Secretory granule content proteins and the luminal domains of granule membrane proteins aggregate in vitro at mildly acidic pH.

A major unresolved issue in the field of secretory granule biogenesis is the extent to which the aggregation of granule content proteins is responsible for the sorting of regulated from constitutively secreted proteins. The aggregation process is postulated to take place in the trans-Golgi network and immature secretory granules as the proteins encounter mildly acidic pH and high calcium concentrations. We have developed in vitro assays that reconstitute the precipitation out of solution of secretory granule content proteins of anterior pituitary gland and adrenal medulla. In the assays, all of the major granule content polypeptides form a precipitate as the pH is titrated below 6.5, and this precipitate can be recovered in the pellet fraction after centrifugation. Addition of calcium is required for the aggregation of chromaffin granule content. In contrast to the proteins secreted by the regulated pathway, the constitutively secreted proteins IgG, albumin, and angiotensinogen, when added to the assays, remain predominantly in the supernatant. Among the individual proteins tested, prolactin is found to aggregate homophilically under these conditions and can drive the co-aggregation of other proteins, such as the chromogranins. Soluble forms of granule membrane proteins, including dopamine beta-hydroxylase and peptidyl glycine alpha-amidating enzyme also co-aggregated with granule content proteins. The results are consistent with the idea that spontaneous aggregation of proteins occurring under ionic conditions similar to those at the sites of granule formation is a property restricted to those proteins packaged in secretory granules. In addition, the association of luminal domains of membrane proteins with content proteins in vitro raises the possibility that analogous interactions between membrane-bound and content proteins also occur during granule formation in intact cells.

A major unresolved issue in the field of secretory granule biogenesis is the extent to which the aggregation of granule content proteins is responsible for the sorting of regulated from constitutively secreted proteins. The aggregation process is postulated to take place in the trans-Golgi network and immature secretory granules as the proteins encounter mildly acidic pH and high calcium concentrations. We have developed in vitro assays that reconstitute the precipitation out of solution of secretory granule content proteins of anterior pituitary gland and adrenal medulla. In the assays, all of the major granule content polypeptides form a precipitate as the pH is titrated below 6.5, and this precipitate can be recovered in the pellet fraction after centrifugation. Addition of calcium is required for the aggregation of chromaffin granule content. In contrast to the proteins secreted by the regulated pathway, the constitutively secreted proteins IgG, albumin, and angiotensinogen, when added to the assays, remain predominantly in the supernatant. Among the individual proteins tested, prolactin is found to aggregate homophilically under these conditions and can drive the co-aggregation of other proteins, such as the chromogranins. Soluble forms of granule membrane proteins, including dopamine ␤-hydroxylase and peptidyl glycine ␣-amidating enzyme also co-aggregated with granule content proteins. The results are consistent with the idea that spontaneous aggregation of proteins occurring under ionic conditions similar to those at the sites of granule formation is a property restricted to those proteins packaged in secretory granules. In addition, the association of luminal domains of membrane proteins with content proteins in vitro raises the possibility that analogous interactions between membranebound and content proteins also occur during granule formation in intact cells.
A hallmark of secretory granules is the presence within them of condensed cores containing aggregated content proteins. Morphological studies have indicated that the formation of these cores begins in dilated extensions of the trans-Golgi network (TGN) 1 (1,2) and continues in the immature secretory granules/condensing vacuoles (ISG) that are intermediates in the granule formation process (3). Progressive condensation of the content proteins occurs during maturation of the ISGs into mature secretory granules.
During the process of secretory granule formation, the granule content and membrane proteins are segregated from molecules constitutively transported to the plasma membrane. This segregation is thought to occur both in the TGN and in the ISGs (4 -6). A major unresolved issue is what role spontaneous protein aggregation has in this sequestration of the granule-specific proteins, particularly for the content proteins. Most of the available information concerning the aggregative properties of granule content proteins has derived from the observed in vitro behavior of the chromo/secretogranins (Cg), which can aggregate at mildly acidic pH in the presence of calcium (4,7,8), conditions thought to resemble those existing in the TGN (9).
If aggregation mediated sorting is to be a general mechanism for the segregation of regulated from constitutively secretory proteins then several criteria should be met. First, granule content proteins should aggregate under the ionic and pH conditions thought to be present in the TGN even in cells where Cgs are minor components or not present. Second, constitutively secreted proteins should not be able to co-aggregate with the granule content proteins. In previous studies where the aggregation of Cgs was examined either in vitro or in detergent-treated cell extracts, the behavior of the other granule content proteins or well characterized constitutively secreted proteins was not examined systematically (4,7,8).
To test whether aggregation of content proteins can potentially mediate the targeting of proteins to secretory granules, we have developed assays to measure protein aggregation in vitro. In these assays, granule content proteins undergo pH-dependent aggregation in a process that excludes constitutively secreted proteins. Some of the content proteins aggregate at low pH when assayed individually in the absence of the other proteins, while others require an aggregating partner to precipitate. Soluble forms of membrane-associated proteins also undergo co-aggregation with the content proteins under conditions where they do not self-associate. The results are consistent with a potential role for protein aggregation in the segregation of both secretory and membrane proteins to storage granules.

MATERIALS AND METHODS
Antibodies and Purified Granule Content Proteins-Rabbit antibodies to dopamine-␤-hydroxylase (DBH) prepared against the purified protein from bovine adrenal gland (10) were generously provided by Dr. R. Angeletti (Albert Einstein College of Medicine, Bronx, NY). Rabbit anti-bovine CgA antibody was a generous gift of Dr. D. Aunis (Centre de Neurochimie du CNRS, Strasbourg, France), and purified rat peptidyl glycine ␣-amidating enzyme (PAM3) and rabbit anti-PAM (PAL) antibodies were a generous gift of Drs. B. Eipper and R. Mains (Johns Hopkins University School of Medicine, Baltimore, MD). Antibody to growth hormone (GH) was purchased from Accurate Chemical (Westbury, NY) while anti-ACTH was obtained both from Accurate and from Sigma, with similar results obtained for each. Antibodies to prolactin (Prl), luteinizing hormone (LH), and follicle stimulating hormone (FSH) were obtained from the NIDDK National Hormone and Pituitary Program as were human GH and LH. Bovine serum albumin (BSA), rabbit and human IgG, Prl, and chymotrypsinogen were purchased from Sigma. Angiotensinogen was from Calbiochem. 125 I-Human recombinant IGF-1 (2000 Ci/mmol) was purchased from Amersham and was reconstituted and stored according to the manufacturer's recommendation. Porcine POMC was obtained from conditioned medium from transfected LLC-PK1 cells expressing the full-length molecule (a gift of Dr. G. Boileau, Université de Montreal, Canada).
Preparation of Granule Content Proteins-Bovine pituitaries were decapsulated, minced, and homogenized in 0.32 M sucrose with protease inhibitors. After centrifugation at 750 ϫ g to remove unbroken cells and nuclei, a crude granule preparation was obtained by centrifugation of the postnuclear supernatant at 8750 ϫ g for 20 min in a Sorvall RC2-B centrifuge (DuPont). Samples were layered onto 1.6 M sucrose and pelleted at 150,000 ϫ g for 120 min using a Ti60 rotor and a L8 -55 centrifuge (Beckman). The granule pellets were lysed by sonication in phosphate-buffered saline, pH 9, with 100 mM KCl and aprotinin (10 KIU/ml) added, and the soluble content was recovered after centrifugation at 100,000 ϫ g for 30 min. Adrenal chromaffin granules were prepared from fresh bovine adrenal medulla using the same procedure (11). Granules were lysed by freeze thawing and sonication in 100 mM KCl, 25 mM HEPES pH 7.5, in the presence of protease inhibitors. The released content proteins were recovered after centrifugation as described previously (12).
In Vitro Aggregation Assay-Granule contents or isolated proteins were de-salted over Bio-Gel P-6 DG columns (Bio-Rad) equilibrated with 5.5 mM HEPES pH 7.5. When necessary, protein stock solutions were concentrated using Centricon (Amicon) or Filtron (Northborough, MA) concentrators. Just before use, the samples were pre-spun for 30 min at 15,000 ϫ g in an Ependorf centrifuge or 56,000 ϫ g for 30 min in a Ti100.3 rotor using a Beckmann TL100 ultracentrifuge, and the pellet was discarded. A procedure similar to that described previously was employed (12). Briefly, 50-l samples containing granule content or the proteins of interest (1-4 mg/ml final concentration) in 5 mM HEPES, 10 mM MES, pH 7.5, were titrated slowly to the indicated pH values by adding 0.125 N HCl while vortexing. KCl or CaCl 2 were added in some experiments where indicated. After incubation for 15-30 min at 23°C, the reaction mixture was centrifuged to separate the pelleted precipitates from the supernatant fractions. Centrifugation in most cases was for 30 min at 15,000 ϫ g in an Ependorf centrifuge. In experiments using adrenal chromaffin extracts in the presence of calcium, centrifugation was conducted at 43-56,000 ϫ g in a TL100 ultracentrifuge. This procedure increased the magnitude of the observed precipitates but did not qualitatively change the overall results. A portion of the supernatant and the entire pellet fractions were then subjected either to a BCA protein determination assay (Pierce) or to SDS-PAGE (run without reducing agent except where indicated) followed by staining of the gels with Coomassie Blue and, when appropriate, autoradiography. As a routine control, samples were subjected to the assay but not centrifuged. The amounts of protein remaining in the tubes after removal of the assay mixture were inconsequential so long as the overall protein concentrations were at least 1.5 mg/ml, indicating that the observed recovery of protein in the pellet fractions was not due to binding to the assay tubes. BSA or hemoglobin were added as carriers to minimize binding to the tubes of test proteins used at low concentrations, such as IGF-1 or PAM.
Immunoblotting and Iodination-Immunoblotting was performed after SDS-PAGE, transfer of the proteins to nitrocellulose paper, and staining with Ponceau S to visualize the transferred proteins. 125 I-Protein A (DuPont NEN) was used in some experiments as the second step as described previously (12). In other cases, the ECL procedure was employed (Amersham). Iodination of proteins was performed using the lactoperoxidase method (13). Quantitation of the radioactivity was performed using a Molecular Dynamics PhosphorImager. In some cases images from the PhosphorImager or from digitalized images (LACIE Ltd., Beaverton, CA) of gels stained with Coomassie Blue or Ponceau S were imported into Quark Express (Denver, CO) for computerized labeling and printing of figures.

Secretory Granule Content Proteins Precipitate in a pH-dependent
Manner-Secretory granule content was obtained from bovine pituitary glands as described under "Materials and Methods." When the pH of the pituitary samples was titrated from 7.5 to below 6.5, content proteins precipitated out of solution and could be recovered in the pellet fraction after centrifugation (Fig. 1). As much as 60% of the total protein aggregated as the pH was reduced to 5.5 under these conditions. Addition of 150 mM KCl and/or 10 mM CaCl 2 (up to 40 mM in some experiments) had no significant effect on the overall aggregation of the pituitary content proteins (Fig. 1).
When analyzed by SDS-PAGE, all of the major proteins in the granule content were observed to precipitate as the pH was reduced (Fig. 2). This effect was also reversed when the pH was retitrated back to 7.5 ( Fig. 2, right panel). The two most prominent proteins were identified as Prl and GH based on the migration positions of the purified proteins and immunoblotting using specific antibodies (data not shown). Thus, these granules, as is typical for purified pituitary granules (14), consist primarily of those from somatotrophs and mammotrophs, the most abundant cells in the pituitary. However, other known pituitary proteins could be identified by immunoblot analysis using specific antibodies. A pH-dependent aggregation was observed for LH, FSH, and chromogranin A (CgA) (Fig. 3). In addition, a ϳ37-kDa protein reacting with anti-ACTH antibodies underwent pH-dependent aggregation as well (Fig. 3). This protein co-migrated on polyacrylamide gels with authentic porcine POMC. In general, all of these pituitary proteins aggregated as well as did GH and Prl in the same experiments.
To extend these results to other types of secretory granule content, the same assay was conducted with purified adrenal chromaffin granule content proteins, consisting almost entirely (Ͼ90%) of the Cgs. Unlike what was observed using the pituitary granule content, the chromaffin granule proteins when used at ϳ2 mg/ml did not undergo pH-dependent aggregation (Fig. 4). Aggregation did occur, however, when calcium was added (Fig. 4) and this effect was greater as the pH was reduced. The aggregation was also reversible. In three separate experiments where the pH was lowered to 5.6 in the presence of 20 mM CaCl 2 and, after a 30-min incubation, titrated back to pH 7.5, the recovery of CgA in the pellets was reduced to the level of the pH 7.5 control (data not shown). The concentration of calcium required for maximal aggregation was ϳ20 mM (not shown), higher than the 10 mM believed to be present in the TGN (4) but lower than the 40 mM present in mature chromaf-FIG. 1. Pituitary granule content proteins aggregate when the pH is reduced. Secretory granule content (2 mg/ml) was prepared from bovine pituitary gland and equilibrated with 5 mM HEPES, 10 mM MES, pH 7.5. The pH was reduced as indicated and the pellet and supernatant fractions recovered as described under "Materials and Methods." The protein in both fractions was measured and is plotted as the mean and standard error of the protein in the pellet as a percentage of the total (pellet ϩ supernatant). 150 mM KCl and 10 mM CaCl 2 were added where indicated. In both cases, aggregation began by ϳpH 6.5 and increased as the pH was reduced further. fin granules (15). The high calcium requirement is likely to be due to the relatively low concentrations of protein employed, far lower than that of mature granules (Ͼ100 mg/ml). It has been already established that aggregation of CgA in vitro requires less calcium as the protein concentration is increased (8). Thus, the overall aggregation of adrenal content is consistent with that expected for the chromogranins themselves.
Constitutively Secreted Proteins Are Excluded from the Precipitate Formed by Granule Content Proteins-If aggregation is to play a role in the segregation of constitutively secreted proteins from granule content proteins, then the constitutively secreted proteins should not aggregate together with the content proteins. To assess whether this would be the case in the in vitro assays, three constitutively secreted proteins were added individually to the pituitary content proteins prior to the reduction of the pH. As shown in Table I, IgG, angiotensinogen, and albumin, all proteins that are known to be constitutively secreted from transfected endocrine cell lines (16 -18), were recovered almost exclusively in the supernatant fractions. The mean aggregation of these proteins was 5% or less as compared to 29 and 37% for Prl and GH, respectively. These observations resemble those we have previously obtained using a soluble version of the pancreatic zymogen granule membrane protein GP2 (12). GP2, which is constitutively secreted from pituitary AtT20 cells, also does not co-aggregate with pituitary content proteins.
A similar set of experiments was conducted using chromaffin granule content proteins induced to aggregate in the presence of calcium. As shown in Fig. 5, none of the three constitutive markers, IgG, angiotensinogen, or BSA, co-aggregated effectively (Ͻ5%) with the adrenal content proteins, whereas CgA, the major content protein, did precipitate well (30 -39% at pH 5.6). Taken together, these data show that protein-protein in-FIG. 3. The major pituitary hormones precipitate at mildly acidic pH. Granule content (3 mg/ml) prepared from bovine pituitary glands was subjected to the aggregation assay at either pH ϳ7.5 (lanes a and b) or 5.6 -5.8 (lanes c and d). Pellet (P; lanes a and c) and 30% of the supernatant (S; lanes b and d) fractions were subjected to SDS-PAGE and transfer to nitrocellulose. The indicated proteins were identified by immunoblotting using specific antibodies followed by the ECL procedure (in the case of POMC) or by 125 I-protein A and autoradiography. All of the content proteins underwent pH-dependent aggregation, including FSH (4% at pH 7.5 versus 50% at pH 5.6), LH (3% at pH 7.5 versus 24% at pH 5.6), and chromogranin A (CgA; 2% at pH 7.5 versus 48% at pH 5.6). As expected, GH, included as a control, was found predominantly in the pellet fraction at low pH (6% at pH 7.5 versus 42% at 5.6). ϳ37-kDa POMC, from a separate experiment, also underwent pH-dependent aggregation (6% at pH 7.5 versus 39% at pH 5.6 after normalization for differences in Prl/GH aggregation). In each case, the results represent one of two similar experiments.
FIG. 4. Adrenal chromaffin granule content proteins aggregate in the presence of calcium. Granule content (2 mg/ml) was obtained from bovine adrenal medulla as described under "Materials and Methods." This content was subjected to the standard aggregation assay in the presence or absence of 40 mM CaCl 2 . Total protein in the pellets was measured and is plotted as the mean Ϯ S.E. Little precipitation of the chromaffin granule content was observed in the absence of CaCl 2 , or in the presence of 150 mM KCl and 10 mM CaCl 2 (not shown). However, when high concentrations of CaCl 2 were used, aggregation did occur, with increased aggregation at lower pH.
FIG. 2. The major pituitary secretory granule content proteins aggregate at mildly acidic pH. Granule content from bovine pituitary gland was utilized in the standard assay (see Fig. 1). Left panel, the pH was reduced to 6.2 or 5.5. The entire pellet (P) (lanes c, f, and i), 20% of the initial sample (I) (lanes a, d, and g), and 20% of the remaining supernatant (S) after centrifugation (lanes b, e, and h) were then subjected to SDS-PAGE and the gel was stained with Coomassie Blue. The migration positions of the indicated proteins, prolactin (Prl) and growth hormone (GH), were determined from those of the corresponding purified proteins and confirmed by immunoblotting using appropriate antibodies. Note the greatly increased appearance of the major proteins in the pellets after pH reduction (lanes f and i). For Prl, the percent aggregation at pH 7.5, 6.2, and 5.5 was 4, 14, and 21%, respectively, and the corresponding values for GH were 7, 35 and 44%. The migration positions of prestained marker proteins are indicated at the right margin. Presented is one of three similar experiments. Right panel, the reversibility of the pH-dependent aggregation was measured after reduction of the pH to 5.5, incubation for 30 min, and retitration back to pH 7.5. Control samples were maintained for 30 min at pH 7.5 and 5.5, respectively. The amounts of Prl and GH in the pellets and supernatants were measured by densitometric scanning of digitalized images of the Coomassie-stained gels and are presented as the overall mean and standard error of three separate experiments.
teractions occurring during protein aggregation are specific and that constitutively secreted proteins aggregate less efficiently than granule content proteins, at least in vitro.
Insulin-like Growth Factor I (IGF-1) Also Co-aggregates with Granule Content Proteins-It has been recently shown that IGF-1 is targeted to the regulated secretory pathway in AtT20 cells (19). However, IGF-1 in secretory granules did not aggregate when detergent-permeabilized AtT20 cells were incubated at pH 6.4 in the presence of 10 mM calcium. Moreover, proinsulin, which is highly similar to IGF-1 in its amino acid sequence, did not undergo pH-dependent co-aggregation with chromogranins in detergent-permeabilized PC12 cells even when accumulated prior to the assay in the TGN (19). To determine whether IGF-1 is an exceptional molecule that is targeted to secretory granules but does not co-aggregate with other granule content proteins, 125 I-IGF-1 was added to the in vitro aggregation assays together with pituitary or adrenal chromaffin content proteins. As shown in Fig. 6, IGF-1 coaggregated well at reduced pH in the pituitary assays (panel A) and reasonably well at reduced pH in the presence of calcium in the adrenal assays (panel B). In control experiments, IGF-1 failed to precipitate significantly when incubated at low pH with only BSA present, indicating that it does not self-associate under these conditions (data not shown). Thus, the targeting of IGF-1 to secretory granules in vivo does correspond to its behavior in the in vitro aggregation assays.
A Role for Self-aggregating Proteins in Content Protein Precipitation-As was shown in Fig. 2, all of the regulated secretory proteins detectable in the pituitary gland content precipitated as the pH was lowered. It is unlikely that this result was due entirely to homophilic self-aggregation of the proteins themselves. CgA, for example, which was found to aggregate in the pituitary content, did not form precipitates under the same conditions when the adrenal content alone was used (Fig. 4). However, to examine this issue in more detail, the inherent capacity of individual polypeptides to undergo pH-dependent precipitation was analyzed using purified granule content proteins. Among the pituitary hormones tested (Fig. 7), Prl was the only one to undergo strong pH-dependent aggregation. LH did not aggregate detectably, while GH did aggregate somewhat in the absence of added salt. BSA and human IgG were also tested in the assay (Fig. 7) and as expected, they did not precipitate as the pH was reduced. It is important to note that the aggregation does not, in general, correlate with the isoelectric point of the proteins. Although bovine prolactin has its pI in this range (5.8), so also do BSA (5.1), angiotensinogen (ϳ6.0), and bovine chymotrypsinogen (ϳ6.0; data not shown), which do not aggregate in the assay.
CgA, GH, and LH, which did not self-associate under these conditions, did undergo substantial aggregation in the pituitary content itself (Fig. 2), implying that their aggregation, at least in vitro, is dependent on their interaction with other proteins in the mixture, perhaps self-aggregating proteins like prolactin. To investigate these types of protein-protein interactions, Cgs, in the form of adrenal content, and purified LH were added to Prl and the pH reduced so as to initiate Prl aggregation. Prl itself was capable of driving the aggregation of the major chromaffin granule proteins, including CgA (data not shown), suggesting that these proteins could potentially interact during granule formation. LH, on the other hand, remained mostly if not entirely soluble when mixed with Prl alone in the assay (Fig. 8A), illustrating the specificity of the protein-protein interactions that occur at low pH. The aggregation of LH could be induced, however, by including LH together with  A and B, bottom). These markers of the constitutive pathway did not co-precipitate significantly with granule content proteins in the presence of CaCl 2 . At pH 5.6, 5% of angiotensinogen, 4% of IgG, and 0% of BSA were detected in the pellet fractions as compared to 30% (panel A) and 39% (panel B) for CgA. For each constitutive protein, the results represent one of two similar experiments.
FIG. 6. IGF-1 also co-aggregates with granule content proteins. 125 I-Labeled IGF-1 (10,000 cpm) was added to the aggregation assays together with content proteins from pituitary (panel A) or adrenal medulla (heat-treated; panel B). The assays were conducted as described in the legends to Figs. 1 and 4 for pituitary and adrenal, respectively, at the pH values indicated with 40 mM calcium in the adrenal samples at pH 6.4 and 5.9. Total protein in the pellet and supernatant fractions was measured using a BCA assay while 125 I was determined by counting on a ␥-counter. The mean and standard error of triplicate samples are presented after correcting for free 125 I (10%). IGF-1 co-aggregated with both the pituitary and adrenal content proteins at low pH. adrenal content and Prl. In other words, the aggregation of LH was dependent upon its interaction with proteins in the adrenal content, most likely chromogranins, which in turn can associate with the aggregating Prl. Addition of the same amount of LH to assays conducted with pituitary extracts led to a similar level of aggregation of LH (not shown). To analyze directly the potential interaction of LH with adrenal content proteins, this hormone was added to adrenal content and Cg aggregation was induced in the presence of calcium. As depicted in Fig. 8B, LH was also induced to co-precipitate together with the Cgs in the presence of calcium, whereas neither LH nor the Cgs aggregated in its absence. These results are in agreement with observations made in vivo concerning the packaging of these proteins in secretory granules. Cgs and Prl are packaged together in the granules of pituitary-derived cell lines in culture and in rat mammotrophs, while LH and chromogranins appear in the same granules in pituitary gonadotrophs (20 -22).

Co-aggregation of the Luminal Domains of Granule Membrane Proteins with Content
Proteins-Protein aggregation could potentially play a role in the segregation of membraneassociated proteins to secretory granules, particularly if these membrane proteins interact with content proteins in the TGN and ISG. For this to occur during granule formation in vivo, the luminal domains of membrane proteins would need to associate with content proteins at acidic pH. To test whether interactions of this nature could occur in vitro, we have taken advantage of the fact that many granule membrane proteins, including DBH and PAM, have soluble forms that are present in the granule content. DBH is a major chromaffin granule membrane protein which consists of a membrane-bound tetramer with two subunits firmly attached to the membrane and two subunits which exchange with a soluble pool found in the content (23). PAM is synthesized as a transmembrane precursor, but a portion of it is processed to yield soluble forms, consisting of segments of the luminal domain, that are present in the granule content of both adrenal and pituitary glands (24 -26). Furthermore, when segments of the luminal domain of PAM were expressed in AtT20 cells as a soluble, secretory proteins, they, like the transmembrane form, were incorporated into secretory granules (27).
As can be seen in Fig. 9, the endogenous soluble form of DBH in adrenal extracts aggregated only modestly when the chromaffin granule content was titrated to ϳpH 6.0 in the absence of calcium. When adrenal content was added to pituitary granule extracts prior to the assay, the major content proteins, including CgA, underwent aggregation at low pH in the absence of calcium (Fig. 9A, lanes k and l). In this case, DBH was also prominent in the pellet fractions. This result indicates that under conditions where DBH does not self-aggregate well, it can interact with other granule content proteins. When adrenal content aggregation was induced by addition of calcium to the assay, DBH sedimented (Fig. 9B). Although homotypic selfaggregation of DBH in the presence of calcium cannot be ruled out, the data suggest that DBH is in fact co-aggregating with the Cgs in the chromaffin extracts.
PAM, in contrast to DBH, could undergo a vigorous pH-dependent aggregation in the absence of other granule components. A significant fraction of purified PAM3 (a soluble isoform consisting of virtually the entire luminal domain) was observed to sediment after reduction of the pH to 5.8 regardless of whether the assay was conducted in the presence of pituitary or adrenal granule content, or nonaggregating proteins such as hemoglobin (Fig. 10) or BSA (not shown). However, PAM selfaggregation was inhibited by CaCl 2 (Fig. 10, lanes e and f). This inhibition was specific for calcium and not observed in the presence of the same molar equivalents of potassium ion (data not shown). When added to the pituitary or adrenal assays in the presence of calcium, PAM did sediment effectively (Fig. 10,  lanes k and l), indicating that it can interact with other granule content proteins. DISCUSSION Proteins traversing the TGN and ISG, the sites of sorting of constitutive from regulated proteins, are exposed to mildly acidic pH and high concentrations of calcium ions (9, 28 -31). The maintenance of an acidic pH in the TGN and ISG is known FIG. 7. Prolactin, but not other pituitary granule content proteins, homotypically self-aggregates at reduced pH. Bovine Prl, human GH, and ovine LH were subjected to the standard aggregation assay (see Fig. 1). 150 mM KCl and 10 mM CaCl 2 was included in one set of the samples as indicated. Total protein in the supernatants and pellets was measured in triplicate samples and is plotted as the mean Ϯ S.E., which in most cases was too small to be represented on the graph. Prolactin underwent a strong self-aggregation as the pH was reduced Ͻ6.5. This aggregation was not affected by the addition of salt. GH did not aggregate well, although some pH-dependent aggregation was detected in the absence of salt (up to 12% in some experiments). By contrast, LH and the constitutively secreted proteins, IgG and BSA, did not self-aggregate detectably under these conditions. to be important for granule biogenesis. Treatment of secretory cells with lysosomotropic amines to dissipate pH gradients blocks maturation of secretory granules, as has been shown in both pancreatic acinar cells (32,33) and islet cells (34). Moreover, incubation of AtT20 cells with high concentrations of chloroquine promotes the diversion of POMC from the regu-lated to the constitutive secretory pathway (35). The results presented here showing spontaneous aggregation of granule proteins in vitro support the idea that the ionic milieu of the TGN and ISG are prime contributors to the selective targeting of proteins to granules and to the condensation of secretory material that is observed to occur by electron microscopy (1, 2).  (lanes a and b), Prl alone (lanes c and d), Prl together with LH (lanes e and f), Prl, LH together with adrenal content (lanes g and h), and LH together with pituitary content (lanes i and j). Pellet (P; lanes a, c, e, g, and i) and 20% of the supernatant (S; lanes b, d, f, h, and j) fractions were subjected to SDS-PAGE in the absence of reducing agent (lanes g and h are from a separate gel). Prl by itself failed to induce the aggregation of LH (lanes e and f). However, LH did sediment when added together to adrenal content in the presence of Prl (lanes g and h). CgA also sedimented when Prl was mixed with adrenal content under these conditions whether LH was present (lanes g and h) or not (not shown). Panel B, the aggregation assays were performed as in Fig. 5. LH (lanes e and f), adrenal content (lanes c and d, g and h), and a mixture of both (lanes c and d, i and j) were incubated in the absence (lanes a-d) or presence (lanes e-j) of 25 mM CaCl 2 . In the absence of CaCl 2 both LH and CgA failed to aggregate. In the presence of calcium, however, LH co-precipitated with the aggregating adrenal proteins, including CgA. The percent aggregation (% Aggreg.) obtained for each of the indicated proteins is listed at the bottom. The results indicate that LH can interact directly with chromaffin content proteins at low pH. It will sediment when the aggregation of Cgs is induced either by calcium or by Prl. Presented is one of two similar experiments.  -h), or 0.5 mg/ml adrenal mixed with 2 mg/ml pituitary content protein (lanes i-l). Pellets (P; lanes a, c, e, g, i, and k) and 30% of the supernatant (S; lanes b, d, f, h, j, and l) fractions were subjected to SDS-PAGE in the absence of reducing agent. The samples were transfered to nitrocellulose and stained with Ponceau S (bottom). The nitrocellulose was probed with rabbit antibodies to bovine DBH and 125 I-protein A and the radioactive bands detected on a PhosphorImager (top). The adrenal content proteins (including CgA) do not aggregate at low pH in the absence of calcium (lanes c and d) but do aggregate when mixed with the pituitary extract (lanes k and l). Top, the 150-kDa adrenal DBH dimer precipitated in the pituitary extract in a pH-dependent fashion but only modestly in the adrenal content itself (compare top, lanes c and d, k, and l). The migration positions of molecular mass standards are noted at the right. Panel B, adrenal chromaffin granule content proteins were subjected to the aggregation assay at pH 7.5 (lanes a and b) or 5.8 (lanes  c and d) in the presence of 25 mM CaCl 2 as described in the legend to Fig. 6. Samples were analyzed as in panel A but using the ECL immunoblotting procedure. DBH aggregated at low pH in the presence of calcium as did CgA. Thus, DBH can co-aggregate with the pituitary content at low pH under conditions where it self-aggregates poorly. It also aggregates in the adrenal extracts in the presence of calcium, conditions that promote self-aggregation of the chromogranins. The percentage of the indicated proteins in the pellet fractions (% Aggreg.) is listed at the bottom of each panel. Presented is one of two similar experiments.
Aggregation of Granule Content Proteins-In this study, we have found that the ability to form large aggregates in vitro at mildly acidic pH is a general property of granule content proteins but not of constitutively secreted proteins. Our results indicate that aggregation is initiated as the pH is reduced to 6 -6.5, the pH range of the TGN (36), Our data would therefore be consistent with the formation of small aggregates in the TGN that continue to coalesce and condense in the ISG, where the pH is lower and the calcium and protein concentrations higher.
The conditions promoting the precipitation are somewhat different in each type of granule content. In pituitary extracts, calcium ion is not required. All of the major pituitary content proteins tested (CgA, prolactin, growth hormone, POMC, FSH, and LH) precipitate. An important finding was that the constitutively secreted proteins IgG, angiotensinogen, and BSA do not co-precipitate with the pituitary content proteins. Moreover, as demonstrated previously (12), a secretory form of the pancreatic membrane protein GP2, which is not packaged in granules in endocrine cells, does not co-aggregate with pituitary content proteins. Thus, the packaging of content proteins in storage granules correlates well with the behavior of the regulated and constitutive secretory proteins in the aggregation assays in vitro.
Adrenal content proteins, consisting primarily of the Cgs (15), do not precipitate under the same ionic conditions as those of the anterior pituitary gland. Their precipitation requires calcium, as does the major constituents of these granules, the chromogranins. Further evidence that it is Cg aggregation being measured in our assay comes from experiments conducted using a preparation that consists almost entirely of soluble Cgs, obtained after chromaffin granule extracts were heated to 100°C and centrifuged (37). The same level of Ca 2ϩdependent aggregation of CgA was observed (e.g. Fig. 6). However, a recent study has provided evidence that Ca 2ϩ is not required for the incorporation of proteins into ISGs (38). The transport of secretogranin II (CgC) from the TGN to ISGs in permeabilized PC12 cells was found to be insensitive to chelation of cytosolic Ca 2ϩ and to the addition of the Ca 2ϩ -H ϩ ionophore A23187. This suggests that either some protein sorting in the TGN can occur without aggregation of the Cgs or that sorting in this system largely occurs later, after formation of ISGs themselves, where Ca 2ϩ could still play a role. In favor of the latter explanation is the appearance in the ISGs, with a sorting index comparable to that of secretogranin II, of the constitutively secreted protein, heparan sulfate proteoglycan in this study (38). It is also possible, however, that interaction of the granule content proteins with membrane proteins, a process that our data using DBH and PAM3 suggest may occur at mildly acidic pH even in the absence of calcium, is sufficient to insure some preferential sorting of these proteins in the TGN. In favor of this view, CgA and CgB have been reported to bind to chromaffin granule membranes at low pH in the absence of Ca 2ϩ (39,40).
Further support for the notion that interactions between granule content proteins themselves are involved in selective packaging of proteins in granules has been inferred from the properties of the Cgs. Cgs have been shown to self-aggregate when calcium is present (7,8). Calcium and low pH can also stabilize the Cg-containing aggregates formed in PC12 and GH4 cells (4). At neutral pH in the absence of Ca 2ϩ , the 35 S O 4 -labeled Cgs, while still in the TGN, were extracted from detergent-permeabilized cells but at pH 6.4 in the presence of 10 mM Ca 2ϩ , they remained sedimentable. The behavior of content proteins other than Cgs was not reported (4). More importantly, the use of BiP/GRP78, protein disulfide isomerase, and free glycosaminoglycans as markers of the constitutive secretory pathway that did not efficiently co-aggregate with Cgs is problematic. BiP and protein disulfide isomerase are normally resident endoplasmic reticulum proteins and have not been shown to exit endocrine secretory cells via the constitutive pathway. Glycosaminoglycans, in contrast to bona fide constitutively secreted proteins, are exported from the cell by both constitutive and regulated pathways (41). Similarly, Gorr et al. (7) reported that ovalbumin did not co-aggregate with CgA. However, ovalbumin has not been expressed in endocrine or neuroendocrine secretory cells; hence it is not known whether it is a regulated or constitutive secretory protein. Thus it is difficult to directly relate aggregation experiments using ovalbumin, BiP, protein disulfide isomerase, or glycosaminoglycan chains to the general problem of protein sorting to secretory granules. It should be noted that the aggregation of pancreatic zymogen granule content proteins in vitro at low pH has been previously reported (42). IgG added to the assay was also shown not to co-aggregate with amylase and other pancreatic zymogens.
The Role of Self-aggregating Content Proteins-Several of the individual granule content proteins were found to homo-  -l) and the standard aggregation assays conducted at pH 7.5 (lanes a and b, g and h) or 5.8 (lanes c-f and i-l) as described under "Materials and Methods." 25 mM CaCl 2 was included in one set of samples in each case (lanes e and f, k and l). Pellet (P; lanes a, c, e, g, i, and k) and 20% of the supernatant (S; lanes b, d, f, h, j, and l) samples were analyzed by SDS-PAGE on a 12% gel followed by transfer to nitrocellulose and staining with Ponceau S (bottom) and immunoblotting using the ECL procedure (top). At low pH, PAM underwent a strong homophilic aggregation (35% at pH 5.8) in the presence of globin, which did not itself precipitate (3%, lanes c and d). In comparison to the samples containing globin, PAM aggregated only slightly better when incubated with pituitary content proteins at pH 5.8 (41%, lanes i and j). However, in the presence of calcium, the self-aggregation of PAM was reduced (13%, lanes e and f) and under these conditions PAM still precipitated strongly in the pituitary samples (57%, lanes k and l). By comparison, the aggregation of Prl and GH together at pH 5.8 was 19% in the absence and 17% in the presence of calcium. Panel B, purified PAM3 was added to 1.5 mg/ml adrenal content proteins or globin and the assay was performed as described in the legend to Fig. 9. Samples were analyzed as described in panel A using a 10% polyacrylamide gel. In the absence of calcium at pH 5.8, 23% of PAM aggregated in the globin sample and 32% was found in the pellets of the adrenal samples. In the presence of calcium, PAM aggregated much better when mixed with adrenal content (55%, lanes k and l) than with globin (3%, lanes e and f). CgA recovered in the pellet was only 2% in the absence of calcium but 33% in its presence. Endogenous PAM did not contribute to the results as it was not readily detectable in immunoblots of pituitary and adrenal content under these conditions (not shown). Thus, PAM is capable of associating with granule content proteins at low pH. The results represent one of two similar experiments. typically aggregate. Aside from the Cgs, Prl underwent strong pH-dependent precipitation. Other major granule content proteins also can self-aggregate. We have found, for example, that insulin will completely precipitate out of solution by lowering the pH from 7.5 to 6.2 (data not shown) similar to what has been documented previously (43). von Willebrand's factor, which is the major component of endothelial cell Weibel-Palade bodies, also aggregates at mildly acidic pH (44). Secretory granule proteins that do not have the inherent ability to selfaggregate must be dependent on other proteins to precipitate. Indeed, it has been shown that parathyroid hormone co-precipitates in vitro with CgA at pH 5.9 in the presence of Ca 2ϩ (45). Similarly, we found that LH, which does not self-aggregate, sedimented efficiently only when Cg aggregation was induced, either by Prl or by addition of Ca 2ϩ . Prl by itself, however, could not bring about the aggregation of LH, highlighting the specificity involved in these types of protein-protein interactions. Taken together, the experiments suggest that self-aggregating proteins, like the Cgs, may be critical for the formation of condensed content in secretory granules.
The Luminal Domains of Membrane Proteins Co-aggregate with Content Proteins-Soluble forms of three membrane-associated proteins PAM, adrenal DBH, and, as we have previously shown, pancreatic GP2 (12), were found to precipitate with content proteins in in vitro assays under conditions where they did not homotypically self-aggregate well. The soluble forms of the membrane proteins used in the assays represent almost their entire luminal domains. In the case of PAM (27) and GP2 (12), the secretory versions also contain granule packaging information when the proteins are expressed in the appropriate cell hosts. In chromaffin granules, DBH is a major membrane constituent (15) and the soluble subunits can exchange with those of the membrane-bound tetramer (23). Thus, it is reasonable to propose that the interaction observed between content and the luminal segments of these membrane proteins could also occur when they are in their membrane-bound state. The ability of exposed luminal domains of membrane proteins to associate with the content proteins, at least in vitro, may be a general feature of many granule membrane proteins. For example, pH-dependent binding of a putative chromaffin granule Ca 2ϩ channel/inositol trisphosphate receptor-related membrane protein to CgA has also been observed (46). We speculate that similar interactions between the luminal segments of membrane and the content proteins may also occur during granule formation. These interactions could potentially play an important role in the incorporation of membrane proteins themselves into granules as they could be segregated in concert with the content proteins.
Protein Aggregation and Selective Packaging of Proteins in Secretory Granules-As noted above, content protein aggregation is a distinctive feature of granule formation that is readily observed by electron microscopy. Our data support those from previous studies and show that constitutively secreted proteins aggregate significantly less effectively in vitro than granule content proteins from pituitary and adrenal medulla. In principle, this would be consistent with an important role for aggregation in excluding nonaggregating constitutively secreted proteins from regions where sorting of granule content proteins is taking place. However, there is evidence that suggests that aggregation alone is not sufficient to ensure the packaging of all proteins in secretory granules. The Ca 2ϩ -induced aggregation of CgB in detergent-permeabilized TGN vesicles, for example, is maintained in a variant whose disulfide bond has been disrupted, even though this protein does not enter the secretory granules of PC12 cells (47). The interpretation of this result is that binding of granule content proteins to the mem-brane in the TGN, possibly mediated by specific receptors, is of greatest consequence. Hence the formation of a protein aggregate would not ensure incorporation into secretory granules but could potentially increase the efficacy of membrane-content protein interactions. Further studies will be necessary to characterize the nature of content protein association with membranes and the potential role of the granule membrane proteins in secretory granule formation in vivo.