Aberrant β-catenin-TCF target gene activation plays a key role in colorectal cancer, both in the initiation stage and during invasion and metastasis. We identified the neuronal cell adhesion molecule L1, as a target gene of β-catenin-TCF signaling in colorectal cancer cells. L1 expression was high in sparse cultures and coregulated with ADAM10, a metalloprotease involved in cleaving and shedding L1's extracellular domain. L1 expression conferred increased cell motility, growth in low serum, transformation and tumorigenesis, whereas its suppression in colon cancer cells decreased motility. L1 was exclusively localized in the invasive front of human colorectal tumors together with ADAM10. The transmembrane localization and shedding of L1 by metalloproteases could be useful for detection and as target for colon cancer therapy.
The development of human cancer is considered a multistage process involving genetic changes endowing the tumor cells first, with proliferative advantage and, at later stages, with the capacity to breakdown cell–cell adhesions, and with motile capacity, enabling the cancer cells to invade neighboring tissues. In colorectal cancer, an early event is the aberrant activation of β-catenin-TCF signaling which results from mutations in β-catenin, or its degradation machinery, thereby leading to the accumulation of β-catenin in the nucleus where it forms transcriptionally active complexes with LEF/TCF factors (Bienz and Clevers 2000; Polakis, 2000; Conacci-Sorrell et al., 2002a). Hyperactivation of growth promoting target genes of β-catenin-TCF signaling, including Cyclin D1 (Shtutman et al., 1999; Tetsu and McCormick, 1999) and c-myc (He et al., 1998) may promote the onset of oncogenesis.
Inappropriate activation of β-catenin signaling also contributes to later stages of tumorigenesis by inducing genes that confer invasive and metastatic capacities. These include metalloproteases (Brabletz et al., 1999; Crawford et al., 1999; Takahashi et al., 2002; Hlubek et al., 2004), ECM components (Gradl et al., 1999; Hlubek et al., 2001) and cell adhesion receptors such as CD44 (Wielenga et al., 1999), Nr-CAM (Conacci-Sorrell et al., 2002b), and uPAR (Mann et al., 1999). In addition to its role as cotranscriptional activator, β-catenin is a major component of adherens junctions linking cadherin family transmembrane receptors to the actin cytoskeleton (Ben-Ze'ev and Geiger, 1998). By playing a dual role in cell adhesion and transcriptional regulation, β-catenin can integrate changes in these two key cellular processes that are disrupted in cancer cells.
Recent studies of human colorectal cancer tissue support this view by demonstrating reversible changes in E-cadherin and β-catenin localization during metastasis. Cells at the central tumor mass in the primary tumor display polarized epithelial organization, and form tubular structures with junctional localization of β-catenin and E-cadherin, whereas cells at the invasive front are characterized by loss of cell-surface E-cadherin and nuclear localization of β-catenin (Brabletz et al., 2001). Interestingly, in lymph node metastases, these two phenotypes of cellular organization seen in the primary tumor are regained, suggesting plasticity in colon cancer cellular morphogenesis during metastasis (Barker and Clevers, 2001; Brabletz et al., 2001). Using a model system of colon cancer cells grown at varying densities in vitro, we could mimic these two different forms of cellular organization and identified signaling pathways linking the negative regulation of E-cadherin expression with nuclear signaling by β-catenin (Conacci-Sorrell et al., 2003). Sparse cultures of colon cancer cells express only small amounts of E-cadherin and high levels of nuclear β-catenin, reminiscent of colon cancer cells at the invasive tumor front, whereas dense cultures have well-developed E-cadherin and β-catenin–containing adherens junctions with little nuclear β-catenin resembling the central, more differentiated part of colorectal tumors (Brabletz et al., 2001; Conacci-Sorrell et al., 2003).
In the present study, we used this cell culture system to characterize a novel target gene of β-catenin signaling involved in human colon cancer invasion, and determined the localization of its gene product in colon cancer tissue. We identified L1, a transmembrane cell adhesion molecule, normally expressed in nerve cells, as a target of β-catenin-TCF signaling, and showed that its expression confers cell motility, invasion and tumorigenesis in fibroblasts and colon cancer cells. In colorectal cancer tissue, L1 was exclusively localized at the invasive front of the tumor tissue that expresses nuclear β-catenin, together with the metalloprotease ADAM10 that is involved in the cleavage and shedding of the L1 extracellular domain.
Cell density–dependent expression of L1 and ADAM10 in colon cancer cells
We used the density-dependent phenotypic conversion displayed by two human colon cancer cells lines, SW480 and HCT116, that mimic the changes in E-cadherin and β-catenin localization, and in E-cadherin expression, displayed by colon cancer cells in the differentiated part and the invasive front of colorectal tumors (Brabletz et al., 2001; Conacci-Sorrell et al., 2003). SW480 cells have mutant APC, and wt β-catenin accumulates in the nuclei of these cells in sparse cell cultures (Fig. 1 C). In HCT116 cells, a stabilizing mutation in the NH2 terminus of β-catenin (on Ser45) is responsible for its accumulation and increased nuclear signaling (Morin et al., 1997). When looking for novel β-catenin target genes that correlate with the invasive phenotype of human colon cancer cells, we reasoned that L1 might be a good candidate, as it is among genes whose expression decreased in a DNA microarray analysis of SW480 cells transduced with wt APC (Lin et al., 2001). In addition, we recently identified another member of the L1 family, Nr-CAM, as a target gene of β-catenin, and showed that its expression is induced in human melanoma and colon carcinoma (Conacci-Sorrell et al., 2002b).
We compared the expression of L1 in sparse and dense cell cultures of both SW480 and HCT116 cells and found that sparse cells express higher levels of L1 than dense cells (Fig. 1 A), whereas E-cadherin expression behaved in the opposite manner. The higher level of L1 in sparse cultures resulted from increased L1 RNA in these cells (Fig. 1 B). Sparse SW480 cells displayed strong cytoplasmic L1 distribution, often in large vesicles (Fig. 1 C, arrows), whereas in dense cultures, a weaker staining of L1 was found in the membrane at cell–cell contact sites, consistent with the capacity of L1 to mediate homophilic cell–cell interactions. However, this localization of L1 in dense cells did not coincide significantly with adherens junctions visualized by staining for β-catenin (Fig. 1 C, dense, merge). The presence of L1 in vesicles and its reported shedding from the surface of some tumor cells mediated by the metalloprotease ADAM10 (Mechtersheimer et al., 2001), prompted us to examine whether L1 can be found in the culture medium of sparse and dense colon cancer cells. When the medium was analyzed from cells expressing equivalent amounts of cell-associated L1 (Fig. 1 D, lanes 3 and 4), we found much more L1 shedded into the medium of sparse than dense cell cultures (Fig. 1 D, lanes 1 and 2). Moreover, in both SW480 and HCT116 cells more ADAM10 was found in sparse cell cultures (Fig. 1 E), and correlated with the higher levels of ADAM10 RNA in such cells (Fig. 1 F). These results indicate that L1 might be a novel target gene of β-catenin signaling, coexpressed with ADAM10 in sparse cell cultures that mimic the invasive phenotype of human colon carcinoma.
The L1 promoter is activated by β-catenin-TCF signaling
We wished to determine whether the L1 gene is a target for activation by β-catenin-TCF signaling. When an L1 promoter reporter plasmid containing 2.9-kb upstream of the transcription initiation site (Fig. 2 C; Kallunki et al., 1997; Meech et al., 1999) was cotransfected with a point mutant (S33Y) stabilized β-catenin into 293 cells, it was activated sixfold (Fig. 2 A), similar to the Cyclin D1 promoter reporter that served as control for a previously characterized β-catenin target gene (Shtutman et al., 1999). Cotransfection of the cytoplasmic tail of cadherin, that binds to and sequesters β-catenin from binding to LEF/TCF factors (Sadot et al., 1998), inhibited L1 transactivation. The involvement of LEF/TCF factors in L1 transactivation is suggested by the inhibition of L1 promoter activation using a dominant negative LEF-1 (ΔNLEF), whereas a dominant positive LEF-1 construct containing the DNA-binding domain of LEF-1 linked to the transactivation domain of viral VP16, activated the L1 promoter (Fig. 2 A, LEF1/VP16). Transactivation of the L1 promoter by endogenous β-catenin-TCF complexes in SW480 and HCT116 colon cancer cells was inhibited by cotransfection of components involved in β-catenin degradation (APC and Axin), and also by the cytoplasmic cadherin tail (Fig. 2 B). Both, the Cyclin D1 promoter and the synthetic TOPFLASH reporter plasmid, containing several LEF/TCF-binding sites, displayed similar responses in these experiments (Fig. 2 A and B). We identified 4 putative TCF-binding sites in the 2.9-kb of the L1 promoter (Fig. 2 C). In DNA electrophoretic mobility shift analyses, all 4 putative TCF-binding sites bound to protein complexes containing β-catenin-TCF derived from nuclear extracts of SW480 cells (the data for radiolabeled site no. 4 are shown in Fig. 2 D). When radiolabeled site no. 4 was incubated with nuclear extracts from SW480 cells and with antibody to β-catenin, a supershift band was detected (Fig. 2 D, lane 8), indicating the presence of β-catenin in this DNA–protein complex. This binding was inhibited by incubation with increasing concentrations of the TCF-binding site of the Cyclin D1 promoter (Fig. 2 D, compare lanes 3–5 with lane 2).
Next, we examined whether endogenous β-catenin-TCF signaling activates the endogenous L1 gene of colon cancer cells, using HCT116 cells expressing doxycycline inducible siRNA to β-catenin and dominant negative TCF4 (ΔNTCF; van de Wetering et al., 2003). The efficiency of this strategy in suppressing β-catenin-TCF signaling was demonstrated by reduced TOPFLASH activity 2 d after doxycyclin-mediated induction of ΔNTCF or β-catenin siRNA (Fig. 2 E). The siRNA to β-catenin also reduced the level of β-catenin (Fig. 2 F, compare lanes 7 and 8 with lanes 5 and 6). Analysis of L1 expression in such HCT116 cells, where β-catenin levels were suppressed, revealed a major decrease in L1 (Fig. 2 F, lanes 7 and 8). Similarly, when TCF-mediated signaling was inhibited by inducible ΔNTCF4, L1 expression was also diminished (Fig. 2 F, compare lanes 3 and 4 with lanes 1 and 2). Together, these data argue that β-catenin-TCF signaling can induce the L1 gene of human colon cancer cells.
Expression of L1 in fibroblasts confers enhanced motility, growth in low serum, transformation, and tumorigenicity in nude mice
To understand the possible role(s) of L1 in development of the tumorigenic phenotype, we considered the known functions of L1 in nerve cells, where it is involved in neuronal migration, axonal growth, and guidance (Kamiguchi et al., 1998). To examine if L1 expression can affect the motility of other cells, we infected NIH3T3 cells with a retrovirus expressing L1, and selected cells resistant to a cotransfected puromycin-resistance gene (Fig. 3 B, puro). As shown in Fig. 3 A, NIH3T3 cells transduced with L1 displayed L1 in the cytoplasm (when grown at low density), mostly in vesicles, and in long cellular processes (Fig. 3 A, arrows). This organization was reminiscent of the vesicles seen in SW480 cells (Fig. 1 C). At higher cell density, L1 localized at cell–cell contacts (Fig. 3 A, dense), as seen in dense cultures of colon cancer cells (Fig. 1 C). Most significantly, L1 expressing NIH3T3 cells formed foci when grown on plastic, whereas puror cells did not (Fig. 3 C). We examined whether L1 expression influences the motility of NIH3T3 cells by introducing a gap in a confluent monolayer of cells, and followed the rate at which control and L1 expressing cells filled the gap. L1 expressing 3T3 cells moved into the denuded area much faster than puror control cells (Fig. 3 D). Because NIH3T3 cells expressing L1 formed foci (Fig. 3 C), indicating that L1 can confer cell transformation, we examined if these cells are tumorigenic in nude mice. We found that all six mice injected with L1 expressing cells formed tumors, whereas only one mouse developed a small tumor when injected with the puror control cells (Fig. 3 E, arrow). When higher numbers of cells were injected into mice, puror-expressing cells also formed tumors (Fig. 3 F), a property inherent to transduction with retroviral constructs. However, tumors formed by puror cells were at least three times smaller than those formed by L1-expressing cells (Fig. 3 F). Analysis of RNA from tumor tissue revealed that L1 is expressed in the tumors obtained with NIH3T3 cells expressing L1 (Fig. 3 G).
We examined the growth properties of NIH3T3 cells expressing L1 and found that in 10% of serum there were no differences between the growth rates of these cells and either control NIH3T3 cells, or puror cells (Fig. 4 A). In 0.5% of serum however, L1 expressing cells continued to grow and formed foci (Fig. 4 C), whereas control NIH3T3 and puror cells died after a few days (Fig. 4 B). We wished to determine which signaling pathways enabled L1 expressing NIH3T3 cells to grow in low serum. Cells were incubated in 0.5% of serum for 24 h and the levels of total ERK and activated (phosphorylated) ERK were determined. L1 expressing cells displayed constitutively activated ERK, whereas puror cells had no detectable P-ERK (Fig. 4 D, compare lane 1 with lane 2). When these cells were grown in 0.5% of serum and incubated with 10% of serum to induce ERK activation, both puror and L1 expressing cells maintained the ability to induce P-ERK upon serum stimulation (Fig. 4 D, lanes 3–8). This transient induction of ERK activation declined to undetectable levels by 24 h in puror cells, whereas L1 expressing cells continued to display sustained levels of P-ERK (Fig. 4 D, lanes 9 and 10). These results on ERK activation by L1 are in agreement with recent studies by Silletti et al. (2004).
In a previous study, expression of L1 was detected among genes suggested to be required for the growth of a variety of tumor, but not normal cells (Primiano et al., 2003). Hence, we examined if L1 is necessary for SW480 and HCT116 cell proliferation by incubating these cells with an antibody recognizing the extracellular domain of L1. As control, cells were incubated with denatured antibody (by boiling to destroy its activity). The L1 antibody blocked very effectively the proliferation of both SW480 and HCT116 cells (Fig. 4, E and F), suggesting that L1 expression is essential for colorectal cancer cell growth.
Overexpression of L1, or its suppression, affect cell motility, invasion, and tumorigenesis of colon cancer cells
To examine if L1 plays a role in determining the motile, invasive, and tumorigenic capacity of human colon cancer cells, we expressed L1 in LS174T human colon cancer cells that do not normally express L1 (Fig. 5 A), and found that L1 is mainly localized at cell–cell contact sites of such dense cell cultures (Fig. 5 B). Two independently isolated LS174T clones expressing L1 displayed an increased capacity to close an artificial wound introduced in a monolayer compared with control, neor cells (Fig. 5 C). When LS174T-neor cells were injected into one side of six nude mice, the L1 expressing cells injected on the other side formed much larger tumors (Fig. 5 D). This was observed with two distinct LS174-L1 cell clones (Fig. 5 E). Using another colon cancer cell line, SW707, that also lacks L1 (Fig. 5 F), we compared the haptotactic motility (toward fibronectin) in transwell filter assays before, and after their stable transfection with L1. We found that such cells display enhanced haptotactic motility compared with empty vector transfected cells (Fig. 5 G). Moreover, L1-transfected SW707 cells were more invasive through filters coated with matrigel than control SW707 cells (Fig. 5 H), and when injected into mice formed larger tumors (Fig. 5 I), indicating that L1 can confer enhanced ECM invasion and tumor growth also in these colon cancer cells. In contrast, suppression of endogenous L1 levels in HCT116 cells by L1 siRNA oligonucleotides (Fig. 5 J) led to decreased haptotactic motility (Fig. 5 K), whereas further elevation of L1 in HCT116 cells, by adenovirus-mediated infection, increased their motility (not depicted). Together, these results imply that L1 expression in human colon cancer cells plays an important role in their motile, invasive and tumorigenic capacities.
Expression of L1 and ADAM10 at the invasive front of human colorectal tumor tissue
We wished to detect and localize the expression of L1 in human colorectal cancer tissue. For this, we performed histochemical staining with antibodies to L1 and β-catenin in serial paraffin sections of both normal and tumor tissue derived from surgical resections. In normal colon epithelium, we observed β-catenin staining at cell–cell contact sites (Fig. 6 B), and only sporadic staining for L1 along intramucosal nerve axons (Fig. 6 A, arrow). In the central, differentiated area of the tumor, β-catenin overexpression was detected in both the membrane and cytoplasm (Fig. 6 D), and often in the nuclei of cells (Fig. 6 D, arrowhead; unpublished data). Yet, we did not detect L1 in this part of the tumor tissue (Fig. 6 C) after analyzing 25 colorectal adenocarcinomas. In contrast, the invasive front of tumors displayed strong L1 expression in budding tumor cells at the interphase between tumor and normal tissue (Fig. 6, E and G, arrowheads) and in nerve axons (Fig. 6 E, arrow). This phenotype was apparent in 68% of the tumors. Higher magnification of the invasive tumor front displayed strong L1 expression in invading tumor cells and also in disseminated, single tumor cells (Fig. 6 G) expressing nuclear β-catenin (Fig. 6 H, arrowhead). Interestingly, in some sections where the nerve axons were more readily detectable, we observed L1 expressing cancer cells (Fig. 6 I, arrowhead) displaying nuclear β-catenin (Fig. 6 J, arrowhead) at the invasive front of the tumor, growing directly along L1 expressing nerve axons (Fig. 6 I, arrow). Because both these cancer and nerve cells express L1 on their surface, L1-mediated adhesive interactions among such cells might guide the further dissemination of invading tumor cells.
Because we found that L1 and ADAM10 expression in colon cancer cells is coregulated by changes in cell culture density (Fig. 1, D–F), and ADAM10 is involved in the cleavage and shedding of the extracellular domain of L1 (Gutwein et al., 2003), we determined the expression of ADAM10 in human colorectal cancer tissue. Staining with antibody to ADAM10 (and in adjacent sections for L1) of colon cancer tissue revealed that in 74% of tumors (19 tumors were analyzed), L1 and ADAM10 colocalized at the invasive front of the tumor tissue (Fig. 7, A–D). These results indicate that L1 might be an important target gene of β-catenin-TCF signaling playing a key role, together with ADAM10, in the invasion of human colorectal cancer cells.
L1 is a novel target of β-catenin-TCF signaling in colon cancer cells
In the present study, we demonstrated that in colon cancer cells the β-catenin-TCF transcriptional complex activates the L1 gene, coding for a cell adhesion molecule mostly expressed in nerve cells. L1 levels were high in sparse cultures of colon cancer cells displaying characteristics of colorectal cancer cells at the invasive front of tumors (Brabletz et al., 2001). These include nuclear localization and signaling by β-catenin and down-regulation of E-cadherin by Slug, a negative transcriptional regulator of E-cadherin induced by β-catenin-TCF signaling (Conacci-Sorrell et al., 2003). Inhibition of β-catenin-TCF signaling in cultured colon cancer cells by inducible ΔNTCF, or suppression of β-catenin levels by inducible β-catenin siRNA, both reduced the expression of endogenous L1 in these cells (Fig. 2 F). In addition, the β-catenin-TCF complex in 293 cells and colorectal cancer cell lines activated the L1 promoter reporter. Together, these results support the view that L1 is a novel target gene of β-catenin signaling. This conclusion is in agreement with a previous study using DNA microarrays to determine genes down-regulated when wt APC was reintroduced into SW480 cells, and detected L1 among such genes (Lin et al., 2001).
The biological activities of L1 in tumor cells
During tumor cell invasion, cancer cells often activate genes that normally promote the growth and migration of normal cells (in the course of various biological processes), to promote tumor cell metastasis. L1 is known for its important function in nerve cell axonal guidance, outgrowth, and pathfinding (Kamiguchi et al., 1998; Kenwrick et al., 2000), and its disruption causes serious damage to the nervous system, including mental retardation (De Angelis et al., 2002) and intestinal aganglionosis (Parisi et al., 2002). In addition, L1 was also found in other specialized sites including a subclass of lymphocytes, intestinal crypt cells, and kidney tubule epithelia (Kowitz et al., 1993; Thor et al., 1987; Debiec et al., 1998). We have shown that expression of L1 in NIH3T3 cells (that lack L1) confers several properties characteristic of cancer cells, including enhanced motility, cell transformation (manifested by the formation of foci in cells grown on plastic), and the ability to grow in low serum. Because NIH3T3 cells are preneoplastic, it remains to be determined whether L1 can transform normal, primary epithelial cells. The ability of L1 to confer NIH3T3 growth in the absence of serum is related to constitutively activated ERK displayed by these cells (Fig. 4 D). Activation of the MAPK pathway by L1 was reported previously (Schaefer et al., 1999), and recently linked to the enhanced motility of 3T3 fibroblasts and melanoma cells (Silletti et al., 2004). Because constitutive ERK activation is considered a key step in cell transformation and the development of various tumors (Cowley et al., 1994; Hoshino et al., 1999; Welsh et al., 2000), its activation by L1 supports the notion that ERK-mediated gene expression is the major survival pathway allowing the proliferation of these cells in low serum. L1 expression also endowed these cells with increased tumorigenic capacity in nude mice, a finding similar to that reported with Nr-CAM, another L1 family member that is also a target gene of β-catenin-TCF signaling, overexpressed in melanoma and colon carcinoma tissue (Conacci-Sorrell et al., 2002b). The role of L1 in human colon cancer cell invasion and tumorigenesis is further supported by our data showing that its expression in human colon cancer cells lacking L1 increases haptotactic and invasive motility, and leads to faster tumor growth, whereas siRNA-mediated suppression of endogenous L1 in human colon cancer cells decreases cell motility (Fig. 5). Hence, L1 family members could function as potential promoters of colon cancer cell growth and invasion by being target genes of hyperactive β-catenin signaling.
Involvement of L1 and ADAM10 in human cancer invasion
Using the cell culture density model of human colon cancer cells, we found that the expression of L1 and of ADAM10 that mediates L1 cleavage and shedding (Mechtersheimer et al., 2001; Gutwein et al., 2003), is coregulated. Sparse cell cultures had much higher levels of L1 and ADAM10 protein and RNA than dense cultures (Fig. 1). Therefore, we analyzed their expression in human colorectal cancer tissue and compared it to that of β-catenin. We observed a very unique localization of L1 in a subpopulation of cancer cells expressing nuclear β-catenin at the invasive tumor front (in ∼70% of the tumors), but not in the central, more differentiated part of the tumor (Fig. 6, C–H). This finding strongly supports a role for L1 in the invasive process of human colorectal cancer cells in vivo. ADAM10 expression was similarly enriched in the invasive front of such tumors (Fig. 7), indicating that L1 shedding into the surrounding microenvironment may play a role in dissemination of colon carcinoma cells. Interestingly, we often observed in very close proximity to the invasive front of tumors nerve axons, the only other cell type expressing L1 in colon mucosa (Fig. 6, E and J). The close apposition of invading tumor cells expressing L1, either on their surface (or as shedded molecules in their microenvironment), to L1 expressing nerve axons, could promote adhesive interactions between tumor and nerve cells, thereby contributing to colon cancer cell invasion using axons as tracks for migration. Studies using cocultures of neuronal and colon cancer cells should shed light on the possible mechanisms involving such interactions and the cross signaling between the different cell types.
β-Catenin-TCF activation of target genes is also known to regulate normal epithelial colon cell proliferation along the crypt-villus interface (van de Wetering et al., 2002). The Wnt/β-catenin target genes involved, include the ephrins and their receptors (Batlle et al., 2002). These molecules help establishing the boundary between proliferating and differentiated cells in the highly dynamic and renewing colon epithelium (Peifer, 2002). These same ephrins and their receptors are known to set segmental boundaries in the brain and determine axonal guidance in the retina. This points to another set of “brain-specific” cell adhesion molecules as key regulators of both normal and perhaps pathological behavior in the human colon.
L1 was previously detected in some normal tissues including leukocytes, intestinal crypt cells and kidney tubule epithelia (for review see Kenwrick et al., 2000), and the involvement of L1 in the motile and invasive properties of a variety of cultured cancer cell lines (lymphoma, lung carcinoma, melanoma) was also suggested (Kowitz et al., 1993; Miyahara et al., 2001; Thies et al., 2002). The shedding of the L1 ectodomain by ADAM10 and its activity in promoting cell adhesion by mechanisms involving integrin receptors was demonstrated in several cancer cell lines (Mechtersheimer et al., 2001). A recent study using transcriptome-scale selection to identify suppressors of breast carcinoma cell growth discovered L1 as a major positive regulator of carcinoma (but not normal) breast epithelial cell growth, and also of some colon and cervical cancer cell lines (Primiano et al., 2003). In agreement with these findings, we found that the proliferation of both SW480 and HCT116 cells is inhibited when they are incubated with anti-L1 antibody (Fig. 4), suggesting a key role for L1 in tumor cell growth. L1 and ADAM10 expression were also recently detected in ovarian and uterine carcinomas, and L1 shedding into the serum and ascites of such patients was suggested to be a predictive marker for their clinical outcome of (Fogel et al., 2003). Because endometrial carcinoma is often associated with aberrant activation of the β-catenin signaling pathway (Palacios and Gamallo, 1998; Moreno-Bueno et al., 2002), it could be that in endometrial tumors, as well, activation of L1 is mediated by hyperactive β-catenin signaling. Together with the present study, these results imply that induction of L1 expression by β-catenin in malignant tumors might represent another case of opportunistic activation of target genes normally expressed in other tissues in the course of various biological processes. This property of cancer cells may facilitate their invasion and metastasis.
Materials And Methods
Cell lines and cell culture
293, NIH3T3, SW480, LS174T, SW707, and HCT116 cells were maintained in DME with 10% bovine calf serum (BCS). NIH3T3 cells expressing puror constructs were cultured in 10 μg/ml puromycin. HCT116 clones expressing the Tet-repressor were obtained from M. van de Wetering and H. Clevers (Hubrecht Laboratory, Utrecht, Netherlands), and were grown in RPMI with 5% FCS and Zeocin (500 μg/ml). HCT116 cells expressing inducible, ΔNTCF4, or siRNA to β-catenin, were prepared as described previously (van de Wetering et al., 2003), and obtained from M. van de Wetering. They were cultured with Zeocin (500 μg/ml) and Blasticidin (10 μg/ml). Transgenes were induced with 1 μg/ml doxycycline. Activation of MAPK was tested in cells starved for 18 h in DME. The cells were collected before, and at different times after incubation with 10% BCS.
Transfections and retroviral infection
Transient transfection of 293T cells was performed using calcium phosphate. SW480 and HCT116 cells were transfected using Lipofectamine 2000 (Invitrogen). In transactivation assays, 0.25 μg of β-galactosidase plasmid was cotransfected with 1 μg reporter plasmid and 2 μg of either β-catenin, or cadherin tail constructs. Cells were plated in duplicates, lysed after 48 h, and luciferase and β-galactosidase levels were determined by enzyme assay kits purchased from Promega. Luciferase activity was normalized to β-galactosidase activity as internal transfection control. Fold induction of the L1 and Cyclin D1 promoters was calculated using empty reporter plasmid (PGL2 and PA3luc, respectively). Retroviral infections were performed with pBabe-puror and pBabe-L1 into 293T cells (2 × 106 cells/10 cm dish), by the calcium phosphate coprecipitation method, together with the ecotropic packaging vector pSV-ψ-E-MLV providing the helper function as described previously (Conacci-Sorrell et al., 2002b). Transfection of the L1 siRNA sequence, 5′-GGAUGGUGUCCACUUCAAA-3′ and 5′-UUUGAAGUGGACACCAUCC-3′ and of control oligonucleotides obtained from QIAGEN was performed with annealed siRNAs using Oligofectamine (Life Technologies) and the transfected cells were analyzed after 72 h. LS174T cells stably expressing human L1 were established by transfection using Lipofectamine 2000, followed by neor selection. SW707 cells stably expressing human L1 were obtained by transfection with Superfect (Stratagene) and neomycin selection, followed by enrichment for L1 expression with mAb L1-11A and magnetic beads (Miltenyi Biotec).
The L1 promoter reporter plasmid containing the transcription initiation site and 2.9-kb upstream sequences (Kallunki et al., 1997; Meech et al., 1999), including the four putative LEF/TCF-binding sites, was cloned into PGL2 (provided by F. Jones, Lois Pope LIFE Center, Miami, FL). Full-length L1 cDNA (U52112) provided by V. Lemmon (University of Miami, Coral Gables, FL; Hlavin and Lemmon, 1991) was subcloned into the EcoRI site of pBabe-puror. ΔNTCF4 was provided by M. van de Wetering and H. Clevers, as were the TOPFLASH and FOPFLASH synthetic promoter reporter plasmids. ΔLEF was a gift from R. Kemler (Max Planck Institute for Immunobiology, Freiburg, Germany). The mutant β-catenin S33Y, Cyclin D1 reporter plasmid, dominant positive LEF-1 construct (Shtutman et al., 1999) and the plasmid coding for the cytoplasmic domain of E-cadherin (Cad tail; Sadot et al., 1998) were described previously.
RT-PCR and electrophoretic mobility shift assays
RT-PCR for L1 was performed using the primers: ACGGGCAACAACAGCAAC and CGGCTTCCTGTCAATCATG, and for GAPDH: ACCACAGTCCATGCCATCAC and TCCACCACCCTGTTGCTGTA. The primers for E-cadherin and ADAM10 and PCR conditions were as described previously (Fogel et al., 2003). For electrophoretic mobility shift assays, double-stranded DNA oligonucleotides containing the putative LEF/TCF-binding sites of the L1 promoter (Fig. 2 C) and 10 adjacent nucleotides upstream and 10 downstream were used. SW480 cells were harvested, incubated for 15 min in low-salt buffer, NP-40 was added, nuclei were pelleted by centrifugation, and nuclear proteins extracted with high-salt buffer as described previously (Shtutman et al., 1999). For DNA-binding assays, 6 μg of nuclear extract were used. Increasing volumes of polyclonal β-catenin antibody were added to the binding reaction to analyze DNA mobility supershift.
Cells cultured on glass coverslips were permeabilized with 0.5% Triton X-100 and fixed with 3% PFA in PBS. The coverslips were incubated with polyclonal antibody against the extracellular domain of L1 provided by V. Lemmon (Schaefer et al., 1999), or the mAb L1-11A (a subclone derived from hybridoma UJ127.11; Mechtersheimer et al., 2001). The mAb against β-catenin was purchased from Transduction Laboratories. The secondary antibodies were Alexa 488–conjugated goat anti–mouse or anti–rabbit IgG (Molecular Probes) and Cy3-labeled goat anti–mouse, or anti–rabbit IgG (Jackson ImmunoResearch Laboratories), as described by Simcha et al. (1998). Images were acquired using the DeltaVision system (Applied Precision) equipped with a microscope (model Axiovert 100; Carl Zeiss MicroImaging, Inc.) and Photometrics 300 series scientific-grade cooled CCD camera, reading 12-bit images, and using the 63×/1.4 NA plan-Neofluar objective. Adjustments of brightness, contrast, color balance, and final size of images was processed using Adobe Photoshop 5.5.
Western blotting, cell growth rate and wound healing
Additional antibodies used in Western blotting were mouse anti-tubulin (Sigma-Aldrich), mouse anti–E-cadherin (Transduction Laboratories), goat anti-ADAM10 (R&D Systems), mouse anti-MAPK and rabbit antiphosphorylated MAPK (Sigma-Aldrich). The Western blots were developed using the ECL method (Amersham Biosciences). For cell growth rate, 5 × 103 cells were plated into 24-well dishes and their number determined every 24 h for 6 d in triplicates. Cells were also grown in the presence of antibody against L1-CAM, at 1:200 dilution with HCT116 and 1:50 with SW480 cells. Anti-L1 antibody (boiled for 10 min) was used as control. A “wound” was introduced into a confluent monolayer of cells with the tip of a micropipette as described previously (Conacci-Sorrell et al., 2002b), the culture medium replaced with fresh medium, and 40 μg/ml of Mitomycin C was added to inhibit cell proliferation. After 24 h, the cells were fixed with ice-cold methanol, stained with Giemsa, and photographed.
FACS analysis, transmigration, and matrigel invasion assays
For FACS analysis, the cells were stained with mAb to L1 and PE-conjugated secondary antibodies, or with secondary antibody alone as control (Mechtersheimer et al., 2001). Stained cells were analyzed with a FACScan using the Cellquest software (Becton Dickinson). Haptotactic cell migration assays (toward 10 μg/ml fibronectin coated on the backside of the filter) was performed in Transwell chambers (Costar; Mechtersheimer et al., 2001). After 16 h, cells that migrated to the backside of the filter were stained with crystal violet solution and the OD of the eluted dye was determined at 595 nm in an ELISA reader. Invasion assays were performed with 24-well BioCoat Matrigel Invasion Chambers (Becton Dickinson) using 5 × 104 cells in serum-free DME and plated onto either control or matrigel-coated filters. Conditioned medium from 3T3 cells was placed in the lower chambers as chemoattractant. After 22 h in culture, cells were removed from the upper surface of the filter by scraping with a cotton swab. Cells that invaded through the matrigel and were adherent to the bottom of the membrane were stained with crystal violet solution. The cell-associated dye was eluted with 10% acetic acid and its OD at 595 nm determined. Each experiment was done in quadruplicates and the mean values ±SEM are presented.
NIH3T3 cells (105 and 2.5 × 105) expressing retrovirally transduced L1, or the empty (puror) vector, or LS174T and LS174T-L1 cells (106 cells), were injected subcutaneously into 6-wk-old CD1 nude male mice. Control neor-LS174T cells were injected on one side, and LS174T-L1 on the other side of each animal in the presence or absence of matrigel (250 μl/animal), generously provided by H. Kleinman (National Institutes of Health, Bethesda, MD). Groups of six mice were followed for 2 wk, when the size of tumors reached ∼1–2 cm in diameter. After sacrificing the mice, the tumors were excised and weighed. Normal subcutaneous tissue was also excised, and RNA was prepared from normal and tumor tissue for RT-PCR. SW707 and SW707-L1 cells (107 cells/animal) were injected into the flanks of 4 scid/nod mice in each group, and tumor volume was determined at different times after injection by measuring tumor diameter.
Immunohistochemistry was performed as previously described (Brabletz et al., 1999). In brief, 3-μm serial sections from 25 formalin-fixed, paraffin-embedded, colorectal adenocarcinomas were deparaffinized, rehydrated, and pretreated for antigen retrieval by microwave treatment for 20 min in 10 μM of citrate buffer, pH 6.0. To detect L1 and ADAM10, the antibodies described above were used. Biotinylated rabbit anti–mouse Ig antiserum, or rabbit anti–goat antiserum (both diluted 1:200; DakoCytomation), served as secondary antibody. To detect L1 and β-catenin (brown staining), the Envision system was used for ADAM10 (red staining), and Strept/AB (for β-catenin) was used according to the manufacturer's protocol (DakoCytomation). After rinsing with water, sections were counterstained with hemalaun (Merck), dehydrated, and covered with coverslips.
We are grateful to our colleagues, Vance Lemmon, Marc van de Wetering, Hans Clevers, Frederick Jones, Rolf Kemler, and Hynda Kleinman for providing reagents.
These studies were supported by grants to A. Ben-Ze'ev from the German-Israeli Foundation for Scientific Research and Development (GIF), the MD Moross Institute for Cancer Research, Israel Cancer Research Fund, a grant from the Estate of Bronia Hacker, La Foundation Raphael et Regina Levy and from the Yad Abraham Center for Cancer Research and Diagnosis. P. Altevogt and T. Brabletz were supported by grants from the Deutsche Krebshilfe (Schwerpunktprogramm: Invasion and Migration) and Deutsche Forschungsgemeinschaft.
Abbreviations used in this paper: ΔNLEF, dominant negative LEF-1; ΔNTCF; dominant negative TCF.