Introduction

The integrin family of transmembrane receptors consists of eight and 18 subunits that assemble as heterodimers to form 24 distinct integrins (Hynes, 2002). The main ligands for integrins in the extracellular space are extracellular matrix proteins, such as laminin and collagen, as well as cellular counter-receptors. On their cytoplasmic domains, integrins are linked to the cytoskeleton, which they regulate and modulate via various sub-membrane linker proteins (Zamir and Geiger, 2001). Integrins transduce signals across the plasma membrane in both directions: integrin binding to its ligands requires its activation by inside-out signals. Conversely, integrin ligation triggers outside-in signals that regulate different aspects of cell behavior, including cell survival, control of transcription, cell proliferation, cell motility and cytoskeletal organization (Hynes, 2002).

Due to their broad spectrum of features, integrins and integrin signaling have been shown to contribute to tumor progression in various ways (Guo and Giancotti, 2004). For example, recent reports have highlighted the role of ₄-integrin in promoting mammary tumor progression by amplifying ErbB2 signaling and thereby activating the transcription factors STAT3 and c-Jun (Guo et al, 2006). On the other hand, _v3, _v5 and several integrins of the ₁ class have been implicated in angiogenesis. Indeed, inhibition of these molecules by antagonist molecules or specific antibodies against integrins efficiently blocks angiogenesis (Brooks et al, 1994, 1995; Friedlander et al, 1995; Sudhakar et al, 2003). Excessive joint signaling by ₁-integrin and tyrosine kinase receptors has been associated with disruption of adherens junctions, a prerequisite for metastatic invasion of cancer cells (Novak et al, 1998; Tan et al, 2001; Fujita et al, 2002; Zhang et al, 2003). Cell type-specific ablation of ₁-integrin function in the mouse has demonstrated that this integrin plays a key role in regulating cell cycle progression of luminal mammary epithelial cells and is critical for the alveolar morphogenesis of glandular epithelium (Li et al, 2005; Naylor et al, 2005). Furthermore, ₁-integrin expression has been proven critical for the initiation of mammary tumorigenesis in vivo, and for maintaining the proliferative capacity of late-stage tumor cells (White et al, 2004).

We have set out to investigate the functional role of ₁-integrin during tumor progression in Rip1Tag2 transgenic mice. In Rip1Tag2 mice (RT2 mice), the Simian Virus 40 large T antigen is expressed under the control of the rat insulin promoter, resulting in the reproducible development of pancreatic cell tumors (insulinomas) following a highly reproducible multistage tumorigenesis pathway (Hanahan, 1985). Our laboratories have previously reported that RT2 mice that are deficient for the expression of neural cell adhesion molecule (NCAM), RT2;NCAM^-/- mice, show striking disintegration of the tumor tissue architecture, and together with upregulated lymphangiogenesis, develop metastases to the local draining lymph nodes of the pancreas (Perl et al, 1999). Subsequent biochemical analyses have revealed that, by binding to and stimulating fibroblast growth factor (FGF) receptors, NCAM activates signaling pathways leading to the activation of ₁-integrin-mediated cell–matrix adhesion (Cavallaro et al, 2001). Based on such epistatic relation between NCAM and ₁-integrin functions, and the fact that the loss of NCAM function leads to lymph node metastasis in Rip1Tag2 mice, we hypothesized that the loss of ₁-integrin may at least in part recapitulate the loss of NCAM function.

We report here that deletion of ₁-integrin function during the development of islets of Langerhans results in a defect of the sorting of endocrine and cells in islets. Ablation of ₁-integrin expression during cell tumorigenesis in Rip1Tag2 mice led to the dissemination of tumor cell emboli into lymphatic vessels but did not induce increased lymphangiogenesis or metastasis. Mice carrying ₁-integrin-deficient tumors exhibited reduced tumor burden, reduced tumor cell proliferation and the induction of cellular senescence, potentially via a pathway involving the upregulation of the cell-dependent kinase inhibitor p21^Cip1. Consistent with these results, tumor cells that lack ₁-integrin were not able to form tumors or metastases in tumor transplantation experiments. Together, the results highlight a critical role of ₁-integrin in the metastatic dissemination of tumor cells and in cellular senescence.

Results

₁-integrin is required for islet cell sorting

Complete loss of ₁-integrin function in ₁-integrin knockout mice results in embryonic lethality (Fassler and Meyer, 1995). We therefore employed mice carrying ₁-integrin alleles flanked by loxP sites (1^(fl/fl) mice) (Potocnik et al, 2000) to ablate ₁-integrin function specifically in normal pancreatic cells and tumor cells, by crossing them to mice expressing the Cre recombinase under the control of the cell-specific rat insulin promoter (RCre) (Ahlgren et al, 1998), and subsequently with Rip1Tag2 mice.

To assess the effects of ₁-integrin deletion during normal islet development, islets of RCre;1^(fl/fl) mice were compared to either wild-type C57Bl/6 or 1^(fl/fl) control mice. Since Cre recombinase is already expressed during islet development in RCre mice, ₁-integrin expression is ablated in insulin-expressing cells at early stages of islet development (Ahlgren et al, 1998; Tan et al, 2001). Histopathological analysis of 8- to 10-week-old RCre;1^(fl/fl) mice showed no apparent changes in islet number, size and architecture (Figure 1A, upper panel and data not shown). However, immunofluorescence stainings of histological sections for insulin and glucagon revealed differences in the organization of islet cells (Figure 1A, lower panel). In control animals, most islets of Langerhans (82.4%) displayed glucagon-producing cells located within the three most peripheral cell layers of the islets, whereas insulin-expressing cells were mainly localized in the center of the islets (Figure 1A, lower left panel). Only in a small percentage of control islets cells were also found within the center of islets (Table I, 17.6%), hereafter referred to as mixed sorting phenotype. In contrast, most islets of RCre;1^(fl/fl) mice were of the mixed sorting phenotype (81% mixed versus 19% normal phenotype, Figure 1A, lower right panel and Table I). Average numbers of cells per islet area were found to be increased in RCre;1^(fl/fl) islets as compared with wild-type controls (1.307 cells/1000 m² versus 2.205 cells/1000 m²; Table I, P<0.005, unpaired t-test). Together, these results indicate that deletion of ₁-integrin in pancreatic cells leads to disturbances in cell type segregation during islet development, a phenotype also observed in islets lacking NCAM expression (Esni et al, 1999) or expressing a dominant-negative version of E-cadherin in cells (Dahl et al, 1996).

Loss of ₁-integrin induces tumor cell dissemination into lymphatics

To investigate the role of ₁-integrin in vivo during Rip1Tag2 tumorigenesis, Rip1Tag2 (RT2) mice were crossed with 1^(fl/fl) and RCre mice. Recombination of the ₁-integrin gene in RCre;RT2;1^(fl/fl) experimental and RT2;1^(fl/fl) control mice was monitored by determining ₁-integrin protein levels in tumors of 12-week-old mice. Dispersed tumor cells were subjected to FACS analysis by performing a CD31 (endothelial cell marker) and ₁-integrin co-staining, excluding endothelial cells from the analysis. More than 95% (96.422.34%) of the cells derived from RT2;1^(fl/fl) tumors show ₁-integrin expression (region R2 in Figure 1B), whereas in RCre;RT2;1^(fl/fl) tumors a second, ₁-integrin negative population appeared (region R1 in Figure 1B). This population represents around 40% (39.243.07%) of all analyzed cells. Thus, the 1^(fl/fl) allele is completely recombined in about 40% of all tumor cells. The ratio between and cells was unchanged between tumors of RT2;1^(fl/fl) and RCre;RT2;1^(fl/fl) mice; in all tumors the majority of cells were insulin-producing tumor cells (Supplementary Figure 2E).

We have previously reported that cell tumors in NCAM-deficient mice show increased tumor-associated lymphangiogenesis (Crnic et al, 2004), as well as tumor tissue disintegration and increased metastasis (Perl et al, 1999). Based on the fact that NCAM modulates ₁-integrin function by stimulating FGF receptor signaling in tumor cells (Cavallaro et al, 2001), we analyzed sections of 11- to 13-week-old RCre;RT2;1^(fl/fl) mice for the presence of lymphatic vessels by immunohistochemical staining of pancreatic sections with antibodies against the lymphatic marker LYVE-1 (Banerji et al, 1999). Tumors were divided into five groups according to the degree of lymphatic vessels surrounding the tumor perimeter (no lymphatic vessels=0%, 1–10%, 10–25%, 25–50% or >50% of tumor perimeter covered by lymphatic vessels; Figure 2A). No significant changes in tumor-associated lymphangiogenesis were observed for any defined group. Moreover, as determined by immunohistochemical staining, expression of the lymphangiogenic factors VEGF-C and VEGF-D was not induced in ₁-integrin-deleted tumors (data not shown). From these results, we conclude that the loss of ₁-integrin function does not induce lymphangiogenesis. Moreover, no changes in blood vessel density, morphology, pericyte coverage and tumor-infiltrating inflammatory cells was observed between tumors of RT2;1^(fl/fl) and RCre;RT2;1^(fl/fl) mice (Supplementary Figure 2A–D).

Ablation of NCAM in the Rip1Tag2 model leads to a severe change in tumor architecture, namely marked tissue disintegration and the appearance of hemorrhagic lacunae (Cavallaro et al, 2001; Xian et al, 2006). Histological analysis of pancreatic sections from RT2;1^(fl/fl) and RCre;RT2;1^(fl/fl) tumors did not reveal alterations in the appearance of hemorrhagic lacunae and tissue disintegration between the two different genotype tumors. However, in these experiments and also during the analysis of sections stained for LYVE-1 (see above), we found a significant increase in the number of disseminated tumor cell emboli within lymphatic vessels in the close vicinity of ₁-integrin-deficient tumors (Figure 2B, left panel) or tumor edges (Figure 2B, right panel). Such tumor cell emboli were detected in 60% of all analyzed RCre;RT2;1^(fl/fl) mice and only in 7.7% of control mice (Figure 2C). However, detailed macroscopic and microscopic analyses did not reveal the occurrence of metastases in any of the mice, neither to local lymph nodes nor to distant organs. These results suggest that the loss of ₁-integrin, while facilitating the displacement of tumor cell emboli into the lymphatic vasculature, is not sufficient to induce metastasis formation.

Loss of ₁-integrin-mediated cell–matrix adhesion

To investigate how the deletion of ₁-integrin contributes to the dissemination of tumor cells, we established cell lines from RCre;RT2;1^(fl/fl) experimental as well as RT2;1^(fl/fl) control tumors. PCR analysis confirmed the presence of two floxed ₁-integrin alleles in control cells (Ti^(fl/fl)), and of two deleted alleles in cells derived from experimental tumors (Ti^(/); Figure 3A, upper panel). Cell lines derived from tumors of normal Rip1Tag2 mice (T2) and of NCAM-deficient Rip1Tag2 mice (TN2; Cavallaro et al, 2001), both carrying wild-type ₁-integrin alleles, were employed as controls (Figure 3A, upper panel). FACS analysis revealed that Ti^(fl/fl), T2 and TN2 expressed comparable levels of ₁-integrin at their cell surface (data not shown) and confirmed that Ti^(/) cells indeed had lost ₁-integrin expression (Figure 3B). When establishing primary tumor cell lines from tumors of Rip1Tag2 mice, some cell lines still express high levels of NCAM, whereas others do not. Since both Ti^(fl/fl) cells and ₁-integrin-deficient Ti^(/) cell lines expressed only low amounts of NCAM, the T2 cell line that expressed high levels of NCAM and the TN2 cell line that did not express any NCAM were included to directly compare the function of NCAM and ₁-integrin in tumor cells side by side (Figure 3A, lower panel).

Previously, we have reported that the activation of ₁-integrin-mediated adhesion of tumor cells to collagen type IV is abrogated in NCAM-deficient TN2 cells and can be re-established by re-introducing NCAM (Cavallaro et al, 2001). We thus compared the adhesion capabilities of Ti^(fl/fl) and Ti^(/) cell lines in the presence or absence of NCAM. Equal amounts of cells were seeded on plates coated with various matrix proteins, and adhering cells were scored (Figure 3C). Adhesion of the Ti^(/) cell line to collagen IV was very low and could not be increased by forced expression of NCAM140. Adhesion of Mock-transfected Ti^(fl/fl) cells was comparable to that of Ti^(/) cells, however, forced expression of NCAM140 significantly stimulated adhesion to collagen IV (P<0.05, unpaired t-test) but not to collagen type I, fibronectin or uncoated plastic (Table III). Cultured tumor cells are known to lack significant migratory and invasive capabilities (Cavallaro et al 2001), and this feature was not changed by the loss of ₁-integrin function. These results further confirm that the genetic ablation of ₁-integrin in tumor cells results in specific deficiencies in their matrix adhesion and that NCAM signaling is epistatic to ₁-integrin-mediated cell–matrix adhesion in tumor cells.

₁-integrin but not NCAM is necessary for metastasis formation

We next asked whether cell lines deficient in either ₁-integrin or NCAM expression were able to form tumors and metastases in syngeneic tumor transplantation experiments in vivo. Equal numbers of Ti^(fl/fl), Ti^(/), TN2 and T2 cells were injected subcutaneously into the flanks of immunocompetent C57Bl/6 mice. After 5 weeks, mice were killed and tumor incidence, tumor size and tumor volumes were determined (Figure 4A and B). Primary tumors formed in 12.5% of all T2-injected sites and in 37.5% of all Ti^(fl/fl)-injected sites (Figure 4B). The average size of tumors from T2 cells was 293.4129.4 mm³ as compared with 174.753.1 mm³ for tumors arising from Ti^(fl/fl) cells (Figure 4A). In contrast, ₁-integrin-deficient Ti^(/) and NCAM-deficient TN2 cells did not give rise to any tumors.

To assess the metastatic potential of these cell lines, we injected equal numbers of cells into the tail vein of athymic nu/nu mice. After 4 weeks, mice were killed, pancreata, lungs and livers of all mice were isolated, and subjected to histological and microscopic examination for metastatic lesions. All tumor cell lines, with the exception of Ti^(/), lacking ₁-integrin expression, induced the formation of metastatic nodules in lungs and/or liver (Table II).

These results indicate that depletion of NCAM and therefore reduction of ₁-integrin activity in TN2 cells diminishes tumor formation but still allows them to colonize distant organs. However, the total loss of ₁-integrin function results in an incapability to form primary tumors and colonizing nodules/metastases, indicating a critical role for ₁-integrin in these processes.

Reduced tumor cell proliferation in the absence of ₁-integrin

We next examined the effects of ₁-integrin depletion on RipTag2 tumorigenesis. Control and experimental mice were killed at the age of 12 weeks, shortly before the mice succumb to hypoglycemia caused by the insulin-expressing tumors. Tumor-bearing pancreata were excised and tumor incidence was scored by counting and measuring macroscopically visible (>1 mm) tumors. Tumor incidence (i.e. number of tumors per mouse; Figure 5A) remained unchanged in RCre;RT2;1^(fl/fl) mice as compared with RT2;1^(fl/fl) control mice. However, total tumor volumes per mouse were found to be significantly reduced by the absence of ₁-integrin in tumors of RCre;RT2;1^(fl/fl) mice (Figure 5B).

We investigated whether the decreased tumor burden was due to reduced proliferation rates or increased apoptosis of tumor cells in ₁-integrin-deficient tumors. Mice were injected with BrdU before being killed, and tumor cell proliferation was determined by visualizing BrdU incorporation in histological sections, with antibodies against BrdU (Figure 5C). Tumor cell proliferation was significantly reduced in the absence of ₁-integrin in RCre;RT2;1^(fl/fl) tumors as compared with tumors of RT2;1^(fl/fl) mice. The apoptotic rate of tumor cells, as determined by TUNEL staining of histological sections, was also reduced in tumors of RCre;RT2;1^(fl/fl) mice as compared with tumors of RT2;1^(fl/fl) mice (Figure 5D). Tumor progression, that is, the incidence of malignant cancer and metastasis, was not affected at all, since tumors of both genotypes showed a similar incidence of adenomas and grade 1, 2 and 3 carcinomas and no apparent metastasis (Figure 5E). It should be also noted that interfering with ₁-integrin function did not affect the onset, extent or quality of tumor angiogenesis, as assessed by staining for the endothelial cell marker CD31 (see above; Supplementary Figure 2).

Consistent with the in vivo observation described above, the growth rates of the ₁-integrin-deleted cell line Ti^(/) were significantly slower than that of the control cell line Ti^(fl/fl) (Figure 5F, k=0.0014 for Ti^(/) versus k=0.0025 for Ti^(fl/fl); P<0.0001, unpaired t-test of linear regression curves). The lack of ₁-integrin also affected the growth behavior of tumor cells in a three-dimensional (3D) Matrigel^TM culture system. Most of the cell spheres formed by Ti^(fl/fl) cells formed filopodia-like protrusions (Figure 6A, left panels), whereas Ti^(/) cells were unable to form these protrusions (Figure 6A, right panels). Quantification of this effect by counting cells with or without protrusion in a defined volume revealed that more than 70% of the Ti^(fl/fl) cells developed protrusions, whereas not a single of the ₁-integrin-deficient cells showed the formation of these structures. A week after plating, all Ti^(/) had died, whereas ₁-integrin-expressing Ti^(fl/fl) cells were still alive (data not shown).

Loss of ₁-integrin induces senescence in cell tumors

The diminished levels of tumor cell proliferation as well as tumor cell apoptosis raises the possibility that ₁-integrin-deficient tumor cells might be impeded in entering and progressing through the cell cycle. To investigate the cell cycle status of cells from RT2;1^(fl/fl) control tumors compared with RCre;RT2;1^(fl/fl) experimental tumors, tumor cell suspensions of tumors from 12-week-old mice were stained with Pyronin Y (staining for RNA) and Hoechst 33342 (staining for DNA), allowing discrimination of cells in cell cycle phase G0, G1 or G2/S/M (Shapiro, 1981). Significantly more cells (P<0.05, unpaired t-test) derived from RCre;RT2;1^(fl/fl) tumors were found to stall in G0 as compared with ₁-integrin-expressing control tumor cells (Figure 7A). Consistent with this observation, cultured ₁-integrin-deficient tumor cells (Ti^(/)) also exhibited a significantly increased cell cycle arrest in G0/G1 (Supplementary Figure 1).

Reduced metabolic activity and cell cycle arrest in G0 are associated with cellular quiescence, senescence and/or tumor dormancy. Senescence-associated -galactosidase (SA--Gal) is expressed in senescent cells and is therefore frequently used as a marker for cellular senescence. It can be specifically detected at pH 6 by histochemical staining (Dimri et al, 1995). In contrast, bacterial -galactosidase is most active at a pH of 7.5. Since the 1^(fl/fl) mouse line used for the experiments described above has been generated in a way that, upon efficient recombination, a bacterial -galactosidase reporter comes under the control of the ₁-integrin promoter, and because bacterial -galactosidase also displays activity at pH 6, we could not test for SA--Gal in these mice. Therefore, we took advantage of an alternative mouse line in which the conditional ₁-integrin alleles do not carry a -galactosidase reporter gene (1EII mice). 1EII mice were intercrossed to the RT2 tumor model, giving rise to RCre;RT2;1EII experimental and RT2;1EII control mice. When compared to RCre;RT2;1^(fl/fl) mice, RCre;RT2;1EII experimental mice exhibited an identical tumor phenotype, including a reduced tumor burden and increased tumor cell embolization in the lymphatics (data not shown). Notably, histological sections of tumors from RCre;RT2;1EII showed positive staining for SA--Gal (Figure 7B, left and middle panels), whereas no staining was detectable in RT2;1EII control mice (Figure 7B, right panels). A total of 57 control and 62 experimental tumors were investigated. Whereas the vast majority of control tumors did not stain for SA--Gal (Figure 7C, 98.2%), SA--Gal activity was detected in approximately 50% of experimental tumors (Figure 7C). Cultured ₁-integrin-deficient tumor cells (Ti^(/)) also exhibited increased activity of SA--Gal, as compared with the control cell line Ti^(fl/fl) (data not shown). These results suggest that deletion of ₁-integrin in RT2 tumors reduces tumor burden by inducing tumor cell senescence.

The cyclin-dependent kinase (CDK) inhibitor p21^Cip1 has been shown to be upregulated in various cell lines during senescence (Xiong et al, 1993; Alcorta et al, 1996), and increased expression of p21^Cip1 leads to the induction of a permanent growth arrest with characteristics of cellular senescence in several cancer cell types (Fang et al, 1999). Recent studies on alveolar epithelial cells correlated ₁-integrin-deficiency with inhibited luminal cell proliferation and a concomitant increase of p21^Cip1 expression (Li et al, 2005). We therefore investigated p21^Cip1 expression levels in tumor lysates derived from control (RT2;1EII and RT2;1^(fl/fl)) and ₁-integrin-deficient experimental (RCre;RT2;1EII and RCre;RT2;1^(fl/fl)) mice by quantitative immunoblotting (Figure 8A). Comparison of the relative p21^Cip1 protein levels revealed an increase of p21^Cip1 expression at an average of 7.7-fold in experimental tumors in vivo (Figure 8B). Upregulated expression of p21^Cip1 was also observed in Ti^(/) cells (Supplementary Figure 1B). No change in the expression of any other cell cycle inhibitors or cell cycle regulatory genes tested (p15, p16, p27, p57, cyclin D2, cyclin E) was observed; only the expression of cyclin B was moderately increased in the absence of ₁-integrin function. This suggests that loss of ₁-integrin in tumor cells induces cellular senescence by a pathway that involves, among other molecular processes, upregulated expression of the CDK inhibitor p21^Cip1.

Discussion

Recent work from several laboratories, including our own, has demonstrated that the loss of NCAM function during cell carcinogenesis in RT2 transgenic mice results in the loss of ₁-integrin activation, tumor tissue disintegration, upregulated lymphangiogenesis and, finally, in metastasis to regional lymph nodes (Perl et al, 1999; Cavallaro et al, 2001; Crnic et al, 2004; Xian et al, 2006). Motivated by these observations, we set out to assess whether the loss of ₁-integrin had any impact on RT2 tumorigenesis and whether it phenocopied the loss of NCAM function. Specific inactivation of ₁-integrin by Cre-mediated recombination of loxP-flanked ₁-integrin alleles occurred in approximately 40% of tumor cells and resulted in the dissemination of tumor cell emboli into the lymphatic vasculature, yet did not induce lymphangiogenesis. Notably, ₁-integrin-deficient tumor cells displayed decreased rates of proliferation and apoptosis by acquiring cellular senescence. Moreover, tumor cell lines lacking ₁-integrin expression were unable to form tumors or to colonize distant organs in tumor transplantation experiments. Together, the results indicate that ₁-integrin plays a critical role in tumor outgrowth by preventing cellular senescence, the lymphogenic dissemination of tumor cell emboli and their colonization of distant metastatic sites.

Defects in endocrine cell sorting during islet development

Besides mild defects in the development of the nervous system, the loss of NCAM function in NCAM^-/- mice results in a cell sorting defect during the development of the islets of Langerhans (Esni et al, 1999). In normal islets of Langerhans, glucagon-producing cells are located at the periphery of the islets, whereas insulin-producing cells localize to the center of the islets. In the absence of NCAM expression, and cells are intermixed in the islets. Here, we report an identical phenotype in RCre;1^(fl/fl) mice, in which ₁-integrin function is specifically ablated in cells during islet development. Also here, the intermixing of and cells does not affect islet physiology or metabolism of the animal. These results indicate that the loss of ₁-integrin recapitulates the loss of NCAM in islet development, further supporting the epistatic link between NCAM and ₁-integrin functions.

Tumor cell emboli but no metastasis

Increased tumor lymphangiogenesis has been demonstrated to correlate with the incidence of lymph node metastasis in many experimental systems and cancer types in patients (Cao, 2005). Interestingly, in tumors of NCAM-deficient RT2 mice, the expression of the two lymphangiogenic factors VEGF-C and D was found to be upregulated, and interference with their activities resulted in a repression of lymphangiogenesis and lymph node metastasis in these mice (Crnic et al, 2004). RT2;NCAM^-/- tumors also show severe tissue disintegration and the appearance of large hemorrhagic lacunae. A recent report linked increased blood hemorrhage and the concomitant tissue disintegration phenotype to decreased pericyte recruitment to endothelial cells (Xian et al, 2006). Notably, interference with pericyte recruitment by ablating PDGF B function during cell carcinogenesis in RT2 mice resulted in blood vessel hemorrhage and tumor metastasis in the absence of lymphangiogenesis (Xian et al, 2006). However, since the observed tissue disintegration of NCAM-deficient cell tumors may potentially also result from a loss of ₁-integrin-mediated cell adhesion (Cavallaro et al, 2001), we assessed whether the inactivation of ₁-integrin might be responsible for the upregulated lymphangiogenesis observed in RT2;NCAM^-/- mice. Our results show that ₁-integrin deletion does not lead to an upregulated expression of the lymphangiogenic factors VEGF-C or VEGF-D and to tumor lymphangiogenesis and, thus, that the increased lymphangiogenesis observed in RT2;NCAM^-/- mice is not directly due to a defect in ₁-integrin function.

In RCre;RT2;1^(fl/fl) mice, tumor architecture per se is not altered, but clusters of tumor cell emboli are found circulating in lymphatic vessels in the absence of activated lymphangiogenesis. Disseminated tumor cell emboli are also found in RT2;NCAM^-/- mice, although in the presence of lymphangiogenesis, suggesting that in NCAM-deficient tumors two different mechanisms are employed: impaired pericyte recruitment (Xian et al, 2006) and loss of cell matrix adhesion (this work). Both of these processes result in tumor cell dissemination and lymph node metastasis.

Taken together, the results reported here support our previous findings that the loss of ₁-integrin function in cells reduces their cell–matrix adhesion and causes tumor cell dissemination. However, the results indicate that the loss of ₁-integrin function by itself does not lead to an induction of tumor lymphangiogenesis and to lymph node metastasis.

₁-integrin signaling is required for metastasis formation

We have employed a variety of tumor cell lines established from RT2 tumors in adhesion, proliferation and transplantation assays, to discriminate the role of adhesive and growth promoting integrin functions (Table III). Control T2 cells have very high NCAM levels and hence show high levels of ₁-integrin activation. TN2 cells lack NCAM expression and thus NCAM-dependent, ₁-integrin-mediated matrix adhesion. In contrast, tumor cell lines established from RCre;RT2;1^(fl/fl) tumors (Ti^(/) cells) lack ₁-integrin expression.

In contrast to the tumor cell emboli in RT2;NCAM^-/- mice, which disseminate to local lymph nodes, tumor cell emboli in RCre;RT2;1^(fl/fl) mice are not able to form metastasis. Consistent with this notion, Ti^(/) cells that lack ₁-integrin expression also do not grow primary tumors or colonize distant organs when injected into the tail veins of immunodeficient mice. This failure may not be due to their loss of ₁-integrin-mediated cell–matrix adhesion, since TN2 cells, which lack NCAM-dependent activation of ₁-integrin and thus cell adhesion, can still metastasize. On the other hand, both cell lines fail to form primary tumors when injected under the skin, suggesting that both NCAM and ₁-integrin are required for primary tumor growth. These results are consistent with published work reporting that ₁-integrin-deficient ES cells fail to efficiently form teratomas in transplanted mice (Bloch et al, 1997).

Together, the data indicate that ₁-integrin is required to allow cells to survive and proliferate at distant sites. This notion is consistent with a previous report that unligated integrins induce caspase 8-dependent integrin-mediated cell death (IMD; Stupack et al, 2001). However, ₁-integrin is required for establishing primary tumors in transplantation experiments, yet not in endogenously growing tumors of RT2 transgenic mice. This difference is likely due to the specific tumor microenvironment that may compensate for the loss of ₁-integrin function.

Loss of ₁-integrin induces cellular senescence

In ₁-integrin-deficient tumors of RCre;RT2;1^(fl/fl) mice, tumor volumes, but not tumor incidence and tumor progression, is reduced as compared with control mice. Lack of cell–matrix adhesion may result in a loss of signals for cell survival and proliferation and the induction of anoikis, that is, apoptosis caused by the lack of cell attachment (Frisch and Ruoslahti, 1997). However, both the rate of tumor cell apoptosis and proliferation are found reduced in ₁-integrin-deficient RCre;RT2;1^(fl/fl) tumors as compared with ₁-integrin-expressing control tumors. A similar effect of ₁-integrin function has been previously reported in various models of breast carcinogenesis, where its loss also represses tumor cell proliferation in the absence of increased apoptosis (White et al, 2004; Li et al, 2005). In these experiments, the loss of ₁-integrin correlates with reduced phosphorylation of focal adhesion kinase (FAK), an increased expression of the cell cycle inhibitory protein p21^Cip1, and the arrest of cells in the G1 phase of the cell cycle. Consistent with these reports, we find that the expression of p21^Cip1 is also upregulated in tumors of RCre;RT2;1^(fl/fl) mice. Moreover, ₁-integrin-deficient tumor cells are retained in the G0/G1 phase of the cell cycle and acquire the expression of SA--Gal, all specific marker for cellular senescence.

Together, the results in RT2 transgenic mice and cell tumor cell lines derived thereof indicate that targeting ₁-integrin function induces cellular senescence and represses the metastatic potential of tumor cell lines. Hence, interfering with the function of ₁-integrin during tumor progression, for example, by neutralizing antibodies or by antagonistic peptides, may offer a valid approach for cancer treatment. In fact, recent studies on mammary epithelial cell lines indicate that blocking ₁-integrin function inhibits proliferation and tumorigenic morphology of cultured breast cancer cells (Weaver et al, 1997; Wang et al, 2002). Our results now raise the intriguing possibility that the ablation of ₁-integrin function results in the induction of cellular senescence in tumor cells, a potential therapeutic goal. Future experimentation is warranted to elucidate the molecular players and pathways underlying the induction of cellular senescence upon loss of ₁-integrin function.

Materials and methods

Transgenic mice

Generation and phenotypic characterization of Rip1Tag2 mice, RCre mice and 1^(fl/fl) mice have been described previously (Hanahan, 1985; Ahlgren et al, 1998; Potocnik et al, 2000). 1EII mice will be described elsewhere (R Fässler, unpublished results). All mouse lines were backcrossed to C57Bl/6 for at least five generations. All experiments involving mice were performed in accordance with the guidelines of the Swiss Federal Veterinary Office (SFVO) and the regulations of the Cantonal Veterinary Office of Basel-Stadt.

Histological analysis

All mice were killed between 12 and 13 weeks of age, unless otherwise stated. Tumor incidence per mouse was determined by counting all macroscopically apparent tumors with a minimal diameter of 1 mm. Tumor volume was defined as total tumor volume per mouse in mm³, calculated by measuring the tumor diameters assuming a spherical shape of the tumors.

Tumors and pancreata were fixed overnight in 4% paraformaldehyde in PBS, dehydrated and paraffin embedded. Tissue sections (5 m thick) were deparaffinized and rehydrated before histological and immunohistochemical stainings.

Tumors were categorized into subclasses by inspection of hematoxylin-/eosin-stained histological sections: normal/hyperplastic, adenoma, carcinoma grade 1 (well differentiated, one invasive tumor edge), carcinoma grade 2 (partially dedifferentiated, tumor capsule largely absent, more than one invasive tumor edge), carcinoma grade 3 or anaplastic tumor (complete loss of tumor cell differentiation).

The following antibodies were used for immunohistochemical or immunofluorescence stainings on paraffin sections: guinea pig anti-insulin (DakoCytomation, Glostrup, Denmark), rabbit anti-mouse LYVE-1 (Reliatech, Braunschweig, Germany), goat anti-glucagon (Santa Cruz, Santa Cruz, CA, USA), biotinylated mouse anti-BrdU (Zymed, South San Francisco, CA, USA), In situ Cell Death Detection kit, (TUNEL, Roche, Basel, Switzerland). All biotinylated secondary antibodies for immunohistochemistry (Vector, Burlingame, CA, USA) were used at a 1:200 dilution, and positive staining was visualized with the ABC horseradish peroxidase kit (Vector, Burlingame, CA, USA) and DAB Peroxidase Substrate (Sigma Chemical Co., St Louis, MO, USA), according to the manufacturers' recommendations. For the analysis of tissue morphology, slides were slightly counterstained with hematoxylin or eosin. Alexa Fluor 568- and 488-labeled secondary antibodies (Molecular Probes, Eugene, OR, USA) diluted 1:400 were used for immunofluorescence analysis. 6-Diamidino-2-phenylindole (DAPI, Sigma) was used for nuclear staining in immunofluorescence stainings. All paraffin-embedded sections were subject to antigen retrieval with 10 mM citrate buffer (microwave), except with stainings for insulin and glucagon (10 min in 0.2% Triton X-100 in PBS), BrdU (1 h in 2 N HCl and subsequently 1 h 1 trypsin at room temperature (RT)) and TUNEL (10 min 2 g/ml proteinase K (Sigma) at RT).

For BrdU labeling, 100 g BrdU (Sigma) per gram body weight was injected 90 min before killing the mice. To determine tumor cell proliferation/apoptotic indices, BrdU/TUNEL-positive nuclei were counted per randomly chosen 40 magnification field of tumor tissue. Approximately 10 fields per mouse were counted.

Lymphangiogenesis was quantified by assessing the extent by which LYVE-1-positive lymphatic vessels surrounded the tumor perimeter: tumors that were not in contact with any lymphatic vessel (0%), tumors that were surrounded less than 10% (<10%), less than 25% (<25%), less than 50% (<50%) and tumors that were surrounded by lymphatics more than 50% of the tumor perimeter (>50%).

For immunohistochemical detection of SA--Gal, 20 m sections of frozen tissues were cut and fixed in 70% ethanol for 3 min and treated as described (Dimri et al, 1995).

Stained sections were viewed on an Axioskop 2 plus light microsope, using axiovision 3.1. software (Zeiss, Feldbach, Switzerland), or a Nikon Diaphot 300 immunofluorescence microscope (Nikon, Egg, Switzerland), using Openlab 3.1.7. software (Improvision, Coventry, England). Insulin and glucagon staining per islet area was determined from images of insulin- and glucagon-stained slides obtained using ImageJ software at the National Institutes of Health (http://rsb.info.nih.gov/ij/). All statistical analyses were performed using GraphPad software (GraphPad, San Diego, CA, USA).

Cell culture

All cell lines were grown in DMEM supplemented with 10% fetal bovine serum, 2 mmol/l glutamine and 100 U/ml penicillin. Tumor cell lines were established from insulinomas of 12-week-old RT2;1^(fl/fl) and RCre;RT2;1^(fl/fl) mice, as described (Cavallaro et al, 2001). For transplantation of tumor cells, 10⁶ cells in PBS were injected subcutaneously into the two flanks of C57Bl/6 mice, or intravenously into athymic nu/nu mice.

Matrix adhesion assays were performed on collagen IV, as described (Cavallaro et al, 2001). Ninety-six-well plates were coated with 5 g/cm² of mouse collagen IV (BD Biosciences, San Jose, CA, USA). A total of 6 10⁴ cells were seeded per well, and after 90 min non-adherent cells were removed by washing with PBS. Adherent cells were fixed for 20 min with 25% glutaraldehyde (Sigma), stained with crystal violet, washed and air-dried. Bound dye was solubilized with 10% acetic acid and absorbance measured at 595 nm. Cell-free wells served as blanks. The assays were performed in triplicates.

For proliferation assays, 10⁵ cells were seeded per well of a 24-well plate. Every 24 h, 100 l MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (Sigma), 5 mg/ml in PBS) was added, and 90 min later medium was removed, 500 l solubilization buffer (95% isopropanol, 5% formic acid) was added and incubated for 5 min at RT. Absorbance of the solution was determined at 570 nm.

For growth in 3D culture, 5 10⁴ cells were mixed in Matrigel™ (BD Biosciences) and seeded on a layer of solidified Matrigel™. After solidification of the upper cell-containing Matrigel™ layer, normal growth medium was added to the culture.

FACS analysis of tumors

Tumors were dissected out of pancreata, placed in ice-cold PBS and minced into small pieces. After washing with PBS, tumor pieces were incubated, for 30 min at 37°C, with a collagenase mix DMEM, 5% NU-serum (BD Biosciences), 0.16 mg/ml DNase I, 1 mg/ml collagenase D, H and collagenase/dispase (Roche), 0.5 mg/ml collagenase I (Sigma), to obtain single-cell suspensions, followed by filtration and washing with FACS–PBS (1