Two proteins associated with secretory granule membranes identified in chicken regulated secretory cells

From yeast to mammals, all types of cells secrete proteins. The secretory pathway, consisting of the rough endoplasmic reticulum (RER), the Golgi apparatus (GA), and various types of secretory vesicles, is involved in two fundamental modes of secretion: the constitutive and the regulated (Palade, 1975; Tartakoff et al., 1978; Kelly, 1985; Burgess and Kelly, 1987; Rothman and Orci, 1992). In the constitutive pathway, newly synthesized proteins are secreted minutes after leaving the Golgi apparatus. In the regulated pathway, which operates in many cell types including exocrine and endocrine cells, neurons and granulocytes, proteins are sorted from those destined for constitutive secretion and appear in a variety of progressively maturing granules, which eventually fuse upon cell activation with the plasma membrane and discharge their contents into the extracellular space (Gumbiner and Kelly, 1982; Moore et al., 1983; Green and Shields, 1984; Burgess and Kelly, 1987).

Earlier morphological studies have revealed the progressive maturation of secretory granules (Bainton and Farquhar, 1966, 1968a,b; Jamieson and Palade, 1971; Farquhar and Palade, 1981). As regulated secretory proteins leave the trans-Golgi network (TGN) (Griffiths and Simons, 1986), they are sorted from constituitively secreted proteins (Tooze and Huttner, 1990), and aggregate, as they are transported from the TGN, into progressively maturing vesicles with electron-dense contents. Immature secretory granules are usually found in the vicinity of the TGN. Mature vesicles are located at the apical side of epithelial secretory cells and undergo exocytosis following cell stimulation.

With the exception of the synaptic vesicle membrane (Matthew et al., 1981; Jahn et al., 1985; Wiedenmann and Franke, 1985; Trimble et al., 1988; Baumert et al., 1989; Fischer von Mollard et al., 1990; Südhof and Jahn, 1991; Brose et al., 1992), the molecular composition of membranes of organelles present in the distal part of the secretory pathway in regulated secretory cells has not been thoroughtly studied. Only a few proteins specific to the secretory granule membrane have so far been identified (Cameron and Castle, 1984; Buckley and Kelly, 1985; Cameron et al., 1986; Bainton and Gerrard, 1991; Brand et al., 1991; Yamashita and Yasuda, 1992) although the granule content has been described extensively (Bainton and Farquhar, 1968a,b; Kiang et al., 1982; Rosa et al., 1985; Winkler et al., 1986). Comparative two-dimensional PAGE analysis of iodinated granule membrane extracts from various exocrine cells (Cameron et al., 1986) showed that these granule membranes share a small number of polypeptides with low turn-over rates, which are partly recovered through recycling for their reutilization in the packaging of new waves of secretory products (Meldolesi, 1974; Herzog and Farquhar, 1977). On the cytoplasmic face, secretory granule membranes display GTP binding proteins (Darchen et al., 1990; Chavrier et al., 1990; Goud and McCaffrey, 1991) and annexins (Crompton et al., 1988; Turgeon et al., 1991; Creutz, 1992). These molecules are involved in targeting of secretory granules to plasma membranes and in exocytosis. On the luminal face, several secretory granule proteins are inserted into the lipid bilayer of the membrane by means of a phosphatidyl inositol anchorage (Fouchier et al., 1988; LeBel and Beattie, 1988; Paul et al., 1991). The best known of these proteins is GP2, the major protein constituent of zymogen granule membrane: GP2 exists in both free and GPI-anchored forms (Fukuoka et al., 1991) and presents pH- and ion-dependent self-association properties, which may be utilized to establish GP2-enriched microdomains in the granule membrane (Fukuoka et al., 1992). Lately, GPI linkage has been proposed to be a specific signal for targeting of proteins to the apical surface of various polarized epithelial cells, such as MDCK cells and other kidney and intestinal cell lines (for review see Nelson, 1992). The sorting events directing the fate of GPI-anchored proteins most probably occur in the TGN, where GPI-anchored proteins and glycosphingolipids are co-clustered in microdomains of the membrane (van Meer and Simons, 1988; Simons and Wandinger-Ness, 1990; Brown and Rose, 1992). However, the GPI-anchored proteins are not exclusively targeted to the apical membrane, since in the epithelial cell line FRT (Fischer rat thyroid) most of the endogenous GPI-anchored proteins are basolateral (Zurzolo et al., 1993).

Several other secretory granule membrane proteins have been recently identified in mammalian endocrine and exocrine cells by means of mAbs (Gerrard et al., 1991; Bainton and Gerrard, 1991; Yamashita and Yasuda, 1992). Using similar approaches, we have detected in chicken granulocytes and thrombocytes two granule membrane proteins of apparent molecular mass of 92-94 kDa (GRL1) and 40-65 kDa (GRL2) (Thomas et al., 1993). During development, GRL1 is restricted to thrombocytes and myelocytes, whereas GRL2 is expressed also in myeloid cells and a subpopulation of erythroid progenitors.

In the present study we report the distribution of GRL1 and GRL2 in a variety of chicken exocrine and endocrine cell types and their co-localization in zymogen granules in acinar pan-creatic cells. The immunogold studies of GRL1 and GRL2 in pancreatic acinar cells show that both antigens are predominantly localized in secretory granules and apical surfaces. The SDS-PAGE analyses show that GRL1 corresponds to a constant 92-94 kDa doublet immunoprecipitated from detergent extracts of granulocytes, thrombocytes, pancreatic and adrenal cells. On the other hand, GRL2 has been recovered from granulocytes, thrombocytes and pancreatic cells as a diffuse 40-65 kDa band. Further biochemical data indicate that GRL1 and GRL2 bind to secretory granule membranes by different mechanisms: GRL1 behaves as a peripheral protein of the secretory granule membrane, while GRL2 displays features of a transmembrane protein. Taken together, the results of our studies are consistent with the conclusion that GRL1 and GRL2 are common constituents of the secretory granule membrane in avian exocrine, endocrine and hemopoietic cells.

Fertilized eggs from chick (Gallus gallus), obtained from commercial sources, were incubated in a rotary incubator at 37.5°C. Stages of development of the embryos were expressed in days of incubation (Ex, embryo of xdays of incubation).

Isolated organs from E12, -14, -16, -18 and -21 embryos and from 2- and 52-day-old chicken were fixed at −20°C by immersion in absolute ethanol containing 1% acetic acid, for 3 hours for small size samples (less than 5 mm long and 2 mm thick), and overnight for larger pieces. Fixed tissues were then frozen in Bright Cryon-M-Bed and 5 μm thick sections were cut. Immunofluorescent labeling was performed as described previously (Thomas et al., 1993).

Also, several tissues (pancreas, adrenal glands, salivary gland, epiphysis and stomach) were fixed for 4 hours at room temperature by immersion in Carnoy’s solution, or in 4% freshly made paraformaldehyde in 0.1 M phospate buffer, pH 7.4 (PBS), then dehydrated and paraffin embedded. Immunocytochemical staining for GRL1 and GRL2 was performed in 5 μm thick sections with a per-oxidase-coupled rabbit anti-mouse antibody (RAM-HRP).

Several tissues were processed according to the method of Brown and Farquhar (1989)with minor modifications. In brief, anesthetized adult chickens were perfused at room temperature (RT) via the left ventricle with approximately 100 ml of buffered physiological saline (PBS), followed by a fixative of freshly made 4% paraformaldehyde and 0.1% glutaraldehyde in PBS. The pancreas was excised and cut into small pieces, about 0.3-0.5 cm³, and fixed at RT for 3-4 more hours. Tissues were washed in PBS containing 50 mM NH₄Cl, and in PBS, cryoprotected with 10% sucrose in PBS for light microscopy (LM) or 30% sucrose in PBS for electron microscopy (EM), and stored at 4°C. On the day of sectioning, tissues were frozen in isopentane cooled in liquid nitrogen. Sections for LM were cut at 10 μm and picked up on poly-L-lysine-coated slides; sections for EM were cut at 40 μm and processed floating.

Before immunostaining, sections were permeabilized with saponin (0.1-0.5%) or Triton X-100 (0.5%) in PBS, and blocked for 30 minutes in 3% fish gelatin (FG) in PBS. Subsequently, sections were incubated overnight in undiluted anti-GRL1 or -GRL2 hybridoma supernatant, washed in PBS, incubated for 1 hour in biotinylated rabbit anti-mouse IgG (70 μg/ml), washed, and incubated for 1 hour in avidin-biotin-peroxidase (ABC) (Vector, Burlingame, CA). Peroxidase was developed by the method of Graham and Karnovsky (1966)using diaminobenzidine (DAB) as substrate with the addition of 10 mM imidazole. Sections for LM were dehydrated and coverslipped: sections for EM were fixed with 1% OsO₄+ 1.5% potassium ferrocyanide, dehydrated, embedded in Araldite, and sectioned according to previously described methods (Stieber et al., 1987).

Adult chicken were perfused with PBS and fixative, as described above, with the exception of the cryoprotection step; instead, tissues were stored at 4°C in PBS + 5% sucrose until use. The method of Tokuyasu (1980, 1989) was followed for polyvinyl pyrrolidone (Sigma) infiltration, cryosectioning, immunostaining, OsO₄and uranyl staining, and PVP embedding. In brief, tissues were infiltrated in 20% PVP + 1.84 M sucrose in PBS, and frozen sections of gold-blue interference colors were cut in a Cryo-Nova ultramicrotome and transfered to grids. The grids were placed face down, and left overnight at 4°C on a layer of 2% gelatin in PBS; the next day the gelatin was melted and the grids washed in PBS.

For immunostaining, incubations and washes were done at RT with the grids floating section-side down; all washes were done in PBS. The following colloidal gold (CG)-labeled antibodies were used: (a) goat anti-mouse IgG-15 nm CG; (b) goat anti-mouse IgG1-15 nm CG (for GRL2); and (c) goat anti-mouse IgG3-5 nm CG (for GRL1). Sections were incubated with undiluted supernatants of the two monoclonal antibodies (mAbs) against GRL1 and GRL2, either separetly or combined for 1 hour or overnight. Colloidal gold-labeled antibodies were used at the dilution of 1:10 or 1:20 in 3% FG in PBS for 1 to 4 hours. To eliminate aggregates of colloidal gold, conjugates were centrifuged (20,000 g) before use and only upper supernatants were collected to perform the immunolabeling. For single labeling, goat anti-mouse IgG-CG was used; for double labeling, the subclass specific anti-IgG1 (GRL2) and IgG3 (GRL1) antibodies were used.

A representative double-labeling experiment involved the followings steps. Non-specific binding was blocked with 3% FG in PBS for 15 minutes. Sections were incubated overnight in both supernatants combined, washed and incubated for 4 hours, with both subclass-specific second antibodies combined. Control sections were incubated separately with each supernatant and each second antibody. After immunostaining, sections were washed and fixed for 5 minutes in 2% glutaraldehyde in PBS. Then they were washed in PBS and distilled water (dH₂O), fixed for 10 minutes in 1% OsO₄+1.5% potassium ferrocyanide, washed in dH₂O, stained for 10 minutes in 2% uranyl acetate in 0.15 M potassium oxalate, pH 7, washed in dH₂O, stained in 2% uranyl acetate in dH₂O, and infiltrated in 3% PVP+0.1% uranyl acetate for at least 5 minutes. Grids were then picked up in a wire loop and dried.

The purpose of this study was to determine the relative labeling of secretory granules with antibodies against GRL1 and GRL2, and to assess the distribution of these two antigens in pancreatic acinar cells. The immunolabeling was performed with hybridoma culture supernatants and affinities of the mAbs for their respective antigens as well as mAbs concentrations in the culture supernatants were not quantified. Therefore, the results of this study represent approximate distributions of GRL antigens in secretory granules. The morphometric analyses were done on the basis of areas rather than linear profiles, which were not uniformly distinct in the cryosections of pancreas.

Electron micrographs of pancreas labeled with the post-embedding immunogold procedure were used at final magnifications ranging from approximately ×10,000 to ×25,000, and results were expressed as number of gold particles per square micrometer in a given organelle or compartment. The surface area was obtained by superimposing a 1 cm × 1 cm array of points on the electron micrographs, and totalling the number of points falling on each specific structure studied. Points were then converted to square micrometers according to the magnification of each picture (Williams, 1977; Hickey et al., 1983). The following compartments or organelles were analyzed: nuclei, mitochondria, rough endoplasmic reticulum, Golgi apparatus, densestaining zymogen granules, pale or empty-looking granules, apical surfaces including the apical lumen and unidentified cytoplasmic components.

Granulocytes, thrombocytes and pancreatic acinar cells were metabolically labeled in vitro with [³⁵S]methionine (Amersham). As previously described (Thomas et al., 1993), granulocytes were isolated from E18 splenocytes on two successive discontinuous Percoll density gradients and recovered at the density of 1.08 g/ml, whereas thrombocytes were obtained from heparinized blood processed on two successive Ficoll-Hypaque gradients according to Traill’s protocol (Traill et al., 1983). Pancreatic acinar cells were isolated according to the procedure described by Amsterdam et al. (1977). For metabolic labeling, cells were incubated in MEM-methionine free medium (Gibco) supplemented with 1% glutamine, 1% penicillin-streptomycin and 5% predialysed foetal calf serum at 40°C in 5% CO₂; [³⁵S]methionine was added at 10 μCi/ml per 10⁷cells for 1 to 3 hours. After labeling, the cells were washed twice in PBS at room temperature, then frozen in liquid nitrogen and thawed in 1% NP40/Tris-HCl, 150 mM NaCl buffer, pH 7.5 (IP buffer), supplemented with freshly prepared protease inhibitors (1% aprotinin, 2 mM phenylmethylsulfonyl fluoride (PMSF), 0.2 mM sodium-p-tosyl-L-lysine chloromethyl ketone (TPCK), 1 mM N-tosyl-L-phenylalanine chloromethyl ketone (TLCK) and trypsin-chymotrypsin common inhibitor (0.5 mg/g of pancreatic tissue). The pellet of broken cells was resuspended in lysis buffer, sonicated (2× 3 seconds), and extracted for 45 minutes on ice. Immunoprecipitations were then performed with anti-GRL1 or -GRL2 mAbs, according to a previously described procedure (Thomas et al., 1993).

Granule-enriched fractions were isolated from [³⁵S]methionine-labeled granulocytes or pancreatic acinar cells and then submitted to several extractions before immunoprecipitation.

Labeled cells were first washed in PBS, frozen in liquid nitrogen, and then thawed in 0.25 M sucrose, 1 mM KH₂PO₄buffer, pH 7.4, supplemented with freshly made protease inhibitors. After incubation in this low-salt buffer for 20 minutes on ice, the broken cells were resuspended and centrifuged for 5 minutes at 500 g; the supernatants collected and centrifuged for 30 minutes at 10,000 g. The resulting pellet, enriched in secretory granules, was thoroughtly mixed and sonicated (3× 1 second) in order to break the granules. Four different extractions were then performed on ice.

Extractions were carried out separately in: 1 M NaCl, 0.1 M NaOH, 0.1 M Na₂CO_3,and in 1% solution of the precondensed non-ionic detergent Triton X-114 according to Bordier (1981). Treatments with high salt or alkaline pH buffers were performed for 30-45 minutes on ice. After centrifugation, supernatants and pellets as well as aqueous and detergent-rich phases from Triton X-114 extraction were collected separately. Supernatants and Triton X-114 aqueous phases were saved and pellets lysed in 1% NP40-IP extraction buffer.

In addition chloroform/methanol (2/1, v/v) extracts were obtained from a pellet of 10⁸granulocytic cells and partitioned according to the Folch’s procedure (Folch et al., 1957). The aqueous phase, rich in glycolipids, and the solvent phase, rich in apolar lipids, were then collected separately and submitted to an ELISA essay, with anti-GRL1 or anti-GRL2 mAbs and with RAM-HRP.

Immunoprecipitations of the GRL1 and GRL2 antigens were performed according to the procedure previously described (Thomas et al., 1993) with detergent extracts from granulocytes, thrombocytes, pancreatic acinar cells and adrenal glands.

Using an anti-GRL2 Ig affinity column, GRL2 protein was purified from secretory granule-enriched fractions isolated from pancreatic cells of adult chicken. Immunoglobulins isolated from ascites produced by hybridoma cells against GRL2 were bound to Protein A-Sepharose CL4B (Pharmacia), and the subsequent purification of the GRL2 protein was carried out according to the procedure of Schneider et al. (1982). Briefly, the membranes were solubilized in 10 mM tri-ethanolamine (TEA), 150 mM NaCl buffer, pH 8.2, containing l% NP40, 0.5% deoxycholate, freshly prepared protease inhibitors as described previously, and trypsin-chymotrypsin common inhibitor (0.2 mg/ml). Membrane extract was loaded on the immunomatrix and the GRL2 antigen was eluted with 10 mM diethanolamine (DEA), 150 mM NaCl buffer, pH 11.4. The yield of GRL2 was 5 μg per 5 g of fresh pancreas.

As previously reported, the two secretory granule antigens GRL1 and GRL2 are present in granulocytes and thrombocytes (Thomas et al., 1993), and in a variety of exocrine and endocrine tissues (Table 1A-B).

The expression of GRL1 and GRL2 antigens was followed during development in histological sections of the entire chicken embryo up to E9 and in isolated organs of embryos, newly hatched and two-month-old chickens. Labeling was performed with anti-GRL1 and -GRL2 mAbs.

Under the light microscope, secretory cells labeled by immunofluorescence display a bright intracytoplasmic granular staining. The brightness of intracytoplasmic stain makes the detection of a possible cell surface staining difficult.

The GRL antigens are not seen in secretory tissues during the first half of the embryonic development. Both antigens are expressed at E18 by pneumocytes of lung alveoli and pancreatic exocrine cells. Catecholamine secreting cells of the adrenal gland are tyrosine hydroxylase⁺and GRL2⁺at E16 (data not shown).

After hatching, the GRL antigens are expressed by exocrine (Table 1A) and various endocrine (Table 1B) secretory cells, but are absent from neurons of the central nervous system. In all exocrine tissues studied (Table 1; Fig. 1A,B), GRL1 and GRL2 are highly expressed. On the other hand, the pattern of distribution in endocrine tissues is more complex. GRL2 is brightly stained in various endocrine tissues (Table 1B; Fig. 1C), but not detected in steroid-secreting tissues such as gonads or corticoid cells, in the thyroid gland and in endocrine cells of the pancreatic islets of Langerhans. The co-expression of GRL1 and GRL2 is only observed in the adenohypophysis, where GRL⁺cells are distributed throughout the gland (Fig. 1C): both in the anterior part of the gland, which secretes prolactin, and in the caudal region of the gland, the site of gonadotrope hormone production. This distribution pattern suggests that various types of pituitary cells can express the GRL antigens. In adrenal cells, GRL1 is not detected by immunofluorescence; however, it can be immunoprecipitated (see Fig. 6), suggesting that GRL1 is probably expressed by more endocrine cell types than immunolabeling studies reveal. In addition to this distribution in regulated secretory cells, we have noted that GRL2 was expressed by other cell types; namely, endothelial cells of adult brain vessels, fibroblasts of the dermis, and chondrocytes. The presence of GRL2 in these cells is transient, indicating that the distribution of GRL2 is not strictly restricted to the granules of regulated secretory cells.

The above observations suggest that GRL antigens are markers of storage granules in various cells with regulated secretion. Moreover, they raise the question of the respective localization of GRL1 and GRL2 within secretory granules. Therefore, we investigated the subcellular localization of GRL1 and GRL2 in pancreatic acinar cells by pre- and post-embedding immunocytochemistry.

On thin (0.5 μm) frozen sections or semithin sections of Araldite-embedded pancreas, stained with mAbs against GRL1 and GRL2, numerous discrete granules are localized at the apical aspect of acinar cells (Fig. 2). The staining pattern is consistent with a localization of the two antigens in zymogen granules.

In order to gain a more precise view of the distribution of the two antigens, fixed frozen sections of pancreas stained first with mAb against GRL1 or GRL2 and then with goat antimouse IgG-15 nm colloidal gold complexes, were examined (Figs 3-5). In the pre-granular compartments of the secretory pathway, namely within the stacks of cisternae of the Golgi apparatus, gold particles are rarely seen (Figs 3, 4). The gold particles are localized almost exclusively at the periphery of large intracytoplasmic granules. Two types of large granules are observed: one with a dense center (Figs 3, 5A) corresponds to a typical zymogen granule (Jamieson and Palade, 1968, 1971); the other, with an apparently empty center, shows an apical localization (Fig. 5B) and is subjected to further discussion. Granules with dense central cores surrounded by peripheral rims that appear empty are probably zymogen granules that, due to the method of fixation and cryo-sectioning, show retraction artefacts (Fig. 5A). In a double-labeling experiment employing GRL1 labeled with 5 nm gold particles and GRL2 labeled with 15 nm gold particles, clusters of both sizes of gold particles are localized at the periphery of zymogen granules (Fig. 5A). In the center of zymogen granules, very few gold particles are observed. In addition to this predominant distribution in intracytoplasmic granules, gold particles are found on apical membranes (Fig. 5A).

Semiquantitative studies, carried out on fixed frozen sections immunolabeled as described above, corroborate the observation on the distribution of GRL1 and GRL2. The analyses of the number of GRL gold particles (Table 2) and of the number of GRL gold particles per unit of surface area (Table 3) show significant similarities as well as some differences between GRL1 and GRL2. The expression of GRL antigens is low, and probably insignificant, in nuclei, mitochondria and rough endo-plasmic reticulum (Table 3). A rapid transit time of GRL1 and GRL2 through the Golgi apparatus is suggested by the absence of significant labeling of this organelle. Obviously, the post-TGN compartments of the regulated secretory pathway are the main sites where GRL antigens are detected. In cells labeled for GRL1, 63.7±1.1% of gold particles (Table 2) are over zymogen granules suggests that GRL1 and GRL2 are concentrated on granule membranes in parallel with the maturation of secretory granules. The distribution pattern of gold particles shows quantitative variations between the different types of granules. Thus, both GRL1 and GRL2 are twice as concentrated at the periphery of light granules as in zymogen granules. On the other hand, the density of GRL2-gold particles expressed on apical areas and in light granules is significantly greater with respect to that of GRL1 (Table 3). This observation may result from a specific topology within the apical and light granule membranes of the GRL2 epitope, which favors its detection.

In Tables 2and 3, gold particles detected in unidendifiable areas are gathered under the heading ‘undetermined structures’. Tangential sectioning through labeled compartments and the less than optimal ultrastructural morphology of the ultrathin frozen sections account for most of the labeling in this category.

GRL1 and GRL2 have been previously identified in detergent extracts of granulocyte membranes, by immunoprecipitation with anti-GRL1 and GRL2 mAbs. The structural features of these antigens are different: GRL1 appears as a doublet of 92-94 kDa apparent molecular mass, whereas GRL2 is a highly N-glycosylated protein corresponding to a diffuse band at 40-65 kDa, a sharper band at 32 kDa, and a polypeptide core of 24 kDa (Thomas et al., 1993).

In order to determine whether the epitopes expressed by hematopoietic cells, exocrine and endocrine cells belong to unique GRL1 and GRL2 molecules, or to a set of different proteins reacting with mAbs against GRL1 or GRL2, immuno-precipitations and immunopurifications have been performed on detergent extracts of crude membranes from granulocytes, thrombocytes, pancreatic cells and adrenal cells (Fig. 6). Irrespective of tissue sources, GRL1 shows identical electrophoretic patterns, suggesting that there are no tissue-specific isoforms of GRL1. A similar conclusion cannot be drawn for GRL2 on the basis of one-dimensional gel analysis. Indeed, from granulocytes thrombocytes and pancreas GRL2 is immunoprecipitated as a long and diffuse band between 40 and 65 kDa; on the other hand, in thrombocyte immunoprecipitates, the 40-65 kDa form of GRL2 is hardly detected, while two 24 and 32 kDa bands are highly expressed.

Ultrastructural immunocytological observations (Figs 3-5) have shown that GRL antigens are expressed at the periphery of secretory granules, close to the membrane. The following experiments introduce information on certain structural features of the GRL1 and GRL2 proteins, and insights on their respective relationship with the granule membrane.

Various agents known to disrupt different membrane-protein interactions have been applied to crude membrane fractions. As shown in Fig. 7A, after carbonate extraction, 100% of GRL1 was found in the supernatant, while extraction in 1 M NaCl released only about 50% of GRL1 into the supernatant (not shown). In contrast to GRL1, GRL2 is not removed from membranes, by high salt (1 M NaCl) or high pH (0.1 M NaOH and 0.1 M Na₂CO₃) reagents (Fig. 7B). The phase partition of crude membrane fractions in Triton X-114 detergent, leads to significant extraction of GRL1 in the aqueous phase, whereas GRL2 is detected only in the detergent phase (Fig. 7A,B).

These results are consistent with the conclusion that the GRL1 and GRL2 antigens have different structural relationships with the lipid bilayer of membranes: GRL2 is an intrinsic protein of granule membrane while GRL1 is probably a peripheral protein. Concerning GRL1, this view is completed by the results of ELISA performed on lipid extracts from crude membranes of granulocytes. Indeed, an anti-GRL1 reactive epitope was detected in lipid extracts of crude membranes, and exclusively in glycolipid fractions, whereas no signal was observed with anti-GRL2 (Fig. 8). This result suggests a copartition of GRL1 protein, or of an anti-GRL1 reactive epitope, with membrane gangliosides. It is also consistent with previous observations reporting the existence of protein contamination in ganglioside extracts (Ledeen and Yu, 1982). Therefore, GRL1 might be a component of a glycoproteolipidic complex formed by a linkage of GRL1 with membrane phospholipids.

Our studies by immunolabeling show that the two GRL proteins are detected predominantly in association with the secretory granule membranes and are expressed in granulocytes, in various exocrine and endocrine cells, but not in neurons of the central nervous system. Taken together, these observations imply a regular relationship between certain proteins of secretory granule membranes, common to hemopoietic, exocrine and endocrine cells. In that regard, GRL1 and GRL2 may belong to recently described membrane constituents of endocrine and exocrine secretory granules (Reinhart et al., 1991; Brand et al., 1991; Yamashita and Yasuda, 1992). In man, R-GRAMP (Reinhart et al., 1991) has been detected in a variety of endocrine and exocrine cells, in granulocytes, but not in cells of neural origin. In the rat, the SG170 antigen was observed in various exocrine and endocrine cells, as well as in liver and absorptive cells of the small intestine and of renal tubules (Yamashita and Yasuda, 1992). Nevertheless, these molecules cannot be defined as common antigens of regulated secretory cells, since their expression is not ubiquitous. In this way, GRL1 and GRL2 are antigens associated with the granule membranes of exocrine cells, whereas the chromogranins/secretogranins (Rosa et al., 1985; Eiden et al., 1987; Huttner et al., 1988) are never expressed by exocrine cells. On the other hand, GRL proteins are not detected in neuronal cells, whereas synaptic vesicles as well as neuroen-docrine granules are endowed with specific proteins such as synaptophysin, synapsins, synaptotagmin and synaptobrevin (for review see Südhof and Jahn, 1991; Greengard et al., 1993). These findings illustrate the diversity inherent in the molecular organization of membranes of secretory granules, conferring a specificity on each different type of hemopoietic, endocrine, exocrine or neuronal secretion. In contrast, molecular analyses of vesicle targeting and fusion processes indicate that both constitutive and regulated fusion of membranes are mediated by common mechanisms (Wilson et al., 1989; Rothman and Orci, 1992; Söllner et al., 1993; Barinaga, 1993; Bennett and Scheller, 1993; Bennett et al., 1993). A common molecular apparatus is implicated in these events and may therefore be also expressed at the level of secretory granule membranes. However, to our knowledge, only granulophysin-type molecules have been described in a variety of regulated secretory cells (Gerrard et al., 1991; Abdelhaleem et al., 1991). Granulophysin (GP) and leukophysin, which react with antigranulophysin mAbs, are present in endocrine, exocrine and neuronal tissues, as well as in endothelial cells and in some leukocytes, but the involvment of these molecules in the physiology of secretory granules has not been clearly defined.

In exocrine or endocrine organs, during the first half of embryonic development the two GRL antigens are not detected by immunofluorescence with anti-GRL mAbs. However, during early stages of embryogenesis, various exocrine and endocrine tissues begin their glandular differentiation (Potvin and Aron, 1927; Wilson, 1949) as well as their secretory activity (Lutz and Case, 1925; Fugo, 1940). We suggest that GRL antigens are not detected by immunofluorescence in these early stages of glandular differentiation, probably because of the small size of the secretory granules (Gage, 1945), which contain a small amount of stored secretory products (Sun, 1932).

The subcellular localization of GRL1 and GRL2, and the associated morphometric studies, were conducted on immunogold-labeled pancreatic acinar cells. Various conditions of fixation and embedding of pancreatic tissues were tested. The best preservation and immunostaining of pancreatic tissues were obtained with a fixative of 4% paraformaldehyde and 0.1% glutaraldehyde. Under those conditions, the immunolabeling of GRL antigens with gold is prominent at the periphery of secretory zymogen granules and at the apical surface of the exocrine cells. Typical zymogen granules are characterized by a size of 0.5-1 μm in diameter and dense cores. The gold-labeling pattern observed at their periphery indicates that both GRL1 and GRL2 are expressed on, or near, the limiting membrane of secretory granules. The frequently observed clustering of gold particles is probably not due to preformed aggre-gates of antibodies, as gold-coupled antibodies were centrifuged immediately before use, in order to eliminate aggregates of gold particles. Rather, this pattern suggests a clustering of antigenic sites at the periphery of secretory granules. Within granules, GRL antigens are not detected in the secretory protein core. However, this observation does not rule out the existence of a secreted form of GRL antigens within the granule matrix. A possible condensation of proteins in the core of the secretory granule may hinder the accessibility of mAbs to the GRL antigens.

For technical reasons, morphometric analysis of cell surface labeling of acini was based on calculation of surface areas, rather than on the more precise calculation of numbers of points of intersections of a linear grid matrix with membranes. Therefore, the number of gold particles at the apical surface was probably underestimated, since most particles are on membranes rather than within the lumen of acini.

The high density of GRL antigens observed in light granules requires a special comment. These granules with clear matrices are usually localized at the apical side of cells. They may be artefactual structures resulting from the method of tissue fixation and cryo-sectioning, which causes retraction of the dense zymogen core from the granule membrane. These light granules may also correspond to zymogen granules after exocytosis of their content, or represent apical endocytic vesicles. The electron-microscopic data alone do not permit a definitive statement on the relationship of GRL antigens with the granule membranes. Biochemical studies have introduced more conclusive information concerning this problem. The two GRL antigens are extracted from membrane pellets by non-ionic detergents and therefore exist in a membrane-bound form. However, GRL1 and GRL2 differ significantly, on the basis of their biochemical features and mechanisms of insertion into membranes.

The GRL1 protein, extracted from granulocytes, thrombocytes, exocrine pancreas and adrenal glands, displays a constant pattern of migration in SDS-PAGE. The GRL1 epitope seems to be restricted to a 92-94 kDa proteic form. Therefore, GRL1 probably does not exhibit differential forms of glycosylation in hemopoietic, endocrine and exocrine tissues, as shown by other granule antigens, such as synaptophysin (Navone et al., 1986), secretogranin II (Rosa et al., 1985) and gGT proteins (Castle et al., 1985). On the other hand, a soluble form of GRL1, detected in the culture super-natant of activated granulocytes (Thomas et al., 1993), presents the same monodimensional electrophoretic pattern as the membrane-bound form. In fact, GRL1 displays the characteristics features of a peripheral membrane molecule. Moreover, the GRL1 protein probably does not anchor to membranes by means of a phosphatidyl-inositol link. Indeed, 100% of GRL1 is extracted from the granule membrane-enriched pellet with carbonate buffer and is soluble in non-ionic detergents, such as NP40 and Triton X-110. Moreover, it is not labeled after incubation with [³H]ethanolamine (data not shown), as phosphatidyl inositol-anchored proteins (GPI-anchored proteins) usually are (Rifkin and Fairlamb, 1985; Medof et al., 1986; Fatemi et al., 1987). Therefore, GRL1 is presumably not a GPI-anchored protein, like the GP2 molecule identified in the membrane of zymogen pancreatic granules (LeBel and Beattie, 1988; Fukuoka et al., 1991, 1992).

On the other hand, GRL2 is a highly N-glycosylated protein with the characteristic features of an integral membrane protein. It remains in membrane pellets after high-salt or alkaline treatments and is partitioned in the detergent phase of Triton X-114 treatment. No soluble form of GRL2 has been detected in the culture medium of activated granulocytes (Thomas et al., 1993).

According to these studies, GRL antigens, like granulophysin (Gerrard et al., 1991; Bainton and Gerrard, 1991), SG 170 (Yamashita and Yasuda, 1992) and GP-2 are proteins associated with storage granule membranes and widely expressed in regulated secretory cells. The membrane topologies of GRL1 and GRL2 suggest different roles with respect to distal events of the regulated secretory pathway. In this sense, further studies of the specific features presented by GRL2 (concentration on apical plasma membrane and at the periphery of apical light granules) may provide clues to the understanding of the complex pathway followed by this protein within secretory cells. Mechanisms such as transcytosis and apical endocytosis (for review, see Rindler, 1992; Kelly, 1993) may be implicated in the GRL2 pathway and are conveniently studied in the model of the chicken pancreatic acinar cell. On the other hand, the existence of soluble and membrane-bound GRL1 molecules poses the questions of the relationship between these two proteic forms and of their respective participation in the processes of secretory granule maturation. The cloning and sequencing of cDNAs for each of the GRL proteins will allow more detailed investigations of their respective functional properties.

We are indebted to Dr N. M. LeDouarin, Dr D. Louvard, Dr I. Maridonneau-Parini, and Dr B. Zalc for their helpful advice. We particulary thank M. F. Hallais for technical assistance, B. Henri and Y. Rantier for photographs, and S. Gournet for artwork. Dr N. Gonatas was supported by grant NS-05572 from NIH (USA). Financial supports were provided by the Centre National de la Recherche Scientifique (CNRS), and the Fondation pour la Recherche Medicale (FRM).