Mammary carcinomas developing in SV40 transgenic WAP-T mice arise in two distinct histological phenotypes: as differentiated low-grade and undifferentiated high-grade tumors. We integrated different types of information such as histological grading, analysis of aCGH-based gene copy number and gene expression profiling to provide a comprehensive molecular description of mammary tumors in WAP-T mice. Applying a novel procedure for the correlation of gene copy number with gene expression on a global scale, we observed in tumor samples a global coherence between genotype and transcription. This coherence can be interpreted as a matched transcriptional regulation inherited from the cells of tumor origin and determined by the activity of cancer driver genes. Despite common recurrent genomic aberrations, e.g. gain of chr. 15 in most WAP-T tumors, loss of chr. 19 frequently occurs only in low-grade tumors. These tumors show features of “basal-like” epithelial differentiation, particularly expression of keratin 14. The high-grade tumors are clearly separated from the low-grade tumors by strong expression of the Met gene and by coexpression of epithelial (e.g. keratin 18) and mesenchymal (e.g. vimentin) markers. In high-grade tumors, the expression of the nonmutated Met protein is associated with Met-locus amplification and Met activity. The role of Met as a cancer driver gene is supported by the contribution of active Met signaling to motility and growth of mammary tumor-derived cells. Finally, we discuss the independent origin of low- and high-grade tumors from distinct cells of tumor origin, possibly luminal progenitors, distinguished by Met gene expression and Met signaling.
Mouse models provide important tools for gaining insight into the biology of breast cancer (BC) and the relationship between histopathology and molecular profiles of cancer subtypes.1, 2 They narrow down the complexity of the disease and allow defining the influence of specific individual molecular factors and pathways on BC development and progression—information that is difficult to obtain from retrospective analyses of human tumor samples. Mammary epithelial cell-specific expression of cellular and viral oncogenes in transgenic mice thus is a suitable approach to study molecular mechanisms of mammary tumor initiation and progression.2
We previously described a transgenic mouse model for mammary tumorigenesis based on murine whey acidic protein (WAP) promoter-dependent expression of the SV40 early gene region in mammary epithelial cells of BALB/c mice.3, 4 The Wap gene (and correspondingly the transgenic WAP promoter) is transiently and weakly expressed during the estrous cycle,5, 6 but strongly induced in late pregnancy with a maximum during lactation.6 Owing to the activity of the WAP promoter in mammary epithelial cells of adult females, mammary gland development in transgenic mice is normal. Mammary carcinogenesis in WAP-T mice is driven by the synergistically acting large (LT) and small (sT) antigens encoded by SV40 early gene region.7 When individually expressed in transgenic mice, the tumor-initiating efficiency of sT is considerably weaker than that of LT, and the fully malignant tumor phenotype typical for coexpression of both antigens is never reached.8 The transforming activity of LT is mainly mediated by direct binding of LT to the p53 and pRb tumor-suppressor proteins. Functional inactivation of p53 results in abrogation of the p53 functions critical for regulating cell cycle checkpoints and for the maintenance of genomic stability, with a consequence that cells become aneuploid and prone to increased recombination and gene amplification.7 The LT-mediated release of pocket proteins from repressive complexes recruited to E2F-responsive promoters drives the cells into S-phase through activation of cell cycle-associated genes.7 In addition, upregulation of cell cycle genes, e.g. Foxm19 negatively regulated by p53, contributes to cell transformation and establishment of a proliferation signature typical for SV40-induced tumors.10, 11
In primiparous WAP-T mice, the LT/st antigens induce multifocal intraepithelial neoplasia in all mammary glands. A very small fraction of these focal lesions further progresses to invasive, but rarely metastatic mammary carcinomas.3, 4 These SV40-induced mammary tumors exhibit either a basal-like, morphologically differentiated phenotype (low-grade tumors), or an undifferentiated phenotype (high-grade tumors) featured by coexpression of epithelial and mesenchymal markers, respectively.12 In comparison to other oncoproteins, e.g. ETV6-NTRK3, whose transgenic expression under control of the WAP promoter results in high tumor incidence already in nulliparous mice,13 strong expression of SV40 proteins during pregnancy and lactation does not result in immediate tumor outgrowth. The WAP-T mouse lines are characterized by a moderately long latency time of up to 8 months after first pregnancy, despite 100% penetrance.3, 4 Similar kinetics of pregnancy-induced tumor development was observed by WAP-SV40 transgene expression in NMRI outbred mice.14 In contrast, in the C57BL/6J inbred background, a low tumor incidence and long latency after multiple pregnancies was observed despite strong LT/sT expression in mammary epithelial cells of the lactating gland.15
The specific gene expression profiles of mammary tumors induced by SV40 proteins in the NMRI genetic background11 and in C57BL/6J WAP-Tag mice16 have been already reported. However, the molecular mechanisms leading to the development of low- and high-grade tumor phenotypes in WAP-T BALB/c mice so far remained unknown. In our study, we aimed to provide an in-depth molecular portrait of the SV40-induced mammary tumors in BALB/c mice and to identify the putative cancer driver genes. Combining the analysis of aCGH-based gene copy number with gene expression profiling, we describe here a set of common and unique molecular features distinguishing high-grade from low-grade tumors. Particularly, recurrent amplification of the Met gene locus and overexpression of an active Met protein kinase accompanied by coexpression of epithelial and mesenchymal proteins in tumor cells are discerning features of WAP-T high-grade tumors. The clear genomic and transcriptomic differences between low- and high-grade WAP-T tumor phenotypes strongly indicate that these tumors develop independently from distinct cells of tumor origin, presumably mammary luminal progenitors, where LT exerts its transforming activity.
Material and Methods
Inbred BALB/c and transgenic WAP-SV40 mouse lines, T1 and NP8,4 were housed under SPF conditions in accordance with official regulations for care and use of laboratory animals (UKCCCR Guidelines for the Welfare of Animals in Experimental Neoplasia) and approved by Hamburg's Authority for Health (Nr. 24/96).
Histology, immunohistochemistry and immunofluorescent staining of cryosections
Processing of paraffin-embedded tissue specimens and cryosections was performed as previously described.12 At least three mammary glands per mouse from each transgenic line were analyzed, and the largest tumor within one mammary gland cross section was evaluated by a pathologist. Detailed procedures are provided in Supporting Information.
Fluorescence in situ hybridization
Fluorescence in situ hybridization (FISH) using probes spanning the Met and Cntn3 (used as internal control) gene loci was performed on touch preparations of tumor samples according to a standard protocol. Detailed procedures are provided in Supporting Information.
Met pathway activity
Detailed procedures are provided in Supporting Information.
Detailed procedures are provided in Supporting Information.
Blood vessel density analysis
Blood vessels were stained with rat monoclonal anti-CD31 antibody (Biolegend; Mec 13.3; 102502) on cryosections of mammary gland and tumor tissue. The 2D images (z = 1.8 μm) were obtained using a laser confocal microscope (Zeiss, LSM 510 Meta, ×10 objective) and were subjected to a particle analysis using the ImageJ software (NIH, v. 1.41o). Only CD31+ particles with a size ≥25 μm2 were considered as blood vessels. Finally, the results were processed in GraphPad Prism (GraphPad software, v. 5.04) and analyzed by a t-test.
Oligo array CGH profiling
Genomic DNA from tumor samples (Supporting Information Table S1) and liver was isolated using the Puregene Genomic DNA Purification Kit according to the manufacturer's instructions (Gentra). aCGH analysis was performed by hybridization of Cy3- and Cy5-labeled samples on the Agilent-015028 Mouse Genome CGH Microarray 44K (Supporting Information Table S2) and Oligator “MEEBO” mouse genome set (Supporting Information Table S3) arrays. Laboratory work and scanning was done at the VU University Medical Center (Amsterdam) and University of Oslo. The genomic positions of 70-mer oligonucleotides on the MEEBO array were updated on the basis of the NCBI Build 37 mouse genome assembly. Oligo sequences were mapped via Ensembl DAS server, GALAXY tools17 and nondistinct mappings curated using the annotation published by the Alizadeh Laboratory (updated version 051705, Stanford University). The obtained log2-ratios (sample vs. reference sample) were Loess normalized (package limma, version 2.12.10). Smoothing and segmentation was done with DNAcopy (version 1.24) using default parameter settings. Segments were classified as gain or loss when the segment mean signal was above or below a 2.5-fold of a robust standard deviation of the 50% quantile of the unsegmented log2-ratios. In analogy, an aberration was classified as amplification or deletion if the absolute segment mean exceeded an 8-fold of the standard deviation. The robust standard deviation of the log2-ratios was calculated as the Tukey's biweight estimator over the standard deviations determined for the single chromosomes. For calculation of aberration frequencies in low- and high-grade tumors, sample 10667-2 was excluded as a histological outlier (atypical grade G2 subtype with areas of G3 grade morphology).
Expression profiling and bioinformatic analysis
Homogenization of mammary gland tissue and tumor samples (Supporting Information Table S1) was performed by Lysin Matrix D Tubes (Qbiogene). Total RNA was extracted using Trizol Reagent (Invitrogen) followed by DNA digestion (TURBO DNA-free; Ambion) according to the manufacturer's instructions. Total RNA was cleaned up using RNeasy mini kit according to the manufacturer's instructions (Qiagen). Labeling, hybridization on the Affymetrix microarray chips (MOE430 2.0) and image data processing were completed by Signature Diagnostics AG (Potsdam) or at the Institute for Clinical Chemistry (Hamburg) according to the Affymetrix standard protocol. Detailed bioinformatic procedures of data processing are provided in Supporting Information.
Integration of gene expression and aCGH data
Global sample correlation of data
For a sample-wide correlation between aCGH and expression data, a z-score-like transformation of the normalized expression signals was performed such that, in the best case, the aberration profiles observed in the aCGH analysis for the complete sample could be visualized completely on the expression level. In analogy to a z-score, the signals of each probe set were shifted by a constant reference value and divided by the probe set signal standard deviation over all samples. The probe set-specific reference values were determined as robust Tukey's biweight estimators for each probe set over all samples. This value should represent some kind of an imaginary “normal level” signal of the probe set. The transformed expression profiles were then analyzed with DNAcopy and compared with the corresponding aCGH profiles.
Driver gene detection
Driver genes were identified by a Welch's t-test comparison of low-grade and high-grade tumor gene expression profiles with the reference samples (normal mammary gland 50 days postweaning). Raw p-values were corrected for multiple testing using Holm's sequential procedure. Differentially expressed genes (corrected p-value below threshold of 0.05 and signal-log-ratio below or above 1) were then screened for the observation of at least a massive amplification (segment mean above or below signal of ±1) in the corresponding low-grade or high-grade tumor group.
The raw data used in our work have been made available in the GEO repository with accession numbers: GSE31212 super series [composed of GSE29117, GSE33038 (gene expression data), GSE31097 (Agilent CGH data)] and GSE7135 (MEEBO CGH data).
WAP-T low- and high-grade tumors show distinct molecular profiles
The generation and characteristics of the SV40 transgenic WAP-T mouse lines (collectively called WAP-T) developed in our laboratory has been described previously.4 The SV40-induced tumors (termed WAP-T tumors) are low-grade (G1 and G2, differentiated histotype) and high-grade (G3 and G4, undifferentiated histotype) adenocarcinomas3 that differ in the degree of mammary-specific differentiation. As the transplanted tumors usually recapitulate the phenotype of the parental tumors,12 we reasoned that each tumor phenotype is defined by phenotype-specific common genetic alterations. Indeed, aCGH analysis of 25 tumor samples (Supporting Information Tables S2 and S3), representing the four histological grades (Fig. 1a), revealed a significant, mouse line independent recurrence of large-scale chromosomal aberrations. Although some alterations are shared by low- and high-grade tumors, e.g. gain of chr. 15 (Figs. 1b and 1c), several aberrations clearly correlate with tumor grade, e.g. loss of chr. 2 and chr. 19 in low-grade tumors (Fig. 1b), and gain of chr. 1 in high-grade tumors (Fig. 1c). Some of these aberrations constitute massive amplifications or deletions (Supporting Information Table S4) and demonstrate a significant association with tumor grade. For instance, all high-grade tumors feature a variably sized amplicon in chromosomal region 6qA2 centered round the Met gene (Fig. 1c, Supporting Information Table S4). Such a Met-containing amplicon occurs only in one low-grade tumor (sample 10667-2) that represented an atypical grade G2 subtype with areas of grade G3 morphology (not shown).
In addition to recurrent genetic alterations, we identified significant coherence between gene copy number and gene expression on a global scale (Fig. 1d, Supporting Information Fig. S1). In this analysis, we applied a novel procedure including a targeted, genome-wide integration of adjusted values from genomic and transcriptomic datasets using data smoothing, segmentation and segment calling steps. As a result, we show that large-scale (Fig. 1e) and local (Fig. 1f) copy number alterations are mirrored at the level of gene transcription. This relation is frequently dosage dependent, e.g. the copy number changes in chr. 18 and chr. 19 are clearly visible at the transcriptomic level in sample 12030-4 (Fig. 1d). Similarly, the gain in chr. 2qC3-qH4, the loss in chr. 6qA1 and the amplification of the Met-locus in sample 12212-2 can be derived from gene expression (Fig. 1d). A certain discordance observed in some tumors (Supporting Information Fig. S1) can be attributed to differing cellular composition and a bias introduced by contribution of the stromal compartment.
The findings emphasize two novel insights regarding SV40-induced mammary carcinomas. First, recurrent genetic alterations indicate phenotype specific progression pathways of WAP-T tumors. Second, a tight linkage between genome composition and gene expression points to a balanced transcriptional regulation, which might be determined by cells of tumor origin and the activity of driver genes.
Met is a driver gene in high-grade WAP-T tumors
Driver genes play a key role in defining the molecular mechanisms underlying the observed tumor phenotype.18 Their consistent tumor specific upregulation or downregulation is frequently accompanied by copy number aberrations (CNA, gene amplifications and deletions) as a consequence of chromosomal region fragility. To identify putative driver genes, we correlated the expression levels of genes within amplicons and deletions with their expression in the reference samples (involuted mammary gland of a BALB/c mouse). In total, eight, respectively, of 13 CNA-associated genes (Supporting Information Table S5) were identified in low- and high-grade tumors, as putative driver genes [adjusted p-value (Holm) < 0.05]. Except for the Met gene, all other putative driver genes are associated with a single or maximally two tumor samples (A430107O13Rik, Cav1 and Srfbp1). Met, however, is by far the most common and consistent CNA-associated gene in high-grade tumors and thus most likely drives their development. The development of low-grade tumors, in contrast, is independent of Met overexpression and Met gene amplification. It obviously proceeds without involvement of recurrent local CNAs.
In accordance with the aCGH data, DNA FISH analysis with Met-specific probes revealed a massive amplification of the Met-locus in a high-grade tumor, but not in a low-grade tumor (Fig. 2a). Aside from Met, these amplicons contain Cav1, Cav2, Capza2, St7, Tes and several other neighboring genes as well (Fig. 2b, left). They demonstrate a large variability in amplicon borders, but are always centered round the Met gene. Except for two samples (11155-7 and 10870-6), gene copy number in the Met-locus correlates with the level of gene expression (Fig. 2b, right), indicating that increased Met expression usually results from Met-locus amplification. In agreement with the gene expression data, a large fraction of cells in high-grade tumors show strong Met immunostaining in contrast to weak Met expression in low-grade tumors (Fig. 2c). The strong expression of Met coincides with prominent phosphorylation at the multifunctional docking site (Tyr1347, corresponding to human Tyr1349) (Fig. 2c), indicating active Met-signaling in high-grade tumors. This is in line with our finding that the Met catalytic domain is mutation-free as determined by sequence analysis (not shown). The intactness of Met-signaling is further supported by several experiments performed in the previously described WAP-T-derived cell line G-2.12 This cell line, derived from an undifferentiated tumor of a WAP-T/mutp53 bi-transgenic mouse, also expresses Met protein lacking mutations in the catalytic domain (not shown), and upon syngeneic transplantation forms tumors that closely resemble high-grade tumors of WAP-T monotransgenic mice.12 Using a small molecule, ATP-competitive Met tyrosine kinase inhibitor (PHA-665752), phosphorylation at Tyr1347 was dramatically reduced (Fig. 3a). Moreover, using PHA-665752, we observed a strong reduction of transphosphorylation of the tyrosine pair Tyr1232/Tyr1233 (corresponding to human Tyr1234/Tyr1235) within the kinase activation loop and an increase in phosphorylation at the juxtamembrane tyrosine residue Tyr1001 (corresponding to human Tyr1003) probably involved in clathrin-dependent receptor internalization19 (Fig. 3b). Moreover, the signaling cascade downstream of Met, e.g. Akt phosphorylation, is significantly affected by inhibiting Met activity by PHA-665752 (Fig. 3b). In accordance with these biochemical data, cell motility, a process controlled by Met signaling,20 is reduced in a wound healing assay by treatment with the PHA-665752 inhibitor (Fig. 3c). Finally, blocking Met activity by PHA-665752 significantly reduced accumulation of cells in the culture (Fig. 3d, left) by abrogation of multilayered cell growth (Fig. 3d, right). The “dome-like” structures typically observed in the control G-2 cultures became very rare in the cultures treated with PHA-665752 inhibitor. This observation points to a contribution of Met signaling to cell density dependent growth control in a standard adherent culture. In analogy, under anchorage-independent conditions Met activity is required for expansion of cell colonies. In a methylcellulose colony formation assay, we observed a significant decrease (from 112 ± 65 μm to 40 ± 22 μm; p-value of 4.6E-14) of colony diameter after seeding of the singularized cells into methylcellulose supplemented with inhibitor PHA-665752 (Fig. 3e). Overall, these data are in line with a role of the Met pathway as a regulator of invasive growth21 and strongly support the necessity for Met signaling for the SV40 protein-induced mammary tumorigenesis.
Expression of Met mirrors the differentiation state of WAP-T tumors
The level of Met gene transcription, alone or in combination with usually concomitant amplification of the Met-locus, is a surrogate marker for the differentiation states of WAP-T tumors. By grouping the samples according to the Met gene expression level (i.e.Metlow and Methigh groups), we generated a “WAP-T/Met” signature (Fig. 4, Supporting Information Table S6). The Metlow tumors (usually low grade with the exception of sample 10870-6) are distinguished by expression of the basal marker Krt14 (Figs. 5a–5b), the absence of EMT features (e.g. vimentin) in tumor cells (shown for example in a low-grade tumor in Wegwitz et al.12) and a higher expression of 18 genes belonging to the cell junction category (Supporting Information Table S6). Another basal marker, keratin 5, a polymerization partner of keratin 14, is only rarely expressed in tumor cells (Fig. 5a). Accordingly, we termed this phenotype “quasi-basal.” The Methigh tumors (usually high grade with exception of the samples 10667-2 and 11155-7) show a feature of epithelial-mesenchymal plasticity (EMP), namely coexpression of epithelial keratins 8/18 and keratin 6 (Figs. 5a–5b) with the mesenchymal protein vimentin (shown for example in a high-grade tumor in Wegwitz et al.12). The EMP phenotype of Methigh tumors is particularly associated with the upregulation of 28 cell development genes and three genes related to pyrimidine metabolism, Upp1, Uck2 and Dctd (Supporting Information Table S6), indicating ongoing developmental processes in these tumors and the necessity for enhanced pyrimidine metabolism. In light of the contribution of active Met signaling to growth and motility of tumor derived G-2 cells, it is likely that the EMP phenotype is determined by a reinforced activity of the Met pathway as well as by enhanced expression of transcriptional regulators controlling the EMT process, particularly the Etv122 and Hmga223 genes (Supporting Information Table S6).
An important factor in the EMT process is a moderate hypoxic microenvironment24 created by insufficient vascularization. In WAP-T tumors, however, the impact of such hypoxic microenvironments on development of the main phenotypes seems rather minor. As shown by CD31 staining of endothelial cells, blood vessel density is higher in the richly vascularized normal mammary gland at different developmental stages (virgin, Day 14 of gestation, and Day 60 of involution) than in low- and high-grade WAP-T tumors (Figs. 6a–6b). Here, the CD31 staining is confined to tumor stroma, and is only slightly higher in high-grade than in low-grade tumors. Independent of tumor grade, poor vascularization of WAP-T tumors correlates with weak expression of genes linked to endothelial differentiation (Fig. 6c).
In conclusion, these results suggest that a hypoxic microenvironment plays a rather minor role in development of the two main WAP-T tumor phenotypes. However, interestingly, features of mesenchymal differentiation occur only in tumor epithelial cells of the EMP phenotype, not in cells of the quasi-basal phenotype, providing a hint that the two main WAP-T tumor phenotypes might have developed independently from distinct cells of tumor origin.
Tumors developing in the WAP-T mouse model feature two distinct phenotypes of SV40 LT-induced adenocarcinomas that are relatively homogeneous within each individual group while exhibiting major molecular intergroup differences. A global coherence between gene copy number and transcription indicates the existence of tight regulatory mechanisms that underlie the interdependent evolution of genotype and transcriptome after the initial, LT-mediated genetic lesions. In line, we assume that chromosomal loss is functionally equivalent to gene repression. Functional analysis of genes located on chr. 19 that is frequently lost in low-grade tumors, supports this hypothesis. According to DAVID-based functional annotation of chr. 19 genes (not shown) 11 and 19 genes, respectively, are associated with the fatty acid biosynthetic process (GO: 0006633) and mitochondrial function (GO: 0044429). Considering that these genes are expected to be active in differentiated mammary epithelial cells that are specialized for secreting activity, loss of these genes in tumor cells might be a plausible alternative for their epigenetic repression.
In contrast, we consider the chromosomal gains and local amplifications as a compensatory mechanism for insufficient gene activation. The observation that ectopic expression of the Myc gene suppresses gain of chr. 15 (contains Myc gene) in the mouse AML model25 supports the proposed link between regulation of transcription and gene dosage. Such dependence on transcription level probably plays a role in the origin of local amplifications, particularly in amplification of the Met locus. In mice, the Met locus is frequently amplified in mammary tumors induced in the Brca1Δ11/co Trp53+/− background,26 in the Metmut knock-in mice,27 in the K14-Cre/p53 conditional knockout mice,28 in the transgenic mice conditionally expressing human PIK3CAH1047R in mammary cells29 and in mice with mammary specific knockout of the Lfng gene.30 It thus seems that in mammary epithelial cells, the Met locus is highly susceptible to amplification, which is most likely initiated by an increased demand for Met transcription during the initial steps of mammary tumorigenesis. Although direct evidence for this assumption is lacking, there is some indirect support. Particularly, transgenic expression of activated Met (Metmut) in a knock-in mouse model induces mammary tumors and Met locus amplification.27 Similarly, amplification of the Erbb2 locus occurs in mammary tumors induced by expression of activated Neu under the endogenous Neu promoter.31 As a possible mechanism for local massive amplifications in the Met gene locus, we suggest a chain of events favored by a loss of functional p53 and driven by enhanced transcription. As a consequence, topoisomerase-mediated torsional tension,32 collision of transcription and replication complexes33 and formation of anaphase bridges34 may potentially contribute to amplicon generation. Interestingly, the human MET gene promoter is negatively regulated by p53, which through transcriptional suppression of MET regulates motility of epithelial cells.35 Loss of the repressive function of p53 on the Met promoter would deregulate Met transcription and contribute to genetic instability in this locus.
Although expression of the mutated Met gene induces mammary tumors in transgenic mice,27, 36 mutations in the human MET gene are extremely rare in BC: 1% of the cases documented in the COSMIC database.37 Also MET amplification is not detected in human BC27, 38 despite the fact that common fragile sites (FRA7G in human39 and Fra6A3.1 in mouse40) encompass the mouse Met and the human MET genes. However, the level of MET expression is high in basal-like and in HER2+ subtypes of human BC,36 and correlates with poor clinical outcome and to a greater degree with the negative ER status of human BC samples.27 Therefore, high expression and activity of nonmutated Met in high-grade WAP-T tumors renders the WAP-T model superior to transgenic mouse models with expression of mutated Met,27, 36 thereby providing a suitable preclinical model for BC with a focus on the oncogenic expression of nonmutated Met.
As Met gene transcription and Met-signaling activity clearly distinguish tumors with an EMP phenotype from tumors with a quasi-basal phenotype, we suppose that the respective cells of tumor origin are featured by an active Met pathway, whereas quasi-basal tumors initiate and progress from cells either with constrained or absent Met activity, because these tumors exhibit a low Met expression level and weak Met phosphorylation. Although the scenario of a limited Met activity and signaling in the precursors of these tumors is most plausible, further studies are required to dissect in mammary glands the cellular basis of Met signaling. In the meantime, it is known that transgenic expression of mutated Met in mammary epithelial cells induces a heterogeneous spectrum of mammary tumors, whose phenotype and gene expression profiles resemble the features of human basal-like BC.27, 36 Furthermore, the coexpression of EMT markers with oncogenic Met in the mixed pathology phenotype36 points to a functional crosstalk between Met signaling and the EMT process in mammary carcinoma. In WAP-T mice, the high expression of EMT regulators such as Hmga223 and Etv122 in high-grade tumors is a strong indication for an intersection of Met activity with the process of EMT. Mouse Etv1 (Er81) is expressed in the mesenchymal cells surrounding growing mammary ducts41 as well as in cap cells of terminal end buds and in luminal cells,42 indicating that this transcription factor might be particularly involved in the regulation of developmentally controlled EMT in mammary gland and be a downstream effector of Met signaling. In mouse mammary glands, Met expression is high in luminal progenitors43 and Met signaling is involved in the regulation of ductal side branching44—a complex morphogenetic process that includes bidirectional EMT at branching points.45 Interestingly, progesterone-induced ductal side branching is more extensive in BALB/c mice than in C57BL/6 mice.46 This observation correlates with a low tumor incidence and a long latency after multiple pregnancies in comparison to our WAP-T BALB/c mice—despite strong SV40 LT/sT-expression in the lactating gland in C57BL/6J mice.15 This fact points to a dependence of the tumor initiation frequency on the extent of ductal side branching. It is also likely that Met signaling at the side branching points acts synergistically with the epistatic effects of modifier genes, e.g. mutations in Cdkn2a and Prkdc genes,47 to promote mammary tumorigenesis in BALB/c mice.
In contrast to high-grade tumors both active Met signaling and features of EMT (e.g. expression of vimentin) are absent in tumor cells of low-grade tumors, which however are distinguished by high expression of 18 cell junction-related genes. Among them the Gjb2 gene is particularly interesting. This gene encodes the gap junction protein connexin 26 (Cx26) that has an opposite, context-dependent influence on breast tumorigenesis,48 probably acting as a factor regulating the epithelial differentiation state of tumor cells. The low-grade tumors are also featured by recurrent loss of chr. 19, which in addition to the already mentioned metabolism related genes carries the Pten gene, a negative regulator of the PI3K/Akt pathway. Loss of Pten gene in mammary epithelial cells reduces the latency of Erbb2-induced tumors and enhances metastasis via overactivated PI3K/Akt signaling.49 Considering that PI3K/Akt signaling is active in both low- and high-grade WAP-T tumors (data not shown), it is suggestive that tumor phenotype-specific constellations of negative and positive regulatory effects may boost this pathway: while loss of Pten in the low-grade tumors may be responsible for a deficiency in negative regulation of this pathway, the active Met signaling may drive activity of the PI3K/Akt signaling in the high-grade tumors. Further studies are necessary to explore the signaling networks operating in the low- and high-grade WAP-T tumors. Owing to multiple occurrences of tumors in the mammary gland of WAP-T mice, the syngeneic transplantation12 of individual, molecularly characterized tumors is the strategy for using this transgenic mouse as a model to study the tumor-relevant signaling pathways, particularly Met and its downstream targets.
The authors thank Julia Abe, Andrea Diesterbeck, Jasmin Oehlmann and Annette Preuss for their extensive and reliable work in the animal facility, Gundula Pilnitz-Stolze for competent technical assistance and the staff of the HPI animal quarters for their help. The authors also thank Kristin Klätschke and Paul P. Eijk for technical assistance in the gene expression and aCGH microarrays and Dr. Frauke Krepulat for providing the mouse cohorts and for help in tumor preparation and classification. This study was supported by grants from the Deutsche Forschungsgemeinschaft (De 212/23-1-3 to W.D. and G.T.), the Deutsche Krebshilfe (Forschungsverbund “Tumorstammzellen”) to W.D. and G.T., the Fonds der Chemischen Industrie to W.D. and the EU (DISMAL). The Senior Professorship of W.D. is supported by the Jung-Foundation for Science, Hamburg. The Heinrich-Pette-Institute is financially supported by the Freie und Hansestadt Hamburg and the Bundesministerium für Gesundheit.