Luminal breast cancers are typically estrogen receptor–positive and generally have the best prognosis. However, a subset of luminal tumors, namely luminal B cancers, frequently metastasize and recur. Unfortunately, the causal events that drive their progression are unknown, and therefore it is difficult to identify individuals who are likely to relapse and should receive escalated treatment. Here, we identify a bifunctional RasGAP tumor suppressor whose expression is lost in almost 50% of luminal B tumors. Moreover, we show that two RasGAP genes are concomitantly suppressed in the most aggressive luminal malignancies. Importantly, these genes cooperatively regulate two major oncogenic pathways, RAS and NF-κB, through distinct domains, and when inactivated drive the metastasis of luminal tumors in vivo. Finally, although the cooperative effects on RAS drive invasion, NF-κB activation triggers epithelial-to-mesenchymal transition and is required for metastasis. Collectively, these studies reveal important mechanistic insight into the pathogenesis of luminal B tumors and provide functionally relevant prognostic biomarkers that may guide treatment decisions.

Significance: The lack of insight into mechanisms that underlie the aggressive behavior of luminal B breast cancers impairs treatment decisions and therapeutic advances. Here, we show that two RasGAP tumor suppressors are concomitantly suppressed in aggressive luminal B tumors and demonstrate that they drive metastasis by activating RAS and NF-κB. Cancer Discov; 7(2); 202–17. ©2016 AACR.

See related commentary by Sears and Gray, p. 131.

This article is highlighted in the In This Issue feature, p. 115

Breast cancer is one of the most common malignancies in women worldwide and the second leading cause of cancer deaths in women in the United States (1). In the clinic, breast tumors are primarily defined by hormone receptor and HER2 status (2). However, transcriptional profiling studies demonstrate that breast tumors can be further classified into subtypes that include luminal cancers (A and B), basal-like, claudin-low, HER2/ERBB2 overexpressing, and normal-like (3–5). Luminal tumors are typically estrogen receptor–positive (ER+) and represent the majority of breast cancers. Notably, although the luminal A subtype of tumors respond well to endocrine therapy, luminal B tumors exhibit a higher proliferative index, are less responsive, and are more likely to become metastatic (6, 7). In fact, the relapse risk rate for luminal B tumors is similar to that of basal and HER2/ERBB2-amplified tumors (8). Genomic efforts have identified many genetic, epigenetic, and transcriptional differences between these two luminal subtypes; however, little is known about the functional events that underlie the aggressive behavior of luminal B tumors (9–11). A better mechanistic understanding of the signals that drive their progression not only would help identify individuals who could benefit from up-front additional adjuvant treatment, but might also provide insight into new therapeutic strategies.

The RAS pathway is one of the most frequently deregulated pathways in human cancer. Activating mutations in RAS genes occur in nearly 30% of all human tumors (12); however, RAS mutations are conspicuously rare in breast cancer, and current data from The Cancer Genome Atlas (TCGA) indicate that RAS is mutated in only 1% of all breast lesions (13, 14). Nevertheless, the RAS pathway is aberrantly activated in more than 50% of human breast tumors and has been suggested to contribute to progression and recurrence (15–17). These observations raise the question of how the RAS pathway becomes hyperactivated in this tumor type.

RAS GTPase activating proteins (RasGAP) promote the hydrolysis of RAS-GTP to RAS-GDP, thereby turning off RAS (18). Accordingly, the inactivation or loss of a RasGAP leads to enhanced RAS-GTP levels, thus providing an alternative means of activating RAS. The NF1 gene is an example of a RasGAP tumor suppressor that is inactivated in a variety of sporadic cancers (reviewed in ref. 19). Here, we show that expression of a unique bifunctional RasGAP tumor suppressor, one that serves as a signaling scaffold to coordinately suppress both RAS and NF-κB, is selectively and frequently lost in luminal B tumors (20). However, we also find that the most aggressive luminal cancers lack expression of two distinct RasGAP genes, which cooperatively trigger the activation of multiple RAS isoforms as well as NF-κB. Collectively, the studies presented here reveal the mechanism by which two major oncogenic pathways are activated and cooperate to drive the progression of luminal B breast cancers, a finding that has important prognostic and therapeutic implications.

Expression of the RasGAP Gene DAB2IP Is Selectively Lost in Luminal B Breast Cancers

Despite the low incidence of oncogenic RAS mutations, the RAS/ERK pathway is hyperactivated in more than half of human breast cancers (16). Interestingly, the RasGAP gene, RASAL2, has been shown to function as a tumor suppressor and is lost in a subset of breast tumors (21). However, given that there are 14 RasGAPs in the human genome, we speculated that other RasGAP genes might also contribute to breast cancer development. Because RasGAPs are frequently inactivated by epigenetic mechanisms in cancer (19), we first evaluated mRNA expression levels of RasGAP genes across breast cancer subtypes. Strikingly, we found that DAB2IP, a RasGAP tumor suppressor that drives prostate cancer progression (20), was selectively suppressed in luminal B breast tumors (Fig. 1A and B). Importantly, a CpG island in the DAB2IP promoter was frequently hypermethylated in the luminal B subtype (Fig. 1C, chr9: 123,331,678), and methylation at this site inversely correlated with mRNA expression levels (Fig. 1D). Moreover, we found that low DAB2IP expression was associated with a significant decrease in relapse-free survival in patients with luminal B tumors (Fig. 1E, log rank P = 0.0028), raising the possibility that loss of DAB2IP expression might contribute to the progression of this breast cancer subtype.

To complement expression analyses, the mutational status of DAB2IP was also examined. Numerous mutations predicted to be deleterious were identified in DAB2IP across several tumor types, suggesting that DAB2IP might function as a broad human tumor suppressor (Fig. 1F). A subset of these alterations were present in breast cancers, including damaging mutations in the catalytic RasGAP domain and the period-like domain, a region shown to be important for NF-κB signaling (to be discussed). However, similar to RASAL2, the frequency of DAB2IP mutations in breast cancer was relatively low, consistent with the observation that RasGAPs are more typically silenced by epigenetic events in cancer (19).

To confirm that DAB2IP protein levels were indeed suppressed in human luminal B tumors, we assessed DAB2IP protein expression in a panel of patient-derived xenografts (PDX). Notably, DAB2IP protein was minimally expressed in all five luminal B tumors examined (Fig. 1G), as compared with a basal PDX control, a subtype in which the DAB2IP gene is infrequently methylated and suppressed (Fig. 1A–C). Immunohistochemical analysis on a larger set of primary human tumors will be described below.

DAB2IP Controls the Activation of Multiple RAS Proteins and Functions as a Tumor Suppressor in Luminal Breast Cancers

To investigate the function of DAB2IP in luminal tumor models, DAB2IP expression was assessed in a panel of ER+ breast cancer cell lines. Although primary luminal A and B tumors can readily be distinguished by transcriptional profiling analysis, there have been discrepancies in classifying luminal cell lines into A and B subtypes. Some suggest that cell lines cannot be molecularly distinguished as luminal A or B and/or that luminal A tumors are not represented by existing cell lines (22, 23), whereas others classify cell lines by different criteria (refs. 24, 25; various annotations of cell lines examined are shown in Supplementary Table S1). Therefore, to select cell lines for functional studies, we merely sought to identity luminal cell lines that either lacked DAB2IP expression (for gain-of-function analysis) or expressed DAB2IP (for loss-of-function studies). In contrast to normal mammary epithelial cells (MCF10As), DAB2IP was absent or minimally expressed in three of six luminal lines (Fig. 1H; Supplementary Fig. S1A). Importantly, we found that the DAB2IP promoter was differentially methylated in breast cancer cell lines with low DAB2IP expression as compared with DAB2IP-expressing cell lines (Supplementary Fig. S1B), and DAB2IP mRNA levels directly correlated with protein expression across a broad panel of human breast cancer cell lines (Supplementary Fig. S1C and S1D; P < 0.0001), further supporting the conclusion that promoter methylation regulates DAB2IP silencing in breast cancer.

The biochemical and biological consequences of either ablating or reconstituting DAB2IP were evaluated in two luminal breast cancer cell lines: CAMA1 and MCF7 cells, respectively. Notably, DAB2IP ablation led to potent RAS pathway activation in CAMA1 cells (Fig. 1I). Specifically, DAB2IP knockdown using two different shRNAs activated HRAS, NRAS, and ERK, although it should be noted that the KRAS protein was not detected in this particular cell line (data not shown). Although shRNA-mediated DAB2IP suppression had no effect on CAMA1 cell proliferation in two-dimensional (2-D) cultures (Supplementary Fig. S1E), it dramatically enhanced anchorage-independent colony formation (1J). Moreover, anchorage-independent growth was suppressed by reconstitution with wild-type DAB2IP, but was not inhibited by a catalytically inactive RasGAP mutant (DAB2IPR413L), demonstrating that the RasGAP domain is required for the tumor-suppressor function of DAB2IP (Fig. 1J; Supplementary Fig. S1F).

DAB2IP was also reconstituted in MCF7 cells, which normally lack DAB2IP (Fig. 1K). Only low levels of DAB2IP expression were tolerated in these cells, potentially due to an evolved dependence on its suppression. Nevertheless, reconstitution potently suppressed HRAS-GTP and KRAS-GTP levels, had more modest effects on NRAS, and suppressed phospho-ERK (Fig. 1K). Consistent with ablation studies, DAB2IP reconstitution did not affect proliferation under 2-D culture conditions (Supplementary Fig. S1G), but reduced colony formation (Fig. 1L), whereas the inactivating mutation in the catalytic RasGAP domain failed to inhibit RAS and ERK activation (Fig. 1M) and had no effect on colony formation (Fig. 1N). Finally, CRISPR-mediated ablation of DAB2IP using two different guide RNAs promoted primary tumor growth in a xenograft model in vivo (Fig. 1O). Collectively, these studies demonstrate that (1) expression of DAB2IP is frequently lost in luminal B breast cancers and its suppression promotes the growth of primary tumors, (2) DAB2IP can regulate multiple RAS isoforms, and (3) its RasGAP activity is required for its tumor-suppressor function.

Concomitant Loss of DAB2IP and RASAL2 Expression Occurs in the Most Aggressive Luminal B Tumors

Because DAB2IP and RASAL2 both regulate RAS (Fig. 1 and ref. 21), we expected that their loss of expression might be mutually exclusive and that tumors would exhibit the loss of either one or the other RasGAP gene. However, immunoblot analysis demonstrated that DAB2IP and RASAL2 proteins were concomitantly suppressed in some luminal cell lines (Fig. 2A). In addition, DAB2IP and RASAL2 were both either minimally expressed or undetectable in three of five luminal B PDX tumors (Fig. 2B). As with DAB2IP, we found that the RASAL2 promoter was hypermethylated in cells that expressed low levels of RASAL2 (Supplementary Fig. S2A, chr1: 176,334,344), and mRNA levels precisely correlated with protein levels (Supplementary Fig. S2B and S2C). To more accurately determine the relative frequency of DAB2IP and/or RASAL2 loss of expression in luminal B tumors, mRNA expression levels were analyzed in primary luminal B tumors from TCGA datasets. DAB2IP was suppressed in a striking 46% of luminal B malignancies, and nearly half of these tumors had also lost expression of RASAL2 (Fig. 2C). Notably, ER+ tumors with low levels of both RASAL2 and DAB2IP more frequently presented as stage III/IV lesions as compared to those with high levels (30% vs. 17%; P < 0.05, Fisher exact test), which more frequently presented as stage I lesions. Moreover, in the context of luminal B disease alone, low levels of both genes together were associated with a significant decrease in relapse-free survival (Fig. 2D, log rank P = 3.1e–08).

To confirm and quantify DAB2IP and RASAL2 protein loss in primary human tumors, immunohistochemical analysis was performed. Antibodies were first validated using human xenografts that expressed or lacked DAB2IP and RASAL2 (Supplementary Fig. S2D and S2E). DAB2IP and RASAL2 expressions were then examined in 63 primary human luminal B tumors (as described in ref. 26). Consistent with mRNA analysis, DAB2IP expression was uniformly absent or minimally detected in 43% of human luminal B tumors, and 21% of luminal B tumors exhibited a loss of both DAB2IP and RASAL2 expression (Fig. 2E). Although outcome data are not yet available, low levels of DAB2IP and RASAL2 in this luminal B cohort were also associated with more advanced stage at the time of diagnosis. Specifically, 90% of tumors that lacked DAB2IP and RASAL2 were categorized as stage II–IV lesions, whereas only 10% were categorized as stage I lesions. In contrast, 45% of luminal B tumors that expressed DAB2IP and RASAL2 were categorized as stage I (P < 0.05, Fisher exact test). Taken altogether, these observations suggest that DAB2IP and RASAL2 are frequently coinactivated in luminal B tumors and raise the possibility that they may functionally cooperate in the development and/or progression of these cancers.

Concomitant Loss of DAB2IP and RASAL2 Cooperatively Drives Invasion, Epithelial-to-Mesenchymal Transition, and Metastasis

To elucidate the precise biological consequences of concomitant DAB2IP and RASAL2 loss, we first investigated whether the suppression of both genes might enhance tumor formation in vivo. Although DAB2IP and RASAL2 suppression individually promoted the growth of human breast cancer xenografts, there was no additional increase in primary tumor growth when both genes were ablated (Fig. 2F). However, because the concomitant loss of DAB2IP and RASAL2 was associated with advanced disease and recurrence in humans, we evaluated the effects of suppression on invasiveness and epithelial-to-mesenchymal transition (EMT): cellular properties associated with tumor progression and metastasis (27). Although the individual ablation of DAB2IP and RASAL2 modestly increased invasiveness, their concomitant loss dramatically enhanced the invasion of human mammary epithelial cells (Fig. 2G and H) and luminal breast cancer cells (Fig. 2I). Results were confirmed using multiple shRNA, siRNA, and sgRNA sequences (Fig. 2G–I; Supplementary Fig. S2F), and this enhanced invasiveness was rescued by reconstitution with RASAL2 and DAB2IP cDNA (Fig. 2J).

Interestingly, the combined loss of DAB2IP and RASAL2 expression also cooperatively enhanced EMT. For example, EMT is characterized by a decrease in E-cadherin (encoded by CDH1) and an increase in N-cadherin (encoded by CDH2) levels. Only the combined ablation of RASAL2 and DAB2IP resulted in both the effective suppression of E-cadherin and upregulation of N-cadherin (Fig. 2K). Immunoblot and/or qPCR analysis also revealed a dramatic cooperative increase in the expression of the EMT transcriptional regulators SLUG, SNAIL, and ZEB1 and other EMT markers such as fibronectin (encoded by FN1) and vimentin (encoded by VIM; Fig. 2K and L). Again, the induction of EMT was rescued by ectopic expression of RASAL2 and DAB2IP cDNA (Supplementary Fig. S2G). Together, these findings suggest that although the loss of one RasGAP gene is sufficient to promote primary tumor development, the concomitant loss of DAB2IP and RASAL2 together potently promotes invasion and EMT, consistent with a potential role for these genes in tumor progression.

To determine whether DAB2IP and RASAL2 regulate metastasis in vivo, both gain-of-function and loss-of-function studies were performed. First, DAB2IP and RASAL2 were reconstituted individually or together in MDA-MB-361 breast cancer cells that lacked endogenous DAB2IP and RASAL2 and were engineered to express luciferase. Following intracardiac injection, metastatic lesions were monitored by Xenogen imaging. Parental MDA-MB-361 cells are metastatic (Fig. 3A and B). Notably, although reconstitution with DAB2IP or RASAL2 alone reduced the number of metastases, only combined reconstitution prevented metastasis formation altogether (Fig. 3A and B). This complete suppression of metastasis was maintained for 4 months, when the experiment was terminated (Fig. 3C, P = 0.004, Mann–Whitney). Importantly, DAB2IP and RASAL2 reconstitution had no effect on the proliferation of these cells, indicating that the absence of metastases was not merely a consequence of impaired proliferation (Fig. 3D and E). To complement these studies, DAB2IP and RASAL2 were also ablated in a luminal mouse cancer cell line using shRNAs, and tumor cells were injected into the tail veins of syngeneic mice as previously described (Fig. 3F and G; ref. 28). Metastases in the lungs were visualized and quantified by counting individual tumors from multiple adjacent histologic sections. Importantly, codepletion of DAB2IP and RASAL2 dramatically enhanced metastasis (Fig. 3G and H; P = 0.0019, Mann–Whitney).

Interestingly, we previously crossed the autochthonous genetically engineered version of this luminal model to a Rasal2 knockout mouse and found that the frequency of metastasis was enhanced (21). Retrospective analysis of DAB2IP expression revealed that nearly 60% of primary tumors from this model had spontaneously lost DAB2IP (Supplementary Fig. S3A) and that animals with primary tumors that lost both RasGAPs developed significantly more metastases (Supplementary Fig. S3B; P = 0.004, Mann–Whitney). Moreover, in two of two instances examined, in which DAB2IP was still expressed in the primary Rasal2−/− tumors, it was selectively lost in metastatic lesions from these same animals (Supplementary Fig. S3C). Together, these gain-of-function and loss-of-function studies demonstrate that DAB2IP and RASAL2 cooperatively regulate the metastasis of luminal cancers. Expression analysis in both human tumors and mouse models confirms that these two tumor suppressors are in fact frequently coinactivated in aggressive luminal malignancies.

RAS Pathway Activation Drives Invasion, Whereas NF-κB Activation Is Required for EMT and Metastasis

To deconstruct the specific signals that drive invasion, EMT, and metastasis, we first evaluated the combined effects of DAB2IP and RASAL2 loss on RAS signaling. Although the suppression of each RasGAP activated KRAS, HRAS, ERK, and AKT, the concomitant ablation of both RasGAPs dramatically enhanced the activation of all of these components, which overall appeared to be greater than an additive effect (Fig. 4A and B). We then assessed the contribution of the RAS pathway to invasion and EMT triggered by DAB2IP/RASAL2 loss. As shown in Fig. 2, MCF10A cells are not normally invasive, but become potently invasive when DAB2IP and RASAL2 are suppressed. Both MEK and PI3K inhibitors suppressed invasion triggered by combined DAB2IP and RASAL2 loss, underscoring the importance of both of these effector pathways in the process of invasion (Fig. 4C; refs. 29–31). Moreover, although reconstitution with wild-type DAB2IP and RASAL2 suppressed RAS pathway activation (Supplementary Fig. S4) and reversed the invasive phenotype (Figs. 2J and 4D), a catalytically inactive RasGAP point mutant (20, 21) failed to do so (Fig. 4D; Supplementary Fig. S4). In direct contrast, PI3K and MEK inhibitors had no effect on EMT triggered by DAB2IP/RASAL2 suppression (Fig. 4E), suggesting that aberrant RAS pathway activation primarily serves to drive invasion in DAB2IP/RASAL2-deficient cells.

In addition to its catalytic RasGAP domain, DAB2IP is unique among RasGAPs in that it also possesses a period-like domain, which has been shown to directly bind TRAF2, and suppress NF-κB signaling (Fig. 4F; refs. 20, 32). Therefore, the effect of DAB2IP and/or RASAL2 loss on NF-κB activation in mammary epithelial cells was examined using an NF-κB transcriptional reporter assay. Although RASAL2 ablation did not affect NF-κB activation, DAB2IP suppression significantly stimulated NF-κB activity (Fig. 4G). Interestingly, the concomitant loss of RASAL2 further enhanced NF-κB activation (Fig. 4G). This finding is consistent with the observation that the RAS pathway can converge with and enhance NF-κB activation in other cell types (33, 34). Nevertheless, in this setting, NF-κB activation first requires DAB2IP loss, which appears to function as a dominant suppressive signal. NF-κB activity can only then be further enhanced by the loss of a second RasGAP, RASAL2.

Although the period-like domain of DAB2IP has been shown to be critical for NF-κB regulation in prostate cancer (20), we also assessed its importance in this setting. Notably, reconstitution with either wild-type RASAL2 and DAB2IP or catalytically inactive RasGAP point mutants (DAB2IPR413L and RASAL2K417E) effectively restored NF-κB activity to baseline levels in DAB2IP/RASAL2-depleted cells (Fig. 4H), supporting the conclusion that RAS pathway activation is not sufficient to induce NF-κB in these cells. However, when a deleterious point mutation in the period-like domain of DAB2IP (DAB2IPR413L/S728A) was also included (32), NF-κB activity remained elevated (Fig. 4H). Taken together, these data demonstrate that the concomitant loss of DAB2IP and RASAL2 triggers a potent activation of both RAS and NF-κB in breast cancers, and illustrate the importance of the RasGAP and period-like domains in regulating these pathways, respectively.

To investigate the contribution of NF-κB activation to invasion, EMT, and metastasis triggered by DAB2IP and RASAL2 loss, we utilized an IKBα super-repressor (IKBα-SR), which effectively suppresses NF-κB activation in DAB2IP/RASAL2-deficient cells (Fig. 5A; ref. 35). In contrast to RAS pathway inhibitors, IKBα suppression had no effect on the invasive properties of DAB2IP/RASAL2-deficient cells (Fig. 5B). However, it potently blocked the suppression of E-cadherin and the induction of N-cadherin (Fig. 5C) and dramatically inhibited the upregulation of all other EMT markers, triggered by DAB2IP and RASAL2 loss of expression (Fig. 5D). Finally, to determine whether NF-κB activation was required for metastasis in vivo, we evaluated its effects in the luminal tumor model shown in Fig. 3G. Although the combined suppression of DAB2IP and RASAL2 triggered metastasis in this model, coexpression of the IKBα-SR significantly suppressed the development of metastatic lesions (Figs. 5E and F, P = 0.037, Mann–Whitney). Taken together, these data suggest that DAB2IP and RASAL2 loss of expression drives the metastasis of luminal breast cancers by potently activating the RAS and NF-κB pathways. These data further suggest that although the RAS pathway primarily regulates invasion, NF-κB drives EMT and is required for metastasis, thus providing a unique mechanism of cooperativity between these two tumor suppressors (Fig. 6).

Luminal tumors comprise approximately 75% of all breast cancers, and many patients are cured by surgery and/or anti-estrogen therapies (1). However, a subset of luminal cancers, namely luminal B tumors, are highly proliferative, metastatic, and frequently recur. Transcriptional, genomic, and epigenetic profiling studies demonstrate that the more aggressive luminal B tumors are distinct from the less aggressive luminal A tumors (3–5). However, the genetic or epigenetic drivers of this aggressive behavior are still largely unknown (9). This lack of insight impedes both treatment decisions and therapeutic development.

In this study, we show that two RasGAP tumor suppressors, DAB2IP and RASAL2, play an important role in the pathogenesis of luminal B tumors. Strikingly, up to 62% of luminal B cancers have lost expression of at least one of these genes. However, we find that tumors that have lost both genes frequently present as advanced disease and are more likely to recur. Importantly, we report that DAB2IP and RASAL2 can individually function as tumor suppressors in breast cancer, and when suppressed promote primary tumor growth in animal models. However, the loss of both genes triggers the potent activation of multiple RAS isoforms, the RAS effectors ERK and AKT, as well as NF-κB. We show that NF-κB is primarily activated by DAB2IP loss, which directly regulates the NF-κB pathway through its period-like domain; however, this signal is further enhanced by the dramatic activation of RAS, caused by the loss of expression of both RasGAPs. Our data further suggest that both the RAS and NF-κB pathways play important and distinct roles in invasion, EMT, and metastasis. Taken together, these studies reveal a mechanism by which these two major oncogenic pathways are activated in luminal B breast cancers and cooperate to drive their progression, which has been a long-standing unanswered question in the field (9).

In addition to providing insight into the pathogenesis of luminal B cancers, this finding may also have important clinical implications. Currently, breast cancers are not functionally classified as luminal A or luminal B tumors in the clinic, and all ER+ luminal tumors are often treated as ER-responsive cancers (9). Although clinical tests have been developed to assess the risk of recurrence, better prognostic markers are urgently needed to identify patients who are at the highest risk, in order to more accurately determine who should receive (or be spared from) escalated treatments such as adjuvant chemotherapy (2). Given that DAB2IP and RASAL2 play a direct causal role in driving metastasis and expression is concurrently lost in almost a quarter of luminal B tumors, our data suggest that DAB2IP and RASAL2 loss of expression should be evaluated as potential predictive biomarkers. The antibodies used and validated in this study may be valuable tools in this context. Moreover, analysis could be coupled with the assessment of downstream effectors such as NF-κB, AKT/mTOR, or ERK, all of which have been implicated in disease progression (9, 17, 36). If successful, these efforts could complement or improve current tests, which could have a significant impact on treatment decisions.

Finally, the mechanism of epigenetically silencing two related tumor suppressors in cancer may have interesting teleologic implications. Notably, the potent activation of RAS through the acquisition of an activating point mutation in a RAS gene directly requires a single genetic event. This may be advantageous for some tumor types in which a burst of RAS activity provides a growth or survival advantage. However despite the frequent activation of the RAS pathway in advanced tumors, RAS mutations are notably rare in breast cancer, raising the intriguing possibility that there may be selection against such an event. This could be due to enhanced sensitivity to a strong oncogenic signal, which might result in replicative stress, senescence, or some other growth-inhibitory response. Thus, the sequential epigenetic silencing of two genes might provide an alternative mechanism of RAS activation, perhaps resulting in a more gradual, progressive activation of the RAS pathway during tumor evolution. The direct convergence of the RAS and NF-κB pathways via DAB2IP also provides a means of further expanding and coupling these important oncogenic signals. Regardless, the studies presented here provide important mechanistic insight into this poorly understood tumor type and provide a framework for investigating this gene family, and more broadly this mechanism of inactivation, in other cancers.

Cell Culture and DNA Constructs

MCF7 cells were purchased from the ATCC in 2013. BT549, MDA-MB-231, MDA-MB-453, SKBR3, and T47D cells were obtained from Dr. William Hahn (Dana-Farber Cancer Institute) in 2011. BT474, SUM149PT, SUM159PT, and ZR-75-1 cells were obtained from Dr. Frank McCormick (University of California, San Francisco) in 2011. CAMA1 cells were obtained from Dr. Marcia Haigis (Harvard Medical School) in 2010. MDA-MB-361 cells were provided by Dr. Charlotte Kupperwasser (Tufts Medical Center) in 2012, and MCF10A cells were provided by Dr. Jayanta Debnath (University of California, San Francisco) in 2013. All human breast cancer cell lines were authenticated in August 2016 by the Molecular Diagnostics Laboratory at the Dana-Farber Cancer Institute according to an American National Standards Institute (ANSI)–recommended protocol. Briefly, cell lines were authenticated using the Promega GenePrint 10 system, and amplified DNA fragments were resolved on the Applied Biosystems 3130xl Genetic Analyzer and were analyzed with Applied Biosystems GeneMapper v 4.0 software. MCNeu cells were obtained from Dr. Sandra McAllister (Brigham and Women's Hospital) in 2015 and were originally established by Dr. Laura Esserman (University of California, San Francisco; ref. 28).

shRNAs were purchased from the RNAi Consortium (Broad Institute, MIT) with the following sequences: RASAL2 shRNA1 (5′-CCCTCGTGTTCTTGCTGATAT-3′), RASAL2 shRNA2 (3′-UTR shRNA) (5′-ATGGAGTGCAATAGGACATTG-3′), DAB2IP shRNA1 (3′-UTR shRNA) (5′-GTAATGTAACTATCTCACCTA-3′), DAB2IP shRNA2 (5′-AGGGATAGGCTAAGGAGTAAG-3′), DAB2IP shRNA3 (5′-GCACATCACTAACCACTACCT-3′). A control shRNA was purchased from Addgene (5′-CCTAAGGTTAAGTCGCCCTCG-3′). siControl and siDAB2IP were purchased from Dharmacon (ON-TARGETplus Nontargeting Control Pool, cat. # D-001810-10, and SMARTpool HUMAN siGENOME DAB2IP, cat. # M-008249-01-0005: 5′-CGCAGUUGUUAGAAGACGA-3′/5′-GGCUAAGGAGUAAGGACGA-3′/5′-GGACCAACAUGCAGCGCUU-3′/5′-GAUAGAUUUCACCCGGUUA-3′). The CRISPR guides [DAB2IP guide-1 (5′-TGGGATGCACCGCTCGCCGGagg-3′) and DAB2IP guide-2 (5′-GATGGAGAACCTCCGGCGAGcgg-3′)] were cloned into the lentiviral Cas9-containing pXPR-BRD001 construct obtained from Dr. William Hahn (Dana-Farber Cancer Institute). The Control pXBR-BRD001 vector contained no guide RNA.

The RASAL2 and RASAL2K417E (GAP point mutant) cDNA in the pHAGE-N-Flag-HA lentiviral expression vector was previously described (21). The wild-type, R413L GAP point mutant, and R413L/S728A double mutant (GAP point mutant and Per-domain point mutant) DAB2IP cDNA, previously described in Minn and colleagues (20), was cloned into the pLX-C-V5 lentiviral expression vector (obtained from Dr. William Hahn, Dana-Farber Cancer Institute) and the pFB retroviral expression vector (generously provided by Pablo Viciana-Rodriguez). The pBABE IKBα-SR construct was obtained from Dr. William Hahn (Dana-Farber Cancer Institute).

Infections and Transfections

shRNA or cDNA constructs were prepared and virus was harvested as previously described (21). Virus was incubated on target cells for 6 to 12 hours at a 1:2–1:10 dilution with 8 μg/mL polybrene. Infected cells were selected in 0.5 to 2.0 μg/mL puromycin, 400 to 800 μg/mL neomycin, or 1 to 2 μg/mL blasticidin, depending on the construct and optimized for each cell line. For transfections, cells were transfected for 6 hours with 10 μmol/L siRNA constructs using a 1:400 dilution of Lipofectamine RNAiMAX (Invitrogen, cat. # 13778-075) in antibiotic-free media.

Proliferation and Soft-Agar Assays

For proliferation assays, 150,000 cells were plated in triplicate in 6-well dishes. The following day, cells were counted for day 0, and then triplicate sets of cells were counted on day 3. Cells for soft-agar assays were plated in six biological replicates in 6-well dishes using the appropriate growth media, cell number, and agarose concentration as previously described (21).

Protein Lysates and Western Blot Analyses

Protein extracts were isolated from cells in 1% SDS boiling lysis buffer. RAS-GTP levels were determined using a RAS Activation Assay Kit (EMD Millipore, cat. # 17-218). The following antibodies were used for immunoblots: phospho-AKT (Ser473, Cell Signaling Technology, cat. # 4060, 1:5,000), AKT (Cell Signaling Technology, cat. # 9272, 1:2,000), DAB2IP (Abcam, cat. # 87811, 1:3,000), E-cadherin (BD Transduction Laboratories, cat. # 610181, 1:1,000), phospho-ERK (Thr202/Thr204, Cell Signaling Technology, cat. # 4370, 1:2,000), ERK (Cell Signaling Technology, cat. # 9102, 1:3,000), GAPDH (Cell Signaling Technology, cat. # 2118, 1:5,000), HA (Covance, cat. # MMS-101P, 1:2,000), N-cadherin (BD Transduction Laboratories, cat. # 610920, 1:1,000), HRAS (Santa Cruz Biotechnology, cat. # SC-520, 1:1,000), KRAS (Santa Cruz Biotechnology, cat. # SC-30, 1:100), NRAS (Santa Cruz Biotechnology, cat. # SC-519, 1:200), RASAL2 (GeneTex, cat. # 120989, 1:2,000), SLUG (Cell Signaling Technology, cat. # 9585, 1:2,000), α-tubulin (Sigma, cat. # T5168, 1:10,000), β-tubulin (Sigma, cat. # T4026, 1:10,000), V-5 (Life Technologies, cat. # 46-0705, 1:2,000), and vinculin (Cell Signaling Technology, cat. # 4650, 1:1,000). For RAS-GTP assays and analysis of EMT markers, cells were typically serum-starved in 1% serum for 36 hours. RAS-GTP levels were assessed by Western blot and quantified using ImageJ software. For inhibitor experiments, drugs were added to the cells [PD-0325901 (MEK inhibitor) was used at 500 nmol/L and GDC-0941 (PI3K inhibitor) was used at 750 nmol/L], and lysates were harvested 48 hours after drug addition.

For the DAB2IP/RASAL2 protein/mRNA correlation analysis, mRNA levels were assessed as described in the “Quantitative Real-Time PCR” section, and samples were normalized to the internal control GAPDH. Protein levels were assessed by Western blot and quantified using ImageJ software, and samples were normalized to GAPDH. Following normalization to GAPDH, mRNA and protein levels were normalized to DAB2IP/RASAL2 levels in normal mammary epithelial cells (MCF10A). The correlation analysis was performed in PRISM.

PDX Analysis

PDXs were produced as previously described by DeRose and colleagues (HCI-003-005, and HCI-011; ref. 37) and Sikora and colleagues (HCI-013; ref. 38). Tumors were harvested and flash-frozen. The frozen tissue was homogenized, and protein extracts were isolated in 1% SDS boiling lysis buffer.

Molecular Subtype Association, Mutational, Survival, and Expression Analysis

Human expression (RNA sequencing; IlluminaHiSeq) and DNA methylation (HumanMethylation 450K) TCGA breast invasive carcinoma data were analyzed using the UCSC TCGA genome browser (39). For statistical analyses, the expression data were downloaded, and the statistical test was performed in PRISM. Mutational data were downloaded from the cBio portal (13, 14). All survival curves were generated using the Kaplan–Meier Plotter browser (40) using the following probes: DAB2IP-225020_at and RASAL2-217201_at.

The frequency of DAB2IP and RASAL2 suppression in human breast cancers was determined using the TCGA breast invasive carcinoma human expression data set (RNA sequencing; IlluminaHiSeq) and defining the lowest 25% expressing samples across all subtypes as DAB2IP/RASAL2 “low.” Conversely, “high” expression was defined as the highest 25% expressing samples across all subtypes. Based on this threshold, luminal B tumors were analyzed to assess the percentage of luminal B tumors that had low RASAL2 and/or DAB2IP expression. The TCGA clinical classifications of the same data set were analyzed, and stage annotations of ER+ tumors with “co-low” versus “co-high” expression (“low” and “high” defined as described above) were compared (Fisher exact test).

Immunohistochemistry and Tissue Microarrays

Human breast tissue samples were acquired and tissue microarrays were constructed as previously described (26). Institutional review board (IRB) approval for the study was originally obtained through the Dana-Farber/Harvard Cancer Center IRB; the Newton Wellesley Hospital IRB; the University of Colorado, Colorado Multiple IRB; the Sunnybrook Health Sciences Center research ethics board; and the Mayo Clinic IRB. The study was conducted in accordance with the Declaration of Helsinki, and every patient participating in the study provided written informed consent. Luminal B tumors were classified as previously described (26). Tissue microarrays were sectioned onto ProbeOn Plus microscope slides (Fisher Scientific) for immunohistochemistry. Briefly, tissues were dewaxed and washed in 100% EtOH. Antigen retrieval was performed with citrate buffer pH 6 for 6 minutes at full power and 6 minutes at medium power in a microwave. Samples were then treated with Redusol (Biomeda Corp.) for 2 minutes at 40°C, incubated with fresh H2O2/MeOH (1% vol/vol) for 2 minutes at 40°C, blocked with appropriate serum for 5 minutes at 40°C, incubated with appropriate antibodies at 4°C overnight, washed, and then incubated with secondary antibodies for 15 minutes at 40°C. Slides were treated by washing with peroxidase enhancer and incubated with AEC chromogen (Biomeda) for 2 minutes at 40°C. All washes were performed with Automation Buffer (Biomeda). Cell nuclei were counterstained with aqueous hematoxylin QS (Vector Laboratories). Slides were mounted with Faramount (Dako Corp.). Antibodies used for staining tissue sections were DAB2IP (1:200; abcam87811) and RASAL2 (1:200; GTX120989). Vectastain Elite ABC system kits were used (Vector Laboratories). Images were captured under indicated magnification with identical exposure and gain for any given experiment using a Nikon Eclipse 90i microscope. We evaluated immunostaining blinded to clinicopathological information and scored the stains as negative or positive. We used the average value from replicate scores for correlation with clinical outcome.

In Vitro Invasion Assays

BioCoat Matrigel Invasion Chambers (BD, cat. # 354480) were prepared according to the manufacturer's instructions. MCF10A cells (50,000) were plated in triplicate in assay media (described by Debnath and colleagues, ref. 41) in the top chamber of the invasion inserts. Assay media with 20 ng/mL EGF were added to the bottom, and cells were incubated overnight. After incubation, cells were fixed and stained using the Diff Stain Kit (IMEB Inc., cat. # K7128) according to the manufacturer's instructions. The number of cells that had invaded through the Matrigel was quantified by counting four random distinct fields using a light microscope. For CAMA1 cells, 100,000 cells were plated per insert in serum-free media, and full serum (10% FBS) was added to the bottom as the chemoattractant. For inhibitor experiments, drugs were added to the top and bottom of the invasion inserts. PD-0325901 (MEK inhibitor) was used at 500 nmol/L, and GDC-0941 (PI3K inhibitor) was used at 750 nmol/L.

Xenograft Assays

All animal procedures were approved by the Center for Animal and Comparative Medicine at Harvard Medical School in accordance with the NIH Guidelines for the Care and Use of Laboratory Animals and the Animal Welfare Act.

Female NOD/SCID mice were purchased from Charles River Laboratories (cat. # 394), and CAMA1 cells (3 × 106 cells, 50% Matrigel from BD Biosciences, cat. # 354234) were injected subcutaneously. Tumor size was measured by caliper, and tumor volume was calculated using the formula volume = (length × width2) × π/6.

For the MDA-MB-316 metastasis experiment, the pLenti PGK V5 Luc vector (Addgene #21471) was introduced into MDA-MB-316 cells, and cells were selected with G418. Cells (1 × 106) were injected intracardically into NOD/SCID mice (Jackson Laboratories, cat. # 001303) using an ultrasound guidance device (VEVO 200). Mice were injected with 150 mg of luciferin intraperitoneally and imaged monthly using a Perkin Elmer IVIS Spectrum CT Biophotonic Imager. Bioluminescence was measured as total flux in photons per second, and a common scale was used for all images across the study.

For the MCNeu metastasis experiment, transgenic FVB/N-Tg (MMTVneu) mice were purchased from Jackson Laboratory (cat. # 002376), and 3 × 105 cells (established previously, ref. 28) were injected into the tail vein. Lungs were harvested 3 weeks after injection, fixed in buffered formalin, stored in 70% ethanol, paraffin embedded, and sectioned. Sections were stained with hematoxylin and eosin, and metastases were quantified by counting all metastases per lung in each animal using a light microscope.

Nuclear Factor-κB Reporter Assays

Cells (2 × 104) were seeded in triplicate in 24-well plates. After 18 hours, cells were transfected with 100 ng of pNF-κB–luciferase (BD Bioscience) and 25 ng Renilla luciferase–encoding plasmid (pRL-TK) with X-tremeGENE 9 DNA Transfection Reagent (Roche, cat. # 06365809001). Thirty-six hours after transfection, luciferase activity was measured followed by Renilla luciferase activity using the Dual Luciferase Reporter Assay Kit (Promega, cat. # E-1910). All data were normalized as relative luciferase light units/Renilla luciferase light units.

Quantitative Real-Time PCR

Total RNA was isolated using TRIzol (Invitrogen) and was subsequently Dnase-treated (Roche). cDNA was synthesized using the Quanta qScript Kit (VWR, cat. # 101414-098). Standard curves were run with each primer set, and sample reactions were run in triplicate using PerfeCTA SYBR Green Supermix (Quanta Script, cat. # 95071-250). Samples were normalized to the internal control glyceraldehyde 3-phosphate dehydrogenase. The primer sequences used were as follows:

  • CDH1

  • 5′ TGCCCAGAAAATGAAAAAGG

  • 3′ GTGTATGTGGCAATGCGTTC

  • FN1

  • 5′ CAGTGGGAGACCTCGAGAAG

  • 3′ TCCCTCGGAACATCAGAAAC

  • GAPDH

  • 5′ CATGTTCGTCATGGGTGTGAACCA

  • 3′ ATGGCATGGACTGTGGTCATGAGT

  • CDH2

  • 5′ ACAGTGGCCACCTACAAAGG

  • 3′ CCGAGATGGGGTTGATAATG

  • SLUG

  • 5′ GGGGAGAAGCCTTTTTCTTG

  • 3′ TCCTCATGTTTGTGCAGGAG

  • SNAIL

  • 5′ CTGGGTGCCCTCAAGATGCA

  • 3′ CCGGACATGGCCTTGTAGCA

  • VIM

  • 3′ GCTTCCTGTAGGTGGCAATC

  • 5′ GAGAACTTTGCCGTTGAAGC

  • ZEB1

  • 3′ TGGTGATGCTGAAAGAGACG

  • 5′ TGCACTGAGTGTGGAAAAGC

  • RASAL2

  • 3′ ACCCTCTTTGCTCGTACAAC

  • 5′ GGGATGTTGACTAGCCCTAC

  • DAB2IP

  • 3′ CGCGCTACCAAACCATCAC

  • 5′ ATCATCAGGTCTGTCAGGAAGT

Bisulfite Sequencing

Genomic DNA was isolated using a Qiagen kit (cat. # 69504), and the bisulfite conversion reaction was performed using a New England BioLabs kit (cat. # E3318S). The primers used for amplifying up the CpG region of interest were designed using the MethPrimer browser (42) and were as follows:

  • DAB2IP

  • 5′ AGAATTCGGGATAGGTTTAAAGTTGTAGT

  • 3′ AGAATTCCCTCTTAACCCCAAATAACTAT

  • RASAL2

  • 5′ AGAATTCTGTTTTTTAGTAAAAGAATGGGATTG

  • 3′ AGAATTCACAAACTCTTACCTAAAATCTACAT

The amplification PCR reaction was carried out using a kit from New England Biolabs (cat. # M0490), and the reaction was set up and run according to the manufacturer's instructions. Following amplification, PCR products were sent out for sequencing.

A. Welm and Y.S. DeRose receive a portion of royalties if the University of Utah licenses PDX models to for-profit companies. No royalties are received for academic work such as that described in this paper. A.H. Partridge is a consultant/advisory board member for Pfizer. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S.N. Olsen, T. De Raedt, K. Cichowski

Development of methodology: S.N. Olsen, Z. Castaño, B. Dake

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.N. Olsen, A. Wronski, B. Dake, C. Malone, W. Zhou, S. Guerra, A. Welm, S.S. McAllister, C. Kuperwasser

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.N. Olsen, T. De Raedt, M. Loda, A.H. Partridge, C. Kuperwasser

Writing, review, and/or revision of the manuscript: S.N. Olsen, A.H. Partridge, S.S. McAllister, C. Kuperwasser, K. Cichowski

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Guerra

Study supervision: C. Kuperwasser

Other (helped to carry out subcutaneous injections): M. Enos

Other (provided the breast cancer PDX tissues to the authors): Y.S. DeRose

We thank Ayana Henderson for technical assistance.

This work was supported by grants from the Ludwig Center at DF/HCC (K. Cichowski) and the NCI (R01 1R01CA188659; K. Cichowski).

1.
Siegel
RL
,
Miller
KD
,
Jemal
A
. 
Cancer statistics, 2016
.
CA: A Cancer J Clin
2016
;
66
:
7
30
.
2.
Redig
AJ
,
McAllister
SS
. 
Breast cancer as a systemic disease: A view of metastasis
.
J Intern Med
2013
;
274
:
113
26
.
3.
Perou
CM
,
Sørlie
T
,
Eisen
MB
,
van de Rijn
M
,
Jeffrey
SS
,
Rees
CA
, et al
Molecular portraits of human breast tumours
.
Nature
2000
;
406
:
747
52
.
4.
Sotiriou
C
,
Neo
S-Y
,
McShane
LM
,
Korn
EL
,
Long
PM
,
Jazaeri
A
, et al
Breast cancer classification and prognosis based on gene expression profiles from a population-based study
.
Proc Natl Acad Sci U S A
2003
;
100
:
10393
8
.
5.
Sørlie
T
,
Perou
CM
,
Tibshirani
R
,
Aas
T
,
Geisler
S
,
Johnsen
H
, et al
Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications
.
Proc Natl Acad Sci U S A
2001
;
98
:
10869
74
.
6.
Nielsen
TO
,
Parker
JS
,
Leung
S
,
Voduc
D
,
Ebbert
M
,
Vickery
T
, et al
A Comparison of PAM50 intrinsic subtyping with immunohistochemistry and clinical prognostic factors in tamoxifen-treated estrogen receptor-positive breast cancer
.
Clin Cancer Res
2010
;
16
:
5222
32
.
7.
Ellis
MJ
,
Ding
L
,
Shen
D
,
Luo
J
,
Suman
VJ
,
Wallis
JW
, et al
Whole-genome analysis informs breast cancer response to aromatase inhibition
.
Nature
2012
;
486
:
353
360
.
8.
Tran
B
,
Bedard
PL
. 
Luminal-B breast cancer and novel therapeutic targets
.
Breast Cancer Res
2011
;
13
:
221
.
9.
Ellis
MJ
,
Perou
CM
. 
The genomic landscape of breast cancer as a therapeutic roadmap
.
Cancer Discov
2013
;
3
:
27
34
.
10.
Hu
Z
,
Fan
C
,
Oh
DS
,
Marron
JS
,
He
X
,
Qaqish
BF
, et al
The molecular portraits of breast tumors are conserved across microarray platforms
.
BMC Genomics
2006
;
7
:
96
.
11.
Koboldt
DC
,
Fulton
RS
,
McLellan
MD
,
Schmidt
H
,
Kalicki-Veizer
J
,
McMichael
JF
, et al
Comprehensive molecular portraits of human breast tumours
.
Nature
2012
;
490
:
61
70
.
12.
Bamford
S
,
Dawson
E
,
Forbes
S
,
Clements
J
,
Pettett
R
,
Dogan
A
, et al
The COSMIC (catalogue of somatic mutations in cancer) database and website
.
Br J Cancer
2004
;
91
:
355
8
.
13.
Gao
J
,
Aksoy
BA
,
Dogrusoz
U
,
Dresdner
G
. 
Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal
.
Sci Signal
2013
;
269
:
p11
.
14.
Cerami
E
,
Gao
J
,
Dogrusoz
U
,
Gross
BE
,
Sumer
SO
,
Aksoy
BA
, et al
The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data
.
Cancer Discov
2012
;
2
:
401
4
.
15.
Sivaraman
VS
,
Wang
H
,
Nuovo
GJ
,
Malbon
CC
. 
Hyperexpression of mitogen-activated protein kinase in human breast cancer
.
J Clin Invest
1997
;
99
:
1478
83
.
16.
Lintig von
FC
,
Dreilinger
AD
,
Varki
NM
,
Wallace
AM
,
Casteel
DE
,
Boss
GR
. 
Ras activation in human breast cancer
.
Breast Cancer Res Treat
2000
;
62
:
51
62
.
17.
Mueller
H
,
Flury
N
,
Eppenberger-Castori
S
,
Kueng
W
,
David
F
,
Eppenberger
U
. 
Potential prognostic value of mitogen-activated protein kinase activity for disease-free survival of primary breast cancer patients
.
Int J Cancer
2000
;
89
:
384
8
.
18.
Bernards
A
. 
GAPs galore! A survey of putative Ras superfamily GTPase activating proteins in man and Drosophila
.
Biochim Biophys Acta
2003
;
1603
:
47
82
.
19.
Maertens
O
,
Cichowski
K
. 
An expanding role for RAS GTPase activating proteins (RAS GAPs) in cancer
.
Adv Biol Regul
2014
;
55
:
1
14
.
20.
Min
J
,
Zaslavsky
A
,
Fedele
G
,
McLaughlin
SK
,
Reczek
EE
,
De Raedt
T
, et al
An oncogene–tumor suppressor cascade drives metastatic prostate cancer by coordinately activating Ras and nuclear factor-κB
.
Nat Med
2010
;
16
:
286
94
.
21.
McLaughlin
SK
,
Olsen
SN
,
Dake
B
,
De Raedt
T
,
Lim
E
,
Bronson
RT
, et al
The RasGAP gene, RASAL2, is a tumor and metastasis suppressor
.
Cancer Cell
2013
;
24
:
365
78
.
22.
Neve
RM
,
Chin
K
,
Fridlyand
J
,
Yeh
J
,
Baehner
FL
,
Fevr
T
, et al
A collection of breast cancer cell lines for the study of functionally distinct cancer subtypes
.
Cancer Cell
2006
;
10
:
515
27
.
23.
Prat
A
,
Karginova
O
,
Parker
JS
,
Fan
C
,
He
X
,
Bixby
L
, et al
Characterization of cell lines derived from breast cancers and normal mammary tissues for the study of the intrinsic molecular subtypes
.
Breast Cancer Res Treat
2013
;
142
:
237
55
.
24.
Subik
K
,
Lee
JF
,
Baxter
L
,
Strzepek
T
. 
The expression patterns of ER, PR, HER2, CK5/6, EGFR, Ki-67 and AR by immunohistochemical analysis in breast cancer cell lines
.
Breast Cancer (Auckl.)
2010
;
4
:
35
41
.
25.
Holliday
DL
,
Speirs
V
. 
Choosing the right cell line for breast cancer research
.
Breast Cancer Res
2011
;
13
:
215
.
26.
Collins
LC
,
Gelber
S
,
Marotti
JD
,
White
S
,
Ruddy
K
,
Brachtel
EF
, et al
Molecular phenotype of breast cancer according to time since last pregnancy in a large cohort of young women
.
Oncologist
2015
;
20
:
713
8
.
27.
Chaffer
CL
,
Weinberg
RA
. 
A perspective on cancer cell metastasis
.
Science
2011
;
331
:
1559
64
.
28.
Campbell
MJ
,
Wollish
WS
,
Lobo
M
,
Esserman
LJ
. 
Epithelial and fibroblast cell lines derived from a spontaneous mammary carcinoma in a MMTV/neu transgenic mouse
.
In Vitro Cell Dev Biol Anim
2002
;
38
:
326
33
.
29.
Shin
S
,
Dimitri
CA
,
Yoon
S-O
,
Dowdle
W
,
Blenis
J
. 
ERK2 but Not ERK1 induces epithelial-to-mesenchymal transformation via DEF motif-dependent signaling events
.
Mol Cell
2010
;
38
:
114
27
.
30.
Virtakoivu
R
,
Pellinen
T
,
Rantala
JK
,
Perälä
M
,
Ivaska
J
. 
Distinct roles of AKT isoforms in regulating β1-integrin activity, migration, and invasion in prostate cancer
.
Mol Biol Cell
2012
;
23
:
3357
69
.
31.
Chin
YR
,
Toker
A
. 
Akt isoform-specific signaling in breast cancer
.
Cell Adh Migr
2014
;
5
:
211
4
.
32.
Zhang
H
. 
AIP1/DAB2IP, a novel member of the ras-GAP family, transduces TRAF2-induced ASK1-JNK activation
.
J Biol Chem
2004
;
279
:
44955
65
.
33.
Hutti
JE
,
Pfefferle
AD
,
Russell
SC
,
Sircar
M
,
Perou
CM
,
Baldwin
AS
. 
Oncogenic PI3K mutations lead to NF- B-dependent cytokine expression following growth factor deprivation
.
Cancer Res
2012
;
72
:
3260
9
.
34.
Dhawan
P
,
Richmond
A
. 
A Novel NF- B-inducing Kinase-MAPK signaling pathway up-regulates NF- B Activity in melanoma cells
.
J Biol Chem
2002
;
277
:
7920
8
.
35.
Boehm
JS
,
Zhao
JJ
,
Yao
J
,
Kim
SY
,
Firestein
R
,
Dunn
IF
, et al
Integrative genomic approaches identify IKBKE as a breast cancer oncogene
.
Cell
2007
;
129
:
1065
79
.
36.
Oida
K
,
Matsuda
A
,
Jung
K
,
Xia
Y
,
Jang
H
,
Amagai
Y
, et al
Nuclear factor-κB plays a critical role in both intrinsic and acquired resistance against endocrine therapy in human breast cancer cells
.
Sci Rep
2014
;
4
:
1
8
.
37.
DeRose
YS
,
Wang
G
,
Lin
Y-C
,
Bernard
PS
,
Buys
SS
,
Ebbert
MTW
, et al
Tumor grafts derived from women with breast cancer authentically reflect tumor pathology, growth, metastasis and disease outcomes
.
Nat Med
2011
;
17
:
1514
20
.
38.
Sikora
MJ
,
Cooper
KL
,
Bahreini
A
,
Luthra
S
,
Wang
G
,
Chandran
UR
, et al
Invasive lobular carcinoma cell lines are characterized by unique estrogen-mediated gene expression patterns and altered tamoxifen response
.
Cancer Res
2014
;
74
:
1463
74
.
39.
Zhu
J
,
Sanborn
JZ
,
Benz
SC
,
Craft
B
,
Szeto
C
. 
The UCSC Cancer Genomics Browser
.
Nat Methods
2009
;
6
:
239
40
.
40.
Györffy
B
,
Lanczky
A
,
Eklund
AC
,
Denkert
C
,
Budczies
J
,
Li
Q
, et al
An online survival analysis tool to rapidly assess the effect of 22,277 genes on breast cancer prognosis using microarray data of 1,809 patients
.
Breast Cancer Res Treat
2009
;
123
:
725
31
.
41.
Debnath
J
,
Muthuswamy
SK
,
Brugge
JS
. 
Morphogenesis and oncogenesis of MCF-10A mammary epithelial acini grown in three-dimensional basement membrane cultures
.
Methods
2003
;
30
:
256
68
.
42.
Li
L-C
,
Dahiya
R
. 
MethPrimer: Designing primers for methylation PCRs
.
Bioinformatics
2002
;
18
:
1427
31
.