(Received for publication, November 13, 1996, and in revised form, February 18, 1997)
From the Department of Medical Biochemistry, University of Göteborg, Medicinaregatan 9, S-413 90 Göteborg, Sweden
This paper describes the biosynthesis of L-CanAg,
a mucin-like glycoprotein which carries the carcinoma-associated
carbohydrate epitope sialyl-Lewis a and is secreted by the colon
adenocarcinoma cell line COLO 205. Recently, it has been shown that
L-CanAg is a novel glycoform of CD43, a surface sialoglycoprotein
normally found only on hematopoietic cells. Immunoprecipitation with
-GPEP18, a novel antiserum against the cytoplasmic domain of CD43,
detected a transmembrane form of L-CanAg carrying sialyl-Lewis a. Cell surface biotinylation experiments demonstrated the presence of transmembrane L-CanAg at the plasma membrane and that COLO 205, unlike
the leukocyte cell line HL-60, contained significant amounts of
glycosylated intracellular CD43.
Immunoprecipitation of phosphate-labeled COLO 205 cells revealed that membrane-bound L-CanAg, like leukocyte CD43, is a phosphoprotein. Interestingly, both surface- and phosphate-labeled L-CanAg were eluted earlier from a gel filtration column than their unlabeled counterparts, indicating that this method could separate membrane-bound L-CanAg from its soluble form. Immunoprecipitations of pulse-chase-labeled COLO 205 lysates fractionated by gel filtration showed that decrease in membrane-bound L-CanAg was concomitant with an increase in the intracellular soluble form. Together, these data indicate that transmembrane L-CanAg is fully glycosylated and phosphorylated before the extracellular domain is cleaved off and stored inside the cell before exocytosis.
Monoclonal antibodies (mAbs)1 detecting carcinoma-associated antigens have often been shown to bind epitopes on mucins. The common denominator of these molecules is the mucin domain, a more or less repetitive sequence rich in proline and serine and/or threonine, the two latter amino acids being extensively O-glycosylated (1, 2). The high abundance of proline and carbohydrate chains makes the mucin domain a very extended, hydrophilic and protease-resistant structure. Mucins are thought to protect mucosae from chemical, mechanical, and microbial damage by restricting the mobility of water in the vicinity of the apical membrane. Glycoproteins containing mucin domains have also been found in non-epithelial cells such as endothelial cells and leukocytes, where they are involved in both adhesive and anti-adhesive events (3-5).
Tumor-associated epitopes on mucins are generated by abnormal
glycosylation. One of these epitopes is the carbohydrate structure sialyl-Lewis a (Si-Lea,
NeuAc2
3Gal
1
3[Fuc
1
4]GlcNAc
1
; Refs. 6 and 7).
Using the colon adenocarcinoma cell line COLO 205 as a model
system, previous studies have identified two mucin-like glycoproteins called H-CanAg and L-CanAg, which are the major
Si-Lea-carrying glycoproteins in these cells (8).
H-CanAg, which has an apparent molecular mass of 600-800 kDa, was
shown to be a MUC1 mucin, i.e. the apoprotein part of the
molecule is the product of the MUC1 gene. MUC1 mucins are
transmembrane glycoproteins common to apical membranes of secretory
epithelia, and overexpression and/or abnormal glycoforms of MUC1 are
often found in carcinomas (9). L-CanAg (apparent molecular mass = 150-300 kDa), which is secreted from COLO 205 cells at a significantly
higher rate than H-CanAg, was recently shown to have an apoprotein that
is identical to that of CD43 (leukosialin) (10), normally a major membrane sialoglycoprotein of leukocytes and platelets (11, 12). That
finding strengthens existing hypotheses (9) that carcinoma cells may
gain a selective advantage, such as immune evasion or facilitated
dissemination, by the expression of mucin-like glycoproteins,
either by overexpressing normally present mucins such as MUC1 or by the
anomalous expression of foreign glycoproteins such as CD43. Since the
identification of L-CanAg as a CD43 glycoform, expression of CD43 in
other non-hematopoietic tumors has also been reported
(13).2
When addressing the question of the possible pathological significance of these glycoproteins, knowledge of their metabolism is important. In the present study, I have investigated events in the post-translational modifications of L-CanAg, which are shown to follow a sequence of glycosylation, phosphorylation, and creation of a soluble mucin-like proteolytic fragment which is retained inside the cell and then released into the medium.
The colon adenocarcinoma
cell line COLO 205 and the promyelocytic cell line HL-60 were cultured
as described previously (10). The Si-Lea-reactive mAbs C241
(IgG1) and C50 IgM were gifts generously provided by CanAg Diagnostics
AB, Göteborg, Sweden. The mAb Leu-22, which recognizes leukocyte
glycoforms of CD43, was purchased from Becton Dickinson, San
José, CA. The peptide GPEP18 (CGRRKSRQGSLAMEELKS), corresponding
to amino acids 327-344 in the CD43 peptide sequence (except for the
N-terminal cysteine) was synthesized and coupled to keyhole limpet
hemocyanin by Scandinavian Peptide Synthesis, Köping, Sweden. The
conjugated peptide was used to generate the antiserum -GPEP18 by
immunizing a New Zealand White rabbit as described earlier (14).
COLO 205 cells grown to approximately 75% confluence in 35-mm Petri dishes or 5 × 106 HL-60 cells were used in each experiment. For methionine incorporation experiments, cells were starved for 1 h in minimal essential medium without methionine (Life Technologies, Paisley, Scotland) containing 5% fetal calf serum and 2 mM glutamine before labeling with Pro-Mix (Amersham, United Kingdom) corresponding to 150 µCi of [35S]methionine per dish. In pulse-chase experiments, chase was initiated by changing the medium to regular culture medium (see above). For phosphate labeling experiments, cells were starved for 1 h in Dulbecco's modified Eagle's medium without phosphate (Life Technologies) containing 5% fetal calf serum and 25 mM HEPES, pH 7.4, before labeling with 200 µCi of [32P]orthophosphate (Amersham) per dish. Cells were washed twice with cold phosphate-buffered saline before being lysed in 1 ml of the lysis buffer described in Ref. 10. In phosphate labeling experiments, the lysis buffer was supplemented with 10 mM sodium fluoride.
Cell-surface BiotinylationBefore surface biotinylation, COLO 205 cells were detached from the substratum using 0.02% EDTA in phosphate-buffered saline. Biotinylation with sulfosuccinimidoyl 6-(biotinamido)hexanoate (NHS-LC-biotin, Pierce, Oud Beijerland, Netherlands) was performed as described by Litvinov and Hilkens (15). Biotinylation of cell-surface glycoconjugates using sodium metaperiodate oxidation followed by treatment with biotin hydrazide (Sigma) was performed according to Orr (16) with the following modifications: the concentration of sodium metaperiodate was 5 mM, biotin hydrazide rather than 2-iminobiotin hydrazide was used at a concentration of 5 mM, and the biotinylation was performed overnight on ice. Biotinylated cells were lysed as described (10).
Gel FiltrationGel filtration was performed in a Superose 6 HR 10/30 column (Pharmacia Biotech, Uppsala, Sweden) using 0.1 M ammonium acetate or 50 mM Tris-HCl, pH 7.5, 4 M guanidinium chloride as eluant at a flow rate of 0.25 ml/min.
ImmunoprecipitationImmunoprecipitations were performed as
described (10). Eluates from gel filtration were immunoprecipitated
afer addition of 0.25 volumes of a 5 × lysis buffer concentrate
(lysis buffer, see Ref. 10). Eluted immunoprecipitates were subjected
to SDS-PAGE with 2.75-15% gradient separation gels and 2.75%
stacking gels. Radiolabeled immunoprecipitates were visualized by
autoradiography of dried gels after fixation in 30% ethanol, 10%
acetic acid for 3 h followed by 30 min treatment with Amplify
solution (Amersham) containing 5% glycerol.
Biotinylated and unlabeled samples were transferred from SDS-PAGE to nitrocellulose (Sartorius, Göttingen, Germany) using a semi-dry Sartoblot cell (Sartorius) and a transfer buffer consisting of 48 mM Tris, 39 mM glycin, 1.3 mM SDS, and 20% (v/v) methanol. Transfer was performed at 2-4 mA/cm2 of membrane for 1-2 h. Membranes were subsequently rinsed with water and blocked with 50 mM Tris-HCl, 0.15 M NaCl, pH 7.5 (TBS), containing 3% bovine serum albumin overnight at room temperature. The membrane was incubated with primary antibody diluted to 1-2 µg/ml in TBS containing 0.05% Tween 20 (TBST) for 4 h. After 3 washes with TBST, the membrane was incubated for 3 h with alkaline phosphatase-conjugated rabbit anti-mouse antiserum (DAKO immunoglobulins, Glostrup, Denmark) diluted 1/2000 in TBST, or with alkaline phosphatase-conjugated streptavidin diluted in TBST containing 0.5 M NaCl. After 3 washes with TBST and one wash with TBS, the membrane was developed using 0.3 mg/ml nitro blue tetrazolium and 0.15 mg/ml 5-bromo-4-chloro-3-indolyl phosphate in 0.1 M Tris-HCl, pH 9.5, 0.1 M NaCl, 5 mM MgCl2.
Rabbits were immunized with a keyhole limpet
hemocyanin-conjugate of GPEP18, a 17-mer peptide identical in sequence
to amino acids 327-344 in the cytoplasmic domain of CD43. The
resulting antiserum, called -GPEP18, was tested for reactivity in
immunoprecipitation experiments with lysates of pulse-chase labeled
cells. In the promyelocytic cell line HL-60, known to express one of
the major leukocyte-associated glycoforms of CD43 (17),
-GPEP18
immunoprecipitated one band of apparent molecular mass of 55 kDa
immediately after the pulse labeling (Fig. 1). After 20 and 60 min of chase, another band of 120 kDa gained in intensity
concomitant with the fading of the 55-kDa band. The apparent molecular
sizes and pulse-chase kinetics of the 55-kDa and 120-kDa bands were in
agreement with those reported for the CD43 primary precursor and fully
glycosylated forms, respectively (17). In the colon adenocarcinoma cell
line COLO 205,
-GPEP18 immunoprecipitated a precursor band with the same electrophoretic mobility as in HL-60 cells. However, in contrast to the sharp 120-kDa band seen in HL-60 cells, the glycosylated CD43 in
COLO 205 appeared as a broad and diffuse smear straddling the 200-kDa
marker position. This electrophoretic mobility is typical for that of
L-CanAg, the Si-Lea-carrying mucin-like molecule which has
been biochemically purified from COLO 205 cells (8) and subsequently
shown to be a novel CD43 glycoform (10). It could therefore be
concluded that
-GPEP18 can be used as a reagent for
glycosylation-independent detection of transmembrane CD43.
Transmembrane CD43 in COLO 205 Cells Is Glycosylated with Si-Lea
CD43 glycosylated with Si-Lea can
be recovered from secretions of COLO 205 cells in the form of the
L-CanAg glycoprotein (10). As the secretion of L-CanAg most likely
entails the proteolytic detachment of the CD43 extracellular domain (or
a part of it), it was of interest to determine whether L-CanAg is still
a transmembrane CD43 molecule at the late biosynthetic stage when
formation of the Si-Lea epitope occurs. Lysates and media
of COLO 205 cells were therefore immunoprecipitated either with the
-GPEP18 antiserum or the mAb C241 directed against the
Si-Lea epitope. The immunoprecipitates were then subjected
to SDS-PAGE and Western blot followed by probing with the mAb C50 which
also recognizes Si-Lea. As can be seen in Fig.
2,
-GPEP18 precipitated a
Si-Lea-containing band in the L-CanAg size range from COLO
205 cell lysate but not from spent medium. In contrast, L-CanAg and
also the Si-Lea-carrying MUC1 mucin H-CanAg were found in
COLO 205 cell lysates when immunoprecipitated with the
anti-Si-Lea mAb C241. Spent medium immunoprecipitated with
mAb C241 contained both L-CanAg and traces of H-CanAg (barely visible
in Fig. 2); materials of lower apparent Mr than
is usually seen for L-CanAg were also detected. These results show that
CD43 in COLO 205 cells acquires the Si-Lea epitope before
the putative cleavage event that generates secreted L-CanAg. The
results also confirm the absence of the CD43 cytoplasmic domain in
secreted L-CanAg.
A Fraction of the Membrane-bound CD43 in COLO 205 Cells Is Present at the Cell Surface
To investigate whether membrane-bound L-CanAg
only exists as an intracellular intermediate or if it is also exposed
at the plasma membrane, surface-bound carbohydrates on COLO 205 cells were labeled by mild periodate oxidation followed by derivatization of
generated aldehyde groups with biotin hydrazide. Lysates of the labeled
cells were immunoprecipitated with the -GPEP18 antiserum or with mAb
C241. After SDS-PAGE and Western blot, biotinylated immunoprecipitates
were detected with alkaline phosphatase-labeled streptavidin. As shown
in Fig. 3A, immunoprecipitation with
-GPEP18 yielded a readily distinguishable band in the L-CanAg size
range, indicating that at least some of the membrane-bound CD43 in this cell line is present at the plasma membrane. In contrast, the Si-Lea-reactive mAb C241 could not immunoprecipitate any
detectable amount of biotinylated L-CanAg. This latter finding
suggested that the periodate oxidation-biotinylation procedure had
rendered all the cell surface-exposed Si-Lea epitopes
unrecognizable to the C241 mAb. Periodate oxidation of
Si-Lea is known to reduce the binding of C241 (18), an
effect which apparently was enhanced by the biotinylation of oxidized
moieties.
This phenomenon was exploited in an attempt to establish whether or not all the mature CD43 in COLO 205 cells is found on the cell surface. Lysates of biotinylated and non-biotinylated cells were immunoprecipitated with the C241 mAb and detected with C50, another mAb against Si-Lea, after SDS-PAGE and Western blot. As shown in Fig. 3B, L-CanAg (and also H-CanAg) are readily detected in untreated cells as well as in oxidized-biotinylated cells. Taken together, the data in Fig. 3, A and B, indicate that a significant proportion of the CD43 present in COLO 205 cells is inaccessible to inactivation of their Si-Lea epitopes by oxidation-biotinylation and therefore most likely represent a pool of intracellularly stored CD43.
Oxidation-biotinylation was also performed with HL-60 cells. As in COLO
205 cells, immunoprecipitation with -GPEP18 yielded a biotinylated
band of the size expected for glycosylated CD43 (Fig. 3A).
When whole cell lysates of untreated and oxidized-biotinylated HL-60
cells were compared in a Western blot probed with the
glycosylation-dependent anti-CD43 mAb Leu-22, only the
untreated cells showed reactivity (Fig. 3C). As the Leu-22
epitope is known to be sensitive to periodate oxidation, the complete
disappearance of Leu-22-reactive CD43 in oxidized-biotinylated HL-60
cells indicates that the majority of CD43 in this cell line is present
on the plasma membrane, in contrast to the situation in COLO 205 cells.
As CD43 is known to
be a phosphoprotein in leukocytes (19), it was of interest to determine
whether phosphorylated CD43 could be detected also in COLO 205 cells.
COLO 205 and HL-60 cells were therefore metabolically labeled with
[32P]orthophosphate and the lysates were
immunoprecipitated with the -GPEP18 antiserum and, in the case of
COLO 205, also with mAb C241. As shown in Fig.
4A, phosphorylated L-CanAg could be detected
in immunoprecipitations both with
-GPEP18 and mAb C241. The
intensity of the staining was comparable to that obtained in the
control immunoprecipitation of HL-60 cells with
-GPEP18. As a
parallel immunopreciptation of [35S]methionine-labeled
cells with
-GPEP18 indicated that the CD43 synthesis rates were
similar in COLO 205 and HL-60 (Fig. 4B), it can be assumed
that the degree of CD43 phosphorylation in COLO 205 is roughly
comparable to that seen in HL-60.
As the GPEP18 peptide contains the putative main phosphorylation site
(Ser-332) of leukocyte CD43 (19), it could be suspected that
phosphorylation at this site might influence the reactivity of the
-GPEP18 antiserum. To test this hypothesis, lysates of [32P]orthophosphate-labeled COLO 205 and HL-60 cells were
repeatedly immunoprecipitated with
-GPEP18. After the depletion of
-GPEP18-reactive CD43, the remaining CD43 was precipitated using mAb
C241 (for COLO 205 lysates) or mAb Leu-22 (for HL-60 lysates). Fig. 4,
C and D, shows that both in COLO 205 and HL-60
cells, a significant amount of phosphorylated CD43 could be recovered
from the lysates after depletion with
-GPEP18. These results show
that all phosphorylated (and thus transmembrane) CD43 is not recognized
by
-GPEP18.
To characterize the size distribution of cell
surface-bound Si-Lea-carrying glycoproteins, COLO 205 cells
were surface-labeled with NHS-LC-biotin, a reagent that couples biotin
to amino groups of polypeptide chains on the cell surface and thus does
not modify carbohydrate structures like Si-Lea. A lysate of
the labeled cells was separated by gel filtration on a Superose 6 column using 4 M guanidinium chloride as eluant. The
fractions were analyzed with immunofluorometric assays which used the
Si-Lea-reactive mAb C50 as solid phase-bound catching
antibody and either C50 or streptavidin as europium-labeled tracer. The
chromatogram obtained with the homologous C50/Eu-C50 assay (Fig.
5) showed the characteristic biphasic distribution of
Si-Lea-carrying glycoproteins observed previously in COLO
205 (8). The early eluting peak, called the E peak, is known to contain the MUC1 mucin H-CanAg, whereas the late eluting L peak has biochemical properties closely similar to those of the secreted CD43 glycoform L-CanAg isolated from spent COLO 205 medium (8). The curve obtained
with the C50/Eu-streptavidin assay showed a markedly different shape
(Fig. 5) with the main peak appearing at a position between the E and L
peaks.
The low C50/Eu-streptavidin reactivity at the position of the L peak
suggested that the L peak consists of predominantly intracellular (and
thus non-biotinylated) material. This raised the question whether the
main C50/Eu-streptavidin-reactive peak represented the position at
which transmembrane L-CanAg elutes in gel filtration. To adress that
question, fractions from Superose 6 gel filtration of a COLO 205 lysate
were combined into pools E1, E2, and L as shown in Fig.
6A and subjected to immunoprecipitation with
the -GPEP18 antiserum. The immunoprecipitates were then probed on a
Western blot with the anti-Si-Lea mAb C50. As shown in Fig.
6B, only the E1 and E2 pools showed detectable amounts of
-GPEP18-reactive L-CanAg, suggesting that the L pool does not
contain transmembrane L-CanAg.
To further test this conclusion, the distribution of phosphorylated (and thus transmembrane) L-CanAg in gel filtration was also analyzed. A lysate of [32P]orthophosphate-labeled COLO 205 cells was subjected to Superose 6 gel filtration and the E1, E2, and L pools were collected and immunoprecipitated with the Si-Lea-reactive mAb C241. As shown in Fig. 6C, the majority of the phosphorylated L-CanAg was found in the E2 pool with only weak staining in the E1 and L pools. This result strengthens the conclusion that the L peak represents an intracellular supply of soluble L-CanAg that can be separated from its membrane-bound counterpart by means of gel filtration.
L-CanAg in the E and L Pools Have Distinct Pulse-Chase KineticsThe capacity of gel filtration to separate
membrane-bound and soluble L-CanAg was exploited to analyze the
sequence of biosynthetic events leading to L-CanAg secretion.
Si-Lea-carrying glycoproteins from lysates of pulse-chase
labeled COLO 205 cells were subjected to Superose 6 gel filtration and
collected in pools E and L (cf. Fig. 6A; pool
E = pool E1 + pool E2). These pools, as well as the spent media
obtained from the pulse-chase experiment, were subjected to
immunoprecipitation with mAb C241. As can be seen in Fig.
7, the appearance and disappearance of labeled L-CanAg
in the E and L pools followed distinctly different time courses. In the
E pool, the L-CanAg staining was most intense after 1 h with
slight fading after 2 h and significant fading after 5 h. In
contrast, the L-CanAg staining in the L pool increased steadily up to
5 h. After 24 h, most of the staining in both the E and L
pools had disappeared, concomitant with a significant increase in
L-CanAg staining in the spent medium (data not shown). As the
immunoprecipitation was performed using a Si-Lea-reactive
mAb, the kinetics of the Si-Lea-carrying MUC1 mucin H-CanAg
appearing in the E pool could also be followed. In striking contrast to
the kinetics of membrane-bound (i.e. pool E) L-CanAg,
labeled H-CanAg was virtually absent after 1 h, maximal at 5 h, and still showing residual staining after 24 h.
In a recent study (10) it was demonstrated that the colon
adenocarcinoma cell line COLO 205 expresses CD43, a surface molecule normally only seen on cells of the hematopoietic lineage. The CD43
glycoform expressed in COLO 205 cells, L-CanAg, was originally isolated
as a secreted mucin-like molecule carrying Si-Lea, a marker
for carcinomas of the intestinal tract (8). The glycans of L-CanAg are
strikingly different from those of leukocyte CD43: whereas the latter
generally contain only 4-6 sugars per chain (11), the former contain
17 on the average (8). It is therefore no surprise that antibodies that
bind the extracellular domain of leukocyte CD43 fail to recognize
L-CanAg (10).2 To facilitate detection of CD43 in a
glycosylation-independent manner, an antiserum was raised against the
synthetic peptide GPEP18 corresponding to a part of the cytoplasmic
domain of CD43. This antiserum, -GPEP18, was indeed capable of
detecting both leukocyte-type CD43 from the promyelocytic cell line
HL-60 and L-CanAg from COLO 205 cells (Fig. 1).
The data presented in Figs. 1 and 2 also show that -GPEP18 can
immunoprecipitate CD43 molecules from COLO 205 cells that have acquired
a high (150-250 kDa) and polydisperse apparent molecular mass as well
as the Si-Lea epitope, i.e. features typical of
the mature L-CanAg mucin. The secretion of L-CanAg almost certainly
entails the proteolytic detachment of the extracellular region from the
transmembrane and intracellular domains, as confirmed by the inability
of
-GPEP18 to immunoprecipitate any Si-Lea-containing
materials from COLO 205 culture supernatant, and on the basis of the
data from Figs. 1 and 2, it can be assumed that this cleavage occurs at
a late biosynthetic stage when most or all of the glycosylation has
taken place. The demonstration of co-expression of Si-Lea
with CD43 peptide epitopes (Fig. 2) also provides additional evidence
that L-CanAg is indeed a CD43 glycoform.
The post-translational modifications of L-CanAg also include phosphorylation, as demonstrated by the metabolic phosphate labeling experiments (Fig. 4). Leukocytes and cell lines derived from hematopoietic cells also phosphorylate CD43 to varying extents (19, 20). The degree of CD43 phosphorylation in COLO 205 cells seemed roughly comparable to that in HL-60, which in turn phosphorylates CD43 to relatively high extent compared with other leukocyte cell lines and to peripheral blood lymphocytes (19). The significance of CD43 phosphorylation is not known, but it may control proteolytic cleavage of the CD43 extracellular domain as both phosphorylation and proteolysis occur upon activation of leukocytes (21-24). The finding that L-CanAg is phosphorylated as well as proteolytically cleaved at a high rate in COLO 205 cells strengthens the hypothesis that there may be a causal relationship between phosphorylation and proteolysis of CD43. One might also speculate that the extensive phosphorylation and proteolysis of L-CanAg are the results of the activated state of these carcinoma cells, which could mimic the situation in activated leukocytes.
As was shown by the sequential immunoprecipitation experiments,
-GPEP18 failed to bind a significant subpopulation of phosphorylated CD43 in both COLO 205 and HL-60 cells. Phosphorylation itself is
probably the cause of
-GPEP18's inability to bind this
subpopulation, as treatment of HL-60 lysates with alkaline phosphatase
has been observed3 to increase the amount
of CD43 recoverable by
-GPEP18. The heterogeneity in
-GPEP18
reactivity toward phosphorylated CD43 suggests that several different
phosphorylation sites may be used to a significant degree in COLO 205 and HL-60; alternatively, phosphorylation at a single site may cause
the surrounding region to assume a number of alternative conformations,
some of which are unrecognizable to the antiserum.
Although the cell-surface biotinylation experiments clearly demonstrated that membrane-bound L-CanAg is present at the plasma membrane (Fig. 3A), they also indicated that there is a significant amount of L-CanAg that is stored inside the COLO 205 cells, as shown by the presence of a pool of molecules that are protected from the Si-Lea inactivation caused by periodate oxidation and biotinylation (Fig. 3B). This was in contrast to the situation in HL-60 cells, where no detectable amount of the Leu-22 epitope remained after oxidation-biotinylation. This difference may be explained by the respective origins of COLO 205 and HL-60 cells: whereas HL-60 is of the hematopoietic cell lineage, COLO 205 is derived from the intestinal epithelium. One class of intestinal epithelial cells, the goblet cells, are known to be capable of storing mucins in intacellular granules from which they can be released by a process of regulated secretion (1). It is therefore possible that COLO 205 cells treat L-CanAg as if it were an intestinal mucin, not only with respect to glycosylation but also with respect to intracellular trafficking. The almost exclusive localization of CD43 to the surface of the hematopoietic cells has been observed before (17).
The finding that membrane-bound and soluble L-CanAg could be separated from each other by gel filtration (Figs. 5 and 6) is quite surprising, as removal of the intracellular and transmembrane domains of L-CanAg would decrease its size by only around 10%. One might speculate that membrane-bound L-CanAg forms noncovalently bound oligomers which are dissociated upon release of the extracellular domain. As the experiment described in Fig. 6, A and B, gives similar results when conducted in the presence of the denaturing solvent 4 M guanidinium chloride (data not shown), such hypothetical oligomers would have to be held together with considerable strength (although being sensitive to SDS).
By separating Si-Lea-carrying glycoproteins from lysates of pulse-chase labeled COLO 205 cells by gel filtration, the kinetics of membrane-bound and soluble L-CanAg could be studied (Fig. 7). From the finding that membrane-bound L-CanAg appeared and disappeared more rapidly than the soluble form, two conclusions may be drawn. First, it shows that the soluble L-CanAg detected in lysates in this and previous experiments is not an artifact created during lysate preparation (this is also indicated by the difference between membrane-bound and soluble L-CanAg in accessibility to surface biotinylation (Fig. 5)). Second, this result indicates that the relatively rapid synthesis of fully glycosylated, membrane-bound L-CanAg (1-2 h) is followed by a slower process involving the conversion to the soluble form, which is retained inside the cell until more than 5 h after synthesis. It is conceivable that membrane-bound L-CanAg cycles between the surface and the interior of the cell in a manner similar to what has been demonstrated earlier for the MUC1 mucin (15). However, experiments aimed at elucidating this question have been inconclusive (data not shown).
In summary, the results presented here suggest the following pathway for the modification and transport of the CD43 anomalously expressed in COLO 205 cells. The primary translation product, following a process of extensive glycosylation (including synthesis of the Si-Lea epitope), is transported to the cell surface where a proportion of the molecules are phosphorylated. Transmembrane L-CanAg is then converted to a soluble form by proteolysis. As a significant amount of soluble L-CanAg is found inside the cells, this proteolytic step presumably follows upon endocytosis of the cell surface-bound molecule. The soluble form of L-CanAg is then stored inside the cell for some time before being secreted. During this storage additional glycosylation may take place, as L-CanAg from COLO 205 cell extracts and from spent medium differ somewhat in carbohydrate composition (8). The present conclusions thus underscore the complexity of mucin expression in carcinomas and illustrate how a single mucin apoprotein species may assume a variety of different biochemical forms with a rich potential for distinct biological functions.
I thank Dr. Gunnar C. Hansson for support and advice throughout this study and Dr. Sven R. Carlsson for supplying unpublished data and helpful suggestions.
HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
All ASBMB Journals | Molecular and Cellular Proteomics |
Journal of Lipid Research | Biochemistry and Molecular Biology Education |