| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
,1
* Max-Planck-Institut für Biochemie, Martinsried bei München, Germany;
LCMB NIH/NCI, Bethesda, Maryland, USA;
Immunopathologie des maladies tumorales et auto-immunes, INSERM U475, Montpellier, France;
EMI 229 INSERM Génotypes et Phénotypes Tumoraux, CRLC Val d’Aurelle, Montpellier, France;
|| Endocrinologie Moléculaire et Cellulaire des Cancers, INSERM U540, Montpellier, France;
¶ Gynaecologic Oncology and Breast Cancer Unit, University of Torino School Medicine, Italy; and
** Institut de Génétique Humaine, CNRS-UPR1142, Montpellier, France
2 Correspondence: Institut de Génétique Humaine, CNRS-UPR1142, 141 rue de la Cardonille, 34396 Montpellier Cedex 5, France. E-mail: vincent.francois{at}igh.cnrs.fr
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Key Words: ovarian carcinoma • BMP antagonist
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
, 2)
. Incessant ovulation has been proposed as a predisposing factor for carcinogenesis because of repeated rupture and repair of this OSE at the ovulation site (3)
.
OSE forms a continuous cell monolayer and, in contrast to most epithelia, it is tenuously attached to its basement membrane and has properties of relatively uncommitted pluripotential cells. At the ovulation site, OSE cells acquire motility, contractility and can proliferate, properties that contribute to the repair of the disrupted ovarian surface. Several growth factors can promote these events in vitro, but the molecular players involved in vivo are still unknown. Epithelial inclusion cysts are found in the ovarian cortex and it has been proposed that they arise from OSE fragments trapped at the time of ovulation (reviewed in ref 1
). These cysts are considered to be sites of metaplastic changes that lead to tumorigenesis (3)
.
BMPs, like the other transforming growth factor ß (TGFß) superfamily, elicit their cellular effects at the membrane as dimeric ligands for Ser/Thr kinase receptors that signal through Smad molecules to the nucleus. With some 20 members in vertebrates, BMPs are involved in a great variety of developmental events, including early embryonic axis determination, organogenesis and vasculogenesis (4)
. BMP signaling generally counteracts the mitogenic effect of other signals, directs cell fate choices or induces apoptosis. BMPs have recently emerged as central players in ovary physiology and female fertility. In vitro studies have established that they have mitogenic effects on granulosa cells and influence their differentiation and steroidogenesis (5
6
7)
, while in vivo genetic inactivation of BMPs expressed in oocytes or in BMP receptors is a cause of infertility (8
9
10
11
12)
. Expression of several BMPs has been reported in rat OSE in situ (5
, 13)
as well as in human cultured OSE (14)
.
During development, BMP activity is regulated by high-affinity binding proteins secreted in the extracellular matrix (15)
. Most of them, like Chrd, Noggin, and DAN family members, behave as antagonists by binding BMPs in the extracellular space or, like Follistatin, by interacting with their receptors. In ovaries, Follistatin is required for follicle growth and fertility (16)
; Gremlin, a DAN family member expressed in mice follicles, can block the effects of BMP4 on granulosa cells in vitro (17)
. Addition of Follistatin, Gremlin, or Chrd to bovine theca cell cultures modifies their androgen production (18)
. As Chrd mice mutants die early in development (19)
, genetic analysis has not shed any light on its possible role in postnatal ovarian development or biology. We previously reported CHRD expression in human ovary (20)
, but its role is still unknown. Chrd binds BMP2 and -4 homodimers as well as BMP4/-7 heterodimers (21)
, and BMP5 and -6, albeit more weakly (22)
; all of these BMPs, except BMP5, are expressed in the ovary.
Similar to other TGFßs, BMP signaling has dual effects on carcinogenesis. BMPs hamper tumorigenesis in situations where they are growth inhibitory or proapoptotic (23
24
25)
, but they can also promote malignancies by stimulation of growth or angiogenesis and attenuation of apoptosis of cancer cell lines (26
, 27)
. BMP signaling was shown to be fully functional in ovarian cancer cells, and exposing these cells to BMP4 was found to induce Id proteins (14)
, which are BMP targets with oncogenic properties in several cancers including epithelial ovarian cancer (EOC) (28)
. Potential inhibition of tumorigenesis has been reported for BMP antagonists (29
30
31)
.
The Chrd/BMP module is a well documented example of extracellular signaling regulation in animal development. Here we undertook a study of the role of CHRD in ovarian physio-pathology. Our results suggest that CHRD participates in OSE biology at ovulation sites and may behave as a suppressor of tumorigenesis in OSE derived carcinoma.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
with a 938 bp cDNA CHRD probe (Image clone AA036834) and a 2 kb cDNA N-myc probe (provided by Dr. F. Alt, ref 33
) separately. Signal intensities were estimated as described previously (32)
.
Fluorescent in situ hybridization (FISH)
To estimate CHRD and bcl6 copy numbers, dual-color FISH was performed on tumor-touch preparations as described in Bautista and Theillet (34)
. bcl6 was used as a control for the 3q 27 region. A 80 kb CHRD PAC genomic clone (HGMPRC, Cambridge) and a 16 kb bcl6 genomic fragment (gift from Dr. Bastard, EMI 9906, Rouen, France) were labeled by nick-translation with biotin-16-dUTP or digoxigenin-11-dUTP (DIG, Roche), respectively. Images were captured on a Zeiss (Le Pecq, France) epifluorescence microscope equipped with a JAI (Glostrup, Denmark) CCD camera run by the Metasystems (Altlussheim, Germany) image analysis software package. Only well separated nuclei were considered for fluorescent spot counting. At least 15 microscope fields were captured for each slide, with each field containing between 1 to 5 interphase nuclei, depending on the extent of DNA decondensation. For each tumor sample, spots were counted on 
40 nuclei. Tumor heterogeneity is observed in each sample, as nuclei present a variable number of fluorescent spots, generally in a factor of two.
RT-PCR experiments on cell lines
Total RNAs were prepared from ovarian cell lines by Trizol extraction (Gibco, Grand Island, NY, USA). cDNA synthesis was performed with 1 µg of total RNA using random hexamer oligonucleotides (Promega, Madison, WI, USA) and M-MLV reverse transcriptase (Gibco). RT efficiency was controlled by HPRT gene amplification. For BMPs, PCR was performed with Taq DNA polymerase (Pharmacia, Piscataway, NJ, USA) on 2 µL of RT product as follows: 5 min at 94°C followed by 35 amplification cycles (30 s at 94°C; 1 min at 60°C; 1 min at 72°C). CHRD and GDF-9 were amplified with HotStart Taq polymerase (Qiagen, Chatsworth, CA, USA). GDF9: 15 min at 94°C followed by 30 s at 94°C, 30 s at 51°C, and 1 min at 72°C (35x). A touchdown protocol was applied for CHRD as described previously (20)
. The sequence of primers used for BMPs and amplicon length are described in Table 1
. CHRD PCR were performed with variant 4 primers as described in Millet et al. (20)
and generated a 462 bp DNA fragment.
|
Real-time PCR
The ovarian normal, cyst and cancer tissue samples were collected from patients at the Department of Obstetrics and Gynecology, Gynaecologic 0ncology Unit, of the University of Torino Medical School. Samples were immediately frozen in liquid nitrogen after surgery. Histological examinations were performed on stained sections, and epithelial cellularity was evaluated. Only samples with an epithelial contingent of at least 50% were retained for further analysis. For normal ovaries, the presence of surface epithelium was checked after staining.
RNAs were extracted from tissues using the RNeasy® kit (Qiagen). 1 µg total RNA was reverse transcribed with 50 units of Expand RT (Roche) using oligodT primers (Sigma, St. Louis, MO, USA). First strand cDNAs were purified with Qiaquick columns (Qiagen). Real-time PCR was performed using SYBR Green mix and a LightCycler apparatus (Roche) with oligonucleotides (Table 1)
at 0.5 µM final concentration. Thermal cycling was initiated with a first denaturation step of 3 min at 95°C. Subsequent PCR conditions involved 45 cycles as follows: 95°C for 1 s, 60 or 65°C (HPRT and CHRD, respectively) for 10 s and 72°C for 15 s. PCRs were performed in triplicate and the data were analyzed by the "fit points" method with the Roche Molecular Biochemicals software package. The amount of mRNA for each sample was calculated from the standard curve performed in the same experiment on serial dilutions of a purified PCR product corresponding to the amplified fragment. The results are expressed as the ratio of the amount of CHRD with HPRT mRNA.
In situ hybridization experiments on rat ovaries
Six- to 12-month-old rat ovaries were fixed for 16 h at 4°C in 4% paraformaldehyde, washed in PBS, equilibrated in PBS with 30% sucrose overnight at 4°C, then embedded in OCT medium (Tissue Tek®) and frozen at –80°C. Dehydrated 14 µm cryosections were overlaid with a solution of 10 µg/mL proteinase K for 6 min, washed with PBS, fixed again in 4% paraformaldehyde for 20 min and dehydrated with graded methanol/PBS solutions. A PCR human CHRD fragment (see primers in Table 1
) cloned into pGEM®T Easy vector (Promega) was used for probe synthesis. DIG-labeled anti-sense and sense cRNA were prepared by in vitro transcription of the linearized plasmid using Sp6 and T7 polymerase (Promega), respectively. In situ hybridization experiments were performed mainly according to Wilkinson (35)
with probes at 50 ng/µL in 50 µL hybridization solution for 16 h at 60°C. After washing with PBS, sections were incubated with mouse anti-DIG antibodies (0.2 µg/mL in 0.1% PBT; Roche) for 2 h at room temperature in the presence of 5% horse serum to lower nonspecific signal. After washing with PBT (PBS with 0.1% Triton X 100), sections were incubated with a secondary biotinylated anti-mouse antibody (1/1000 dilution) for 16 h at 4°C and washed with PBT, 0.3% Triton for 2 h. The vectastain detection system (Vector, Burlingame, CA, USA) was then used for visualization. Coloration was obtained with a solution of BCIP/NBT in the presence of 0.024% levamisol (w/v) and the revelation was stopped by washing the sections with PBS when the anti-sense probe signal was sufficient. Sections were immersed overnight in 4% paraformaldehyde, dehydrated in methanol and mounted with Aquamount (BDH) under a coverslip. Tissues hybridized with the sense probe were used as negative control. Figure 5
shows pictures obtained with ovaries from one female.
|
Cell lines and human OSE cell culture
BG1, OVCAR3, SKOV3, PEO4 and PEO14 human ovarian cancer cell lines were used in this study. BG1 (36)
was kindly provided by Dr. Welander (Emory University, Atlanta, GA, USA), PEO4 and PEO14 cells (37)
by Dr. Langdon (Hospital of Edinburgh, UK). BG1 cells were cultured in McCoy’s medium (Cambrex) supplemented with 10% fetal calf serum (FCS, Invitrogen, San Diego, CA, USA), whereas PEO14, PEO4, and OVCAR3 cells were cultured in RPMI-1640 (Invitrogen) supplemented with 10% FCS and SKOV3 in Dulbecco’s modified Eagle medium/F12 (DMEM/F12, Invitrogen). Chinese hamster ovary (CHO) (ATCC) cells secreting the CHRD protein were used in this study. The CHRD cDNA was tagged with a c-myc epitope and cloned in the pcDNA3 mammalian expression vector (Invitrogen): In brief, we first replaced the 5'-noncoding region of the full-length human CHRD cDNA (GenBankTM accession no. AF209928) (20)
with a short DNA fragment containing the CAAA sequence as translation start site upstream of the ATG initiation codon. A c-Myc tag (EQKLISEEDL), recognized by antibody 9E10 (Roche Applied Science), was then introduced by PCR upstream of the CR1 domain of the protein (after amino acid 40). Stably transfected CHO cells with this construct were shown to secrete the c-Myc-tagged CHRD, which was demonstrated to antagonize BMP2 in a cell differentiation assay (Noël et al., unpublished results). Cells were cultured in DMEM supplemented with 10% FCS. CHO cells secreting Xenopus Noggin (38)
is a gift from Dr. Harland (University of California, Berkeley). Murine mesenchymal C3H10T1/2 cells secreting human BMP2 (39)
is a gift from Dr. Gazit (Hebrew University-Hadassah Medical Center, Jerusalem, Israel) and were cultured in DMEM supplemented with 10% FCS. The human OSE cell culture conditions are from Auersperg et al. (40)
. Briefly, normal ovarian tissue was obtained in the operating room from consenting donors. OSE cells were brushed off the surface of the ovaries and placed in M199:MCDB 105 (1:1) medium (Invitrogen) supplemented with 20% FCS. The OSE nature of the cells in the primary culture was assessed by immunohistochemistry, which showed positive staining for cytokeratin using a pan-cytokeratin antibody (KL1), for vimentin, typical of the mesothelial nature of OSE, and negative staining for factor VIII for absence of endothelial cells.
Migration and Matrigel invasion assay
One day before transfection, the cells were trypsinized and spread in a 6-well plate at 80% confluency. Transfections were then performed overnight with Lipofectamin 2000 (Invitrogen). Typically, 0.55 µg DNA of the pcDNA3-CHRD vector or the pcDNA3 control plasmid were cotransfected with 0.55 µg of pGl3 vector coding for firefly luciferase (Promega). For the migration assay, a suspension of 3 x 105 transfected cells in 200 µL of culture medium supplemented with 1% FCS was layered in triplicate on a polyethylene terephtalate (PET) membrane in the upper compartment of a cell culture insert (BD) placed in a 24-well plate. For the invasion assay, transfected cells were layered on the insert precoated with Matrigel basement membrane (BD). For this, 30 µg of Matrigel was diluted in 100 µL of culture medium, then coated on the PET filter and left to polymerize for 1 h at 37°C with 5% CO2 before adding the cells. Plates were then incubated for 24 h at 37°C with 800 µL of culture medium supplemented with 1% FCS or 5% FCS in the lower compartment to detect random motility or chemotaxis, respectively (41)
. In parallel to transwell seeding, aliquots of 105 transfected cells were plated in triplicate in a 24-well plates and cultured in the same conditions to quantify total transfected cells. After 24 h incubation, the cells remaining on the upper surface of the inserts were removed using a cotton swab, and those remaining on the lower surface of the PET membrane were lysed, in parallel to the cells used as control in the 24-well plate. Luciferase activities were determined on each cell sample by measuring luminescence. The % of migrating or Matrigel invading cells was calculated as the ratio of luciferase activity of invasive cells to luciferase activity of control cells (corresponding to total activity in the 24-well plate) x 100.
Cell adhesion assay
After trypsinization, BG1 or PEO14 cells were resuspended in the conditioned medium of CHO cells secreting CHRD or not, diluted 1:1 with culture medium supplemented with 10% FCS. 105 cells were then added in triplicate to 24-well plates and allowed to adhere for 0 to 60 min at 37°C, 5% CO2. After gentle rinsing with PBS, adherent cells were quantified by evaluating the mitochondrial deshydrogenase enzymatic activity: the cells were incubated in a 3 (4,5-dimethylthiazol-2-yl) 2,5-diphenol tetrazolium bromide (MTT) solution (0.5 mg/mL) for 4 h before dissolution in 200 µL of an ethanol-DMSO 1:1 solution, and absorbance was read at 540 nm. To confirm the results, experiments were repeated and cell number estimations were obtained by measuring the DNA content by the diaminobenzoic acid (DABA) fluorescence assay (not shown). Conditioned medium of CHO cells secreting Noggin was used in the same conditions on PEO14 cells.
Modulation of BMP signaling pathway of cancer cells and Western blot
Suspensions of 106 PEO14 cells in 3 mL of culture medium supplemented with 10% FCS were spread in 6-well plates and left to seed for 90 min. 1.5 mL of culture medium was then replaced by the same volume of conditioned medium from CHO or C3H10T1/2 cells secreting either CHRD, Noggin or BMP2 proteins. After 4 h incubation at 37°C, cells were washed briefly with PBS and proteins were extracted by three freeze-thaw cycles in lysis buffer (50 mM HEPES pH 8, 150 mM NaCl, EDTA 1 mM, EGTA 7.5 mM, 0.1% NP-40, glycerol 10%, 10 mM sodium glycerophosphate, 1 mM NaF, 1 mM NaVO3, and 1 mM DTT) supplemented with protease inhibitors (Complete, Roche) and centrifugation at 10,000 x g for 15 min. For Western blot analysis, proteins were fractionated by SDS-PAGE and transferred to PVDF in TBST (20 mM Tris pH 7.6, 137 mM NaCl, 0.1% Tween 20). The membrane was blocked in TBST with 5% non-fat milk powder, washed with TBS-T and probed with a phospho-Smad1/5/8 antiserum (42)
, a generous gift from Dr. ten Dijke (Ludwig Institute for Cancer Research, Uppsala, Sweden). After washing, the membrane was incubated with horseradish peroxidase-conjugated anti-rabbit immunoglobulin (Amersham, Arlington Heights, IL, USA). A chemiluminescent substrate (ECL detection system, Amersham) was used for visualization.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
, 43)
. Gains of chromosome 3q and in particular of its distal portion are common in ovarian tumors (44)
. We therefore carried out a Southern analysis of the CHRD genomic region in 386 human ovarian tumors (Fig. 1
). The CHRD signal was found to be increased relative to reference probes, indicating a low to moderate copy number increase (CNI) in 33 tumors (8.5%). Analysis of clinico-pathological data showed a higher prevalence of CHRD CNI in high-grade tumors (14/17 tumors of grade 3 or 4) compared with low grade (3/17). No association was found with tumor type or stage; 6/15 ovarian tumors showing CNI at CHRD presented amplification of C-MYC, a gene frequently amplified in ovarian tumors (45)
. We assessed whether CHRD expression was altered as a consequence of copy number changes by monitoring CHRD mRNA expression levels by real-time PCR in another group of 47 macrodissected epithelial ovary tumors (for which tissue material was available). We found 3/47 (6%) CHRD-positive tumors with a CHRD mRNA level significantly higher than the overall tumor average (Fig. 2
). Therefore, we monitored CHRD genomic content by fluorescent in situ hybridization (FISH) on tumor-touch preparations in these three tumors along with three tumors with low CHRD mRNA levels. CHRD showed increased copy numbers in all tumors analyzed, regardless of its expression level (Fig. 3
A, B). A mild to moderate CHRD copy number increase was thus observed in all six tumors analyzed by FISH, but no correlation could be established with mRNA expression levels.
|
|
|
CHRD is underexpressed in ovarian cancer tissue or cell lines compared with cysts and normal ovaries
CHRD expression was found at the same level in normal ovaries (n=8) and ovarian cysts (n=6), as assessed by real-time RT-PCR analysis (Fig. 2)
. In the carcinoma tumor group (n=47), excluding the three CHRD-positive tumors, CHRD was either undetected or expressed at a very low level in comparison to the group of normal and cystic ovaries (n=14). On average, the CHRD/HPRT ratio was 15-fold lower in carcinoma vs. normal and cystic ovaries (Fig. 2)
. No correlation was found between CHRD level and histological subtypes, stage or grade in the bulk of the carcinoma.
To establish whether or not the very low CHRD expression in tumors was conserved in cancer cell lines, we used analytical RT-PCR to evaluate its expression in five different ovarian cell lines derived from human ovarian adenocarcinomas. Indeed, CHRD was absent or barely detectable in the cell lines (Fig. 4
). We also analyzed BMP2, -3, -4, -6, -7, -15, and GDF9 since their expression has already been described in different cell types in rat ovaries (11
, 13
, 14)
. The cell lines were found to express different combinations of BMPs, but they all consistently expressed BMP4 and GDF9.
|
CHRD is expressed in normal human and rat ovary surface epithelium
The absence of CHRD expression in most tumors could be a feature of transformed epithelial cells acquired during the tumorigenic process or reflect constitutively low or absent CHRD expression in normal OSE. Therefore, we addressed the question of CHRD expression in normal ovarian epithelium. First, we analyzed primary cultures of normal human OSE and, in contrast with cancer cell lines, we found CHRD mRNA in these cells (Fig. 4)
. Conversely, in cultured human OSE, we found the same BMP4 expression status as in cancer cell lines. BMP2, -6, and -15 were also detected at low levels, whereas GDF9 (which is expressed in all cancer cell lines) was not found in OSE.
Second, to determine the Chrd expression pattern in normal ovaries, we examined its expression in rat ovaries by in situ hybridization. Chrd mRNA expression was found in OSE, but almost exclusively limited to corpus luteus (Fig. 5
). Expression was consistently observed in luteal cells and more rarely in granulosa cells of antral follicles. The results of these experiments most likely indicated that Chrd is not continuously expressed throughout the epithelium during the cycle but induced in OSE surrounding post-ovulatory follicles.
CHRD reduces haptotactic motility and invasiveness of ovarian cancer cells
The low CHRD mRNA expression noted in ovarian tumor tissues and cancer cell lines as compared with normal ovary and OSE led us to hypothesize that this gene could influence epithelial cell behavior during ovarian tumor progression. We thus tested the effect of CHRD on the behavior of cancer cell lines in migration and invasion assays. We focused on BG1 and PEO14, two cell lines for which migration and invasion properties have already been described (46)
.
Quantitative analyses of the effect of CHRD on cell migration were performed using a Boyden chamber-type assay (Transwell) after transient cotransfection of the cells with an expression vector coding for CHRD (pcDNA3-CHRD) or with the vector alone, together with a luciferase-encoding pGl3 vector as a marker of transfected cells (41)
. We used the same concentrations of serum on each side of the membrane onto which cells were seeded in order to evaluate random motility by measuring the amount of luciferase activity in cells that had moved to the lower surface of the membrane. A different serum concentration in the upper (1%) vs. lower (5%) compartment was used to determine whether CHRD would have an effect on chemotaxis, with the serum being the chemoattractant here. We found that BG1 and PEO14 had similar migration rates when exposed to identical serum conditions, as described previously (46)
. For both cell lines, serum-induced cell migration conditions enhanced by 3-fold the number of cells migrating to the lower compartment (Fig. 6
A). Following CHRD expression, migration was reduced by 40 to 60% for either cell line in both serum conditions (Fig. 6A
). Similar results were obtained in migration experiments using conditioned medium from CHRD-expressing cells as a source of CHRD protein (data not shown).
|
Invasion assays were performed in the same conditions as the migration assays, except that the Transwells were coated with Matrigel, a solubilized basement membrane preparation, to mimic basal lamina. We observed that the % of Matrigel-invading cells was approximately half for BG1 cells compared with PEO14 cells. For both cell lines, 2-fold more cells passed through the PET-Matrigel barrier in the serum-induced cell migration conditions. Similar to migration, invasiveness of Matrigel by CHRD-expressing cells was reduced by 40% to 60% by comparison with the control cells not exposed to the BMP antagonist (Fig. 6B
). From these results, we thus conclude that CHRD significantly reduces both motility and invasiveness when expressed in ovarian carcinoma cells.
CHRD enhances adhesion of PEO14 and BG1 cells to the support
Since we have shown that CHRD could behave as an oncosuppressor by decreasing cell migration and invasion toward Matrigel, we wanted to establish whether CHRD influences cell adhesion, one of the key factors in ovarian tumor progression (1)
. Microtiter well adhesion analysis indicated that after 10 min seeding, almost 2-fold more cells were adhering to the plastic well when exposed to culture medium containing CHRD in comparison to control cells without CHRD (Fig. 6C
). This effect of CHRD on cell adhesion diminished over the time course, and the difference between CHRD-exposed cells and control cells was 15% after 1 h of cell seeding. Over longer time periods, the number of cells attached to the support remained slightly higher in the presence of CHRD (not shown).
Noggin inhibits BMP signaling of PEO14 cells and enhances their adhesion to the support
To gain insight into the mechanism by which CHRD influences cellular behavior of cancer cells, we first investigated whether the ovarian cancer cells have a functional BMP signaling pathway and whether the addition of CHRD, Noggin or BMP2 to the cells could modulate this pathway. This was done by exposing PEO14 cells to conditioned medium containing either BMP antagonist or BMP2, followed by Western blot analysis of protein extracts using a phosphoSmad1/5/8 antiserum to reveal activation of the BMP pathway (42)
by endogenous or exogenous BMPs that would be present in the serum. The BMP pathway was found to be activated in the presence of conditioned medium (Fig. 7
A, lane 1), and slightly reduced when CHRD or Noggin were present in the medium (Fig. 7A
, lanes 2 and 3). Conversely, addition of BMP2 enhanced the phosphoSmad signal (Fig. 7A
, lanes 4 and 5). In a second step, we observed that adhesion of PEO14 cells to the support was enhanced by Noggin to a similar extent as for CHRD (Fig. 7B
). Altogether, these results strongly suggest that the effect of CHRD on adhesion can be at least partly explained by the neutralization of a BMP activity.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
, 48)
. At 3q26, two candidate oncogenes, PIK3CA (49
, 50)
and e1F-5A2 (51)
were identified for their CNI, overexpression and effects on cell proliferation and apoptosis. Similar to PIK3CA, we found low-level CHRD amplification. CNI at CHRD prompted us to verify whether its expression pattern could be modified in relation to copy number. We found that 6% of the carcinomas (3/47) had higher CHRD expression than the average level in cysts and normal tissue. FISH analysis of these tumors, along with CHRD nonexpressing tumors, revealed no correlation between the degree of genomic amplification of CHRD and its mRNA expression level, which indicates that CHRD is not selected for increased expression by copy number gains. A similar situation has been noted for many breast cancer genes whose mRNA level is not correlated with copy number changes (52)
. Significantly, the overall carcinoma population had a 15-fold lower level of CHRD expression in comparison to normal and cystic tissue, which differentiates CHRD from neighboring oncogene candidate genes in 3q26, PIK3CA, and e1F-5A2. Our functional studies provided further evidence that CHRD does not act as an oncogene. Note that normal and cystic ovaries are more heterogeneous tissues than carcinomas and may thus contain CHRD expressing cells other than OSE cells. However, the significance of this lack of CHRD in most carcinomas was boosted by the fact that CHRD was found to be expressed in normal human cultured OSE and hardly detectable in cancer cell lines. This indicates that the difference of CHRD expression in normal tissues and tumors is not solely due to the different proportion of OSE in these tissues. As malignant tumors generally arise in OSE, we reasoned that a loss of CHRD expression in this tissue might favor the neoplasic process.
The role of BMPs in the epithelium has not yet been evaluated, but is suggested by the strong BMP3b, -4, and -6 expression noted in rat OSE (13)
and by BMP signaling in cultured human OSE cells (14)
. In these cells, we detected BMP2, -4, and -6 mRNAs. BMP2 was not found in rat OSE on ovary sections (13)
, which likely reflects a difference between the two species, but we cannot exclude that BMP2 expression is induced in human OSE cells upon culture.
We found CHRD expression in cultured human OSE and it still has to be determined if its expression is solely limited to a subset of cells of this ovary compartment in humans. It is considered that the response of OSE cells to explantation in culture mimics their normal response to ovulatory rupture (1)
, which suggests that CHRD expression in human OSE may be restricted to ovulatory sites, i.e., sites of OSE regeneration. Indeed, from our in situ hybridization experiments it appears that in rat ovaries, Chrd mRNA is most likely induced in the OSE restricted to corpora lutea, as well as in underlying luteanized cells of these post-ovulatory follicles. BMP3 and -6 appear to be constitutively expressed throughout OSE during the rat ovarian cycle, whereas BMP4 expression is only induced in a subset of OSE cells covering the ovulated follicle and newly formed corpus luteus-I, as well as in theca cells underlying the OSE of dominant follicles (13)
. This gave rise to the hypothesis that BMP4 may participate in the OSE repair process after ovulation (13
, 14)
. The Chrd expression pattern in rat OSE thus roughly overlaps that of BMP4. We propose that Chrd induction could then be part of a regulatory mechanism that would stall BMP activity at the site of OSE healing — a mechanism that would be lost in carcinogenesis. This hypothesis is supported by the lack of CHRD in ovarian tumors and cell lines as compared with normal tissue and OSE.
We detected CHRD expression in granulosa cells in follicles, where it could regulate the activity of at least four BMPs during follicle growth: BMP2 strongly expressed in granulosa, BMP4 and –7 in covering theca cells (13)
, and BMP6 coming from oocytes. Ten percent of ovarian cancers are granulosa cell tumors (53)
. We thus found CHRD expression in both normal OSE and granulosa cells, the two main cell types from which ovarian cancers originate.
We indirectly tested the hypothesis that the absence of CHRD from OSE after ovulation may be a carcinoma promoting factor by analyzing CHRD effects on the motility and adhesion of several ovarian cancer cell lines. These characteristics are of particular interest, as ovarian cancer cells have the ability to disseminate throughout the peritoneal cavity from the ovary, which is the major cause of death of this type of cancer. More generally, loss of adhesion and migration are two properties acquired by cells escaping from a primary tumor to metastatic sites. Consistent with the situation in carcinomas, we found no CHRD expression in cancer cell lines, whereas BMP4 and GDF9 were found in all the cell lines we analyzed. It is noteworthy that mRNAs for BMP4 were found in primary cultured human OSE, but not for GDF9, which is not known to be a Chrd target. These results are in accordance with those of Shepherd and Nachtigal who documented BMP4 expression in OSE and EOC cells (14)
.
This CHRD negative/BMP positive general expression profile was appropriate for testing the effects of CHRD on cancer cell behavior. Indeed, we found that exposure of BG1 and PEO14 cells to CHRD reduced motility and invasion through the extracellular matrix by 2-fold. Inhibition of BMP activity by Chrd or Noggin has been shown to account for opposing epithelial cell migration in various situations during vertebrate development. A recent report shows that BMP4 and -7 are increased in malignant melanoma and that Chrd expression reduces the migratory and invasive properties of melanoma cell lines into which it was introduced (54)
.
Exposure to CHRD had the effect of enhancing adhesion of BG1 and PEO14 cancer cells to the support. We have shown that BMP signaling is functional in PEO14 and is reduced by exogenous CHRD and Noggin. Furthermore, this latter BMP antagonist enhances adhesion of the cancer cell line, which supports the hypothesis that CHRD influences cancer cell behavior by antagonism of BMP signaling. These results are in apparent contradiction with those of Shepherd and Nachtigal, who reported increased adherence of primary EOC cells after long-term BMP4 treatment (14)
. One possible explanation is that treatment of the cells with exogenous BMP4 may induce CHRD expression in a negative feedback loop. BMP antagonists were already shown to be up-regulated by the BMP ligands they antagonize, such as Gremlin in granulosa cells, where its expression is induced by BMP4 (17)
. The effect of CHRD on adhesion could be explained by neutralization of a BMP activity that usually down-regulates expression of an adhesion molecule. Nevertheless, we cannot exclude that CHRD influences cell adhesion by directly interacting with ECM components. Noteworthy, while screening for extracellular surface proteins interacting with sog/Chrd (sog is the ortholog of Chrd in Drosophila) (55)
, two groups have recently identified genetic interactions or binding of the antagonist with integrins (56
, 57)
. Chrd thus appears as a possible ligand for integrins, which are expressed in OSE (58)
. It is becoming clear that enhanced expression of integrins in neoplastic cells favors their proliferation, survival and migration during the multistep progression from tumor growth to metastasis (59)
. Ovarian cancer cells express high levels of ß1 integrin, and laminins have been shown to promote migration of ovarian cancer cells through ß1 integrin (60
, 61)
. Binding of CHRD to integrins might thus partially explain the effects we observed on cancer cell line motility and adhesion, in a mechanism that would not require interactions of CHRD with BMP ligands, but maybe by competition with or modification of the ß1 integrin-laminin interaction.
By its lack of expression in ovarian carcinomas as compared with noncancerous ovarian tissue and normal OSE, and by its effects of reducing motility and enhancing adhesion of cancer cell lines, CHRD emerges as a potential inhibitor of cancer progression in the ovary. It will be of great interest to put CHRD potential tumor suppressor behavior to the test of in vivo mouse cancer models.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Received for publication April 5, 2005. Accepted for publication October 3, 2005.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  B. L. Theriault, T. G. Shepherd, M. L. Mujoomdar, and M. W. Nachtigal BMP4 induces EMT and Rho GTPase activation in human ovarian cancer cells Carcinogenesis, June 1, 2007; 28(6): 1153 - 1162. [Abstract] [Full Text] [PDF]
|
|
