Transcriptomic Profiling of Sequential Tumors from Breast Cancer Patients Provides a Global View of Metastatic Expression Changes Following Endocrine Therapy

Published OnlineFirst August 3, 2015; DOI: 10.1158/1078-0432.CCR-14-2155

Biology of Human Tumors Clinical Cancer Research Transcriptomic Proﬁling of Sequential Tumors from Breast Cancer Patients Provides a Global View of Metastatic Expression Changes Following Endocrine Therapy Jean McBryan1, Ailís Fagan1, Damian McCartan1, Fiona T. Bane1, Damir Vareslija1, Sinead Cocchiglia1, Christopher Byrne1, Jarlath Bolger1, Marie McIlroy1, Lance Hudson1, Paul Tibbitts1, Peadar OGaora 2, Arnold D. Hill1, and Leonie S. Young1

Abstract

Purpose: Disease recurrence is a common problem in breast between sequential tumors were common to all patients. Larger cancer and yet the mechanisms enabling tumor cells to evade divergence was observed between primary and liver tumors than therapy and colonize distant organs remain unclear. We sought to between primary and nodal tumors, reflecting both the latency characterize global expression changes occurring with metastatic to disease progression and the genetic impact of intervening disease progression in the endocrine-resistant setting. therapy. Furthermore, an endocrine-resistant in vivo mouse model Experimental Design: Here, for the first time, RNAsequencing demonstrated that tamoxifen treatment has the potential to has been performed on matched primary, nodal, and liver met- drive disease progression and establish distant metastatic disease. astatic tumors from tamoxifen-treated patients following disease Common functional pathways altered during metastatic, endo- progression. Expression of genes commonly elevated in the crine-resistant progression included extracellular matrix receptor metastases of sequenced patients was subsequently examined in interactions and focal adhesions. an extended matched patient cohort with metastatic disease from Conclusions: This novel global analysis highlights the influence multiple sites. The impact of tamoxifen treatment on endocrine- of primary tumor biology in determining the transcriptomic pro- resistant tumors in vivo was investigated in a xenograft model. file of metastatic tumors, as well as the need for adaptations in cell– Results: The extent of patient heterogeneity at the gene level cell communications to facilitate successful tumor cell colonization was striking. Less than 3% of the genes differentially expressed of distant host organs. Clin Cancer Res; 1–9. 2015 AACR.

Introduction organ disease. The dominant site of metastatic involvement can have a significant bearing on clinical outcome. Metastatic disease The vast majority of breast cancers are steroid receptor positive of the liver has been reported to be a predictor of poor outcome, and depend on estrogen for growth. The use of estrogen receptor with a median survival of 18 months (1), whereas metastatic modulators, such as tamoxifen and estrogen depletion strategies, burden confined to the bone is thought to be more indolent (2). including aromatase inhibitors, has been widely successful. Disease Colonization of the host organ is the ultimate step of a recurrence, however, is common, with between 10% and 25% of progressive disease path. Most breast cancer cells that enter the patients developing extensive local or distant metastases. Most circulation and infiltrate distant organs die due to the hostile patients with metastases will succumb to their disease, as targeted nature of the host microenvironment (3). For metastatic growth therapies that provide significant clinical benefit remain elusive. to be established, disseminated breast cancer cells need to survive Metastatic breast cancer secondary to endocrine treatment is a period of latency, prolonged exposure to therapy, and subse- heterogeneous, ranging from discreet lesions to diffuse multi- quently reinitiate growth when appropriate. Evidence of the impact of steroid treatment on tumor progression is emerging. Steroid receptor switching and in particular loss of the progester- 1Endocrine Oncology Research Group, Department of Surgery, Royal one receptor (PGR), an estrogen receptor (ER) target, is observed 2 College of Surgeons in Ireland, Dublin, Ireland. UCD School of in approximately 30% of patients (4, 5). Furthermore, gain-of- Biomolecular and Biomedical Science, Conway Institute, University ESR1 fi College Dublin, Dublin, Ireland. function mutations in the gene have been identi ed in metastatic lesions from endocrine treated patients (6, 7). Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/). The mechanisms required to enable cancer cells to overcome hostile forces and reinitiate growth at distant organs in endocrine- J. McBryan and A. Fagan contributed equally to this article. treated patients have not been fully elucidated. Lessons from Corresponding Author: Leonie S. Young, Endocrine Oncology Research Group, in vitro studies and murine models have provided important Department of Surgery, Royal College of Surgeons in Ireland, Dublin 2, Ireland. information about discrete tools used by breast cancer cells to Phone: 353-402-8576; Fax: 353-402-8551; E-mail: [email protected] prosper at distant metastatic sites. For example, the CXCR4/SDF1 doi: 10.1158/1078-0432.CCR-14-2155 pairing has been identified as a mediator of metastatic cell survival 2015 American Association for Cancer Research. in the bone (8, 9); transcriptional inhibitors of differentiation, Id1

www.aacrjournals.org OF1

McBryan et al.

cDNA was synthesized as previously described (13). Gene Translational Relevance expression was assessed using TaqMan technology (Applied To date, next-generation sequencing has been completed for Biosystems) on the ABI PRISM 7500 platform with the follow- a substantial number of primary breast tumors and provides ing probes: EPPK1 Hs01104050_s1; COL4A2 Hs01098873_ an in-depth understanding of the biology and heterogeneity of m1; COL4A1 Hs00266237_m1; SPDEF Hs01026050_m1; this disease. Here, primary tumors together with matched KRT19 Hs00761767_s1; AHDC1 Hs00210424_m1; NCOR2 nodal and liver metastatic patient tumors have been Hs00196955_m1; AKAP8L Hs00205106_m1. EIF2B Hs00426752_ fi sequenced, offering the rst in-depth view of the transcrip- ml was used as an internal control. The comparative CT (DDCT) tomic changes occurring during disease progression. Patient method was applied to analyze relative mRNA expression levels. heterogeneity was strongly evident in both primary and met- – astatic tumors. In addition, common adaptations in cell cell IHC communication were detected as metastatic endocrine-resis- Of note, 5-mm tumor sections were deparaffinized and rehy- fi tant disease took hold. These ndings emphasize the need for drated as previously described (4). Antigen retrieval was ongoing personalized medicine throughout disease manage- performed with sodium citrate at pH6 or EDTA at pH9 (Ki- ment. In particular, it is important to reevaluate the biology of 67). Dako Envisionþ HRP kit was used as per manufacturer's metastatic disease, following treatment intervention, in order instructions. Primary antibodies incubated for 1 hour at room to monitor adaptations and identify the optimal treatment temperature were: Ki-67 (MKI67, 1:200, MIB-1, M7240, Dako); strategy. KRT19 (1:200, HPA002465, Sigma); LMNB2 (1:75, MAB3536, Merck Millipore); KIF12 (1:100, NBPI-86029, Novus); EVPL (1:30, Ab49397, Abcam); MCM4 (1:50, HPA004873, Sigma); p-mTOR (1:100, Ser2448, #2976, Cell Signaling Technology); and Id3, have been associated with lung metastasis (10, 11); and and active b-catenin (1:150, 05-665, Merck Millipore). The fol- serpins, as metastatic functionaries in the brain (12). However, to lowing primary antibodies were incubated overnight at 4 C: p- date, no global overview of gene expression alterations that are p70S6K (1:100, Thr389, #9205, Cell Signaling Technology); p- essential for successful colonization of cells at metastatic sites in HER2 (1:400, Tyr1248, 06-229, Merck Millipore); p-STAT3 breast cancer patients has been undertaken. (1:100, Tyr705, #9131, Cell Signaling Technology); and p-ERK To clearly define mediators of colonization in endocrine-trea- 1/2 (1:400, Thr202/Tyr204, #4370, Cell Signaling Technology). ted patients, we undertook gene expression analysis from sequen- ER (1:60, RM9101S, Thermo Fisher Scientific), PGR (1:60, tial tumor samples. RNAseq analysis was performed on matched H3569, DAKO), and HER2 (neat, 4B5, 790-2991, Ventana) were primary, nodal, and distant metastatic tumors from breast cancer stained on an automated program. H-score was assigned by two patients, all of whom had developed a liver metastasis following independent assessors on a scale of 0–300 with 300 representing tamoxifen treatment. At a functional level, studies in xenograft the highest intensity staining in 100% of cells. models of endocrine-resistant tumors were undertaken to examine the role of tamoxifen in the development of metastatic disease Xenograft studies secondary to therapy resistance. Over-represented pathways All mouse experiments were performed in accordance with identified by the RNAseq analysis were further studied at the the European Communities Council Directive (86/609/EEC) transcript and protein level in an expanded cohort of patients with and were reviewed and approved by Research Ethics Commit- metastases of multiple distant organs. tee under license from the Department of Health. Ten 6-week- oldfemaleBALB/cSCIDmice(Harlan)wereimplantedwith Materials and Methods 17-b-estradiol pellets (0.36 mg/pellet, 60-day release) alone or Gene expression and bioinformatic analysis in combination with 4-hydroxytamoxifen pellets (5 mg/pellet, Following ethical approval, formalin-fixed paraffin-embed- 60-day release). All pellets were obtained from Innovative ded (FFPE) tumor samples were macrodissected to select Research of America and were readministered at week 8. 6 regions with >70% tumor cells. RNA isolation was performed Luciferase-tagged LY2 cells (1 10 cells) were mixed with with High Pure FFPE RNA Micro Kit (Roche) and subjected to 50% Matrigel (BD biosciences) and implanted into the fourth duplex specific thermostable nuclease (DSN) treatment. RNA inguinal mammary gland by injection. Tumors were imaged SequencingwasperformedbyBGI(HongKong)usingthe using an IVIS whole body imaging system (Xenogen Corp) to Illumina HiSeq 2000 with >40 million reads. Insufficient RNA detect luciferase activity. Five weeks after cell implantation, fi was obtained from the nodal sample of patient 1 for sequenc- primary tumors were surgically removed, formalin xed, and fi ing. Reads were aligned to the hg19 genome using two different paraf n embedded. The mice were monitored by IVIS once per types of analysis: one for the differential expression and the week for the next 10 weeks to detect metastasis of the labeled fl t other for use with The Cancer Genome Atlas (TCGA) data. For cells. The fold increase in metastatic ux was compared by the the differential expression analysis, the reads were aligned test between each group of animals. using TopHat and genes were quantified using HTseq. For usage with TCGA data, the data were aligned using MapSplice Results and the genes were quantified using RSEM. Full details of Gene expression profiles of breast cancer patients reflect the bioinformatics analysis are provided in Supplementary Meth- histopathology of the original primary tumor ods. Data from this study have been deposited in the NCBI RNA was extracted from matched breast cancer primary, node, Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih. and liver metastatic tumor tissue from three tamoxifen-treated gov/geo/) under accession number GSE58708. luminal breast cancer patients. The patients represented the

OF2 Clin Cancer Res; 2015 Clinical Cancer Research

Transcriptomic Proﬁles of Metastatic Breast Cancer

Table 1. Tumor and disease characteristics for sequenced patient samples Patient 1 Patient 2 Patient 3 Age at diagnosis 34 y 61 y 37 y Time to distant metastasis 11 y 5 y 6 y Time to death 12 y 5 y 9 y Primary breast pathology ER þþþ PGR þþ HER2 þ TP53 Ki-67 10% 16% 38% Tumor size 50 mm 26 mm NA Tumor grade 2 2 3 Lymph node status þþþ Lobular/ductal Lobular Ductal Ductal OncotypeDX risk Low High High Luminal classification A B1 B2 Surgery, adjuvant treatment Surgery WLE þ ANC mastectomya WLE þ ANC Mastectomy þ ANC Chemotherapy CMF 6, AC 4a, taxane 6a AC 4, taxane 4 CMF 6 Radiotherapy Yes Yes Yes Endocrine therapy 5 y tamoxifen, 5 y arimidexa, 4 y zoladexa 5 y tamoxifen 5 y tamoxifen Herceptin No No No Metastatic tumor pathology Sites of metastasis Liver, bone Liver, lung Liver, lung, brainb Biopsy Liver Liver Liver ER (liver biopsy) þ NA þ PGR (liver biopsy) þ NA þ HER2 (liver biopsy) NA þ TP53 (liver biopsy) NA NA Post-metastatic therapy Fulvestrant Arimidex Herceptin, Letrozole, Fulvestrant Abbreviations: y, years; WLE, wide local excision; ANC, axillary node clearance; CMF, cyclophosphamide methotrexate fluorouracil; AC, adriamycin cyclophosphamide; NA, not available. aTreatment given 5 years following initial diagnosis. bBrain metastasis diagnosed 8 years following initial diagnosis. spectrum of the luminal classification, from luminal A (patient 1) Metastatic gene expression alterations are heterogeneous in to luminal B2 (patient 3; Table 1). Despite the diverse pathology endocrine-treated breast cancer patients of each of the primary breast tumors (Table 1; Fig. 1A; Supple- To define the changes involved in disease progression, differ- mentary Fig. S1), positive steroid receptor status qualified each of entially expressed genes (DEG) were identified between matched the patients for adjuvant endocrine treatment. primary and liver metastatic tumors from each patient. This Correspondence analysis of the sequenced RNA from the analysis identified 1,268 upregulated and 2,606 downregulated individual tissue samples revealed that each of the sequential genes between matched primary and metastatic tumors. Remark- patient samples clustered together, rather than with the met- ably, less than 3% of DEGs were common to all three patients astatic site (Fig 1B). The relative divergence in gene expression highlighting the extent of patient heterogeneity (Fig. 2A). Only 31 from the primary to the metastatic tumors was reflective of upregulated and 69 downregulated genes were shared by the 3 the disease latency period. Patient 1 with a recorded metas- patients (Supplementary Tables S1 and S2). The individuality of tasis-free survival period of 11 years had the greatest difference each patient is evident in the heatmaps of DEGs where each in gene expression profile between the primary and metastatic patient elevated their own unique set of genes in the liver met- tumors (Fig. 1B). Patient 2 and patient 3 had shorter disease- astatic sample (Fig. 2B). Of note, little difference in gene expres- free periods and correspondingly substantially less alterations sion was detected between the primary tumor and the nodal in gene expression, suggesting that some of the genes metastases, both of which were treatment na€ve and were resected necessary for successful progression are present in the primary at the time of initial surgery (Fig. 2B). tumor. Analysis of enriched KEGG pathway terms was performed for RNAseq data from the breast cancer sequential samples were the DEGs of each patient and highlighted a number of important analyzed with breast cancer samples from the TCGA by corre- pathways that were unique to each patient (Supplementary Fig. S2 spondence analysis of almost 16,000 genes (Fig. 1C). Nodal and and S3). The luminal A tumor from patient 1 was highly endo- metastatic samples clustered in amongst the primary tumors. crine dependent with strong expression of both ER and PGR. The PAM50 genes were used to classify the breast cancer sequential large divergence from primary to metastasis on correspondence samples and the TCGA samples into the major breast cancer analysis as well as the large number of differentially expressed subtypes. Tumors from patients 1 and 2 clustered on the luminal genes is indicative of the multiple adaptations that were required end of the horizontal axis. The tumors from patient 3 straddled the for successful tumor progression and colonization. Following luminal/HER2 border. Interestingly, while the primary tumor endocrine treatment, steroid receptor status remained strong; from patient 1 was luminal A, the metastatic tumor from this however, alterations in growth factor signaling networks, includ- patient was classified as luminal B ing MAPkinase, Jak-STAT, and ErbB signaling pathways were

www.aacrjournals.org Clin Cancer Res; 2015 OF3

McBryan et al.

A Paent 1 Paent 2 Paent 3 B

ER Paent 1 M

M PGR P N Paent 2 M P Figure 1. Primary tumor HER2 P N Molecular proﬁling of matched tumor samples from 3 heterogeneous breast Paent 3 cancer patients. A, IHC staining of classic biomarkers in primary breast tumor tissue (top) and H&E staining of Ki67 matched liver metastatic tumor tissue (bottom) showing strong heterogeneity Correspondence analysis of matched tumors between the 3 patients. ER, estrogen receptor; PGR, progesterone receptor; from 3 paents H&E, hematoxylin and eosin staining. H&E B, correspondence analysis of RNAseq

Liver mets 50 μm data from primary (P), nodal (N), and metastatic liver (M) tumors of the 3 patients. Patient heterogeneity is evident and tumors do not cluster based C d = 0.05 Luminal B on site of origin. Patient 1 shows the largest divergence between primary and metastatic tumors. C, correspondence HER2-enriched analysis of RNAseq data from the primary, nodal, and metastatic samples from the 3 patients (n ¼ 8) and breast cancer patient samples from the TCGA (n ¼ 840). Each dot or shape represents 2n an individual tumor, colored based on PAM50 proﬁling. 2m 1m 3p 3m Basal 2p 1p

Luminal A 3n

Normal-like

Correspondence analysis of tumors from 848 paent samples

observed (Supplementary Fig. S2). As protein function cannot analysis, genes from the primary, node and liver metastasis always be directly inferred from gene expression analysis, expres- clustered with the HER2 subtype using the PAM50 classification. sion of phosphoproteins from these pathways was examined and Alterations in the WNT signaling pathway observed in the metas- confirmed in matched primary and metastatic tumors from a tasis, at both the gene and phosphoprotein level (Supplementary luminal A patient (Supplementary Fig. S4). In contrast with Fig. S2 and S4), signify tumor adaptation to endocrine therapy patient 1, the luminal B1 primary tumor from patient 2 displayed and activation of the epithelial to mesenchymal transition limited endocrine dependence, with low steroid receptor expres- network. sion and no HER2. Limited divergence on correspondence anal- Of note, the mTOR pathway, which is a known predictor of ysis and relatively low number of gene alterations suggest that the endocrine resistance (14), was found to be elevated in the liver genes required for metastatic progression may have been present metastases of patients 1 and 3 at the transcript level in comparison in the primary tumor. Pathways that were altered in the liver with the primary tumor (Supplementary Fig. S2). Expression metastatic tumor include DNA replication and cell cycle (Sup- levels of phospho-mTOR and phospho-p70S6 kinase in matched plementary Fig. S2), also confirmed at the protein level (MCM4; primary and metastatic tumors from endocrine-treated patients Supplementary Fig. S4). The luminal B2 tumor from patient 3 was confirmed activation of the mTOR pathway (Supplementary strongly positive for ER, PGR, and HER2. From correspondence Fig. S5).

OF4 Clin Cancer Res; 2015 Clinical Cancer Research

Downloaded from clincancerres.aacrjournals.org on September 23, 2021. © 2015 American Association for Cancer Research. www.aacrjournals.org nihdKG ahascmo oal3ptet nld ubro acrptwy n locl neato ahassc sfclahsosadECM and adhesions focal as such pathways S3. interaction and cell S2 also Figs. and Supplementary pathways pathways in KEGG cancer provided enriched of are showing number patient maps a each network include to C, to patients unique highlighted. relative 3 pathways as interactions. tumor all of expression receptor metastatic to Details gene in common DEGs. of downregulated pathways (right) pattern genes KEGG downregulated distinct Right, Enriched and a tumor. tumor. (left) displayed primary primary upregulated patient matched the the Each to than from A. relative less part tumor was from metastatic sample DEGs in liver change (right) upregulated normal fold a genes a exhibited Left, DEGs in tumor. tumors. expression primary metastatic matched genes, cell and these in primary of involved their all genes between For of patient individual expression each altered for display DEGs tumors metastatic and primary Matched 2. Figure A C B Downloaded from 2012 –2 Row Z-score Row Color key Patient 1 Patient Patient 3 Patient interaction receptor Published OnlineFirstAugust3,2015;DOI:10.1158/1078-0432.CCR-14-2155 ECM Genes upregulatedinmetastasis Amoebiasis clincancerres.aacrjournals.org Normal liver 3 Paent Paent 1 Paent adhesion cell lung Paent 1 Paent Focal cancer Small

Primary 387 Pathways in carcinoma Renal cell Mets cancer 45

Paent 2 Paent Primary 509 67 31 > . and 1.5 Node 18 Patient 2 Patient

Mets 211 > Paent 2 Paent 0cut e ilo.B etasdslyn h 28urgltd(et n ,0 downregulated 2,606 and (left) upregulated 1268 the displaying heatmaps B, million. per counts 50

Paent 3 Paent Primary Cancer Research. 21,958 Node on September 23, 2021. © 2015American Association for Mets – elitrcin n acrptwy.A endarm showing diagrams Venn A, pathways. cancer and interactions cell 2– 2 1 0 –1 –2 Row Z-score Row Color key Patient 3 rncitmcPro Transcriptomic Patient 1 interaction Endocytosis pathway Genes downregulatedinmetastasis ECM R MAPK Paent 1 Paent Paent 3 Paent signaling T-cell R

Primary 1 Paent adhesion Pathways in Colorectal Focal cancer cancer Mets Pancreatic 856 cancer plasmosis e fMtsai ratCancer Breast Metastatic of ﬁles Toxo- Primary Paent 2 Paent cell lung cell 89 cancer CML Small 924 203 Node 69

Mets 109 lnCne e;2015 Res; Cancer Clin 356 Paent 2 Paent Paent 3 Paent

Patient 2 Primary 20,620 Node Mets OF5 Published OnlineFirst August 3, 2015; DOI: 10.1158/1078-0432.CCR-14-2155

McBryan et al.

A Tumor removed B Week 4 Week 5 Week 9 Week 12 Week 15 Bone Lung Liver

E2 only E2 only E2 + 4-OHT

1,000 200 800 p/s) E2 + 3 500 100 400 Total ﬂux 4-OHT (×10 0 0 0 E2 E2 + E2 E2 + E2 E2 + 4-OHT 4-OHT 4-OHT

15 0 Ki67 Keran 19 E2 + 4-OHT (n = 5) 12 0 E2 only (n = 5) Primary breast

90 tumor

60 * metastac ﬂux Fold increase in 30 Metastac liver 0 6 7 8 9 10 11 12 13 14 15 Weeks following cell implantaon

C D RNA expression fold changes:Mets/Primary

RNA-seq TaqMan®

EPPK1 COL4A2 COL4A1 SPDEF e KRT19 AHDC1 NCOR2 AKAP8L Paent: 1 2 3 A BCDEF G* H* Mets site: reviL Perit.contr. enoB Ax.

Increased in metastasis Color key Unchanging Decreased in metastasis Not available Mets Mets Mets Node Node e Conservave esmate based on Primary Primary Primary undetectable expression in primary –2 0 1 2 Paent 1 Paent 2 Paent 3 Untreated metastasis Row Z-score * Normal liver

E KRT19 LMNB2 KIF12 EVPL PT MRT MNT PT MRT MNT PT MRT MNT PT MRT MNT PT MRT MNT PT MRT MNT PT MRT MNT

Hscore Paent A Paent B Paent C Paent D Paent I Paent J Paent K 0 liver liver peritoneum contralateral contralateral lung presacral 1–49 50–149 PT: Primary tumor ssue >150 MRT: Matched recurrent tumor ssue Not available MNT: Matched normal ssue at recurrence site

Figure 3. Tamoxifen contributes to the development of metastasis secondary to endocrine resistance. A, representative in vivo bioluminescence images of mice following orthotopic injection of luciferase expressing endocrine resistant breast cancer cells. (Continued on the following page.)

OF6 Clin Cancer Res; 2015 Clinical Cancer Research

Transcriptomic Proﬁles of Metastatic Breast Cancer

Cell-to-cell communication is a common pathway in liver Cellular communication and cell structure proteins are evident metastasis at other metastatic sites Although at the gene level, patient heterogeneity was very An extended cohort of matched primary and metastatic patient apparent in the endocrine-resistant metastatic tumors, a number samples was used to examine the RNA and protein expression of of functional pathways were common to all three patients (Fig. shared genes from the RNAseq data. The patient cohort included a 2C). Alterations in several cancer pathways, including small cell range of metastatic sites, including the liver, lung, bone, and lung carcinoma, renal cell carcinoma, and pathways in cancer peritoneum of endocrine-treated patients (Supplementary Table were observed. Of interest, pathways important in cell to cell S4). Real-time PCR confirmed the elevated expression of genes in interaction were represented in gene sets that were both enhanced liver metastases relative to matched primary tumors (Fig. 3D). and suppressed in the metastatic liver. Alterations in genes impor- Increases in gene expression were also detected at other metastatic tant in ECM receptor interactions and focal adhesion indicate the sites (Fig. 3D). importance of communication between the tumor and the host At the protein level, IHC analysis of matched primary and liver for successful colonization. metastatic tumors revealed strong keratin 19 protein expression, consistent with that observed in the metastatic tumors from the Endocrine treatment contributes to the development of tamoxifen-treated xenografts (Fig. 3E). Protein expression of the metastasis secondary to endocrine resistance cellular communication proteins, lamin B2 (LMNB2), KIF12, and Of the sequential tumor samples sequenced here, primary and envoplakin (EVPL) were confirmed in the metastatic tumors nodal tumors were biopsied at the same time. Liver tumors were of the endocrine-treated patients. Greater expression of these biopsied following a latency period and also following treatment proteins was found in the metastatic tumor cells in comparison intervention. To determine the impact of endocrine treatment on with the surrounding normal host tissue (Fig. 3E). Furthermore, the development of distant metastasis in endocrine-resistant analysis of unmatched metastatic brain datasets revealed tumors, we used a xenograft model. Endocrine-resistant cells were enhanced expression of our defined 31 gene set in the metastatic implanted into the mammary fat pad of immune-compromised tumors compared with normal brain tissue (Supplementary Fig. mice in the presence of estrogen (Fig. 3A). Once established, the S7). Taken together, these data support a role for cellular com- primary tumors were resected. Local disease recurrence occurred munication in breast cancer colonization of distant organs fol- in animals treated and untreated with the SERM tamoxifen. lowing treatment resistance. Distant metastasis, however, was only established in the tamoxifen-treated group (Fig. 3A). Metastatic disease was observed in the bone, lung, and liver (Fig. 3B). Expression of the proliferation Discussion marker, Ki-67 and keratin 19 (KRT19), a marker of disseminated The ability of selected tumors to adapt and evade endocrine cancer cells, which was elevated in all of the patient liver metas- therapy is well established. Enhanced expression of plasticity tases, was found to be enhanced in the xenograft liver metastatic networks (4, 15, 16), as well as mediators of EMT (17, 18), in tissue compared with the primary (Fig. 3B). These data demon- the primary tumor has been associated with treatment failure. strate that tamoxifen has the potential to contribute to the Information about changes that occur with tumor progression on development of metastatic disease progression in endocrine- treatment, however, is not readily available. Here, we describe the resistant tumors. Analysis of the patient RNAseq data revealed first global transcriptomic analysis of sequential primary and that the majority of the identified upregulated DEGs were in fact metastatic tumor samples from endocrine-resistant breast cancer elevated in metastatic tumors relative to both matched primary patients. and nodal tumors (Fig. 3C). These data suggest that for the Several initiating events of tumor adaptability have been pro- majority of these genes, expression did not increase when the posed, including the degree of intra-tumor heterogeneity (19, 20) cells first moved away from the primary site to the node but did as well as the presence of stem cell populations (21). Though the increase following treatment with tamoxifen, the latency period, source of the treatment adaptability is not addressed in this and intravasation of the distant organ. Remarkably, of the com- study, the resultant mediators that enable the successful coloni- mon DEGs displaying this pattern of expression (Supplementary zation of the tumor at a distant site are elucidated. Table S3), almost half have defined roles in cell-to-cell commu- Here, global mapping of gene expression alterations that occur nication and cell structure (Supplementary Fig. S6). in metastatic breast cancers revealed a significant degree of patient

(Continued.) Five mice were treated with estrogen pellets (E2 only) and 5 mice had estrogen and 4-hydroxytamoxifen pellets (E2þ4-OHT). Primary tumors were resected at week 5. Local recurrences developed in both groups but distant metastases were detectable only in tamoxifen-treated mice. Graph displays the fold increase in metastatic ﬂux with each line representing an individual mouse. , P < 0.05 by an unpaired t test at week 15. B, ex vivo bioluminescence imaging of organs from the mouse xenograft model. All estrogen plus tamoxifen-treated mice displayed bioluminescence in at least one organ of their bone, lung, and livers (n ¼ 5). No bioluminescence was detected in the same organs from estrogen-only treated animals (n ¼ 5). Scatter plots show total ﬂux from each ex vivo organ. Representative images of IHC staining of matched primary and metastatic tumors from estrogen plus tamoxifen treated mouse xenografts are shown. The metastatic marker Keratin 19 was more strongly expressed in metastatic tumors than matched primary tumors (n ¼ 3). Ki-67 staining is also shown to help identify tumor cells. C, heatmap listing the common 31 DEGs that are upregulated in liver metastatic tumors relative to matched primary tumors for all 3 patients. The majority of these genes are more strongly expressed in the liver metastasis, sampled after tamoxifen treatment, than in the nodal tumor, sampled before tamoxifen. Those genes with a fold change >1.5-fold between liver and node are highlighted in bold (n ¼ 21 genes). D, heatmap of TaqMan gene expression fold changes of genes upregulated in metastasis from the RNA-seq data analysis in an extended cohort of matched primary breast cancer and metastases samples. To the left of the TaqMan expression is the original gene expression fold changes for the RNA-seq data. Metastatic sites of the extended cohort include liver, peritoneum (Perit.), contralateral breast (Contr.), bone and axilla (Ax.). The liver metastases patients in the extended cohort exhibited a similar gene expression pattern to the RNA-seq data. E, IHC staining of four DEGs in matched tumors of 7 independent endocrine-treated patients. H-scores for each sample are reported as a heatmap. Details of patient samples and treatments are provided in Supplementary Table S4.

www.aacrjournals.org Clin Cancer Res; 2015 OF7

McBryan et al.

heterogeneity. The relative divergence in gene expression reflected Though several elegant studies have used in vitro and in vivo the histology of the original tumor and the disease latency period. models to determine key players in breast cancer colonization of Although disease progression in some cancers is rapid (22, 23), in distant organs, including the bone, lung, and brain (8–12), there ER-positive breast cancer, metastasis may occur decades after is little information about signaling networks important in met- resection of the primary tumor (24). The greatest degree of astatic liver disease. In this study, common pathways elevated in divergence in gene expression was observed in the ER-positive liver metastatic tumors included pathways in cancer, ECM recep- PGR-positive luminal A patient (patient 1) with a metastasis-free tor interactions, and focal adhesions. Extracellular matrix proteins survival period of >10 years. Differences in gene expression have previously been shown to be important in colon cancer liver suggest that the primary tumor underwent significant alterations metastasis (32), moreover enhanced Claudin 2-extracellular over an extended period of time to successfully develop liver matrix interactions have been demonstrated in liver-aggressive metastasis. In contrast, the low ER-positive, PGR/HER2-negative breast cancer in vivo models (33). The ability of breast tumor cells tumor (patient 2) displayed relatively little alteration in gene to communicate with and affect host cells would therefore appear expression between the primary and the metastatic tumors. These to be essential to its ability to colonize and survive at a distant site. data suggest that the primary tumor from patient 2 may harbor Elucidation of differential gene expression in metastatic tumors metastatic progression genes (8), which could have altered func- is a key step to understanding the complex mechanisms control- tions at the primary and metastatic sites. The extent of divergence ling tumor adaptability. Here, we define gene expression altera- in gene expression between primary and metastatic tumors would tions that occur in endocrine treated metastatic liver tumors. therefore appear to be inversely related to the aggressiveness of the These global studies build on model systems and clinical obser- primary tumor. vations, which along with further metastatic sequencing studies Signal pathways enriched at the metastatic site were also patient will enable rational personalized treatment strategies to be devel- specific and closely related to the biology of the primary tumor. oped for endocrine-resistant breast cancer patients. The ER/PGR/HER2-positive patient (patient 3) displayed elevations in plasticity networks, including the WNT signaling path- Disclosure of Potential Conflicts of Interest ways, whereas the ER/PGR-positive, HER2-negative patient No potential conflicts of interest were disclosed. (patient 1) had enhanced growth factor signaling. Increases in de-differentiation signals and in particular induction of steroid Authors' Contributions receptor/growth factor receptor cross-talk have been well Conception and design: D. McCartan, L.S. Young described as mechanisms of tumor adaptability to endocrine Development of methodology: J. McBryan, A. Fagan, D. McCartan treatment (16, 25, 26). Furthermore, elevations in key compo- Acquisition of data (provided animals, acquired and managed patients, nents of the mTOR pathway were observed in both patient 1 and provided facilities, etc.): J. McBryan, A. Fagan, D. McCartan, D. Vareslija, patient 3 in metastatic liver tissue. Activation of the mTOR S. Cocchiglia, C. Byrne, J. Bolger, M. McIlroy, L. Hudson, A.D. Hill pathway has previously been associated in endocrine resistance Analysis and interpretation of data (e.g., statistical analysis, biostatistics, (14, 27) and elevated expression of p-mTOR, p-4EBP1, and computational analysis): J. McBryan, D. Vareslija, C. Byrne, J. Bolger, P. Tibbitts, P. O Gaora, L.S. Young p-p70S6K has been reported in metastatic tumors in comparison Writing, review, and/or revision of the manuscript: J. McBryan, A. Fagan, with matched primary tissue (28). S. Cocchiglia, J. Bolger, M. McIlroy, P. Tibbitts, P. O Gaora, L.S. Young The impact of sequential endocrine therapy in the treatment of Administrative, technical, or material support (i.e., reporting or organizing endocrine-resistant disease in patients, however, remains to be data, constructing databases): A. Fagan, F.T. Bane, S. Cocchiglia, L. Hudson fully clarified. Though several clinical studies suggest that some Study supervision: J. McBryan, P. O Gaora, L.S. Young metastatic patients respond to continued endocrine treatment, a subset of patients receives little or no benefit (29–31). Here, using Acknowledgments endocrine-resistant xenografts to model sequential treatment The authors thank Robert Clarke, Georgetown University, for his gift of LY2 strategies, we found that the development of distant metastasis cells. was dependent on continued tamoxifen treatment. Though the molecular mechanisms driving this continued resistance have not Grant Support ex vivo been described to date, studies demonstrating gain-of- Funding support was provided by Science Foundation Ireland (08-IN1- function mutations in breast metastatic patient tumors following B1853 and 12/1A/1294), the Health Research Board of Ireland (HRB/POR/ endocrine treatment suggest that ER may remain an important 2012/101) and Breast Cancer Campaign (2013MaySP022). This study is also player (7). In this study, the majority of DEGs were elevated in the based upon works supported by the Irish Cancer Society Collaborative Cancer metastatic tumors relative to both the primary and the node, with Research Centre grant, CCRC13GAL. The costs of publication of this article were defrayed in part by the payment of little alteration observed between the primary and the matched page charges. This article must therefore be hereby marked advertisement in nodal tissue. The gene expression changes that occur during the accordance with 18 U.S.C. Section 1734 solely to indicate this fact. latency period, under the pressure of treatment are therefore those that are important for the establishment of distant metastatic Received August 19, 2014; revised June 26, 2015; accepted July 21, 2015; disease. published OnlineFirst August 3, 2015.

References 1. Nicolini A, Giardino R, Carpi A, Ferrari P, Anselmi L, Colosimo S, et al. 2. Sherry MM, Greco FA, Johnson DH, Hainsworth JD. Metastatic breast Metastatic breast cancer: an updating. Biomed Pharmacother 2006;60: cancer conﬁned to the skeletal system. An indolent disease. Am J Med 548–56. 1986;81:381–6.

OF8 Clin Cancer Res; 2015 Clinical Cancer Research

Transcriptomic Proﬁles of Metastatic Breast Cancer

3. Luzzi KJ, MacDonald IC, Schmidt EE, Kerkvliet N, Morris VL, Chambers AF, antiestrogen resistance in LY2 human breast cancer cells. PLoS One 2013;8: et al. Multistep nature of metastatic inefficiency: dormancy of solitary cells e62334. after successful extravasation and limited survival of early micrometastases. 19. Lawrence MS, Stojanov P, Polak P, Kryukov GV, Cibulskis K, Sivachenko A, Am J Pathol 1998;153:865–73. et al. Mutational heterogeneity in cancer and the search for new cancer- 4. McBryan J, Theissen SM, Byrne C, Hughes E, Cocchiglia S, Sande S, et al. associated genes. Nature 2013;499:214–8. Metastatic progression with resistance to aromatase inhibitors is driven by 20. Kleppe M, Levine RL. Tumor heterogeneity confounds and illuminates: the steroid receptor coactivator SRC-1. Cancer Res 2012;72:548–59. assessing the implications. Nat Med 2014;20:342–4. 5. Broom RJ, Tang PA, Simmons C, Bordeleau L, Mulligan AM, O'Malley FP, 21. Meacham CE, Morrison SJ. Tumour heterogeneity and cancer cell plasticity. et al. Changes in estrogen receptor, progesterone receptor and Her-2/neu Nature 2013;501:328–37. status with time: discordance rates between primary and metastatic breast 22. Janne PA, Freidlin B, Saxman S, Johnson DH, Livingston RB, Shepherd cancer. Anticancer Res 2009;29:1557–62. FA, et al. Twenty-five years of clinical research for patients with limited- 6. Robinson DR, Wu YM, Vats P, Su F, Lonigro RJ, Cao X, et al. Activating ESR1 stage small cell lung carcinoma in North America. Cancer 2002;95: mutations in hormone-resistant metastatic breast cancer. Nat Genet 1528–38. 2013;45:1446–51. 23. Briele HA, Das Gupta TK. Natural history of cutaneous malignant mela- 7. Jeselsohn R, Yelensky R, Buchwalter G, Frampton G, Meric-Bernstam F, noma. World J Surg 1979;3:255–70. Gonzalez-Angulo AM, et al. Emergence of constitutively active estrogen 24. Lee YT. Patterns of metastasis and natural courses of breast carcinoma. receptor-alpha mutations in pretreated advanced estrogen receptor-posi- Cancer Metastasis Rev 1985;4:153–72. tive breast cancer. Clin Cancer Res 2014;20:1757–67. 25. Arpino G, Wiechmann L, Osborne CK, Schiff R. Crosstalk between the 8. Nguyen DX, Bos PD, Massague J. Metastasis: from dissemination to organ- estrogen receptor and the HER tyrosine kinase receptor family: molecular specific colonization. Nat Rev Cancer 2009;9:274–84. mechanism and clinical implications for endocrine therapy resistance. 9. Kang Y, Siegel PM, Shu W, Drobnjak M, Kakonen SM, Cordon-Cardo C, Endocr Rev 2008;29:217–33. et al. A multigenic program mediating breast cancer metastasis to bone. 26. Montemurro F, Di Cosimo S, Arpino G. Human epidermal growth factor Cancer Cell 2003;3:537–49. receptor 2 (HER2)-positive and hormone receptor-positive breast cancer: 10. Gupta GP, Perk J, Acharyya S, de Candia P, Mittal V, Todorova-Manova K, new insights into molecular interactions and clinical implications. Ann et al. ID genes mediate tumor reinitiation during breast cancer lung Oncol 2013;24:2715–24. metastasis. Proc Natl Acad Sci U S A 2007;104:19506–11. 27. Tokunaga E, Hisamatsu Y, Tanaka K, Yamashita N, Saeki H, Oki E, et al. 11. Swarbrick A, Roy E, Allen T, Bishop JM. Id1 cooperates with oncogenic Ras Molecular mechanisms regulating the hormone sensitivity of breast cancer. to induce metastatic mammary carcinoma by subversion of the cellular Cancer Sci 2014;105:1377–83. senescence response. Proc Natl Acad Sci U S A 2008;105:5402–7. 28. Beelen K, Hoefnagel LD, Opdam M, Wesseling J, Sanders J, Vincent AD, 12. Valiente M, Obenauf AC, Jin X, Chen Q, Zhang XH, Lee DJ, et al. Serpins et al. PI3K/AKT/mTOR pathway activation in primary and corresponding promote cancer cell survival and vascular co-option in brain metastasis. metastatic breast tumors after adjuvant endocrine therapy. Int J Cancer Cell 2014;156:1002–16. 2014;135:1257–63. 13. O'Hara J, Vareslija D, McBryan J, Bane F, Tibbitts P, Byrne C, et al. AIB1: 29. Barrios C, Forbes JF, Jonat W, Conte P, Gradishar W, Buzdar A, et al. The ERalpha transcriptional activity is selectively enhanced in aromatase sequential use of endocrine treatment for advanced breast cancer: where are inhibitor-resistant breast cancer cells. Clin Cancer Res 2012;18:3305–15. we? Ann Oncol 2012;23:1378–86. 14. Beelen K, Opdam M, Severson TM, Koornstra RH, Vincent AD, Wesseling J, 30. Goss PE, Ingle JN, Pater JL, Martino S, Robert NJ, Muss HB, et al. Late et al. Phosphorylated p-70S6K predicts tamoxifen resistance in postmen- extended adjuvant treatment with letrozole improves outcome in women opausal breast cancer patients randomized between adjuvant tamoxifen with early-stage breast cancer who complete 5 years of tamoxifen. J Clin versus no systemic treatment. Breast Cancer Res 2014;16:R6. Oncol 2008;26:1948–55. 15. Pathiraja TN, Nayak SR, Xi Y, Jiang S, Garee JP, Edwards DP, et al. Epigenetic 31. Mauri D, Pavlidis N, Polyzos NP, Ioannidis JP. Survival with aromatase reprogramming of HOXC10 in endocrine-resistant breast cancer. Sci Transl inhibitors and inactivators versus standard hormonal therapy in Med 2014;6:229ra41. advanced breast cancer: meta-analysis. J Natl Cancer Inst 2006;98: 16. McIlroy M, McCartan D, Early S, O Gaora P, Pennington S, Hill AD, et al. 1285–91. Interaction of developmental transcription factor HOXC11 with steroid 32. Van den Eynden GG, Majeed AW, Illemann M, Vermeulen PB, Bird NC, receptor coactivator SRC-1 mediates resistance to endocrine therapy in Hoyer-Hansen G, et al. The multifaceted role of the microenvironment in breast cancer [corrected]. Cancer Res 2010;70:1585–94. liver metastasis: biology and clinical implications. Cancer Res 2013;73: 17. Loh YN, Hedditch EL, Baker LA, Jary E, Ward RL, Ford CE. The Wnt 2031–43. signalling pathway is upregulated in an in vitro model of acquired 33. Tabaries S, Dong Z, Annis MG, Omeroglu A, Pepin F, Ouellet V, et al. tamoxifen resistant breast cancer. BMC Cancer 2013;13:174. Claudin-2 is selectively enriched in and promotes the formation of breast 18. Manavalan TT, Teng Y, Litchfield LM, Muluhngwi P, Al-Rayyan N, Klinge cancer liver metastases through engagement of integrin complexes. Onco- CM. Reduced expression of miR-200 family members contributes to gene 2011;30:1318–28.

www.aacrjournals.org Clin Cancer Res; 2015 OF9

Transcriptomic Profiling of Sequential Tumors from Breast Cancer Patients Provides a Global View of Metastatic Expression Changes Following Endocrine Therapy

Jean McBryan, Ailís Fagan, Damian McCartan, et al.

Clin Cancer Res Published OnlineFirst August 3, 2015.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-14-2155

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2015/08/08/1078-0432.CCR-14-2155.DC1

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/early/2015/09/30/1078-0432.CCR-14-2155. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.