Abstract
Human carcinoembryonic antigen (CEA), a widely used tumor marker, and CEACAM6 [formerly nonspecific cross-reacting antigen (NCA)] are up-regulated in many types of human cancers, whereas family member CEACAM1 [formerly biliary glycoprotein (BGP)] is usually down-regulated. Deregulated overexpression of CEA/CEACAM6 but not CEACAM1 can inhibit the differentiation and disrupt the polarization and tissue architecture of many different types of cells. In this report, we show that CEA and CEACAM6, but not CEACAM1, markedly inhibit the apoptosis of cells when deprived of their anchorage to the extracellular matrix, a process known as anoikis. By blocking this tissue architecture surveillance mechanism, the architectural perturbation initiated by CEA/CEACAM6 can thus be maintained.
Introduction
CEA 3 and highly related CEA family member CEACAM6, members of the immunoglobulin superfamily, show deregulated cell surface overexpression in about 50% of human cancers (1) ; CEACAM1, on the other hand, which differs from CEA/CEACAM6 chiefly by its mode of cell membrane anchorage (transmembrane versus GPI), is usually down-regulated (2) . These CEA family members function, in vitro at least, as intercellular adhesion molecules (3, 4, 5) . A common view that CEA represents a “differentiation marker” that merely reflects the differentiation status of the tumors that express it has been challenged recently by evidence that CEA and CEACAM6, but not CEACAM1, could play an instrumental role in tumorigenesis by the disruption of cell polarity and tissue architecture and the inhibition of cell differentiation of many different types of cells 4 ,(6, 7, 8) . If deregulated CEA/CEACAM6 overexpression in colonocytes at the base of colonic crypts can indeed cause the aberrations in tissue architecture observed in human colonic carcinomas, CEA/CEACAM6 overexpression must also overcome the control mechanisms that normally preserve the tissue architecture of the crypts. We have obtained evidence that the function of specific integrins is perturbed by CEA/CEACAM6 overexpression, 5 thus affecting cell-ECM interactions that are necessary for the establishment of tissue architecture and the correct deployment of differentiation programs. It has been proposed that anoikis, a type of apoptotic program triggered when cells lose contact with the ECM (9, 10, 11) , functions in vivo as a surveillance mechanism that prevents dysplasia and preserves normal tissue architecture by destroying any cells that attempt to deviate from their normally operative spatial constraints (10) . Our model (Fig. 4) ⇓ would require that this process be inhibited, if aberrant multilayered tissue architecture initiated by CEA/CEACAM6 overexpression were to be maintained. We demonstrate here that forced overexpression of CEA/CEACAM6 but not CEACAM1 on the surface of L6 rat myoblasts, MDCK epithelial cells, and human SW1222 and Caco-2 colorectal cancer cells inhibited their anoikis in vitro.
Materials and Methods
Cell Lines.
L6 rat myoblasts, SW1222 and Caco-2 human colorectal carcinoma cells, and MDCK epithelial cells were grown as monolayer cultures in DMEM (L6 and MDCK cells) or α-MEM (SW1222 and Caco-2 cells) containing 10% fetal bovine serum (GM; Life Technologies, Inc., Gaithersburg, MD) supplemented with 100 μg/ml streptomycin and 100 units/ml penicillin (Life Technologies, Inc.) at 37°C in a humidified atmosphere with 5% CO2. All cell lines were subcultured before confluence and seeded at a density of 1 × 104 cells/cm2.
cDNA Transfections and Infections.
Transfection procedures and the transfected cell lines used in this study have been described previously (6 , 7) .4, 5 Briefly, stable transfectants of rat L6 myoblasts were obtained by the calcium phosphate precipitation method using p91023B expression vector (courtesy of R. Kaufman; Genetics Institute, Boston, MA) containing full-length cDNAs encoding cell adhesion proteins CEA, CEACAM6, CEACAM1-4L (formerly splice variant BGPa), CEA deletion mutant ΔNCEA [lacking the last 75 amino acids of the NH2-terminal domain (6 , 12)] , NCAM-125 [GPI-linked NCAM splice variant with a muscle-specific domain (13)] and pSV2neo plasmid as a dominant selectable marker. SW1222 and Caco-2 transfectants were obtained with Zn2+-inducible episomal expression vector pML1 containing the hygromycin gene plus full-length cDNAs encoding either CEACAM1, CEA, or ΔNCEA or a cDNA containing the entire coding region of CEACAM6.4 SW1222-Hygro and Caco-2-Hygro (vector alone) controls were obtained with pML1 containing the hygromycin gene. Pooled clones of stably transfected cells were selected with 400 μg/ml neomycin (G418; L6 cells) or 200μ g/ml hygromycin B (SW1222 and Caco-2 cells). L6 myoblasts expressing comparably high levels (FACS mean value, 150–225)5 of the proteins encoded by the transfected cDNAs were selected by FACS using specific mAbs. After promoter induction with Zn2+, the SW1222-CEACAM6 transfectant and the Caco-2-CEA/CEACAM6 double transfectant cells expressed about 9- and 20-fold higher cell surface levels of CEACAM6 or CEA/CEACAM6, respectively, than control cells transfected with the vector alone.4 Although G418 (L6) and hygromycin B (SW1222 and Caco-2) were removed from the culture media during functional assays, no loss of cell surface expression of the transfected cDNAs was observed (data not shown).
L6 and MDCK cells in the exponential phase of growth were infected with a replication-defective recombinant retrovirus containing either pBabe(human bcl-2)puro [L6 (7)] or pLXSN(CEA cDNA)neo (MDCK) or the vector alone as a control. Stably transfected MDCK cells were selected with 400 μg/ml G418, and cells expressing high levels of CEA on the cell surface were selected by FACS using specific anti-CEA mAbs.
To avoid phenotypic perturbations due to clonal variation, all transfected cell lines used in this study were prepared as pooled (total) populations consisting of multiple clones selected with G418 or hygromycin. The polyclonality of the CEA-expressing L6 cell population was demonstrated by Southern blot (7) .
Apoptotic Assays.
To induce overexpression of CEA/CEACAM6, SW1222 and Caco-2 transfected cells were cultured in GM supplemented with 0.1 mm ZnSO4 for 24 h before the experiment. To measure the anoikis of L6, MDCK, and Zn2+-induced SW1222 and Caco-2 cells (control and transfected cells), 0.2 × 106 cells/ml of each cell line were suspended in PolyHEMA (Aldrich Chemicals, Milwaukee, WI)-coated 6-well tissue culture plates for a period of 12–72 h in the presence (GM) or absence (serum-free DMEM) of growth factors. The percentage of apoptotic cells was estimated by staining the nuclei with DAPI (Boehringer Mannheim, Roche Diagnostics, Laval, Quebec, Canada) or by using the TUNEL assay (Oncor; S7100-KIT, Intergen, Boston, MA) following the instructions from the manufacturer. Briefly, to stain the cells with DAPI, the cells were fixed on ProbeOnPlus microscope slides (Fisher Biotech) with 4% paraformaldehyde for 20 min, washed, permeabilized for 5 min with 0.1% Triton X-100 (T-8787; Sigma, St. Louis, MO), and stained with 10 μg/ml DAPI in PBS. Cells with fragmented (DAPI) or stained (TUNEL) nuclei were scored as apoptotic cells. The apoptotic index was calculated by scoring no less than 1000 cells. All observations were reproduced at least twice by independent experiments.
All cell lines used in the anoikis assays formed cellular aggregates in suspension over PolyHEMA-treated surfaces. The average size of these aggregates was determined by dividing the total number of cells per sample by the total number of aggregates, which was determined in triplicate. To determine the total number of cells per sample, the aggregates were dissociated at 37°C for 30 min with a collagenase enzyme mixture that contained 690 units crude collagenase/ml (code CLS-1; Worthington Biochemical Corp., Freehold, NJ) in PBS. The cell concentration in the resulting single cell suspension was measured using a particle counter (Coulter Electronics Inc., Hialeah, FL). The average aggregate size of the parental L6 cells was 30% less than that of the transfectant populations, which had average aggregate sizes that were within 5% of each other.
Statistical Analysis.
Differences between groups (CEA/CEACAM6expressing cells versus control cells) were analyzed by using Student’s t test. P < 0.05 was considered significant.
Results
Effects of CEA and CEACAM6 Expression on Anoikis of Rat L6 Myoblasts.
To test whether CEA and CEACAM6 overexpression on the cell surface inhibits anoikis in vitro, we used pooled stably transfected L6 rat myoblast clones as an ectopic model. These transfectants, together with appropriate controls, were cultured in suspension in serum-free DMEM or GM in tissue culture dishes coated with PolyHEMA to prevent cell attachment to the substratum. Under these conditions, parental L6 myoblasts underwent anoikis due to the lack of survival signals provided by the ECM or substratum, showing characteristic fragmentation of their nuclei (Fig. 1A) ⇓ and positive staining by the TUNEL assay (Fig. 2B) ⇓ . L6 myoblasts expressing CEA or CEACAM6 on their surfaces, on the other hand, were much less prone to undergo anoikis in GM (P = 0.001 and P = 0.004, respectively; Fig. 1, E–G ⇓ ) or serum-free DMEM (P = 0.003 and P < 0.001, respectively; Fig. 2, D and E ⇓ ); control L6 myoblasts expressing CEACAM1 (Figs. 1B ⇓ and 2, C and E ⇓ ) or cells transfected with the vector alone (data not shown) showed about the same level of anoikis as the parental cells.
A−F, DAPI staining of L6 parental and transfected myoblasts suspended in GM for 48 h on PolyHEMA-coated surfaces. Apoptotic cells showed characteristic fragmented nuclei, whereas survivors showed an intact nuclear morphology. A, L6 parental cells. L6 myoblasts expressing human CEACAM1 (B), human Bcl-2 protein (C), the human GPI-linked isoform of NCAM (NCAM-125) (D), CEA (E), and CEACAM6 (F). G, the apoptotic index (see the graph) was calculated by scoring the percentage of apoptotic cells in three independent samples of 500–1000 cells each. The statistical average and SD of three independent experiments are indicated. Both CEA and CEACAM6 expression on the surface of L6 myoblasts significantly inhibited anoikis (P = 0.005 and P = 0.004, respectively). A second independent pooled population of stably transfected L6 myoblasts (L6-CEA-2) showed results similar (P = 0.001) to those seen in L6-CEA transfected cells. These results were confirmed using the TUNEL assay in two separate experiments (data not shown).
Apoptotic cells detected by positive staining using the TUNEL assay. A, control attached L6 parental myoblasts (L6-att) were cultured attached to the tissue culture plastic surface and did not undergo apoptosis. B−D, parental and transfected L6 myoblasts suspended in serum-free DMEM on PolyHEMA-coated surfaces for 24 h (B, L6 parental myoblasts; C, CEACAM1-expressing L6 myoblasts; D, CEACAM6-expressing L6 myoblasts). E, the apoptotic index was calculated as described in the Fig. 1 ⇓ legend. These results were corroborated by DAPI staining in four independent experiments. F, quantitation of anoikis of transfected MDCK cells suspended in GM on PolyHEMA-coated surfaces for 24 h. The average and SDs of two independent experiments are presented.
To test for nonspecific effects due to intercellular adhesion, L6 transfected myoblasts expressing high levels of the NCAM isoform NCAM-125, another GPI-anchored intercellular adhesion molecule of the immunoglobulin superfamily, were used as a control. NCAM-125, despite being expressed at higher levels than CEA [FACS mean value, 225 (NCAM-125) versus 190 (CEA)],5 did not inhibit anoikis (Fig. 1, D and G ⇓ ; Fig. 2E ⇓ ). This result indicates that the inhibitory effects of CEA and CEACAM6 are not due to indirect effects of their intercellular adhesion function or GPI anchorage per se. As a further test of specificity, L6 myoblasts expressing ΔNCEA, a deletion mutant lacking the last two-thirds of the NH2-terminal domain (6 , 12) , were tested for anoikis. These transfectants were as prone to undergo anoikis as L6 parental cells (Fig. 2E) ⇓ . The deleted portion of the NH2-terminal domain has been shown to be required for both intercellular adhesion (12) and the differentiation inhibitory effects of CEA (6) .
Parental L6 myoblasts suspended in the absence of growth factors (serum-free DMEM) showed apoptotic features as early as 12 h (data not shown). However, in the presence of growth factors (GM), apoptosis of suspended L6 myoblasts was only 5% after 24 h (data not shown) and increased to 40% after 48 h (Fig. 1G) ⇓ . As mentioned above, CEA and CEACAM6 expression inhibited anoikis of L6 myoblasts in both the presence and absence of growth factors. Suspended L6 myoblasts expressing human Bcl-2, a known inhibitor of anoikis (9 , 10) , were less prone to undergo cell death (P < 0.001) than L6 parental myoblasts (Fig. 1, C and G ⇓ ; Fig. 2E ⇓ ). Interestingly, Bcl-2-mediated inhibition of anoikis was more prominent in the presence of growth factors (Fig. 1G) ⇓ than in the absence of growth factors (Fig. 2E) ⇓ . The ectopic expression of CEA on the surface of MDCK cells inhibited anoikis (P = 0.001) when these cells were suspended in PolyHEMA-coated dishes (Fig. 2F) ⇓ , thus confirming the effect.
Effect of CEA and CEACAM6 Overexpression on Anoikis of Human Colonocytes.
CEA and CEACAM6 are not endogenously expressed by either skeletal muscle cells or MDCK cells. To test the inhibitory effects of CEA/CEACAM6 in a medically relevant model system, anoikis of stably transfected human colorectal carcinoma SW1222 and Caco-2 cell lines, both of which are capable of colonic epithelial differentiation4 (14 , 15) , was measured. Parental SW1222 and Caco-2 cells express relatively low levels of endogenous CEA and CEACAM6 (16) . After promoter induction with Zn2+, the SW1222-CEACAM6 transfectant and the Caco-2-CEA/CEACAM6 double transfectant cells from subconfluent growing cultures expressed about 9- and 20-fold higher cell surface levels of CEACAM6 and CEA/CEACAM6, respectively, than control cells transfected with the vector alone.4 These levels of CEA/CEACAM6 expression correspond to those observed in vivo in human colonic tumors (17) . SW1222 cells overexpressing CEACAM6 were found to be unable to form glandular-like spheroids of polarized cells in collagen gels;4 Caco-2 cells overexpressing CEA/CEACAM6 lost their ability to form monolayers of polarized cells and instead formed stratified layers of disorganized cells, closely resembling dysplastic colorectal carcinomas.4
CEACAM6-overexpressing SW1222 cells and CEA/CEACAM6-overexpressing Caco2 cells were 2- and 4-fold less prone to undergo anoikis, respectively, than transfectants expressing CEACAM1 or ΔNCEA or cells transfected with the vector alone (P < 0.01; Fig. 3 ⇓ ). CEACAM6 alone also inhibited anoikis (P = 0.01) of transfected Caco-2 cells (Fig. 3F) ⇓ . Transfected Caco-2 cells overexpressing only CEA could not be tested because of a pronounced tendency to lose cell surface CEA expression during culture (data not shown). These results show that deregulated overexpression of CEA and CEACAM6, but not CEACAM1, can not only disrupt cellular polarization and tissue architecture of human epithelial colonocytes4 but can also inhibit their architectural quality control mechanism, anoikis.
A−D, DAPI staining of SW1222 parental and transfected cells. A, control SW1222 parental cells attached to uncoated substratum. SW1222 parental cells (B), SW1222-CEACAM6 cells (C), and SW1222-Hygro (vector alone) (D) control cells cultured on PolyHEMA-coated dishes for 24 h in GM. E, quantitation of anoikis of transfected SW1222 cells (24 h). F, quantitation of anoikis of Caco-2 human colorectal cancer cells suspended in GM on PolyHEMA-coated dishes for 48 h (left) and 72 h (right). The data represent the average of two typical experiments of six independent experiments with similar results.
Discussion
The observed inhibition of a quality control mechanism preserving tissue architecture by CEA/CEACAM6 cell surface overexpression has important implications for the role of these molecules in human cancer, especially in light of the fact that so many cancers show overexpression of these CEA family members. The question of mechanism arises. L6, MDCK, SW1222, and Caco-2 cells formed aggregates when cultured in suspension on PolyHEMA-coated surfaces (data not shown). Apoptotic cells were detected in these cellular aggregates, but most of the apoptotic cells detached from them. In contrast, most of the cells that survived anoikis were found in the aggregates, suggesting that cell-cell interactions could provide a survival signal. This phenomenon has been observed before in 3T3 cells that become resistant to anoikis in the presence of insulin-like growth factor I (18) . This raises the question of whether the intercellular adhesion activity of molecules such as CEA and CEACAM6 could indirectly induce resistance to anoikis as a consequence of increasing the number of cells in the aggregates. This is not the case, however, because CEACAM1 and NCAM-125 are also intercellular adhesion molecules, and cells expressing them showed the same number and size of cellular aggregates as CEA/CEACAM6-expressing cells (see “Materials and Methods”) but no inhibition of anoikis. We suggest that intercellular interactions could nevertheless play a role in inhibiting anoikis by fostering clustering of CEA/CEACAM6 molecules by both antiparallel and parallel binding on the cell surface, thus amplifying a CEA/CEACAM6-mediated antiapoptotic signal. In support of this suggestion, the deletion mutant ΔNCEA, which is defective in mediating both intercellular adhesion (12) and the myogenic differentiation block (6) , is also defective in inhibiting anoikis.
CEACAM1 overexpression in L6 and Caco-2 transfected cells did not inhibit anoikis. These results are consonant with the observation that CEA and CEACAM6 cell surface expression is up-regulated in many cancers, whereas CEACAM1 expression is down-regulated (see“ Introduction”). L6 myoblasts expressing GPI-anchored NCAM-125 also underwent anoikis as readily as parental and CEACAM1expressing L6 myoblasts. This result suggests that the CEA/CEACAM6 inhibitory function is specific and is not an effect of GPI anchorage per se. CEA/CEACAM6 inhibition of anoikis seems to be specific for this type of apoptosis because CEA expression on the surface of L6 myoblasts does not inhibit v-myc-induced apoptosis (7) .
The α5β1 integrin receptor, when bound to fibronectin, triggers a survival signal in many tissue culture cell lines (10) . In the absence of theα 5β1 integrin/fibronectin interaction, many cell types undergo anoikis. We have recently demonstrated that CEA/CEACAM6 expression on the surface of L6 myoblasts modifies cell adhesion to fibronectin due to a change in the functional status of theα 5β1 integrin receptor.5 The antiapoptotic intracellular signals generated by CEA/CEACAM6-mediated activation of theα 5β1 integrin receptor are currently under investigation.
We hypothesize that the CEA/CEACAM6-mediated inhibition of anoikis could contribute to the disruption of normal tissue architecture that accompanies malignant transformation (Fig. 4) ⇓ . We propose that the inhibitory effect of CEA/CEACAM6 on anoikis occurs only when these glycoproteins are overexpressed over the entire colonocyte surface (Fig. 4, D and E) ⇓ , as seen in dysplastic cells, but not when CEA is restricted to the apical surface of normal colonocytes (Fig. 4A) ⇓ , where interference with integrin (localized on the basal membrane) functions is unlikely. CEA/CEACAM6-mediated inhibition of anoikis could allow malignant colonocytes to survive out of the plane of the monolayer in the absence of cell-basement membrane adhesion, a condition that would cause the death of normal colonocytes (19) , thus permitting the persistence of aberrant tissue architecture (Fig. 4F) ⇓ . This disruption of colonic tissue architecture would also inhibit cellular differentiation, further contributing to malignant progression. Considering the fact that CEA and/or CEACAM6 overexpression are observed at other major sites, such as the breast and lung, the inhibitory effects of these molecules on anoikis could be of major importance for the understanding of human cancer.
Tissue architecture model. A−C, anoikis functions as a surveillance mechanism preserving the normal architecture of human colonic crypts. D−G, CEA/CEACAM6-mediated inhibition of anoikis disrupts tissue architecture. A, CEA (denoted as thick dark lines) and CEACAM6 are apically expressed in the top third of normal human colonic crypts (17) . To simplify the model, only CEA is indicated. B, in normal crypts, colonocytes that have lost contact with the basement membrane (ECM) undergo anoikis and are unable to survive out of the epithelial monolayer plane, thus (C) tissue architecture is preserved. D, unlike the expression pattern in normal colonic crypts, CEA and CEACAM6 are overexpressed over the entire surface of malignant colonocytes (circles). E, a CEA/CEACAM6-overexpressing cell that has lost cell-ECM adhesion has a longer survival capacity and proliferates (F) out of the plane of the single columnar epithelium, disrupting normal tissue architecture and inhibiting cellular differentiation. The latter event, together with other genetic lesions, will contribute to tumor formation and progression (G).
Footnotes
-
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
-
↵1 Supported by grants from the National Cancer Institute of Canada and the Medical Research Council of Canada. C. O. was supported by a Studentship from the Cancer Research Society of Canada.
-
↵2 To whom requests for reprints should be addressed, at McGill Cancer Centre, 3655 Drummond Street, Montreal, Quebec, H3G1Y6 Canada. Phone: (514) 398-3535; Fax: (514) 398-6769; E-mail: stanners{at}med.mcgill.ca
-
↵3 The abbreviations used are: CEA, carcinoembryonic antigen; GM, growth medium, ECM, extracellular matrix; TUNEL, terminal deoxynucleotidyl transferase-mediated nick end labeling; DAPI, 4′,6-diamidino-2-phenylindole; MDCK, Madin-Darby canine kidney; GPI, glycosylphosphatidyl inositol; FACS, fluorescence-activated cell sorting; NCAM, neural cell adhesion molecule; PolyHEMA, poly(2-hydroxyethylmethacrylate); mAb, monoclonal antibody.
-
↵4 C. Ilantzis, L. DeMarte, R. A. Screaton, and C. P. Stanners. Deregulated expression of the human tumor marker CEA, and CEA family member NCA disrupts tissue architecture and blocks colonocyte differentiation, submitted for publication.
-
↵5 C. Ordoñez, R. A. Screaton, C. Ilantzis, M. Fan, and C. P. Stanners. Carcinoembryonic antigen, a human tumor marker, inhibits cell differentiation and apoptosis by perturbing the function of the α5β1 integrin receptor, submitted for publication.
- Received October 22, 1999.
- Accepted May 12, 2000.
- ©2000 American Association for Cancer Research.
References
Volume 60, Issue 13
