The Disulfide-bonded Loop Is Essential for Sorting of CgB during Vaccinia Expression
In our study we provide biochemical and morphological evidence that the 22 amino acids of hCgB comprising the disulfide-bonded loop are essential for sorting of CgB to secretory granules and that they affect sorting at the level of the TGN. This result was accomplished using a vaccinia virus–based expression system which inhibits host cell protein synthesis. We have shown by different methods that under these conditions wt-hCgB but not Δcys-hCgB is sorted to secretory granules. First, sucrose gradient analysis showed a differential exit of wt-hCgB and Δcys-hCgB from the TGN. Whereas wt-hCgB was sorted mainly into ISG, which were shown to undergo maturation, Δcys-hCgB was found exclusively in CV (Fig. 10). The reduction in sorting efficiency of wt-hCgB in the vaccinia system (~70% versus ~100% after transfection) is most likely due to the shut-off of host cell protein synthesis, which could make factors involved in sorting rate-limiting. Consistent with this explanation, we observed that the sorting efficiency of virally expressed wt-hCgB decreased with increasing infection times (data not shown). Second, after pulse-chase labeling, Δcys-hCgB was secreted like the constitutive marker protein AT whereas wt-hCgB was stored ~4.5-fold more (Fig. 11). Third, immunofluorescence analysis after 90 min of cycloheximide treatment revealed a pattern for wt-hCgB, which was reminiscent of granule staining. For the deletion mutant, no signal was detected after 90 min of cycloheximide treatment reflecting its efficient secretion (Fig. 12). Furthermore, we have demonstrated that sorting of Δcys-hCgB to ISG can be rescued by coexpression with full-length CgB (Fig. 13), resembling the presence of endogenous granins, as in the case of transfection.
From the finding that the disulfide-bonded loop is necessary for the sorting of hCgB to secretory granules, we conclude that the disulfide-bonded loop comprises either a part or the entire sorting signal. This explanation is consistent with the data obtained by Chanat et al. (1993), who showed that reductive cleavage of the disulfide bond of CgB by DTT caused its missorting to the constitutive secretory pathway. However, the previous DTT data did not strictly exclude an effect of DTT on a component of the secretory machinery for CgB, rather than CgB itself. Our results obtained by expression of Δcys-hCgB clearly indicate that missorting is due to the mutation in CgB.
Sorting of Transfected Δcys-hCgB by Coaggregation
In contrast to the data obtained with the vaccinia expression system, Δcys-hCgB was efficiently sorted to the regulated pathway after transfection. This was determined by pulse-chase labeling (Fig. 2), sucrose gradient centrifugation (Fig. 3), depolarization-induced calcium-dependent release (Fig. 4), and confocal double immunofluorescence analysis (Fig. 5). How can one reconcile the opposing results for Δcys-hCgB obtained by the two different expression systems? The major difference between the two systems is the ongoing synthesis of endogenous proteins after transfection, in contrast to a shut-off of host cell protein synthesis during viral infection, resulting in the presence or absence of endogenous granins in the TGN (Fig. 14). Ongoing protein synthesis during transfection would allow the formation of aggregates between transfected and endogenous regulated secretory proteins enabling the former to enter the regulated secretory pathway by coaggregation. The fact that approximately one-third of Δcys-hCgB was missorted to CV after transfection may be explained by the formation of homotypic aggregates of the mutant. Because, under transfection, Δcys-hCgB is expressed approximately two to three times higher than endogenous granins (not shown), the probability of homotypic Δcys-hCgB aggregates is increased. Indeed the absence and presence of rescuing molecules (full-length CgB) as the principal difference between the two expression systems is clearly shown by coexpression in the vaccinia system of Δcys-hCgB with full-length CgB, which rescues sorting of the mutant to secretory granules (Fig. 13). This implies that sorting of the deletion mutant is accomplished by coaggregation with loop-bearing CgB, i.e., endogenous rCgB in the case of transfection and recombinant wt-hCgB in the vaccinia system (Fig. 14).
This explanation is supported by the demonstration of heterophilic interactions between regulated secretory proteins in vitro (Gorr et al., 1989; Palmer and Christie, 1992; Colomer et al., 1996). In particular, it was shown that CgB interacts with CgA at low pH (Yoo, 1996). Because the aggregative properties of CgB are maintained after reduction of the disulfide bond (Chanat et al., 1994) and deletion of the disulfide-bonded loop (A. Krömer and H.-H. Gerdes, unpublished observations), Δcys-hCgB, like wt-hCgB, can be expected to interact with other regulated secretory proteins. Taken together, our data suggest that Δcys-hCgB lacking the disulfide-bonded loop, when expressed by transfection, is sorted to the regulated pathway of secretion by coaggregation with endogenous regulated secretory protein, the latter bearing the sorting signal.
Constitutive-like Secretion in PC12 Cells
Gradient analysis of transfected PC12 cells showed that wt-hCgB and Δcys-hCgB pulse labeled for 5 min with [35S]sulfate exited completely within 12 min from the TGN into post-TGN vesicles. At the end of a 12-min chase, transfected wt-hCgB was detected exclusively in ISG, whereas Δcys-hCgB was found in ISG and to a minor extent in CV (Fig. 3), indicating that ~30% of the mutant had entered the constitutive pathway of secretion upon exit from the TGN. These data are consistent with a role of the disulfide-bonded loop in sorting to ISG. We assume that the proportion of Δcys-hCgB entering CV did not coaggregate with endogenous granins but formed homotypic aggregates (Fig. 14). After 90 min of chase, a time point at which ISG have largely been converted to mature secretory granules (t1/2 ~45 min; Tooze et al., 1991), 70.9 and 40.8% of transfected wt-hCgB and Δcys-hCgB, respectively, were stored in secretory granules (Fig. 2 C). Comparison of the storage efficiencies determined after 90 min of chase with the proportion sorted to ISG after 12 min suggests that a similar proportion of both proteins was secreted in an unstimulated manner after entering the ISG. Two mechanisms, basal release of ISG (Matsuuchi and Kelly, 1991) and constitutive-like secretion (Kuliawat and Arvan, 1992), have been described which could account for this effect. For PC12 cells it has been shown, following rSgII as a marker protein, that basal release occurred at a very low level if at all (Gerdes et al., 1989; Tooze and Huttner, 1990). Similar results were obtained in this study for transfected PC12 cells (Seff of rSgII: 93.4% ± 2.7 SD). Therefore, it is unlikely to be basal release but rather constitutive-like secretion that accounts for unstimulated secretion of wt-hCgB and Δcys-hCgB between 12 and 90 min of chase. This explanation is consistent with data obtained from pancreatic islet cells, in which constitutive-like secretion was restricted to the maturation period of ISG (Kuliawat and Arvan, 1992). During vaccinia expression, unstimulated secretion of wt-hCgB (sorting into ISG ~65% versus Seff ~32%) from ISG was observed to a similar extent as in the case of transfection (sorting into ISG ~100% versus Seff ~71%). This suggests that in both systems, transfection and infection, the difference between sorting and storage of wt-hCgB is caused by constitutive-like secretion.
Sorting during Host Cell Synthesis Shut-off
The observation that virally expressed wt-hCgB is sorted and packaged into secretory granules under host cell protein synthesis shut-off allows the following conclusions: First, sorting of wt-hCgB to secretory granules is independent of other regulated secretory proteins, i.e., it carries all signals necessary for sorting to the regulated pathway of secretion. Second, involvement of chaperones with a short half-life in the secretory pathway assisting transport or aggregation of wt-hCgB is unlikely. Third, recruitment of other components necessary for granule formation, e.g., membrane proteins and cytosolic factors is, at least for the infection time analyzed, mostly independent of ongoing protein synthesis, and thus can be obtained from preexisting pools.
In the past, the vaccinia virus has been used frequently as a mammalian expression system because of its ability to infect a wide spectrum of cell types yielding high expression levels of the recombinant proteins (Hruby et al., 1986). In particular, processing and sorting of prohormones and proneuropeptides have been studied by infecting fibroblast-like cells, which have only a constitutive pathway of protein secretion, and AtT20 cells, with recombinant vaccinia viruses usually at very low viral multiplicity (Thomas et al., 1986; Edwards et al., 1988; Thomas et al., 1988; Seethaler et al., 1991). However, inhibition of host cell protein synthesis was not analyzed under the conditions used. Because the degree of inhibition varies significantly with viral multiplicity (Moss, 1968) and with cell type (Bablanian et al., 1978), it remains unclear whether processing and sorting was determined in the presence or absence of host cell protein synthesis.
Secretory granule biogenesis under protein synthesis arrest was subject of a study by Moore and co-workers (Brion et al., 1992). Using cycloheximide to block protein synthesis and analyze glycosaminoglycan chain trafficking, Brion et al. found that formation of secretory vesicles was severely impaired. In contrast to a total block of protein synthesis after cycloheximide application, vaccinia infection inhibits only host cell protein synthesis. Under the latter condition, we have shown that expression of a single regulated secretory protein is sufficient for the formation of secretory granules.
Analysis of Sorting Signals in Other Regulated Secretory Proteins
Prosomatostatin is a regulated secretory protein whose sorting to secretory granules has been studied intensively. Its sorting information was shown to reside in the first 82 (Stoller and Shields, 1989) or 78 amino acids (Sevarino et al., 1989) of prosomatostatin. A more detailed mutational analysis of prosomatostatin did not reveal a unique amino acid sequence sufficient for sorting (Sevarino and Stork, 1991). Taking into account that the proregion of somatostatin encodes for more than half of the prohormone, the defined region might exhibit aggregative properties rather than functioning as a sorting signal. Furthermore, by using stimulated secretion to monitor sorting to the regulated pathway of protein secretion, it is not clear to which extent sorting in the TGN or sorting by retention in mature secretory granules had occurred.
Proopiomelanocorticotrophic hormone (POMC) is another model protein for studying sorting to the regulated pathway. When mutants of this hormone with a 25– or 26– amino acid deletion at the NH2 terminus of the proregion were analyzed in Neuro 2A cells, opposing results for the existence of an NH2-terminal–sorting motif were reported (Roy et al., 1991; Chevrier et al., 1993; Cool and Peng Loh, 1994; Cool et al., 1995). Using a retroviral expression system (which does not shut off host cell protein synthesis) Roy et al. showed that the NH2-terminal 26 amino acids are not necessary for sorting. In addition, they provided evidence that sorting information is present in multiple domains of POMC and speculated that cooperation between these domains is needed for sorting of the precursor to secretory granules (Chevrier et al., 1993). In contrast, Cool et al. reported that transfection of a deletion mutant of POMC lacking amino acids 2–26 or 8–20 resulted in its constitutive secretion (Cool and Peng Loh, 1994; Cool et al., 1995). Furthermore, the authors present data that in AtT20 cells the NH2-terminal 26 amino acids of POMC are sufficient to sort the constitutive reporter protein chloramphenicol acetyltransferase to secretory granules (Tam et al., 1993). However, it remains unclear to what degree homo- or heterophilic interactions of the transfected proteins with endogenously regulated secretory proteins were involved in the sorting process. It is possible that such interactions vary considerably with the expression level of the transfected proteins and might account for the opposing results. In addition, as for prosomatostatin, sorting was studied at the level of mature secretory granules, and not in the TGN. Nevertheless, it is interesting that the proposed sorting motif of POMC like the loop of CgB is stabilized by a disulfide bond. This raises the question of whether an exposed loop of amino acids is a general feature of the signal for granular sorting. From theoretical calculations based on the primary sequence analysis of regulated proteins it was suggested that the sorting signal may contain a degenerated amphipathic helix (Kizer and Tropsha, 1991). Consistent with the latter suggestion, the disulfide-bonded loop in hCgB forms an amphipathic helix, as determined by secondary structure prediction.
In summary, our data show that the vaccinia-based expression is a powerful tool for the identification of sorting signals of regulated secretory proteins. Most importantly, it allows us to distinguish sorting signals from aggregative properties. Using this system it should be possible to identify two classes of regulated secretory proteins: (a) proteins containing all the structural information necessary for sorting to secretory granules and which, therefore, are properly sorted when they are present in the TGN as the only cargo protein and (b) proteins lacking a specific sorting signal, but which are competent for aggregation, and whose sorting therefore depends on the former class of regulated secretory proteins. A paradigm for the former class is CgB analyzed in this study. Recently, Natori and Huttner (1996) provided in vivo evidence for a helper function of CgB in sorting of proteins to secretory granules. In AtT20 cells, CgB was shown to promote sorting of the 23-K fragment of POMC, which was further processed in ISG to ACTH. Other candidates that may rely on helper proteins for sorting may be thyroid hormone and luteinizing hormone whose aggregative properties have been analyzed in vitro by Cohn and co-workers (Gorr et al., 1989), and Rindler and co-workers (Colomer et al., 1996), respectively.
We have demonstrated here using the vaccinia expression system that the NH2-terminal disulfide-bonded loop of CgB is necessary for its sorting to secretory granules. Likewise, this system can be used in future studies to analyze sorting of regulated secretory proteins for which transfection studies failed in the identification of a sorting signal. In particular, it will be interesting to identify the nature of sorting signals in regulated secretory proteins that do not contain cysteins, e.g., SgII. It may turn out that exposed loops of amino acids stabilized by other means than disulfide-bonds serve as signal in these proteins.