|Home | About | Journals | Submit | Contact Us | Français|
Integrins regulate cell-cell and cell-matrix adhesion and thereby play critical roles in tumor progression and metastasis. Although work in preclinical models suggests that β1 integrins may stimulate metastasis of a number of cancers, expression of the β1 subunit alone has not been shown to be a useful prognostic indicator in human cancer patients. Here we have demonstrated that the α2β1 integrin suppresses metastasis in a clinically relevant spontaneous mouse model of breast cancer. These data are consistent with previous studies indicating high expression of α2β1 integrin in normal breast epithelium and loss of α2β1 in poorly differentiated breast cancer. They are also consistent with our systematic analysis of microarray databases of human breast and prostate cancer, which revealed that decreased expression of the gene encoding α2 integrin, but not genes encoding α1, α3, or β1 integrin, was predictive of metastatic dissemination and decreased survival. The predictive value of α2 expression persisted within both good-risk and poor-risk cohorts defined by estrogen receptor and lymph node status. Thus, the α2β1 integrin functionally inhibits breast tumor metastasis, and α2 expression may serve as an important biomarker of metastatic potential and patient survival.
Cancer progression involves fundamental alterations of cell-cell and cell-matrix interactions and the cues derived from them that contribute to the invasive and metastatic phenotype. Death from cancer is usually caused by metastatic disease rather than the primary tumor. Tumor metastasis is not a single event but requires a series of steps that allow tumor cells to detach and invade through the basement membrane (1, 2). The detached cells then intravasate into the circulation via blood vessels or lymphatics where they are transported throughout the body, arrest, and survive (3).
Hynes first argued that defects in cell surface adhesive receptors, in addition to altered expression of matrix components in the microenvironment, could lead to the altered cell behavior that typifies cancer progression and metastasis (4–7). As important mediators of cell-adhesive behavior, integrins play a critical role in tumor progression and metastasis. Based on these observations, integrins have become attractive therapeutic targets (8–13). A selective inhibitor targeting the ligand-binding RGD sequence within the αvβ3 and αvβ5 integrins is in phase II and phase III clinical trials for glioblastoma multiforme and lung and breast cancer. These αv integrins are expressed both by some tumor cells and by the tumor-associated neovasculature and therefore have an impact on tumor growth and progression in 2 ways. Not only does the inhibitor suppress tumor growth and invasion, but it inhibits tumor angiogenesis.
Much attention has been devoted to the β1 family of integrins. Early work by Weaver et al. demonstrated that inhibition of β1 integrin ligation produced reversion of the malignant phenotype of a human breast tumor cell line to a normal, growth-arrested phenotype (9). Work using numerous in vitro and in vivo orthotopic or xenograft models indicated that the β1 integrins mediate drug resistance and stimulate metastasis in gastric, ovarian, and lung cancer (14, 15). Similarly, animals with targeted deletion of the β1 gene in mammary epithelium failed to develop cancer, suggesting that the β1 integrins are essential for cancer initiation (16). Strategies to target the β1 integrin subunit for therapy have not progressed rapidly, in part due to the heterogeneity of the β1 integrin family.
The β1 integrin subunit pairs with numerous α subunits to produce heterodimers that mediate adhesion to a wide variety of ligands including matrix molecules (collagen, laminin, fibronectin, etc.) and cell surface proteins such as E-cadherin and vascular cell adhesion molecule (6). Despite the impressive experimental evidence described above demonstrating that collectively the family of β1 integrins is essential for tumor initiation and progression, a well-defined role for any single αβ-heterodimer in tumor initiation and cancer metastasis is not known.
The α2β1 integrin, a receptor for collagen and other matrix molecules, is highly expressed on normal breast epithelium (17). Early studies provided the initial correlative suggestion that loss of α2β1 integrin expression might play an important role in cancer progression (18). Alternatively, loss of α2β1 integrin expression may simply be a consequence of malignant progression. Other studies, however, using cell lines either in vivo or in vitro, suggest that the α2β1 integrin may serve to enhance metastasis to different organs (19–22).
The results of in vitro and in vivo studies of various designs may be subject to conflicting interpretation. Therefore, we have explored the role of the α2β1 integrin in cancer initiation and progression using a clinically relevant, spontaneous mouse model, the mouse mammary tumor virus–Neu (MMTV-Neu) model of breast cancer progression and metastasis. Furthermore, we have validated and extended the conclusions of our model system with a detailed analysis of α2 integrin gene expression and its significance in human breast and prostate cancer. We demonstrate that the α2β1 integrin is a metastasis suppressor of breast cancer, and we identify key mechanistic steps in the metastatic cascade enhanced by loss of α2β1 integrin expression.
To determine the role of the α2β1 integrin in breast cancer initiation, progression, and metastasis, we crossed the α2β1 integrin–deficient mouse with transgenic animals carrying the MMTV-c-erbB2/Neu oncogene to generate WT/MMTV-Neu (WT/Neu) and α2-null/MMTV-Neu (α2-null/Neu) animals (23–25). Tumor latency was significantly reduced in α2-null/Neu mice when compared with WT/Neu animals; however, the difference in latency was quite small (P = 0.02, Figure Figure1A).1A). Tumors were first identified at 22 weeks of age in the α2-null/Neu mice (n = 22) and at 24 weeks of age in the WT/Neu (n = 24). The median time to palpable tumor development was 31 and 33 weeks in α2-null/Neu and WT/Neu mice, respectively. The hazard ratio for developing tumors in the α2-null/Neu animals relative to WT/Neu mice was 2.09 (95% confidence interval 1.1–4.0).
Spontaneous metastasis to the lungs was evaluated when the tumors reached 10% of the body weight of the animal; therefore, tumor metastasis was evaluated when the tumors in α2-null/Neu and WT/Neu mice were of similar size. The number and size of metastatic foci in the lungs were significantly increased in the α2-null/Neu mice (Figure (Figure1B).1B). The metastatic lesions remained within the vasculature of both α2-null/Neu and WT/Neu lungs and did not extravasate, a feature characteristic of the MMTV-Neu mouse model of tumor metastasis. As expected from previous studies, 60% of the WT/Neu mice (n = 10) developed grossly visible or microscopic metastasis. In contrast, 80% of the α2-null/Neu animals (n = 15) developed metastases in the same time frame. Animals lacking the α2β1 integrin developed a significantly greater number of metastatic nodules per animal (P = 0.02), and the metastatic foci were larger in the α2-null/Neu animals compared with the WT/Neu mice (P = 0.02) (Figure (Figure1,1, C and D). These data demonstrate that the α2β1 integrin plays a critical role not only in regulating tumor latency but also in determining metastatic tumor cell dissemination to the lung (26).
The increased rate of metastasis in α2-null/Neu mice was not a result of increased primary tumor growth because the rates of growth of primary tumors spontaneously arising in WT/Neu and α2-null/Neu animals were similar (Figure (Figure2A2A and Supplemental Figure 1; supplemental material available online with this article; doi: 10.1172/JCI42328DS1). Thus, the increased rate of metastasis was not a result of primary tumor growth. The number of tumors per animal also did not differ significantly between the 2 genotypes (Figure (Figure2B).2B). Consistent with the similar growth kinetics of tumors in the WT/Neu and α2-null/Neu mice, histologic and immunohistochemical analyses of the primary tumors demonstrated similar morphology (Figure (Figure2C),2C), comparable rates of proliferation and apoptosis, as revealed by the number of Ki67 (Figure (Figure2,2, D and E) and TUNEL-positive (Figure (Figure2,2, D and F) cells, and insignificant differences in collagen deposition (Supplemental Figure 2A and Supplemental Table 1). In addition, recruitment of B cells, T cells, macrophages, and granulocytes to WT/Neu and α2-null/Neu tumors was similar (Supplemental Figure 2B). As expected, the tumors expressed the α2β1 integrin in WT/Neu mice, but not in α2-null/Neu mice (Figure (Figure2G).2G). Expression of the α1, α4, α5, α6, β3, and β4 integrin subunits by WT/Neu and α2-null/Neu tumor cells was similar (Supplemental Figure 2C and Supplemental Table 2). Together these data demonstrate that the loss of α2β1 integrin increases the rate of breast cancer metastasis without altering tumor growth, collagen deposition, or the immune response in vivo.
Since the in vivo data suggested that α2β1 integrin expression regulates one or more steps in tumor metastasis, we next focused on tumor cell intravasation. To determine the ability of tumor cells from WT/Neu and α2-null/Neu to intravasate, blood was collected by cardiac puncture at the time of sacrifice. The colony-forming ability of tumor cells in the cellular fraction of 1 ml of blood was analyzed ex vivo (Figure (Figure3,3, A and B) (27). As shown in Figure Figure3,3, A and B, the number of tumor colonies generated from the blood of α2-null/Neu mice (18+/– colonies) was significantly greater than from the blood of WT/Neu mice (5+/– colonies) (P = 0.002). The increased number of viable, circulating tumor cells correlated with the number of lung metastases.
Increased colony growth ex vivo may result from increased numbers of viable, circulating tumor cells or from an increased ability of isolated tumor cells to proliferate ex vivo. To differentiate between these possibilities, we determined the relative number of circulating, Neu mRNA–positive cells in the blood by quantitative real-time PCR (qRT-PCR). Figure Figure3C3C shows the ratio of circulating Neu-expressing cells relative to GAPDH in the blood at the time of sacrifice when the primary tumors in the WT/Neu and α2-null/Neu mice were of similar size; α2-null/Neu mice had a significantly larger proportion of these cells than WT/Neu mice (P = 0.005). Since c-ErbB2/Her2 is expressed in normal hematopoietic cells, the number of circulating tumor cells was also characterized by qRT-PCR for cytokeratin 19 (28). As shown in Supplemental Figure 3, increased numbers of circulating, cytokeratin 19–expressing cells were identified in the peripheral blood of tumor-bearing, α2-null/Neu in comparison with WT/Neu mice at the time of sacrifice. These results imply that the greater number of tumor colonies arising from blood of α2-null/Neu mice was due to increased numbers of circulating tumor cells in these mice. At the time of sacrifice, when the tumors were of similar size, tumor cell dissemination was significantly increased in mice lacking the α2β1 integrin.
We hypothesized that differences in tumor cell dissemination occurred not only at the time of sacrifice but also at earlier time points in α2-null/Neu mice. To determine the time course of tumor cell dissemination, the number of circulating tumor cells was quantitated at approximately 10 to 11 weeks of age and every other week thereafter. There was no difference between the genotypes in the proportion of Neu-positive cells between 10 and 25 weeks of age, a time prior to the development of palpable tumors (data not shown). However, between 26 and 30 weeks of age, the time that tumors first became palpable in both the WT/Neu and α2-null/Neu mice, the number of circulating Neu-positive tumor cells was significantly increased in α2-null/Neu mice relative to WT/Neu mice (Figure (Figure3D)3D) (P = 0.002). These results demonstrate that not only was tumor cell dissemination significantly enhanced in the absence of the α2β1 integrin, but that tumor cell intravasation occurred at an earlier time point in α2-null/Neu mice. Although we cannot exclude a contribution of enhanced tumor cell survival, the data support the hypothesis that enhanced metastasis in the α2-null/Neu mice is, in part, due to increased tumor cell intravasation.
The level of circulating Neu-positive tumor cells was elevated in both WT/Neu and α2-null/Neu mice. The large number of Neu-expressing cells in the circulation of both young WT/Neu and α2-null/Neu mice was unexpected. Therefore, the level of detectable circulating Neu-expressing cells in nontransgenic littermates was determined. In comparison with WT/Neu mice, there was no detectable Neu mRNA in the circulation of nontransgenic animals, an important negative control (Figure (Figure3E). 3E).
Since the α2β1 integrin is expressed on activated endothelial cells and lack of the α2β1 integrin can alter tumor angiogenesis, enhanced intravasation could be due to changes in the host tumor vessels. To examine this possibility, expression of CD31-positive vessels in WT/Neu and α2-null/Neu tumors was evaluated. As shown in Figure Figure3F3F and Supplemental Figure 4, there was no difference in vascular density between the 2 subgroups.
To further evaluate the contribution of enhanced tumor cell intravasation to the metastatic phenotype observed in vivo, we employed the classic experimental metastasis model, tail vein injection. This model examines the later stages of the metastatic cascade. Primary neoplastic cells were isolated from the tumors of WT/Neu and α2-null/Neu mice and reinjected i.v. into the tail veins of either WT/Neu or α2-null/Neu animals, thereby bypassing the intravasation step. The lungs were evaluated histologically and the number of metastatic foci within the lungs was quantitated. As shown in Figure Figure4,4, A and B, there was no difference in the percentage of WT animals that developed metastases (P = 0.34) and no difference in the number of metastases (P = 0.35) per animal when WT mice were injected with WT/Neu or α2-null/Neu tumor cells. Similarly, when the α2-null animals were used as the host, there was also no difference in the percentage of α2-null animals that developed metastases (P = 0.72) or number of metastatic foci (P = 0.85) (Figure (Figure4,4, C and D). These data from the i.v. tail vein injection model indicate that the later stages of the metastatic cascade are not altered by tumor cell expression of the α2β1 integrin and that there is no difference in the ability of the α2-null/Neu cells to colonize the lung. Extravasation into the lung is not evaluable in the MMTV-Neu mouse model. These data provide additional support that the α2β1 integrin regulates tumor metastasis at the step of tumor cell intravasation.
Intravasation requires tumor cells to migrate and invade through the basement membrane and into the vasculature. To compare the malignant behavior of α2-null/Neu and WT/Neu tumor cells, primary tumor cells were isolated. In monolayer culture, primary WT/Neu tumor cells exhibited an epithelial morphology with cobblestone-appearing colonies and cell-cell junctions. In contrast, the α2-null/Neu tumor cells were mesenchymal in shape and failed to cluster into epithelioid colonies (Figure (Figure5A).5A). To determine whether cells lacking α2β1 integrin expression were more motile, migration of WT/Neu cells and α2-null/Neu cell was evaluated in an in vitro wound closure assay. The α2-null/Neu tumor cells migrated more rapidly than the WT/Neu cells (Figure (Figure5,5, B and C, and Supplemental Figure 5). The mesenchymal morphology of the α2-null/Neu cells was apparent in migrating cells.
Morphogenesis of primary, WT/Neu, and α2-null/Neu tumor cells was evaluated in 3D gels of reconstituted basement membrane (Matrigel). In this system, nontransformed and Neu-transformed mammary epithelial cells undergo rapid reorganization into acinar structures that recapitulate mammary morphogenesis in vivo (29). Primary tumor cells from WT/Neu animals formed typical acinar structures with small, well-circumscribed cell aggregates, as previously described by several investigators. In contrast, the tumor cells from the α2-null/Neu mice formed large, disorganized, and irregular-shaped acini with protrusions that extended into the surrounding matrix. Such protrusions were entirely lacking from WT/Neu acini (Figure (Figure5D). 5D).
The α2-null/Neu acini also appeared slightly larger than the WT/Neu acini. Malignant potential is associated with enhanced proliferation under conditions of both anchorage dependence and anchorage independence. Proliferation of WT/Neu and α2-null/Neu tumor cells was not different when tumor cells were plated on plastic in either serum-free or complete media (Supplemental Figure 6). The α2-null/Neu tumor cells demonstrated a decreased proliferative potential when plated on type I collagen.
To assess anchorage-independent growth, the ability of primary WT/Neu and α2-null/Neu tumor cells to form colonies in soft agar was determined. After 3 weeks in culture, the number of colonies equal to or larger than 50 mm was enumerated. Tumor cells from α2-null/Neu mice generated significantly greater numbers of colonies than tumor cells isolated from WT/Neu mice (Figure (Figure5E).5E). Therefore, although there was no difference in tumor growth in vivo (Figure (Figure2),2), tumor cells lacking the α2β1 integrin demonstrate enhanced anchorage-independent, but not anchorage-dependent, growth in vitro.
In order to disseminate into the blood, tumor cells must transit across the endothelial cell barrier. We compared the ability of α2-null/Neu and WT/Neu tumor cells to intravasate in vitro by migrating across a barrier of WT endothelial cells in a Transwell chamber. The α2-null/Neu cells more effectively intravasated through the endothelial cell barrier than WT/Neu cells. The number of α2-null/Neu cells that successfully intravasated through the endothelial cell barrier was significantly increased in comparison with the number of WT/Neu tumor cells (P = 0.0003) (Figure (Figure5,5, F and G). Therefore, lack of α2β1 integrin expression may contribute to the metastatic phenotype by potentially 2 separate mechanisms. Although the α2-null/Neu tumors did not demonstrate a growth advantage in vivo, enhanced anchorage-independent growth in vitro raises the possibility that the tumor cells survive in the circulation or once lodged in the lung vasculature, leading to enhanced metastasis. In contrast, increased α2-null/Neu tumor cell intravasation both in vitro and in vivo is well supported by the data.
To validate our findings from the murine breast cancer model and to extend the relevance to human breast cancer, publicly accessible microarray data were analyzed for a correlation between α2 integrin gene expression and clinically relevant parameters. Expression of the α2 integrin was significantly reduced in breast cancer versus normal breast tissue in the Richardson cohort (P = 0.038; Figure Figure6A)6A) (30). Not only was the loss of α2 integrin gene expression highly associated with malignant disease, the decrement in α2 integrin expression consistently correlated with disease progression in 2 different public data sets for which stage-specific data were available (P ≤ 0.01) (31, 32). All data sets are shown in Supplemental Table 3.
To determine whether the loss of α2 integrin expression was an indicator of prognosis, the correlation between α2 integrin expression, metastasis, and patient survival was analyzed in publicly accessible breast cancer cohort NKI-295 organized by van de Vijver (33). The expression of α2 integrin was greatly reduced in the patients with distant metastasis when compared with patients without distant metastases (P < 0.0084; Figure Figure6B).6B). To further evaluate its predictive value, the correlation between α2 integrin expression, metastasis-free survival, and overall survival was determined using log-rank analysis. The expression of α2 integrin correlated significantly with improved metastasis-free survival (P = 0.0022) and overall patient survival (P < 0.0001) as depicted in Kaplan-Meier plots of 295 patients divided evenly according to high-versus-low α2 integrin expression (Figure (Figure6,6, C and D). Further subdivision of the cohort into terciles demonstrated that increasing α2 integrin expression is strongly associated with improved patient survival (log-rank test for trend P = 0.0008; Supplemental Figure 7A). In contrast, expression of the integrin subunits α1, α3, and β1 did not correlate with survival (Figure (Figure6,6, E, F, and G; P = 0.2639, 0.9509, and 0.5, respectively).
Both lymph node–negative and lymph node–positive patients were included in the NKI-295 cohort. In accordance to the standard of care, the majority of node-positive patients had received systemic adjuvant treatment while node-negative patients had not. These 2 classes of patients were analyzed separately to evaluate any correlation with α2 integrin expression. Elevated expression of the integrin subunit continued to correlate strongly with improved survival in both node-positive and node-negative patients (Figure (Figure7,7, A and B; P = 0.0049 and 0.0084 respectively).
The ability of α2 integrin expression to predict distant metastasis was further assessed in the context of estrogen receptor (ER) status. ER status has been particularly important in the treatment of breast cancer because ER-negative status corresponds with poor patient prognosis (34–36). In 15 independent studies in which ER status was available, decreased α2 integrin expression correlated with decreased ER expression. The individual studies analyzed are outlined in Supplemental Table 3. A detailed survival analysis was performed on the NKI-295 patients. As shown in Figure Figure7,7, C and D, high and low α2 integrin expression subdivided both the ER-positive patients and the high risk, ER-negative patients into prognostically distinct categories. High α2 integrin expression was significantly associated with survival in the ER-positive patient cohort (P = 0.001). A similar difference in estimated survival between high and low α2 expression was also observed in ER-negative patients. However, this difference did not achieve statistical significance (P = 0.15) due to the modest number of ER-negative women in these studies. In patients with known metastatic disease at presentation, survival was also shortened in patients with decreased α2 integrin expression (P = 0.004; Figure Figure7E). 7E).
To validate the clinical findings from the NKI-295 cohort, the correlation between α2 integrin expression and improved clinical performance of the patients was confirmed in a separate clinical investigation of 286 lymph node–negative patients with invasive breast carcinoma in the Wang cohort (37). Disease-free survival in this lymph node–negative cohort was significantly decreased in patients with reduced α2 integrin expression (P = 0.0317; Figure Figure7F).7F). In addition, among all patients with metastatic disease, disease relapse involving brain metastasis occurred only in patients with reduced α2 integrin expression (P = 0.0049; Supplemental Figure 7B). Overall survival information was not available for this study.
The strong clinical association between loss of α2 integrin expression and breast cancer progression and metastasis suggested that α2 integrin expression may be a metastasis suppressor in other models of cancer. To assess this possibility, we evaluated α2 integrin expression in normal prostate, prostate intraepithelial neoplasia (PIN), prostate cancer, and prostate cancer metastasis. As shown in Figure Figure8,8, A and B, the α2 integrin was expressed at high levels in normal prostate (38). The level of α2 integrin gene expression was diminished in PIN and prostate cancer and was more greatly decreased in metastatic prostate cancer. These findings in a second model of human cancer further validate the role of the α2 integrin as a metastasis suppressor in many subtypes of human cancer.
The studies reported in this manuscript identify the α2β1 integrin as the first integrin known to serve as a metastasis suppressor in breast cancer and prostate cancer. We found this to be the case in both a spontaneous, clinically relevant mouse model of breast cancer and in human breast cancer. In MMTV-Neu transgenic mice, lack of α2β1 integrin expression resulted in modestly decreased mammary tumor latency and markedly increased cancer metastasis. Deletion of the α2β1 integrin enhanced tumor cell intravasation both in vivo and in vitro. The increased tumor cell dissemination detected in the peripheral blood of the α2-null/Neu, but not the WT/Neu mice, at both early time points when the tumors were just barely palpable and later when the primary tumors were of comparable size at sacrifice suggests that intravasation is a key step in the metastatic cascade where loss of integrin expression acts to enhance metastasis. Loss of the α2β1 integrin does not affect tumor growth or tumor cell proliferation in vivo or in vitro. However, in vitro, the α2-null/Neu tumor cells demonstrate enhanced anchorage-independent growth. We therefore propose that enhanced anchorage-independent growth and the slightly shorter latency period in mice lacking the α2 subunit contribute to the metastatic phenotype. It was surprising that a difference in metastasis to the lung was not observed in the classic experimental metastasis model. Although we did not see a difference between the ability of WT and α2-null tumor cells to colonize the lung, there was a slight trend in favor of colonization by the α2-null tumor cells. However, our sample size was too small to show that this difference was significant. The primary tumor cells when reinjected were not highly metastatic, and increasing the number of injected cells was limited by animal viability.
The α2β1 integrin was globally deleted in all cells of the α2-null mouse, not just in the mammary gland. Therefore, although our in vitro and in vivo data strongly support the important role that α2β1 integrin expression by the tumor cells plays in tumor metastasis, the possibility exists that α2β1 integrin expression by cells of the tumor microenvironment, i.e., endothelial cells, fibroblasts, or inflammatory cells, and not just by the tumor cells themselves, also plays a role in inhibiting tumor metastasis in vivo. Although we cannot exclude the role of the α2β1 integrin in the tumor microenvironment, our studies define the critically important role that α2β1 integrin expression by the tumor cells plays in controlling tumor cell metastasis. However, the biologic studies and the human expression data focus on integrin expression or lack of integrin expression by the mammary epithelial cells. All in vitro experiments, including cell migration, epithelial morphogenesis, and tumor cell intravasation, solely define the importance of integrin expression by the tumor cells to the malignant phenotype. In addition, the in vivo classic metastasis model using primary tumor cells also supports the role of the integrin in modulating tumor cell intravasation. Finally, the extensive microarray analysis further supports the significance of α2β1 integrin loss from the malignant tumor in tumor cell progression, metastasis, and patient outcome.
Previous studies by White et al. demonstrate that animals lacking the β1 integrin due to targeted deletion of the gene in the mammary epithelium fail to develop cancer (14, 16). Since the β1 integrin subunit forms heterodimers with at least 11 α subunits that mediate adhesion to multiple different matrix molecules and other cell adhesion receptors, it is not surprising that the suppressive phenotype of any individual heterodimer may be masked. Our finding that decreased α2 integrin gene expression is associated with increased metastatic disease and decreased patient survival does not exclude additional specific roles for other heterodimers. However, expression of the α1, α3, or β1 integrin genes did not correlate with metastasis or patient survival in the same cohort of breast cancer patients. Since integrins have become attractive targets for adjuvant therapy, understanding the influence of each specific heterodimer is critical to understanding the potential ramification of such therapy (14). Indeed, the biologic consequences of genetic deletion versus inhibition of a specific integrin leave a number of open questions regarding the applicability of anti–β1 integrin therapy in breast cancer. More selective agents for specific heterodimers deemed biologically relevant in human disease may have greater efficacy.
In patients with breast cancer, α2 integrin expression correlated with ER positivity. In both the ER-positive patient subgroup and the ER-negative cohort, α2 integrin expression was associated with a survival advantage. Therefore, α2 integrin expression is not simply a surrogate marker of ER expression. Although ER-negative breast cancer accounts for only 15%–20% of cases (34), this subgroup is very important clinically because these patients have a poorer prognosis and there are fewer therapeutic targets. Expression of the α2 integrin subunit may serve as an important biomarker and clinical determinant for future clinical trials in ER-negative breast cancer patients by identifying the subset likely to benefit from novel or more aggressive therapies.
The identification of the α2β1 integrin as a metastasis suppressor opens up a large number of exciting and unanswered questions for future basic and clinical studies. It remains to be determined whether occupancy of the integrin with ligand is required for the metastasis suppressor effect. The potential role for the collagenous matrix in regulating tumor progression is an exciting avenue of current investigation (39–41). The mechanism by which α2 integrin subunit gene expression is diminished during oncogenesis and tumor progression is unknown. The classic metastasis suppressor genes are lost due to genetic and epigenetic changes (3, 42–44). Loss of metastasis suppressors augments tumor cell procession through the metastatic cascade. In the mouse model, loss of α2β1 integrin expression was accomplished by targeted gene deletion. In human disease, the mechanisms by which α2 integrin subunit gene expression is decreased in cancer are unknown. The human α2 integrin gene lies on chromosome 5q23, a site of frequent deletions in breast cancer (45). However, deletion of the α2 integrin gene in cancer has not been reported. Genetic polymorphisms in the α2 gene that potentially alter expression have been associated with prostate cancer susceptibility (46). Epigenetic modification of the α2 gene promoter and regulation by microRNA are possibilities, since there is a progressive loss of expression in humans as tumor cells become less differentiated. Since reexpression of the α2 subunit in patients with cancer may be beneficial, defining the mechanisms of α2 gene silencing may lead to novel therapeutic strategies to restore gene expression in tumor cells.
Hunter and colleagues have suggested that constitutional genetic heterogeneity modulates tumor efficiency and determines those individuals at increased risk of metastasis once a primary cancer arises (47). It is well recognized that genetic polymorphisms within the α2 gene regulate α2β1 integrin expression by platelets and perhaps other cell types (reviewed in ref. 48). The variable number of α2β1 integrin molecules in different individuals correlates with the ability of platelets to adhere to type I collagen (49–51). Individuals with low baseline levels of cellular α2β1 integrin expression due to allelic differences in the α2 integrin gene may be at increased risk for metastatic disease or earlier age of tumor onset.
In conclusion, we demonstrate that the α2β1 integrin, which serves as a receptor for collagens, laminins, and other matrix and nonmatrix proteins, functions as a metastasis suppressor in part by regulating tumor cell intravasation. The impact of deleting the α2 integrin subunit, hence α2β1 integrin heterodimer expression in cancer biology, is distinctly different and in some ways counter to the consequences of deleting the β1 subunit and thus all the heterodimers associated with the β1 integrins. We provide both experimental evidence and analysis of gene-profiling data that establish the contribution of the α2β1 integrin in regulating tumor progression and the metastatic phenotype. Future studies will evaluate the genetic and epigenetic modifications of the gene that lead to loss of integrin expression in breast cancer. In addition, it will be critical to determine whether similar changes in α2β1 integrin expression are seen in subtypes of cancer other than breast and prostate and whether they are predictive of metastasis and/or poor prognosis in other tumor types.
Animals were housed in pathogen-free conditions at Vanderbilt University Medical Center in compliance with IACUC regulations. All animal studies were approved by the Vanderbilt Institution Animal Care and Use Committee at Vanderbilt Medical Center. The α2β1 integrin–deficient mice (52) were backcrossed 10 generations onto the FVB/N background to obtain animals that were 99% genetically FVB/N and crossed with transgenic mice carrying the Neu (c-neu) gene under the transcriptional control of the MMTV promoter, obtained from Jackson Laboratories (24, 25) to generate WT/Neu and α2-null/Neu mice. Mice positive for the transgene were identified by PCR following Jackson Laboratories’ recommendations. Female mice were examined after birth for tumor development by palpation. Once detected, primary tumors were measured weekly with digital calipers. Total tumor volume was calculated using the equation: volume = a × b2 × 0.52, where a is the longest dimension and b is the shortest. Blood was collected every 2 weeks after 6 weeks of age by saphenous vein or at the time of sacrifice by cardiac puncture. Animals were sacrificed when total tumor volume reached 10% of their body weight. The lungs were perfused with formalin. Blood, tumor tissue, and lung were collected. Tumor nodules in lung were quantitated.
Tumor and lung morphology were evaluated on formalin-fixed, paraffin-embedded, H&E-stained sections. Immunohistochemical identification of Ki67, TUNEL, type IV collagen, B220, CD3, and F4/80 was performed on paraffin-embedded tissues with a biotin-conjugated secondary antibody and diaminobenzidine substrate (Vectastain ABC Kit; Vector Laboratories). The number of Ki67- and TUNEL-positive cells was determined quantitatively using the Metamorph Imaging System (Molecular Devices). Primary antibodies included CD45/B220 (BD Biosciences), Ki67 and collagen IV (Abcam), TUNEL assay (R&D Systems), CD3 (Santa Cruz Biotechnology Inc.), and F4/80 (Novus Biologicals). Immunofluorescence analysis was conducted on 7-μm frozen sections with the primary antibodies, anti-mouse CD31, anti–α2 integrin subunit (BD Biosciences — Pharmingen), and secondary antibodies Alexa Fluor 594– or Alexa Fluor 488–conjugated IgG and DAPI, all from Molecular Probes. Quantification of the CD31+ vessels was carried out using the Scion Image system. The number of CD31+ structures in 5 high-power fields/tumor with 5 animals in each group were elucidated. Tumor cells were also analyzed by flow cytometry using the following antibodies (all from BD Biosciences): hamster anti-rat/mouse CD49a, APC anti-mouse CD49b, FITC anti-mouse CD49d, PE anti-mouse CD49e, and anti-mouse CD104. Additional antibodies included anti-mouse CD29 and PE anti-mouse CD61 from Pharmingen.
Primary WT/Neu and α2-null/Neu tumor cells were isolated, as previously described (53), with modifications. Briefly, primary tumors were dissected under sterile conditions and incubated overnight at 4°C in growth media (DMEM/F12, 10% FBS, 20 ng/ml EGF, 0.5 μg/ml hydrocortisone, 0.1 μg/ml cholera toxin, 10 μg/ml insulin, 1× Pen/Strep solution) supplemented with 2 mg/ml collagenase type I, 100 units/ml hyaluronidase, and 50 μg/ml gentamicin. Minced tumors were digested at 37°C for 2 hours. After digestion, cells were centrifuged for 10 minutes at 500 g, followed by differential centrifugation (5 times for 45 seconds at 500 g). Cells were cultured in serum-free growth media and incubated at 37°C, 5% CO2 for 3–4 days. The subsequent generations were cultured in growth media with 5% FBS. Primary tumor cells in 3D Matrigel cultures were established as previously published (29). For 3D gel culture, 10,000 cells were overlain on growth factor–reduced Matrigel (BD Biosciences) with complete growth media containing 2% Matrigel and incubated at 37°C.
Circulating tumor cells were cultured as described by Biswas et al. (27). 500 μl of blood supplemented with DMEM, 10% FBS, was added onto 6-well plates coated with growth factor-–reduced Matrigel and incubated at 37°C, 5% CO2. The next day, wells were washed with lysis buffer (4.15 g NH4Cl, 0.5 g NaHCO3, 0.0186 g disodium EDTA in 200 ml water) and PBS. At 12 days, colonies of 50 μm or greater were counted.
WT/Neu or α2-null/Neu primary tumor cells were cultured in 24-well plates with complete growth media until confluence. The cell monolayers were scratched with a P200 pipette tip and incubated with serum-free media at 37°C. Wound closure was quantitated by digital imaging using T-Scratch (54).
Intravasation assays were carried out with WT primary pulmonary endothelial cells, prepared as previously described (55). Endothelial cells (5 × 104) were seeded on the bottom side of a Transwell filter, precoated with growth factor–reduced Matrigel (1:20), and cultured in growth medium overnight. On day 2, primary WT/Neu or α2-null/Neu tumor cells were labeled with CellTracker Red CMTPX (Invitrogen), and the endothelial cells were labeled with CellTracer Vybrant CFDA SE. Both tumor and endothelial cells were starved in DMEM medium with 1% FBS overnight. On day 3, the endothelial cells were stimulated with TNF-α (20 ng/ml) for 2 hours and then washed in DMEM medium with 1% FBS. WT/Neu or α2-null/Neu primary tumor cells (5 × 104) in DMEM plus 1% FBS were placed on the upper membrane of the Transwell filter. DMEM with 10% FBS was placed in the bottom well. After 8 hours, all remaining tumor cells in the upper chamber of Transwell were removed. Endothelial cells and intravasated tumor cells on the lower surface of the Transwell were fixed with 4% paraformaldehyde. The membrane was analyzed and the number of intravasated tumor cells was quantitated using a Nikon Eclipse 80i fluorescence microscope.
For evaluation of anchorage-independent growth, 10,000 primary tumor cells were plated in 1 ml growth media with 0.4% agarose in 35-mm petri dishes coated with 0.8% agarose. Plates were incubated at 37°C and colonies 50 μm or greater were counted at 21 days of culture. All assays were carried out in triplicate.
Total RNA was extracted from blood using the RiboPure-Blood Kit (Ambion) following the manufacturer’s instructions. cDNA was synthesized from 1 μg of total RNA using iScript cDNA Synthesis kit (Bio-Rad). qRT-PCR analysis was performed on an iCycler system (Bio-Rad) using mouse Neu (sense, CAGTGTCTTCCAGAACCTTCG; anti-sense, CGGTGAATGAGAGCCAATCC) or cytokeratin 19 (sense, CCCAGCTCAGCATGAAAGC; anti-sense, CCTGGTTCTGGCGCTCTATG) cDNA primers. Number of circulating tumor cells expressing Neu or cytokeratin 19 mRNA was quantitated by external standard curve. Serial dilutions of blood containing known number of tumor cells were carried out in duplicate in a decreasing range of 1 × 107 to 1 to generate the standard curve. Samples were normalized using GAPDH.
Total RNA was isolated from primary tumors, reverse transcribed using oligo-dT primers, and analyzed by qRT-PCR for gene expression of the following genes: collagen I, α1 (sense, GTATCACCAAACTCAGAAG; anti-sense, ATCATAGCCATAGGACATC) and α2 (sense, AGTATTCTGACAAAGGAGTTTC; anti-sense, CCAGGAAGTCCAGGAGTC) chains, collagen IV (sense, AAAGGGTCGGTGGGAGAG; anti-sense, CGGGAAAGCCAGGGTTAC), and integrin α1 (sense, CACCTTTCAAACTGAGCCCGCCA; anti-sense, GCTGCCCAGCGATGTAGAGCACAT), α5 (sense, TGCTACCTCTCCACAGAAAAC; anti-sense, GCCAGTCTTGGTGAACTCAG), α6 (sense, CTGGCTTCCTCGTTTGGCTATG; anti-sense, TGCCTTGCTGGTTAATGTAGACGT), αv (sense, TTGGGGACGACAACCCTCTGACAC; anti-sense, TGCGGCGGGATAGAAACGATGAG), β1 (sense, TTCAGACTTCCGCATTGGCTTTGG; anti-sense, TGGGCTGGTGCAGTTTTGTTCAC), β3 (sense, CACGGATGCCAAGACCCATATTG; anti-sense, GTGGAGGCAGAGTAGTGGYYGTC), and β4 (sense, CGTCCCACAGACAGACATGAG; anti-sense, GCGTTCTCCTTGCAGTTTGTTC) subunits. Expression of each gene was normalized to the expression of GAPDH.
Eight- to twelve-week-old female WT or α2-null FVB mice received i.v. tail vein injections of either 5 × 105 WT/Neu or α2-null/Neu primary tumor cells. The animals were sacrificed at 120 days after tumor cell injections, and their lungs were fixed in Bouin solution. After 24 hours of fixation, lungs were transferred to 70% ethanol and tumors visible on the lung surface were counted. In addition to gross examination, lungs were subjected to H&E staining to confirm the presence of metastatic tumors.
Survival for both mice and women are depicted with Kaplan-Meier curves, and log-rank tests were used to compare both relapse-free and overall survival in groups defined by genotype or α2 integrin expression levels (56). Proportional hazards regression analysis was used to derive the hazard rate ratio for α2-null/Neu mice relative to WT/Neu mice (56). Tests of continuous response measures in mice grouped by genotype were performed using the Wilcoxon rank sum test (56). A response feature approach was taken for the analysis of longitudinal data (56, 57). Differences in the rate of growth of tumor volume by genotype were assessed by fitting separate linear regression curves to the volume-age plots for each animal. This provided a separate slope estimate for each animal; these slopes were compared in mice with different genotypes using the Wilcoxon rank-sum test. A similar approach was used to analyze the wound closer assay data. In these analyses, linear regression lines through the origin were fit to the extent of wound closure for the cells from each animal as a function of time since the start of the assay. The difference in these slopes for mice with WT/Neu and α2-null/Neu was assessed using a Wilcoxon rank-sum test. Graphs of longitudinal data were depicted by spaghetti plots in which observations from the same animal are connected by straight lines. Longitudinal measures of mRNA expression were analyzed by first averaging the expression from each mouse over time and then performing Wilcoxon rank-sum tests on these averages to compare expression in the WT/Neu and α2-null/Neu genotypes. Statistical analyses were performed using GraphPad Prism Software and Stata (58). All P values were derived with respect to 2-sides alternative hypotheses.
The NKI-295 cohort published by van de Vijver et al. and the Wang cohort were analyzed for patient survival and disease-free survival, respectively (33, 37). Expression and clinical data for the NKI-205 cohort and the Wang cohort were obtained from Rosetta InPharmatics (33) and NCBI Gene Expression Omnibus (GEO datasets GSE2034), respectively. In metastasis-free survival analyses, patients who died of causes other than breast cancer were censored at the time of their death. In the original studies, gene expression analysis was performed on RNA obtained from frozen samples taken within 1 hour of the initial surgery. The level of α2 integrin expression was set as a dichotomous variable with a “high” α2 integrin expression level defined to be twice the average expression level for all patients in the cohort. All other patients were classified as having “low” expression levels. Trend analysis was performed by dividing the cohort into terciles. Additional gene expression profiling studies analyzed for a correlation among α2 integrin expression, tumor grade, ER status, and metastasis (Supplemental Table 3) were obtained from Oncomine (www.oncomine.org) and Gene Expression Omnibus (GEO) at the National Center for Biotechnology Information (NCBI). Statistical analyses including unpaired Student’s t test, ANOVA, log-rank test, and Kaplan-Meier were performed using GraphPad Prism Software.
Gene expression profiling studies analyzed for the prostate cancer data were obtained from the Stanford Microarray Database or NCBI GEO datasets (Lapointe, ref. 59; Tomlins, ref. 38). Prostate patients in the Lapointe and Tomlins datasets (n = 109 and 87 respectively) were categorized according to their clinical diagnosis and analyzed by ANOVA with post-test analysis for linear trend.
We thank Samuel A. Santoro, Harold Moses, Jin Chen, and Vivien Siegel for advice and encouragement in many helpful discussions, and Jean McClure for excellent photographic and secretarial assistance. This work was supported by grants CA115984, CA098027, CA70275, and CA098131-07 from the NIH.
Conflict of interest: The authors have declared that no conflict of interest exists.
Citation for this article: J Clin Invest. 2011;121(1):226–237. doi:10.1172/JCI42328.