2011;71:4653-4663. Published OnlineFirst April 25, 2011.
Cancer Res
I-Wen Teng, Pei-Chi Hou, Kuan-Der Lee, et al.
Cancer Stem/Initiating Cells
Sufficient to Transform Mesenchymal Stem Cells into Targeted Methylation of Two Tumor Suppressor Genes Is
Updated Version
10.1158/0008-5472.CAN-10-3418 doi:
Access the most recent version of this article at:
Material Supplementary
html
http://cancerres.aacrjournals.org/content/suppl/2011/04/25/0008-5472.CAN-10-3418.DC1.
Access the most recent supplemental material at:
Cited Articles
http://cancerres.aacrjournals.org/content/71/13/4653.full.html#ref-list-1 This article cites 44 articles, 12 of which you can access for free at:
E-mail alerts Sign up to receive free email-alerts related to this article or journal.
Subscriptions Reprints and
. [email protected] Publications Department at
To order reprints of this article or to subscribe to the journal, contact the AACR
Permissions
. [email protected] Department at
To request permission to re-use all or part of this article, contact the AACR Publications
Targeted Methylation of Two Tumor Suppressor Genes Is Sufficient to Transform Mesenchymal Stem Cells into Cancer Stem/Initiating Cells
I-Wen Teng
1, Pei-Chi Hou
1, Kuan-Der Lee
2, Pei-Yi Chu
3, Kun-Tu Yeh
3, Victor X. Jin
4, Min-Jen Tseng
1, Shaw-Jenq Tsai
5, Yu-Sun Chang
6, Chi-Sheng Wu
6, H. Sunny Sun
7, Kuen-daw Tsai
1,8, Long-Bin Jeng
8, Kenneth P. Nephew
9, Tim H.-M. Huang
10, Shu-Huei Hsiao
1, and Yu-Wei Leu
1Abstract
Although DNA hypermethylation within promoter CpG islands is highly correlated with tumorigenesis, it has not been established whether DNA hypermethylation within a specific tumor suppressor gene (TSG) is sufficient to fully transform a somatic stem cell. In this study, we addressed this question using a novel targeted DNA methylation technique to methylate the promoters of HIC1 and RassF1A, two well-established TSGs, along with a two-component reporter system to visualize successful targeting of human bone marrow –derived mesenchymal stem cells (MSC) as a model cell system. MSCs harboring targeted promoter methylations of HIC1/RassF1A displayed several features of cancer stem/initiating cells including loss of anchorage dependence, increased colony formation capability, drug resistance, and pluripotency. Notably, inoculation of immunodeficient mice with low numbers of targeted MSC resulted in tumor formation, and subsequent serial xenotransplantation and immunohistochemistry confirmed the presence of stem cell markers and MSC lineage in tumor xenografts.
Consistent with the expected mechanism of TSG hypermethylation, treatment of the targeted MSC with a DNA methyltransferase inhibitor reversed their tumorigenic phenotype. To our knowledge, this is the first direct demonstration that aberrant TSG hypermethylation is sufficient to transform a somatic stem cell into a fully malignant cell with cancer stem/initiating properties. Cancer Res; 71(13); 4653–63. 2011 AACR.
Introduction
DNA methylation, a tightly regulated process during normal development, frequently becomes dysregulated during disease development including cancer (1 –3). Although methylation- induced tumorigenesis has yet to be recapitulated experimen-
tally, during somatic cell proliferation, environmental and extracellular signals can initiate changes in DNA methylation that contribute to clonal selection, altered cellular behavior, and ultimately tumorigenesis (4 –6). Hypomethylation and/or hypermethylation of specific loci, including tumor suppressor loci, were strongly associated with transformation and carci- nogenesis (7, 8), and genetic knockout of the DNA methyl- transferases (DNMT), resulted in global hypomethylation and tumorigenesis (9, 10). Although a causative role for altered methylation at specific loci, particularly as an initiating neo- plastic event, remains poorly understood, dormant stem cells, either pre-existing in tissues or arising from somatic cells, may play a role in cancer origin and prognosis (4). Furthermore, because DNA hypermethylation of tumor suppressor gene (TSG) has been documented in many cancers, and its spread- ing correlates with cancer progression (4), we hypothesized that abnormal DNA hypermethylation can disrupt somatic stem cell proliferation and differentiation, resulting in the development of neoplasia.
To directly examine the effect of aberrant DNA methylation on cellular physiology, we established a Targeted DNA Methy- lation method called "TDM" (11, 12), in which transfection established an in vitro methylated DNA complementary to the target gene promoter region initiated recruitment of DNMT to the endogenous target loci. Consequently, DNMT-mediated methylation spreads within the promoter region of the target loci, ultimately silencing the target gene after cellular passages.
Authors' Affiliations:
1Human Epigenomics Center, Department of Life Science, Institute of Molecular Biology and Institute of Biomedical Science, National Chung Cheng University;
2Chang Gung Memorial Hospital, Chia- Yi;
3Department of Pathology, Changhua Christian Hospital, Changhua, Taiwan;
4Department of Biomedical Informatics, The Ohio State University Columbus, Ohio;
5Department of Physiology,
6Institute of Molecular Med- icine, College of Medicine, National Cheng Kung University, Tainan;
7Basic Medical Sciences, Chang Gung University, Tao-Yuan;
8Department of Internal Medicine, China Medical University Beigang Hospital, Yunlin, Taiwan;
9Medical Sciences and Department of Cellular and Integrative Physiology, Indiana University Simon Cancer Center, School of Medicine, Bloomington, Indiana; and
10Division of Human Cancer Genetics, Depart- ment of Molecular Virology, Immunology, and Medical Genetics, and the Comprehensive Cancer Center, The Ohio State University, Columbus, Ohio Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
I-W. Teng and P.-C. Hou contributed equally to this work.
Corresponding Author: Shu-Huei Hsiao, National Chung Cheng Univer- sity, 168 University Road, Min Hsiung, Chia-Yi, 621 Taiwan. Phone: 886-5- 2720411 ext 53202; Fax: 886-5-2722871; E-mail: [email protected] or Yu-Wei Leu, ext 66507; E-mail: [email protected]
doi: 10.1158/0008-5472.CAN-10-3418
2011 American Association for Cancer Research.
To monitor progression of TDM and cellular transformation, we also developed a 2-component system (11, 12). The first component consisted of regulation of tetracycline repressor (Tet) expression by the cloned target promoter sequence. A CMV promoter driving expression of a reporter (enhanced green fluorescence protein or EGFP) comprised the second component. A Tet repressor binding site, Tet operator, placed between the CMV and the EGFP, regulated EGFP expression. In the absence of DNA methylation, Tet expression was observed, and the expression of EGFP was silenced. Induction of DNA methylation silenced the Tet repressor and activated EGFP expression, allowing us to observe progression of DNA methy- lation in a living cell. Furthermore, if target loci DNA methyla- tion was sufficient to induce cellular transformation, an EGFP- expressing cell behaved like a tumor (11, 12). It was of interest to use this system to initiate TDM in a somatic stem cell and monitor the subsequent effects on cellular transformation.
Although spontaneous transformation of human mesenchymal stem cells (MSC) in vitro was recently described (13, 14), genetic disruptions of the p53 pathway, but not retinoblastoma (Rb), was sufficient to transform a fat-derived MSC (15), supporting the possibility that sarcoma could be initiated from MSC.
In the current study (work flow is illustrated in Supple- mentary Fig. S1A), we aimed to test the hypothesis that targeted DNA methylation is sufficient for cellular transfor- mation, thus the promoter regions of HIC1 (hypermethylated in cancer 1) and RassF1A (ras-associated family protein iso- forms 1A), 2 TSGs reported to be frequently silenced by DNA methylation in cancer (16–18), were cloned, methylated in vitro, and then transfected into the MSCs, individually or concurrently. HIC1 and RassF1A are involved in highly diverse, interacting cellular networks (16, 19, 20), and their loss of function could result in the recently described phenomenon of oncogenic addiction through p53 pathway (5). We thus hypothesized that hypermethylation of HIC1 and RassF1A would not only to directly suppress their tumor suppressor function but also disrupt multiple cellular networks resulting in tumorigenesis and cancer progression.
Materials and Methods
MSC isolation and characterization
Human MSC isolation and culture were carried out as described by Lee and colleagues (21). MSC expansion medium, passages, and culture condition were as described by Hsiao and colleagues (11).
In vitro DNA methylation
Four micrograms of PCR-amplified HIC1 and RassF1A promoters were incubated with 20 units of CpG methyltrans- ferase (New England BioLabs) at 37
C for 4 hours in the presence of 160 mmol/L S-adenosylmethionine to induce methylation.
Validation of in vitro DNA methylation
Methylated DNA showing resistance to methylation-sensi- tive restriction enzymes (BstUI) indicated completed conver- sion (Supplementary Fig. S2).
Cy5 labelling of the HIC1 promoter fragment
HIC1 DNA was labeled with LabelIT tracker Reagents (Mirus) according to the manufacturer's instruction.
Transfection
The methylated PCR products (0.4 mg) were denatured at 95
C and then transfected into 5 10
5cells using DMRIE-C (Invitrogen), according to the manufacturer's instructions.
Unmethylated PCR products were transfected as control. Cells were transfected three times at days 1, 3, and 5.
Semiquantitative real-time methylation-specific PCR The semiquantitative methylation-specific PCR (qMSP) experiment was conducted and products were quantified according to the protocol described in Yan and colleagues (22). Briefly, bisulfite-converted genomic DNAs (0.5 mg) were subject to real-time PCR (RT-PCR) with methylation-specific primers (Supplementary Table S2). The qMSP reactions were carried out using the SYBR Green I PCR Kit (Toyobo) in an iQ5 Real-Time PCR instrument (Bio-Rad). Melting analysis was conducted followed by all of the PCR reactions to ensure a specific amplicon was generated. Col2A1 was used for stan- dard curve construction and as loading control. Serial dilution of Col2A1 amplified bisulfite-converted DNA was used to generate standard curve. Methylation percentage was calcu- lated as follows: (means of target gene)/(means of Col2A1);
fold change was calculated as follows: (TDM methylation percentage)/(mock methylation percentage). Endogenous and exogenous HIC1 promoters were discerned by the reverse primers indicated in Supplementary Figures S1B and S3A.
Differential methylation hybridization
All procedures for the differential methylation hybridization (DMH) microarray were conducted as described in Leu and colleagues (18) using a human CpG island microarray (Agi- lent). Briefly, me_H&R-treated and control MSC genomic DNAs were digested into small fragments and then ligated with designed adaptors. Methylation-sensitive restriction enzymes (BstUI and HpaII) were used to discriminate the methylated and the unmethylated DNA fragments. Differences in methylation status were then amplified by PCR using the adaptor as primer. The amplicons from control and me_H&R- transfected cells were labeled with Cy3 and Cy5, respectively and then cohybridized onto the CpG microarray. After mea- suring the Cy3 and Cy5 intensity, the M value [M ¼ log
2(Cy5 intensity/Cy3 intensity)] was used to indicate the difference in DNA methylation between 2 sources, and the L value [L ¼ 0.5
log
2(Cy5 Cy3)] was used to indicate the intensity of individual loci. Both values were adjusted and normalized by LOWESS. A cutoff value of 4 based on the M value was used to identify the target loci.
Immunostaining
Cells were fixed in 2% formaldehyde/PBS, then permeabi- lized with 0.5% NP40/PBS. After blocked with horse serum/
PBS (1:100), the slides were incubated with primary antibody
in 3% bovine serum albumin (BSA)/PBS followed by 3 times
of PBS washes. The cells were incubated with secondary
antibodies conjugated with Fluorescein or Texas Red (Vector Lab) in 3% BSA/PBS. After several PBS washes, the slides were mounted in mounting medium with 4
0,6-diamidino-2-pheny- lindole (DAPI; Vector Lab). The primary antibodies used were as follows: anti-HIC1 (Millipore), anti-RassF1A (Bioscience), anti-CD133 (Abcam), anti-Oct4 (Cell Signaling), and anti-Neu- ronal nuclei (NeuN; Chemicon).
5-Aza-dc-2
0-deoxycytidine treatment
Control and me_H&R transfected MSCs were treated with 20 mmol/L (Fig. 1) or 5 mmol/L (Supplementary Fig. S6D) of 5- aza-dc-2
0-deoxycytidine (5-aza-dc) or an equal final volume of dimethyl sulfoxide (DMSO) for 5 consecutive days.
Cloning of the human HIC1 and RassF1A promoters Primer sequences for human HIC1 and RassF1A promoters are listed in Supplementary Table S1. Genomic DNA purified from human MSCs served as a template for PCR. Purified PCR products were ligated into the pyT&A cloning vector (Yeastern Biotech) according to the manufacturer's protocol. Inserts were confirmed by restriction digests and sequencing. Cloning
and TDM for the Salvador–Warts–Hippo (SWH) signaling pathway components were conducted using the same proto- cols and the primers were listed in Supplementary Table S1.
Semiquantitative RT-PCR
Total RNA isolation, first-strand cDNA synthesis, and detec- tion of the transcripts were carried out as described (18).
Briefly, total RNA (2 mg) was reverse transcribed using the SuperScript III reverse transcriptase (Invitrogen). The semi- quantitative RT-PCR (qRT-PCR) was then carried out by SYBR Green I PCR Kit (Toyobo) in an iQ5 Real-Time PCR instrument (Bio-Rad). A serial dilution of glyceraldehyde-3-phosphate dehydrogenase (GAPDH)-amplified cDNA was used as control to generate standard curve and GAPDH from each samples was used as loading control. The primers are listed in Sup- plementary Table S2.
Cell survival assay
In 96-well plates, 20 mL of MTT solution (Sigma-Aldrich; 5 mg/mL) was added to each well containing different number of cells and incubated at 37
C for 5 hours. The reaction was
A
3
2
1
MSC me_H&R
me_H&R 0 HIC1
HIC1
RassF1A
RassF1A
Meth ylation (f old) Meth ylated
Meth ylation (f old) 6
4
2
0 MSC HIC1
DAPI
5 4 3 2 1 0
3
2
1
– + 0 – +
5-Aza-dc 5-Aza-dc
DAPI DAPI DAPI
HIC1 RassF1A RassF1A me_H&R
me_H&R
me_H&R
MSC
MSC
Ctrl 20 µM 5-Aza-dc Attached
Low attachment
Ctr l HIC1 RassF1A H&R
MSC
C D
B
Figure 1. Concurrent HIC1 and RassF1A methylation in transformed MSCs. A, HIC1 and RassF1A TDM. Denatured methylated or unmethylated (mock control) HIC1 and/or RassF1A promoter DNAs were transfected individually or together into human MSCs. The promoter methylation of endogenous HIC1 (top left) and endogenous RassF1A (top right) were detected by qMSP and protein expression was detected by immunostaining (bottom).
B, transformation of MSCs by concurrent HIC1 and RassF1A methylation. MSCs transfected with unmethylated DNA only (control), me_HIC1, me_RassF1A,
or me_H&R were cultured in attachment (left) or low attachment dishes (right). Spheroid formation was observed for all 4 treatments on low attachment
dishes; however, only the me_H&R MSCs showed loss of contact inhibition. C, 5-aza-dc treatment reverses me_H&R-induced endogenous HIC1 and
RassF1A hypermethylation as measured by qMSP. D, 5-aza-dc treatment represses me_H&R-induced MSC aggregates. Loss of DNA methylation after
5-aza-dc treatment was correlated with reversion of the loss of contact inhibition phenotype in me_H&R treated MSCs (right bottom).
terminated by adding 200 mL of DMSO, and absorbance was measured at 595 nm.
Soft agar assay
Cells were plated at a density of 5 10
4per well in soft agar.
After 2 weeks of culture, cells were stained with 0.01% crystal violet, and the number of spheres ( >50 cells) from each dish was counted.
Transwell study
Cells were plated at a density of 5 10
4per well into the hanging cell culture insert (Millipore), and the redistribution of the cells on the other side of the insert was observed, stained, and quantified.
In vivo tumorigenesis and serial transplantation of xenografts
Six-week-old nude mice (Narl: ICR-Foxn1
nu) were inocu- lated subcutaneously with 1 10
4me_H&R-transfected or control MSCs. Growth of tumours was monitored until they reached 0.8 cm in diameter. Then, tumors were surgically removed and subcultured in MSC medium on low attachment plates until spheres were observed again. The same number of subcultured cells was inoculated into a new nude mouse, and the entire procedure (n ¼ 9) was repeated 4 more times (n ¼ 36 in total).
Immunohistochemistry
Tumor masses surgically removed from nude mice inocu- lated with me_H&R-transfected MSCs were paraffin embedded and sectioned at 4 mm or embedded in optimum cutting temperature (OCT) and sectioned on a cryostat (Leica) at 12 mm. Sections were stained with the indicated antibodies, and detection was carried out with Vectastain (Vector Lab) for the paraffin sections and Fluorescein- or Texas red–conju- gated anti-mouse or rabbit IgG (Vector Lab) for the cryosec- tions, followed by DAPI staining. Sections were also stained with hematoxylin and eosin (H&E; Vector Lab) for pathologic exams.
Lineage-specific induction of MSCs
Transfected MSCs (mock or me_H&R) were plated onto 6- well plates at 5 10
4cells per well. After attachment, the medium was replaced with neuronal preinduction medium [Dulbecco's modified Eagle's media (DMEM) with 20% FBS, 10 ng/mL basic fibroblast growth factor (bFGF), and 1 mmol/L b-mercaptoethanol] for 24 hours, followed by neuronal induc- tion medium [DMEM with 100 mmol/L butylated hydroxya- nisole (BHA), 10 mmol/L forskolin, 2% DMSO, 25 mmol/L KCl, 2 mmol/L valproic acid, 1 B27 supplement, 10 ng/mL bFGF, 10 ng/mL platelet-derived growth factor (PDGF)]. Morpholo- gic changes and NeuN expression were used to validate neuronal induction. Osteocyte induction medium consisted of DMEM, 10% FBS, 10 mg/mL penicillin/streptomycin, 100 nmol/L dexamethasone, 10 mmol/L b-glycerophosphate, and 50 mmol/L
L-ascorbic acid-2-phosphate. Cells were treated with the osteocyte induction medium for 10 days and then subject to alkaline phosphatase (Sigma-Aldrich) staining.
Construction of two-component reporter system The construction of HIC1 2-component reporter system is described and illustrated in Supplementary Figure S4. Both constructs were sequence validated and used to transfect the MSCs and transfected cells were selected with hygromycin and G418 resistance. PCR were carried out to validate the integrations of both constructs (HIC1-TR and EG1; Supple- mentary Fig. S4B). MSC clones carrying both reporter con- structs were treated with doxycycline (Supplementary Fig. S4B). Doxycycline-induced EGFP expression indicates the reporter system is functional.
Human subjects
Isolation and characterization of human MSCs were con- ducted under Institutional Review Board (IRB) regulations of the Chang Gung Memorial Hospital.
Animals
The use of mice followed the regulations and protocols reviewed and approved by the Institutional Animal Care and Use Committee at the National Chung Cheng University.
Results
Targeted HIC1 and RassF1A methylation transforms MSCs
In vitro methylated (validation of in vitro methylation is shown in Supplementary Fig. S2) or unmethylated (control) HIC1 and RassF1A promoter DNA fragments were transfected into human bone marrow –derived MSCs mixed population (MSC_MP) alone or in combination. Transfection of methy- lated HIC1 (me_HIC1) or methylated RassF1A (me_RassF1A) increased endogenous HIC1 or RassF1A promoter methyla- tion, as detected by qMSP (Fig. 1A, top). MSP amplification was confirmed by sequencing in Supplementary Fig. S3A and B) and the expression of endogenous HIC1 or RassF1A decreased accordingly, as detected by immunostaining (Fig. 1A, bottom). TDM was confirmed by bisulfite sequencing (Supplementary Fig. S3C and D).
Loss of contact inhibition was used as a screening criterion for MSC transformation (Fig. 1B, left). Transfection and selec- tion of transformed MSCs are illustrated in Supplementary Figure S6A. Cotransfection of methylated HIC1 and RassF1A (me_H&R) caused formation of anchorage-independent aggre- gates, similar to the transformed phenotype in attached cultures, whereas the controls and cells transfected with either me_HIC1 or me_RassF1A alone remained contact inhib- ited (Fig. 1B, left middle two). When cultured in low attach- ment dishes, both methylated- and mock-transfected MSCs formed spherical aggregates (Fig. 1B, right), suggesting that these me_H&R transfected MSCs retained their self-renewal property.
To confirm that loss of anchorage dependence was due to
DNA hypermethylation, cells were treated with the DNMT
inhibitor 5-aza-dc. 5-Aza-dc treatment decreased the hyper-
methylation level of endogenous HIC1 and RassF1A in the
me_H&R-transfected MSCs (Fig. 1C and Supplementary
Fig. S6D) and abrogated the formation of me_H&R-induced
aggregates (Fig. 1D, right and Supplementary Fig. S6D, top). These data indicate that DNA hypermethylation of both HIC1 and RassF1A potentially transformed somatic MSCs and rendered them anchorage independent. More-
over, the requirement for methylation of both HIC1 and RassF1A for MSC transformation supports the notion that cancer initiation and development is a multistep process.
Figure 2. Visualization of targeted DNA methylation and MSC transformation. A, schematic diagram of the reporter system (left; construction and validation of the obtained clones in
Supplementary Fig. S4). Targeted HIC1 methylation was visualized by EGFP fluorescence (right middle) including RassF1A TDM led to EGFP-expressing cell aggregation (right). B, tracking TDM. Unmethylated (control) and/
or methylated HIC1 and RassF1A DNA fragments were transfected into the MSCs harboring both constructs shown in A. me_HIC1 was labeled with Cy5 to track the distribution of transfected DNA in the MSCs, after induction of EGFP expression. Only the Cy5- containing cells expressed EGFP, confirming the specificity of the targeted methylation. C, detecting the methylation state of the exogenous HIC1 promoter (HIC1- TR) and the expression of EGFP.
qMSP was carried out to quantify exogenous HIC1 promoter region methylation using HIC1-specific and vector-specific primers (left).
Increased HIC1-TR methylation was observed only in the me_HIC1-targeted MSCs. EGFP expression increased accordingly, as detected by qRT-PCR (right). D, validation of HIC1 and/or RassF1A TDM. MSCs were treated with methylated HIC1 and/or RassF1A as C. qMSP was used to detect methylation changes at endogenous HIC1 or RassF1A promoter (left). qRT-PCR was used to determine changes in expression of endogenous HIC1 or RassF1A (right).
A
neo
neo hygro
hygro
HIC1 promoter
HIC1 promoter Tet O
2Tet O
2Tet O
2Tet O
2EGFP reporter
EGFP reporter TATA
48 h EGFP
50 40 30 20 10 0
Meth ylation (f old) Meth ylation (f old) Expression (f old) Meth ylation (f old)
200,000 150,000 100,000 50,000 0
1.6
2.0 1.5 1.0 0.5 0 1.2 0.8 0.4 0 1 × 10 6
5 × 10 5 1 × 10 5 5 × 10 0 4 30 20 10 0
EGFP –
– – +
– – – +
– – –
+ –
– – +
– – + + – – +
+ – – + + –
– – –
– – – +
– – – +
– – – + –
– –
+ –
– + + – – + +
– – +
+ –
– – – RassF1A
RassF1A me_RassF1A
me_RassF1A HIC1-TR
HIC1-TR
HIC1-TR
RassF1A RassF1A
HIC1 HIC1
Cy5-me_HIC1-TR
me_HIC1-TR
– – – +
– – – +
– – – + –
– –
+ –
– + + – – + +
– – +
+ –
– – – RassF1A me_RassF1A HIC1-TR me_HIC1-TR
– – – +
– – – +
– – – + –
– –
+ –
– + + – – + +
– – +
+ –
– – – RassF1A me_RassF1A HIC1-TR me_HIC1-TR
– – – +
– – – +
– – – + –
– –
+ –
– + + – – + +
– – +
+ –
– – – RassF1A me_RassF1A HIC1-TR me_HIC1-TR
Cy5 Vis .
Unmethylated HIC1-TR
Methylated HIC1-TR
Methylated HIC1 &
RassF1A & HIC1-TR
EGFP Vis .
Tet repressor
Tet repressor CMV
CMV
B
C
D
Validation of TDM by two-component reporter system To confirm that the TDM caused silencing at the transcrip- tional level and induced aggregates within the targeted cells, we used a 2-component reporter system (11, 12) to mark and track methylation-mediated silencing of the HIC1 promoter in live cells (single colonies, SC). In this system, the HIC1 promoter regulates expression of the EGFP reporter construct (Fig. 2A flow diagram on left; reporter system construction is illustrated in Supplementary Fig. S4). Methylation of the HIC1 promoter alone did not result in phenotypic changes but caused HIC1 promoter silencing, as indicated by increased EGFP expression (Fig. 2A, right middle). However, concomi- tant methylation of RassF1A and HIC1 led to cell aggregates containing EGFP-expressing (i.e., HIC1 silenced) cells (Fig. 2A, right and Supplementary Fig. S6B and C).
To track the distribution of transfected DNAs, me_HIC1 construct was labeled with Cy5 prior to transfection. Cy5 signals localized mainly in the nuclear region (Fig. 2B and Supplementary Fig. S5), indicating that the me_HIC1 entered
the MSC nuclei. EGFP was detected in these Cy5-containing/
transfected MSC (Fig. 2B, columns 3, 7, 8). When me_HIC1 was used to transfect the cells, qMSP analysis further showed increased methylation of exogenous HIC1 promoter (HIC1- TR; Fig. 2C, left), and EGFP expression increased accordingly (Fig. 2C, right). The me_RassF1A, RassF1A, or control (unmethylated) HIC1 DNAs failed to induce the EGFP expres- sion, further indicating that we successfully and specifically methylated the HIC1 promoter. Targeted HIC1 methylation had no effect on the methylation state of the endogenous RassF1A promoter and vice versa (Fig. 2D), consistent with our previous report that the TDM is locus-specific (11, 12). Com- bined transfection of me_HIC1 and me_RassF1A resulted in methylation of both endogenous loci, as determined by qMSP (Fig. 2D, left), as well as concomitant silencing of both genes (qRT-PCR; Fig. 2D, right). Data from the 2-component reporter system further confirmed that our targeting method caused gene silencing at the transcriptional level and that aberrant DNA methylation of both HIC1 and RassF1A might transform
SNX26
LIN7C MPP5
ARID3A PPAP2B DDX6 PRIC285
ERCC5 MLYOD
TP53 PP2A UPF1 PSRC1 CK1
TBX21 FABP5
XTP3TPA EIF5
PHF17 PDCD5 SSSCA1
HNRNPL ATE1 VHL INPP4A LHX1
TFAP2A NFIB SLC2A4
EXO1 TMEPAI
AMPK Nuclear
factor 1
C8ORF32 Smad
A
B
C
TNFRSF19L SNX26
LIN7C MPP5
ARID3A PPAP2B DDX6 PRIC285
ERCC5 MLYOD
TP53 PP2A UPF1
PSRC1 CK1
TBX21 FABP5
XTP3TPA EIF5
PHF17 PDCD5 SSSCA1
HNRNPL ATE1 VHL INPP4A LHX1
TFAP2A NFIB SLC2A4
EXO1 TMEPAI
AMPK Nuclear
factor 1
C8ORF32 Smad IGF2BP2
Slide 1_Cy5 Slide 2_Cy5 Slide 1_Cy3 Slide 2_Cy3
HIC1
MGC11257
SSBP4
SOX7
ADAM9
GATA6
OSBP
TBX2
20,000 10,000 5,000 0.15 0.10 0.05 0
2.5 2.0 1.5 1.0 0.5 0
Meth ylation (f old)
40 30
AP staining D API CD133 D API Oct4 Ctr l Pre-NIM NIM NeuN D API
MSC me_H&R MSC me_H&R
D
20 10 0
7060 50 40 30 20 10 0
0.8 0.6 0.4 0.2 0 2.0 1.5 1.0 0.5 0
MSC me_H&R MSC me_H&R
CXXC1
Figure 3. Methylation changes and stemness of me_H&R- transfected MSCs. A, left, altered methylation levels of the MSC methylome depicted by the heatmap of the DMH data. Mock- transfected cells were labeled with Cy3, me_H&R-transfected cells were labeled with Cy5. Red and green lines correspond to hyper- and hypomethylation,
respectively. Right, arrows highlight hypermethylated genes in selected array blocks. B, validation of altered methylation by qMSP. C, visualization of affected loci by unsupervised pathway finding (PathVisio; ref.
44). D, stemness of me_H&R- transfected MSCs.
Immunostaining for Oct4 and CD133 stem cell markers (top left).
Phase contrast and
immunostaining of MSCs during neuronal induction (right) and osteocyte induction (bottom left).
pre-NIM, neuronal preinduction medium; NIM, neuronal induction medium. The neuronal
differentiation was indicated by
NeuN immunostaining. AP,
alkaline phosphatase.
normal somatic stem cells. Furthermore, our data indicated that methylation of HIC1 initiated the accumulation of epi- genetic changes in MSCs and increased the potential for transformation. In normal cells, the transformation process can be prohibited by gatekeepers or roadblocks, such as RassF1A; thus, gatekeeper silencing can allow for excess cell proliferation and/or genome corruption, increasing the prob- ability of neoplastic transformation.
Transfection of me_H&R induces genome-wide DNA methylation changes in MSC
To further investigate changes in DNA methylation follow- ing me_H&R targeting, genome-wide methylation profiling of me_H&R-transfected and control MSCs was carried out using DMH microarrays (23). As shown in the methylation heatmap, transfection of me_H&R induced extensive disruption in the MSC epigenome (Fig. 3A, left). Individual loci (Fig. 3A, right) displaying hypermethylation (SOX7, ADAM9, and GATA6) or hypomethylation (CXXC1, OSBP, and TBX2) were validated by qMSP (Fig. 3B). On the basis of the wide range of cellular
functions associated with HIC1 and RassF1A (16, 19, 24), it was not surprising to observe global changes in the MSC methy- lome. Furthermore, as both HIC1 and RassF1A were reported to function as TSGs that act via p53, we conducted unsuper- vised ontological analysis of the array target loci and found that p53 and its associated signaling components were sig- nificantly altered by me_H&R transfection (Fig. 3C and Sup- plementary Fig. S7, target loci with the p53 binding domain are listed in Supplementary Table S3). Our finding that MSC transformation is strongly associated with altered p53 func- tion is consistent with a previous report using genetic approaches (15).
Characteristics of me_H&R-transformed MSCs
To determine whether the me_H&R-transformed MSCs retained stemness, me_H&R transfected MSCs were labeled with antibodies against stem cell surface markers CD133 and Oct4 and induced to differentiate. Expression levels of CD133 and Oct4 were unaffected by me_H&R transfection (Fig. 3D, top left). Multipotency was assayed by neuronal induction and
C
3.0 2.5 2.0 1.5 1.0 0.5 0
Colon y for mation (f old) T ransition (f old)
12 10 8 6 4 2 0
Cell sur viv al (f old)
1.2 1.0 0.8 0.6 0.4 0.2 0
MSC MSC
me_H&R me_H&R MSC me_H&R
– – –
–
– –
– – –
– – +
+ +
+ +
+ +
Isolated me_H&R +
+ –
– – – RassF1A me_RassF1A me_HIC1-TR 20 µmol/L 5-Aza-dc
– – –
–
– –
– – –
– – +
+ +
+ +
+ +
Isolated me_H&R +
+ –
– – – RassF1A me_RassF1A me_HIC1-TR 20 µmol/L 5-Aza-dc
10 µmol/L
CD133 100 µmol/L Cisplatin
D
A B
100 µmol/L
Figure 4. In vitro and in vivo tumorigenesis of me_H&R-transfected MSCs. A, in soft agar assays, me_H&R-transfected MSCs formed a greater number of colonies versus controls (5th and 7th columns), and colony formation was inhibited by 5-aza-dc treatment (6th column). B, quantification of migrated cells. Me_H&R-transfected MSCs migrated to the bottom of the Transwell inserts (control cells did not) and this migration was blocked by 5-aza-dc treatment. C, increased drug resistance in me_H&R-transfected MSCs. Cisplatin-induced cell death was measured by MTT assay. High dose cisplatin treatment (100 mmol/L) caused massive cell death in control MSCs but not in me_H&R-transfected MSCs (histogram). Representative images of cisplatin- induced cell death (right). D, tumor development in mice inoculated with me_H&R-transfected MSCs (n ¼ 9; left top). H&E staining (right) and
immunohistochemistry of CD133 expression (bottom left) were carried out on xenograft tumors.
osteogenic differentiation. MSCs transfected with me_H&R were able to differentiate into neuron-like cells (Fig. 3D, right, indicated by NeuN staining) and osteoblasts (Fig. 3D, left bottom, alkaline phosphatase positive). These data indicate that although concurrent methylation of HIC1 and RassF1A TSGs was sufficient to transform MSCs, as indicated by altered anchorage dependence and methylation patterns, the trans- formed MSCs retained somatic stem cell characteristics.
The proliferation and invasion capability of transformed MSCs were evaluated by using colony formation and Trans- well studies. Colony formation was substantially increased (8-fold, n ¼ 6) in me_H&R-transfected cells versus controls (Fig. 4A and Supplementary Figs. S8A and S9A), and this was inhibited by 5-aza-dc treatment (Fig. 4A and Supplementary Fig. S8A). In Transwell studies, MSCs transfected with me_H&R exhibited greater migratory capability (Fig. 4B and
A
B
C
Oct4 CD133 CD44 Vimentin NSE Desmin Xenog raft MSC
CD133
CD133 CD133
Vimentin DAPI
NSE LCA
DAPI DAPI
Merged
Merged Merged
me_H&R
Variations in primary
EMT or MET?
Selection
More transformation and selection
Selections by niches in primary or secondary
Liver GI Brain Fat Lung
LCA CK S-100
D
Figure 5. Immunohistochemistry of xenografts from me_H&R-transfected MSCs. A, stem cell marker expression in xenografts. Representative images
of cells positive for CD44, CD133, or Oct4 from the serial sections of xenograft tumors. B, representative images from xenografts (n ¼ 36) containing
heterogeneous cell types of either epithelial or mesenchymal lineage (vimentin). C, expression of lineage-specific markers (left: NSE, right: LCA) was observed
in many CD133-positive cells, or lineage-specific marker expressing cells surrounded CD133-positive cells. D, top, images of coexpression of CD133
and vimentin in a subpopulation of control MSCs (top), me_H&R-transfected MSCs (middle), and subcultures derived from me_H&R-transfected MSC
xenografts (bottom). Bottom, simplified model for mechanisms by which normal somatic stem cells (primary) can gradually become tissue-specific tumor
cells, via transformation and clonal selection occurring at niches in primary or secondary sites. Colored circles represent different lineages. EMT: epithelial-to-
mesenchymal transition; GI, gastrointestine.
Supplementary Figs. S8B and S9B). Taken together, these results indicated that MSCs with concurrent methylation of HIC1 and RassF1A acquired a cancer phenotype.
Acquired drug resistance is a hallmark of malignancy and a characteristic of a cancer stem cell (CSC) phenotype. Cisplatin treatment induced cell death in control MSCs in a dose- dependent manner, but the drug was much less effective in me_H&R-transfected MSCs, even at a very high dose (100 mmol/L; Fig. 4C), further indicating that concurrent HIC1 and RassF1A methylation transformed normal somatic stem cells into CSC like.
To examine the tumorigenic capacity of the me_H&R- transfected MSCs in vivo, immunodeficient nude mice were inoculated with me_H&R-treated MSCs. As shown in Figure 4D, 100% of these mice developed tumors (n ¼ 9), whereas the mice inoculated with control MSCs remained tumor free.
Immunohistochemistry revealed the presence of CD133
þstem cells in the tumors (Fig. 4D, bottom left). Furthermore, the tumors were soft tissue sarcomas like (Figs. 4D, right and 5A and B), consistent with this type of malignancy in mice with heterozygous disruption of HIC1 (17).
Expression of stem cell markers in me_H&R MSC- derived tumors
We examined xenografts from serial transplantations for expression of known stem cell markers by immunohistochem- istry. As shown in Figure 5A, expression of CD44, CD133, and Oct4 was observed, substantiating the CSC-like phenotype.
Vimentin expression was also detected, confirming the mesenchymal origin of the xenografts. Clonal expression of a panel of epithelial markers, including neuron-specific eno- lase (NSE), S-100, cytokeratin, desmin, and leukocyte common antigen (LCA; Fig. 5B), further showed the ability of CSCs to differentiate into heterogeneous tumors. In addition, expres- sion of both stem cell (CD133) and mesenchymal (vimentin) markers in sparsely scattered cells of me_H&R tumors showed that the xenografts were derived from inoculated me_H&R MSCs (Fig. 5C). Although we observed slight enrichment of CD133
þ/vimentin
þcells in serial transplantation experiments (Supplementary Fig. S10A), the overall percentage remained low, in agreement with previous observations (25) that CSCs continue to form a small proportion of the overall tumor even after in vitro enrichment and xenograft transplantation.
Figure 6. TDM of SWH signaling pathway is not sufficient for full MSC transformation. In vitro methylation of the main components within SWH pathway ( me_SWH) was carried out and transfected cells were characterized. A, methylation of the 9 SWH pathway components was detected by qMSP. In the physical maps, short, filled boxes indicate the target sites in each promoter and the arrow heads indicate the primer sites used to detect the TDM on the left. B, images, methylation of the SWH pathway caused the MSC to lose contact inhibition (top right).
Treatment with 5-aza-dc reversed the aggregate phenotype caused by TDM (bottom right).
Histograms, me_SWH-treated MSCs exhibited higher growth rate in soft agar assay versus mock- treated cells. C, me_SWH-treated MSCs retained stemness properties. Me_SWH-treated MSCs expressed stem cell markers CD133 and Oct4 (top left) and neuronal (right) and osteogenic (bottom left) lineages could be induced. D, no tumor formation was observed after subcutaneous injection of me_SWH-treated MSCs in nude mice (middle; n ¼ 8); in contrast, subcutaneously implanted me_H&R-treated MSCs formed tumors (right). TSS, transcriptional start site.
A MSC
CXCR4 LATS1 LATS2 FAT4
FAT MOBKL1A MOBKL1B NF2
0 0 0 0 0 0 0 0
0 2.5 5.0 7.5
0.9 1.8 2.7
4 8 12
1.5 3.0 4.5
1.5
5 10 15
3.0 4.5
5 10 15
2.5 5.0 7.5
6 12
Methylation (fold)
MSC MSC
B D
Ctr l +5-Aza-dc
me_SWH
me_SWH me_H&R 18
Gro wth r a te (f old)
2.5 2.0 1.5 1.0 0.5 0
MSC AP Staining D API CD133 D API Oct4
Methylation target site
C
MSP primer
TSS 100bps
