Prominent Oncogenic Roles of EVI1 in Breast Carcinoma

Breast carcinoma is the most common malignant tumor and predominant cause of cancer-related death in women worldwide. During the last years, increasing breast carcinoma heterogeneity has been documented concerning mutational background, histopathology, dissemination patterns and efficacy of surgical, antihormonal, chemotherapy, or radiotherapies. Despite high initial remission rates, especially in early-stage disease, breast carcinoma patients carry a significant life-long risk for disease relapse (1). Recent research has focused on so-called breast carcinoma stem cells (CSC) as mediators of tumor relapse after long latency (2) as well as on stemness proteins as CSC biomarkers and potential drug targets (3, 4).

The EVI1 gene is part of the complex MECOM locus on human chromosome 3q26 and encodes a zinc finger transcription factor that is expressed in long-term repopulating hematopoietic stem cells (HSC; refs. 5, 6). In acute myeloid leukemia (AML), EVI1 overexpression can occur due to chromosomal rearrangements or as a reflection of the stem cell origin of the disease, but in either case predicts very adverse prognosis (7). EVI1 expression has also been reported in solid tumors including breast carcinoma (8, 9), where it is still largely understudied with respect to relevance, functional roles, and molecular regulation.

Here, we performed a comprehensive expression and functional analysis of EVI1 in human breast carcinoma. By analyzing a tissue microarray (TMA) of 608 patient samples, we found high EVI1 protein expression in breast carcinoma regardless of the ER status. A detailed clinicopathologic investigation uncovered a prognostic significance of EVI1 expression in ER- and especially triple-negative breast carcinoma, which was however not observed in HER2⁺ tumor subsets. Although EVI1 depletion impaired cell-cycle progression, apoptosis resistance, and MAPK signaling in both estrogen receptor–negative (ER⁻) and estrogen receptor–positive (ER⁺) breast carcinoma cells, addition of estrogen could rescue these effects only in ER⁺ cells. Moreover, similar as in patients, HER2 overexpression appeared to overrule EVI1 effects on MAPK signaling, explaining why EVI1 expression is of particular clinical relevance in the ER⁻ HER2⁻ tumor subset. Finally, we identified the GPR54-ligand KISS1 as a novel transcriptional target of EVI1, which promotes breast carcinoma cell migration. In sum, our report identifies EVI1 as an oncogene that profoundly regulates breast carcinoma biology and that is of particular importance for estrogen-independent HER2-negative tumors.

Handling of patient samples and data analyses were performed in accordance with federal and state laws and approved by the local ethics committee. The TMA included samples from 608 human primary breast carcinoma (primary or recurrent) histologically processed and diagnosed at the Institute for Pathology and Molecular Pathology, University Hospital Zurich (Zurich, Switzerland) as described (10). Immunohistochemistry using rabbit anti-EVI1 antibodies (Cell Signaling Technology) and digital expression analysis were performed as published (11). Briefly, a semi-quantitative image analysis software (Tissue Studio v.2, Definiens AG) was applied to digitalized TMA slides, obtaining a continuous spectrum of average brown staining intensity of tumor cell nuclei in arbitrary units. Subsequently, EVI1 expression was categorized in low, medium, or high according to the 25th and 75th percentiles of all measured expression values. FISH was used to detect EVI1 copy-number gains and rearrangements using the EVI1-flanking BAC clones CTD-2079P9 and RP11-264O10 for probe labeling (12).

Breast carcinoma cell lines (DSMZ) were bought in 2012 and reauthenticated by DSMZ in September 2014 and August 2015, respectively, using a nanoplex PCR for specific DNA profiles in eight different highly polymorphic short tandem repeat loci. In addition, samples were tested for the presence of rodent mitochondrial DNA from mouse, rat, Chinese, and Syrian hamster. Cells were cultivated according to data sheet. Breast carcinoma primary tissue samples were dissociated to single cells and cultured as described (4). Estradiol (Sigma-Aldrich), Kisspeptin-10 (Kp-10; Santa Cruz Biotechnology), and RKI-1447 (Selleckchem) were used as indicated.

EVI1-specific or control shRNAs were designed using the MISSION TRC shRNA software tool and integrated into the pLKO.1-Puro vector system (Sigma) for lentiviral production. EVI1 overexpression and control vectors (13) were kindly provided by Olga Kustikova and Christopher Baum (Hannover Medical School, Hannover, Germany). For inducible overexpression, EVI1 or KISS1 cDNAs (the latter cloned from MDA-MB-231 cells using primers as indicated in Supplementary Table S2) were integrated into a pLVX vector system to drive expression by doxycycline (Sigma) from a Tet_on lentiviral system (Clontech). Lentiviral particles were produced and cells transduced as described (14).

Primary breast carcinoma cells were cultured for 24 hours with a mixture of 3 independent siRNAs against EVI1 and respectively 2 control siRNAs (Life Technologies) together with lipofectamine (Invitrogen) in penicillin/streptomycin-free culture medium as described (14, 15). Cells were then cultured under standard conditions for another 24 hours and then harvested for mRNA and functional assays.

RNA was extracted with an RNeasy kit (Qiagen) and cDNA synthesized using a Thermo Script RT-PCR System (Invitrogen). Reverse transcripts were amplified by qRT-PCR and quantified upon incorporation of SYBR Green on an ABI 7500 workstation. Relative expression levels were calculated after normalization to the reference gene GAPDH using the ΔΔC_T method. CDKN1A, CDKN1B, BIK, and BBC3 primers were purchased from Qiagen (SYBR Green QuantiTect Primer Assays). Other primer sequences are given in Supplementary Table S2.

To assess cell growth, 50,000 cells were plated and quantified after trypsinization on days 3, 6, and 9 postseeding. Cell proliferation was investigated by incorporation of BrdUrd or EdU as detailed in the manufacturers’ protocols (BrdUrd: BD Biosciences; Click-iT, EdU kit: baseclick). Cell-cycle and apoptosis assays were performed as described (15). Cells were analyzed by flow cytometry for their DNA content on a FACS Fortessa machine using FlowJo software (FlowJo enterprise). For apoptosis assays, 5 × 10⁴ cells/mL were incubated overnight and then treated either for 16 hours with staurosporine (2.5 μmol/L; Sigma-Aldrich) or for 24 hours with SuperKiller TRAIL (50 ng/mL; Enzo Life Sciences).

Immunoblotting was performed as described (4) using the following primary antibodies (Cell Signaling Technology): anti-pan AKT (#4691S), anti-pAKT (pSer473, #4060S), anti-ERK1/2 (#4695), anti-pERK1/2 (Thr202/Tyr204, #4377), anti-GAPDH (#5174P), anti-EVI1 (#2593), anti-p21 (#2947), anti-p27 (#3688), anti-CDK2 (#2546), and anti-β-actin (#3700S). Fluorescently labeled or HRP-conjugated secondary antibodies were used as described (14, 15).

Microarrays analyses were performed in triplicates from control and EVI1 knockdown MDA-MB-231 cells (obtained with either one of two independent EVI1-specific shRNAs). RNA was extracted with an RNeasy Mini kit (Qiagen). Concentration and purity of RNA samples were determined with a NanoDrop photometer (peqlab), and integrity confirmed on a 2100 Bioanalyzer (Agilent Technologies). Only RNA samples with RIN values ≥ 7.5 were considered. Per condition, 100 ng of RNAs were used to prepare cyanine-3–labeled cRNA for hybridizations, which were performed according to standard protocols using Agilent SurePrint G3 Human Gene Expression 8×60K v2 Microarrays. After extensive washing, fluorescence intensities were detected with the Scan Control A.8.4.1 software (Agilent) on an Agilent DNA Microarray Scanner and extracted from images using Feature Extraction 10.7.31 software (Agilent). Quantile normalization was applied to the data set, and correlation analysis was performed. Fold-change calculations identified differentially expressed genes, and Panther analysis most prominently affected pathways in EVI1 knockdown versus control cells. Array data have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE95272 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE95272.

Chromatin immunoprecipitation (ChIP) was performed as described (16). Briefly, 1 × 10⁷ control or EVI1-overexpressing Hs 578T cells were fixed in 1% formaldehyde, sonicated, precleared, and incubated with 10 μg anti-EVI1 or isotype control antibodies overnight at 4°C. Complexes were washed, DNA-extracted, precipitated, and amplified by RT-PCR using primers sets homologous to regions of the human KISS1 promoter. Nonimmunoprecipitated chromatin was used as input control. Primers flanking the EVI1-binding site in the BCL2L1 promoter and at a previously described nonbinding site served as positive and negative controls, respectively (17).

The established “wound healing assay” was performed to assess cell migration (18). Briefly, cells were grown to confluence in 24-well plates and incubated with 5 μg/mL aphidicolin (Sigma-Aldrich, A4487) and reduced FCS concentrations (2%) to stall proliferation. Subsequently, the monolayer was injured with a pipette tip and detached cells removed by iterative washing, leaving an approximately 200-μm wide unsettled zone free for lateral repopulation. Migration into these “wound areas” was followed on an Axio Vert.A1 microscope (Zeiss) and quantified by Fiji Imaging software at 0, 12, and 24 hours of incubation with or without addition of doxycycline, Kp-10, or RKI-1447 as indicated.

Animal experiments and zebrafish husbandry were approved by the “Kantonales Veterinaeramt Basel-Stadt.” Xenotransplantation and assessment of tumor cell engraftment were performed as described (4, 19). In brief, 75 to 100 breast carcinoma cells labeled with the fluorescent CellTracker (Life Technologies) were micro-injected at 48 hours postfertilization into the vessel-free area of the yolk sac of transgenic Tg(flk1:eGFP) zebrafish embryos anesthetized in 0.4% tricaine (20). For rescue experiments, the fish water was supplemented with 100 nmol/L estradiol (Sigma) or carrier (DMSO) at days 0 and 2.5 posttransplantation. Tumor development was assessed microscopically at day 5 postinjection (19, 21). For pERK inhibition, the fish water was supplemented with 200 nmol/L of CI-1040 at days 1 and 2 posttransplantation.

NOD.Cg-Prkdc^scid IL2rg^tmWjl/Sz(NSG) mice purchased from The Jackson Laboratory were maintained under pathogen-free conditions according to federal and state regulations. Control and EVI1 knockdown MDA-MB-231 cells (1 × 10⁶) mixed with Matrigel (1:1; BD Biosciences) were co-laterally implanted by subcutaneous injection into the flanks of individual mice and occurrence of tumors monitored by palpation as reported (15). Tumor area was assessed in situ using the Fiji software, and tumor weight was measured after excision.

Unless otherwise indicated, data from ≥3 independent biological experiments performed in technical triplicates were analyzed. Results are shown as mean ± SD. P values were calculated by two-tailed, unpaired Student t tests or as specified and P values indicated with * for <0.05, ** for <0.01, and *** for <0.001. Retrospective survival analyses of breast carcinoma patients were performed by the Kaplan–Meier method using log-rank (Mantel–Cox), Breslow, and Tarone–Ware tests.

First, we assessed EVI1 gene and protein expression in 12 primary human breast carcinoma samples (Fig. 1A) and 8 breast carcinoma cell lines (Fig. 1B and C; ER⁻: MDA-MB-231, BT-549, Hs 578T, MDA-MB-468, SK-BR-3, and ER⁺: BT-474, T-47D, MCF7). EVI1 expression was detected in several samples irrespective of the ER status. To cover a comprehensive range of endogenous EVI1 expression for subsequent functional investigations, two ER⁻ (MDA-MB-231 and Hs 578T) and two ER⁺ breast carcinoma cell lines (T-47D and MCF7) were chosen and investigated alongside with four primary patient-derived cell samples of different ER status (P1–P4).

Furthermore, we employed immunohistochemistry to investigate EVI1 protein expression on a TMA of 608 breast carcinoma samples (10). Reliable and biologically interpretable results were obtained from 527 samples, 512 of which information on ER status was available. Consistent with our previous data, EVI1 protein was detected at variable degrees (Fig. 1D) in both ER⁻ (n = 91) and ER⁺ (n = 421) tumors (Fig. 1D and Supplementary Table S1). However, a significant correlation between EVI1 expression levels and survival was only observed in the ER⁻ subgroup (n = 91 patients; 5-year survival: P = 0.011, overall survival: P = 0.026) but not in the ER⁺ subgroup (n = 421 patients) or the whole patient cohort analyzed together (Supplementary Fig. S1A). Interestingly, the influence of EVI1 expression on overall survival was most pronounced in triple-negative breast carcinoma (P = 0.006), but lost when ER⁻/HER2⁺ subsets were separately analyzed (Supplementary Fig. S1A). Together, these data suggest that EVI1 expression is of particular significance in breast carcinoma that is not driven by active ER and HER2 signaling.

ER- and triple-negative breast carcinoma subgroups, in which EVI1 showed prognostic relevance, were subjected to further analysis of clinico-pathologic parameters. High EVI1 expression associated indeed with enhanced distant metastasis rate (P = 0.046 and P = 0.027, respectively; Supplementary Fig. S1A), indicating a putative functional contribution to tumor cell dissemination/migration. To investigate the mechanisms responsible for EVI1 overexpression, we performed FISH analyses. Unlike in leukemia (22), we could not detect EVI1 gene rearrangements or copy-number gains except in 2 of 512 breast carcinoma patients (Fig. 1E; Supplementary Fig. S1B).

Based on these data, we conclude that EVI1 expression is frequently observed in human breast carcinoma, where it is mostly driven by yet unknown regulatory events, and might be particularly relevant for estrogen-independent HER2-negative tumors.

To examine the functional significance of EVI1 in breast carcinoma, we performed EVI1 knockdown experiments in two ER⁻ (MDA-MB-231 and Hs 578T) and one ER⁺ (T-47D) breast carcinoma cell line and two patient-derived primary breast carcinoma samples per condition (ER⁺: P1, P2; ER⁻: P3, P4). Cells were transduced with lentiviral particles carrying either noncoding or two alternate EVI1 shRNAs. Transduction with either shRNAs downregulated EVI1 protein and mRNA expression when compared with controls (Fig. 2A; Supplementary Figs. S2A–S2B and S3A). Throughout all analyzed samples, EVI1 knockdown cells showed a clear growth defect when compared with corresponding control cells (Fig. 2B; Supplementary Figs. S2C and S3B).

The lower growth rates observed in EVI1 knockdown cells could be caused by decreased proliferation or enhanced apoptosis rates, both of which are modulated by EVI1 in other cell types (22). Indeed, knockdown of EVI1 enhanced basal breast carcinoma cell apoptosis (Fig. 3A; Supplementary Figs. S2D and Fig. S3C) as well as apoptosis sensitivity in response to staurosporine or the death ligand TRAIL (Fig. 3B; Supplementary Fig. S3D). In addition, cell-cycle analyses revealed a G₁–S phase transition defect upon EVI1 knockdown (Fig. 3C; Supplementary Fig. S3E), indicating a proliferation defect. Supporting this notion, BrdUrd incorporation was also diminished (Fig. 3D; Supplementary Fig. S3F). In line, key checkpoint regulators that block G₁–S phase transition, such as the cyclin-dependent kinase inhibitors 1A and 1B (p21^Cip1 and p27^Kip1), were upregulated in EVI1 knockdown MDA-MB-231 cells, while their mutual downstream target CDK2 was decreased (Supplementary Fig. S4).

Intriguingly, the profound growth-modulatory effects of EVI1 were observed independent of the ER status, which is in apparent contrast to the prognostic significance of EVI1 expression especially in ER⁻ breast carcinoma patients. Indeed, the in vitro findings in ER⁺ breast carcinoma cells could be biased by lack or reduced ER stimulation under standard cultivation conditions. Confirming this hypothesis, addition of estradiol greatly restored growth of EVI1 knockdown ER⁺ T-47D but not ER⁻ MDA-MB-231 breast carcinoma cells (Fig. 4A). Consistently, estradiol fostered the incorporation of EdU in T-47D but not in MDA-MB-231 EVI1 knockdown cells (Fig. 4B). We conclude that active estrogen signaling overrules EVI1-mediated growth effects, and, therefore, EVI1-mediated growth induction may be more critical for patients with ER⁻ tumors that do not equally respond to natural estrogen.

ERK and AKT kinases are key regulators of cell proliferation and survival downstream of estrogen signaling (23–25). We thus wondered whether EVI1 also acts via activation of these kinases in breast carcinoma. No consistent pAKT suppression was observed in EVI1 knockdown cells (Fig. 4C and F), although EVI1 overexpression indeed induced pAKT levels (Fig. 4D). However, an overt decrease in phosphorylated (i.e., activated) ERK levels was reproducibly noted upon EVI1 knockdown throughout the analyzed ER⁻ and ER⁺ breast carcinoma samples (Fig. 4C and Supplementary Fig. S5A), indicating the ERK pathway as a dominant growth axis regulated by EVI1. Indeed, treatment with MEK inhibitors (CI-1040, trametinib, or AZD6244) that act upstream of ERK (26) similarly suppressed cell growth and cycle progression of MDA-MB-231 and T-47D cells (Supplementary Fig. S5B–S5F). Further supporting this notion, addition of estradiol enhanced ERK, but not AKT, phosphorylation in EVI1 knockdown ER⁺ T-47D but not ER⁻ MDA-MB-231 breast carcinoma cells (Fig. 4C). EVI1 overexpression consistently upregulated pERK in MCF7, T-47D, and MDA-MB-468 cells, and displayed synergistic effects with estradiol in ER⁺ T-47D cells (Fig. 4D–E). Interestingly, the rescue of cell growth induced by β-estradiol in EVI1 knockdown cells was abrogated by cotreatment with either the ER-blocking reagent tamoxifen or the MEK inhibitor CI-1040 (Supplementary Fig. S5G).

As also HER2 mediates growth-stimulatory effects via the MAPK/ERK signaling axis in breast carcinoma, we further examined the significance of EVI1 knockdown on HER-dependent ERK phosphorylation and found that, although loss of EVI1 signaling effectively depleted pERK in HER2⁻ breast carcinoma cells, ERK phosphorylation remained essentially unaltered in the investigated HER2⁺ samples (Fig. 4F, left vs. right plots). Together, these data suggest that EVI1, ER, and HER2 signaling functionally impinge on phosphor-modulation of ERK as a common downstream pathway.

Next, we used xenotransplantation assays to examine the relevance of EVI1 for in vivo tumorigenesis from human breast carcinoma cells. Equal numbers of EVI1 knockdown and control MDA-MB-231 cells (ER⁻HER2⁻) were injected subcutaneously into contralateral flanks of immuno-permissive NSG mice, and tumor formation was assayed over time. At 12 days posttransplantation, smaller tumors were documented from EVI1 knockdown cells versus control cells (Fig. 5A–C), indicating that EVI1 influences in vivo tumorigenicity. Immunoblot analysis confirmed persistent knockdown of EVI1 and impaired phosphorylation of ERK in excised tumors (Supplementary Fig. S6A). These data were confirmed in a previously established zebrafish xenotransplant model (4, 19). Consistent with the results obtained in mouse, both EVI1 knockdown ER⁺ T-47D and ER⁻ MDA-MB-231 cells induced fewer tumors than corresponding control cells, whereas estrogen supplementation rescued in vivo tumor formation selectively from ER⁺ cells (Fig. 5D–E; Supplementary Fig. S6B). Moreover, the MEK inhibitor CI-1040 was able to block the rescue effect of β-estradiol in vivo (Supplementary Fig. S6C). In line with their reduced in vivo tumorigenicity, EVI1 knockdown cells also displayed impaired mammosphere formation in vitro (Supplementary Fig. S6D). These data indicate that, although EVI1 may not be a specific CSC marker in breast carcinoma, it coregulates the stem cell compartment.

To further explore the molecular mechanisms underlying EVI1-driven effects in breast carcinoma, we analyzed the transcriptome of control and EVI1 knockdown MDA-MB-231 cells using gene expression microarrays. A total of 816 differentially expressed genes were identified in EVI1 knockdown versus control cells, of which 324 were up- and 492 downregulated. Panther analysis identified cell(-cell) adhesion, cell communication, signal transduction, developmental and immune system process regulation as the predominantly influenced biological processes, and receptors, cell-adhesion, and respectively extracellular matrix proteins as the most significantly affected protein classes (Fig. 6A). Furthermore, GeneSpring analyses revealed “G protein–coupled receptor (GPCR) signaling” molecules, such as KISS1, EDN1, PTGFR, and PIK3CG (Fig. 6B and C), as the most influenced pathway, next to cell-cycle control and progression (with perturbed expression levels of several key regulators such as CDKN1A, CDKN1C, CCNA1, and CDK1), apoptosis resistance (with upregulation of proapoptotic genes such as BIK, BMF, and BBC3), and ERBB signaling–related molecules (e.g., EREG, DUSP5, and NRG2). Heat maps of these individual categories are depicted in Fig. 6C and Supplementary Fig. S7A and S7B with a cut-off of 2-fold and 1.5-fold expression changes, respectively. EVI1-dependent expression changes of 15 exemplary candidate genes were further validated by qRT-PCR (Supplementary Fig. S7C).

To identify potential direct target genes of EVI1 in breast carcinoma, we next investigated the expression of candidate genes in response to EVI1 overexpression (Supplementary Fig. S7D). Among these, the GPR54-ligand KISS1 stood out as one of the most strongly influenced genes. In addition, the induction of KISS1 mRNA displayed a clear dose-dependency on EVI1 transcript levels (Fig. 7A). Furthermore, codepletion of EVI1 and KISS1 mRNA was observed in primary breast carcinoma cells treated with siRNAs against EVI1 versus corresponding control siRNA–treated cells (Supplementary Fig. S7E). Moreover, promoter analysis of KISS1 revealed several potential EVI1-binding sites within KISS1 regulatory elements (Supplementary Fig. S8A), reinforcing KISS1 as a putative direct transcriptional target of EVI1. Based on this analysis, four promoter regions of KISS1 were selected (Supplementary Fig. S8A) and assessed for EVI1 binding in ChIP assays. Higher enrichment rates were indeed observed in EVI1-overexpressing versus control cells especially at the most distal promoter site (–4880 to –4761 bp, Fig. 7B). Thus, these data identify the KISS1 promoter as a yet-unrecognized target region for EVI1 in breast carcinoma (see also Supplementary Fig. S8 for control and schematic illustration of binding sites). We therefore conclude that, next to modulating expression of cell-cycle– and apoptosis-relevant genes (Fig. 6B and C and Supplementary Fig. S7B–S7C), EVI1 directly influences GPCR signaling via transcriptional modulation of the GPR54 ligand KISS1.

KISS1 was originally identified as a metastasis suppressor (27, 28), and more recently described to enhance motility and invasiveness of ER⁻ breast carcinoma cells (29, 30). We thus hypothesized that EVI1 contributes to these processes at least in part via transcriptional regulation of KISS1. Indeed, migration assays revealed that knockdown of EVI1 strongly impaired the motility of ER⁻ MDA-MB-231 (Fig. 7C–F) and Hs 578T cells (Supplementary Fig. S9A–S9B), whereas overexpression of EVI1 overtly increased cell mobility (Fig. 7G and H; Supplementary Fig. S9C–S9D). Supporting the role of KISS1 as a downstream target in EVI1-dependent migration, exposure of cells to the GPR54-ligand Kisspeptin-10 (Kp-10), a gene product of KISS1 shown to enhance ER⁻ breast carcinoma cell motility (Supplementary Fig. S10A), indeed rescued the migration defects observed in EVI1 knockdown MDA-MB-231 cells (Fig. 7C and D). Supporting these data, overexpression of KISS1 itself also rescued migration in EVI1 knockdown MDA-MB-231 (Fig. 7E and F and Supplementary Fig. S10A–S1B) and Hs 578T cells (Supplementary Fig. S9A–S9B).

Noteworthy, several further modulators of cell motility and adhesion were found to be regulated by EVI1 in our microarray analysis, including RhoJ and TIE1, two molecules that had not been linked to EVI1 or breast carcinoma cell migration before. Exemplifying the functional relevance of also these findings, inhibition of RHO/ROCK signaling with RKI-1447 impaired EVI1-induced breast carcinoma cell mobility in migration assays (Fig. 7G and H and Supplementary Fig. S9C–S9D).

Interestingly, supplementation with Kisspeptin (Kp-10), which effectively rescued migration (Fig. 7C and D), could not restore cell growth in EVI1 knockdown cells (Supplementary Fig. S10C). Consistently, neither treatment with Kisspeptin (Kp-10) nor KISS1 overexpression influenced pERK activity (Supplementary Fig. S10D). Vice versa, treatment of breast carcinoma cells with the MAPK inhibitor CI-1040 effectively suppressed pERK and cell growth (Supplementary Fig. S5B, S5C, and S5F) but did not influence breast carcinoma cell migration (Supplementary Fig. S10F–G). In particular, treatment with CI-1040 did not abrogate EVI1-induced KISS1 overexpression (Supplementary Fig. S10E) and related breast carcinoma cell migration, reinforcing the idea that these effects are independent of the MAPK pathway (Fig. 7I).

Taken together, we demonstrate a functional relevance of EVI1 gene expression for breast carcinoma cell growth that involves modulation of pERK signaling (see Fig. 7I for schematic illustration). In part complementary to ER signaling and eventually overruled by constitutive ERK activity in HER2⁺ breast carcinoma, EVI1-mediated effects achieve pivotal significance in ER⁻ HER2⁻ breast carcinoma, where EVI1 expression is of prognostic significance. Moreover, we present evidence for a hitherto unrecognized EVI1-KISS1-GPR54 axis that—independently of ERK signaling—modulates breast carcinoma cell motility, suggesting that also the capacity to induce metastasis may be intimately linked to EVI1. Thus, our work identifies EVI1 as a novel critical determinant of breast carcinoma cell biology that is of particular importance for estrogen-independent HER2⁻ breast carcinoma.

Initially identified as a retroviral insertion region in hematopoietic cells (31), EVI1 has been intensely studied in HSCs and AML where it represents a marker of adverse prognosis. EVI1 is also expressed in other tissues such as brain, lung, and kidney (32–34). Pointing toward its significance in early organogenesis, Evi1 knockout mice are embryonically lethal and show broad hypocellularity and patterning defects (35, 36). The molecular regulation and functional relevance of EVI1 expression in breast carcinoma however are largely unexplored.

Analysis of a large cohort of primary samples did not detect significant gene rearrangements or copy-number gains, indicating that activation of the EVI1 locus in breast carcinoma follows different principles than in myeloid leukemia, such as regulation via miRNAs (8). Consistently, a common EVI1 polymorphism (rs6774494 A>G) targeted by miR-206/133b was suggested to predict adverse outcome in postmenopausal breast carcinoma patients (37). Interestingly, immunohistochemical analyses of our patient cohort identified EVI1 protein as a prognostic marker in ER⁻ but not ER⁺ breast carcinoma, supporting previous mRNA-based studies (8, 38). Importantly, when the ER⁻ cohort was further subdivided in ER⁻HER2⁺ and triple-negative breast carcinoma, EVI1 expression influenced survival specifically in the latter. In this subgroup, high EVI1 expression further associated with enhanced distant metastases.

Functional studies documented a profound growth defect in EVI1 knockdown versus control cells, resulting from impaired proliferation, cell-cycle progression, and apoptosis resistance. Interestingly, addition of estrogen to ER⁺ but not ER⁻ cells restored the impaired ERK activation and proliferation. Furthermore, both effects of β-estradiol were abrogated by cotreatment with either the ER-blocking reagent tamoxifen or the MAPK inhibitor CI-1040, which highlights the specificity of the observed effects and indicates that EVI1 and β-estradiol merge in pERK activation to regulate breast carcinoma cell growth. However, an inverse correlation was documented between EVI1 expression and tumor size in triple-negative breast carcinoma and by trend also in the ER⁻ subgroup. We hypothesize that the subgroup of breast carcinoma presenting with large primary tumor size and negative to low EVI1 expression is driven by aggressive, yet EVI1-unrelated molecular mechanisms.

Our analyses reliably identified the significance of EVI1 expression in the absence of endogenous estrogen signaling. In contrast, in the ER⁺ breast carcinoma cohort, such analyses might be complicated by the fact that these patients receive antiestrogen treatments. The importance of EVI1 expression might differ depending on the patient response and efficacy of such treatments. Our functional data show that EVI1 also severely regulates the growth of ER⁺ breast carcinoma cells, if estrogen is not provided. Unfortunately, we have no detailed and complete information on the mode and efficacy of antiestrogen treatments of the ER⁺ breast carcinoma patients. Thus, although our analyses support the notion that EVI1 could affect breast carcinoma independently of ER signaling, assessment of the relevance of EVI1 in ER⁺ breast carcinoma requires further patient stratification according to the response to antiestrogen treatment.

Suppression of EVI1 expression consistently inhibited MAPK activation in HER2⁻ but not HER2⁺ breast carcinoma. Thus, potential inhibitory effects of EVI1 knockdown on MAPK signaling might be overruled by constitutive HER2 activity. This assumption is consistent with the results in patients where EVI1 expression levels were prognostically relevant in triple-negative breast carcinoma, but not in ER⁻ HER2⁺ subsets. Thus, EVI1, estrogen, and HER2 signaling might converge on MAPK signaling as a common downstream effector controlling breast carcinoma cell growth. Growth-stimulatory properties of estrogen in breast carcinoma also involve transcriptional induction of cyclin D1 (39) and suppression of CDK inhibitors, such as p21^Cip1 or p27^Kip1 (40). Indeed, our investigations uncovered that also these growth-regulatory events are influenced by EVI1 and, moreover, that EVI1 modulates the expression of several further key cell-cycle regulators (e.g., CDKN1A, CDKN1B, CDKN1C, CCNA1, and CDK1).

In addition, our data suggest that EVI1 enhances apoptosis resistance in breast carcinoma by inducing a concerted suppression of pro- and induction of antiapoptotic genes. In line with previous data (14, 17), we found a physical association of EVI1 with the BCL-XL promoter. Previous links between EVI1 and apoptosis include direct blocking interactions with JNK in hematopoietic cells, and the inhibition of apoptosis through a PI3K-dependent mechanism in colon cancer cells (41, 42). EVI1 is further discussed as a stem cell factor in hematopoiesis and leukemia (22), but in breast carcinoma, it rather homogenously marked tumor cells, at least in cases of high EVI1 expression. Nevertheless, EVI1 knockdown affected the frequency of in vivo tumor– as well as in vitro mammosphere–initiating breast carcinoma cells. These data suggest that, even though not confined to breast CSCs, EVI1 expression might also regulate this compartment.

Intriguingly, we identified migration as a novel cellular function promoted by EVI1 in breast carcinoma. In particular, EVI1 knockdown impaired the breast carcinoma cell mobility, whereas its overexpression enhanced migration. Gene microarray and qRT-PCR experiments surprisingly uncovered, next to regulators of cell cycle and apoptosis, several factors implicated in cell communication and GPCR signaling as downstream effectors of EVI1. Of these, we analyzed in more detailed the GPR54-ligand KISS1, which ChIP assays identified as a novel transcriptional target of EVI1. The EVI1-KISS1 ligand axis promoted motility of ER⁻ breast carcinoma cells, in line with previously reported roles of KISS1 on mobility and adhesion in this disease entity (29, 30). Interestingly, although KISS1 has been reported as an upstream regulator of ERK (43), stimulation with the GPR54-ligand Kp-10 was not able to restore pERK and proliferation of EVI1 knockdown breast carcinoma cells, although it did influence breast carcinoma cell migration. Vice versa, MAPK inhibition effectively suppressed cell growth but did not alter migration, again reinforcing the idea that the EVI1-KISS1 migratory axis acts independently of pERK. Besides KISS1, we identified additional established (e.g., CXCR4, CCR1, AKAP12; refs. 44–46) as well as novel factors in breast carcinoma cell migration as targets of EVI1. For instance, TIE1, a modulator of angiogenesis and cell adhesion (47), and RhoJ, a modulator of the RHO/ROCK-dependent cell motility (48), were found to be modulated by EVI1, suggesting that this transcription factor serves as a master regulator of breast carcinoma cell motility.

Taken together, our data identify EVI1 as a potent oncoprotein regulating breast carcinoma cell proliferation, apoptosis resistance, and migration. Interestingly, estrogen and HER2 signaling as well as EVI1-mediated transcriptional modulation seemingly merge to stimulate MAPK signaling. This functional convergence identifies EVI1 as a major driver of cell growth acting independently of estrogen signaling. EVI1 and downstream MAPK activation might represent therapeutic targets in patients suffering from HER2⁻ ER⁻ or ER⁺ breast carcinoma resistant to antiestrogen therapies. Finally, targeting the newly identified EVI1-GPR54-KISS1 axis, for example by GPR54 inhibitors, may be considered for the treatment of metastasizing ER⁻ breast carcinoma. Effective targeting of EVI1-induced breast carcinoma cell migration might however require either inhibition of EVI1 itself or joint suppression of additional migratory pathways (e.g., RHO/ROCK signaling).

C. Lengerke reports receiving a commercial research grant from Roche Postdoctoral Fellowship Grant and has provided expert testimony for Celgene, Amgen, and Gilead (contributions to travel expenses to scientific conferences). No potential conflicts of interest were disclosed by the other authors.

Conception and design: H. Wang, T. Schaefer, C. Lengerke

Development of methodology: H. Wang, T. Schaefer, M. Konantz, C. Lengerke

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Wang, T. Schaefer, M. Konantz, M. Braun, Z. Varga, A.M. Paczulla, S. Reich, S. Perner, H. Moch, T.N. Fehm, L. Kanz, C. Lengerke

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Wang, T. Schaefer, M. Konantz, M. Braun, Z. Varga, A.M. Paczulla, F. Jacob, S. Perner, T.N. Fehm, K. Schulze-Osthoff, C. Lengerke

Writing, review, and/or revision of the manuscript: H. Wang, T. Schaefer, M. Braun, F. Jacob, S. Perner, H. Moch, T.N. Fehm, K. Schulze-Osthoff, Claudia Lengerke

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Braun, H. Moch, L. Kanz, C. Lengerke

Study supervision: H. Moch, C. Lengerke

We thank Olga Kustikova and Christopher Baum for provision of EVI1 plasmids, and Joelle Müller for technical support in zebrafish experiments.

This study was performed as contracted research supported by the Baden-Württemberg Stiftung (Adult Stem Cells Program II) and by grants from the Deutsche Forschungsgemeinschaft to S. Perner (PE 1179/4-1) and the Schweizer Nationalfonds to C. Lengerke (310030_149735).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.