Skip to main content
NCBI home page
Bioinformatics logoLink to Bioinformatics
. 2014 Jul 15;30(21):3036–3043. doi: 10.1093/bioinformatics/btu400

Modeling time-dependent transcription effects of HER2 oncogene and discovery of a role for E2F2 in breast cancer cell-matrix adhesion

Aliccia Bollig-Fischer 1,2,*, Luca Marchetti 3,4, Cristina Mitrea 5, Jiusheng Wu 1,2, Adéle Kruger 6, Vincenzo Manca 3, Sorin Drăghici 5,6
PMCID: PMC4201152  PMID: 25028721

Abstract

Motivation: Oncogenes are known drivers of cancer phenotypes and targets of molecular therapies; however, the complex and diverse signaling mechanisms regulated by oncogenes and potential routes to targeted therapy resistance remain to be fully understood. To this end, we present an approach to infer regulatory mechanisms downstream of the HER2 driver oncogene in SUM-225 metastatic breast cancer cells from dynamic gene expression patterns using a succession of analytical techniques, including a novel MP grammars method to mathematically model putative regulatory interactions among sets of clustered genes.

Results: Our method highlighted regulatory interactions previously identified in the cell line and a novel finding that the HER2 oncogene, as opposed to the proto-oncogene, upregulates expression of the E2F2 transcription factor. By targeted gene knockdown we show the significance of this, demonstrating that cancer cell-matrix adhesion and outgrowth were markedly inhibited when E2F2 levels were reduced. Thus, validating in this context that upregulation of E2F2 represents a key intermediate event in a HER2 oncogene-directed gene expression-based signaling circuit. This work demonstrates how predictive modeling of longitudinal gene expression data combined with multiple systems-level analyses can be used to accurately predict downstream signaling pathways. Here, our integrated method was applied to reveal insights as to how the HER2 oncogene drives a specific cancer cell phenotype, but it is adaptable to investigate other oncogenes and model systems.

Availability and implementation: Accessibility of various tools is listed in methods; the Log-Gain Stoichiometric Stepwise algorithm is accessible at http://www.cbmc.it/software/Software.php.

Contact: bollig@karmanos.org

Supplementary information: Supplementary data are available at Bioinformatics online.

1 INTRODUCTION

Cancer cells demonstrate unlimited and unmitigated proliferation and resistance to anti-proliferative and cell death cues, phenotypes that set them apart from normal cells. Underlying these cancer-specific phenotypes are aberrantly activated reconfigured signaling networks that control malignant growth and survival (Luo et al., 2009). Cancer-specific signaling networks and carcinogenesis are governed by the activation of key cancer driver genes, termed oncogenes. Based on this understanding, much progress has been made to develop anti-cancer drugs, with great clinical gains made through efforts to target oncogenes that possess kinase function; including the epidermal growth factor receptor family member HER2 in the treatment of breast cancer (BC; Sharma and Settleman, 2007). The advantages of molecularly targeted systemic treatments over traditional chemotherapy are clear: targeted treatments yield improved patient outcomes and lesser side effects (Jordan, 2008).

Despite advancements made to improve cancer treatments, targeted therapies eventually fail for patients, leading to developed drug resistance and cancer recurrence. A number of molecular mechanisms are described to underlie de novo and developed resistance to drugs. Included among them is the event of an activating mutation in an effector protein downstream of the targeted oncogenic kinase (Wagle et al., 2011). The complexity of oncogene activation and the prevalence of resistance underscore the fact that although it is appreciated that activated oncogenes are the dominant drivers of malignant progression in cancer, it remains unclear how activated oncogenes regulate the expression of transformed phenotypes. Beyond well-studied biochemical signaling pathways, it is especially true for yet unknown transcriptional regulatory networks. Clearly, elucidating the signaling networks regulated by oncogenes that result in cancer-specific phenotypes will provide insight to improve therapeutic strategies and overcome resistance.

To this end, biological processes are increasingly being described by network modeling of gene expression data. Various approaches have been used for purposes of investigating transcription networks to identify expression-based classifiers for cancer subtypes (Ramaswamy et al., 2001); or to integrate DNA-level genetic variation and expression data to infer causality (Jornsten et al., 2011); or to model before and after perturbation expression data to infer causality (Saeki et al., 2009). In step with these efforts, an ever-increasing variety of formalisms have been proposed for modeling and simulating gene-interaction networks (Bansal et al., 2007).

Here, we developed a multi-pronged approach to investigate HER2 oncogene signaling mechanisms using perturbed, time-dependent, gene expression data from microarray analysis. At this stage in the development of our modeling system, clustering of genes was required to reduce the dimensionality or complexity of the data. Our working premises were that genes displaying similar dynamic patterns of expression are similarly regulated and functionally related, a concept supported in the literature (Lockhart and Winzeler, 2000), and furthermore, that the expression level changes of a gene or genes in one cluster impacts the expression level changes of genes in another. We started with a method to cluster genes displaying similar expression dynamics after HER2 oncogene signaling was inhibited in the HER2 amplified metastatic BC cell line SUM-225 (Forozan et al., 1999; Kuperwasser et al., 2005). Then we derived mathematical models to support predicted regulatory relationships among clusters using a novel MP grammars method, which identifies algebraic notation that can simulate the dynamics of one cluster as a function of the gene expression level changes displayed by another cluster (Manca, 2008). Here, we have adapted the application of P systems that was developed for the purpose of expressing metabolic systems in discrete mathematical terms (Manca, 2008, 2013). A promoter analysis reinforced indications for relationships between clusters. A pathway impact analysis was used to identify cancer pathways/functions specific to a clustered gene set. All told, the analysis predicted a key intermediary role for the E2F transcription factor 2 (E2F2) in HER2 oncogene signaling. In follow-up functional studies to test the prediction, targeted E2F2 expression knockdown and 3D on-top Matrigel assays showed that matrix adhesion and outgrowth, which is indicative of cancer cell potential for metastatic colonization (Shibue and Weinberg, 2009), was significantly inhibited in SUM-225 cells when E2F2 levels were reduced.

2 METHODS

2.1 Overview of gene expression data

Gene expression data for MCF10A, MCF10HER2 and SUM-225 cells are from previous microarray analysis of global gene expression (Bollig-Fischer et al., 2011). Briefly described here, cultures were treated with the HER2 kinase inhibitor CP724 714 (1 µM), and RNA was isolated from parallel plated cultures at 0 h and every 3 h after addition of inhibitor for 45 h, yielding 16 time points. Expression levels were determined by microarray analyses using the Illumina human Ref8v2 array (Illumina, San Diego, CA). Data were processed for quality control in BeadStudio and quantile normalized, and are available through NCBI’s Gene Expression Omnibus accession numbers GSE22955, GSE23137 and GSE23138 (http://www.ncbi.nlm.nih.gov/geo).

2.2 Modeling gene cluster regulatory interactions

Raw microarray data with a detection P ≤ 0.01 were log2 transformed and quantile normalized. The normalized time series data were further filtered based on a curve fitting procedure that discards genes whose expression does not exhibit time-dependent change (Marchetti and Manca, 2012). For this purpose, we calculated polynomial models approximating the gene expression profile, similar to work by Ramoni et al. (2002), by using polynomial regression and least-square estimate. These models were used to filter out genes that exhibit constant or chaotic expression. The differences between the polynomial models and the gene profiles can be because of error or lack of fit. These two types of differences were squared and summed up to values that follow two independent chi-square distributions. A Fisher F-test (P < 0.001) was then used to select the best fit. Finally, genes that exhibited polynomial-like dynamics were filtered according to their minimum–maximum log2 fold-change to arrive at a gene set that well represents the three types of polynomial shapes considered: polynomial of first degree (linear profiles, log2 fold-change >1), of second degree (parabolic profiles, log2 fold-change >0.7) and third degree (cubic profiles, log2 fold-change >0.8). At the end of this phase, the number of time series decreased from >20 000 to 1175 (Supplementary Table S1). Filtered genes were clustered according to the similarity of their corresponding polynomial model using a method based on a classic hierarchical clustering technique (Eisen et al., 1998; Marchetti and Manca, 2012). The clustering algorithm was wrapped in an iterative procedure that ran the algorithm multiple times, arriving at different total numbers of clusters. Ultimately, the procedure chose the best division among those computed according to a user-defined threshold for the minimum number of genes in a single cluster (10 genes for this application). Interactions among clusters were then explored by deriving an MP grammar. This was achieved by application of the Log-Gain Stoichiometric Stepwise (LGSS) algorithm, which infers system parameters from discrete observations (Manca and Marchetti, 2011, 2012). The size of the systems of equations solved by LGSS depends on the number of variables and potential number of rules of the system under examination. In the specific case of the application presented in this article, the regression performed by LGSS calculated the MP grammar with a limitation on the rules, such that genes in clusters fitting with polynomials of lower degree (linear) were not regulated by genes fitting with polynomials of high degree (for example, cubic). The LGSS algorithm and user guide are accessible at http://www.cbmc.it/software/Software.php.

Promoter analysis of genes for experimentally verified transcription factor binding site over-representation used the BioBase ExPlain 3.1 tool and TRANSFAC® database (Beverly, MA). For this purpose, the promoter region included 500 base pairs on each side of the transcription start site. Gene-annotation enrichment analysis was done using the Genes-to-Systems Breast Cancer Database (http://www.itb.cnr.it/breastcancer/). The pathway impact analysis was implemented using the Pathway-Express analysis package (Khatri et al., 2005). Further explanation of MP grammars and MP grammar notation as well as BioBase, Gene-to-Systems database and Impact Factor Analysis is also made available in the Supplementary Description of Methods.

2.3 Cells, cell culture and E2F2 knockdown

SUM-225 cells were derived from a HER2-amplified BC described in (Forozan et al., 1999). The MCF10A human mammary epithelial cell line was developed at the Karmanos Cancer Institute, originally the Michigan Cancer Foundation (Miller et al., 1993). The authenticity of the cell lines was validated during the course of these studies by arrayed comparative genomic hybridization, confirming HER2 amplification in the SUM-225 cell line and lack thereof but amplification of FGFR1 in MCF10A (Marella et al., 2009). Stable lentivirus-mediated shRNA knockdown of E2F2 expression in SUM-225 and MCF10A cells was done using the Expression Arrest GIPZ lentiviral shRNAmir system following manufacturer instruction (Thermo Scientific, Huntsville, AL). GIPZ is a commercially available bicistronic vector that expresses green fluorescent protein and the shRNA of interest. Further information including culture conditions are detailed in Supplementary Description of Methods.

2.4 Cell proliferation, 3D on-top Matrigel assays, immunoblot analysis and real-time RT-PCR

Analysis of cell adhesion and outgrowth on BD Matrigel™ (via 3D on-top Matrigel assays) was done as described in (Shibue and Weinberg, 2009). Immunoblot analysis was done as previously described (Bollig-Fischer et al., 2010). Greater detail surrounding 3D on-top Matrigel assays, immunoblot analysis and real-time PCR that used TaqMan® primers (Life Technologies, Carlsbad, CA) are given in Supplementary Description of Methods.

3 RESULTS

3.1 Identifying genes regulated by the HER2 oncogene

The gene expression data that we analyzed in this study came from time-course perturbation experiments that used the SUM-225 BC cell line (Bollig-Fischer et al., 2011). The SUM-225 cell line is derived from a chest wall recurrence of ductal carcinoma, and SUM-225 cell proliferation is driven by HER2 amplification and overexpression (Forozan et al., 1999). The cells exhibit strong growth inhibition throughout treatment with the HER2-specific small molecule kinase inhibitor CP724 714, and growth inhibition is reversed when CP724 714 treatment is discontinued. Gene expression data were from Illumina platform-based microarray analysis of RNA collected at 16 time points over the course of 45 h, before and after HER2 signaling was blocked by treatment with CP724 714. During this time, cell counts were taken to observe that CP724 714 treatment was inhibiting the proliferation of the HER2-amplified/overexpressed BC cell line (Bollig-Fischer et al., 2010).

From the ∼20 000 genes analyzed by microarray, 1175 showed significant time-dependent expression-level changes in the SUM-225 cell line treated with CP724 714 (see Section 2 for filtering algorithm), suggesting they were regulated, directly or indirectly, by the HER2 oncogene. We found that the time-course expression signature for these genes could be reasonably divided into eight clusters (see Section 2 for clustering details) represented by their prototypical expression profile shown in Figure 1A: linear-quick-up, linear-slow-up, linear-quick-down, linear-slow-down, parabolic-up, parabolic-down, cubic-up-down and cubic-down-up. Genes and their assigned cluster are listed in Supplementary Table S1, and examples of gene expression dynamics that passed or failed our novel filtering procedure are shown in Figure 1B.

Fig. 1.

Fig. 1.

Open in a new tab

Defining the genes regulated by the HER2 oncogene and predicted gene cluster interactions. (A) HER2-regulated gene expression profiles fit eight clusters as determined by similarity of expression. The clusters are represented here by their prototypical expression profile (o) and the profile approximated by simulating the MP grammar (–). (B) Separating signal from noise. The applied filtering method yielded examples of time-course gene expression profiles that were included (upper panel) and excluded (lower panel) in data analysis. (C) A graphical representation of putative gene cluster regulatory interactions supported by the MP grammar. The graph edges represent predicted activating (arrow) or inhibiting (T intersect) relationships among clustered gene expression data

Using the filtered and clustered dataset, an MP grammar was calculated by means of a LGSS algorithm. The resulting MP grammar that predicts gene cluster regulatory interactions for our measured or observed data is defined by rules and regulator functions listed in Supplementary Figure S1A and in graphical form in Supplementary Figure S1B. The approximations of the prototypical expression profiles of each cluster, which were computed by simulating the MP grammar, are displayed superimposed onto the observed data in Figure 1A. Here, it was apparent that the MP grammar-simulated data resembles the observed data, thus reinforcing a putative mathematically defined regulatory interaction between clusters. An initial exploration of gene identities in each cluster was done using the Genes-to-Systems BC gene database (Mosca et al., 2010). This step highlighted genes associated with BC in the filtered dataset (listed in Supplementary Table S1).

3.2 Predicted gene cluster regulatory interactions correspond with mechanisms known to be activated in SUM-225 cells

Among the genes in the analysis linked to BC via the Genes-to-Systems BC gene database, two are already known to be of key importance for the HER2-amplified SUM-225 BC cell line—P38MAPK and ALDH1A3 (Charafe-Jauffret et al., 2009; Diehl et al., 2007; Korkaya et al., 2008)—identified in clusters numbered 5 and 8, respectively. Based on this, we elected to focus the current study on investigating mechanisms that may regulate these genes and/or others in the clusters. The putative gene cluster regulatory interactions supported by the MP grammar that involve cluster 5 and 8 are shown in Figure 1C.

Molecular biology approaches previously demonstrated that HER2 signaling in SUM-225 cells induces P38MAPK activity that is important for cancer cell anchorage-independent growth (Diehl et al., 2007). Here we report that HER2 signaling upregulated expression of P38MAPK, the MAPK13 isoform specifically, which was among genes assigned to cluster 5. To our knowledge, it is not previously reported that P38MAPK gene expression is regulated by the HER2 oncogene. Furthermore, MP grammar modeling predicted that genes in cluster 1 would negatively influence P38MAPK/cluster 5 gene expression (Fig. 1C). In line with that prediction, repression of the dual specificity phosphatase 1 gene (DUSP1, in cluster 1)—which promotes p38MAPK phosphorylation, activity and expression (Tibbles and Woodgett, 1999)—was observed in cluster 1.

ALDH1A3 is a member of a family of aldehyde dehydrogenase isozymes that are associated with stem cell-like characteristics in BC cells. Although ALDH activity in SUM-225 cells was previously reported (Charafe-Jauffret et al., 2009; Korkaya et al., 2008), our work here adds new understanding that ALDH1A3 mRNA expression is upregulated by an intracellular HER2 oncogene-directed signaling mechanism. The predicted signaling circuit that P38MAPK upregulates ALDH1A3 gene expression is already supported by the literature (Hanson et al., 2004; Norris and Baldwin, 1999). The knowledge of these expression-level targets for HER2 oncogene signaling is integrated into the MP grammar predicted regulatory interaction model shown in Figure 2.

Fig. 2.

Fig. 2.

Open in a new tab

The regulation of indicated genes according to the data modeling are consistent with known signaling in SUM-225 cells. ALDH1A3 and P38MAPK are known to be activated in SUM-225 cells. The graph indicates novel discoveries that ALDH1A3 and P38MAPK gene expression were upregulated (up arrows) by HER2 signaling and DUSP1 was downregulated (down arrow)

3.3 Deriving a novel functional prediction from the proposed gene cluster regulatory interactions

Moving past identifying support for what is already known for HER2 signaling in SUM-225 cells and what is reinforcing the robustness of the gene expression data, we next examined the prediction that the upregulation of P38MAPK-containing cluster (cluster 5) was also influenced by a gene or set of genes in cluster 3 as predicted by the MP grammar, depicted in Figure 1C. We tested the promoters of genes in the P38MAPK-containing cluster for transcription factor binding element enrichment. Results from this analysis using the BioBase ExPlain™ 3.1 tool (Matys et al., 2003), which uses the TRANSFAC® database of empirical evidence, showed that the P38MAPK-associated cluster was significantly enriched for genes that are regulated by E2F transcription factor family members. Of the 196 total genes in cluster 5, 140 were identified to contain an E2F factor binding motif (listed in Supplementary Table S1). The over-represented E2F binding matrices identified in the promoters of this clustered gene set are in Figure 3A. E2F matrices were over-represented and contributed 5 of the top 15 most significant binding motifs (Fig. 3A). We examined genes in the putative upstream cluster (cluster 3) and discovered the E2F transcription factor family member E2F2 in that cluster. No other E2F family member was among the total 1175 genes discovered to be regulated by HER2 signaling. Members of the E2F family share binding motifs and functional homology, and thus, the significance of the E2F1 binding sequence shown in Figure 3A can also be attributed to the action of E2F2 (Lees et al., 1993).

Fig. 3.

Fig. 3.

Open in a new tab

Analysis of the cluster 5 gene set and E2F2 expression. (A) Transcription factor binding site enrichment for genes in cluster 5. The results are based on analysis using the BioBase ExPlain™ 3.1 tool and TRANSFAC® database of empirical evidence. An enrichment of genes regulated by E2F family of transcription factors and multiple E2F consensus binding matrices (5 of the top 15) was significant. Columns list the transcription factor binding matrix overrepresented in the query set with respect to the background set (genes in all other clusters). Consensus sequence is the IUPAC consensus of the transcription factor binding matrix. The Yes/No field is the ratio of the average number of putative binding sites per 1000 bp for the query set (Yes) and the average number of putative binding sites per 1000 bp of the background set (No). The P-value is the significance of this representation. False discovery rate (FDR)-adjusted P-values are listed as FDR. (B) Genes in cluster 5 are functionally enriched. Results of Pathway Impact analysis ranking the most significantly enriched KEGG pathways for genes in cluster 5. Pathways are ranked according to highest IF with corrected P-value in parentheses. (C) E2F2 expression levels in patient specimens. The Oncomine™ data mining platform was used to query expression-level data for normal breast (n = 25) and breast tumor (n = 245) tissue made available by TCGA consortium. Results showed that E2F2 is upregulated 3.8-fold in invasive ductal breast carcinoma compared with normal breast tissue (P = 5.05–15)

We then investigated whether these similarly regulated clusters of genes were functionally integrated too. Results of pathway impact analysis, developed by us and described in (Draghici et al., 2007), indicated that they were and identified the most significant Kyoto Encyclopedia of Genes and Genomes (KEGG) signaling pathways as adherens junction and cell cycle and included pathways in cancer (Fig. 3B). The impact analysis was used as implemented in the Pathway-Express analysis package (Khatri et al., 2005). This technique identifies the most significantly affected canonical pathways by use of an impact factor (IF), calculated for each identified pathway. The IF incorporates parameters such as the normalized fold-change of the differentially expressed genes, the statistical significance of the set of pathway genes and the position and type of each gene in the given pathway (see Section 2 for a complete description). We further confirmed the significance of E2F2 specifically for BC with a query of The Cancer Genome Atlas (TCGA) normal breast and breast tumor-derived gene expression data using Oncomine™ (Rhodes et al., 2004). Results showed significant E2F2 overexpression in invasive ductal breast carcinoma compared with normal breast tissue (Fig. 3C, 3.8-fold increase in cancer, P = 5.05–15; 25 normal breast and 245 Invasive Ductal Carcinoma). Also, the TCGA data query showed that E2F2 expression levels ranked among the top 4% of BC genes.

3.4 Evidence for the importance of E2F2 in SUM-225 cells

To this point in our study, accumulating evidence suggested that E2F2 was a key intermediary in a HER2 oncogene-driven signaling circuit in SUM-225 cells. To test the significance of this, we designed experiments to examine whether targeted knockdown of E2F2 had a SUM-225 cancer cell-specific impact on proliferation. Lentiviral-based shRNA vectors and methods were used to stably knockdown E2F2 in the SUM-225 cell line and in the non-transformed breast epithelial cell line—MCF10A. The on-target effects of the shRNA vectors were confirmed by real-time RT-PCR (Supplementary Fig. S2). And consistent with a comparison of normal breast and breast tumor data from TCGA, immunoblot analysis of E2F2 in cell lysates showed that steady-state levels of E2F2 were lower in non-transformed MCF10A cells than in SUM-225 cancer cells. Also, according to immunoblot analysis, a substantial E2F2 knockdown was achieved using targeted shRNA vectors in both cell lines (Fig. 4A). Knockdown using either of two shRNA vectors against E2F2 had no effect on the proliferation of SUM-225 or MCF10A cells growing on tissue culture plasticware (Supplementary Fig. S3).

Fig. 4.

Fig. 4.

Open in a new tab

Targeted E2F2 gene knockdown effects on SUM-225 cancer cell growth. (A) Immunoblot analysis of E2F2 protein levels in cell lines transduced with two unique E2F2-targeted shRNA expressing vectors with respect to control levels (lanes 1 and 4). (B) The effect of E2F2 knockdown in SUM-225 cells cultured on Matrigel. E2F2 knockdown caused a marked reduction in viable colony numbers and size (4× magnification). The figure is representative of two independent experiments, in triplicate. Viable cells were stained dark purple with MTT (3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) reagent. (C) Quantitation of the effect of E2F2 knockdown in SUM-225 cells cultured on Matrigel. The graph shows the total number of viable transduced colonies per well of a 96-well plate. Results are the mean of three biological replicates and representative of replicated experiments (bars, SD)

We then proceeded to test the impact of E2F2 knockdown on SUM-225 cells cultured on BD Matrigel™ in a 3D on-top model system. As such, Matrigel is used as a substrate to test cancer cell extracellular matrix adhesion and outgrowth, and positive results predict a metastatic phenotype (Ivanova et al., 2013; Shibue and Weinberg, 2009). Non-transformed cells like MCF10A form small growth-limited structures at low efficiencies on Matrigel (Chakraborty et al., 2012; Pradeep et al., 2012). In contrast, metastatic cancer cells like SUM-225 take shape as large-sized expanding colonies and at relatively high efficiency when cultured on Matrigel, as in Figure 4B left panel. Single cell suspensions were plated equally, and results after 7 days showed that E2F2 knockdown in SUM-225 cells reduced the numbers of adherent colonies and caused a marked reduction in colony size (Fig. 4B). The experiment using Matrigel was done twice comprising three biological replicates of cultures independently transduced with two unique E2F2-targeted shRNA sequence vectors. Each time, for all replicates of each vector, the results were the same, supporting a highly confident assessment that HER2 oncogene-regulated E2F2 did not simply promote proliferation but that E2F2 had a role in promoting the cancer cell-specific matrix adhesion and outgrowth. The biologically validated role of E2F2 is annotated in the updated HER2 oncogene signaling model in Figure 5.

Fig. 5.

Fig. 5.

Open in a new tab

Gene-annotated HER2-oncogene regulated signaling circuits predicted by the proposed network model. The indicated genes and their role in HER2 oncogene signaling are supported by previously published data (blue), or as in the case of E2F2, they are supported by computational and in vivo cell biology experiments presented in this study (orange). Black arrows indicate that according to our data, the gene expression was upregulated (up arrow) or downregulated (down arrow) by HER2 signaling

3.5 Evidence for a HER2-E2F2 cancer-specific signaling axis

To gain greater insight into the SUM-225 cancer cell-specific effect of E2F2 knockdown, we compared the time-course gene expression microarray data for E2F2 in non-transformed MCF10A cells and SUM-225 BC cells, both before and after treatment with HER2 inhibitor CP724 714. By comparison, in the context of SUM-225 cells where HER2 functions as a driver oncogene, markedly high levels of E2F2 expression were observed and were dependent on HER2 signaling (Fig. 6, t = 0). The relatively high expression levels in SUM-225 cells were consistent with relatively high steady-state E2F2 protein levels in lysates from SUM-225 cells compared with MCF10A cells (Fig. 4A). Analysis of expression data from MCF10A cells transduced to stably overexpress HER2 (referred to as MCF10HER2 cells) showed that forced overexpression alone did not result in the high E2F2 expression levels observed in the cancer cells (Fig. 6). Moreover, treatment of all three cell lines with the HER2 inhibitor produced a HER2-regulated E2F2 time-dependent expression signature in SUM-225 cancer cells that was distinct from an E2F2 expression signature that was similar for MCF10A and MCF10HER2 cells (Fig. 6).

Fig. 6.

Fig. 6.

Open in a new tab

Cancer cell-specific HER2 oncogene regulation of E2F2 expression. Time-course E2F2 expression data before and after treatment with the HER2-targeted inhibitor CP724 714. Time-dependent (hours, H) RNA levels from BC-derived cell line SUM-225 and non-transformed MCF10A and MCF10A overexpressing HER2 (MCF10HER2) cell lines demonstrate existence of a unique HER2-dependent expression profile in cancer cells

4 DISCUSSION

The success of driver oncogene-targeted therapeutic strategies can be attributed to the fact that oncogene-directed signaling promotes the many phenotypes that define the patho-biology of cancer cells, termed the hallmarks of cancer that include among others unrestricted growth, immune system evasion, metabolic transformation, cancer cell tumorigenicity and the various stages of metastases (Hanahan and Weinberg, 2011). The challenge remains to understand the specifics for oncogene-directed signaling processes that contribute to each of these phenotypes.

In earlier work using engineered cell lines to identify genes whose expression is regulated by the driver HER2 oncogene versus the proto-oncogene, we explored gene expression dynamics using a Bayesian clustering method (Bollig-Fischer et al., 2010). This previous analysis revealed overlapping and unique clusters of genes for each condition, suggesting that driver oncogenes engage unique regulatory mechanisms that cause distinguishing gene expression dynamics and signaling circuits in cancer cells. The work presented here represents considerably advanced efforts to discover unknown regulatory paths in transcription programs. The use of cell lines, where the functional status of HER2 is known, is necessary to distinguish and model gene expression features that were regulated, directly or indirectly, by HER2 driver oncogene function. Time-course perturbation studies were used because rigorous statistical analyses of gene expression changes over time can be applied to filter noise (Ideker et al., 2011; Venet et al., 2011); also, algorithms that reverse engineer gene networks from dynamic gene expression data outperform classic clustering algorithms (Bansal et al., 2007). Polynomial models were used for the purpose of curve fitting analysis and were also used in the calculation of the eight clusters, all necessary to reduce data complexity. This methodology might be improved by using more complex clustering algorithms based on machine learning techniques such as hidden Markov models or Support Vector Machines, which hold potential to distinguish more refined clusters.

Nevertheless, in the present iteration, our approach using simple polynomial models to investigate HER2 oncogene signaling networks in SUM-225 BC cells resulted in a model with features that were concordant with both previous reported results and data from patient tumors. The integration of various analytical tools such as Impact Factor analysis, BioBase/Explain and Oncomine, with the results of MP grammar modeling supported specific predicted regulatory circuits and pointed to novel insights surrounding oncogene regulation of P38MAPK, ALDH1A3 and E2F2 genes in the HER2-amplified BC-derived cell line. Definitive contributions of this study are that HER2 driver oncogene signaling upregulated the expression of E2F2, and that E2F2 potentiated the HER2 oncogene signal for maintaining a transformed phenotype.

The overwhelming focus of existing literature surrounding the E2F family of transcription factors has characterized their function as master regulators of cell proliferation, but Chen and his colleagues are correct to point out that knowledge is lacking as to the biological processes that they collectively or individually regulate in cancer (Chen et al., 2009). Among the limited literature specifically on E2F2 and its aberrant role in BC, work by Miller et al. showed that an E2F2 activation signature correlates with resistance to anti-estrogen treatment for estrogen receptor alpha (ERα)-positive BC (Miller et al., 2011). The study herein now indicates the importance of E2F2 in a novel HER2 oncogene-directed signaling pathway where HER2 oncogene-regulated E2F2 expression impacts cell-matrix adhesion function, with potential consequences for metastatic colonization (Shibue and Weinberg, 2009).

Another consideration arising from our study surrounds the nature of oncogene activation and the status of HER2 as a driver oncogene. Simply overexpressing the HER2 oncogene in the non-transformed MCF10A cell did not cause the dominant regulatory effect for HER2 action on E2F2 expression as seen in SUM-225 cells. These observations support the concept that the status and function of a driver oncogene is not simply the result of signal amplification but that additional context-dependent variables are involved to evoke aberrant pathway signaling.

Funding: United States Department of Defense (W81XWH-10-2-0068) and National Institutes of Health (NIH P30CA022453) awarded to the Karmanos Cancer Institute at Wayne State University. Additional funding support was provided by the NIH (RO1 DK089167 and STTR R42GM087013 to S.D.) and the National Science Foundation (DBI-0965741 to S.D.).

Conflicts of Interest: none declared.

Supplementary Material

Supplementary Data
supp_30_21_3036__index.html (1KB, html)

REFERENCES

  1. Bansal M, et al. How to infer gene networks from expression profiles. Mol. Syst. Biol. 2007;3:78. doi: 10.1038/msb4100120. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Bollig-Fischer A, et al. HER-2 signaling, acquisition of growth factor independence, and regulation of biological networks associated with cell transformation. Cancer Res. 2010;70:7862–7873. doi: 10.1158/0008-5472.CAN-10-1529. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Bollig-Fischer A, et al. Oncogene activation induces metabolic transformation resulting in insulin-independence in human breast cancer cells. PLoS One. 2011;6:e17959. doi: 10.1371/journal.pone.0017959. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Chakraborty G, et al. Semaphorin 3A suppresses tumor growth and metastasis in mice melanoma model. PLoS One. 2012;7:e33633. doi: 10.1371/journal.pone.0033633. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Charafe-Jauffret E, et al. Breast cancer cell lines contain functional cancer stem cells with metastatic capacity and a distinct molecular signature. Cancer Res. 2009;69:1302–1313. doi: 10.1158/0008-5472.CAN-08-2741. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Chen HZ, et al. Emerging roles of E2Fs in cancer: an exit from cell cycle control. Nat. Rev. Cancer. 2009;9:785–797. doi: 10.1038/nrc2696. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Diehl KM, et al. p38MAPK-activated AKT in HER-2 overexpressing human breast cancer cells acts as an EGF-independent survival signal. J. Surg. Res. 2007;142:162–169. doi: 10.1016/j.jss.2007.01.025. [DOI] [PubMed] [Google Scholar]
  8. Draghici S, et al. A systems biology approach for pathway level analysis. Genome Res. 2007;17:1537–1545. doi: 10.1101/gr.6202607. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Eisen MB, et al. Cluster analysis and display of genome-wide expression patterns. Proc. Natl Acad. Sci. USA. 1998;95:14863–14868. doi: 10.1073/pnas.95.25.14863. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Forozan F, et al. Molecular cytogenetic analysis of 11 new breast cancer cell lines. Br. J. Cancer. 1999;81:1328–1334. doi: 10.1038/sj.bjc.6695007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell. 2011;144:646–674. doi: 10.1016/j.cell.2011.02.013. [DOI] [PubMed] [Google Scholar]
  12. Hanson JL, et al. The nuclear factor kappaB subunits RelA/p65 and c-Rel potentiate but are not required for Ras-induced cellular transformation. Cancer Res. 2004;64:7248–7255. doi: 10.1158/0008-5472.CAN-03-3898. [DOI] [PubMed] [Google Scholar]
  13. Ideker T, et al. Boosting signal-to-noise in complex biology: prior knowledge is power. Cell. 2011;144:860–863. doi: 10.1016/j.cell.2011.03.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Ivanova IA, et al. FER kinase promotes breast cancer metastasis by regulating α6- and β1-integrin-dependent cell adhesion and anoikis resistance. Oncogene. 2013;32:5582–5592. doi: 10.1038/onc.2013.277. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Jordan VC. Tamoxifen: catalyst for the change to targeted therapy. Eur. J. Cancer. 2008;44:30–38. doi: 10.1016/j.ejca.2007.11.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Jornsten R, et al. Network modeling of the transcriptional effects of copy number aberrations in glioblastoma. Mol. Syst. Biol. 2011;7:486–503. doi: 10.1038/msb.2011.17. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Khatri P, et al. Recent additions and improvements to the Onto-Tools. Nucleic Acids Res. 2005;33:W762–W765. doi: 10.1093/nar/gki472. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Korkaya H, et al. HER2 regulates the mammary stem/progenitor cell population driving tumorigenesis and invasion. Oncogene. 2008;27:6120–6130. doi: 10.1038/onc.2008.207. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Kuperwasser C, et al. A mouse model of human breast cancer metastasis to human bone. Cancer Res. 2005;65:6130–6138. doi: 10.1158/0008-5472.CAN-04-1408. [DOI] [PubMed] [Google Scholar]
  20. Lees JA, et al. The retinoblastoma protein binds to a family of E2F transcription factors. Mol. Cell. Biol. 1993;13:7813–7825. doi: 10.1128/mcb.13.12.7813. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Lockhart DJ, Winzeler EA. Genomics, gene expression and DNA arrays. Nature. 2000;405:827–836. doi: 10.1038/35015701. [DOI] [PubMed] [Google Scholar]
  22. Luo J, et al. Principles of cancer therapy: oncogene and non-oncogene addiction. Cell. 2009;136:823–837. doi: 10.1016/j.cell.2009.02.024. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Manca V. The metabolic algorithm for P systems: principles and applications. Theor. Comput. Sci. 2008;404:142–155. [Google Scholar]
  24. Manca V. Infobiotics: Information in Biotic Systems. Vol. 3. Berlin Heidelberg: Springer; 2013. [Google Scholar]
  25. Manca V, Marchetti L. Log-gain stoichiometric stepwise regression for MP systems. Int. J. Found. Comput. Sci. 2011;22:97–106. [Google Scholar]
  26. Manca V, Marchetti L. Solving dynamical inverse problems by means of metabolic P systems. Biosystems. 2012;109:78–86. doi: 10.1016/j.biosystems.2011.12.006. [DOI] [PubMed] [Google Scholar]
  27. Marchetti L, Manca V. A methodology based on MP theory for gene expression analysis. Lect. Notes Comput. Sci. 2012;7184:300–313. [Google Scholar]
  28. Marella NV, et al. Cytogenetic and cDNA microarray expression analysis of MCF10 human breast cancer progression cell lines. Cancer Res. 2009;69:5946–5953. doi: 10.1158/0008-5472.CAN-09-0420. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Matys V, et al. TRANSFAC: transcriptional regulation, from patterns to profiles. Nucleic Acids Res. 2003;31:374–378. doi: 10.1093/nar/gkg108. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Miller FR, et al. Xenograft model of progressive human proliferative breast disease. J. Natl Cancer. Inst. 1993;85:1725–1732. doi: 10.1093/jnci/85.21.1725. [DOI] [PubMed] [Google Scholar]
  31. Miller TW, et al. ERalpha-dependent E2F transcription can mediate resistance to estrogen deprivation in human breast cancer. Cancer Discov. 2011;1:338–351. doi: 10.1158/2159-8290.CD-11-0101. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Mosca E, et al. A multilevel data integration resource for breast cancer study. BMC Syst. Biol. 2010;4:76. doi: 10.1186/1752-0509-4-76. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Norris JL, Baldwin AS., Jr. Oncogenic Ras enhances NF-kappaB transcriptional activity through Raf-dependent and Raf-independent mitogen-activated protein kinase signaling pathways. J. Biol. Chem. 1999;274:13841–13846. doi: 10.1074/jbc.274.20.13841. [DOI] [PubMed] [Google Scholar]
  34. Pradeep CR, et al. Modeling ductal carcinoma in situ: a HER2-Notch3 collaboration enables luminal filling. Oncogene. 2012;31:907–917. doi: 10.1038/onc.2011.279. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Ramaswamy S, et al. Multiclass cancer diagnosis using tumor gene expression signatures. Proc. Natl Acad. Sci. USA. 2001;98:15149–15154. doi: 10.1073/pnas.211566398. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Ramoni MF, et al. Cluster analysis of gene expression dynamics. Proc. Natl Acad. Sci. USA. 2002;99:9121–9126. doi: 10.1073/pnas.132656399. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Rhodes DR, et al. ONCOMINE: a cancer microarray database and integrated data-mining platform. Neoplasia. 2004;6:1–6. doi: 10.1016/s1476-5586(04)80047-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Saeki Y, et al. Ligand-specific sequential regulation of transcription factors for differentiation of MCF-7 cells. BMC Genomics. 2009;10:545–561. doi: 10.1186/1471-2164-10-545. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Sharma SV, Settleman J. Oncogene addiction: setting the stage for molecularly targeted cancer therapy. Genes Dev. 2007;21:3214–3231. doi: 10.1101/gad.1609907. [DOI] [PubMed] [Google Scholar]
  40. Shibue T, Weinberg RA. Integrin beta1-focal adhesion kinase signaling directs the proliferation of metastatic cancer cells disseminated in the lungs. Proc. Natl Acad. Sci. USA. 2009;106:10290–10295. doi: 10.1073/pnas.0904227106. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Tibbles LA, Woodgett JR. The stress-activated protein kinase pathways. Cell. Mol. Life Sci. 1999;55:1230–1254. doi: 10.1007/s000180050369. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Venet D, et al. Most random gene expression signatures are significantly associated with breast cancer outcome. PLoS Comput. Biol. 2011;7:e1002240. doi: 10.1371/journal.pcbi.1002240. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Wagle N, et al. Dissecting Therapeutic Resistance to RAF Inhibition in Melanoma by Tumor Genomic Profiling. J. Clin. Oncol. 2011;29:3085–3896. doi: 10.1200/JCO.2010.33.2312. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Data
supp_30_21_3036__index.html (1KB, html)
supp_btu400_Supplementary_Description_of_Methods-revised.pdf (429.3KB, pdf)

Articles from Bioinformatics are provided here courtesy of Oxford University Press

RESOURCES