cis-Golgi resident proteins and O-glycans are abnormally compartmentalized in the RER of colon cancer cells

Cellular transformation is commonly associated with alterations in the glycosylation pattern and cellular distribution of glycolipids and glycoproteins (Hakomori, 1989; Schoenenberger and Matlin, 1991). In tumors derived from glandular epithelia, mucins are a major fraction of the glycoproteins synthesized by the tumor cells. Mucins are characterized by a high carbohydrate content (70-80% of their molecular mass) and their protein core is rich in serine, threonine and proline. Complex carbohydrate moieties are O-glycosidically linked to Ser or Thr residues in the peptide backbone (Neutra and Forstner, 1987). Glycosylation of mucins in tumors is both quantitatively and qualitatively abnormal: incomplete glycosylation leads to the accessibility of carbohydrate and peptide epitopes which are cryptic in mucins synthesized by normal epithelial cells (Springer, 1984; Burchell et al., 1987; Gendler et al., 1988; Longenecker et al., 1988; Takahashi et al., 1988; Schüssler et al., 1991); abnormal glycosylation leads to the expression of novel epitopes that are absent from normal mucins (Hako-mori, 1989). Some of the epitopes thus generated are preferentially expressed in tumor versus normal cells, although they are not strictly tumor-specific, since they are also expressed during development or in a few normal adult epithelial tissues (Hakomori, 1989). The abnormal expression of carbohydrate epitopes on membrane glycoproteins and glycolipids of tumor cells is likely to play a major role in cellular interactions (reviewed by Brandley et al., 1990).

The basis for the alterations in mucin glycosylation in epithelial tumors is not well known and could be both genetic and epigenetic, involving changes in the expression, localization and/or activity of the glycosyltransferases that participate in the synthesis of the carbohydrate moieties.

To understand better the events responsible for altered mucin glycosylation we have examined the subcellular distribution of an epitope, GalNAcα-O-Ser/Thr, which is expressed at higher levels in tumor mucins than in normal mucins. GalNAcα-O-Ser/Thr results from the first glycosylation step of the core peptide and is the precursor structure of more complex O-glycosidically linked carbohydrate chains. Several studies have shown enhanced expression of this precursor molecule in tumor cells (Takahashi et al., 1988; Itzkowitz et al., 1989; Schüssler et al., 1991) and the studies presented here demonstrate the abnormal compartmentalization of the GalNAcα-O-Ser/Thr epitope in tumor cells. On the basis of these findings, we have examined the subcellular localization of RER and Golgi markers in normal colon and in colon cancer tissues. Altered distribution of resident Golgi proteins and enzymes involved in mucin synthesis may contribute to the abnormal glycosylation of these molecules in tumor cells (Hull and Carraway, 1988).

Tetrachloroauric acid, trisodium citrate, polyethylene glycol (M_r 20,000), paraformaldehyde, Tween-20 and ammonium chloride were from Merck (Darmstadt, Germany); glutaraldehyde (25% in water) was from Fluka (Buchs, Switzerland). Lectins: DBA and VVA were from Vector Laboratories (Burlingham, CA); LFA was from Calbiochem-Behring (La Jolla, CA); HPL was purchased from Pharmacia (Lund, Sweden). Exo-α-N-acetylgalactosaminidase from chicken liver (EC 3.2.1.49), Triton X-100, fetuin, N-acetylgalactosamine (GalNAc) and N-acetyl-glucosamine (GlcNAc) were from Sigma (St. Louis, MO). Rabbit anti-mouse Ig was obtained from Dako (Dakopatts, Glostrup, Denmark), fluorescein isothiocyanate-labeled goat anti-mouse Ig, rhodamine-labeled goat anti-mouse Ig and goat anti-rabbit Ig were from Zymed Laboratories (South San Francisco, CA), and streptavidin-FITC was from Vector Laboratories. Protein A was obtained from Pharmacia and streptavidin-gold complexes were purchased from Biocell Laboratories (Cardiff, United Kingdom).

The mAb Cu-l (IgG3) detects the GalNAcα-O-Ser/Thr epitope (Takahashi et al., 1988). It was obtained from Dr S.I. Hakomori (The Biomembrane Institute, Seattle, WA) and was purified from hybridoma culture medium using Protein A-Sepharose affinity chromatography. mAb LDQ10 (IgM) was raised against deglycosylated mucin isolated from LS174T colon cancer xenografts and detects a peptide epitope of the MUC2 intestinal mucin tandem repeat sequence (Gambús et al., 1993). mAb LDQ10 was biotinylated as previously described (Goding, 1983). Rabbit anti-protein disulphide isomerase (PDI) was a generous gift from Dr José G. Castaño (Instituto de Investigaciones Biomédicas, CSIC, Madrid); it detects PDI, a resident RER protein (Nieto et al., 1990). Rabbit serum 15C8 was generously provided by Dr Ignacio Sandoval (Centro de Biologia Molecular, CSIC, Madrid, Spain); it detects a Golgi resident protein of 130 kDa localized to the cis cisternae of the Golgi apparatus (Yuan et al., 1987).

HT-29 MTX cells are a subpopulation of mucus-secreting cells selected by adaptation of the HT-29 cell line (Fogh and Trempe, 1975) to 10⁻⁶ M methotrexate (MTX); these cells, which maintain their differentiation characteristics when subcultured in the absence of drug, were routinely grown in Dulbecco’s minimal essential medium as described (Lesuffleur et al., 1990). In other experiments, as indicated in the text, subconfluent HT-29 MTX cultures were used. HT-29 MTX cells were studied 15 days after achieving confluence, since at this stage they exhibit a goblet celllike phenotype (Lesuffleur et al., 1990). WiDr colon cancer cells were cultured in RPMI medium supplemented with insulin, transferrin and selenium (WiDr-ITS) as previously described (Real et al., 1991). Cell monolayers (HT-29 MTX) or cell pellets (WiDr-ITS) were fixed with 3% paraformaldehyde, 0.1-0.5% glutaraldehyde in 0.1 M phosphate buffer, pH 7.4, for 30 min at 22°C. Fixed cells were briefly rinsed with the same buffer and free aldehyde groups were quenched with 50 mM NH₄Cl in PBS for 1 h. Cells were embedded in Lowicryl K4M at −35°C as previously described (Roth et al., 1981; Carlemalm et al., 1982). Low-temperature embedding with Lowicryl K4M was used because it provides an acceptable compromise between antigenic and morphologic preservation. Three independent specimens of normal human colon and colon cancer tissues were collected by endoscopy; samples selected for analysis were obtained from the periphery of the lesion to ensure that areas of viable tumor were taken. Samples were rinsed with PBS and fixed with 3% paraformaldehyde/0.1% glutaraldehyde in PBS for 2 h at room temperature. Fixed colon samples were rinsed several times with PBS and free aldehyde groups were quenched with NH₄Cl (50 mM in PBS) for 60 min. After rinsing with PBS, samples were processed for embedding in Lowicryl K4M as described above. Ultrathin sections (60-80 nm) were cut with glass knives, and placed on parlodion/carboncoated nickel grids, and stored until used.

Monodisperse colloidal gold particles with an average diameter of 14 nm or 10 nm were prepared according to previously described procedures (Frens, 1973; Slot and Geuze, 1985), respectively. Protein A, HPL and DBA were conjugated to gold as previously described (Roth, 1983). Fetuin was coupled to gold particles as reported (Roth et al., 1984). To detect GalNAc and GalNAcα-O-Ser/Thr, VVA-gold complexes were prepared (G. Egea, unpublished results).

Ultrathin sections were placed on parlodion/carbon-coated nickel grids, floated on a drop of PBS for 5-10 min, and transferred to a drop of 14 nm lectin-gold complexes. Grids were rinsed twice with PBS, once with distilled water, and air dried. To detect sialic acid the two-step cytochemical affinity technique was used (Roth et al., 1984). Incubation times were 30 min for HPL-, VVA-, DBA- and fetuin-gold complexes, and 45 min for LFA. Lectin concentrations used are listed in Table 1. Thin sections were stained with uranyl acetate (3% in distilled water) for 6 min and lead citrate for 45 s.

Thin sections were floated on a droplet of PBS for 5-10 min at 22°C and then incubated with mAb Cu-l or mAb LDQ10. Sections were incubated with mAbs for 2 h at 22°C or for 18 h at 4°C in a moist chamber, rinsed twice with PBS, and then incubated with affinity-purified rabbit anti-mouse IgG (25-50 μg/ml in 0.1% Tween-20 in PBS) for 45-60 min at 22°C. After two washes with PBS, grids were floated on a droplet of Protein A-gold for 1 h. Sections were then washed twice with PBS, once with distilled water, and then allowed to air dry. Staining with uranyl acetate and lead citrate was as described above. Rabbit sera 15C8 and anti-PDI were used at a 1/10 to 1/100 dilution in PBS containing 2% skim milk. Reactions with rabbit sera were developed using Protein A-gold complexes, as indicated above (Roth et al., 1978).

The first labeling reaction was performed as described above. Subsequently, sections were floated on a droplet of glutaraldehyde (1% in PBS) for 10-15 min, washed twice with PBS, and transferred to a droplet of 50 mM NH₄Cl in PBS for 45 min. After two further washes with PBS sections were processed for the second labeling reaction. When the double-labeling technique included one lectin (HPL, LFA) and one mAb (Cu-l), the procedure used for the second labeling reaction was the same described above. When two mAbs were used in double-labeling experiments, the reaction of the first mAb (Cu-l) was developed using rabbit anti-mouse Ig and Protein A-gold (10 nm) and the second reaction was performed using biotinylated mAb LDQ10 and streptavidin-gold complexes (14 nm). When a rabbit serum (PDI or 15C8) and mouse mAb Cu-l were used in double-labeling experiments, the reactivity of the rabbit serum was developed with Protein A-gold; tissues were then fixed with 1% glutaraldehyde, free aldehyde groups were blocked with NH₄Cl and mAb Cu-l immunolabeling was performed as described above. When the two rabbit sera were used in double-labeling experiments, the same procedure as for single labeling was used but glutaraldehyde fixation and NH₄Cl quenching steps were added in between the two labeling procedures, as described above.

To assess the specificity of the reactions of the lectins and mAbs used in these studies, several controls were used. Inhibition with monosaccharides was performed by pre-incubating the lectins (45 min) or mAb Cu-l (2 h) with α-GalNAc (10-500 mM) or α-GlcNAc (50-500 mM). Enzymatic treatment of ultrathin sections with exo-α-N-acetylgalactosaminidase was performed for 18 h at 37°C (1 unit/ml in 0.1 M citrate, pH 4.5). Following enzyme treatment thin sections were extensively rinsed in distilled water and processed for lectin or mAb cytochemistry as described above. Positive controls always corresponded to identical treatments in the absence of monosaccharides or enzymes. Other controls included incubation of sections with fetuin-, streptavidinor Protein A-gold complexes in the absence of first or second antibody, use of isotype-matched unrelated first antibodies, and use of unlabeled lectin to inhibit the binding of lectin-gold complexes.

Quantitative estimations were performed on enlarged negatives at a final magnification of ×55,000. The section areas were measured using a plamimeter (Digicard Plus, Kontron Daten System, Rotkenz, Switzerland) and the number of gold particles was counted manually.

Most human colon cancer cell lines do not show differentiated properties (Zweibaum et al., 1991). Less than 5% of HT-29 cells and less than 5% of WiDr-FBS cells are differentiated. In contrast, 100% of HT-29 cells selected by adaptation to 10⁻⁶ methotrexate (HT-29 MTX) and approximately 30% of WiDr-ITS cells produce and secrete mucins (Lesuffleur et al., 1990; Real et al., 1991).

The general ultrastructural features of the mucin-producing HT-29 MTX and WiDr-ITS cell lines have previously been described (Lesuffleur et al., 1990; Real et al., 1991). HT-29 MTX cells only display a goblet cell phenotype in the post-confluent state. In subconfluent HT-29 MTX cultures, 10% of cells showed swollen RER cisternae; this phenotype was observed in 80% of cells in late post-confluent cultures. In most cells, the Golgi complex was located along the basal-apical axis, as in intestinal absorptive cells. In HT-29 MTX cells, mucus droplets were clearly polarized in a supranuclear position. The RER of 5% of WiDr-ITS cells was also markedly swollen and sometimes extended to the nuclear envelope. In more than 90% of WiDr-ITS cells, typical Golgi stacks were not observed, probably due to the lack of polarization of these cells, and mucus droplets were distributed in a non-polarized fashion.

Table 1 shows the specificity of the antibodies and lectins used in this study. mAb Cu-l and lectins that recognize terminal non-reduced GalNAc showed a heterogeneous pattern of labeling of mucin droplets in both cultured cell types (Fig. 1A-C). Helix pomatia lectin (HPL) labeled some mucin droplets (Fig. 1B), whereas none of them were labeled by Dolichos biflorus agglutinin (DBA) or Vicia vil -losa agglutinin (VVA) (data not shown).

Double-labeling experiments were performed to establish if the epitopes recognized by mAb Cu-l and lectins colocalized. When mAb Cu-l was used in combination with HPL or LFA (Fig. 1D) three patterns of mucin staining were observed: (1) some droplets were labeled with both mAb and lectin; (2) some droplets were labeled with the lectin but not with the mAb; and (3) some droplets were not labeled at all. As the pattern of labeling of mAb Cu-l was always more restricted than the lectin pattern, the results indicate that the glycosylation of cellular mucins includes terminal non-reduced GalNAc and NeuAc residues different from GalNAcα-O-Ser/Thr.

In addition to the labeling of mucin droplets, mAb Cu-l also labeled the RER in 90% of HT-29 MTX cells and in a lower proportion of WiDr-ITS cells (Fig. 2A,C) The RER was identified morphologically on the basis of the presence of membrane-associated ribosomes; this characteristic allowed proper recognition of swollen RER cisternae and discrimination from other electron-lucid subcellular structures such as mucin droplets. To confirm the presence of terminal GalNAc in the lumen of the RER of both cultured colon cancer cells types, GalNAc-specific lectins were used (Fig. 2B,D). Gold particles were observed in the RER and, in some cases, at sites where the nuclear envelope was connected with the RER (Fig. 2B). To confirm further the specificity of the RER labeling we used the enzyme exo-α-N-acetylgalactosaminidase on ultrathin sections of HT-29 MTX cells. As shown in Fig. 2E and F, labeling of the RER by mAb Cu-l specifically disappeared after enzyme treatment. These results were also confirmed using the DBA lectin. The effect of exo-α-N-acetylgalactosaminidase was substantiated by the specific disappearance of cis-Golgi labeling in ovine submaxillary gland mucous cells after enzyme treatment (data not shown). In addition, the presence of GalNAc in the RER of HT-29 MTX cells was demonstrated by the specific inhibition of the reactivity of mAb Cu-l by GalNAc but not by GlcNAc (data not shown). On the basis of these findings we conclude that GalNAcα-O-Ser/Thr is present in the RER of cultured colon cancer cells. As described above, the proportion of WiDr-ITS cells showing swollen RER cisternae containing GalNAc-O-Ser/Thr was lower than the proportion of HT-29 MTX cells, possibly reflecting the greater cell heterogeneity observed in the former cultures (Real et al., 1991).

To determine if the prior findings were related to in vitro culture of colon cancer cells, we examined the reactivity of mAb Cu-l with normal human colon epithelium and with samples from three independent colon cancers. To identify the RER and the Golgi complex, both ultrastructural features and marker proteins were used. In goblet cells of normal colonic epithelium mAb Cu-l labeled the cis-Golgi stacks and small vesicles situated between the RER and the first continuous Golgi cisterna. In double-labeling experiments, GalNAc-O-Ser/Thr was detected with mAb Cu-1 and sialic acid with the LFA-fetuin gold complex. These markers always showed reciprocal labeling patterns, indicating that GalNAc-O-Ser/Thr is restricted to the cis-Golgi (Fig. 3A). PDI, a resident RER protein, and GalNAcα-O-Ser/Thr were located in the RER and the cis-Golgi of normal colon, respectively (Fig. 3B). In contrast, both molecules co-localized in swollen cisternae in colon cancer tissues (Fig. 3C). mAb LDQ10 detects a peptide epitope of the tandem repeat sequence of the MUC2 intestinal mucin. In normal colon, mAb LDQ10 labeled the RER but not the cis-Golgi, which was exclusively labeled by mAb Cu-1 (Fig. 3D; Table 2). In colon cancers, the epitopes detected by both mAbs co-localized in the RER (Fig. 3E). Since RER swelling is often accompanied by a decreased density of membrane-associated ribosomes, double-labeling experiments were performed to distinguish the RER clearly from mucin droplets. Sialic acid (detected with LFA) and the MUC2 apomucin tandem repeat (detected with mAb LDQ10) showed reciprocal labeling patterns (Fig. 4). Altogether these experiments demonstrate that GalNAcα-O-Ser/Thr abnormally colocalizes with RER markers in swollen RER cisternae in colon cancer tissues.

The previous results suggest that a reorganization of the Golgi complex occurs in cultured colon cancer cells and in colon cancer tissues. To examine this hypothesis, the localization of a 130 kDa GRPc (Yuan et al., 1987) was analysed. Double indirect immunofluorescence microscopy studies of WiDr-ITS and HT-29 MTX, simultaneously stained with mAb Cu-l and polyclonal rabbit serum 15C8, displayed a coincident fluorescence Golgi pattern (data not shown). In addition, immunoelectron microscopy with rabbit serum 15C8 on WiDr-ITS cells and HT-29 MTX cells showed gold particles associated with the Golgi complex and the RER (Table 2). Double-labeling experiments with mAb Cu-l and rabbit serum 15C8 demonstrated the simultaneous presence of GalNAcα-O-Ser/Thr and GRPc in a morphologically well-structured Golgi apparatus (Fig. 5A) and in the swollen RER of the two cultured colon cancer cell lines studied (Fig. 5B,C). To confirm that serum 15C8 indeed recognizes a GRPc, normal human colon tissue was examined. Serum 15C8 labeled the cis-Golgi cisternae of normal colonic goblet cells (Fig. 6A), but not the RER (Table 2). In colon cancer tissues, the GRPc colocalized in the lumen of swollen RER cisternae with Gal-NAcα-O-Ser/Thr (Fig. 6B), PDI (Fig. 6C) and the MUC2 tandem repeat sequence detected by mAbLDQ10 (Fig. 6D). These findings demonstrate that a cis-Golgi resident protein is abnormally redistributed to the RER in colon cancer tissues.

To rule out the possibility that the abnormal compartmentalization of Golgi markers in the RER of differentiated HT-29 MTX cells could be related to stress or to ‘aging’ of the cultures, we analyzed the subcellular localization of GalNacα-O-Ser/Thr and the 130 kDa GRPC in subconfluent cultures. The two markers were found in the RER cisternae (Fig. 7).

We have used immunocytochemical techniques in an attempt to understand the subcellular basis for the abnormal glycosylation of mucins in epithelial cancer cells. The experiments reported here demonstrate that a disorganization of the Golgi apparatus occurs in cultured tumor cells and in tumor tissues, leading to the abnormal localization of GalNAcα-O-Ser/Thr and a resident protein of the cis-Golgi in the RER. The observations were supported by the quantification of gold immunolabeling in the various cellular compartments examined.

In mucin-producing cultured colon cancer cells, and in colon cancer tissues, GalNAcα-O-Ser/Thr was found in the RER, whereas this epitope was only detected in the cis-Golgi of normal colonic goblet cells.

The subcellular compartment where O-glycosylation is initiated is still debated, possibly because of cell-, tissueor species-specific differences (Carraway and Hull, 1989). In addition, the diverse limitations of different methodological approaches (i.e. biochemical versus ultrastructural) may account for discordant findings. The first indication that O-glycosylation is initiated in an early compartment came from the isolation of O-glycans from purified peptidyl-tRAS in gastric polysomes (Strous, 1979); subcellular fractionation analysis of the distribution of UDP-GalNAc:polypeptidetransferase activity in hen oviduct microsomes (Hanover et al., 1980) and biochemical analysis of the maturation of human chorionic gonadotropin hormone in BeWo cells labeled in vitro (Hanover et al., 1982) suggested that O-glycosylation takes place after N-glycosylation has been initiated. Other studies using electron microscopic cytochemical techniques have demonstrated the presence of GalNAc in the Golgi apparatus (but not in the RER) of rat intestinal goblet cells (Roth, 1984), porcine submaxillary gland (Deschuyteneer et al., 1988), and the human erythroleukemic cell line K562 (Piller et al., 1989). Finally, some studies have shown the participation of both early (presumably RER) and late (presumably Golgi complex) compartments in the first step of O-glycosylation (Spielman et al., 1987, 1988). In normal human colonic goblet cells, GalNAcα-O-Ser/Thr was not found in the RER using immunocytochemical techniques, indicating that most GalNAc is transferred in the Golgi complex, although initiation of O-glycosylation in earlier compartments cannot be excluded. Several mecanisms might explain the predominant location of the GalNAcα-O-Ser/Thr epitope in the RER of colon cancer cells: (a) O-glycosylation might be a co-translational event resulting from the abnormal localization of the enzyme UDP-GalNAc:polypeptidetransferase in the RER; (b) O-glycosylation could be initiated in a putative intermediate compartment (Tooze et al., 1988) but further maturation would be blocked, resulting in an accumulation in the RER along with the salvage pathway of the RER membrane and resident proteins (Pelham, 1989); and (c) O-glycosylation would start in the cis-Golgi stacks and accumulation in the RER would result from an unbalanced anterograde and retrograde RER-Golgi flow. These options are not necessarily mutually exclusive, since recycling of the active UDP-GalNAc:polypeptidetransferase to the RER might lead to the initiation of O-glycosylation in this compartment. Several studies have described the activity of Golgi enzymes in the RER in cells treated with BFA. Galactosylation and sialylation of glycophorin precursors in the RER has been reported in erythroleukemia cells (Ulmer and Palade, 1989), and the acquisition of resistance to endoglycosidase H in the RER in the absence of acquisition of sialic acid has been demonstrated in CHO cells and in the 2B4 T cell hybridoma (Doms et al., 1989; Lippincott-Schwartz et al., 1989). Abnormal O-glycosylation of ribophorin I and of galactosyltransferase have also been described in cells treated with BFA (Bosshart et al., 1991; Ivessa et al., 1992). Preliminary data indicate that trans-Golgi markers (i.e. β-galactose and sialic acid) can also be detected in the RER of HT-29 MTX cells and in colon cancer tissues (unpublished observations).

To determine if redistribution of Golgi resident proteins could contribute to the presence of GalNAcα-O-Ser/Thr epitope in the RER, we examined the subcellular distribution of a 130 kDa GRPc in cultured colon cancer cells, normal colon and colon cancer tissues. Rabbit serum 15C8 detecting GRPc showed reactivity with the RER of cultured colon cancer cells and colon cancer tissues, but not with normal colonic cells. That this compartment is indeed the RER was further demonstrated by the co-localization of PDI (a resident RER protein) and the core protein sequence of the MUC2 intestinal mucin. The colocalization of Gal-NAcα-O-Ser/Thr and a GRPc in the RER of cultured colon cancer cells and tissues demonstrates the aberrant redistribution of GRPc to this compartment. If UDP-GalNAc:polypeptidetransferase is located in the cis-Golgi compartment of normal colonic goblet cells, it may also be redistributed to the RER in colon cancer cells, where it may be active, as in BFA-treated cells (Bosshart et al., 1991; Ivessa et al., 1992). Antibodies detecting this enzyme and/or subcellular fractionation studies may be of help in understanding the accumulation of GalNAcα-O-Ser/Thr in the RER of colon cancer cells.

Irrespective of where O-glycosylation is initiated, our findings indicate that cis-Golgi resident proteins are abnormally distributed in colon cancer cells.

The abnormal compartmentalization of Golgi resident proteins and GalNAcα-O-Ser/Thr was associated with a morphological swelling of RER cisternae in post-confluent HT-29 MTX cultures, but not in subconfluent cultures. In addition, an increased number of uncoated vesicles of variable size, and a disorganized Golgi complex were observed. Similar morphological abnormalities were present in a variable proportion of cells in colon cancer biopsies, probably reflecting the phenotypic heterogeneity of tumors (Heppner, 1984). Similar swelling of the RER, disruption of the Golgi apparatus and altered topography of Golgi resident proteins have been described in a variety of cell types cultured in the presence of BFA (Lippincott-Schwartz et al., 1989; Ulmer and Palade, 1989; Fujiwara et al., 1988; Chege and Pfeffer, 1990; Shite et al., 1990). In these cells, the retrograde pathway of membrane cycling from Golgi to the RER becomes evident due to the effective blockade that BFA exerts on the exit of proteins from the RER to the Golgi apparatus and the return of Golgi contents to the RER in a microtubule-dependent process (Lippincott-Schwartz et al., 1990). Recent data suggest that this effect could be related to the ability of BFA to induce the release of a 110 kDa protein, β-COP (Waters et al., 1991), from the Golgi membranes prior to the appearance of any morphological changes (Donaldson et al., 1990; Serafini et al., 1991).

Altered structure and/or function of the gene products involved in intracellular traffic, or the presence of endogenous BFA-like molecules, might be involved in the generation of the abnormal phenotype observed in colon cancer cells. The availability of reagents to detect the proteins involved in the exocytic pathway should facilitate the analysis of their contribution to the abnormal phenotype described here.

The accumulation of glycosylated mucin in the RER might in itself contribute to the abnormal phenotype of this compartment in tumor cells. The extensive O-glycosylation of the mucin polypeptide leads to an unusually high radius of gyration of these molecules, which is associated with a stiff conformation. Light-scattering and NMR studies of ovine submaxillary mucin and its deglycosylated derivatives have shown that the acquisition of GalNAc has a major role in determining this inflexible conformation (Shogren et al., 1989; Gerken et al., 1989; Jentoft, 1990). In the RER, these molecules could be recognized as improperly folded structures leading to a heat-shock response, as has been described in other cell systems. The elevated level of heat-shock proteins in some tumor cells supports this notion (Tandom et al., 1990). The accumulation of heatshock proteins in the RER might lead to a further blockade in the RER to Golgi traffic as well as to further alterations in protein glycosylation (Lee, 1987).

In this study, two types of abnormalities were observed: (1) redistribution of Golgi markers to the RER and (2) morphological swelling of RER cisternae. While the former was clearly unrelated to culture conditions (prevs. postconfluent), the latter was much more prominent in the postconfluent state. In this state, cells were viable, could be maintained for more than 60 days after reaching confluence (T. Lesuffleur, unpublished observations) and could be subcultured. However, it is possible that RER swelling may be related to cellular stress, neoplastic transformation, or both.

Swelling of RER cisternae and accumulation of Golgi enzymes and oligosaccharide sequences in the RER have recently been described in a temperature-sensitive CHO secretory mutant (Zuber et al., 1991). In addition, abnormal accumulation of proteins in the RER has been described in storage diseases (reviewed by Carlson, 1990). Fibroblasts from individuals with familial hypercholesteremia bearing mutations in the LDL receptor display an abnormal accumulation of this protein, as well as the presence of O-linked sugars, in the RER (Pathak et al., 1988). Accumulation of secretory proteins in the RER has also been described in patients with α-1-antitrypsin and α-1-antichymotrypsin deficiencies (Sifers et al., 1989; Lindmark et al., 1990). In other cases in which synthesis of abnormal proteins occurs, the RER seems to be the major site of degradation of these molecules (reviewed by Klausner and Sitia, 1990). The abnormal accumulation of proteins in RER cisternae is a selective event, as recent studies on Russell bodies in plasmacytoma cells have shown (Valetti et al., 1991), and does not represent a complete blockade in the transport from the RER to later compartments (Colley et al., 1991). Furthermore, structural alterations of the Golgi apparatus similar to those described here have recently been reported to occur in two human epidemoid cancer cell lines, although those studies were carried out using indirect immunofluorescence, a technique with lower resolution capacity (Seguchi et al., 1992). Altogether, these findings suggest that abnormal accumulation and degradation of proteins of the exocytic pathway in the RER may be a common finding in states of cellular pathology, including neoplastic transformation.

This work was supported in part by grants from the Fondo de Investigaciones Sanitarias, DGICYT, and the Fundació Knickerbocker. The authors would like to acknowledge Drs J. G. Castaño and I. Sandoval for providing antibodies, Drs T. Thomson, A. Garcia de Herreros, I. Sandoval, O. Stutman and G. Griffiths for valuable discussions and critical reading the manuscript, and N. Llinás for help in the preparation of the manuscript.