derived fluorescence, the expression level of HASPB-N18-GFP does not significantly differ between wild-type and K3 cells as analyzed by FACS (Fig. 6A). However, HASPB-N18-GFP cell surface staining is largely reduced to about 30% as compared to wild-type levels (Fig. 6A). These data were confirmed by a biochemical assessment of the extracellular population of HASPB-N18-GFP in the wild type compared to K3 mutant cells using the biotinylation assay described in Fig. 4. The amount of biotinylated cell surface HASPB-N18-GFP (Fig. 6B, upper panel; lane 3) derived from K3 cells is reduced to about 20% when compared to CHO wild-type cells (lower panel; lane 3).

As dual acylation of HASPB-N18-GFP is a critical determinant for plasma membrane targeting and export, it was important to analyze whether the reporter molecule HASPB-N18-GFP itself is properly modified in CHO K3 cells. Therefore, in a first set of experiments, overall membrane association of HASPB-N18-GFP was tested in CHO K3 cells by a carbonate extraction analysis. The input material of HASPB-N18-GFP was comparable in wild-type CHO and K3 cells, which is consistent with the FACS experiments shown in Fig. 6, demonstrating that the expression level of HASPB-N18-GFP does not differ significantly in the two cell lines (Fig. 7A). Similarly, there are no apparent differences in the distribution of HASPB-N18-GFP between cytosol and membranes when CHO wild-type and K3 cells are compared (Fig. 7B, lanes 1 and 2). Finally, when the membrane fractions of wild-type and K3 cells were subjected to carbonate extraction, the distribution of HASPB-N18-GFP between the membrane-associated and the soluble pool was identical between the two cell types (Fig. 7B, lanes 3 and 4). These data suggest that the acylation status of HASPB-N18-GFP does not differ between CHO wild-type and K3 cells. This conclusion was confirmed by metabolic labeling of HASPB-N18-GFP in wild-type and K3 cells using [³H]-labeled myristate and [³H]-labeled palmitate. Incorporation of both fatty acids into HASPB-N18-GFP can be detected in both wild-type and K3 cells whereas the corresponding negative controls are either labeled only with [³H]myristate (in the case of HASPB-N18-GFP-Δpalm) or not labeled at all (for HASPB-N18-GFP-Δmyr/palm and GFP) (Fig. 7C). These experiments demonstrate that HASPB-N18-GFP processing in terms of acylation occurs normally in the CHO K3 mutant cell line and, therefore, the lack of HASPB-N18-GFP present on the cell surface of these cells must be due to the translocation machinery itself from which at least one component is apparently disrupted in K3 cells.

In order to define the subcellular site of HASPB-N18-GFP membrane translocation, we analyzed its localization in CHO wild-type cells versus CHO K3 cells using confocal microscopy and subcellular fractionation. There was virtually no difference in the subcellular distribution of HASPB-N18-GFP in the two cell types and in both cases the majority of the material is localized to the plasma membrane (Fig. 8A,B). These morphological data could be confirmed by a biochemical analysis employing subcellular fractionation (Fig. 8C). A protocol to purify plasma membrane vesicles (Schäfer et al., 2004) was used to compare the enrichment of the HASPB-N18-GFP reporter molecule in plasma membranes of CHO wild-type cells (Fig. 8C, lanes 1-4) and CHO K3 cells (Fig. 8C, lanes 5-8), respectively. When compared to the homogenate (lanes 1), HASPB-N18-GFP (lane 5) was found to be enriched in gradient-purified plasma membrane vesicles using the plasma membrane marker transferrin receptor (TfR) (Futter et al., 1998) in both CHO wild-type and CHO K3 cells (lanes 4 and 8, respectively). Using the Golgi marker GM130 (Nakamura et al., 1995), the plasma membrane vesicle fraction was depleted of Golgi membranes demonstrating the significance of the findings described above. The association of HASPB-N18-GFP with plasma membranes in both CHO wild-type cells and K3 cells is considered to be quite striking as it demonstrates a defect in CHO K3 cells directly in the translocation machinery rather than some kind of intracellular segregation of HASPB-N18-GFP that prevents access to the site of membrane translocation. Thus, the HASPB-N18-GFP translocation apparatus is a plasma membrane-resident machinery.

In order to analyze whether the disrupted export component in CHO K3 cells is a specific factor for HASPB export or rather a general component for non-classical secretory processes, we tested whether FGF-2, as a classical example for unconventional secretory proteins (Nickel, 2003), is exported from CHO K3 cells. For this purpose, we transduced both CHO K3 and the parental CHO wild-type cells expressing the HASPB-N18-GFP fusion protein with retroviral particles encoding a FGF-2-GFP fusion protein (Engling et al., 2002; Backhaus et al., 2004). Owing to the different molecular weights (25 and 45 kDa, respectively), the two reporter molecules could be easily distinguished in the cell surface biotinylation assay (Figs 4 and 6). There was no difference in FGF-2 cell surface expression between CHO wild-type and CHO K3 cells (Fig. 9, compare lanes 3 and 6; upper panels), respectively. By contrast, HASPB-N18-GFP analyzed from the same cell preparations is exported only from CHO wild-type cells (compare lanes 3 and 6; lower panels). These results demonstrate that the component disrupted in CHO K3 cells is a specific machinery molecule of the HASPB export pathway.

The specific aim of this study was to develop an experimental approach designed to identify the subcellular membrane system that contains the molecular machinery required for cell surface expression of Leishmania HASPB in mammalian cells. This question is of general interest because, during its biogenesis, HASPB has been shown to contact intracellular membranes such as the Golgi (Denny et al., 2000). Based on the evidence currently available, the most logical explanation for this observation is that Golgi membranes contain the palmitoylacyltransferase required for the thioester-based acylation of cysteine 5 in the N-terminal SH4 domain of HASPB. From this point on, it is not clear whether fully acylated HASPB translocates across the membrane of the Golgi or whether it is first transported to the plasma membrane associated with the cytoplasmic leaflet of secretory vesicles. Even though overall cell surface expression of HASPB has been shown not to be affected by brefeldin A (Denny et al., 2000), this result does not unequivocally rule out the possibility that HASPB is traveling to the plasma membrane associated with the cytoplasmic leaflet of TGN-derived secretory vesicles. It is also possible that transfer to the plasma membrane involves an intermediate step in which HASPB is not at all associated with intracellular membranes or HASPB might be transported from the TGN to the endosomal system from where it might get access to the plasma membrane.

Based on the uncertainties discussed above, we were interested in establishing an experimental system that allows for a quantitative analysis of HASPB cell surface expression under conditions where defined steps in HASPB biogenesis are blocked. In particular, we were looking for experimental conditions where dual acylation of the N-terminal SH4 domain of HASPB occurs normally but HASPB export to the outer leaflet of the plasma membrane is blocked. In this context, we decided to use CHO cells as a model system as they are well suited to the generation of random somatic mutants. To efficiently mutate CHO cells expressing HASPB-N18-GFP in a doxicycline-dependent manner, we used retroviral insertion mutagenesis with CD4 as a cell surface marker to efficiently enrich mutated cells by FACS. Using the FACS-based experimental system described, we succeeded in isolating clonal CHO mutants with the desired phenotype, one of which has been characterized in detail in the current study.

In the described CHO mutant strain (K3), HASPB-N18-GFP is expressed in a doxicycline-dependent manner at a level comparable to that of wild-type cells. In the context of this work, it was most critical to isolate CHO mutants that do not have any defect in the co- and/or post-translational processing of HASPB, i.e. we were looking for mutants that are still capable of adding both myristate and palmitate to the SH4 domain of HASPB. In the case of CHO K3 cells, this was shown to be the case by carbonate extraction experiments to probe overall membrane association of HASPB-N18-GFP as well as metabolic labeling experiments to demonstrate directly the incorporation of [³H]-labeled myristate and palmitate into HASPB-N18-GFP when expressed in CHO K3 cells. These experiments unequivocally demonstrated that HASPB-N18-GFP is processed normally in CHO K3 mutant cells resulting in a membrane association that is indistinguishable from HASPB-N18-GFP expressed in CHO wild-type cells.

The negative HASPB-N18-GFP export phenotype by which the clonal CHO K3 cell line was isolated could be confirmed by both FACS-based cell surface staining experiments and a biochemical assessment of HASPB-N18-GFP export using a membrane-impermeable biotinylation reagent. Both methods consistently demonstrated that HASPB-N18-GFP cell surface expression in CHO K3 cells is greatly reduced to ∼10-30% of the population found on the cell surface of CHO wild-type cells. As the expression level of HASPB-N18-GFP does not differ significantly between CHO K3 and wild-type cells, as shown by both western blotting and GFP-derived fluorescence determined by FACS, we conclude that the process of HASPB-N18-GFP membrane translocation is perturbed in the mutant cell line. Based on this observation, it was then crucial to analyze whether the steady-state distribution of the reporter molecule is changed in CHO K3 cells as compared to wild-type cells. As demonstrated by confocal microscopy and subcellular fractionation, this is not the case as the majority of the HASPB-N18-GFP population localizes in both cell lines to the plasma membrane. Intriguingly, a FGF-2-GFP fusion protein is secreted equally well from the CHO K3 mutant cell line and the parental CHO wild-type cells, demonstrating that the disrupted factor in K3 cells is a specific component of the HASPB export pathway.

From the combined data presented in this study, we conclude that the CHO mutant cell line K3 has a direct defect in the molecular machinery promoting membrane translocation of HASPB. The subcellular distribution of the reporter molecule is unchanged when wild-type and K3 mutants cells are compared, with HASPB-N18-GFP detected in both cases as a plasma membrane-resident protein. As cell surface exposure of the reporter is largely reduced in the mutant cell line, we conclude that the plasma membrane is the subcellular site of membrane translocation of HASPB-N18-GFP. Thus, along with the angiogenic growth factors FGF-1 and FGF-2 (Prudovsky et al., 2002; Schäfer et al., 2004), HASPB represents another example of an unconventional secretory protein that is translocated directly across the plasma membrane of mammalian cells in order to be exposed to the extracellular space. Our data allow us to describe a two-step process for the overall biogenesis of HASPB that is defined by acylation-dependent targeting to the inner leaflet of the plasma membrane followed by translocation across the plasma membrane, resulting in a membrane-anchored lipoprotein on the surface of both parasites and mammalian cells.

The experimental approach presented in this study shows great promise for addressing a number of questions related to HASPB biogenesis in the future. In this regard, our goals are not restricted to the molecular analysis of the HASPB membrane translocation apparatus itself but also concern the preceding steps such as the molecular identity and the exact subcellular localization of the HASPB palmitoylacyltransferase. Moreover, it remains a completely open question how HASPB is transferred from the Golgi to the plasma membrane. As the experimental approach described gave rise to a whole collection of CHO mutant cell lines characterized by a negative HASPB export phenotype, we hope to also identify mutants that are defective in early steps of HASPB biogenesis. The most important future goal will be to look for the defective genes by functional expression cloning and/or genome insertion analysis of the retrovirus used as a mutagen as well as to assign the corresponding gene products to defined steps of the HASPB biogenesis pathway on a functional basis.

We thank Britta Brügger (Heidelberg University Biochemistry Center) for critical comments on the manuscript. We would like to thank Lars Dietrich and Christian Ungermann (Heidelberg University Biochemistry Center) for many interesting discussions on protein acylation and SH4 domains. This study was supported by a grant of the German Research Council (DFG Ni 423/3-5).