Cancer Letters 356 (2015) 880–890

Contents lists available at ScienceDirect

Cancer Letters j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / c a n l e t

Original Articles

A pathway-based approach for identifying biomarkers of tumor progression to trastuzumab-resistant breast cancer Seungyoon Nam a, Hae Ryung Chang a, Hae Rim Jung a, Youme Gim a, Nam Youl Kim b, Regis Grailhe b, Haeng Ran Seo c, Hee Seo Park d, Curt Balch e, Jinhyuk Lee f, Inhae Park g, So Youn Jung g, Kyung-Chae Jeong h, Garth Powis i, Han Liang j, Eun Sook Lee a, Jungsil Ro g, Yon Hui Kim a,* a

New Experimental Therapeutics Branch, National Cancer Center, Goyang-si, Gyeonggi-do 410-769, Republic of Korea Core Technology, Institut Pasteur Korea, Bundang-gu, Seongnam-si, Gyeonggi-do 463-400, Republic of Korea c Functional Morphometry II, Institut Pasteur Korea, Bundang-gu, Seongnam-si, Gyeonggi-do 463-400, Republic of Korea d Animal Sciences Branch, National Cancer Center, Goyang-si, Gyeonggi-do 410-769, Republic of Korea e Bioscience Advising, Indianapolis, IN 46227, USA f Korean Bioinformation Center (KOBIC), Korea Research Institute of Bioscience and Biotechnology, Daejeon 305-806, Republic of Korea g Center for Breast Cancer, National Cancer Center of Korea, Goyang-si, Gyeonggi-do 410-769, Republic of Korea h Biomolecular Function Research Branch, National Cancer Center, Goyang-si, Gyeonggi-do 410-769, Republic of Korea i Cancer Center, Sanford-Burnham Medical Research Institute, La Jolla, CA 92037, USA j Department of Bioinformatics and Computational Biology, The University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA b

A R T I C L E

I N F O

Article history: Received 24 September 2014 Received in revised form 30 October 2014 Accepted 30 October 2014 Keywords: Trastuzumab HER2 Biomarker discovery Drug resistance Breast cancer

A B S T R A C T

Although trastuzumab is a successful targeted therapy for breast cancer patients with tumors expressing HER2 (ERBB2), many patients eventually progress to drug resistance. Here, we identified subpathways differentially expressed between trastuzumab-resistant vs. -sensitive breast cancer cells, in conjunction with additional transcriptomic preclinical and clinical gene datasets, to rigorously identify overexpressed, resistance-associated genes. From this approach, we identified 32 genes reproducibly upregulated in trastuzumab resistance. 25 genes were upregulated in drug-resistant JIMT-1 cells, which also downregulated HER2 protein by >80% in the presence of trastuzumab. 24 genes were downregulated in trastuzumabsensitive SKBR3 cells. Trastuzumab sensitivity was restored by siRNA knockdown of these genes in the resistant cells, and overexpression of 5 of the 25 genes was found in at least one of five refractory HER2 + breast cancer. In summary, our rigorous computational approach, followed by experimental validation, significantly implicate ATF4, CHEK2, ENAH, ICOSLG, and RAD51 as potential biomarkers of trastuzumab resistance. These results provide further proof-of-concept of our methodology for successfully identifying potential biomarkers and druggable signal pathways involved in tumor progression to drug resistance. © 2014 The Authors. Published by Elsevier Ireland Ltd. This is an open access article under the CC BYNC-SA license (http://creativecommons.org/licenses/by-nc-sa/3.0/).

Introduction Worldwide, breast cancer kills over 521,000 women annually, associated with over 1.7 million new cases [1]. Approximately 25% of cases express the human epidermal growth factor receptor-2 (HER2, ERBB2), a 185-kDa transmembrane tyrosine kinase. HER2 has no known ligand, but plays a major role in the progression of breast and other cancers [2]. A humanized monoclonal antibody against HER2, trastuzumab (Herceptin®), was approved for the treatment

* Corresponding author. Tel.: +82 31 920 2519; fax: +82 31 920 2542. E-mail address: [email protected] (Y.H. Kim).

of HER2-positive breast cancer in 1998 [3]. The HERA (Herceptin Adjuvant) phase III trial showed that in early HER2-positive breast cancer patients, one-year trastuzumab therapy following surgery resulted in 32% greater disease-free survival [4]. However, an additional 1 year of adjuvant trastuzumab therapy showed no additional benefit [5]. While the therapy is initially effective against metastatic breast cancer, drug-resistant relapse is almost universal [6]. Due to the cost vs. benefit of trastuzumab, and the emerging study of other HER2-targeted therapies, drug-response-predictive biomarkers are strongly needed. A number of studies have now reported various possible cellular/ molecular mechanisms of trastuzumab resistance, largely from comparing sensitive vs. resistant breast cancer cell lines. Of these is overactivity of the mitogenic PI3K/Akt pathway. In particular,

http://dx.doi.org/10.1016/j.canlet.2014.10.038 0304-3835/© 2014 The Authors. Published by Elsevier Ireland Ltd. This is an open access article under the CC BY-NC-SA license (http://creativecommons.org/licenses/bync-sa/3.0/).

S. Nam et al./Cancer Letters 356 (2015) 880–890

activating mutations in PI3K and loss of its antagonist, PTEN, have been identified in both preclinical [7,8] and clinical [9,10] studies of trastuzumab-resistant breast cancer. However, another study, examining both trastuzumab-treated and -untreated patients over three years, found no association of resistance with either of these PI3K pathway members [11]. Other reported resistance contributors include DARPP, HSP90, cytosolic sequestration of the cell cycle inhibitor CDKN1B (p27KIP1), and downregulated tubulin III [3,12,13]. In addition to the above-described intracellular signal mediators, several cell surface transmembrane proteins and kinases have been implicated in trastuzumab resistance via either indirect or direct interaction with HER2. These include the hyaluronate receptor CD44, MUC4, c-MET, integrins α6β1 and α6β4, and HER2 heterodimerization with other EGFR family members [3]. As CD44 expression combined with low or absent CD24 has been reported to demarcate breast “cancer stem cells” [14] it has also been hypothesized that trastuzumab resistance may arise from a failure to eradicate tumor progenitors [3]. Analogously, the cancer stem cellassociated epithelial-to-mesenchymal transition has also been implicated in trastuzumab resistance [15]. In the current study, rather than examining HER2 signal effectors or pathways in isolation, we used a systems biology [16–18] approach, and applied our recently established subpathway identification [16] and network permutation method [17], which resulted in identification of 32 upregulated KEGG subpathway genes common to two preclinical and two clinical datasets. This discovery method allows the consideration of pathway “edges” as well as individual “nodes” [16]. The roles of these genes in trastuzumab resistance were experimentally validated by their upregulation in refractory HER2 + breast cancer cells and tumors. Based on these results, we believe our subpathway and network permutation approach is effective for identifying biomarker genes and pathways responsible for resistance to trastuzumab and other targeted, antineoplastic agents.

Materials and Methods Trastuzumab-resistant network and its reproducibly upregulated, resistance-associated genes To delineate a trastuzumab resistance network, we applied an association rule [16], based on gene expression and KEGG pathways, to identify all possible subsets of signaling pathways. This rule was applied to a preclinical dataset [GEO: GSE15043] composed of two BT474 drug-sensitive parent cells and eight BT474 daughter subclones selected for resistance by continuous growth in 0.2 and 1.0 μM trastuzumab [19]. The subsets, i.e., subpathways, included not only nodes but also edges such as activation or inhibition among the nodes [16]. The subpathways associated with drug resistance are represented by a network (Fig. 1), visualized using Cytoscape [20]. This network consisted of 4502 subpathways, with 916 gene entries. The statistical significance of the network was measured by “the mean correlation over all edges in the network”, henceforth designated as “r” [17,21], to describe interactions between entities, rather than a single entity [22]. A correlation of an edge was calculated from the Pearson correlation coefficient of the expressions between two neighboring nodes. The r value was obtained from the mean of the correlations of all the edges. With the network topology conserved, we randomly permuted the node labels in the network to obtain the null distribution of the statistic r. The true mean correlation prior to permutation was set to r0. We performed 5000 permutations, and the empirical p-value was calculated from Pr(r > r0). Reproducible gene entries belonging to the network were identified using multiple trastuzumab-resistant preclinical ([GEO: GSE15376] [23] [ArrayExpress: E-TABM-157] [24]) and resistance-associated clinical datasets. We also compared Herceptin-nonresponsive and -responsive cells from the two preclinical datasets. The HER2-positive control cells, SKBR3 and BT474 are Herceptinresponsive, while two HER2-negative control cell lines, MCF7 and MDA-MB-231 are Herceptin-nonresponsive [25]. Additional clinical datasets were used to assess survival associated with gene expression profiles in HER2 + recurrent breast cancer from The Cancer Genome Atlas (TCGA) [26] dataset under the heading “Breast Invasive Carcinoma (TCGA, Nature 2012)” provided by cBioPortal [27], including 47 disease-free HER2 + patients vs. 7 recurred HER2 + patients, and the Netherlands Cancer Institute dataset contained 25 disease-free

881

HER2 + patients vs. 24 recurrent HER2 + patients [28]. These four trastuzumab resistance-associated upregulated gene intersections (Fig. 2) resulted in the identification of 32 genes (Table 1) with their corresponding KEGG pathways listed in Appendix S1: Supplementary Table S1, including “Hallmarks of Cancer” in Appendix S1: Supplementary Table S2. We performed overall survival analyses, based on 66 TCGA HER2 + breast cancer patients. The 66 patients were selected based on PAM50 classification that was annotated in the dataset TCGA_BRCA_exp_HiSeqV2-2014-05-02 (column PAM50Call_RNAseq in file clinical_data) from the UCSC Cancer Genomics Browser (https://genome-cancer.ucsc.edu) [29]. The expression subset of the 32 genes among the 66 patients was extracted from the dataset, and we divided the patients into two groups (high, low) in terms of the expression of each gene in the subset: gene expression greater than the median expression of the patient was denoted as “high;” otherwise it was denoted “low.” Subsequently, we performed log-rank tests between the two groups for each gene. We also generated Kaplan– Meier curves for the gene MMP9 as well as PER2, a gene of interest due to the recent emergence of circadian rhythm dysfunction as a risk factor for breast cancer [30].

Procurement and processing of tumor tissues Patients consented to the use of the tissue specimens for research purposes, as approved by the Institutional Review Board of National Cancer Center, Republic of Korea (Ilsan, Korea). Histologic classification and tumor stage were reviewed by a pathologist in the Department of Pathology at the National Cancer Center. Included in our analysis were eight breast tumor samples. Clinical characteristics of 6 out of the 8 patients, including Herceptin sensitivity, are summarized in Table 2 (Appendix S1: Supplementary Fig. S1A and S1B).

Cell lines and reagents We obtained human breast cancer cell lines SKBR3 and JIMT-1 cells from the American Type Culture Collection (ATCC, Manassas, VA, USA) and The German Collection of Microorganisms and Cell Cultures (DSMZ, Braunschweig, Germany). ATCC and DSMZ uses short tandem repeat profiling to authenticate all cell lines [31]. Reagents used on SKBR3 and JIMT-1 breast cancer cells, including a HER2 monoclonal antibody, were purchased from Abgent (San Diego, CA).

Cell culture SKBR3 and JIMT-1 human breast cancer cell line studies were performed within six months of cell resuscitation. SKBR3 cells were grown in McCoy’s 5A medium (Sigma-Aldrich, St. Louis, MO) and JIMT-1 cells were grown in DMEM (Hyclone, Logan, UT) with 10% FBS (Hyclone) at 37 °C in 5% CO2. Cells (2.5 × 105) were seeded and grown under normoxic conditions to 70–80% confluency, and then incubated with or without 10 μg/ml trastuzumab [32] for up to 4 days, according to the required time study.

Western blotting Cells grown with or without 10 μg/ml trastuzumab were harvested, washed twice in phosphate-buffered saline solution, lysed, and subjected to Western blot, as we previously described [33]. The blots were quantified using Image Lab software (BioRad, Hercules, CA).

Real-time RT-PCR analysis Total RNA was isolated from cell lysates using Isol-RNA Lysis Reagent (5PRIME, Hamburg, Germany) and processed with ReverTra Ace® qPCR RT Master Mix with gDNA Remover Kit (Toyobo, Osaka, Japan) to synthesize cDNA. Quantitative PCR was conducted using iQ™ SYBR® Green Supermix (Bio-Rad), according to the manufacturer’s protocol, using a CFX384 TouchTM Real-Time PCR Detection System (BioRad). Primer sequences were obtained from RTPrimerDB (http://rtprimerdb.org/), designed via GenScript Real-time PCR (TaqMan) Primer Design (https:// www.genscript.com/ssl-bin/app/primer) or manually designed, as indicated in Appendix S1: Supplementary Table S3. Measurements of all 32 genes were initially analyzed. Twenty-five met the following quality standards: (1) No template control of primer sets must read a null (no signal detected) Ct value. (2) Resulting Ct value must be between 32 and 37. Normalization procedures and folds-change were carried out using β-actin, as we have previously reported [33] using the 2(-delta-delta C(T)) method as the internal reference [34]. P-values for tumor samples were calculated between each trastuzumab-resistant patient sample with each of the trastuzumabresponding patient sample. Trastuzumab-resistant patient samples with statistically significantly (i.e. p < 0.05 with all three trastuzumab-responding patients)

882

S. Nam et al./Cancer Letters 356 (2015) 880–890

Fig. 1. The subpathway-network relating to the trastuzumab resistance. We applied our previous systems biology approach [16] to trastuzumab-resistant cells versus trastuzumabresponsive cells [19] in order to obtain all the subpathways relating to resistance. The network consisting of the 4502 subpathways is represented. The network implicated in resistance shows not only multiple pathways but also complex cross-talk among the pathways. The yellow shaded oval indicates “crosstalk” from which upregulated genes represented in Appendix S1: Supplementary Fig. S3A. In this figure, we noted 19 KEGG pathways involved in resistance; the pathways are provided in detail in Appendix S1: Supplementary Fig. S3.

increased gene expression have been indicated with an asterisk (*). All measurements were performed in triplicate.

Results Trastuzumab-resistance network and resistance-associated genes

siRNA knockdown of trastuzumab-treated cells 4000 (SKBR3) or 2000 (JIMT-1) cells were plated in 384-well plates (Greiner BioOne, Monroe, NC). Cells were plated in triplicate and 6 images were taken per well. siRNA was purchased from Thermo Scientific (Waltham, MA) and incubated at a concentration of 50 nM. 0.5 μl of DharmaFECT1 was used per well. After 48 hr siRNA treatment, cells were treated for 48 hr with 10 μg/ml trastuzumab, fixed (4% paraformaldehyde), stained with HCS Cellmask (Molecular Probes, Eugene, OR), and nuclei stained with Draq5 (Cell Signaling Technology, Danvers, MA) for viability counting. OPERA (Opera High Content Screening System, PerkinElmer, Waltham, MA) was used for imaging and cell counting. PRISM (Version 5) software was used for analysis of the siRNA data.

To identify candidate biomarkers or “signatures” of trastuzumabresistant breast cancer, we applied a systems biology approach, using multiple datasets. First, we obtained a trastuzumab-resistance global network by applying our published systems biology approach [16] to BT474 drug-sensitive parent cells and four BT474 resistant daughter subclones [GEO: GSE15043] [19]. The network consisted of 129 edges and 916 resistance-related genes. The permutation-based test of the network was statistically significant (p-value 0.000; Appendix S1: Supplementary Fig. S2). The network (Fig. 1) revealed various

S. Nam et al./Cancer Letters 356 (2015) 880–890

883

datasets, we identified 32 genes which were upregulated in association with trastuzumab resistance.

Characteristics and clinical relevance of the 32 resistance-associated genes and their associated pathways

Fig. 2. Reproducibly upregulated genes, common among multiple datasets, and validation of the drug resistance network. The network of subpathways inferred from GSE15043 was compared to two other preclinical trastuzumab resistance datasets ([GEO: GSE15376] [23] and [ArrayExpress: ETABM-157] [24]) and two clinical datasets (the TCGA [26] and the Netherlands Cancer Institute (NKI) [28]) to establish a validated predictor gene set. The comparison result (blue box in the Venn diagram) consists of 32 genes commonly upregulated in trastuzumab resistance. Each set in the Venn diagram represents commonly upregulated genes (resistance-related) between its corresponding dataset and the network genes. The Venn diagram was generated by Venny (bioinfogp.cnb.csic.es/tools/venny/).

As shown in Table 1, the 32 genes commonly upregulated in trastuzumab resistance encode a diverse group of proteins, including oncoproteins (TIAM1, FRAT1), signal transducers (STAT3, SMAD1), and catenin family members of cadherin-binding proteins (CTNNA1, JUP), in association with the “hallmarks” of cancer [35] (Appendix S1: Supplementary Tables S1 and S2). Further, some of these 32 genes reside in pathways previously implicated in HER2 signaling, including MMP9 and MAPK10 [3,13,38,39], as corroborated by our bioinformatics identification of the MAPK, TGF-beta, and Wnt KEGG pathway groups (Appendix S1: Supplementary Fig. S3C–E). We also identified a guanine nucleotide exchange factor gene, TIAM1, that in the brain, recruits α6β1, an integrin thought to interfere with trastuzumab binding to HER2 [3]. Interestingly, we also identified PER2, a component gene of the circadian clock oscillatory system regulated by p53. Circadian clock dysfunction, another identified KEGG pathway, hsa04710 (Appendix S1: Supplementary Fig. S3B) has been recently linked to breast carcinogenesis [40]. Overall survival analysis for all the 32 genes was performed by log-rank test (Appendix S1: Supplementary Table S4), and the two genes (PER2, MMP9) were associated with prognosis (Appendix S1: Supplementary Fig. S4A and B). We also identified PTPRF, encoding a protein phosphatase, in the trastuzumab resistance dataset, in similarity to previous studies implicating another phosphatase, DARPP, in trastuzumab resistance (through the PI3K pathway) [12]. Interestingly, PTPRF is regulated by EGFR signaling [41]. We also identified several mediators of inflammatory responses, including STAT3, ICOSLG, and RPS6KA5 (cytokine-cytokine receptor interactions, KEGG pathway hsa04672, Appendix S1: Supplementary Fig. S3E). Inflammation has been linked to breast cancer, largely in association with overexpression of specific cytokines secreted by infiltrating monocytes [42].

Experimental validation of overexpression of the 25 least-studied trastuzumab-resistance genes in cell lines and tumor specimens signaling pathways, e.g., tight junctions, axon guidance, etc., not previously implicated in trastuzumab resistance. Also, the pathways closely interconnected by “crosstalk” genes (light yellow-shaded ovals in Fig. 1; for upregulated, red-dotted crosstalk genes, refer to expanded view in Appendix S1: Supplementary Fig. S3A). As shown in Fig. 1, the total network comprised 19 distinct KEGG pathways (expanded views in Appendix S1: Supplementary Fig. S3B–F), including pathways related to the “hallmarks of cancer” [35] (Appendix S1: Supplementary Table S2). These computational findings likely reflect clonal selection of a tumor cell subpopulation utilizing signal pathways autonomous of the drugspecific target (in this case, HER2), similar to breast or prostate cancer hormone independence [36,37]. Subsequently, we further explored reproducible gene entries belonging to the network throughout multiple trastuzumab-resistance preclinical ([GEO: GSE15376] [23], [ArrayExpress: E-TABM-157]) [24] and clinical datasets. In particular, the clinical datasets were inspected in terms of HER2-positive recurrent breast cancer vs. extended disease-free survival, from The Cancer Genome Atlas [26] and The Netherlands Cancer Institute [28]. Through analyzing these

For further characterization of the 32 genes, we used two well-established breast cancer cell models of trastuzumab resistance: the trastuzumab-resistant cell line JIMT-1 and the -sensitive cell line SKBR3 [32,43]. Drug sensitivity and HER2 expression of these cells was confirmed by time-dependent growth curves, showing only a 15% loss of viability of JIMT-1 resistant cells, following a four-day culture in 10 μg/ml trastuzumab. The widely used SKBR3 cell line, by contrast, exhibited >30% loss of viability after four days of trastuzumab treatment, compared to no treatment (Fig. 3A), in agreement with a previous study demonstrating 10 μg/ml trastuzumab as a “saturable” dose effective for SKBR3 sensitivity [32]. Moreover, JIMT-1 cells demonstrated >90% downregulation of HER2 protein, after each day of trastuzumab treatment, and by day 4, even untreated JIMT-1 cells showed 60% downregulation (Fig. 3B). In the same time period, SKBR3 cells showed negligible changes in HER2 expression. These results agree with previous studies showing HER2 internalization or “shedding” of its extracellular domain (thus precluding antibody binding and/or halting inhibition of normal HER2 pathway signaling) as mechanisms of trastuzumab resistance [3,43].

884

S. Nam et al./Cancer Letters 356 (2015) 880–890

Table 1 List of 32 trastuzumab resistance-associated genes. The genes in italic were filtered out to further experimental validation due to quality control (see Materials and Methods). Gene

Location

Gene product/function

ARHGEF12 ATF4 CCL22 CHEK2 CTNNA1 CYCS EGF EGLN2 ENAH EPAS1 FARP2 FES FRAT1 ICOSLG (ICOSL) JAM3 JUP LIMK2 MAPK10 MAPKAPK5 MMP9 NFATC4 PER2 PTPRF RAD51 RPS6KA5 SMAD1 STAT3 TIAM1 TICAM1 (TRIF) TJP2 TPM1 VAMP4

11q23.3 22q13.1 17q11.2 22q12.1 5q31.2 7p15.3 4q25 19q13.2 1q42.12 2p21-p16 2p37.3 15q26.1 10q24.1 21q22.3 11q25 17q21 22q12.2 4q22.1 12q24.13 20q11.2 14q11.2 2q37.3 1p32 15q15.1 14q31 4q31 17q21.31 21q22.11 19p13.3 9q13-q21 15q22.1 1q24-q25

Rho guanine nucleotide exchange factor; interacts with IGF1R Activating transcription factor; regulates glucose homeostasis Monocyte chemotactic protein; role in inflammation Checkpoint kinase-2; blocks cell cycling upon DNA damage Cadherin-interacting protein (possible role in cell differentiation) Mitochondrial heme protein electron carrier; participates in apoptosis Epidermal growth factor Prolyl hydroxylase of hypoxia inducing factor-1 Actin-associated protein; cytoskeleton remodeling Hypoxia-inducible transcription factor Guanine nucleotide exchange factor; cytoskeleton remodeling Feline sarcoma oncogene; tyrosine kinase Inhibitor of GSK3 (likely oncogene) Inducible T-cell co-stimulator ligand Cell-cell adhesion; mediator of angiogenesis Junction Plakoglobin; binds cadherins; complexes to adherens junctions LIM kinase-2; inhibits actin depolymerization; possible cytoskeleton reordering Mitogen-activated protein kinase-10; growth signaling MAPK-activated protein kinase 5; HSP27 kinase (tumor suppressor) Degradation of extracellular matrix; cleavage of cell surface proteins Nuclear factor of activated T cells (T-cell stimulator) Period-2; component of the circadian core oscillator Protein tyrosine phosphatase DNA repair protein; interacts with BRCA1 and BRCA2 Ribosomal Protein S6 Kinase-5; phosphorylates histones; role in inflammation Transducer and transcriptional modulator of multiple signaling pathways Signal Transducer and Activator of Transcription 3; growth factor response Guanine nucleotide exchange factor; modulates Rac1 Toll-like receptor adaptor molecule 1; signal transducer Tight Junction Protein-2; assembler and component of tight junction barrier Tropomyosin 1; involved in the contraction of striated and smooth muscles Vesicle-Associated Membrane Protein-4; role in exocytosis and secretion

Of the 32 genes 25 genes were acceptable according to the qRT-PCR quality control condition (Materials and Methods, Realtime RT-PCR analysis). As noted above, Fig. 3B demonstrated that the resistance of JIMT-1 cells associated with downregulation of HER2 in the presence of trastuzumab, in a time-dependent manner. Thus, we hypothesized that SKBR3 cells, having an intact HER2 signal transduction pathway (thus amenable to trastuzumab blockade), would differentially express the 25 resistance-correlated genes in the presence of trastuzumab. After 24 hr trastuzumab treatment, the 25 genes were downregulated in SKBR3 cells, and upregulated in JIMT-1 cells (Fig. 3C). After 48 hr trastuzumab treatment, however, profound effects on gene expression were demonstrated, with greater downregulation of 24 of the 25

resistance-associated genes in SKBR3 cells, concomitant with upregulation, with significance, ARHGEF12, EGF, ENAH, ICOSLG, TICAM1, and TPM1 in trastuzumab-resistant JIMT-1 cells (Fig. 3D, heat map in Appendix S1: Supplementary Fig. S5). These results suggest that the expression of these genes plays a significant role in trastuzumab resistance, in conjunction with downregulation of the HER2 protein. As shown in Fig. 4, siRNA knock-down of each specific gene in JIMT-1 cells restored trastuzumab sensitivity in 27 out of 32 genes, and loss of HER2 degradation, 48 hr after trastuzumab treatment (data not shown). Finally, we examined expression of the 25 out of 32 genes in 6–8 clinical specimens of HER2-positive tumors. Three were

Table 2 Trastuzumab-sensitive and -resistance tumors and patient clinical data. The first three patients are trastuzumab-responsive and the last five patients are trastuzumab non-responsive. Patient no.

Age at biopsy

Initial diagnosis/ subtype

Initial stage

1

66

8/1996/HER2

III

2 3 4

46 57 27

7/2008/TN 3/2007/LA 8/2012/HER2

III III II or III

5

48

8/2009/HER2

6

67

7 8

47 56

Metastasis in

Organ biopsied

ER

PR

HER2

Ki67

Treatment

Treatment outcome

1/2013

Liver

8

4

3+

36%

Trastuzumab/docetaxel

5/2010 11/2013 2/2014

Liver Liver Liver

8 0 0

8 0 3

3+ 3+ 3+

32% 48% 73%

Trastuzumab/anastrozole Trastuzumab/paclitaxel Trastuzumab adjuvant

III

2/2011

Liver

2

3

3+

50%

7/2000/HER2

III

10/2006

LN

0

0

3+

ND

1/2013/HER2 2/2012/HER2

III III

2/2013 4/2013

Lungs Lungs

0 0

0 0

3+ 3+

72% 39%

Trastuzumab adjuvant and metastasis Trastuzumab/paclitaxel for metastasis Trastuzumab/docetaxel Trastuzumab adjuvant

Improved and maintained on trastuzumab monotherapy Improved Improved Retrying trastuzumab/paclitaxel since metastasis Failed on every chemotherapeutic agents (primary failure) Failed on trastuzumab monotherapy Disease progression after 4th cycles Metastasis during adjuvant treatment

S. Nam et al./Cancer Letters 356 (2015) 880–890

A

885

B

C

Fig. 3. Validation of overexpression of 25 of the 32 genes in trastuzumab-resistant vs. -sensitive cells. SKBR3 and JIMT-1 cells (trastuzumab-sensitive, and -resistant, respectively) were grown to 70% confluency and treated with 10 μg/ml trastuzumab, with a portion of the cells harvested for isolation of total RNA after each day. (A) Cell viability curves for JIMT-1 and SKBR3 cells grown for four days in 10 μg/ml trastuzumab. (B) Western blot of untreated vs. trastuzumab-treated JIMT-1 or SKBR3 cells following each 24 hr period. (C) Expression levels of 25 of the 32 genes after 24 hr culture with or without 10 μg/ml trastuzumab, as determined by qRT-PCR. (D) Expression levels of the same 25 genes following 48 hr culture with or without 10 μg/ml trastuzumab, as determined by qRT-PCR. Control, β-actin, n = 3, *p < 0.05.

886

S. Nam et al./Cancer Letters 356 (2015) 880–890

D

Fig. 3. (continued)

trastuzumab-responsive patients and five unresponsive -patients (see clinical characteristics for each patient, Table 2 and treatment timelines in Appendix S1: Supplementary Fig. S1A and B). We note that these biopsies were taken from metastatic lesions, and HER2-positive metastatic breast cancer almost universally progresses due to Herceptin resistance [3,6,44]. Of these, we found ATF4, CHEK2, ENAH, ICOSLG and RAD51 to each be upregulated in one or more trastuzumab-resistant tumors, with concurrent downregulation in responsive tumors (Fig. 5A–C). We note that while patients 1–3 (Table 2) were classified as trastuzumab-responsive, it is possible these patients may develop resistance in the future (Appendix S1: Supplementary Fig. S1A) [45–49]. Taken together, our rigorous bioinformatics approach reproducibly implicated the expression of 32 genes in breast cancer trastuzumab resistance. Overexpression of 25 of these genes was validated in the trastuzumab-resistant cell line JIMT-1. While JIMT-1 cells are HER2-positive, we note that in SKBR3 (trastuzumab-sensitive) cells, trastuzumab is much more

inhibitory to HER2 heterodimerization with other EGFR family members [50]. Moreover, HER2 was strongly downregulated by the presence of trastuzumab in JIMT-1 cells, in stark contrast to SKBR3 cells. These findings lead us to hypothesize that expression of the resistance-conferring genes may allow autonomous HER2 downstream signaling in the absence of HER2 tyrosine kinase activity. Moreover, it has also been shown that trastuzumab sensitivity positively correlates with levels of HER2 expression [44], and JIMT-1 cells express only ~10% of the HER2 protein level of SKBR3 cells (Fig. 3B).

Discussion In this study, we used a subpathway-based and network permutation method to analyze numerous preclinical and clinical datasets to identify several KEGG pathways associated with HER2

S. Nam et al./Cancer Letters 356 (2015) 880–890

887

Fig. 4. Validation of trastuzumab-resistance reversible through RNA interference of the 32 genes in JIMT-1. SKBR3 and JIMT-1 following 48 hr culture in 10 μg/ml trastuzumab. This demonstrates that in 27 genes, Herceptin sensitivity is restored in JIMT-1 cells after siRNA treatment of each genes.

downstream signaling, including the MAPK, WNT, and TGF-beta cascades (Fig. 1 and Appendix S1: Supplementary Table S1). We previously demonstrated that this method was superior to standard Gene Set Enrichment Analysis (GSEA), based on its specific consideration of cancer-related pathways [16]. In support of its effectiveness, subpathway-based analysis revealed several pathways previously implicated in trastuzumab resistance in breast cancer [3]. Additionally, however, we found several pathways/processes not previously linked to Herceptin resistance, including circadian rhythms, axon guidance, and tight junctions (Appendix S1: Supplementary Tables S1 and S2). Of these, circadian dysregulation has been previously noted as a risk for breast cancer [40], while various axon guidance molecules have been associated with various cancers, perhaps due to their regulation of cytoskeleton structure [51].

Similarly, claudins, critical components of tight junctions, have been proposed as an additional method for stratifying breast cancer subtypes [52]. With regard to specific genes that might represent potential Herceptin response biomarkers, examination of clinical specimens validated overexpression of ATF4, CHEK2, ENAH, ICOSLG, and RAD51 in trastuzumab resistance. We concede that our findings our quite preliminary, with respect to the REMARK criteria for biomarker clinical translation [53]. Nonetheless, there exists some precedence for the respective pathways, composed of these genes, in facilitating possible autonomy from HER2 signaling and evasion of trastuzumab-associated, antibody-dependent cellular cytoxicity [3]. In particular, trastuzumab was found to synergize with cisplatin-induced DNA damage; consequently, one hypothesized

888

S. Nam et al./Cancer Letters 356 (2015) 880–890

A

B

Fig. 5. Clinical validation in eight trastuzumab responsive and non-responsive breast cancer tumors. (A–C) Expression of the 25 selected genes in 6–8 clinical specimens (see clinical information in Table 2). Three trastuzumab-responsive tumors and five trastuzumab-resistant tumors. *p < 0.05 in resistant tumors vs. each trastuzumabresponsive tumor.

antineoplastic mechanism of Herceptin is inhibition of DNA repair, consistent with the activities of RAD51 and CHEK2 in facilitating cell cycle inhibition and repair of DNA double-strand breaks [44]. Moreover, ENAH is a facilitator of the epidermal-tomesenchymal transition (EMT), a phenomenon [54] strongly implicated in Herceptin resistance [3,15]. Similarly, ENAH and FARP2 in particular are involved in cytoskeleton remodeling, an event essential for cell motility and metastasis [45,46], in agreement with our identification of KEGG pathway hsa04810 (regulation of actin cytoskeleton), another hallmark of EMT [55]. Interestingly, members of the insulin-like growth factor family, another contributor to EMT [55], were examined as trastuzumab resistance markers in a phase I clinical trial (http://www.clinicaltrials.gov) [3]. In summary, our now-established subpathway identification approach is highly effective for accurate pathologic biomarker and pathway discovery. Further retrospective and prospective study of these resistance-associated pathways and gene products will provide greater insight into breast cancer signal transduction and the possible prediction (and perhaps, reversal) of resistance to targeted therapies such as trastuzumab.

Acknowledgements The Korea Institute of Science and Technology Information generously provided SN with high performance computing resources. The work was supported by grants from the National Cancer Center, Korea (NCC-1210350-2 to YHK; NCC-1210460 to SN).

Conflict of interest CB is the Chair of Bioscience Advising (Indianapolis, IN, USA). This entity has no financial interest in the results of this study or public deposition of any data. All authors declare no potential competing interests.

Appendix: Supplementary material Supplementary data to this article can be found online at doi:10.1016/j.canlet.2014.10.038.

S. Nam et al./Cancer Letters 356 (2015) 880–890

889

C

Fig. 5. (continued)

References [11] [1] World Health Organization Media Centre, Fact sheet N°297, Cancer, World Heath Oganizaton, (2014) . [2] D. Harari, Y. Yarden, Molecular mechanisms underlying ErbB2/HER2 action in breast cancer, Oncogene 19 (2000) 6102–6114. [3] P.R. Pohlmann, I.A. Mayer, R. Mernaugh, Resistance to trastuzumab in breast cancer, Clin. Cancer Res. 15 (2009) 7479–7491. [4] I. Smith, M. Procter, R.D. Gelber, S. Guillaume, A. Feyereislova, M. Dowsett, et al., H.s. team, 2-year follow-up of trastuzumab after adjuvant chemotherapy in HER2-positive breast cancer: a randomised controlled trial, Lancet 369 (2007) 29–36. [5] A. Goldhirsch, R.D. Gelber, M.J. Piccart-Gebhart, E. de Azambuja, M. Procter, T.M. Suter, et al., Herceptin Adjuvant Trial Study, 2 years versus 1 year of adjuvant trastuzumab for HER2-positive breast cancer (HERA): an open-label, randomised controlled trial, Lancet 382 (2013) 1021–1028. [6] R. Bartsch, C. Wenzel, G.G. Steger, Trastuzumab in the management of early and advanced stage breast cancer, Biologics 1 (2007) 19–31. [7] C.H. Lu, S.L. Wyszomierski, L.M. Tseng, M.H. Sun, K.H. Lan, C.L. Neal, et al., Preclinical testing of clinically applicable strategies for overcoming trastuzumab resistance caused by PTEN deficiency, Clin. Cancer Res. 13 (2007) 5883–5888. [8] C.G. Murphy, P.G. Morris, Recent advances in novel targeted therapies for HER2-positive breast cancer, Anticancer Drugs 23 (2012) 765–776. [9] K. Berns, H.M. Horlings, B.T. Hennessy, M. Madiredjo, E.M. Hijmans, K. Beelen, et al., A functional genetic approach identifies the PI3K pathway as a major determinant of trastuzumab resistance in breast cancer, Cancer Cell 12 (2007) 395–402. [10] Y. Nagata, K.H. Lan, X. Zhou, M. Tan, F.J. Esteva, A.A. Sahin, et al., PTEN activation contributes to tumor inhibition by trastuzumab, and loss of

[12]

[13]

[14]

[15]

[16]

[17]

[18]

[19]

PTEN predicts trastuzumab resistance in patients, Cancer Cell 6 (2004) 117– 127. F. Vegran, R. Boidot, B. Coudert, P. Fumoleau, L. Arnould, J. Garnier, et al., Gene expression profile and response to trastuzumab-docetaxel-based treatment in breast carcinoma, Br. J. Cancer 101 (2009) 1357–1364. A. Belkhiri, A.A. Dar, D.F. Peng, M.H. Razvi, C. Rinehart, C.L. Arteaga, et al., Expression of t-DARPP mediates trastuzumab resistance in breast cancer cells, Clin. Cancer Res. 14 (2008) 4564–4571. M. Jung, J.S. Koo, Y.W. Moon, B.W. Park, S.I. Kim, S. Park, et al., Overexpression of class III beta tubulin and amplified HER2 gene predict good response to paclitaxel and trastuzumab therapy, PLoS ONE 7 (2012) e45127. M. Al-Hajj, M.S. Wicha, A. Benito-Hernandez, S.J. Morrison, M.F. Clarke, Prospective identification of tumorigenic breast cancer cells, Proc. Natl. Acad. Sci. U. S. A. 100 (2003) 3983–3988. C. Oliveras-Ferraros, B. Corominas-Faja, S. Cufi, A. Vazquez-Martin, B. Martin-Castillo, J.M. Iglesias, et al., Epithelial-to-mesenchymal transition (EMT) confers primary resistance to trastuzumab (Herceptin), Cell Cycle 11 (2012) 4020–4032. S. Nam, T. Park, Pathway-based evaluation in early onset colorectal cancer suggests focal adhesion and immunosuppression along with epithelialmesenchymal transition, PLoS ONE 7 (2012) e31685. S. Nam, X. Long, C. Kwon, S. Kim, K.P. Nephew, An integrative analysis of cellular contexts, miRNAs and mRNAs reveals network clusters associated with antiestrogen-resistant breast cancer cells, BMC Genomics 13 (2012) 732. S. Nam, H.R. Chang, K.T. Kim, M.C. Kook, D. Hong, C.H. Kwon, et al., PATHOME: an algorithm for accurately detecting differentially expressed subpathways, Oncogene 33 (2014) 4941–4951. L. Gu, S. Waliany, S.E. Kane, Darpp-32 and its truncated variant t-Darpp have antagonistic effects on breast cancer cell growth and Herceptin resistance, PLoS ONE 4 (2009) e6220.

890

S. Nam et al./Cancer Letters 356 (2015) 880–890

[20] P. Shannon, A. Markiel, O. Ozier, N.S. Baliga, J.T. Wang, D. Ramage, et al., Cytoscape: a software environment for integrated models of biomolecular interaction networks, Genome Res. 13 (2003) 2498–2504. [21] J. Nunez-Iglesias, C.C. Liu, T.E. Morgan, C.E. Finch, X.J. Zhou, Joint genome-wide profiling of miRNA and mRNA expression in Alzheimer’s disease cortex reveals altered miRNA regulation, PLoS ONE 5 (2010) e8898. [22] A.L. Barabasi, Z.N. Oltvai, Network biology: understanding the cell’s functional organization, Nat. Rev. Genet. 5 (2004) 101–113. [23] J. Kao, K. Salari, M. Bocanegra, Y.L. Choi, L. Girard, J. Gandhi, et al., Molecular profiling of breast cancer cell lines defines relevant tumor models and provides a resource for cancer gene discovery, PLoS ONE 4 (2009) e6146. [24] R.M. Neve, K. Chin, J. Fridlyand, J. Yeh, F.L. Baehner, T. Fevr, et al., A collection of breast cancer cell lines for the study of functionally distinct cancer subtypes, Cancer Cell 10 (2006) 515–527. [25] T. Ichikawa, F. Sato, K. Terasawa, S. Tsuchiya, M. Toi, G. Tsujimoto, et al., Trastuzumab produces therapeutic actions by upregulating miR-26a and miR-30b in breast cancer cells, PLoS ONE 7 (2012) e31422. [26] The cancer genome atlas, The Cancer Genome Atlas Networks, Comprehensive molecular portraits of human breast tumours, Nature 490 (2012) 61–70. . [27] E. Cerami, J. Gao, U. Dogrusoz, B.E. Gross, S.O. Sumer, B.A. Aksoy, et al., The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data, Cancer Discov. 2 (2012) 401–404. [28] L.J. van’t Veer, H. Dai, M.J. van de Vijver, Y.D. He, A.A. Hart, M. Mao, et al., Gene expression profiling predicts clinical outcome of breast cancer, Nature 415 (2002) 530–536. [29] M. Goldman, B. Craft, T. Swatloski, K. Ellrott, M. Cline, M. Diekhans, et al., The UCSC Cancer Genomics Browser: update, Nucleic Acids Res. 41 (2013) (2013) D949–D954. [30] T. Miki, T. Matsumoto, Z. Zhao, C.C. Lee, p53 regulates Period2 expression and the circadian clock, Nat. Commun. 4 (2013) 2444. [31] R. Barallon, S.R. Bauer, J. Butler, A. Capes-Davis, W.G. Dirks, E. Elmore, et al., Recommendation of short tandem repeat profiling for authenticating human cell lines, stem cells, and tissues, In Vitro Cell. Dev. Biol. Anim. 46 (2010) 727–732. [32] G.D.I. Cara, G. Marengo, N.N. Albanese, M.R. Marabeti, R. Musso, P. Cancemi, et al., Proteomic profiling of trastuzumab (Herceptin®)-sensitive and -resistant SKBR-3 breast cancer cells, Anticancer Res. 33 (2013) 489–503. [33] Y.H. Kim, A. Coon, A.F. Baker, G. Powis, Antitumor agent PX-12 inhibits HIF1alpha protein levels through an Nrf2/PMF-1-mediated increase in spermidine/ spermine acetyl transferase, Cancer Chemother. Pharmacol. 68 (2011) 405–413. [34] T.D. Schmittgen, K.J. Livak, Analyzing real-time PCR data by the comparative C(T) method, Nat. Protoc. 3 (2008) 1101–1108. [35] D. Hanahan, R.A. Weinberg, Hallmarks of cancer: the next generation, Cell 144 (2011) 646–674. [36] B.J. Feldman, D. Feldman, The development of androgen-independent prostate cancer, Nat. Rev. Cancer 1 (2001) 34–45. [37] C.K. Osborne, R. Schiff, Mechanisms of endocrine resistance in breast cancer, Annu. Rev. Med. 62 (2011) 233–247. [38] H.H. Jung, Y.H. Park, H.J. Jun, J. Kong, J.H. Kim, J.A. Kim, et al., Matrix metalloproteinase-1 expression can be upregulated through mitogen-activated protein kinase pathway under the influence of human epidermal growth factor

[39]

[40] [41]

[42]

[43]

[44]

[45]

[46]

[47] [48] [49]

[50]

[51]

[52]

[53]

[54]

[55]

receptor 2 synergized with estrogen receptor, Mol. Cancer Res. 8 (2010) 1037–1047. T.A. Bailey, H. Luan, R.J. Clubb, M. Naramura, V. Band, S.M. Raja, et al., Mechanisms of trastuzumab resistance in ErbB2-driven breast cancer and newer opportunities to overcome therapy resistance, J. Carcinog. 10 (2011) 28. S. Sahar, P. Sassone-Corsi, Circadian clock and breast cancer: a molecular link, Cell Cycle 6 (2007) 1329–1331. J.E. Ruhe, S. Streit, S. Hart, A. Ullrich, EGFR signaling leads to downregulation of PTP-LAR via TACE-mediated proteolytic processing, Cell. Signal. 18 (2006) 1515–1527. E. Elinav, R. Nowarski, C.A. Thaiss, B. Hu, C. Jin, R.A. Flavell, Inflammation-induced cancer: crosstalk between tumours, immune cells and microorganisms, Nat. Rev. Cancer 13 (2013) 759–771. E.A. Mittendorf, C.E. Storrer, C.D. Shriver, S. Ponniah, G.E. Peoples, Investigating the combination of trastuzumab and HER2/neu peptide vaccines for the treatment of breast cancer, Ann. Surg. Oncol. 13 (2006) 1085–1098. R. Nahta, D. Yu, M.C. Hung, G.N. Hortobagyi, F.J. Esteva, Mechanisms of disease: understanding resistance to HER2-targeted therapy in human breast cancer, Nat. Clin. Pract. Oncol. 3 (2006) 269–280. F. Di Modugno, L. DeMonte, M. Balsamo, G. Bronzi, M.R. Nicotra, M. Alessio, et al., Molecular cloning of hMena (ENAH) and its splice variant hMena+11a: epidermal growth factor increases their expression and stimulates hMena+11a phosphorylation in breast cancer cell lines, Cancer Res. 67 (2007) 2657– 2665. N. Takegahara, S. Kang, S. Nojima, H. Takamatsu, T. Okuno, H. Kikutani, et al., Integral roles of a guanine nucleotide exchange factor, FARP2, in osteoclast podosome rearrangements, FASEB J. 24 (2010) 4782–4792. D.C. Singleton, A.L. Harris, Targeting the ATF4 pathway in cancer therapy, Expert Opin. Ther. Targets 16 (2012) 1189–1202. K.K. Khanna, S.P. Jackson, DNA double-strand breaks: signaling, repair and the cancer connection, Nat. Genet. 27 (2001) 247–254. J. Zhang, H. Willers, Z. Feng, J.C. Ghosh, S. Kim, D.T. Weaver, et al., Chk2 phosphorylation of BRCA1 regulates DNA double-strand break repair, Mol. Cell. Biol. 24 (2004) 708–718. M. Tanner, A.I. Kapanen, T. Junttila, O. Raheem, S. Grenman, J. Elo, et al., Characterization of a novel cell line established from a patient with Herceptinresistant breast cancer, Mol. Cancer Ther. 3 (2004) 1585–1592. A. Chedotal, G. Kerjan, C. Moreau-Fauvarque, The brain within the tumor: new roles for axon guidance molecules in cancers, Cell Death Differ. 12 (2005) 1044–1056. S. Ricardo, R. Gerhard, J.F. Cameselle-Teijeiro, F. Schmitt, J. Paredes, Claudin expression in breast cancer: high or low, what to expect?, Histol. Histopathol. 27 (2012) 1283–1295. L.M. McShane, D.G. Altman, W. Sauerbrei, S.E. Taube, M. Gion, G.M. Clark, et al., Reporting recommendations for tumor marker prognostic studies (REMARK), J. Natl. Cancer Inst. 97 (2005) 1180–1184. K. Polyak, R.A. Weinberg, Transitions between epithelial and mesenchymal states: acquisition of malignant and stem cell traits, Nat. Rev. Cancer 9 (2009) 265–273. S. Lamouille, J. Xu, R. Derynck, Molecular mechanisms of epithelialmesenchymal transition, Nat. Rev. Mol. Cell Biol. 15 (2014) 178–196.