Progressive tumor features accompany epithelial–mesenchymal transition induced in mitochondrial DNA–depleted cells

Abstract

The growth of LNCaP, a human prostate adenocarcinoma cell line, and MCF‐7, a human breast adenocarcinoma cell line, is initially hormone dependent. We previously demonstrated that LNρ0‐8 and MCFρ0, derived from LNCaP and MCF‐7 by depleting mitochondrial DNA (mtDNA), exhibited hormone‐independent growth that led to progressed phenotypes. Here, we demonstrate that LNρ0‐8 and MCFρ0 have invasive characters as evaluated by the ability of invasion through the extracellular matrix (ECM) in vitro. In addition, the induction of vimentin and the repression of E‐cadherin expression in ρ0 cells indicate that they are mesenchymal cells. Since LNρ0‐8 and MCFρ0 were derived from epithelial cancer cell lines, LNCaP and MCF‐7 must have lost epithelial features and gained the mesenchymal phenotype by epithelial‐mesenchymal transition (EMT) during the mtDNA depletion. In the ρ0 cell lines, the Raf/MAPK signaling cascade was highly activated together with the expressions of transforming growth factor‐beta (TGF‐β) and type I TGF‐β receptor (TGF‐βRI). EMT requires cooperation of TGF‐β signaling with activation of the Raf/MAPK cascade, suggesting that EMT was induced in mtDNA depleted cells resulting in the acquisition of progressive tumor features, such as higher invasiveness and loss of hormone dependent growth. Our results indicate that decreasing mtDNA content induces EMT, enabling the progressive phenotypes observed in cancer. (Cancer Sci 2008; 99: 1584–1588)

Abbreviations:

AGPC

acid guanidium‐phenol‐chloroform

BSA

bovine serum albumin

DAPI

4′,6‐diamidino‐2‐phenylindole

DMEM

Dulbecco's Modified Eagle Medium

ECM

extracellular matrix

ELISA

enzyme‐linked immunosorbent assay

EMT

epithelial–mesenchymal transition

EtBr

ethidium bromide

FCS

fetal calf serum

GAPDH

glyceraldehyde‐3‐phosphate dehydrogenase

MAPK

mitogen activated protein kinase

MET

mesenchymal–epithelial transition

mtDNA

mitochondrial DNA

mtDNA

mitochondrial DNA

OXPHOS

oxidative phosphorylation

PCR

polymerase chain reaction

RIPA

radioimmuno precipitation assay

RFU

relative fluorescence units

ROS

reactive oxygen species

RT‐PCR

reverse transcription–PCR

TGF‐β

transforming growth factor‐beta

TGF‐βRI

type I transforming growth factor‐1 receptor

It is known that prostate and breast cancers originate in hormone‐dependent forms. However, they recur as hormone‐independent phenotypes during cancer progression. The mechanisms involved in this phenotypic change are not completely understood. Previously, we have demonstrated that C4‐2 cells, established from LNCaP cells (a human prostate adenocarcinoma cell line) in an androgen‐deprived environment, and 4‐hydoroxy tamoxifen (antiestrogen)–resistant MCF‐7 cells (a human mammary adenocarcinoma cell line), have a greatly reduced amount of mtDNA. Moreover, mtDNA depleted LNCaP (designated as LNρ0‐8 cells) and MCF‐7 (designated as MCFρ0 cells) changed from androgen‐dependent to androgen‐independent, and from antiestrogen‐susceptible to antiestrogen‐resistant, respectively. MtDNA transferred ρ0 cells (cybrids) were recovered from hormone‐independent to hormone‐dependent forms. Collectively, these facts suggest that mtDNA regulates hormone‐dependent growth.^{( 1} , ^{2 )}

Epithelial cells are the cells that cover surfaces, line a cavity, perform secretion, transport, or regulate functions. They (1) have cohesive interactions among cells; (2) facilitate the formation of continuous cell layers; (3) possess three membrane domains: apical, lateral, and basal; (4) are connected through tight junctions between apical and lateral domains; (5) have an apicobasal polarized distribution of the various organelles and cytoskeleton components; and (6) lack mobility as individual cells with respect to their local environment. Mesenchymal cells are the cells of mesodermal origin that are capable of developing into connective tissues, blood, and lymphatic and blood vessels. In contrast to epithelial cells, mesenchymal cells (1) have loose or no interactions among each other, so a continuous cell layer cannot be formed; (2) have no clear apical and lateral membranes; (3) possess no apcicobasal polarized distribution of organelles and cytoskeleton components; and (4) are motile cells that may even have invasive properties.^{( 3} , ⁴ , ^{5 )}

The EMT represents the transition of epithelial cells to a mesenchymal phenotype, and it features a loss of epithelial cell markers such as junctional and cell–cell adhesion proteins. In development, EMT is observed in mesoderm formation and in emigration of neural crest cells from the neural tube in avian and in mammalian embryo. In adults it has been implicated that EMT plays an important role in tumor formation and progression to metastatic carcinomas.^{( 6 )} The autocrine TGF‐β loop cooperating with oncogenic ras activation is required for the maintenance of EMT in epithelial cells and for metastasis in a mouse model.^{( 7 )}

In this report, we demonstrate that LNρ0‐8 and MCFρ0 are of the mesenchymal phenotype by invasive features with induction of vimentin and suppression of E‐cadherin expression. In these cells, the Raf/MAPK signaling cascade was highly activated together with TGF‐β1 and TGF‐βRI expression. Those results suggest that EMT was induced in mitochondrial DNA–depleted cells, and that resulted in the acquisition of progressive features, such as increased invasiveness and hormonally deregulated cell growth. The association between altered mtDNA copy number, especially a decrease^{( 8} , ⁹ , ¹⁰ , ^{11 )} with cancer initiation and progression, has been reported. Our results provide a mechanism for the cancer progression, suggesting that a decrease in mtDNA induces EMT, thereby shifting the cancer cells into a progressive phenotype.

Materials and Methods

Cell culture.  LNCaP was purchased from UROCOR (Oklahoma City, OK, USA). MCF‐7 was obtained from ATCC (Manassas, VA, USA). The cells were cultured at 37°C in a humidified atmosphere of 5% CO₂ with RPMI‐1640 media (ATCC) containing 10% FCS (Hyclone, Logan, UT, USA) for LNCaP and with DMEM (Invitrogen, Carlsbad, CA, USA) containing 5% FCS supplemented with 0.01‐mg/mL bovine insulin (Sigma‐Aldrich, St. Louis, MO, USA) for MCF‐7. The establishment of mtDNA‐depleted cell lines (LNρ0‐8 and MCF7ρ0) are described elsewhere.^{( 1} , ^{2 )}ρ0 cells were maintained in DMEM supplemented with 10% FCS, 50 µg/mL of uridine, and 100 µg/mL of sodium pyruvate.

Invasion assay.  A Fluorometric QCM 24‐Well Cell Invasion Assay kit (Chemicon, Temecula, CA, USA) was used according to the manufacturer's instructions to evaluate the invasive ability of cells. Briefly, the concentration of cells was adjusted to 0.5 × 106 cells/mL in FCS‐free DMEM containing 5% BSA (Sigma‐Aldrich) after the starvation in FCS‐free DMEM for 24 h. A total of 250 µL of cell suspension was seeded in the inner chamber and 500 µL of DMEM containing 10% FCS was added to the outer chamber. After culturing for 72 h in a CO₂ incubator, the invaded cells on the bottom of the insert membrane were dissociated from the membrane and lyzed by lysis buffer. The RFU of cellular nucleic acid in the lysates by binding to CyQUANT dye was measured using 480/520 nm filter set. A total of 50 µg/mL uridine and 100 µg/mL sodium pyruvate were supplemented for ρ0 cells. Assays were performed in triplicate.

Immunocytochemistry.  Cells were plated with submerged cover slips. Primary antibodies used were rabbit anti‐E‐cadherin antibody (Cell Signaling Danvers, MA, USA), mouse antivimentin monoclonal antibody (Sigma‐Aldrich), or mouse anti‐SMAD2/3 antibody (BD Biosciences, San Jose, CA, USA). Alexa‐488‐conjugated secondary antibodies (Invitrogen) were used for the visualization of the primary antibodies. The nuclei were counter stained by 300‐nM DAPI (Invitrogen). The cover slips were mounted onto glass slides with Pro Long Gold antifadant mounting medium (Invitrogen). The images were obtained with a confocal microscope (LSM410; Carl Zeiss, Oberkochen, Germany) with ×63 objective.

Western blotting.  Cells were lyzed with RIPA buffer containing a protease inhibitor cocktail (Sigma‐Aldrich). The primary antibodies used were mouse monoclonal antiphospho c‐Raf (Ser338; Millipore), rabbit antiphospho c‐Raf (Ser259; Cell Signaling), rabbit antiphospho MEK1/2 (Cell Signaling), rabbit antitotal MEK1/2 (Cell Signaling), rabbit antiphospho p38 (Cell Signaling), rabbit antitotal p38 (Cell Signaling), rabbit antiphospho JNK (Cell Signaling), rabbit antitotal JNK (Cell Signaling), rabbit antiphospho p44/42 (Cell Signaling), and rabbit antitotal p44/42 antibody (Cell Signaling). The primary antibodies were detected with matched secondary antibodies labeled with horseradish‐peroxidase (Cell Signaling). The signals were visualized using ECL Plus (GE Healthcare Bio‐Sciences, Uppsala, Sweden) by exposure to X‐ray films.

ELISA.  Cells were seeded at a concentration of 2 × 10⁴ cells/mL into a 96‐well plate (200 µL/well) with serum‐free DMEM for 16 h. The supernatants were collected and TGF‐β1 in the supernatant was measured by a Quantikine Human TGF‐β 1 ELISA kit (R&D Systems, Minneapolis, MN, USA) according to the manufacturer's instructions.

Reverse transcrptase coupled–PCR.  Total RNA was isolated by the modified AGPC method (TRI reagents; Ambion, Austin, TX, USA). One‐µg of RNA primed with oligo (dT)_18–20 (Invitrogen) was reverse transcribed by Superscipt III (Invitrogen). Primers were designed to amplify from 234 to 647 of TGF‐βRΙ cDNA. Ex Taq (Takara Mirus, Madison, WI, USA) was used for the amplification. The amplification protocol comprised 32 cycles of denaturation for 20 s at 94°C, annealing for 20 s at 56°C, and elongation for 30 s at 72°C. The entire coding sequence of GAPDH cDNA was amplified as a control.

Results

Mitochondrial DNA depleted ρ0 cells showed higher motility.  We have previously shown that the amount of mtDNA regulates hormone‐dependent growth of prostate and mammary cancer cells.^{( 1} , ^{2 )} However, the mechanisms underlying this are still unclear. Hormone‐independent growth is one of the characteristics of progressed cancers, as is invasive ability, a feature we have not previously investigated in mtDNA‐compromised cells. Parental LNCaP and MCF‐7 cells did not show any invasiveness through an ECM matrix; however, when we looked at LNρ0‐8 and MCFρ0 cells, we found them to be highly invasive (Fig. 1). Thus, we can identify ρ0 cells as progressed cancer cells and utilize them as a model for cancer progression.

Figure 1.

Cell invasion assay. LNCaP, MCF‐7, and their ρ0 cells were cultured in an inner chamber with an 8‐µm pore membrane coated with extracellular matrix (ECM). The neighboring outer chamber was filled with 10% fetal calf serum containing medium. The cells were cultured for 72 h. Invaded cells on the bottom of the insert membrane were dissociated from the membrane and lyzed. The relative fluorescence units (RFU) of cellular nucleic acid in the lysates by binding to CyQUANT dye were measured using a 480/520 nm filter set. The assay was performed in triplicate and the graph shows the triplicate averages. Error bars represent standard deviations.

ρ0 cells express mesenchymal markers.  Motility of the cells with invasiveness is the fundamental characteristic of mesenchymal cells; therefore, we speculated that ρ0 cells might have acquired mesenchymal phenotypes. We analyzed expression of E‐cadherin, a marker of epithelial cells, and vimentin, a marker of mesenchymal cells. As we hypothesized, both ρ0 cells showed mesenchymal features (Fig. 2). The vimentin was highly expressed in cytosol with a three‐dimensional structure and no E‐cadherin was expressed. MCF‐7 cells exhibited clearly epithelial properties; E‐cadherin was expressed on the cell surface and no vimentin was expressed. Although LNCaP cells clearly expressed E‐cadherin on their cell surfaces, vimentin signals were also observed slightly with nonstructural staining pattern in the cytosol. Because the staining pattern of vimentin in LNCaP did not look like intermediate filament structure (compared with that in ρ0 cells), the signals of vimentin in LNCaP cells might be artifacts, Since ρ0 cells originated from epithelial cells, it is clear that the epithelial cells transitioned to mesenchymal cells during the course of mtDNA depletion. Indeed, we have found the E‐cadherin promoter was hypermethylated in both LNρ0‐8 and MCFρ0 cells.^{( 12 )}

Figure 2.

Immunocytochemistry for E‐cadherin and vimentin. The indicated cells were cultured with submerged cover slips and stained with anti‐E‐cadherin or antivimentin antibodies. The primary antibodies were visualized by Alexa‐488‐conjugated secondary antibodies. The nuclei were counter stained with 4′,6‐diamidino‐2‐phenylindole (DAPI). The magnification bars represent 20 µm.

MAPK, JNK signaling cascade, and TGF‐β production are up‐regulated in ρ0 cells.  The cooperation of TGF‐β signaling with ras activation induces EMT.^{( 7 )} We analyzed the Raf/MAPK cascade as an endpoint of ras activation. Phosphorylation of c‐Raf on Ser259 has been proposed to inactivate the p44/42 pathway,^{( 13 )} while phosphorylation of c‐Raf on Ser338 is required for p44/42 phosphorylation.^{( 14 )} Phosphorylation of c‐Raf on Ser259 was only observed in parental LNCaP and MCF‐7 cells but not in the ρ0 cells (Fig. 3a). On the other hand, c‐Raf on Ser338, together with MEK1/2 and p44/42 were highly phosphorylated in ρ0 cells (Fig. 3a), suggesting that the Raf/MAPK pathway was highly activated in ρ0 cells. We have not yet identified the molecules responsible for activation of the Raf/MAPK pathway, nor have we determined whether or how oncogenic ras and RTKs are involved. When we measured the production of TGF‐β, we found that TGF‐β production in both LNρ0 and MCFρ0 cells is about 20‐times higher than in parental cells (Fig. 3b). In addition, we also investigated the JNK and p38 pathway. As shown in Fig. 4, JUNK, but not the p38 pathway, was up‐regulated in both LNρ0 and MCFρ0 cells.

Figure 3.

Signal transduction leading to epithelial–mesenchymal transition. (a) Western blotting for Raf/MAPK signal transduction. A total of 20 µm of the indicated lysates were loaded into each lane of an SDS‐PAGE gel; samples were electrophoresed and membrane transferred. The membranes were blotted with antiphospho c‐Raf (Ser338), antiphospho c‐Raf (Ser259), antiphospho MEK1/2, and antiphospho p44/42. The membrane blotted with antiphospho p44/42 was reprobed and blotted with total p44/42 antibody. (b) Transforming growth factor‐beta (TGF‐β) production measured by enzyme‐linked immunosorbent assay (ELISA). The supernatants were collected from 4 × 103 cells culture with serum‐free Dulbecco's Modified Eagle's Medium (DMEM) for 16 h. The concentration of TGF‐β1 in the supernatant was measured by the ELISA. The average from triplicate assays was plotted on the bar graph. Standard deviations were represented as Y‐error bars. Each value is summarized below the plot.

Figure 4.

Western blotting for JNK and p38 signal transduction. A total of 20 µm of the indicated lysates were loaded into each lane of an SDS‐PAGE gel; samples were electrophoresed and membrane transferred. The membranes were blotted with anti phospho JNK, antitotal JNK, antiphospho p38, and antitotal p38.

TGF‐β receptor expression is restored in ρ0 cells.  It has been reported that TGF‐β has opposite functions, including inducing cell growth arrest and having a mitogenic effect. LNCaP losing expression of TGF‐βRI resulted in escape from cell‐growth arrest. ^{( 15 )} Thus, we analyzed signal transduction induced by TGF‐β in ρ0 cells. First, TGF‐βRI expression was analyzed by RT‐PCR. Both ρ0 cells showed TGF‐βRI expression although LNCaP and MCF‐7 showed no expression of TGF‐βRI (Fig. 5a). SMAD‐2/3 translocated into nuclei within 30 min after addition of TGF‐β (final concentration, 10 ng/mL) to ρ0 cells but not in parental cells (Fig. 5b). Thus the TGF‐β response with expression of TGF‐βRI and TGF‐β in ρ0 cells suggests that the TGF‐β signaling pathway is functional in ρ0 cells. However, without addition of exogenous TGF‐β, spontaneous SMAD‐2/3 translocation to the nucleus could not be detected (Fig. 5b, left panels).

Figure 5.

Signal transduction by transforming growth factor‐beta (TGF‐β). (a) Reverse transcription–polymerase chain reaction (RT‐PCR) for type I TGF‐β receptor. Total RNA was isolated from the indicated cells and 1 µg was reverse transcribed using oligo dT. Primers were designed to amplify from 234 to 647 of type I TGF‐β receptor cDNA. The entire coding sequence of glyceraldehyde‐3‐phosphate dehydrogenase (GAPDH) cDNA was amplified as a control. (b) Immunocytochemical staining of SMAD2/3. The indicated cells were fixed without addition of TGF‐β (–TGF‐β, left panels) or 30 min after TGF‐β addition (+TGFβ, right panels), and then stained with anti‐SMAD2/3. The primary antibodies were visualized by Alexa‐488‐conjugated secondary antibodies. The nuclei were counterstained by DAPI. The magnification bars represent 20 µm.

Discussion

Although we have found that the down‐regulation of E‐cadherin is due to hypermethylation of its promoter region in LNρ0‐8 and MCFρ0 cells,^{( 12 )} the transcription factor is responsible for the expression of vimentin in both ρ0 cells. A couple of molecules and pathways are thought to be involved in EMT induction in addition to the cooperation of TGF‐β signaling with Raf/MAPK. Transcription factors, such as SNAIL and TWIST, are proposed to enhance vimentin transcription and down‐regulate E‐cadherin.^{( 16} , ^{17 )} In the E‐cadherin promoter, there is a repressor element called E‐pal, where SNAIL binds to suppress E‐cadherin expression.^{( 18 )} SNAIL is regulated by GSK3β through WNT signaling,^{( 19 )} and hypoxia is proposed to regulate EMT through SNAIL.^{( 20 )} In the case of LNρ0‐8 and MCFρ0 cells, we could not detect any up‐regulation of Wnt signaling, indicating that SNAIL may not be involved in the EMT induced by reduction of mtDNA content (data not shown). Hypoxia cannot stabilize hypoxia inducible factor‐1‐alpha (HIF‐1α) in ρ0 cells,^{( 21 )} although there are contradictions about the hypoxia response in ρ0 cells.^{( 22 )} Therefore the involvement of HIF‐1α is also unlikely. Thus TGF‐β signaling with Raf/MAPK signaling may play a significant role in EMT in ρ0 cells, though the transcriptional regulation of vimentin should be more precisely investigated.

Since inhibitors of mitochondrial respiratory chains and uncouplers can induce MAPK activation,^{( 23 )} the mechanisms underlying Raf/MAPK activation by mtDNA depletion have been implicated. We speculate that the alteration of mitochondrial membrane potential, cytosolic calcium concentration, shift of ATP generation from respiration to glycolysis, redox regulation, O₂ sensoring, or other mitochondrial changes may up‐regulate Raf/MAPK signaling.

Here, we have demonstrated that mtDNA‐depleted cell lines acquire invasive features and hormonal‐independent growth through EMT caused by activation of the Raf/MAP kinase pathway together with TGF‐β signal transduction. Indeed, we have proven that mtDNA content was deregulated in prostate cancer cells compared with normal adjacent cells using the laser microdissection coupled with real‐time PCR.^{( 24 )} In addition, breast,^{( 9 )} renal,^{( 11 )} and liver^{( 8} , ^{10 )} cancers reportedly have reduced mtDNA contents, while head and neck,^{( 25} , ^{26 )} lung,^{( 26 )} thyroid,^{( 9 )} and pancreas^{( 27 )} cancers have increased. It is possible that abnormality of p53 observed in many cancers deregulate mtDNA contents. This idea is supported by the fact that p53 accumulates in mitochondria in a transcription‐independent manner^{( 28 )} and maintains mtDNA stability;^{( 29 )} and/or that environmental pressures for disturbing the maintenance of mtDNA can also cause mtDNA decrease. For example, tamoxifen causes mtDNA decrease in mammary cancer cells and hepatocytes,^{( 2} , ^{30 )} and androgen ablation causes mtDNA decrease in prostate cancer.^{( 1 )} ROS, by products of respiration in mitochondria, can also damage mtDNA.

The role of EMT in development has been widely accepted; however, the role of EMT in cancer progression remains a big debate because it has not been proven in vivo; the criteria for the identification of carcinoma (cancer derived from epithelial cells) and sarcoma (cancer derived from mesenchymal cells) have been pathologically well defined and are not thought to be convertible, and the mixed type of the tumor, known as sarcomatoid carcinoma, is rarely found. Here, we propose that a decrease in mtDNA by any number of mechanisms including ROS, hypoxia, and chemotherapeutic agents, induce EMT to result in a progressed phenotype. After escaping from a growth disadvantage due to the environmental pressure by metastasis or deregulation from hormone dependence, mtDNA copy number can then be increased.^{( 2} , ^{31 )} Finally the cells may morphologically regain epithelial features (MET) at metastatic sites. That is the possible reason that we can not find EMT in vivo. Our idea will be supported by the identification of increased or reduced mtDNA content in vivo. We propose the theory that cancers achieve their growth advantage/progression via EMT/MET induced by the alteration of mtDNA amounts.