The cyclin-dependent kinase inhibitor p21WAF1 has been characterized as an important effector of the tumor suppressor p53 and has been linked to various growth-regulatory processes. To identify a potential role of p21 in anchorage-dependent growth control, we analyzed a pair of HCT116 human colon carcinoma cell lines that differed only in their p21 status. We found that during suspension culture, HCT116 cells (which contain wild-type p53 and p21) continued to proliferate and formed compact multicellular spheroids (MCSs). In contrast, HCT116 cells engineered to lack functional p21 (HCTp21−/−)were unable to form MCSs in suspension culture, ceased proliferation,and eventually died through apoptosis. The parental HCT116 cells underwent the same fate when treated with hyaluronidase, indicating that cell-cell contact might be required for survival in suspension culture. We established that E-cadherin was induced in HCT116 but not in HCTp21−/− cells and accounted for the formation of MCSs. Forced expression of E-cadherin or p21 in HCTp21−/− cells restored the ability to form MCSs and to grow independently of anchorage. Moreover,HCTp21−/− cells exhibited a severely reduced transformed phenotype and demonstrated greatly enhanced chemosensitivity in suspension culture. Thus, our results link an important regulator of the cell cycle machinery to the expression of a cell-cell adhesion molecule involved in tumor formation. Because our results indicate that loss of p21 severely impairs the ability of HCT cells to grow independently of anchorage, it may not be coincidental that inactivating mutations of this gene are very rarely found in tumor cells.
Cell cycle checkpoints play crucial roles in maintaining tissue homeostasis, and loss of these control mechanisms may contribute to the development of the malignant phenotype. The CKI3p21WAF1 has been characterized as an important component of these events and was found to be a crucial executioner of p53-induced cell cycle arrest (1, 2, 3). It has been shown in many examples that activation of p53 in response to various external insults that threaten the integrity of the cell leads to the increased expression of p21 (1, 4, 5, 6, 7). Elevated levels of p21, in turn, bind to and inhibit the activity of CDKs and proliferating cell nuclear antigen, both of which are essential for cell cycle progression(8, 9, 10). As a consequence, cellular proliferation is halted, and the cells are arrested in the cell cycle.
To further determine the exact contribution of p21 to cell-regulatory events, cell lines without the p21 gene were established and investigated. In such studies, it was found that p21-negative mouse fibroblasts exhibit a partial deficiency in G1 cell cycle arrest after radiation or nucleotide depletion (11, 12). In human HCT116 colon carcinoma cells engineered to lack both copies of the p21gene, a complete loss of G1 cell cycle arrest became evident, and this was associated with a predisposition to S- and M-phase uncoupling after exposure to DNA-damaging drugs(13). Furthermore, an altered sensitivity of these p21-negative HCT116 cells to certain DNA-damaging chemotherapeutic drugs has been demonstrated both in vitro and in vivo (7, 13, 14, 15, 16). In addition, p21-negative cells have been shown to be defective in nucleotide excision repair, and it has been suggested that this may underlie their increased sensitivity to certain chemotherapeutic drugs (16). However, despite the demonstrated role of p21 as a cell cycle inhibitor, inactivating mutations of this gene are very rarely found in tumor cells (17, 18), and mice lacking the p21 gene do not exhibit an increased rate of spontaneous tumors (11). Therefore, the potential role of p21 in tumor development has remained largely unclear.
The in vitro culture of cells in suspension, where cells are allowed to form three-dimensional MCSs, is believed to more closely mimic in vivo conditions than the culture of cells in two-dimensional monolayers (19). However, except for some hematopoietic cells, normal cells do not proliferate in suspension culture, a phenomenon called anchorage dependence. In contrast, tumor cells are able to continue to grow in suspension culture (or embedded in soft agar), and this anchorage-independent phenotype has been found to closely correlate with their ability to form tumors in animals(20, 21). In general, the growth rate in suspension culture varies greatly between different tumor cell lines(22), which may indicate the function of cell type-specific processes modulating anchorage-independent growth. In this regard, it was shown in one study (23) that intercellular adhesions mediated by the cell surface receptor E-cadherin were required for the survival and anchorage-independent growth of human HSC-3 oral squamous carcinoma cells. In contrast,another study (24) described that transfected E-cadherin resulted in the suppression of anchorage-independent growth of EMT/6 mouse mammary carcinoma cells. Although the underlying basis for this discrepancy is unclear, the detailed analysis of E-cadherin in these processes is of importance because this molecule not only plays a crucial role in normal tissue morphogenesis and organization (25, 26) but presumably also in invasive processes and the metastatic spread of tumor cells (27, 28, 29, 30).
The contribution of CKIs to anchorage-dependent growth control has been recognized as well. After transfer of monolayer cells to suspension culture, the expression of several CKIs, such as p21WAF1, p27KIP1,p16INK4a, or p18INK4c, has been found elevated and implicated in the subsequent growth arrest of anchorage-dependent cells (22, 31, 32, 33, 34, 35, 36). Curiously, a very similar induction of p21 and p27 has also been detected in many anchorage-independent tumor cells (22, 33). It appears that in this latter case, the CKIs are sequestered in the cytoplasm and therefore prevented from exerting their inhibitory effect in the nucleus (33). However, this neutralizing mechanism seems only partially effective because most, if not all, tumor cells in suspension exhibit reduced proliferation rates when compared with monolayer culture (22). In addition, there are striking cell type-specific differences; in a review of 39 different tumor cell lines, only 22 were found to exhibit increased expression of either p16, p21, or p27 after transfer to suspension culture conditions(22).
In an attempt to further explore the role of p21 in anchorage-independent growth, we comparatively analyzed the HCT116 human colon carcinoma cell line and derivatives thereof that were engineered previously (37) to lack both alleles of the p21 gene (HCTp21−/−). Here, we show that HCTp21−/− cells have lost the ability to grow independently of anchorage and die through apoptosis when transferred to suspension culture. This phenotype is attributable to the loss of induction of E-cadherin and insufficient cell-cell adhesion after detachment and can be overcome by transfection with either p21 or E-cadherin expression vectors. Moreover, HCTp21−/− cells exhibit severely reduced capacity for tumor formation in chicken embryos and are unable to form foci in monolayer culture. In addition, the absence of tight cell-cell interactions potentiates the sensitivity of these cells toward chemotherapeutic agents when assessed in suspension culture. Thus, our study links a major regulator of the cell cycle to the expression of a cell-cell adhesion molecule with implications for the tumorigenic potential and chemosensitivity of these cells. It further indicates that p21 is required for anchorage-independent growth, and tumorigenic potential,of HCT116 colon carcinoma cells.
MATERIALS AND METHODS
HCT116 and the corresponding HCTp21−/− cells were kindly provided by B. Vogelstein (Johns Hopkins, Baltimore, MD) and cultured as described previously (37). For suspension cultures, tissue culture dishes or 96-well plates were coated with poly-HEMA (Sigma). For the disruption of cell-matrix interactions, cells grown as a monolayer were trypsinized, dispersed by pipetting, and seeded onto poly-HEMA-coated dishes.
Soft Agar Assays.
Assays were performed as described previously (38) with 104 cells/well in a six-well plate.
Immunoblotting was performed as described previously (35). Monoclonal and polyclonal antibodies against E-cadherin were obtained from Santa Cruz Biotechnology, Zymed, and ICN/Cappel. Antibodies against p21 and polyclonal antibodies against P-cadherin and N-cadherin were purchased from Santa Cruz Biotechnology. Anti-integrin antibodies were a gift from P. C. Brooks (University of Southern California, Los Angeles, CA).
Cell Cycle Analysis.
Cells from monolayer or suspension culture were harvested by trypsinization and treated with 10 mg/ml hyaluronidase for 15 min at 37°C to disrupt cell-cell adhesions. Cells were then fixed in 90%ethanol, treated with 40 μg/ml RNase A, stained with 20 μg/ml propidium iodide, and analyzed by flow cytometry on a FACScan analyzer(Becton Dickinson).
Synthetic phosphorothioate-modified oligonucleotides were provided by the Core Facility of the K. Norris Jr. Comprehensive Cancer Center. The sequences of anti-p21 oligonucleotides AS-MID and AS-IC, as well the control oligonucleotides 18 MER and 21 MER, have been published elsewhere and shown to be effective at inhibiting p21 expression(41). The cytofectin GS2888 (Ref. 42; a gift from W. M. Flanagan, Gilead Sciences, Foster City, CA) was used for transfection. One μm oligonucleotide was incubated with 105 cells/ml 24 h before detachment and again immediately after detachment.
Cytotoxicity and Proliferation Assays.
Cell number and cytotoxicity were determined using an assay kit that is based on the cellular conversion of a tetrazolium salt (MTT) into a blue formazan product that is detected using a 96-well plate reader at 570 nm. The kit was obtained from Boehringer Mannheim and used according to the manufacturer’s instructions.
To measure apoptotic cell death, cells were labeled with FITC, and 104 cells/point were analyzed with the use of FACStar (Becton Dickinson). In addition, cell death was confirmed by preparing cryosections of MCSs and staining of cells in situ. For this purpose, MCSs of cells cultured in suspension for 48 h were collected, embedded in Tissue-Tek (Sakura), and frozen on dry ice. Cryosections were air dried and fixed in 3% formalin. TUNEL reaction (43) was performed using a kit according to manufacturer’s instructions (Boehringer Mannheim). After staining,samples were processed for examination with a Nikon fluorescence microscope using filters for FITC.
Annexin V Assay.
Externalized phosphatidylserine was detected by analyzing the binding of FITC-labeled annexin V (44) with the use of a commercially available kit (Calbiochem). Cells were costained with propidium iodide and analyzed by flow cytometry for necrosis and apoptosis.
E-Cadherin Blocking Antibody.
Cells were seeded at a density of 105 cells/ml in 96-well plates in the presence of mouse anti-human E-cadherin monoclonal antibody (clone SHE78-7; Zymed Laboratories) at 1 μg/ml and incubated for up to 72 h. A mouse anti-p53 monoclonal antibody added to monolayer and suspension cultures at up to 10 μg/ml was used as control.
Chicken Embryo Tumorigenicity Assay.
The assay was performed as reported previously (45) with the following modifications. Tumor cells at a density of 106 cells/40 μl were transferred onto the chorioallantoic membrane of 10-day-old chicken embryos. The eggs were further incubated for 7 days, and the tumors were harvested. Tumor volume (V) was calculated using the formula V = d2D/2, where d is the smallest tumor diameter and D is the largest diameter (in millimeters).
Adhesion assays were performed according to procedures described recently (46) with some modification as follows. HCTp21 or HCTp21−/− cells (4 × 104) were seeded onto 96-well microplates coated with different ECM proteins. The wells had been coated with up to 50 μl of ECM protein solution until evaporation had finished with the following concentrations: 25 μg/ml vitronectin in PBS, 25 μg/ml fibronectin in PBS, 25 μg/ml laminin in PBS, 0.1 mg/ml collagen in 0.2 n acetic acid, and 200μg/ml BSA in serum-free medium. After coating, all wells were rinsed three times with serum-free medium containing 200 μg/ml BSA. The cells were allowed to adhere for 15–30 min (depending on the ECM protein) in a 5% CO2 incubator. Incubation time was stopped as soon as cells started to adhere. The wells were then washed three times with serum-free medium to remove unattached cells. Cell number was determined using the MTT assay as described above.
HCTp21−/− Cells Do Not Form Compact MCSs.
Our experiments were done with the human colon carcinoma cell line HCT116 and two independent isogenic derivatives thereof that lacked both alleles of the p21 gene, called HCTp21−/−. These cell lines were characterized previously and exhibited identical morphology,growth rates, and cell cycle distribution under monolayer cell culture conditions (37). They are near-diploid and harbor wild-type p53. To study their proliferation and morphology under anchorage-independent conditions, we transferred these cells from monolayer to suspension culture. Surprisingly, under these conditions,the cells exhibited obvious morphological differences. HCT116 cells formed tight, densely packed MCSs where single cells could not be discriminated. In contrast, HCTp21−/− cells aggregated only very loosely with fewer, easily discernible cells (Fig. 1).
HCTp21−/− Cease Proliferation in Suspension Culture and Eventually Die through Apoptosis.
When the proliferation of these cells was measured, there was a smaller increase in cell number for the HCTp21−/− cells within the first 48 h of suspension culture, and their proliferation ceased thereafter (Fig. 2,A). FACS analysis indicated that the slower growth of HCTp21−/− cells was not attributable to cell cycle blockage, because the percentage of these cells in S phase was 34% as compared with 20%for the HCT116 cells (Fig. 3 A). This is in agreement with earlier observations that HCTp21−/− cells have a reduced ability to arrest in G1 of the cell cycle (13).
Because previous observations have indicated that under certain conditions HCTp21−/− cells are more prone to undergo apoptosis(13), we investigated this aspect. When cells in suspension culture were stained with the tetrazolium salt MTT, cell death became obvious in cultures of HCTp21−/− cells (not shown). Morphological changes of these cells included shrinkage and rounding,and the condensed, fragmented nuclei were stained intensely by TUNEL(Fig. 3,B). Greatly increased apoptotic cell death of HCTp21−/− cells as compared with HCT116 cells became obvious at 48 h after detachment and increased further over the course of several days. At day 7, more than half of the whole population of HCTp21−/− cells were apoptotic, whereas only 12% of HCT116 cells stained positively (Fig. 3 B). Apoptosis as the main cause of cell death in HCTp21−/− cells was further confirmed with the use of the FITC-labeled annexin V, which binds to phosphatidylserine of apoptotic cells (not shown), as well as by in situ TUNEL staining of cryosections from MCS of each cell type (not shown).
Cell-Cell Adhesion Is Crucial for the Survival of HCT Cells in Suspension Culture.
To investigate whether the pronounced differences in MCS formation were involved in the induction of apoptosis, we used the enzyme hyaluronidase to break and prevent cell-cell interactions(47). When added to suspension cultures, hyaluronidase completely prevented cell-cell aggregation of both HCT116 and HCTp21−/− cells (Fig. 2,B) and caused cell death in either cell type (Fig. 2 A). This was observed at enzyme concentrations that were nontoxic per se, i.e.,that neither interfered with the attachment of these cells to cell culture plates nor their growth rate in monolayer culture (not shown;Ref. 47). Therefore, these results indicated that the ability of HCT116 cells to form compact MCSs allowed them to evade cell death.
MCS Formation of HCT Cells Is Mediated by Induction of E-Cadherin.
Cell-cell interactions by cadherins have been shown recently to modulate survival of tumor cells under anchorage-independent conditions(23). Therefore, studies were undertaken to assess expression and function of cadherins. Using immunoblotting techniques,a striking variance in the expression of E-cadherin in suspension culture cells was found. In HCT116 cells, the amount of E-cadherin protein was strongly increased when cells were transferred from monolayer to suspension culture over a period of 48 h (Fig. 4,A). In comparison, HCTp21−/− cells exhibited a substantially weaker increase in E-cadherin. Two other cadherins,P-cadherin and N-cadherin, were not detected in either cell type (not shown), which is consistent with their preferred expression in different groups of cells. The expression of members of the integrin family (αv-β3 and αv-β5), which are cell surface proteins that determine cell-matrix interactions, was not significantly altered in either cell type (Fig. 4,A). Additionally, the binding of cells before and after suspension culture was quantitated with a functional assay using plates coated with individual ECM proteins. No difference in adhesion to these distinct matrices was observed between the two cell types (Fig. 4 B), which indicates the absence of alterations in these types of cell-matrix interacting proteins. Thus,these results further emphasize the specificity of differential E-cadherin induction.
The observed difference in E-cadherin expression in suspension culture was not attributable to different basal levels of this protein, because HCT116 cells and HCTp21−/− cells grown as monolayer cultures contained comparable amounts of E-cadherin (Fig. 4,C). Importantly, HCTp21+/− cells, which retained one allele of the p21 gene, had similar basal levels of E-cadherin as well and exhibited an induction somewhat weaker than HCT116 cells (Fig. 4 C). In suspension, these cells formed MCSs with slightly less density than the HCT116 cells (not shown).
To determine whether E-cadherin was indeed the crucial determinant of the observed cell-cell interactions, HCT116 cells in suspension were incubated with a monoclonal antihuman E-cadherin antibody (clone SHE78-7). This antibody had been shown before to perturb the function of E-cadherin and prevent cell-cell aggregation of tumor cells(23). Indeed, when added to HCT116 cells, this antibody completely blocked the formation of compact MCSs. The microscopic appearance of these cells was essentially identical to the appearance of nonantibody-treated HCTp21−/− cells as shown in Fig. 1, i.e., there were weakly aggregated spheroids of small size and with individual cells clearly visible (not shown). The presence of this antibody did not affect cell growth or morphology of these cells in monolayer culture. As a control, an unrelated monoclonal antibody was used. In this case, even at up to 10-fold higher concentrations, no effects on spheroid formation were observed. Moreover, the stable transfection of E-cadherin cDNA into HCTp21−/− cells conferred the ability to form dense MCSs with the same morphological appearance as those formed by HCT116 cells (Fig. 6, Lane 4). Taken together, these results indicate that the increased expression of E-cadherin is the major determinant of MCS formation in these cells.
Reduction of p21 Expression in HCT116 Cells by Antisense Oligonucleotides Reduces E-Cadherin Induction.
Because the above results demonstrated a lack of E-cadherin induction in HCTp21−/− cells in suspension, we investigated next whether the manipulation of p21 levels by antisense oligonucleotides would generate a similar effect in the parental HCT116 cells. For this purpose, HCT116 cells were incubated with two different p21-antisense nucleotides that have been shown earlier to be highly specific and effective in inhibiting p21 expression (41). As a control, two oligonucleotides were used that targeted sequences from Mycobacterium tuberculosis. As shown in Fig. 5,A, cells treated with p21-antisense oligonucleotides exhibited decreased E cadherin expression after cellular detachment,whereas control oligonucleotides had no inhibitory effect. Upon microscopic inspection, it became apparent that the cells treated with p21-antisense oligonucleotides were unable to form compact MCSs (not shown), and cell death was observed leading to significant differences in cell number (Fig. 5 B). Thus, the HCT116 cells treated with p21-antisense oligonucleotides displayed a phenotype that was very similar to HCTp21−/− cells.
Reintroduction of p21 into HCTp21−/− Cells Restores E-Cadherin Expression.
Because the above experiments indicated that inhibition of p21 in HCT116 cells could generate a phenotype that was similar to the HCTp21−/− cells with respect to MCS formation and expression of E-cadherin, we next investigated whether the reverse could be achieved as well. Therefore, we introduced p21 cDNA into HCTp21−/− cells and investigated whether this would restore E-cadherin induction and dense MCS formation. HCTp21−/− cells stably transfected with an expression vector containing p21 cDNA were generated, and 10 individual clones were isolated and analyzed. Five of these clones exhibited increased levels of p21 that, in four cases, correlated with increased levels of E-cadherin and the formation of dense MCSs. One representative clone is shown in Fig. 6, Lane 1. The other five clones did not harbor elevated levels of p21 and did not exhibit increased levels of E-cadherin nor formation of dense MCSs. Thus, 80% of p21-transfected clones (four of five) displayed elevated levels of E-cadherin and at the same time had gained the ability to form dense MCSs. Mock-transfected cells exhibited the same phenotype as the respective parental cells (i.e.,no alterations), thus excluding transfection per se as the cause of the observed alterations. Therefore, the above experiments demonstrate a causative link between the expression of p21 and the induction of E-cadherin in suspension cells.
Loss of p21 Potentiates Chemosensitivity in Suspension Culture.
Differences in the sensitivity toward anticancer drugs between HCT116 and HCTp21−/− cells grown as monolayer have been reported previously(13). Moreover, in other tumor cells, it has been indicated that cell-cell adhesion may affect cellular chemoresistance as well (48). Therefore, to assess the impact of loss of p21 on the chemosensitivity of detached cells, we incubated suspension cultures of HCT116 and HCTp21−/− cells with either daunorubicin, a clinically used anticancer drug, or with 3,6-diaziridinyl-1,4-benzoquinone (DZQ), a quinoid model compound commonly used to study the redox-cycling abilities of anticancer quinones (49). The addition of either drug to cells immediately at the onset of suspension culture (before spheroid formation) revealed a 2- and 2.1-fold higher chemosensitivity of HCTp21−/− cells, which was comparable with what we observed in monolayer culture. However, when the drugs were added 24 h after the onset of suspension culture (when spheroids had formed), the difference was 10- and 26.7-fold, respectively (Fig. 7). Therefore, in suspension culture, the differences in chemosensitivity between these two cell types were greatly enhanced, indicating that p21 might contribute to this process via its effects on the expression of E-cadherin and spheroid formation.
HCTp21−/− Cells Exhibit a Substantially Weakened Transformed Phenotype.
The greatly reduced ability of HCTp21−/− cells to form MCSs suggested that these cells might not be able to grow independently of anchorage,and thus, that they may have lost their transformed phenotype. Therefore, three characteristic indicators of transformed cells, i.e., anchorage-independent growth in soft agar, focus formation in monolayer culture, and tumor development in chicken embryos, were analyzed. As shown in Fig. 8, HCT116 cells were strongly positive in all of these three assays. In contrast, HCTp21−/− were negative; they did not form foci, they lacked the ability to grow independently of anchorage, and they formed a >10 times smaller tumor mass in chicken embryos (Fig. 8). Therefore,our results clearly indicate a correlation between the loss of the p21 gene and a greatly reduced tumorigenic potential in these cells.
The main finding of our paper is that the CKI p21 is required to maintain the transformed phenotype of HCT116 colon carcinoma cells. Homozygous deletion of p21 in these cells results in the loss of their ability to form foci in monolayer culture, to grow colonies in soft agar, and to develop tumors on chicken chorioallantoic membranes. Furthermore, it appears that the expression of E-cadherin is intimately involved in theses processes. Because the ability of cells to grow in suspension (anchorage-independent growth) very closely correlates with their tumorigenic potential (20, 21), we have chosen this aspect to further investigate the underlying differences between HCT116 cells (which harbor both copies of the p21 gene) and HCTp21−/− cells (which were engineered to lack both copies of the p21 gene; Ref. 37). Both cell lines were characterized previously and shown to exhibit identical morphology,growth rates, and cell cycle distribution under monolayer culture conditions (37).
We found, however, that upon transfer of these cells to suspension culture conditions, obvious differences became apparent. Microscopically, there was a striking difference in their morphological appearance. Whereas HCT116 cells form compact, dense MCSs, HCTp21−/−cells aggregate only very loosely (Fig. 1). This difference is extended to their proliferative capacity, because HCT116 cells continue to proliferate independently of anchorage (for at least 3 weeks in soft agar), whereas HCTp21−/− cells cease proliferation in suspension and die through apoptosis within 5 days. Thus, our finding was quite surprising, because it suggested “pro-tumorigenic” functions of the cell cycle inhibitor p21. This interpretation seems rather contradictive, because p21 has been characterized as a major executioner of p53-induced cell growth arrest and tumor suppression(1, 40, 50, 51). Interestingly, however, in contrast to p53, which is very frequently found mutated, p21 mutations are extremely rare in human tumors (17, 18). Furthermore,whereas p53-knockout mice develop spontaneous tumors (52, 53), this phenomenon is not observed in p21-knockout mice(11). Additionally, in experimental mouse skin carcinogenesis assays, it was shown that the genetic deletion of p21 enhances the formation of papillomas but not their malignant conversion(54). From these studies, it appears that the contribution of p21 to tumorigenesis is clearly distinct from p53 and may be restricted to early phases of neoplastic development.
Intriguingly, comparable to our results with p21, a “paradoxical tumor inhibitory effect of p53 loss” has been described(55). In this study, p53-knockout mice were crossed with transgenic mice expressing the oncogenes v-Ha-ras,v-fos, or transforming growth factor α targeted to the epidermis. In keeping with the established role of p53 as a tumor suppressor, it was anticipated that the resulting progeny would display further increased hyperplasia and papillomatogenesis in response to various insults such as phorbol ester treatment or wounding. However, contrary to expectation, it was found that these mice, now harboring the respective oncogene in the background of a p53-negative genome, exhibited a block of spontaneous, phorbol ester-promoted, and wound-induced hyperplasia (55). Although the processes underlying these surprising effects are unclear,it was speculated that cell type specificities, differentiation states,or a critical combination of the synergistic neoplastic insults may play a role. In this regard, it is interesting to note that Li-Fraumeni individuals, who are hemizygous with respect to p53 deletions, have a high spectrum of sarcomas, lymphomas, and breast tumors but only few epidermal tumors (56). Alternatively, it was suggested that there may be a requirement for p53 expression during the early stages of skin tumor development, e.g., during the response to phorbol ester-promoted or wound-induced stimuli(55).
In comparison with our results and others discussed above, it could be hypothesized that p21 may be required during later stages of tumor development, at least in certain types of tumors, as exemplified by the HCT116 cell line. Such a requirement of p21 could be attributable to its documented apoptosis-inhibitory function (41, 57, 58, 59, 60, 61),which is indeed observed in the present study; when HCT cells with and without p21 are transferred to suspension culture, HCT116 cells continue to proliferate independently of anchorage, whereas HCTp21−/−cells die through apoptosis (Figs. 2 and 3).
Our studies further demonstrate that the antiapoptotic effects of p21 during suspension culture of HCT116 cells appear to be mediated via the induced expression of the cell surface molecule E-cadherin and closely correlate with the ability of the cells to form dense MCSs. This is indicated by the following observations: (a) expression of E-cadherin is induced in HCT116 but not in HCTp21−/− cells after detachment (Fig. 4,A); (b) in the presence of anti-E-cadherin antibodies, HCT116 cells loose the ability to form dense MCSs and die through apoptosis (not shown); (c)overexpression of transfected E-cadherin in HCTp21−/− cells results in dense MCS formation and anchorage-independent growth (Fig. 6). Together, these results indicate that MCS formation and anchorage-independent proliferation of HCT116 cells is dependent on the expression of E-cadherin. In this regard, our results are in agreement with findings by others (23), who established that in HSC-3 human squamous carcinoma cells, E-cadherin-mediated intercellular adhesions were required for anchorage-independent growth and survival of cells in suspension.
In addition, we show that the induced expression of E-cadherin is dependent on the presence of p21: (a) E-cadherin is induced in HCT116 but not in HCTp21−/− cells after detachment (Fig. 4,A); (b) in HCT116 cells, inhibition of p21 by antisense-p21 oligonucleotides results in the concurrent reduction of E-cadherin expression during suspension culture (Fig. 5,A);(c) transfection of p21 cDNA into HCTp21−/− cells restores elevated expression of E-cadherin (Fig. 6). Taken together, the above observations strongly indicate that p21 regulates anchorage-independent growth and survival via the increased expression of E-cadherin. It should be noted that although HCTp21−/− cells do exhibit reduced capability for DNA repair (16), it is unlikely that the lack of E-cadherin regulation in these cells is attributable to unrecognized mutations in the E-cadherin gene:(a) we have used two different HCTp21−/− cell lines that were selected independently of each other (37);(b) the level of E-cadherin expression is the same in HCT116, HCTp21+/−, and both HCTp21−/− cell lines during monolayer culture (Fig. 4,C), indicating that the basal level regulation is intact; and (c) transfection of HCTp21−/−with p21 cDNA restores the elevated expression of E-cadherin in suspension culture (Fig. 6).
A further finding of our analysis is the greatly enhanced chemosensitivity of HCTp21−/− cells in suspension culture. Differential chemosensitivity of HCT116 and HCTp21−/− cells to anticancer agents have been described before (13). However, those studies used only cells grown as a monolayer. In our present investigation, we confirm that HCTp21−/− cells are significantly more sensitive than HCT116 cells when grown as monolayer;however, when transferred to suspension culture, this difference is greatly augmented, for example, from 2.1- to 26.7-fold for daunorubicin(Fig. 7). Because the ability of suspension cells to form MCSs via E-cadherin has been proposed to condition their sensitivity to chemotherapeutic drugs (48, 62), it is conceivable that the differential expression of E-cadherin in our HCT cell model is responsible for the observed effect. Furthermore, because we show that elevated E-cadherin expression is dependent on the presence of p21, our results further indicate that in suspension culture, p21 might determine cellular chemosensitivity, at least in part, via the expression level of E-cadherin.
The indicated involvement of p21 in the chemosensitivity of suspension culture cells is related to results by others (63),implicating another CKI, p27KIP1, in the resistance of tumor cells to anticancer agents. Similar to our findings with p21, it was established that the down-regulation of p27 expression with antisense oligonucleotides in suspension culture resulted in reduced intercellular adhesion and increased cellular sensitivity to the drug 4-hydroperoxycyclophosphamide.
Overall, however, it appears that the exact contribution of E-cadherin and of p21 and p27 to anchorage-independent cellular growth is strongly dependent on the cell type under investigation. For example, E-cadherin has been characterized as a major growth suppressor of EMT/6 mouse mammary carcinoma cells in suspension culture (24),whereas in other studies it was shown to be required for anchorage-dependent proliferation and survival of HSC-3 human squamous carcinoma cells (23). Similarly, the induction of p21 or p27 that has been observed after transfer of cells from monolayer to suspension culture can be detected in many but not all cell types(22, 33, 35). It is likely that differences in the genetic background contribute to the observed discrepancies between different cell types. Thus, for the future analysis of these processes, it would be helpful to comparatively investigate more cell lines where defined genes of interest have been deleted by targeted disruption.
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.
This work was supported by Research Grant CN-82601 from the American Cancer Society (to A. H. S.) and by Grant 1RO1HL53467 (to E. C.) and Grant R29CA74278 (to A. H. S.) from the NIH. S. M. is a Feodor-Lynen-Fellow of the Alexander-von-Humboldt Foundation.
The abbreviations used are: CKI,cyclin-dependent kinase inhibitor; MCS, multicellular spheroid;poly-HEMA, poly-(2-hydroxyethyl methacrylate); MTT,3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; TUNEL,terminal deoxynucleotidyl transferase-mediated nick end labeling; ECM,extracellular matrix; FACS, fluorescence-activated cell sorter.
We thank B. Vogelstein for the HCT116 cell lines, W. El-Deiry for p21 plasmids, B. M. Gumbiner, M. J. G. Bussemakers, and F. Van Roy for E-cadherin clones, and P. C. Brooks and E. Petitclerc for anti-integrin antibodies and technical assistance with the chicken embryo tumorigenicity assays. The technical assistance of Z. Baharians, S. Campos, and R-C. Wu is acknowledged.