A transposon-based analysis reveals RASA1 is involved in triple negative breast cancer

RAS genes are mutated in 20% of human tumors, but these mutations are very rare in breast cancer. Here, we used a mouse model to generate tumors upon activation of a mutagenic T2Onc2 transposon via expression of a transposase driven by the keratin K5 promoter in a p53+/- background. These animals mainly developed mammary tumors, most of which had transposon insertions in one of two RASGAP genes, neurofibromin1 (Nf1) and RAS p21 protein activator (Rasa1). Immunohistochemical analysis of a collection of human breast tumors confirmed that low expression of RASA1 is frequent in basal (triple-negative) and estrogen receptor negative tumors. Bioinformatic analysis of human breast tumors in The Cancer Genome Atlas database showed that although RASA1 mutations are rare, allelic loss is frequent, particularly in basal tumors (80%) and in association with TP53 mutation. Inactivation of RASA1 in MCF10A cells resulted in the appearance of a malignant phenotype in the context of mutated p53. Our results suggest that alterations in the Ras pathway due to the loss of negative regulators of RAS may be a common event in basal breast cancer. Cancer Res; 77(6); 1357-68. ©2017 AACR.


INTRODUCTION
Ras genes are some of the most frequently mutated in human cancer. According to the catalogue of somatic mutations in cancer (COSMIC v77) (1), which represents the most comprehensive database on human cancer mutations currently available, around 20% of the analyzed tumors have activating mutations in any of the three Ras genes, with a maximum of 57% incidence for KRAS in pancreatic tumors. Ras signaling may also be activated by other means, notably by inactivation of molecules that limit Ras activity, such as Ras GTPase-activating proteins (RasGAPs). Ras proteins are molecular switches that cycle between an inactive GDP-bound form and an active GTP-bound form. They signal through several effector pathways, and RasGAPs stimulate the weak intrinsic GTPase activity of normal (but not mutant) Ras proteins, effectively acting as suppressors of Ras function. Interestingly, less than 1% of the near 10,000 breast cancer samples sequenced in COSMIC have mutations in Ras genes, however the Ras pathway is significantly activated in a number of breast tumors, in particular of the triple-negative type (2).
Breast cancer is by far the most frequent tumor type in the female population worldwide (25% of all new cases in 2012), and although its mortality rate is not the highest, it is the most frequent cause of cancer death in women (14.7% of all deaths in 2012) (3). Triple negative breast cancers (TNBC, which are negative for HER2, ERα and the progesterone receptor) constitute a heterogeneous group of tumors which very often exhibit a basal-like signature (4). Although they represent approximately 15% of all breast cancers, they account for a much higher mortality: they are tumors with a poor prognostic, mainly due to the lack of specific targets for treatment. These triple-negative tumors are enriched for mutations in TP53. Indeed, TP53 is mutated in 36% of all breast cancers, but this proportion rises to 86% in PAM50 basal-like tumors (5).
TNBC tumors also bear a highly variable number of genomic alterations, including the presence of a large number of somatic mutations and copy number aberrations (CNA) (6), suggesting that i) combinations of mutations interact to drive tumor formation, and ii) most of the mutations found are "passengers", not related to development of the tumor.
Transposon-induced mutagenesis is an excellent method to identify cancer driver genes. For example, when mobilized by Sleeping Beauty (SB) transposase, the mutagenic T2Onc2 transposon integrates throughout the genome, and cells with insertions in a gene or combination of genes that favour tumorigenesis are positively selected. Using this method, genes involved in multiple tumor types have been identified (reviewed in (7)). This analysis has not yet been performed for breast cancer.
In this manuscript, we report on generation of transposon-bearing mice that develop mammary tumors, identification of RasGAP genes as the major target of transposonmediated mutagenesis, and identification of RASA1 hemizygous deletion in human triple negative breast cancer. Together with the identification of hemizygous deletions in NF1 and RASAL2 in breast cancer (8,9), our results highlight the importance of RasGAP gene hemizygosity as a driver of elevated Ras signalling and breast cancer in humans.

Tumor collection
Tumors were excised from the mice and were fractionated. One part was included in formaldehyde for subsequent paraffin embedding for immunohistochemistry analysis, and the rest was snap frozen in liquid N2 for protein/nucleic acid extraction.

Human Tumors
Samples and data from patients included in this study were provided by the Biobanco i+12 in the Hospital  Spain), who purchased them from ATCC and provided vials after first division. All cells lines were authenticated by ATCC by isoenzyme analysis and STR profiles, and were kept in culture for less than six months. Cells were periodically analyzed for contamination with a mycoplasma detection kit. All cells were passaged and cultured using the media and recommendations provided by ATCC. MCF10A-p53 R175H/shRasa1 cells were grown in plates coated with 0.1% gelatin due to their limited adhesion. For experimentation, all other types of MCF10A cells were also grown in gelatin-coated plates to avoid substrate effect.

Immunohistochemistry
Mouse tissues were fixed in 10% buffered formalin and embedded in paraffin. Five μmthick sections were used for H&E staining or immunohistochemical preparations. Most of the tumors originated in the mice were fixed and classified by morphology after sectioning and staining with H&E. For immunohistochemistry, slides were deparaffinized and antigen retrieval was performed in microwave with citric acid buffer (pH 6) for mouse tissues, and in pressure cooker with Dako Target

Statistic analysis
For western blotting, bands were quantified by Quantity One software and normalized with respect to beta-actin, total ERK or total AKT expression. P values were determined by using the unpaired, two-tailed Student t test. For immunohistochemistry, samples were assessed in a blind manner by two experts. RASA1 and NF1 staining was quantified from 0 to 3 comparing with normal tissue. When staining was heterogeneous within tumors, five different areas were assessed and an average value was calculated. P values were determined by using the Chi-square test. In both cases, p values < 0.05 were considered significant and data are expressed as mean ± SEM.

SB/T2/p53+/-mice develop spontaneous mammary tumors.
To generate tumors, double transgenic mice bearing both a concatemer of T2Onc2 mutagenic transposons and the SB11 transposase under control of the keratin K5 promoter (10) were mated with heterozygous Trp53+/-mice (Fig. 1A). As expected, these Trp53+/-animals are prone to the development of lymphoma. Also, as K5 is expressed in the skin, they also develop skin tumors (10). Interestingly, SB/T2/p53+/triple transgenic mice preferentially developed mammary tumors ( Figure 1B). This occurs at a higher frequency (41% vs. 19%) and shorter latency (49 vs. 60 weeks) than in control Trp53+/-mice lacking transposition (Fig. 1C). Keratin K5 is expressed in the myoepithelial layer of the mammary gland and SB11 transposase was detected in this layer of transgenic animals (Fig. 1D). Likewise, mammary tumors showed concurrent expression of K5 and SB11 transposase (Fig. 1D). Thus, transposition of T2Onc2 in K5expressing cells facilitates development of mammary gland tumors in a Trp53+/background.
RasGAPs are the most frequently mutated genes in transposon-induced mammary tumors.
DNA from these 33 mammary tumors was extracted, and sequences flanking the transposon insertion sites were amplified by PCR and then subjected to nextgeneration Illumina sequencing to identify all transposon integrations. Scrutiny of these integrations using gene-centric Common Insertion Site (gCIS) analysis (14) resulted in identification of 16 CIS or Common Insertion Sites, which represent specific sites in the genome that accumulate transposon insertions in independent tumors at a rate significantly higher than expected by chance, and are therefore likely the result of positive selection during tumor development ( Table I) (15)(16)(17)(18), and NFIB, which is disrupted by SB in 40% (13/33) of tumors undergoes translocations in breast cancers (19). Gene ontology analysis showed selection for insertion into genes related to cell matrix adhesion, oncogenesis, apoptosis, cell migration and angiogenesis (Table II). Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Author Manuscript Published OnlineFirst on January 20, 2017; DOI: 10.1158/0008-5472.CAN-  In our murine tumors, the RasGAPs Nf1 and Rasa1 were among the most frequently mutated genes. Of note, 18 of the 33 analyzed tumors (54%) had transposon insertions in either one or both of the RasGAP genes Nf1 and Rasa1, strongly suggesting activation of the Ras pathway in generation of these tumors. The position and orientation of T2Onc2 insertions in each target gene hints at the type of alteration produced (20); when orientation of the transposon was analyzed, we found that 100% of the transposon insertions in Nf1 and 70% of insertions in Rasa1 were in the opposite orientation, suggesting that gene inactivation has occurred. Both in Nf1 and Rasa1, all insertions were located before or into the functional RASGAP domain ( Fig. 2A), effectively disrupting the function of these proteins and resulting in the activation of Ras signalling.

Mouse mammary tumors show reduced RasGAP expression
To validate these results, we analyzed a number of tumors by western blot (Fig. 2B). In Since disruption of Rasa1 or Nf1 should result in enhanced Ras signaling, we checked for activation of ERK (a known Ras effector) and AKT (an effector of Ras and PI3K) in these tumors. ERK activity (measured as pERK accumulation) was present in 85% (6/7) of tumors with Rasa1 and/or Nf1 insertions, as compared to 38% (3/8) of tumors without insertions in either gene ( Fig. 2B and 2E). A correlation was also detected between activation of AKT and diminution of RASA1 ( Fig. 2B and 2F), suggesting activation of the Ras pathway in tumors with Rasa1 or Nf1 insertions. These partial correlations may well be associated with ERK activation through oncogenic pathways distinct from elevated Ras signaling associated with disruption of Rasa1 or Nf1.
Interestingly, in some tumors with high pERK but lacking insertions in Rasa1 or Nf1 we found insertions (not included in our gCIS list, since they were found in less than three tumors) in other Ras-related genes, such as Rasgrf1 (see Supplementary Table SI). We did not find any correlation between insertions in Rasa1 or Nf1 and expression of EGFR, ERBB2 or ESR1 (data not shown).

RASA1 and NF1 genes are frequently lost in triple negative breast cancer
While NF1 is a well-known tumor suppressor gene, and NF1 deletions and mutations have been reported in breast cancer (8,18), RASA1 has not been previously considered as a breast cancer gene. We used immunohistochemistry to assess expression of  Tables S III and S IV). This trend was even more evident when ER-negative tumors were compared to ER-positive tumors (p<0.01, Fig 3D). For NF1 staining, a trend towards lower expression in ER-negative tumors was also seen (supplementary Tables S III and S IV). TP53 is frequently mutated in triple negative breast cancer, and TP53 staining of these tumors allowed us to identify a significant correlation between low or absent RASA1 staining and mutation in TP53 (p<0.05, Fig   3A and (Fig 4B). Interestingly, we also detected a correlation between copy number of RASA1 and ESR1 expression, both at mRNA and protein levels (Figs. 4C and D), and also between high methylation of ESR1 and hemizygous deletion of RASA1 (Fig. 4E). A similar correlation was found between RASA1 copy loss and PGR1 (but not ERBB2) expression (data not shown). In addition, 75% of PAM50 basal-like tumors showed coincident allelic loss at RASA1 and mutation of TP53 (vs 21% when all tumors are considered). Enrichment analysis on 974 tumors also revealed a significant correlation between RASA1 copy number loss and mutation of TP53 (p= 3,50E-60, We also performed expression meta-analysis using the GOBO online database (http://co.bmc.lu.se/gobo) (22), which includes more than 1800 patients from 10 breast cancer studies. Again, in agreement with our results, expression of NF1 and  Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Concomitant TP53 mutation and inactivation of RASA1 malignize human breast cells
To validate in vitro these observations, we permanently inactivated RASA1 by lentiviral-mediated shRNA interference in several breast cell lines. In T-47D, BT-474 and HCC-1954 cells, inactivation of RASA1 with two different shRNAs (sh1 and sh2) resulted in activation of ERK and/or AKT, confirming that partial deletion of RASA1 is enough to activate the Ras pathway (Fig. 5A). Interestingly, in SUM159 or MDA-MB-231 cells, which already have constitutive activation of the Ras pathway, this effect was not seen (data not shown). Since all the breast cancer cell lines tested had mutations in TP53, and most triple negative tumors present concomitant deletion of RASA1 and mutations in the TP53 gene, we next investigated the relation of TP53 and RASA1 using MCF10A cells, aTP53-wild type immortalized non-malignant mammary epithelial cell line which is often used as a model of normal human mammary gland. In these cells, interference using sh1 almost completely suppressed the expression of RASA1, while sh2 reduced its expression to around 50% (Fig. 5B). For both shRNAs, phosphorilation of ERK was weak, as also was upon introduction of a R175H TP53 mutation (the most frequent mutation in breast tumors). However, combination of CD49f expression (Fig. 5D), which is indicative of a transition from an epithelial to a mesenchymal phenotype (23). Moreover, p53 R175H /shRASA1 cells lost E-cadherin and upregulated N-cadherin expression ( Fig. 5B and E). Finally, double p53 R175H /shRASA1 cells also acquired invasive properties, as seen by invasion chamber assays using matrigel (Fig. 5F). Interestingly, the two interfering shRNAs exerted similar but not identical effects on MCF10A cells in the context of the p53 mutation: almost total inactivation of RASA1 by sh1 resulted in almost total disappearance of EpCAM and Ecadherin, while partial inactivation of RASA1 by sh2 resulted in partial EpCAM loss and partial inactivation of E-cadherin, with the remaining E-cadherin expressing cells exhibiting a spiky, discontinuous pattern of E-cadherin, instead of the continuous staining seen in the control cells (Fig. 5E). On the contrary, both total and partial inactivation of RASA1 coupled with p53 R175H resulted in activation of the Ras pathway,

DISCUSSION
Transposon-mediated generation of tumors in a Trp53-heterozygous background has allowed us to identify inactivation of RasGAP genes as a frequent event in murine mammary tumors. Since its initial development as a tool for identification of cancerpromoting genes in transgenic mice (24,25), transposon technology has been successfully used by many laboratories to identify genes causing cancer in a variety of tissues, and many of these have proved to be of clinical significance (reviewed in (7)).
Our screen has identified a number of gCIS that are known to function as tumor suppressor genes, and are subject to loss-of-function mutations in human breast cancer. 30% of the genes identified in our screen are included in the cancer gene census, among them PTEN, which is mutated in a high percentage of human breast tumors, in particular in TNBC (26). Combined deletion of Pten and Trp53 in mouse mammary epithelium results in the development of claudin-low type tumors (27). tumors, which appear earlier and are more malign in a Trp53+/-background (reviewed in (31)). Furthermore, there are several reports that hint in particular to RASA1 as a breast cancer gene: for instance, downregulation of RASA1 is associated with poor survival of breast invasive ductal carcinoma patients (32), and loss of Chr 5q14 (where the RASA1 gene is located) has been noted before in breast tumors (33-35). Moreover, chromosome 5q loss is a characteristic marker of the integrative cluster IntClust 10.
This IntClust 10 is one of 10 groups that result from the characterization of human breast tumors according to genomic and transcriptomic landscapes (36,37), and includes mostly triple negative tumors from the core basal-like intrinsic subtype.
Interestingly, these tumors have the highest rate of TP53 mutations (37).
More than 50% of the murine tumors that we have analyzed showed transposon insertions in either one or both of the RasGAP genes Nf1 and Rasa1, suggesting a strong selective pressure towards RasGAPs inactivation for tumorigenesis. There are some classical studies that have already established a relationship between NF1 and human breast cancer (38-41), and recently, high throughput studies have confirmed involvement of NF1 mutations in this disease (8,18), so it is conceivable that RASA1 (which is a protein that shares its main function with NF1) is also involved in the development of breast tumors. Our results also synergize with recent reports on RasGAPs alterations in other tumor types, as for instance RASA2 in melanoma (42), RASALl1 in colorectal cancer (43) and RASAL2 also in breast cancer (9,44 negative regulator of ERK activity, acts as a tumor suppressor in basal-like breast cancer (48). Interestingly, a recently reported breast cancer transposon screen has also identified Rasa1 as a potential breast tumor suppressor gene in a Pten-mutant background (49).
Our results also confirm that downregulation of RASA1, when accompanied by the presence of a mutated TP53, is sufficient to induce an EMT response in MCF10A cells.
This response probably requires a mutation in TP53, since cells bearing wild-type TP53 did not show any sign of malignancy upon RASA1 inactivation, nor did cells with p53 R175H and wild-type RASA1 ( Figure 5 and data not shown). These results agree with the strong correlation between RASA1 loss and mutation of TP53 that we have detected (Supplementary Table S      transposon insertion in the Rasa1 gene. Panels C and D were normalized using β-actin signal, panels E and F were normalized using total ERK and AKT signals, respectively.      T23  T29  T2   T1  T19  T8   TP53   T23  T29  T2   T1  T19