Oct4轉錄調控網路參與肺癌惡化及抗藥性之機制探討

(1)

I

科技部補助專題研究計畫成果報告

期末報告

Oct4 轉錄調控網路參與肺癌惡化及抗藥性之機制探討

(3/3)

計畫類別：



個別型計畫

計畫編號：MOST103-2320-B-006-045-MY3

執行期間：104 年 8 月 1 日至 104 年 7 月 31 日（第三年）

執行機構及系所：成功大學醫學院藥理所

計畫主持人：王憶卿

共同主持人：賴吾為

計畫參與人員：曾鴻泰、湯硯安、任婕羽、呂英鴻、吳莉婷、林哲仲

期中報告處理方式：

1. 公開方式：

非列管計畫亦不具下列情形，立即公開查詢

□涉及專利或其他智慧財產權，□一年□二年後可公開查詢

2.「本研究」是否已有嚴重損及公共利益之發現：否 □是

3.「本報告」是否建議提供政府單位施政參考 否 □是

中華民國 106 年 10 月 22 日

(2)

1 中文關鍵詞：Oct4、轉錄調控網路、PTEN、TNC、長鏈非編碼 RNA、NEAT1、

MALAT1、UCA1、肺癌。

研究背景: 幹細胞轉錄因子 Oct4 在肺癌以及其他多種癌症中有高表達的

現象。長鏈非編碼 RNAs (long non-coding RNAs, lncRNAs) 於近年的研究也發

現會參與在癌症的進程。

研究目的: 本研究旨在(1)透過辨識 Oct4 在肺癌中轉錄調控的基因群，並

進一步觀察基因表現調控異常時所導致的肺癌惡化情形，來探討 Oct4 在肺癌

中所扮演的角色；(2)探討 Oct4 是如何透過轉錄層次去調控長鏈非編碼 RNAs 的

表現，而進一步參與在肺癌的惡化。

研究結果: 藉由全面性的基因 chromatin-immunoprecipitation-sequencing

(ChIP-seq) 檢測，我們辨識出 Oct4 結合位置以及可能為 Oct4 轉錄調控的基

因。透過定量反轉錄 PCR (qRT-PCR) 實驗驗證 29 個基因，Oct4 會透過轉錄

調控使抑癌基因 PTEN 表現量下降，致癌基因 TNC 表現量上升。臨床病人中

PTEN 蛋白表現與 Oct4 存在負相關性 (P=0.006)，且高度表現 Oct4 同時低表

現 PTEN 的病人和其他病人相較之下會有較差的存活率 (P=0.049)。我們發現

在 PTEN 以及 TNC 的基因上游 promoter 位置皆具有 Sp1 轉錄因子的 DNA

結合序列，並鄰近於 Oct4 的 DNA 結合序列。實驗結果顯示 Oct4 會和 Sp1 及

HDAC1/2 形成複合體，Oct4 需透過 Sp1 蛋白質結合到 PTEN 的 promoter 位

置，而相反地 Oct4 可以獨立結合到 TNC promoter。藉由 ChIP-seq 檢測，我們

辨識出 Oct4 所調控的 lncRNAs，篩選出了 14 個距離 Oct4 結合位置小於 100kb

的 lncRNAs，進一步透過在 A549 及 CL1-0 肺癌細胞株中進行 ChIP-PCR 和

qRT-PCR 分析，確定 Oct4 會結合到 7 個致癌型 lncRNA 基因的啟動子 (promoter)

和加強子 (enhancer) 去增進它們的表現。針對三個已在其他癌症 lncRNAs

(NEAT1、MALAT1、UCA1)，以 dual luciferase activity assay 證明了在 A549 和

CL1-0 肺癌細胞株中，Oct4 會直接促進 NEAT1 啟動子及 MALAT1 和 UCA1 加

強子的轉錄活性；然而在 Oct4 結合的位置進行突變則會消除 Oct4 對此三個

lncRNAs 的轉錄活化。

結論：我們的研究結果顯示，Oct4 可以協同其他轉錄因子調控許多與肺癌

進展相關的基因表現，包括細胞增生、轉移，與產生抗藥性相關的 PTEN、TNC

基因，來促使肺癌的惡化 (本篇論文已發表於 Nucleic Acids Res. 43(3): 1593–

1608, 2015)。Oct4 透過增強 NEAT1 啟動子和 MALAT1 加強子之轉錄活性來促

進 NEAT1 和 MALAT1 的 lncRNAs 表達進而導致肺癌的惡化 (本篇論文已發表

於 Mol Cancer 16(1):104, 2017)。

(3)

2 英文摘要

英文關鍵詞：Oct4; transcriptional network; PTEN; TNC; long non-coding RNAs;

NEAT1; MALAT1; UCA1; lung cancer.

Background: The precise transcriptional control of Oct4, a key stemness

transcription factor, in tumorigenesis remains unclear. Long non-coding RNAs

(lncRNAs) emerge recently to play vital roles in regulating multiple cellular functions.

However, there is no evidence that Oct4 modulates lncRNAs expression leading to

tumorigenesis.

Purpose: We aim to elucidate (1) the interacting partners and transcriptional

downstream targets of Oct4, then correlating Oct4 transcriptional regulation with

tumorigenesis events involving Oct4 target genes; (2) elucidate the transcription

regulation of Oct4 on lncRNAs in promoting lung tumorigenesis using.

Results: Oct4 binding sites and putative transcriptional targets were identified via

genome-wide

chromatin-immunoprecipitation-sequencing

(ChIP-seq)

analysis.

Validation of 29 putative target genes by qRT-PCR showed that Oct4 downregulated

tumor suppressor gene PTEN and induced expression of oncogene TNC. Clinical PTEN

protein expression had an inverse correlation with Oct4 (P=0.006), and patients with

Oct4 overexpression and PTEN low expression had poor prognosis (P=0.049). Notably,

the transcription factor Sp1 binding motifs are adjacent to the Oct4 binding sites in

PTEN and TNC promoter regions. We confirmed that Oct4, Sp1, HDAC1 and HDAC2

formed a complex in Oct4 stable cells. Oct4 bound to PTEN promoter in a

Sp1-dependent manner, while Oct4 bound to TNC promoter independently of Sp1.

Among Oct4 putative transcriptional targets, we identified several lncRNAs that the

Oct4 binding sites located within 100kb from the lncRNA genomic locus. We further

focused Oct4 regulation on NEAT1, MALAT1 and UCA1 lncRNAs. Using dual

luciferase activity assay, we demonstrated that Oct4 potentiated transcription of NEAT1

through targeting on its promoter while potentiated transcription of MALAT1 and UCA1

through targeting on their enhancers in lung cancer cells. Furthermore, NEAT1 and

MALAT1 acted as downstream regulators of Oct4-mediated tumorigenesis evident by

reconstitution experiments.

Conclusion: Our results suggest Oct4 cooperates with Sp1 transcription factor in

regulating expressions of critical genes PTEN and TNC in promoting lung

tumorigenesis

(

Nucleic Acids Res. 43(3): 1593–1608, 2015.

).

Our study provides novel

results that Oct4 upregulates NEAT1 and MALAT1 transcription via promoter and

enhancer activation, respectively, to enhance lung cancer proliferation, migration and

and invasion (Mol Cancer 16(1):104, 2017).

(4)

3 Oct4, also called POU5F1 (POU domain, class 5, transcription factor 1), is a

homeodomain transcription factor of the POU family. Oct4 is a key regulator for

maintaining pluripotency of embryonic stem cells (ESCs) by regulating

transcriptional programs in coordination with various transcription factors, such as

Sox2, Nanog and Klf4

1-3

, or with transcriptional repression complexes, such as

histone deacetlyases (HDACs)-containing complex

4

. In addition to maintaining

ESCs, recent studies have reported overexpression of Oct4 in breast

5

, bladder

6

,

glioma

7

, oral squamous cell carcinoma

8

, esophageal squamous cell carcinoma

9

,

cervical cancer

10

and lung cancer

11,12

. Moreover, study showed that coexpression of

Oct4 and Nanog confers hepatocellular carcinoma cells cancer stem cell (CSC)

properties and induces epithelial-to-mesenchymal (EMT) transition through

Stat3-mediated activation of Snail expression

13

. Altogether, these studies suggest

that Oct4 is an oncogenic transcription factor that promotes CSC properties.

However, Oct4-mediated transcription reprogramming in CSC remain largely

unknown.

LncRNAs is a subset of non-coding RNAs with length ranging from 200bp to

100kbp. LncRNAs have been demonstrated to exert their functions through the

mechanisms, which are (1) binding with chromatin-modifying complexes to guide

these chromatin modifiers to chromosome to regulate gene expressions, (2) acting as

scaffolds between proteins to maintain the integrity of protein complexes, (3)

decoying transcription factors from promoters to modulate gene expressions, and (4)

sponging miRNAs from binding with their target mRNA and thus allowing the

mRNA translation

14

. The next generation RNA-seq studies have identified

dysregulation of lncRNAs in cancers and some of them may be potential prognosis

(5)

4 markers due to their expression changes in specific cancer types

15,16

. LncRNAs have

been characterized as oncogene-like or tumor suppressor-like lncRNAs and linked

with the hallmarks of cancer, including extensive proliferation, invasiveness,

angiogenesis and chemoresistance by functional studies. Although studies had

reported the association of dysregulated lncRNA expression with cancer progression,

less studies investigate how expressions of lncRNAs are regulated, especially at

transcriptional level.

In our previously established Oct4-mediated genome-wide ChIP-seq datasets,

we identified several lncRNAs which may be regulated by Oct4 in lung cancer

12

. In

current study, we validated the transcriptional regulation of Oct4 on lncRNAs,

including NEAT1, MALAT1 and UCA1. Functional studies further confirmed the

oncogenic roles of Oct4-mediated NEAT1 and MALAT1 up-regulation in promoting

proliferation and metastasis abilities of lung cancer cells.

(6)

5 Oct4 acts as an oncogenic transcription factor in cancers. In this three-year

study, we performed Oct4 chromatin-immunoprecipitation coupled with

high-throughput sequencing (ChIP-seq) in comparison of A549 vector control and

Oct4 stable cells. We identified 5,370 Oct4 binding sites, which corresponded to

3,350 genes including the genomic loci of lncRNAs. As eluted in the previous

section, some lncRNAs have been shown to modulate expressions of target genes to

cause broad effects in cancer cells.

Till now, the controls of lncRNA expressions are still unclear although they

have been emerging as critical regulators in cancers. Therefore, further studies on

transcriptional regulation of lncRNAs, as Oct4 downstream target genes, not only

address the oncogenic effects of Oct4 in lung cancer but also elucidate the

transcription regulation of lncRNAs. The present study proposed the following aims

to clarify the transcription regulation of Oct4 on lncRNAs in lung cancer.

I. Identification and validation of Oct4-regulated lncRNAs

According to our previous genome-wide Oct4 ChIP-seq dataset, Oct4 can bind

around genomic loci of some lncRNAs. Therefore, this study further validated the

transcription regulation of Oct4 on these lncRNAs using the approaches described

below.

1. Putative Oct4-targeted lncRNAs were identified by integration of in-house

ChIP-seq with functional lncRNA database, lncRNA db and literature search. We

focused on those lncRNAs with Oct4 binding sites located within 100kb from

lncRNAs genomic loci and classified them into promoter controlled and enhancer

controlled lncRNAs by Oct4.

(7)

6 defined by ALGGEN PROMO software and TFSEARCH software. ChIP-PCR was

further used to validate Oct4 binding to the lncRNA promoter or enhancer.

3. Oct4 transcription regulation on lncRNAs was confirmed by qRT-PCR

and dual luciferase assays.

II. Demonstration of Oct4-lncRNAs axis promoting cancer cells proliferation

and migration

LncRNAs have been reported to be involved in cancer progression. In this

study, we used gain and loss of function approaches to elucidate if lncRNAs

participated in Oct4 promoting lung cancer proliferation, migration and invasion.

1. The roles of lncRNAs in cell proliferation, migration and invasion were

examined in lung cancer cells manipulated with the expression level of lncRNAs.

2. Oct4-regulated lncRNAs in promoting oncogenic effects were confirmed

by reconstitution after overexpression of lncRNAs in Oct4 knockdown lung cancer

cells.

3. The correlation between Oct4 mRNA and lncRNAs expression levels was

examined in clinical specimens through our in-house cohort and public datasets of

lung cancer patients.

(8)

7 1. Y.-A. Tang, C.-H. Chen, H. S. Sun, C.-P. Cheng, V. S. Tseng, H.-S. Hsu, W.-C.

Su, W.-W. Lai, Yi-Ching Wang. 2015. Global Oct4 target gene analysis

reveals novel downstream PTEN and TNC genes required for drug-resistance

and metastasis in lung cancer. Nucleic Acids Res. 43(3):1593-608 (IF 10.162,

14/290 in “Biochemistry & Molecular Biology”, cited 10 times)

2. Jayu Jen, Y.-A. Tang, Y.-H. Lu, C.-C. Lin, H.-S. Hsu, W.-W. Lai, Yi-Ching

Wang. 2017. Oct4 transcriptionally regulates the expression of long

non-coding RNAs NEAT1 and MALAT1 to promote lung cancer progression.

Mol Cancer 16(1):104. doi: 10.1186/s12943-017-0674-z (SCI 6.204, 31/217 in

“ONCOLOGY”)

(9)

Published online 21 January 2015 Nucleic Acids Research, 2015, Vol. 43, No. 3 1593–1608

doi: 10.1093/nar/gkv024

Global Oct4 target gene analysis reveals novel

downstream

PTEN

and

TNC

genes required for

drug-resistance and metastasis in lung cancer

Yen-An Tang

1,2

_{, Chi-Hsin Chen}

2

_{, H. Sunny Sun}

3

_{, Chun-Pei Cheng}

4

_{, Vincent S. Tseng}

4,5

_,

Han-Shui Hsu

6

_{, Wu-Chou Su}

7

_{, Wu-Wei Lai}

8

_{and Yi-Ching Wang}

1,2,*

1_{Institute of Basic Medical Sciences, National Cheng Kung University, No.1, University Road, Tainan 701, Taiwan,} 2_{Department of Pharmacology, National Cheng Kung University, No.1, University Road, Tainan 701, Taiwan,}3_Institute

of Molecular Medicine, College of Medicine, National Cheng Kung University, No.1, University Road, Tainan 701, Taiwan,4_{Department of Computer Science and Information Engineering, National Cheng Kung University, No.1,}

University Road, Tainan 701, Taiwan,5_{Institute of Medical Informatics, College of Electrical Engineering and}

Computer Science, National Cheng Kung University, No.1, University Road, Tainan 701, Taiwan,6_{Division of Thoracic}

Surgery, Taipei Veterans General Hospital; Institute of Emergency and Critical Care Medicine, National Yang-Ming University School of Medicine, No.155, Sec.2, Linong Street, Taipei 112, Taiwan,7_{Department of Internal Medicine,}

National Cheng Kung University Hospital, No.138, Sheng Li Road, Tainan 704, Taiwan and8_{Department of Surgery,}

National Cheng Kung University Hospital, No.138, Sheng Li Road, Tainan 704, Taiwan

Received October 15, 2014; Revised December 21, 2014; Accepted January 10, 2015

ABSTRACT

Overexpression of Oct4, a stemness gene encoding a transcription factor, has been reported in several cancers. However, the mechanism by which Oct4 di-rects transcriptional program that leads to somatic cancer progression remains unclear. In this study, we provide mechanistic insight into Oct4-driven tran-scriptional network promoting drug-resistance and metastasis in lung cancer cell, animal and clinical studies. Through an integrative approach combining our Oct4 chromatin-immunoprecipitation sequenc-ing and ENCODE datasets, we identified the genome-wide binding regions of Oct4 in lung cancer at pro-moter and enhancer of numerous genes involved in critical pathways which promote tumorigenesis. Notably, PTEN andTNC were previously undefined targets of Oct4. In addition, novel Oct4-binding mo-tifs were found to overlap with DNA elements for Sp1 transcription factor. We provided evidence that Oct4 suppressedPTENin an Sp1-dependent manner by recruitment of HDAC1/2, leading to activation of AKT signaling and drug-resistance. In contrast, Oct4 transactivatedTNCindependent of Sp1 and resulted in cancer metastasis. Clinically, lung cancer patients with Oct4 high, PTEN low and TNC high expression profile significantly correlated with poor disease-free survival. Our study reveals a critical Oct4-driven

tran-scriptional program that promotes lung cancer pro-gression, illustrating the therapeutic potential of tar-geting Oc4 transcriptionally regulated genes. INTRODUCTION

Lung cancer is the leading cause of cancer deaths (1). The 5-year survival of non-small cell lung cancer (NSCLC) is

<16%, partly due to metastasis and relapse from current

therapeutic strategies (1). Traditionally, platinum-based chemotherapy is considered the first line treatment but it shows limited effectiveness (2). Recently, new encouraging targeted therapies for sub-populations of NSCLC patients harboring epidermal growth factor receptor or anaplas-tic lymphoma kinase mutations have been developed (3,4). Therefore, further understanding of the molecular mecha-nisms that drive NSCLC will be useful for development of innovative therapeutic interventions.

Tumor-initiating cells (TICs) are defined as a sub-population of tumors that have the ability to self-renew and differentiate into nontumorigenic cells, thereby con-tributing to the functional and phenotypic heterogeneity in diverse cancer types (5). TICs have been shown to be more resistant to chemotherapy and radiotherapy and pos-sess metastatic potential (5). Previous study demonstrated that regional pulmonary stem cell population, termed bronchioalveolar stem cells (BASCs), exhibit properties of self-renewal and multipotency. Oncogenically transformed BASCs give rise to lung adenocarcinoma, suggesting the ex-istence of lung TICs (6). Primary lung tumors contain

sub-*_{To whom correspondence should be addressed. Tel: +886 6 2353535 (Ext 5502); Fax: +886 6 2749296; Email: [email protected]}

C

The Author(s) 2015. Published by Oxford University Press on behalf of Nucleic Acids Research.

This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact [email protected]

at National Cheng Kung University, Medical Library on March 1, 2015

http://nar.oxfordjournals.org/

Downloaded from

論文

1 、

Nucleic Acids Research 2015

(10)

populations of cells, which are characterized by CD133+

surface marker, showing stem-like features with increased tumorigenicity and drug-resistance (7–9). The CD133+

tu-morigenic cancer-initiating cells are maintained by highly expressed Oct4 transcription factor (8). Additionally, ec-topically co-expression of Oct4 and Nanog in lung cancer cell lines induced TICs properties and promoted epithelial-mesenchymal transition (10). These findings suggest that Oct4-driven lung TICs may be effectively targeted to ben-efit therapeutic response. Recently, accumulative evidences demonstrate that Oct4 is overexpressed in various solid tu-mors, including breast cancer (11), bladder cancer (12) and lung cancer (10,13). Targeting of Oct4 promotes cell death in breast and lung cancer cells (14,15), and sensitizes drug-resistant hepatocellular carcinoma to chemotherapy (16). However, the mechanism by which Oct4 influences tran-scriptional reprogramming that leads to somatic cancer pro-gression remains largely unclear. The target gene networks of Oct4 in somatic cancer are also undiscovered.

Oct4, encoded by POU5F1, belongs to a member of the POU (Pit-1, Oct1/2 and UNC-86) family of transcription factors. Oct4 is a key regulator for maintaining pluripo-tency of embryonic stem cells (ESCs) by regulating scriptional programs in coordination with various tran-scription factors, such as Sox2, Nanog and Klf4 (17–19), or with transcriptional repression complexes, such as his-tone deacetlyases (HDACs)-containing complex (20). The genome-wide binding profile of Oct4 in ESCs has been re-vealed in independent studies using different genome-wide approaches (17,21–23). The functions of Oct4 in ESCs are well defined, however, little is known about those in somatic cancer cells or TICs. Investigation of Oct4-centered tran-scriptional networks in somatic cancers will not only un-cover how Oct4 cooperates with other transcriptional fac-tors to regulate downstream genes, but also reveal the cel-lular mechanisms of Oct4-induced stem-like characteristics, such as drug resistance and metastasis.

Here, we demonstrate the first genome-wide binding profile of Oct4 in lung cancer cells using chromatin-immunoprecipitaion followed by deep sequencing (ChIP-seq). We identified that Oct4 binds to thousands genomic re-gions in lung cancer cells. De novo motif and sequence sim-ilarity analyses showed that some novel Oct4-binding mo-tifs contained the DNA elements for transcription factors such as Sp1, Klf4, ZNF219 and Stat3. Pathway analyses showed that Oct4 downstream target genes play key roles in several tumorigenesis events and important signaling path-ways, such as phosphatase and tensin homolog (PTEN) sig-naling. Mechanistically, Oct4 suppressed tumor suppressor genes, such as PTEN, whereas Oct4 activated oncogenes, such as Tenascin-C (TNC) via differential interplay with the Sp1 transcription factor. Oct4-mediated downregulation of

PTEN contributed to drug resistance, while TNC induction

was required for Oct4-elicited cell invasion and metastasis in cell and animal models. Lung cancer patients with ad-vanced disease showed high Oct4 protein expression coin-ciding with low PTEN and high TNC in tumors.

MATERIALS AND METHODS

Cell lines and culture conditions

Human lung adenocarcinoma cell lines A549 were pur-chased from American Tissue Culture Company (ATCC). Human lung adenocarcinoma cell line CL1–0 and CL1–5 was obtained from Dr Pan-Chyr Yang (Department of In-ternal Medicine, Medical College, National Taiwan Uni-versity, Taiwan). All media were supplemented with 10% Fetal Bovine Serum (Gibco, Carlsbad, CA, USA) and 1% penicillin/streptomycin (Gibco). Stable cell line expressing Oct4 (stable Oct4 expressing cells) or vector was established by ectopic transfection of Flag-Oct4 or empty vector plas-mid into A549 cells with puromycin selection.

Plasmid, RNAi and transfection

The plasmids and interference RNA (RNAi) used in the study are listed in Supplementary Tables S1 and S2. Plas-mids and RNAi transfections were carried out with lipofe-tamine 2000 (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s protocols.

Tumor-sphere formation assay

Cells were expanded as spheres in a 10-cm ultra-low adhe-sion culture dish (Corning Inc., Corning, New York, NY, USA) containing DMEM/F-12 with N2 supplement (In-vitrogen), 20 ng/ml epithelial growth factor and 20 ng/ml basic fibroblast growth factor (PeproTech Inc., Rocky Hill, NJ, USA), referred to as stem cell medium. Tumor spheres consisting of>30 cells were counted and expressed as the means± SEM of triplicate within the same experiment.

Tumor formation assay

The 5–6-week-old BALB/c nude female mice were subcu-taneously implanted with varying number (1 × 102_{, 1} _×

103_{or 5}_{× 10}3_{) of vector control (vector) or Oct4-stably}

ex-pressing A549 cells (Oct4#1). Vector and Oct4#1 cells in 50 ␮l Hanks’ balanced salt solution (HBSS) were mixed with 50␮l matrigel (2.5 mg/ml, Sigma-Aldrich, St Louis, MO, USA) and then subcutaneously injected into each flank of a mouse. The incidence of tumor formation was monitored within 8 weeks after implantation.

Anchorage-independent growth assay

Anchorage-independent colony formation assay was car-ried out by growing 3× 103_{cells in 0.4% bactoagar on a}

bot-tom layer of solidified 0.6% bactoagar in 6-well plates. After 12–17 days, colonies consisting of>30 cells were counted and expressed as the means± SEM of triplicate within the same experiment.

Transwell migration and invasion assay

The transwell insert with millipore membrane (pore size of 8 ␮m, Falcon, BD, Franklin Lakes, NJ, USA) was used. For transwell migration assay, 2× 105 _{cells were seeded onto}

the upper chamber with 1 ml serum-free medium. For in-vasion assay, the transwell insert membrane was pre-coated

(11)

Nucleic Acids Research, 2015, Vol. 43, No. 3 1595

with Matrigel (2.5 mg/ml, Sigma-Aldrich) one day before cells were seeded. Complete medium containing 20% FBS was added to the lower chamber as chemoattractants and cells were incubated for 24 h. The cells attached on the re-verse side of the membrane were then fixed and stained with 1% crystal violet/MeOH for 15 min at room temper-ature. Seven random views were photographed and quanti-fied under an upright microscope (Nikon E400, Yurakucho, Tokyo, Japan).

ChIP-seq assay

Stable Oct4 expressing cells or empty vector control (1× 107_{cells) were cross-linked with 1% formaldehyde, followed}

by preparation of nuclear lysates using Magna ChIPTM

pro-tein G Kit (Millipore Co., Billerica, MA, USA) according to the manufacturer’s protocols. Nuclear lysates were son-icated to shear crosslinked DNA, followed by immunopre-cipitation with anti-Oct4 antibody using the conditions de-scribed in Supplementary Table S3. Purified ChIP DNA was used for the preparation of fragment libraries, followed by high-throughput sequencing using a SOLiDTM 5500xl sequencer (Applied Biosystems, Foster City, CA, USA). We obtained about 24–29 million raw reads from vector and stable Oct4 cell line. The raw reads were further analyzed using LifeScopeTM Genomic Analysis Software (version 2.5) and mapped to human genome (hg19) released from UCSC database. To find the significant peak, the mapped profiles were analyzed using the ChIP-seq tool in CLC Ge-nomics Workbench (version 4.9). Window size and false dis-covery rate (FDR) were set to 200 bp and 5%, respectively. To determine the high confidence Oct4 binding loci, the ChIP-region was identified by scanning the peaks with sig-nificantly higher read count in stable cells expressing Oct4 compared to those in the vector control cells. Oct4 ChIP-seq data can be viewed online under GEO accession number GSE58462.

Quantitative ChIP-PCR assay

ChIP assays were performed as described above using anti-Oct4, anti-Sp1, anti-HDAC1, anti-HDAC2 or anti-Rabbit IgG antibody in vector control and A549 cells stably-expressing Oct4. Quantitative polymerase chain reaction (PCR) analysis was performed using Fast SYBR Green Master Mix and StepOnePlusTM_{System (Applied}

Biosys-tems). The primer sequences are listed in Supplementary Table S4.

DNA sequence motif analysis

For de novo motif discovery algorithm, the top 150 ranked Oct4 ChIP-seq peaks (±100 bp from center of the ChIP-seq peaks) were analyzed by MEME software (24) using the rec-ommended default settings. To identify specific transcrip-tion factors that potentially interact with Oct4-targeting motifs, the STAMP software (25) (using the recommended default settings), which compares the motifs discovered by MEME with two existing databases of known motifs, TRANSFAC (v11.3) and JASPAR (v3), was used.

ENCODE datasets analysis

To identify functional DNA elements targeted by Oct4, we correlated our Oct4 ChIP-seq results with datasets from ENCODE project (26) that were deposited to UCSC web-site. The genomic regions of DNase I hypersensitive sites (DHS), H3K4me3 and H3K4me1 obtained from A549 lung cancer cells were downloaded and analyzed. We used a 1-bp minimum cutoff for the overlap between regions to de-fine common genomic targets across all datasets. The Oct4-targeted functional promoters were obtained by integration of four datasets, including Oct4 ChIP-seq, DHS, H3K4me3, and gene promoter regions (within 2 kb upstream from transcriptional start site) annotated by ENCODE (release v17). The Oct4-targeted enhancers were revealed by inte-gration of three datasets, including Oct4 ChIP-seq, DHS, and H3K4me1. To investigate the coordinates of Oct4 and Sp1 binding regions in A549 lung cancer cells, we inte-grated Oct4 ChIP-seq and Sp1 ChIP-seq, which is obtained from ENCODE project, using a 1-bp minimum cutoff as described above.

Gene set enrichment analysis

Gene set enrichment analysis (GSEA) was conducted to determine whether Oct4-target gene set was highly en-riched in the previously reported gene signatures of aggres-sive lung adenocarcinoma (27) or lung TICs (28). GSEA was performed using Oct4-target gene set comprising Oct4-bound top 150, promoter-associated and enhancer-associated genes and using default settings (29). The genes with 1.5-fold increased or decreased change on the microar-rays from two previous studies (27,28) were ranked accord-ing to their differential expression levels across the two dis-tinct phenotypes using a t-Test metric. Gene sets were con-sidered to be highly enriched if FDR q< 0.25. The P-values were determined by a random permutation test.

RNA extraction and qRT-PCR assays

The primers used for quantitative RT-PCR analyses are de-scribed in Supplementary Table S4.

Western blot analysis

Samples containing equal amounts of protein (50 ␮g) were separated on a 10% sodium dodecylsulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and electroblotted onto Immobilon-P membranes (Millipore). Immunoblotting was performed using the conditions described in Supplementary Table S3.

Immunoprecipitation assay

For immunoprecipitation, 600␮g nuclear lysates were incu-bated with 4␮g of the appropriate antibody, including Oct4, Sp1, Rabbit-IgG and Mouse-IgG, then 1× wash buffer was added to a final volume of 480␮l. After incubation at 4◦C for 3 h, 20 ␮l Protein G/Protein A agarose beads (Cal-biochem Co., Darmstadt, Germany) were added. After 1 h of incubation at 4◦C, complexes were washed 3× with 1× immonoprecepitation buffer (50 mM Tris–HCl, pH 7.5,

(12)

150 mM NaCl, 20 mM ␣-glycerol-phosphate, 1% NP-40, 5 mM ethylenediaminetetraacetic acid (EDTA)). Proteins were eluted by boiling in 2× SDS loading buffer, separated by 8% SDS-PAGE, then blotted with Oct4, Sp1, HDAC1 or HDAC2 antibody.

Cell cytotoxicity analyses

Cells were incubated with solvent control or various concentrations of cisplatin, Suberoylanilide hydrox-amic acid (SAHA), MS275 or LY294002 compound for indicated times. Cell cytotoxicity was assayed with 3-(4.5-dimethylthiazol-2-ly)-2,5-diphenyl tetrazolium bromide (MTT, Sigma-Aldrich) according to the manufacturer’s instructions.

Site-directed mutagenesis

Mutations of Sp1 binding sites within PTEN-1 and/or PTEN-2 regions were generated by site-directed mutagene-sis using wild-type (WT) PTEN promoter vector and spe-cific primers listed in Supplementary Table S4. The Oct4 binding site or Sp1 binding site was mutated by site-directed mutagenesis using WT TNC promoter vector and specific primers listed in Supplementary Table S4.

Dual luciferase promoter assay

Cells were plated in 12-well plates the day before transfec-tion. The pGL4-Renilla construct was included as an inter-nal control. After 16 h co-transfection with empty vector or gene promoter vector and pGL4-Renilla, the dual luciferase reporter assay kit (Promega, Madison, WI, USA) was used to determine gene promoter activity according to the pro-tocols provided by the manufacturer. The data were repre-sented as the means of ratio of firefly luciferase to Renilla luciferase activity by triplicate experiments.

DNA affinity precipitation assay (DAPA)

Biotin-labeled DNA probes containing Oct4 and Sp1 bind-ing sites on PTEN and TNC promoter regions were used in DAPA assay. The Oct4 and Sp1 binding motifs within these probes were either WT or mutated (Mut), and the sequences are shown in Supplementary Table S5. Nuclear lysates of A549 cells transfected with empty vector or Oct4 expres-sion vector were incubated with 10␮l streptavidin agarose beads (Thermo Fisher Scientific Inc., Waltham, MA, USA) with wash buffer (20 mM 4-(2-Hydroxyethyl)piperazine-1-ethanesulfonicacid, 0.1 mM KCl, 2 mM MgCl2, 0.2 mM

EDTA. 1 mM DTT, 10% glycerol, 0.01% NP-40 (pH 7.9)) at room temperature for 20 min. Nuclear lysates were then in-cubated with 2␮g biotin-probes for 30 min on ice, followed by rotation at room temperature for 1 h with streptavidin agarose beads. The DNA–protein complexes were washed 3× with wash buffer (with additional 0.25% Triton X-100). Proteins were eluted by boiling in 2× SDS loading buffer, separated by 8% SDS-PAGE, then blotted with Oct4, Sp1, HDAC1 or HDAC2 antibody.

In vivo tumor growth assay

For subcutaneous xenograft model, 5–6-week-old BALB/c nude female mice were subcutaneously implanted with vec-tor control cells or A549 cells stably expressing Oct4 (1 × 107 _{cells per mouse). When tumor xenograft volume}

reached 50 mm3_{, animals were treated intraperitoneally}

with LY294002 (25 mg/kg) and/or SAHA (30 mg/kg) on days 1, 3 and 5 for three weeks. The volume of the xenograft were measured and calculated as (length× width square)/2 in mm3_.

In vivo tumor metastasis assay

Parental A549 cells were transfected with control or Oct4 siRNAs for 24 h, followed by ectopic expression of TNC for another 24 h. Cells (1× 106_cells_{/200 ␮l) were injected}

intravenously into NOD/SCID mice via the tail vein. Mice were sacrificed at 9 weeks after tail vein injection. The lungs with colonized tumor nodules were dissected and stained for further confirmation.

Study population

We recruited 91 lung cancer patients from Veterans General Hospital, Taipei, Taiwan and 42 lung cancer patients from National Cheng Kung University Hospital after obtaining appropriate institutional review board permission and in-formed consent from the patients. Paraffin blocks of tumors were collected. The detailed clinicopathological character-istics of the enrolled patients are listed in Supplementary Table S6.

Immunohistochemistry assay

Antibodies used and their experimental conditions are listed in Supplementary Table S3. Staining was scored as +++ if >75% tumor cells were immunostaining-positive; ++ for 50–75%; + for 25–50%; +/− for 10–25% cells and − if<10% were positive. For Oct4 or TNC protein expression level, the stains were graded as ‘overexpression’ if the score were ++ and +++. For PTEN protein expression level, the stains were graded as ‘low expression’ if the score was equal or less than +.

Statistical analysis

Pearson’s␹2_{t-test was used to determine the significance of}

difference in cell and animal model experiments. Chi-square test was conducted to examine the association between Oct4, PTEN and TNC protein expression levels and clinical pathological parameters. Overall and progression-free sur-vival curves were calculated according to the Kaplan-Meier method using the log-rank test. P<0.05 was considered to be statistically significant.

RESULTS

Identification of global target regions of Oct4 by ChIP-seq in lung cancer cells

Oct4 is a master regulator of pluripotent ESCs and main-tains stem cell-like properties of TICs. The genomic bind-ing pattern of Oct4 in somatic cancer is still unknown

(13)

and uncovering it could reveal how Oct4 drives somatic cancer into TIC-like states. To this end, we established stable Oct4-expressing cell line using A549 lung adeno-carcinoma cells. Interestingly, ectopic expression of Oct4 alone promoted dedifferentiation of A549 cells into TIC-like cells, which possessed increased in vitro self-renewal ca-pacity and increased expression of stemness-related genes, such as SOX2 and NOTCH (Supplementary Figure S1A and SB). Limited numbers of stable Oct4 expressing cells formed tumors in immunodeficiency mice compared to vec-tor control cells (Supplementary Figure S1C). Enhanced cell proliferation, cell motility, as well as drug-resistance to chemotherapy reagent (cisplatin) and HDAC inhibitors (SAHA and MS275) were observed in stable Oct4 express-ing cells (Supplementary Figure S2). In contrast, knock-down of Oct4 diminished anchorage-independent growth and cell invasion/migration ability in parental A549 cells, while restoration of Oct4 expression in Oct4-depleted A549 cells completely reversed the effect of Oct4 siRNA on such phenotypes (Supplementary Figure S3). These data suggest that enforced Oct4 expression alone in A549 cells induces TIC-like states.

Since TICs account for less than 1% in somatic cancer cell lines, it is difficult to address the genome-wide binding regions of Oct4 using A549 lung cancer cell lines. There-fore, we chose the stable Oct4 expressing A549 cells, which showed TIC-like properties (Supplementary Figures S1 and S2), to interrogate the Oct4 occupancy using ChIP-seq ap-proach. Based on two biological replicates, a total of 5,380 ChIP regions corresponding to 3,350 unique RefSeq genes were identified (Supplementary Table S7). A large propor-tion of Oct4 occupancy located at gene body and, to a lesser extent, distal region (Figure1A). With respect to gene tar-geting, our ChIP-seq data showed that Oct4 heavily occu-pied its own promoter and gene body regions (POU5F1) (Figure 1B) as reported in ESC. Our ChIP-seq data re-vealed new Oct4-targeted sites in many cancer-related genes (Supplementary Table S7) in addition to several Oct4 bind-ing loci previously identified in ESCs, includbind-ing ESRRB,

JARID2, TCF3 (21), ZFP42, SIRPA, MYBL2, PCSK6 (30), EOMES, and PAX6 (31). For example, newly iden-tified Oct4-targeted sites were found at promoter region of oncogenic gene GSK3A and intragenic region of tumor sup-pressor gene STAG3 (Figure1B).

Oct4-centered transcriptional network contributes to tumori-genesis

To delineate the importance of Oct4-driven transcriptional network in cancer progression, we conducted Ingenuity Pathway Analysis (IPA) for network and canonic pathway analyses. As shown in Figure 1C, Oct4 bound genes in the top 150 binding sites were enriched in several cellular functions related to tumorigenesis events, such as cellular growth and proliferation, cell death and survival, and cellu-lar movement (P<0.05) in addition to development control. Importantly, these genes are directly involved in several crit-ical pathways, including Wnt/␤-catenin, ErbB2-ErbB3 and PTEN signaling (P<0.05) (Figure1D).

To further verify the effects of Oct4 occupancy on gene expression relevant to tumorigenesis, we next classified

these top Oct4 bound cancer-related genes into two groups, oncogenic-like and tumor suppressor-like genes, based on literature survey. qChIP-PCR confirmed the enriched occu-pancy of endogenous Oct4 at these target genes in parental A549 cells (Figure2A). Interestingly, among 29 genes val-idated, Oct4 was found to regulate many genes, most of which were reported for the first time, such as transacti-vation of oncogenes TNC, HDAC4, LAMB1 and GSK3A (Figure2B, left), or transcriptional suppression of tumor suppressor-like genes, including PTEN, DKK3, FBXO31 and RAB37 (Figure2B, right). Knockdown of Oct4 by two independent Oct4 siRNAs in A549 cells showed opposite effects on gene expression compared to overexpression of Oct4 (Figure2C), thus confirming differential modulation of target genes by Oct4. Collectively, these data showed high confidence of our ChIP-seq results and suggested the important roles of Oct4 in transcriptional regulation of oncogenic-like and tumor suppressor-like genes in lung can-cer cells.

Oct4 occupies promoter and enhancer regions of genes that play key roles in critical signaling pathways involved in cancer progression

Since DNA regulatory elements (i.e. promoter and en-hancer) comprise important information that determines gene expression, we correlated Oct4 binding profiles with datasets from the ENCODE project (26), which deposit in-formative ChIP-seq of histone modifications and chromatin accessibility characterized by DHS of numerous cell lines, including A549 lung cancer cells. We found that Oct4 oc-cupied 258 functional promoters, including PTEN gene, which are characterized by simultaneous DHS and his-tone H3 lysine 4 tri-methylation (H3K4me3) (Figure3A). In addition, Oct4 binding sites were well associated with 431 enhancers, characterized by simultaneous DHS and hi-stone H3 lysine 4 mono-methylation (H3K4me1) (Figure

3B). IPA analysis revealed that promoter-associated and enhancer-associated genes were both significantly associ-ated with cancer disease (P<0.05) and important signal-ing pathways, such as PTEN signalsignal-ing (P<0.001) and in-tegrin signaling (P<0.01) (Figure3C and D). Moreover, we conducted GSEA using a gene set comprising Oct4-bound top 150, promoter- and enhancer-associated genes to com-pare with those aggressive gene signatures of lung cancer published previously. GSEA demonstrated that our Oct4-target gene set correlated with an aggressive gene signature of lung adenocarcinoma (27) and a gene signature of lung TICs (28) (Supplementary Figure S4). These data suggested that Oct4 occupies functional regulatory elements of critical genes and is involved in tumorigenesis.

Oct4 co-occupies distinct binding motifs in lung cancer with transcription factors

The interacting partners of Oct4 have not been defined clearly in somatic cancer. Thus, we adopted MEME soft-ware (24) for de novo motif analysis followed by STAMP software (25) to explore the Oct4 binding DNA sequences and potential Oct4-interacting transcription factors that may mediate the preferential recruitment of Oct4. We ana-lyzed the top 150 Oct4 binding sites and identified not only

(14)

Figure 1. Identification of global binding sites and potential interacting partners of Oct4 in lung cancer cells. (A) Distribution of Oct4 binding sites from ChIP-seq. Schematic diagram illustrates the definition of the location of a binding site in relation to a transcription unit (upper). Pie diagram shows the location of Oct4 binding sites relative to the nearest transcription unit (lower). Gene desert was defined as loci location>100 kb away from the nearest gene. (B) Snapshots of the ChIP-seq binding profiles of Oct4 at POU5F1, GSK3A and STAG3 genes from Integrative Genomics Viewer. (C) Oct4-centered transcriptional network analyzed by Ingenuity-IPA software is shown. Genes validated by qChIP-PCR and qRT-PCR in Figure2are in red circle. (D) Top five canonical signaling pathways analyzed by Ingenuity-IPA using gene set from (C). (E) Potential interaction partners of Oct4. Oct4 binding motifs were obtained by MEME software. The potential transcription factor binding sites related to Oct4 targeting motifs were analyzed by STAMP software. The

E-value for specific transcription factor is shown.

(15)

Figure 2. Oct4 induces expression of oncogenic genes while represses tumor suppressor-like genes. Oct4 target genes were classified into two groups: oncogenic genes (left panel) and tumor suppressor-like genes (right panel). (A) qChIP-PCR analysis confirmed the endogenous Oct4 occupancy at the binding sites of genes obtained from Oct4 ChIP-seq. Oct4-bound chromatin was immunoprecipitated (IP) in parental A549 lung cancer cells using anti-Oct4 antibody. Normal IgG served as negative control. (B) qRT-PCR analysis of genes expression in A549 stable clones (vector, anti-Oct4#1 and anti-Oct4#2).

GAPDH gene served as internal control. (C) qRT-PCR analysis of genes expression in parental A549 cells after knockdown of Oct4 (Oct4#1 and

si-Oct4#2) or control (si-Ctrl). For all graphs, data are mean± SEM. (n = 3). P-values determined using two-tailed Student’s t-test. *P<0.05; **P<0.01; ***P_<0.001.

the known Oct4 consensus binding motif in ESCs (Motif 6, E-value: 1.4e-04) but also several novel binding motifs (Motifs 1–5) containing DNA binding sequence of known transcription factors such as Klf4, Sp1, Stat3 and ZNF219 (Figure1E). Klf4 has been shown experimentally to interact with Oct4 (19) while Sp1 and ZNF219 have been reported as candidate partners of Oct4 in ESCs using mass spectrome-try without further validation (30,32). These data supported our predicted results and suggested that combinatorial ac-tion of Oct4 with other co-regulators is also prevalent in lung cancer.

Differential interplay of Oct4 and Sp1 leads to downregula-tion of PTEN and upreguladownregula-tion of TNC

We identified that important tumor suppressor PTEN is a novel Oct4-repressed gene and PTEN signaling was a signif-icantly enriched pathway by bioinformatics analyses using different subsets of Oct4 bound genes (Figures1D and3C). Therefore, we investigated the mechanism by which Oct4 transcriptionally downregulated PTEN. PTEN promoter contained two Oct4 ChIP-seq binding regions (indicated by PTEN-1 and PTEN-2, Figure4A) belonged to ‘Motif 1’, which contained Sp1 transcription factor binding sites (Fig-ure1E, E-value: 2.1e-04). Indeed, we confirmed that Oct4

(16)

Figure 3. Involvement of Oct4-targeted promoter- and enhancer-associated genes in critical signaling pathways relevant to cancer progression. (A and B) Configurations of Oct4 binding (data from Oct4 ChIP-seq), DHS, H3K4me3 and H3K4me1 profiles (data from ENCODE datasets) at promoter-associated (A) and enhancer-associated (B) genes. Snapshots of the binding profiles from Integrative Genomics Viewer. (C and D) Diseases and disorders (upper) and canonical pathways analyses (lower) of promoter-associated (C) and enhancer-associated (D) genes targeted by Oct4 conducted by Ingenuity-IPA.

interacted with Sp1 in the nuclear fractions of A549 cells by IP-western assay (Supplementary Figure S5A). To gain in-sight into the effect of interplay between Oct4 and Sp1 on target genes expression, we included another novel Oct4-transactivated oncogene TNC, which contains a separate Sp1 site from the Oct4 consensus binding site (Motif 6) in its promoter region (Figure4B), serving as a comparison in the following transcriptional regulation assays.

The cell-based qChIP-PCR assays were performed using Oct4 and Sp1 antibodies in cells with Sp1 knockdown (Fig-ure 4C) or Oct4 knockdown (Figure4D). In the context of PTEN promoter, knockdown of Sp1 led to loss of Oct4 binding (Figure4C, PTEN-1 and PTEN-2 panels), whereas knockdown of Oct4 did not affect Sp1 binding (Figure4D, PTEN-1 and PTEN-2 panels), suggesting that Sp1 protein binding is essential for Oct4 binding at PTEN promoter. Re-ChIP results further supported the co-occupancy of Oct4 and Sp1 at PTEN promoter (Figure4E). Notably, it is re-ported that HDAC1/2-containing repressive complex par-ticipates in Oct4- or Sp1-mediated gene suppression (20,33). Indeed, we found that Oct4 facilitated formation of pro-tein complex composed of Sp1 and HDAC1/2 (Supplemen-tary Figure S5A), thereby recruiting HDAC1/2 to PTEN promoter as evident in the loss of HDAC1/2 binding in si-Oct4 cells (Supplementary Figure S5B and SC, PTEN-1 and PTEN-2 panels).

In contrast, at the TNC promoter, knockdown of Sp1 did not affect Oct4 binding (Figure4C, TNC panel), and

vice versa (Figure4D, TNC panel), suggesting that Oct4 could recognize and bind to its own consensus binding site independent of Sp1. Consistently, we did not find co-occupancy of Oct4 and Sp1 at TNC promoter by Re-ChIP assay (Figure4E). Moreover, Sp1, but not Oct4, recruited the HDAC1/2 complex to TNC promoter (Supplementary Figure S5B and SC, TNC panel), which was in agreement with positive regulation of TNC by Oct4. Similar results were observed by in vitro DNA affinity precipitation assay (DAPA) (Supplementary Figure S6A-D). Interestingly, in-tegration of our Oct4 ChIP-seq data with Sp1 ChIP-seq in A549 cells from ENCODE dataset further confirmed the co-occupancy of Oct4 and Sp1 at PTEN promoter, but not at TNC promoter (Supplementary Figure S7). Together, these results indicated that, in the context of PTEN pro-moter, Sp1 serves as a platform for Oct4 binding, then Oct4 recruits HDAC1/2 complex to the PTEN promoter. On the other hand, Oct4 binds to TNC promoter independent of Sp1 (Figure4F).

To investigate the differential effects of Oct4 and Sp1 on promoter activity of PTEN and TNC, dual luciferase ac-tivity assays were conducted. As shown in Figure4G, Oct4 suppressed the activity of PTEN promoters containing ei-ther one of WT Sp1 binding sites. However, Oct4 did not affect the activity of PTEN promoter with double mutant Sp1 binding sites (Figure4G). Consistently, the repressive effect of Oct4 relied on the HDAC1/2 complex as evident in restored PTEN expression and promoter activity upon knockdown of HDAC1/2 (Supplementary Figure S8). In addition, Oct4 significantly induced activity of TNC pro-moter in the presence of WT Oct4 binding site, but not the TNC promoter containing mutant Oct4 binding site (Figure4H). Collectively, these data further supported our model that Oct4 cooperates with Sp1 and HDAC1/2 com-plex to suppress PTEN expression while Oct4 can transac-tivate TNC expression independent of Sp1.

Oct4 transcriptionally represses PTEN, leading to AKT sig-naling activation and drug-resistance

Dysregulation of PTEN/AKT pathway has been impli-cated in gain of drug-resistance in cancers (34). Therefore, we speculated that Oct4 could modulate drug-resistance (Supplementary Figure S2C) through PTEN. We confirmed that overexpression of Oct4 decreased PTEN protein ex-pression and subsequently induced phosphorylation of AKT in A549 and CL1–0 lung cancer cell lines (Supplemen-tary Figure S9A), while knockdown of Oct4 by two inde-pendent siRNAs showed opposite effects on PTEN/AKT signaling (Figure5A). In contrast, expression of Oct4 did not affect phospho-AKT levels in CL1–5 cells (Figure5A; Supplementary Figure S9B), which are PTEN deleted cells (35), suggesting that Oct4 up-regulated AKT signaling via suppression of PTEN expression.

To address the involvement of PTEN/AKT signaling in Oct4-mediated drug-resistance, we performed cytotoxi-city and western blot assays in A549 cells stably express-ing Oct4 treated with PI3K/AKT inhibitor LY294002 in combination with cisplatin (Figure5B) or SAHA (Figure

(17)

Figure 4. Differential targeting and regulation of Oct4 and Sp1 on PTEN and TNC promoters. (A and B) Promoter regions of PTEN (A) and TNC (B) were analyzed. Oct4 ChIP-seq regions (blue; PTEN-1 and PTEN-2), Sp1 binding site (green) and Oct4 binding site (red) are shown. qChIP-PCR primers for amplification of PTEN-1, PTEN-2 and TNC promoters are indicated by arrows. (C and D) qChIP-PCR analysis of Oct4 and Sp1 occupancies at

PTEN and TNC promoters in A549 cells stably expressing Oct4 after knockdown of Sp1 Sp1) (C) or Oct4 Oct4) (D) compared to si-control

(si-Ctrl). (E) Re-ChIP assay was performed in control (vector) and A549 cells stably expressing Oct4 (Oct4) with overexpression of HA-tagged Sp1 (HA-Sp1). Sequential IP was performed using anti-HA antibody, followed by anti-Oct4 antibody. #, no detectable signal; ns, non-significant. (F) Working models of differential targeting of Oct4, Sp1 and HDAC1_{/2 complex at PTEN and TNC promoters. (G and H) Dual luciferase activity assays were performed} using PTEN promoter (G) or TNC promoter (H) containing WT (white circle), mutant Sp1 binding site (red circle) and/or mutant Oct4 binding site (red square) in A549 cells overexpressing vector control or Oct4. Data are mean± SEM. (n = 3). P-values determined using two-tailed Student’s t-test. *P<0.05; **P<0.01; ***P<0.001.

5C). The results showed that combined treatment signifi-cantly abolished drug resistance in stable Oct4 expressing cells (Oct4#1 and Oct4#2). Inhibition of AKT/GSK3␤ sig-naling by combinatorial treatment further triggered apop-tosis signaling compared to single drug treatment (Figure

5D and E, lane 5 versus lane 6). Importantly, re-enforced PTEN expression (Figure5F) or knockdown of AKT (Fig-ure5G) sensitized stable Oct4 expressing drug-resistant cells to SAHA treatment. Knockdown of Oct4 sensitized sta-ble Oct4 expressing drug-resistant cells to SAHA treatment, whereas simultaneous knockdown of PTEN rescued the cell viability of si-Oct4 cells upon SAHA treatment (Figure5H). To further characterize Oct4-mediated drug-resistance in

vivo, Balb/c nude mice bearing vector or stable Oct4

ex-pressing (Oct4#1) xenograft were established. Low dose

of LY294002 or SAHA single treatment significantly sup-pressed tumor growth of vector xenograft (Figure6A, left), but had no effects on Oct4#1 xenograft (Figure6A, right). Notably, anti-tumor activity was significantly increased by combinatorial therapeutics in Oct4#1 xenograft as well as vector xenograft (Figure 6A and B). In addition, im-munohistochemistry (IHC) analyses confirmed that Oct4 expression was inversely correlated with PTEN expression in xenograft tissues (Figure6C). Combined LY294002 and SAHA treatment effectively increased apoptosis in vivo, in-dicated by Caspase-3 activation (Figure6D). These results collectively suggested that Oct4 induces drug-resistance, in part, by triggering loss of PTEN and activation of AKT sig-naling.

(18)

Figure 5. Oct4 induces drug-resistance via dysregulation of PTEN/AKT signaling. (A) A549, CL1–0 and CL1–5 lung cancer cells were transiently trans-fected with control (si-Control) and Oct4 (si-Oct4#1 and si-Oct4#2) siRNAs for 48 h and then subjected to western blot analysis of PTEN, AKT and AKT phosphorylation levels.␤-actin was used as internal control. (B and C) Cell viability in vector control (vector) or A549 cells stably expressing Oct4 (Oct4#1 and Oct4#2) after treatment with solvent control, LY294002 (LY, 5 or 10␮M) and (B) cisplatin (5 ␮M), or (C) SAHA (5 ␮M) for 48 h. Cell viability was normalized to that of solvent control. (D and E) Western blot analysis of AKT signaling and apoptosis-related proteins in vector and Oct4#1 cells after treatment with LY294002 (LY) and (D) cisplatin (Cis), or (E) SAHA for 24 h. (F and G) A549 stable clone cells were transfected with PTEN expression vector (PTEN, F) or with AKT siRNA (si-AKT, G), followed by treatment with SAHA for 48 h. Western blot shows the efficiency of PTEN overexpression or AKT knockdown (upper). Cell viability was analyzed after drug treatment (lower). (H) A549 cells stably expressing Oct4 cells (Oct4#2) were transfected with Oct4 (si-Oct4#1) and/or PTEN siRNA (si-PTEN), followed by treatment with different concentration of SAHA for 48 h. Western blots show the efficiency of Oct4 and_{/or PTEN knockdown (upper). Cell viability was analyzed after drug treatment (lower). For all graphs, data are mean} ± SEM. (n = 3). P-values determined using two-tailed Student’s t-test. *P<0.05; **P<0.01; ***P<0.001.

(19)

Figure 6. Combinatorial treatment of LY294002 and SAHA overcomes Oct4-mediated drug resistance in vivo. (A) Nude mice bearing xenografts of A549 stable clone, vector (left) or Oct4#1 (right), were treated with DMSO, LY294002 (25 mg/kg) and/or SAHA (30 mg/kg). Tumor volume and image of xenografts are shown. (B) Tumor weight of vector and Oct4#1 xenografts. (C) IHC analyses of Oct4 and PTEN proteins. (D) IHC analysis of cleaved Caspase-3 proteins. For all IHC images, the insets are a higher magnification of the boxed areas. Original magnification× 200. For all graphs, data are mean± SEM. (n = 5 mice per group). P-values determined using two-tailed Student’s t-test. *P<0.05; **P<0.01; ***P<0.001.

TNC is required for Oct4-mediated cancer cell invasiveness

Previous studies showed that overexpression of Oct4 in-creased cell motility in cancer cells (12), however, little is known about the underlying mechanism. TNC has been reported to be involved in lung metastasis of breast can-cer (36). We found that Oct4 increased TNC protein level in multiple lung cancer cell lines (Supplementary Figure S9), this prompted us to investigate whether Oct4-induced

TNC is required for cancer invasion and metastasis. We

first performed in vitro transwell invasion/migration assay in vector control (vector) and stable Oct4 expressing cells (Oct4#1 and Oct4#2). Knockdown of TNC significantly abrogated Oct4-mediated cell invasion and migration (Fig-ure7A–C). Conversely, knockdown of Oct4 led to decrease in cell motility, which could be recovered by reconstitution of TNC expression in A549 cells (Figure 7D–F). The re-quirement of TNC expression for Oct4-induced cell inva-sion and migration was confirmed in CL1–5 cells (Supple-mentary Figure S10). Notably, we performed in vivo ex-travasation assay via tail vein injection model and verified that ectopically expressed TNC reestablished lung coloniza-tion of Oct4 knocked down A549 cells in immunodeficiency mice (Figure7G). These data suggested that Oct4 transac-tivates TNC, in part, leading to increased invasiveness and extravasation of lung cancer.

Lung cancer patients with high Oct4, low PTEN and high TNC expression profiles correlate with poor prognosis

We have demonstrated above that Oct4 transcriptionally re-pressed PTEN to induce drug resistance and activated TNC to promote tumor invasiveness. Since the relationship be-tween Oct4, PTEN and TNC protein expression has never been examined in human cancer patient, we investigated whether high Oct4 protein expression correlates with low PTEN and high TNC levels in lung cancer patients with poor prognosis or cancer metastasis. For this purpose, we examined the expression profiles of Oct4, PTEN and TNC proteins in lung tumor specimens from 133 lung cancer patients by IHC analyses (Figure8A; Supplementary Ta-ble S6). Importantly, an inverse correlation was found be-tween Oct4 and PTEN protein expressions in lung can-cer patients (P=0.022, Figure 8B; Supplementary Table S8), while patients with high Oct4 showed concordantly in-creased TNC (P<0.0001, Figure8C; Supplementary Table S8). In addition, 69.4% (43/62) patients with high Oct4 and low PTEN expression levels also demonstrated high TNC expression levels. Consistently, 72.7% (16/22) patients with normal Oct4 and PTEN expression levels demonstrated normal TNC expression (P=0.003, Supplementary Table S9). These data further supported our finding that Oct4 neg-atively regulates PTEN and positively regulates TNC ex-pression in lung cancer.

(20)

Figure 7. TNC expression is required for Oct4-mediated cancer invasiveness in vitro and in vivo. (A and B) Transwell invasion/migration assay was per-formed in A549 stable clones (Oct4#1 and Oct4#2) after TNC knockdown (si-TNC) or control siRNA (si-Ctrl). Representative image (A) and quantitation (B) are shown. Data are mean± SEM. (n = 3). (C) Western blot analysis of TNC and Oct4 protein levels in cells from (A and B). (D and E) Transwell invasion/migration assay was performed in si-Ctrl, TNC, si-Oct4, si-Oct4/TNC cells. Representative image (D) and quantitation (E) are shown. Data are mean± SEM. (n = 3). (F) Western blot analysis of TNC and Oct4 protein levels in cells from (D and E). (G) Representative lung images of mice intravenously injected via tail vein with si-Ctrl, TNC, si-Oct4, si-Oct4_{/TNC A549 cells (left). Tumor nodules were indicated by arrows. Number of tumor} nodules in lungs per mice is shown (right). Data are mean± SEM. (n = 8 mice per group). P-values determined using two-tailed Student’s t-test. *P<0.05; **P_{<0.01; ***P<0.001. Original magnification × 100 for all microscopic images.}

Of note, patients with high Oct4 expression (P=0.025, Pearson␹ 2 test) or high TNC expression (P=0.001, Pear-son␹ 2 test) were significantly associated with lymph node metastasis (Figure 8D and E; Supplementary Table S6). Kaplan–Meier analysis showed high expression of Oct4 (P=0.022), low expression of PTEN (P=0.045) or high ex-pression of TNC (P<0.0001) in lung cancer patients corre-lated with poor overall survival (Figure8F). Patients with high Oct4, low PTEN and high TNC levels showed worse progression-free survival compared to patients with nor-mal expression of the corresponding proteins (P=0.0002, Figure 8G). To further define the relative risk of death

of Oct4, TNC and PTEN expression, we performed the multivariate Cox-regression analysis and the data indicated that patients with Oct4 overexpression had poor outcome (P=0.052, hazard ratio [HR] = 1.83, 95% confidence in-terval= 1.00–3.35) after adjusting for all the clinicopatho-logical characteristics (Table1, left panel). Notably, the pa-tients with aberrant expression of all three proteins Oct4, PTEN and TNC showed an HR of 3.24 (P<0.001) (Table

1, right panel). Altogether, these clinical data demonstrated that high Oct4 protein coincides with low PTEN and high TNC levels in surgical tumor tissues, which could be prog-nostic biomarkers for lung cancer.

(21)

Figure 8. High Oct4 protein coincides with low PTEN and high TNC levels and poor survival of lung cancer patients. (A) Representative IHC images of Oct4, PTEN and TNC proteins in tumor specimen of lung cancer patients. Oct4 positive immunoreactivity (+), PTEN negative immunoreactivity (–) and TNC positive immunoreactivity (+) were found in patient 1, whereas patient 2 shows a reverse pattern. Original magnification× 200. The insets are a higher magnification of the boxed areas. (B and C) Concordance analysis between Oct4 and PTEN (B) or TNC (C) proteins expression (+, positive immunoreactivity; –, negative immunoreactivity) according to the four molecular subtypes. The percentage of the concordant group (left columns) and discordant group (right columns) is indicated above. (D and E) Histogram showing frequency of high Oct4 (D) or high TNC (E) protein expression in patients with different lymph node metastasis status. (F) Overall survival curves of Kaplan–Meier method indicated that patients with high Oct4, low PTEN or high TNC expression had significantly poorer survival than patients with normal expression of the corresponding protein. (G) Progression-free survival analyses indicated that patients with high Oct4 combined with both low PTEN and high TNC expression had significantly poorer survival than other patients. P-values for correlation analyses (B–E) were determined using Pearson␹ 2 test; and for survival analyses (F and G) using log-rank test.

DISCUSSION

Here, we provide compelling evidence from lung cells, an-imal and clinical studies that Oct4-driven transcriptional program promotes lung cancer progression. Through Oct4 ChIP-seq and cross-validation of ENCODE datasets, we identify important Oct4 target genes involved in tumorige-nesis and critical signaling pathways, especially PTEN sig-naling. We further provide the mechanistic insight into the roles of Oct4 in modulating downstream target genes. Oct4 suppresses PTEN and activates TNC expression via differ-ential interplay with Sp1 transcription factor, resulting in drug-resistance and cancer metastasis, respectively. Clini-cally, lung cancer patients with Oct4 high, PTEN low and TNC high expression profile correlate with cancer progres-sion and poor disease-free survival (see schematic in Sup-plementary Figure S6E and SF).

Recent study addressed the upstream signaling (i.e. IGF signaling) that initiates stem-like features in lung cancer for which Oct4 is an indispensable effector (13,28), but the rel-evant functions of Oct4 are not defined. In this study, we present the first genome-wide Oct4 binding profile in so-matic cancer using ChIP-seq approach and strengthen the critical role of Oct4 in lung cancer progression. Interest-ingly, we found high Oct4 occupancy at gene body and, to a lesser extent, distal region in lung cancer cells (Figure1A). Such distribution pattern was different from that in human ESCs (37), in which Oct4 mainly occupied distal region. In addition, only 75 common genes out of 3350 genes in our ChIP-seq have been identified in human ESCs dataset (17) (Supplementary Table S7), suggesting cell-type speci-ficity of Oct4 binding profile. Although such discrepancy could be explained, in part, by the different platforms and analysis tools employed in our and ESC studies, our GSEA analysis showed that Oct4-target gene set correlated with

(22)

Table 1. Multivariate Cox regression analysis of risk factors for cancer-related death in lung cancer patients

Characteristics Multivariate analysis Characteristics Multivariate analysis

HRa_{(95% CI}b₎ _P-valuec _HRa_{(95% CI}b₎ _P-valuec

Oct4 protein expression

Oct4/PTEN/TNC protein expression

Normal expression 1.00 Others 1.00

Overexpression 1.83 (1.00–3.35) 0.052 Aberrant expression 3.24 (1.77–5.94) <0.001

Age Age <60 1.00 _<60 1.00 ≥60 1.44 (0.75–2.75) 0.273 ≥60 1.14 (0.59–2.22) 0.700 Gender Gender Female 1.00 Female 1.00 Male 1.20 (0.48–3.01) 0.693 Male 1.12 (0.44–2.87) 0.816

Smoking habit Smoking habit

Non-smoker 1.00 Non-smoker 1.00

Smoker 1.79 (0.80–4.00) 0.154 Smoker 1.67 (0.72–3.84) 0.230

Tumor typed Tumor typed

SCC 1.00 SCC 1.00

ADC 2.83 (1.34–5.98) 0.006 ADC 1.75 (0.81–3.79) 0.156

Stage Stage

Stage I-II 1.00 Stage I-II 1.00

Stage III-IV 1.29 (0.55–3.03) 0.561 Stage III-IV 1.95 (0.80–4.77) 0.144

T stage T stage Stage 1–2 1.00 Stage 1–2 1.00 Stage 3–4 2.99 (1.49–5.98) 0.002 Stage 3–4 2.30 (1.15–4.58) 0.018 N stage N stage N0 1.00 N0 1.00 ≥N1 3.11 (1.31–7.40) 0.010 ≥N1 2.17 (0.88–5.37) 0.093 M stage M stage M0 1.00 M0 1.00 ≥M1 0.75 (0.26–2.16) 0.591 ≥M1 0.46 (0.15–1.36) 0.159 a_{HR, Hazard ratio.} b_{CI, Confidence interval.}

c_{Bold values indicatestatisticalsignificance (P}_<0.05). d_{ADC, Adenocarcinoma; SCC, Squamous cell carcinoma.}

previously published gene signature of lung TICs (Supple-mentary Figure S4B), supporting that our Oct4 target genes were specific to lung cancer cells. In addition, we identi-fied that many cancer-related genes were previously unde-fined direct targets of Oct4 in lung cancer (Figure2). The validation rate of Oct4 occupancy at cancer-related genes was 100% (20/20) by qChIP-PCR assay, while the valida-tion rate was 58% (29/50) by qRT-PCR assay demonstrat-ing the differential regulation between oncogene and tu-mor suppressor-like genes by Oct4 (Figure2). For example, oncogenic HDAC4 is able to activate Stat1 and promote drug-resistance in ovarian cancer (38), and is involved in bone-metastasis of lung cancer (39). Enhanced translation of LABM1 is required for epithelial to mesenchymal tran-sition in hepatocellular carcinoma (40). In terms of Oct4-suppressed genes, DKK3, a well-known Wnt antagonist, ex-hibits low expression level in many cancer types and serves as a poor prognostic marker (41). RAB37 is a novel metas-tasis suppressor that regulates exocytosis machinery in lung cancer (42,43). In addition, we confirmed the roles of Oct4 in transcriptional regulation of the tumor suppressor gene

PTEN and oncogene TNC that lead to drug-resistance and

metastasis. Together, we propose that these cancer-related genes comprise the Oct4-centered transcriptional network that collectively drives lung cancer progression.

PTEN is a bona fide tumor suppressor gene, which acts as

negative regulator of PI3K/AKT signaling pathway (34). However, somatic PTEN mutations occur at a low

fre-quency in lung cancer compared to other cancer types (34). Here, we show a new dysregulation mechanism of PTEN in cancer by discovering that Oct4 transcriptionally sup-pressed PTEN via cooperation with Sp1 and HDAC1/2 in lung cancer cells (Figure4; Supplementary Figures S5– S8). Previous study showed that mice with lung-specific ho-mozygous deletion of Pten in alveolar type II cells develop lung adenocarcinoma (44), suggesting Pten is critical for prevention of onset of lung tumorigenesis. However, there was no evidence of the mechanism leading to PTEN loss in lung TICs. Our finding and previous study indicate that Oct4-mediated transcriptional repression of PTEN may ac-count for expansion of lung TICs and gain of drug-resistant capacity. Whether lung cancer patients with both high Oct4 and low PTEN expressions show poor response to current chemotherapy needs further interrogation.

TNC is a secreted extracellular matrix molecule and is highly expressed in the microenvironment of several solid tumors, including breast and lung cancers (45). TNC se-creted by breast cancer cells acts as a metastatic niche com-ponent for colonization to lungs (36). TNC downregulates

Dickkopf-1 expression by blocking actin stress fiber

forma-tion, thereby activating Wnt signaling in neuroendocrine tumor cells (46). However, little is known about the mech-anisms or stimulations leading to TNC overexpression in cancers. In this study, we provide evidences that Oct4 tran-scriptionally activated TNC expression (Figure4), which was required for cancer invasion and metastasis (Figures