Unconventional Secretory Routes: Direct Protein Export Across the Plasma Membrane of Mammalian Cells



The vast majority of extracellular proteins are exported from mammalian cells by the endoplasmic reticulum/Golgi-dependent secretory pathway. For poorly understood reasons, however, a heterogenous group of extracellular proteins has been discovered that does not make use of signal peptide-dependent secretory transport. Both the release mechanisms and the molecular identity of the secretory machines involved have remained elusive. Recent studies now have established a subgroup of unconventional secretory proteins capable of translocating from the cytoplasm directly across the plasma membrane to get access to the exterior of eukaryotic cells. This review aims to focus on a detailed comparison of the subcellular site of membrane translocation of various unconventional secretory proteins such as the proangiogenic molecule fibroblast growth factor-2 (FGF-2) and Leishmania hydrophilic acylated surface protein B (HASP B). A potential link between membrane translocation and quality control as an integral part of unconventional secretory processes is discussed.

Classical Versus Non-Classical Mechanisms of Eukaryotic Protein Secretion

All eukaryotic cells from yeast to man are characterized by an elaborate secretory machine that recognizes signal peptide-bearing proteins resulting in their translocation across the membrane of the endoplasmic reticulum (ER) (1). Once localized to the lumen of the ER, secretory proteins are packaged into transport vesicles provided they pass ER quality control measures. Following cargo delivery to the Golgi apparatus (2), post-Golgi transport carriers fuse with the plasma membrane, a process that eventually results in the release of classical secretory proteins into the extracellular space (3,4). However, a number of secretory proteins with defined extracellular functions have been shown not to contain functional signal peptides and do not represent substrates for the ER membrane translocation machinery (5–8). Furthermore, the extracellular appearance of such molecules is not compromised in the presence of brefeldin A, a drug that blocks ER/Golgi-dependent secretory transport. These observations led to the postulation of alternative secretory mechanisms in eukaryotic cells that are fully functional in the absence of an intact ER/Golgi system and therefore have been collectively termed unconventional secretory processes (5–8). Intriguingly, unconventional secretory proteins comprise a group of molecules of significant biomedical relevance such as the proangiogenic growth factor fibroblast growth factor-2 (FGF-2). FGF-2 is a tumor-produced, direct-acting stimulator of angiogenesis, a process that is essential for tumor growth and metastasis (9). Other examples include lectins of the extracellular matrix (galectin family) involved in tumor-mediated immune suppression (10), inflammatory cytokines such as interleukin-1β (IL-1β) (11) and macrophage migration inhibitory factor (MIF) (12) as well as a family of stage-regulated surface molecules from Leishmania parasites termed hydrophilic acylated surface proteins (HASPs) implicated in host cell infection (13).

The Subcellular Site of Membrane Translocation

The focus of this review is a specific aspect of unconventional secretory processes, namely the subcellular site of membrane translocation. The bulk population of unconventional secretory proteins in general is typically localized to the cytoplasm (5–8). As depicted in Figure 1, four potential mechanisms of unconventional protein export have been discussed in the literature to mediate translocation of cytosolic factors into the extracellular space (6). Two of these (Figure 1; mechanisms 1 and 3) involve intracellular vesicles of the endocytic membrane system such as secretory lysosomes (14,15) and exosomes (16), the latter ones being internal vesicles of multivesicular bodies (17). Under suitable conditions, lysosomal contents gain access to the exterior of cells when specialized endocytic structures such as secretory lysosomes of cytotoxic T lymphocytes or melanosomes of melanocytes fuse with the plasma membrane (15). Similarly, lumenal contents of endocytic structures can be released into the extracellular space when multivesicular bodies fuse with the plasma membrane, a process that results in the release of exosomal vesicles along with their cargo molecules (16). Two alternative unconventional secretory mechanisms are characterized by a direct translocation of cytosolic factors across the plasma membrane using either protein-conducting channels such as adenosine triphosphate-binding cassette (ABC) transporters (18) (Figure 1; mechanism 2) or a process called membrane blebbing (Figure 1; mechanism 4), the latter one being characterized by shedding of plasma membrane-derived microvesicles that are released into the extracellular space (19–21) (Figure 1).

Figure 1.

Vesicular and non-vesicular pathways potentially involved in unconventional secretory processes. 1, export by secretory lysosomes; 2, export mediated by plasma membrane-resident transporters; 3, export through the release of exosomes derived from multivesicular bodies; 4, export mediated by plasma membrane shedding of microvesicles. For details, see main text.

A classical example of an unconventional secretory protein whose export mechanism involves intracellular vesicles is IL-1β (22). When homogenates of activated monocytes were analyzed by gradient centrifugation and protease protection experiments, an IL-1β subpopulation could be detected in the lumen of intracellular vesicles (23). On the basis of immunolocalization studies employing electron microscopy, these subcellular vesicles have been identified as an endolysosomal subcompartment because IL-1β-positive vesicles display the typical morphology of endocytic organelles and are positive for cathepsin D (CD) and Lamp-1, classical markers of late endosomes and lysosomes. However, only a fraction of the total population of CD- and Lamp-1-positive vesicles was also labeled by anti-IL-1β antibodies suggesting that this population represents a specialized subspecies of endolysosomes (23). It is believed that following appropriate stimulation during the onset of inflammatory processes, IL-1β-containing vesicles undergo fusion with the plasma membrane resulting in the release of IL-1β into the extracellular space (23,24). Uptake of IL-1β into secretory lysosomes might be mediated by a protein-conducting ABC transporter as the overall process of IL-1β secretion is sensitive to glyburide, a drug targeted against the ABC1 family of membrane transporters (25,26). However, it is also quite possible that glyburide inhibtion of IL-1β secretion is based on an indirect mechanism, as ABC1 transporters have also been implicated in the membrane translocation of cholesterol (27), and therefore, it seems unlikely that a defined type of ABC transporter is capable of translocating two classes of molecules structurally as different as a small protein and a membrane lipid. Other unconventional secretory proteins such as high mobility group 1 protein (HMGB1) and possibly MIF are released by secretory lysosomes as well (28–30); however, in other cases such as for FGF-2, galectin-1 or Leishmania HASPB, lumenal localization in intracellular vesicles as an obligatory intermediate of the secretory process in question has not been observed.

Two recent experimental approaches have been used to analyze the subcellular site of membrane translocation of FGF-2 and Leishmania HASPB. In the case of FGF-2, an in vitro assay has been exploited that is based on immunopurified plasma membrane vesicles with inside-out topology (31). The rationale of this system is that translocation of a given factor into the lumen of inside-out vesicles would mimic its secretion, as the lumen of these vesicles is topologically equivalent to the extracellular space (Figure 2). When recombinant FGF-2 was incubated in the presence of inside-out vesicles, translocation of FGF-2 into the lumen of these vesicles could be observed in a temperature- and incubation time-dependent manner. Intriguingly, recombinant FGF-2 was capable of translocating only across the membrane of inside-out vesicles while right-side-out vesicles rejected FGF-2 as a substrate for membrane translocation. As depicted in Figure 2, galectin-1 added as a recombinant protein was also found to traverse the membrane of inside-out vesicles. By contrast, the unconventional secretory protein MIF and FGF-4, a classical secretory protein with an N-terminal signal peptide, were found not to be substrates for membrane translocation (31; Wuttke, Zehe and Nickel, unpublished data). These data establish that plasma membrane vesicles highly enriched by immunopurification contain the molecular machinery required for FGF-2 membrane translocation into the extracellular space. Indeed, removal of vesicle-associated proteins under high salt conditions or through protease treatment resulted in a loss of FGF-2 membrane translocation across the membrane of inside-out vesicles (31; Wuttke, Zehe and Nickel, unpublished data). Thus, the combined data suggest that FGF-2 secretion is mediated by a plasma membrane-resident transporter (Figure 1; mechanism 2).

Figure 2.

In vitro reconstitution of fibroblast growth factor (FGF)-2 membrane translocation (31). For details, see main text.

In the case of the Leishmania cell-surface molecule HASPB, an in vivo approach has been used to analyze the subcellular site of membrane translocation (32). Interestingly, HASPB has been shown not only to be exported from Leishmania parasites but, upon heterologous expression, is also found to be secreted from mammalian cells (33). The HASPB primary structure differs from all other unconventional secretory proteins known to date, as it contains an N-terminal SH4 domain commonly found in src kinases that is a substrate for N-terminal protein acylation (34). As depicted in Figure 3, HASPB biogenesis starts with cotranslational myristoylation of its N-terminus. A second acylation step involves palmitoylation at cysteine 5 of the SH4 domain of HASPB. HASPB palmitoylation mutants (C5A) localize to Golgi membranes suggesting that the putative palmitoylacyltransferase is a resident enzyme of the Golgi apparatus (32,33). Dual acylation of the SH4 domain of HASPB mediates stable membrane association of the molecule. Following transient association with the Golgi, HASPB is transported to the inner leaflet of the plasma membrane. As indicated in Figure 3, it is currently unclear how this transport step is mediated; however, there are in principle three options: (i) HASPB might be transported to the plasma membrane associated with the cytoplasmic leaflet of secretory vesicles, (ii) HASPB might be targeted first to endosomal structures followed by translocation to the plasma membrane or (iii) HASPB transport from the Golgi to the plasma membrane might not rely on transport vesicles. In any case, palmitoylation of HASPB is strictly required for plasma membrane targeting as palmitoylation-deficient mutants of HASPB are efficiently retained at the level of the Golgi (32,33). The final localization of HASPB is characterized by its stable association with the outer leaflet of the plasma membrane with the protein moiety being exposed to the extracellular space (Figure 3). Therefore, HASPB must translocate across at least one membrane during its biogenesis pathway. In case intracellular HASPB transport relies on vesicular intermediates, membrane translocation could occur at any membrane HASPB is associated with, during its transport to the cell surface, e.g. the Golgi, the plasma membrane and, potentially, secretory or endosomal vesicles. The site of HASPB membrane translocation has recently been addressed by a somatic mutagenesis approach in which CHO mutants were screened for a phenotype characterized by the inability to export HASPB to the outer leaflet of the plasma membrane (32). Following isolation of such CHO clones using flow cytometry, the fate of HASPB in one such mutant cell line was analyzed in detail. It was shown that HASPB is expressed at a level comparable with that of CHO wild-type cells, and HASPB was found to be efficiently acylated by both myristate and palmitate residues. On the basis of confocal microscopy and subcellular fractionation, accumulation of HASPB was observed at the level of the plasma membrane; however, employing both flow cytometry and cell-surface biotinylation assays, the amount of HASPB on the cell surface was drastically reduced. These data were taken to mean that a component of the HASPB translocation machinery is compromised in this CHO-mutant cell line, and therefore, membrane translocation occurs at the level of the plasma membrane rather than at intracellular sites (32). Thus, like FGF-2, HASPB membrane translocation is likely to be mediated by a plasma membrane-resident transporter (Figure 1; mechanism 2). Intriguingly, FGF-2 was shown to get normally exported from the HASPB export mutant cell line suggesting that the factor disrupted is a specific component of the HASPB export pathway.

Figure 3.

Biogenesis of hydrophilic acylated surface proteins (HASPs) expressed on the surface of Leishmania parasites (32,33). For details, see main text.

How do the findings described for FGF-2 and HASPB compare with other reports in the literature? FGF-1, a close relative of FGF-2, is believed to be secreted by a mechanism distinct from FGF-2 export as FGF-1 release can be triggered under stress conditions such as heat shock treatment (35,36), an experimental condition without any impact on FGF-2 export (37). Also, FGF-1 has been found to associate with the S100 family of Ca2+-binding proteins and additional components resulting in the release of an FGF-1-containing hetero-oligomeric complex (38), an export-related property of FGF-1 that has not been reported for FGF-2. Despite the apparent mechanistic differences in the modes of export of FGF-1 and FGF-2, it appears quite likely that both FGF-1 (39) and FGF-2 (31) are exported by directed translocation across the plasma membrane of mammalian cells. Employing a real-time analysis of various FGF-1-expressing cell lines, Prudovsky et al. have demonstrated that heat shock treatment triggers the redistribution of FGF-1 along with other components of the S100 release complex from a diffuse cytoplasmic pattern into a population that accumulates at the inner leaflet of the plasma membrane (39). This subcellular redistribution has been interpreted as the initial step of heat shock-triggered FGF-1 secretion. Even though it remains elusive how the FGF-1-containing hetero-oligomeric complex gets translocated across the plasma membrane, the data clearly suggest that the overall process of FGF-1 secretion does not involve intracellular vesicles (as depicted in Figure 1 for mechanisms 1 and 3) but rather relies on a direct transport mechanism at the level of the plasma membrane (39) (Figure 1; mechanism 2).

Unconventional Secretory Processes without the Involvement of Intracellular Vesicle Intermediates

What could be the transport mechanism of unconventional secretory proteins such as FGF-1, FGF-2, galectin-1 and Leishmania HASPB whose export apparently does not involve intracellular vesicles? As illustrated in Figure 1, there are two options, i.e. translocation mediated by plasma membrane-resident transporters (mechanism 2) and protein release by membrane shedding or blebbing (mechanism 4), the latter process being characterized by the formation of so-called exovesicles. The data on FGF-1 described above (39) do not allow discrimination between these two options; however, evidence in a number of studies indicates that the secretory process for members of the galectin family involves the formation of exovesicles generated by membrane blebbing (6). Using a model system in which the developmental regulation of galectin-1 secretion can be studied, it has been shown that differentiation of muscle cells from myoblasts to myotubes is accompanied by the extracellular appearance of galectin-1 (40). Intriguingly, in immunolocalization studies at intermediate stages of the differentiation process, galectin-1 was reported to accumulate in evaginations of the plasma membrane. Additionally, galectin-1 was detected in exovesicles apparently in the process of detaching from the plasma membrane (40). Similarly, galectin-3 secretion has been proposed to be mediated by membrane shedding of labile vesicles into the supernatant of COS cells (41); however, in these experiments, an engineered version of galectin-3 has been used that is artificially targeted to plasma membranes using an N-terminal acylation motif. In contrast to the reports discussed above, a systematic proteomic analysis of exosomes does not point to a role of membrane blebbing in galectin secretion as galectin-3 has been identified as a major component of these structures that were isolated from cell culture supernatants of dendritic cells (42). As the major components of exosomes are derived from endocytic compartments and resident proteins from plasma membranes have been reported to be depleted from exosomes (16), it appears likely that the galectin-3-containing exovesicles described by Thery et al. are not generated by membrane blebbing but rather are derived from multivesicular bodies (Figure 1; mechanism 3). Even more confusing, galectin-1 is capable of translocating across the membrane of plasma membrane-derived inside-out vesicles (31), an observation that points to a potential role of a plasma membrane-derived transporter in the process of galectin-1 secretion (Figure 1; mechanism 2). Thus, the various studies on the mechanism of galectin secretion do not provide a consistent picture, and therefore, future studies using functional assays combined with high-resolution imaging techniques are required to gain insight into the molecular mechanisms of galectin export from mammalian cells.

In a recent report by Taverna et al., the shedding of plasma membrane vesicles has also been proposed as a main mechanism for FGF-2 secretion (43). This work was initiated based on the previous finding that following secretion, FGF-2 – GFP concentrates in heparan sulfate proteoglycan (HSPG)-containing clusters on the outer leaflet of the plasma membrane (44). These structures were interpreted as potential exovesicles originating from the plasma membrane (43); however, FGF-2 – GFP can be eluted from these structures employing heparin and is directly available for antibody staining without a need for detergent treatment. These results demonstrate that the extracellular FGF-2 – GFP population is not localized to the lumen of vesicles but is rather associated with cell-surface HSPGs (44). However, as depicted in Figure 1 (mechanism 4), shedding of plasma membrane-derived exovesicles is in principle a potential mechanism for unconventional secretory processes. Consistent with that, Taverna et al. have purified FGF-2-containing vesicles from cell culture supernatants which would support this hypothesis. However, these vesicles can be affinity-purified using immobilized annexin V, a well-established binding protein of phosphatidylserine. Thus, annexin V affinity-purified vesicles are likely to be derived from apoptotic cells as translocation of phosphatidylserine to the outer leaflet of the plasma membrane is a hallmark of programmed cell death. Additionally, Taverna et al. report that FGF-2-containing vesicles appear in the supernatant of cells that have been serum starved for long periods of time followed by readdition of serum (43), a procedure commonly used to artificially induce apoptosis. Nevertheless, it certainly cannot be excluded that FGF-2 secretion involves shedding of plasma membrane vesicles. In this context, it is notable that shear stress, another kind of cell activation promoting microvesicle shedding (19), has been implicated in the release of FGF-2 (45). However, further work is needed to establish a role for this mechanism in the release of FGF-2 from mammalian cells, especially because in the absence of various forms of cell activation, FGF-2 export has been found to occur in a constitutive manner from many cell lines (8).

Quality Control during Unconventional Secretory Processes

Despite the uncertainties discussed above, it is evident from recent work that there are alternative secretory routes that do not rely on intracellular vesicles as obligatory intermediates. Thus, release of unconventional secretory proteins such as IL-1β and HMGB1 by secretory lysosomes (23,24,30) is not a common theme for all extracellular factors without access to the classical ER/Golgi-dependent secretory route. This is especially true for FGF-2 and Leishmania HASPB whose export mechanisms are likely to rely on plasma membrane-resident transporters (31,32). Because membrane translocation of FGF-2 and Leishmania HASPB occurs at the level of the plasma membrane, the fundamental question arises of how quality control is ensured during these processes. This is particularly evident from the fact that eukaryotic cells have developed elaborate mechanisms of quality control within their classical ER/Golgi-dependent secretory pathway; hence, it appears rather unlikely that they do not apply quality control measures for unconventional secretory factors. In case of classical secretory transport, quality control occurs at the level of the ER in that secretory proteins not being folded properly do not have access to transport vesicle-mediated exit from this compartment but rather are targeted for degradation (46,47). On the basis of the ability of recombinant FGF-2 to translocate across the membrane of plasma membrane-derived inside-out vesicles, it appears likely that FGF-2 secretion is mediated by some kind of membrane-associated transporter system. In case FGF-2 membrane translocation would be mechanistically similar to ER-mediated membrane translocation of signal peptide-containing secretory proteins, the result would be the release of a non-functional protein into the extracellular space as ER-mediated membrane translocation occurs in an unfolded state (1). Similarly, membrane translocation in the course of protein import into the mitochondrial matrix occurs in an unfolded state, and chaperone systems ensure proper folding of newly translocated proteins in the lumen of both the ER and mitochondria (1). However, once secretory proteins are released into the extracellular space, quality control measures can no longer be applied. Thus, it appears quite reasonable that translocation of, for example, FGF-2 is in some way coupled to a mechanism that ensures secretion only of functional (i.e. properly folded) FGF-2. These considerations imply that FGF-2 might not be released in an unfolded state but rather is exported from cells in a functional form that has passed quality control measures. This hypothesis has been put to a first test using FGF-2 fusion proteins with dihydrofolate reductase (DHFR) as ligands of this enzyme can stabilize its conformation in a way such that chaperones are not able to break up its tertiary structure. This system has been used successfully to demonstrate the unfolded state of matrix proteins during import into mitochondria (48,49). Intriguingly, ligand-mediated stabilization of FGF-2 – DHFR fusion proteins did not have any impact on export efficiency suggesting that FGF-2 membrane translocation does not occur in an unfolded state (50). It is now important to challenge this observation using independent approaches. Most importantly, it needs to be addressed whether engineered FGF-2 fusion proteins are able to carry artificial interaction partners into the extracellular space based on non-covalent interactions. Also, the targeting of FGF-2 to its unconventional secretory pathway needs to be elucidated in molecular terms. In light of the issues discussed above, it is particularly tempting to speculate that the secretory targeting motif of FGF-2 is inseparably connected to its functional three-dimensional structure. If this were true, the export machinery would reject FGF-2 molecules that are not properly folded and therefore quality control of FGF-2 secretion would be in place. Like FGF-2, protein import into peroxisomes has also been shown to occur in a folded state (51). Interestingly, as opposed to mitochondria (1), the existence of chaperones in the lumenal matrix of peroxisomes has never been reported. Thus, matrix components of peroxisomes are targeted from the cytoplasm into a location in which chaperone-assisted folding reactions are unlikely to be possible. In principle, this is a scenario similar to what has been outlined above for FGF-2 export, and therefore, the fact that both protein translocation pathways appear to operate with folded substrates implies that the corresponding machineries might be related. In this context, it is also of note that bacterial transporters involved in protein secretion as well as protein import systems of plant chloroplasts have been reported to translocate proteins in a fully folded state (52). A hallmark of the twin-arginine translocation (Tat) system is its ability to move large, fully folded and even cofactor-containing oligomeric protein complexes across the plasma membrane into the bacterial periplasm (53). Most intriguingly, strong evidence has been reported that the Tat system can sense the folding state of cargo molecules, a feature that has been interpreted as a quality control mechanism that ensures a transport restriction to folded substrates (54). In this study, disulfide-dependent dimer formation of Tat-dependent cargo recognition was used as a tool, which is particularly interesting because the unconventional secretory mechanism of FGF-1 is also likely to be linked to dimer formation (36,55).


On the basis of quantitative in vitro and in vivo assays (31,44,50), advanced imaging techniques (39) as well as genetic studies on mammalian mutant cell lines (32), the molecular mechanisms of various unconventional secretory processes are now beginning to emerge. In particular, as opposed to IL-1β and HMGB1 secretion (23,24,30), it is now well established that a subset of unconventional secretory factors are exported by direct translocation from the cytoplasm across the plasma membrane into the extracellular space. These findings imply that quality control measures operate at this level of the export pathway, and indeed, evidence has been reported that FGF-2 secretion occurs in a folded state (50). These findings offer the fascinating possibility that the so-far-uncharacterized targeting signal of FGF-2 is somehow integrated in its three-dimensional structure rather than being a linear stretch of amino acids. If this were true, the export apparatus might only accept cargo molecules that are properly folded and, in turn, would ensure quality control during protein translocation at the level of the plasma membrane. Unfortunately, more than 15 years following the discovery of unconventional secretory processes (22,40,56), the field still awaits the molecular identification of the first membrane translocation-specific components of the export machineries of, for example, IL-1β, galectin-1, FGF-1/2 or Leishmania HASPB which is the main roadblock for the design of experiments to elucidate the mechanistic details of these processes. New tools in gene silencing such as genome-wide RNAi-screening procedures in combination with the in vivo reconstitution of unconventional secretory processes are likely to overcome this situation and, eventually, will allow us to analyze the molecular mechanisms involved.


I thank Britta Brügger (Heidelberg University Biochemistry Center), Blanche Schwappach (Zentrum für Molekulare Biologie, University of Heidelberg), Oliver Fackler (Department of Virology, University of Heidelberg), Jürgen Bernhagen (University Hospital RWTH Aachen) and Markus Künzler (ETH Zürich) for critical comments on the manuscript as well as Christophe Zehe and Carolin Stegmayer from my laboratory for many helpful discussions. Work in the laboratory of the author is supported by grants from the German Research Council (SFB 638, Project A8 and SFB 544, Project B15) and the Ministry of Science, Research and the Arts of the State of Baden-Württemberg.