Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
Most of the examples of protein translocation across a membrane (such as the import of classical secretory proteins into the endoplasmic reticulum, import of proteins into mitochondria and peroxisomes, as well as protein import into and export from the nucleus), are understood in great detail. In striking contrast, the phenomenon of unconventional protein secretion (also known as nonclassical protein export or ER/Golgi-independent protein secretion) from eukaryotic cells was discovered more than 10 years ago and yet the molecular mechanism and the molecular identity of machinery components that mediate this process remain elusive. This problem appears to be even more complex as several lines of evidence indicate that various kinds of mechanistically distinct nonclassical export routes may exist. In most cases these secretory mechanisms are gated in a tightly controlled fashion. This review aims to provide a comprehensive overview of our current knowledge as a basis for the development of new experimental strategies designed to unravel the molecular machineries mediating ER/Golgi-independent protein secretion. Beyond solving a fundamental problem in current cell biology, the molecular analysis of these processes is of major biomedical importance as these export routes are taken by proteins such as angiogenic growth factors, inflammatory cytokines, components of the extracellular matrix which regulate cell differentiation, proliferation and apoptosis, viral proteins, and parasite surface proteins potentially involved in host infection.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Soluble secretory proteins typically contain N-terminal signal peptides that direct them to the translocation apparatus of the endoplasmic reticulum (ER) . Following vesicular transport from the ER via the Golgi to the cell surface, lumenal proteins are released into the extracellular space by fusion of Golgi-derived secretory vesicles with the plasma membrane [2–5]. This pathway of protein export from eukaryotic cells is known as the classical or ER/Golgi-dependent secretory pathway. However, more than 10 years ago, it was reported that interleukin 1β (IL1β) and galectin-1 (also referred to as L-14) could be exported from cells in the absence of a functional ER/Golgi system [6,7]. Since then, the list of proteins demonstrated to be secreted by unconventional means is steadily growing. Figure 1 gives an overview of cellular, viral and parasitic proteins that have been shown to be exported by mechanisms that are operational in the absence of a functional ER/Golgi system. The basic observations (summarized previously in [8,9]) that led to the proposal of alternative pathways of eukaryotic protein secretion are (a) the lack of conventional signal peptides in the secretory proteins in question, (b) the exclusion of these proteins from classical secretory organelles such as the ER and the Golgi combined with the lack of ER/Golgi-dependent post-translational modifications such as N-glycosylation and (c) resistance of these export processes to brefeldin A, a classical inhibitor of ER/Golgi-dependent protein secretion [10–12]. Because the secretory proteins discussed here are soluble factors synthesized on free ribosomes in the cytoplasm, various experimental strategies have been pursued in order to exclude unspecific release based on cell death under the experimental conditions applied. As described in detail in the following sections, these experiments included parallel quantitative measurements of the appearance of unrelated cytoplasmic proteins in cellular supernatants [8,9] as well as the identification of mutants that are deficient in nonclassical export . Moreover, nonconventional protein secretion was shown to be dependent on both energy and temperature and is stimulated or inhibited by various treatments [8,9]. Finally, nonconventional protein secretion processes were shown to be regulated for example by cell differentiation [7,14], NF-κB-dependent signalling pathways , and post-translational modifications such as phosphorylation . Based on these observations, it has to be concluded that the secretory proteins discussed in this review exit eukaryotic cells in a controlled manner mediated by proteinaceous machineries. In the following sections, the various cargo proteins known to be secreted by unconventional means will be discussed in detail.
Cytokines: interleukin-1β, thioredoxin and macrophage migration inhibitory factor
In 1987, Dinarello and colleagues demonstrated that interleukin 1, a cytokine [17,18] lacking a classical signal peptide for ER/Golgi-mediated protein secretion , is exported from activated human monocytes . Two isoforms of interleukin 1 termed IL-1α and IL-1β have been described which represent proteolytically processed forms derived from two related but distinct precursors . The processing of IL-1α involves myristoylation and, following insertion into the plasma membrane, calpain-dependent cleavage that is thought to cause release of the mature form of IL-1α into the extracellular space [21,22]. In the case of IL-1β, interleukin-converting enzyme produces mature IL-1β[23,24], which is then exported .
Most studies targeted toward the molecular mechanism of interleukin 1 export have been focused on the β-isoform. A detailed molecular analysis of the export process revealed that IL-1β does not make use of an unconventional pathway of translocation into the lumen of the endoplasmic reticulum but rather appears to utilize a secretory mechanism independent of ER/Golgi-related vesicular transport . This process was shown to be distinct from unspecific release as for example only the processed form of IL-1β (17 kDa) can be detected in cellular supernatants whereas the precursor (33 kDa) is retained by IL-1β-expressing cells . Moreover, under the experimental conditions applied, only the β-isoform was found to be secreted, whereas the α-isoform could not be detected in cellular supernatants . However, despite apparently utilizing a distinct secretory mechanism, it was later found that IL-1α is also exported .
Though IL-1β is found in certain intracellular vesicles, as judged by protease protection experiments, these structures appear to be unrelated to the ER/Golgi system as IL-1β secretion was not inhibited but rather stimulated by brefeldin A, a drug that compromises the structure and function of the Golgi apparatus [10–12]. Consistently, IL-1β was found not to be glycosylated, despite bearing corresponding consensus sequences. Intracellular vesicles proposed to play a role in IL-1β secretion have been shown to be related to an endolysosomal compartment that releases its content upon fusion with the plasma membrane . These observations are consistent with the fact that IL-1β secretion is sensitive to methylamine , a drug that disturbs endocytosis . Based on pharmacological studies employing the sulfonylurea glyburide (10 µm) along with expression-inhibition studies employing antisense techniques, an ABC transporter, ABC1, has been implicated in the overall process of IL-1β secretion [27,28] and therefore might mediate IL-1β translocation from the cytoplasm to the lumen of the endolysosomal compartment. Interestingly, glyburide also appears to inhibit nonclassical secretion of macrophage migration inhibitory factor (J. Bernhagen, RWTH Aachen, Germany, personal communication), an inflammatory cytokine mediating a number of immune and inflammatory diseases, e.g. bacterial septic shock [29–31]. The potential function of ABC transporters in these processes might be related to that of bacterial ABC transporters that mediate protein secretion of, for example, hemolysin [32–34].
Thioredoxins are ubiquitous intracellular enzymes that catalyze thiol-disulfide exchange reactions . Additionally, extracellular populations of thioredoxin have been detected that, similar to IL-1β and migration inhibiting factor, follow an ER/Golgi-independent route of secretion [36–39]. This observation is consistent with additional physiological roles of thioredoxin such as its function as a mitogenic cytokine that requires extracellular localization [40,41]. Secretion of thioredoxin appears to be mediated by a pathway distinct from IL-1β as it could neither be detected in intracellular vesicles, nor was the secretion process reported to be inhibited by reagents that interfere with the function of ABC transporters. However, as with IL-1β, secretion of thioredoxin is inhibited by methylamine and stimulated by brefeldin A . Interestingly, the redox state of thioredoxin does not influence its unconventional export .
Pro-angiogenic growth factors: FGF-1 and FGF-2
Fibroblast growth factor 1 and 2 (FGF-1 and FGF-2) belong to a large family of heparin-binding growth factors  that, apart from their mitogenic activity [43,44], are key activators of tumor-induced angiogenesis . The majority of the members of the FGF family are exported by ER/Golgi-dependent secretory transport. However, FGF-1 and the 18 kDa isoform of FGF-2 have been shown to be secreted by an alternative pathway [46–48]. While it was first assumed that angiogenic growth factors might be released from mechanically injured tissue to promote wound healing , a process that requires angiogenesis, various lines of evidence demonstrate that FGF-1 and FGF-2 are exported from cultured cells in the absence of appreciable amounts of cell death [46–48,50,51]. Like IL-1β, FGF-1 is increasingly secreted under stress conditions such as heat shock treatment [46,52]. In contrast, FGF-2 export is not affected under these experimental conditions . While serum starvation has been reported to inhibit export of both FGF-2  and IL-1β, it was found to actually induce secretion of FGF-1 . Similarly, methylamine has been found to block export only of FGF-2  and IL-1β with no apparent effect on FGF-1 export . Recently, it was reported that expression of the IL-1α precursor inhibits FGF-1 release in response to temperature stress . In contrast, expression of the mature form of IL-1α did not affect FGF-1 export, suggesting that IL-1α processing is somehow related to FGF-1 biogenesis. However, whether FGF-1 and IL-1α utilize similar export mechanisms remains an open question.
These observations point to some common characteristics in the export of the cargo proteins discussed, but it seems unlikely that one and the same machinery mediates secretion of these factors. Consistent with this view, IL-1β has been reported to be secreted by a vesicular nonclassical export pathway [6,25], while FGF-1 and FGF-2 are likely to be directly translocated from the cytoplasm into the extracellular space (Fig. 1). While there is also one report pointing to a role of large granules involved in FGF-2 export based on immuno-EM analysis of mast cells , this issue remains controversial as intracellular FGF-2 has been localized to the cytoplasm in many FGF-2-secreting cell types with no apparent localization in vesicular structures [50,57–59]. Similar findings have been reported for FGF-1 [60–62].
With regard to the protein components involved in the overall processes of nonclassical export pathways, most is known about the secretion of FGF-1. As noted above, FGF-1 export is significantly increased in response to stress conditions such as heat shock treatment  and serum starvation . Based on these experimental conditions, it was shown that secreted FGF-1 isolated from cell culture supernatants represents a latent (inactive form that can be reactivated) homodimer  that can also be formed upon chemical oxidation of FGF-1 in vitro. These observations led to the discovery of a specific cysteine residue (Cys30) in FGF-1 that is required for both dimer formation and nonclassical export of FGF-1 [13,54]. Upon heat shock treatment, two intracellular proteins have been shown to associate with the latent FGF-1 homodimer in the cytoplasm. These are a cleavage product of the transmembrane protein synaptotagmin consisting of its cytoplasmic domain (p40-Syt1) and the Ca2+-binding protein S100A13. Apparently, they are exported together with FGF-1 [64–66]. A direct role of p40-Syt1 and S100A13 in FGF-1 export has been proposed as both repression of p40-Syt1 expression by antisense techniques and the expression of a dominant-negative S100A13 mutant attenuate FGF-1 export [64,66]. As with FGF-1 dimer formation , oxidation by Cu2+ cations has been demonstrated to trigger the formation of a complex consisting of FGF-1, p40-Syt1 and S100A13 . Consistent with the view that p40-Syt1 and S100A13 are involved in the export of FGF-1, tetrathiomolybdate, a Cu2+ chelator, has been shown to inhibit heat shock-induced FGF-1 export . More recently, stress-induced formation of the intracellular FGF-1–p40-Syt1–S100A13 complex has been demonstrated to cause a redistribution of cytoplasmic FGF-1 to the inner surface of the plasma membrane . These results suggest that FGF-1–p40-Syt1–S100A13 complex formation is the first step in the FGF-1 export pathway, followed by direct translocation of this protein complex across the plasma membrane. However, the machinery that mediates membrane translocation of this protein complex remains unknown.
Compared to FGF-1 export, much less is known about the mechanism and the role of specific proteins with regard to the overall process of FGF-2 export from mammalian cells. To date the only protein that has been proposed to play a role in FGF-2 export is the Na+/K+-ATPase . This conclusion was based initially on the observation that cardiac glycosides such as ouabain partially inhibit FGF-2 export [50,68,69]. This was further strengthened by experiments demonstrating that the expression of an ouabain-resistent α-subunit mutant of the Na+/K+-ATPase rescues FGF-2 export in the presence of ouabain . Moreover, a direct or indirect physical interaction between the α subunit and FGF-2 has been detected based on coimmunoprecipitation though this association could only be observed upon co-overexpression of both proteins . Together with the result that overexpression of the α subunit interferes with FGF-2 export , these observations are reasonably supportive of a role for the Na-K-ATPase in the overall process of FGF-2 export. On the other hand, ouabain treatment (typically used at 10–100 µm) causes only partial inhibition of FGF-2 export, whereas concentrations of ouabain of less than 5 µm (IC50 ≈ 1 µm) completely inhibit the ATP-dependent translocation of cations catalyzed by the Na+/K+-ATPase [71,72]. Interestingly, the membrane potential generated by the Na+/K+-ATPase is not required for FGF-2 export . Based on these observations, it has been proposed that the α/β heterodimers that constitute a functional Na+/K+-ATPase in terms of ion transport might be able to form higher ordered complexes that catalyze FGF-2 export in a membrane potential-independent manner . Alternatively, the α subunit alone might associate with other so far uncharacterized factors as part of a novel complex that mediates FGF-2 export . Unfortunately, no progress has yet been made in identifying such molecular structures.
Galectins: components of the extracellular matrix
The members of the galectin protein family are abundant β-galactoside-specific lectins of the extracellular matrix implicated in many cellular processes such as regulation of cell proliferation, differentiation and apoptosis [73–76]. The best characterized members of this family are galectin-1 and galectin-3 which are present as soluble proteins in the cytoplasm in a wide range of vertebrate cell lines and tissues [7,14,77–81]. Secreted galectins are found either bound to the extracellular surface of the plasma membrane or as abundant components of the extracellular matrix [7,14,77,79–81]. Cell surface association of galectins is mediated by both N- and O-glycosylated β-galactose-terminated oligosaccharide side chains of glycoproteins [9,73] as well as by galactose-containing glycolipids such as GM1[73,82]. As galectin-1 and galectin-3 can form homodimers [9,83,84], it has been proposed that secreted galectins affect their glycosylated cell-surface counter receptors by inducing conformational changes of their extracellular domains and/or by clustering galectin counter receptors based on noncovalent crosslinking of oligosaccharide moieties . In this way, secreted galectins are thought to affect processes such as cell differentiation by cell surface counter receptor-mediated signalling [73,85]. While classical counter receptors of, for example, galectin-1 include laminin , fibronectin  and cell-type specific receptors such as T cell CD43 and CD45 , it has been shown more recently that the tumor-specific cell surface antigen CA125 also represents a galectin counter receptor that preferentially binds galectin-1 . This latter example is of particular interest as it provides a potential molecular mechanism for how tumor cells can differentially interact with the extracellular matrix, a process crucial for tumor progression.
Similar to interleukin 1β, FGF-1 and FGF-2, galectins apparently do not contain signal peptides in their primary structure suitable for ER/Golgi-mediated secretion . Consistently, galectins are synthesized on free ribosomes in the cytoplasm  and galectin secretion has been shown not to be blocked by inhibitors of the ER/Golgi-dependent pathway such as brefeldin A and monensin [9,80,90]. Unlike interleukin 1β, galectin-1 and galectin-3 do not appear to be packaged into intracellular vesicles prior to export [7,9,80,81]. Rather, galectin-1 and galectin-3 have been shown to accumulate directly below the plasma membrane, followed by an export mechanism that appears to involve the formation of membrane-bound vesicles (also called exosomes but not to be confused with structures involved in RNA processing ) that pinch off before being released into the extracellular space [7,9,80,81]. This mechanism also distinguishes galectin export from FGF-1 and FGF-2 export, as there is no evidence that these proteins are packaged into exosomes (see above). An engineered version of galectin-3 containing an N-terminal acylation motif derived from a protein tyrosine kinase (p56lck) has been shown to be secreted more efficiently than wild-type galectin-3 . These results indicate that targeting to the plasma membrane is a rate-limiting step in galectin secretion. However, there is no information describing exactly what causes galectin-1 and galectin-3 to accumulate at specific spots underneath the plasma membrane, and what actually causes the formation of exosomes into which these proteins appear to be packaged in an active fashion.
Other secretory proteins exported by nonconventional means
HIV-Tat, Herpes simplex VP22 and foamy virus Bet
Besides the classical examples of ER/Golgi-independent protein secretion described above, a whole variety of proteins has been reported to be secreted by nonconventional means. Among them are many factors whose localization-dependent functions, akin to those noted above, are of tremendous biomedical importance. Such proteins include virus-encoded factors that are critical for the viral replication cycle. The most prominent example is HIV-Tat, one of the auxiliary proteins required by HIV in addition to structural and enzymatic proteins to replicate its genome . HIV-Tat has been shown to be released from both HIV-infected and HIV-Tat-transfected cells in the absence of appreciable amounts of cell death [93,94]. Intriguingly, HIV-Tat contains a region in its primary structure termed the basic transduction domain that appears to enable the protein to traverse membranes [95,96]. The molecular mechanism of this translocation process does not seem to involve a proteinaceous machinery as another HIV-Tat-like protein transduction domain, the antennapedia third helix domain , has been shown to cross artificial protein-free membranes . Another unusual feature of protein transduction domains is their apparent ability to translocate across membranes even at 4 °C [96,98], an observation consistent with a membrane translocation mechanism independent of proteinaceous machinery. In all cases, however, protein transduction domains appear to function in unconventional modes of protein uptake by mammalian cells. Specifically, in the cases of HIV-Tat and Herpes simplex tegument protein VP22, it is rather unlikely that their ER/Golgi-independent export mechanisms are based on their protein transduction domains. For example, HIV-Tat secretion from cultured cells is a temperature-dependent process . Thus, protein transduction-dependent uptake and ER/Golgi-independent export of protein transduction domain-containing factors might be mechanistically distinct processes. Similar to herpes simplex VP22, a secreted auxiliary protein termed Bet encoded by foamy viruses  has been shown to spread between cultured cells . Interestingly, both VP22 and Bet are found in the cytoplasm of expressing cells whereas they are targeted to the nucleus of cells that received the protein by intercellular spreading [98,100]. In both cases, this process is not affected by brefeldin A, suggesting that export of VP22 and Bet from expressing cells does not involve the ER/Golgi system [98,100]. In conclusion, it appears likely that uptake by mammalian cells of HIV-Tat, VP22 and possibly Bet is mediated by transduction domains in a temperature-insensitive manner whereas export is mediated in a temperature-sensitive manner by proteinaceous machineries that are insensitive to brefeldin A.
Another quite remarkable example of nonclassical protein export from eukaryotic cells is the mechanism of cell surface expression of Leishmania HASPB (hydrophilic acylated surface protein B) which is found associated with the outer leaflet of the plasma membrane only in the infectious stages of the parasite lifecycle [101,102]. The protein is synthesized on free ribosomes in the cytoplasm and becomes both myristoylated and palmitoylated at its N-terminus, which is the molecular basis of how HASPB is anchored in the membrane . Mutational analysis revealed that an HASPB construct lacking its 18 N-terminal amino acids is redistributed into the cytoplasm . The same is true for a mutant that retains the N-terminus but lacks the myristoylation site . Interestingly, a mutant that lacks the palmitoylation site but continues to be myristoylated has been found associated with the cytoplasmic surface of the Golgi apparatus . Based on these observations, a model has been proposed in which HASPB is transferred from the cytoplasm to the outer leaflet of the Golgi membrane, from where it is transported to the plasma membrane via conventional vesicular transport. This process would insert HASPB into the inner leaflet of the plasma membrane. At present it is completely unclear how HASPB is then translocated across the membrane, resulting in the insertion of the two acyl chains in the outer leaflet of the plasma membrane. Intriguingly, heterologous expression of various HASPB fusion proteins in mammalian cells revealed the existence of a machinery that is capable of translocating the protein across the plasma membrane , demonstrating a conserved pathway among lower and higher eukaryotes. No endogenous mammalian cargo proteins that make use of this type of export system have been identified.
Homeodomain-containing transcription factors and HMG (high mobility group) chromatin-binding proteins
As another example of nonclassical protein export, two classes of proteins involved in the overall process of regulated gene transcription have been proposed to operate as extracellular factors even though they are normally localized to the nucleus of mammalian cells [104–106]. For the transcription factor Engrailed homeoprotein isoform 2 (En2), a potential paracrine signalling activity was postulated as a subpopulation of En2 has been localized to the cell periphery in caveolae-like structures . In addition, a small but significant portion of total cellular En2 was found to reside in membrane-bound vesicles as judged by protease protection experiments . Therefore, it was reasoned that En2, despite lacking a conventional ER signal peptide, might be secreted at a certain rate. This hypothesis was tested experimentally by coculturing COS cells expressing the chicken orthologue of En2 (cEn2) with rat primary neurons demonstrating intercellular transfer of cEn2 . Interestingly, not only export from COS cells but also import by cocultured primary neurons appears to rely on nonclassical mechanisms, as cellular uptake of cEn2 was shown to depend on an unusual WF motif in position 48–49 . This sequence is also required for the uptake of other homeodomain-containing proteins [107–109]. The internalization of homeodomain-containing proteins apparently differs from classical endocytosis, as it seems to occur by direct translocation across the plasma membrane . This process might be similar to the uptake mechanism of some viral proteins such as HIV-Tat and Herpes simplex VP22 as discussed above .
About 5% of total cellular En2 becomes externalized by COS cells which is about the portion that is also found to be protected against protease treatment. An 11-amino acid sequence within the homeodomain of En2 has been identified that, when removed, causes a block in export of the corresponding mutant protein . This phenotype correlates with the disappearance of the mutant protein from the protease-protecting organelle, which probably represents a kind of a secretory compartment . The homeodomain-derived peptide was later shown to be part of a nuclear export signal and therefore promotes retrotranslocation of En2 from the nucleus into the cytoplasm . These results have been taken to mean that retrotranslocation of En2 from the nucleus to the cytoplasm is a prerequisite for nonclassical export of En2 . While the homeodomain-derived peptide was originally thought to represent a signal for nonclassical export, this view has to be re-evaluated as it might only trigger cytoplasmic localization of En2 and may not be required afterwards for externalization of En2.
HMG proteins are intranuclear factors that mediate the assembly of site-specific DNA-binding proteins within chromatin . As a surprising finding, but similar to the homeodomain-containing transcription factors described above, HMGB1 is secreted during certain physiological processes such as inflammation. Specifically, monocytes have been shown to export HMGB1 upon stimulation with bacterial lipopolysaccharides . Because antibodies against HMGB1 suppress LPS-induced endotoxemia, and injection of HMGB1 protein into mice causes toxic shock, HMGB1 apparently acts as a mediator of endotoxin lethality in mice . Interestingly, HMGB1 export competence appears to be a special property of a limited number of cell types (such as monocytes and macrophages) as many cell types including lymphocytes are not capable of secreting HMGB1 . Again, similar to homeodomain-containing transcription factors, extracellular HMGB1 has also been shown to act as both an autocrine and paracrine signalling molecule promoting differentiation processes of the HMGB1-secreting cell [113,114] or other cells nearby .
As with all the examples of unconventional protein secretion discussed in this review, HMGB1 does not contain a signal peptide for translocation into the ER . Similar to IL-1β, FGF-2 and galectin-3 , a rise in intracellular Ca2+ triggers HMGB1 export [114,115]. Akin to the mobilization of homeodomain-containing transcription factors, HMGB1 has been observed to redistribute from the nucleus to the cytoplasm upon activation of monocytes . A detailed ultrastructural analysis revealed that redistributed HMGB1 localizes to an endolysosomal compartment from which secretion can be triggered by stimuli known to promote lysosomal exocytosis . These characteristics are strikingly similar to the process of IL-1β secretion . However, IL-1β secretion from monocytes can be triggered by adding exogenous ATP, whereas HMGB1 release is induced by lysophosphatidylcholine. Moreover, the kinetics of IL-1β and HMGB1 release from monocytes differ significantly, with IL-1β being secreted early after monocyte activation and HMGB1 at a later stage. IL-1β consistently acts at an early phase of inflammation whereas HMGB1 functions as a late mediator of inflammation (see above). These results have been taken to indicate that lysosomal exocytosis might involve distinct populations of endolysosomal vesicles, thereby allowing different kinetics of cargo release .
Direct translocation of proteins from the cytoplasm into the lumen of lysosomes has been reported  but this pathway appears to function primarily for enhanced degradation of these proteins . The corresponding targeting motif KFERQ  is not found in the primary structure of IL-1β, En2 or HMGB1. It therefore appears more likely that these factors are translocated by a different mechanism. A potential candidate is ABC1, an ABC transporter that has been implicated to play a role in the overall process of IL-1β secretion [27,28].
Cytoplasmic clearance of unfolded proteins by nonclassical secretion
The mitochondrial matrix protein rhodanese, a monomeric sulfotransferase, that, following synthesis on free ribosomes in the cytoplasm, is normally imported into mitochondria, represents another unusual example of nonclassical protein export from mammalian cells. When overexpressed in HEK-293 cells from a strong viral promotor, about 40% of total rhodanese was found to be secreted into the culture medium . Export was shown to occur in the absence of appreciable amounts of cell death and to depend on neither the mitochondrial targeting sequence of rhodanese nor a functional ER/Golgi system . Based on the observation that rhodanese acquires its enzymatic activity only after import into the mitochondrial matrix (and that the signal peptide cannot be an inhibitor of enzymatic activity as it is not cleaved off in the matrix), it was concluded that the population present in the cytoplasm remains unfolded before import into mitochondria. Therefore, it has been postulated that the export pathway detected for rhodanese represents a mechanism for clearing the cytoplasm of unfolded proteins that apparently accumulate upon overexpression . More recently, a potentially similar example of cytoplasmic clearance of an unfolded protein population possibly generated by overexpression has been observed . In this case, an unfolded subpopulation of transiently overexpressed GFP was found to be secreted in a brefeldin A-insensitive manner. This effect has not been observed in stable cell lines that express moderate levels of GFP in a doxicycline-dependent manner . However, these different observations are not necessarily inconsistent as in the latter case an unfolded population of GFP is unlikely to exist. Interestingly, methylamine and other drugs known to inhibit nonclassical export of substrates such as IL-1β, FGF-2, thioredoxin, and the galectins (; see above) do not block externalization of rhodanese or unfolded GFP [119,120], again suggesting the existence of distinct molecular mechanisms of unconventional protein secretion.
Targeting motifs and regulation of nonclassical protein export
In many cases of intracellular protein sorting, short, linear amino acid sequences have been identified that serve as sorting motifs, including N-terminal signal peptides for ER translocation and the N-terminal targeting signals of mitochondrial proteins [118,121]. Currently, very limited information is available about motifs directing proteins to the various pathways of unconventional protein secretion described above. The most defined one is that of Leishmania HASPB which consists of a linear sequence of 18 amino acids at the extreme N-terminus referred to as HASPB-N18 . This sequence is both necessary and sufficient to direct a corresponding fusion protein to the HASPB export pathway in both parasites and mammalian cells. HASPB-N18 is myristoylated at a glycine residue in position 2 and palmitoylated at a cysteine residue in position 5. HASPB externalization requires that both residues are acylated. However, a construct termed HASPB-N10, which contains both acylation sites but lacks the amino acids 11–18, fails to translocate across the plasma membrane . These results suggest that acylation might only be required to initially insert the protein into the membrane, and the translocation that follows requires an interaction of the proteinaceous part of HASPB-N18 with the putative export machinery. Based on these characteristics, the HASPB export pathway appears to be unrelated to other examples of nonclassical protein export described here. As the pathway is functional in mammalian cells, endogenous substrates are likely to exist. However, the 18-amino acid sequence found at the N-terminus of HASPB is not only absent from other secretory proteins exported by unconventional means but is also not found in any mammalian protein.
Akin to Leishmania HASPB, the N-terminus of galectin-3 has been proposed to contain targeting information for nonclassical export [122,123]. When the first 120 amino acids of galectin-3 are deleted, the residual portion of the protein is no longer secreted. Conversely, addition of this N-terminal segment to a cytosolic protein directs the corresponding fusion protein to the galectin-3 export pathway. A short sequence comprising residues 89–96 (based on the hamster amino acid sequence) was identified that, upon deletion, causes a breakdown of galectin-3 export. However, the addition of this small peptide to a cytosolic protein is not sufficient to direct the resulting fusion protein to the galectin-3 export pathway suggesting that, besides the critical role of this short segment, other determinants for nonclassical export exist in the N-terminal part of galectin-3 . When compared to the galectin-1 amino acid sequence, no significant homologies can be found within the N-terminal 120 amino acids of galectin-3.
In contrast to HASPB and galectin-3, the C-terminal half of FGF-1 has been implicated in its temperature stress-induced release . A domain comprising a stretch of amino acids from position 83–154 (based on the human FGF-1 orthologue) appears to prevent the protein from entering the nucleus, which has been suggested to be a prerequisite for unconventional export. When the corresponding domain of FGF-2 was transferred to FGF-1, secretion of the resulting hybrid protein was no longer observed. These data have been taken to mean that FGF-1 and FGF-2 are exported by distinct pathways , which is consistent with the observation that only FGF-1 release can be triggered by temperature stress [46,54]. The actual targeting motifs for nonclassical export have not been revealed for either FGF-1 or FGF-2.
For the homeodomain-containing transcription factor En2, it has been suggested that an 11-amino acid motif within the homeodomain may function as a signal for nonclassical export . As discussed above, this sequence was later found to be part of a nuclear export signal suggesting that nuclear export of En2 is a prerequisite for its unconventional secretion . Therefore, it is rather unlikely that this signal is required for the export process of En2. Interestingly, En2 has been shown to be a substrate for protein kinase CK2 which, upon phosphorylation of En2 within a serine-rich domain, causes attenuation of En2 secretion . At this point, it is not clear whether this segment of En2 (residues 146–169) is part of a signal sequence for nonclassical export or whether this domain regulates access of En2 to its export pathway. In either case, the information for En2 export must lie within the En2 homeodomain, as this part of the protein alone is an efficient substrate for intercellular transfer . Phosphorylation-dependent regulation might be a general principle for the regulation of intercellular transfer, at least for a subset of these cargo proteins, as it has also been suggested to play a role in the intercellular transfer of VP22 [125,126].
Similar to En2 export, many of the proteins described here are exported in a regulated fashion. For example, IL-1β and HMGB1 can be released from monocytes upon stimulation with reagents that induce an inflammatory response [6,104]. At the same time, En2, IL-1β and HMGB1 are those factors among unconventionally secreted proteins that appear to be exported from an endosomal subcompartment [25,106,116], which might be interpreted as some kind of storage mechanism from which regulated secretion of these factors can be triggered. On the other hand, as discussed above, nonclassical export of for example FGF-1 and galectin-1 are also regulated inducible processes, yet there is no evidence that these factors are packaged into intracellular vesicles prior to secretion. The export process of galectin-1 has been shown to be regulated based on cell differentiation. For example, during the differentiation of muscle cells a massive increase of galectin-1 export has been observed in correlation with the transition from myoblasts to myotubes . Similarly, galectin-1 export from the leukemia cell line K-562 can be stimulated by the addition of differentiation-inducing agents such as erythropoietin . It appears likely that differentiation-correlated galectin-1 export requires the synthesis of specific proteins, which is in line with the time delay of induced galectin-1 export. Similarly, FGF-2 export has recently been shown to be triggered by the expression of the Epstein–Barr virus protein LMP-1 . This mechanism requires a functional NF-κB signalling pathway , possibly indicating that LMP-1-mediated stimulation of FGF-2 export involves the induction of a protein machinery based on de novo synthesis.
As illustrated in Fig. 1, at least four distinct pathways of unconventional protein secretion exist in mammalian cells that are fully functional in the absence of an intact ER/Golgi system. Various fundamental questions arise from these observations such as why do mammalian cells actually need additional secretory mechanisms besides the classical pathway? As noted previously, for the galectins it is relatively obvious that the alternative secretory pathway prevents their premature binding to glycolipids and glycoproteins within the lumen of the classical secretory pathway . However, in other cases it is less clear why these cargo proteins are exported by unconventional means. Other fundamental questions are: What are the molecular components that drive various mechanisms of nonclassical export? Why do proteins such as HMGB1 with completely unrelated functions also serve as paracrine signalling molecules upon unconventional release into the extracellular space? The answers to these questions are of exceptional interest as the cargo proteins secreted by unconventional means are factors whose biological functions are of tremendous importance to biomedical research. For example, FGF-2 has been identified as a major target protein for the development of antiangiogenic drugs, as it has been shown that inhibitors of ternary complex formation between FGF-2 and its high and low affinity receptors  on the surface of target cells display antiangiogenic activity in vivo. Similarly, unconventional cell-surface expression of HASPB by Leishmania parasites appears to be tightly correlated with host cell infection [101,102] and therefore the HASPB export pathway might be an excellent target for the development of drugs against tropical and subtropical diseases termed the leishmanias. These pathways are in general attractive targets, because it may be possible to identify inhibitors that do not interfere with the essential function of the classical secretory pathway. Therefore, elucidation of the molecular machineries controlling the various kinds of nonclassical export might provide a whole variety of novel target proteins suitable for drug design.
So far, a biochemical analysis of the molecular machineries of nonclassical protein export has proven difficult because in many cases the export process is relatively inefficient. Also, it is not yet clear whether unfolding of the various cargo proteins is required for unconventional secretion. If so, classical methods employing dihydrofolate reductase  domains to trap cargo proteins at the site of translocation could be used in order to allow a biochemical analysis of the export apparatus. It is also a major problem that, in most cases, only very limited information is available about the motifs that target cargo proteins to the nonclassical export routes. Recently, novel assays have been developed which deploy fluorescence activated cell sorting to reconstitute unconventional protein export pathways on a quantitative basis [50,79]. This new approach might facilitate genetic screens in mammalian cells and is compatible with systematic high throughput screening technologies for the identification of low molecular mass inhibitors of the processes described here. Thus, the elucidation of the molecular mechanisms of nonclassical protein export from eukaryotic cells will not only solve a fundamental problem in current cell biology but will also lead to the identification of novel target proteins, with great value for biomedical research.
I would like to thank Britta Brügger (Biochemie-Zentrum Heidelberg), Tracy LaGrassa (Biochemie-Zentrum Heidelberg), Blanche Schwappach (Zentrum für Molekulare Biologie Heidelberg), Jürgen Bernhagen (University Hospital RWTH Aachen) and Markus Künzler (ETH Zürich) for critical comments on the manuscript, as well as all members of my laboratory for helpful discussions. Work in the laboratory of the author is supported by grants from the German Research Foundation (DFG) and the Ministry of Science, Research and the Arts of the State of Baden-Württemberg.