Article Text
Abstract
Background Gastric cancer (GC) is a highly prevalent disease, being the fourth most common cancer and the second leading cause of cancer-associated deaths worldwide. Although many genes have been implicated in its development, many cases remain genetically unexplained. Hence, there is an urgent need to find new disease-related genes.
Methods A transgenic Drosophila model was used to screen for novel genes putatively involved in GC. The authors evaluated the expression of the most interesting candidates in GC cell lines and primary tumours by semi-quantitative reverse transcription PCR, dissected the molecular mechanisms responsible for the deregulation of the most relevant one, and analysed its functional role in vitro and in a chicken embryo model.
Results Six candidate genes were identified, of which cytoplasmic polyadenylation element binding protein 1 (CPEB1) was downregulated in all GC cell lines and in 11 of 12 primary GC tumours. The pivotal CPEB1 promoter CpG site was determined, and it was found that methylation at this 79th CpG site was associated with CPEB1 silencing in GC cell lines and primary tumours. It was also discovered that methylation of this site was significantly more prevalent in diffuse type GC (p=0.007) and in cases with lymph node metastases (p=0.042). In vitro, CPEB1 impaired invasion. Its antiangiogenic role was also discovered, which was associated with downregulation of MMP14 and VEGFA.
Conclusions The first evidence of CPEB1 involvement in GC is presented, along with the molecular mechanism underlying the regulation of its expression and its potential role in invasion and angiogenesis.
- Methylation
- metastases
- E-cadherin
- cancer
- CPEB1
- cadherins
- angiogenesis
- cancer genetics
- RNA expression
- cell biology
- cell adhesion molecules
- colorectal diseases
- colorectal carcinoma
- colorectal cancer screening
- mutation screening
Statistics from Altmetric.com
- Methylation
- metastases
- E-cadherin
- cancer
- CPEB1
- cadherins
- angiogenesis
- cancer genetics
- RNA expression
- cell biology
- cell adhesion molecules
- colorectal diseases
- colorectal carcinoma
- colorectal cancer screening
- mutation screening
Significance of this study
What is already known about this subject?
Many genes have been pinpointed to be important for the development of gastric cancer (GC).
However, for a large proportion of cases, the genetic events responsible for disease progression are still undetermined.
Two recent reports have shown CPEB1 downregulation in ovarian cancer and multiple myelomas—in the latter, through methylation.
What are the new findings?
Using a Drosophila model, this work has identified CPEB1 as a novel GC-related gene. We show that it is downregulated in more than 90% of the primary GC tumours analysed in terms of mRNA. We also found decreased CPEB1 levels in colorectal and breast cancer cell lines, suggesting a broader implication of CPEB1 in cancer.
In addition, we present the first evidence for CPEB1's silencing mechanism in GC, through promoter methylation, particularly at a pivotal CPEB1 promoter CpG site.
We have uncovered an anti-invasive role for CPEB1 in vitro. In addition, by analysing the effect of CPEB1 overexpression in the formation of new blood vessels in the chick chorioallantoic membrane assay, we have found a new role for CPEB1 as a putative angiogenic modulator. Accordingly, two main angiogenic effectors (VEGFA and MMP14) were downregulated after CPEB1 overexpression.
How might it impact on clinical practice in the foreseeable future?
This study has important clinical implications in both GC and many other types of cancer also affected by CPEB1 silencing, with respect to the value of CPEB1 as either a tumour or prognostic marker.
Our findings may also have an impact on GC therapeutics, given that we have successfully reversed CPEB1 epigenetic inactivation by demethylating agents.
Introduction
Gastric cancer (GC) is still the second leading cause of cancer-related deaths worldwide, despite a decrease in incidence during the last century.1 ,2
Many genes have been associated with the development of the distinct histological and molecular subtypes of gastric carcinomas.3 CDH1 (the gene that codes for epithelial cadherin (E-cadherin)) is an example of a crucial tumour suppressor gene in gastric carcinogenesis, which, when its function is lost, promotes invasion, migration, epithelial–mesenchymal transition and metastasis, being involved in the initiation and progression of both hereditary and sporadic forms of GC.4–11
However, a large proportion of GC cases remain genetically unexplained.12 Therefore the identification and characterisation of new GC-related genes is crucial.
Simple animal models are being used to unveil novel human cancer pathways. Drosophila melanogaster (fruit fly) has been useful in identifying candidate genes for neurodegenerative disorders,13 muscular dystrophies,14 infectious15 and cardiovascular disease,16 as well as cancer.17 ,18 The fruit fly has been widely used because of its small size, short generation time, low cost and easy genetic manipulation, and, most importantly, because of conservation of basic mechanisms and signalling pathways even at the molecular level between flies and humans.17 ,18
Gain-of-function screening is an easy and rapid alternative to identifying interactor genes through their misexpression, which is especially advantageous for detecting genes missed by loss-of-function approaches because of being expressed at low levels, in a few cells, or transiently during development.19 One of the most popular places to conduct such screens is the adult fly eye, since it is not required for viability or fertility and is easy to score.20
Here, we used a previously described Drosophila model21 ,22 to functionally screen for novel GC-related genes, focusing on those showing differential interaction with two causative GC mutations (affecting the extracellular (A634V) or intracellular (V832M) domain of human E-cadherin (hEcad)). This was the way chosen to create a GC-sensitised background.
The collection of flies used was kindly provided by José Felix de Celis. These fly lines were generated by random P-GS (Gene Search) insertion. The insertions, the gain-of-function mutant phenotype of which affected wing vein formation, were subsequently selected and mapped by Molnar and co-workers.23 They observed that most of the genes affected were also required in other developmental processes controlled by similar interaction networks. This GS library is extremely useful because the process of growth and patterning of the fly wing involves a considerable number of genes implicated in a wide range of cellular processes that are closely associated with cancer development, from regulation of adhesion and extracellular matrix to control of cell death.
Methods
Drosophila strains and genetic manipulations
The Gal4/UAS system was used to target gene expression in Drosophila tissues. This genetic tool relies on two components: Gal4, a yeast transcriptional activator, the ectopic expression of which is driven by tissue-specific promoters; UAS (upstream activating sequences) target genes, the expression of which is Gal4 dependent because they bear Gal4-binding sites. The two elements, initially separated into two distinct transgenic lines, are brought together in a simple genetic cross. In the progeny, the UAS responders will be activated in those cells where Gal4 is expressed, in a pattern that mimics that of the Gal4 promoter.24
For this, the Gal4 coding sequence is fused to a P-transposable promoter and injected into flies for transformation. The Glass multimer reporter (GMR)-Gal4 line used was constructed using pGMR-1, which contains five copies of a Glass response element from the Rh1 rhodopsin gene and drives expression in all the fly differentiating retina cells (behind the morphogenetic furrow).25
UAS-green fluorescent protein (GFP)26 was used as expression control. GMR-Gal4; UAS-hEcad, GMR-Gal4; UAS-hEcadA634V and GMR-Gal4; UAS-hEcadV832M were generated as tester strains.22 Females from the tester genotypes were crossed with males from a collection of GS lines previously selected as affecting growth and patterning of the fly wing when misexpressed (provided by José Felix de Celis).23 These GS lines were created by random insertion of a P-element that contains two UAS at both 5′ and 3′, oriented to target the expression of the genes outwards.20 Hence, they allow assay of more genes at the same time than other P-element insertion lines with only one UAS sequence.
Flies were maintained at room temperature, and the eye phenotype of the progeny was analysed. When the resulting products affected biological systems, phenotypes that could be easily scored were induced. Enhancers or suppressor GS lines were then selected and rescreened to validate the interaction, because an interacting GS line identifies as potential targets the two genes located in the immediate vicinity of the insertion point.19 GS insertion sites had been previously mapped by the de Celis laboratory. Eyes were examined under a Leica compound microscope, and digital images were processed with Adobe Photoshop.
Cell lines and tumour samples
To determine the expression levels of the six human homologues (MEF2A, ENSG00000130177/PAM, ENSG00000145730; SLC17A5, ENSG00000119899/BMP2, ENSG00000125845 and CPEB1, ENSG00000214575/CDC16, ENSG00000130177) of the gene pairs identified in the Drosophila screen as candidates in cancer-derived cell lines, we used nine GC (IPA220, MKN28, NCI-N87, GP202, AGS, MKN45, KATOIII, SNU638 and SNU1), seven colon cancer (LS174T, LOVO, HCT116, HT29, SW480, DLD1 and HCT15) and seven breast cancer (SKBR3, BT474, HS578T, MDAMB231, MDAMB468, BT20 and MCF7AZ) cell lines. SNU1, SNU638 and all breast cell lines were grown in Dulbecco's modified Eagle's medium (Gibco–Invitrogen, Barcelona, Spain). The other cell lines were maintained in RPMI 1640 (Gibco–Invitrogen).
To test the effect of CPEB1 in vivo, we used a series of primary GCs (n=43), collected in the Department of General Surgery and Surgical Oncology of the University of Siena (Italy). The study protocol was reviewed and approved by the appropriate ethics committees. All samples were obtained with informed consent and in compliance with the Helsinki Declaration (http://www.wma.net/e/policy/b3.htm). Written consent was obtained from every patient affected by primary GC. These consents were archived at the Department of General Surgery and Surgical Oncology (Hospital Santa Maria alle Scotte, Siena, Italy). In accord with the Italian D.L. n. 196 of the 30 June 2003, namely the Privacy Law, we cannot distribute personal information such as name, health condition and specific information on patients. Therefore information has been sufficiently anonymised, meaning that neither the patient nor anyone else can identify the patient with certainty.
DNA and RNA were isolated using Gentra Systems (Minneapolis, Minnesota, USA) and Qiagen RNeasy Kit (Qiagen, Hilden, Germany). SuperScript II RT and random primers (Invitrogen, Carlsbad, California, USA) were used for complementary DNA (cDNA) synthesis.
Expression of GC candidate genes in cancer-derived samples
cDNA (50 ng) from each GC line and commercially available RNA derived from normal stomach (Human Stomach Total RNA; Ambion, Austin, TX, USA) were amplified by semi-quantitative reverse transcription PCR to test the expression of the following candidate genes: MEF2A, PAM, SLC17A5, BMP2, CPEB1 and CDC16 (primer sequences are summarised in online supplementary table 1). GAPDH was amplified in combination with each of the candidate genes. PCR products were run in a 2% agarose gel, and bands were quantified using Quantity One 4.6.8 Software (Bio-Rad, Amadora, Portugal). The expression level of each candidate gene was normalised to the expression presented by normal stomach.
Quantitative real-time PCR (qRT-PCR) was performed to determine messenger RNA (mRNA) expression levels of CPEB1 in both primary GC and cell lines. PCRs were prepared using TaqMan Universal PCR Master Mix, No AmpErase UNG and Inventoried TaqMan Gene Expression Assay (Hs00229015_m1; Applied Biosystems, Foster City, CA, USA) for CPEB1 and GAPDH (Human GAPD Endogenous Control; Applied Biosystems) quantification. CPEB1 expression levels were normalised to the mean expression level of GAPDH. For cell lines, technical and biological triplicates were performed. Regarding primary tumours, three technical replicates were performed. For normal breast and colon controls, commercially available normal RNA was used (MVP Breast Human Total RNA and Human Colon Total RNA from Stratagene, La Jolla, California, USA and Ambion, respectively).
An identical approach was used to evaluate MMP14 and VEFGFA (Hs00237119_m1 and Hs00900055_m1, respectively).
CPEB1 mutation and methylation analyses
CPEB1 mutation analysis was performed in nine GC cell lines and in a series of 12 primary GCs. All CPEB1 exons annotated in the Ensembl database (ENSG00000214575) were analysed using intronic flanking primers designed for each region. PCRs were performed using 50 ng DNA. The amplified products were analysed by direct sequencing. At least two independent PCRs were performed for every sample. Primer sequences and PCR conditions are available upon request.
Methylation analysis was performed in the same series of samples (nine GC cell lines and in a series of 12 primary GCs). CpG island searcher program was used to identify the CPEB1 promoter CpG island (http://cpgislands.usc.edu/). Eighty-five CpG sites that localise between coordinates chr15: 83316406 and 83317074 were analysed in GC cell lines, and only 45 at the 3′-end of this region were evaluated in primary tumours because of limitations of biological material. From each sample, 200–300 ng DNA was treated using an EpiTect Bisulfite Kit (Qiagen, Hilden, Germany). Unmethylated cytosines were converted into uracil, and methylated cytosines were left unmodified. Bisulphite-treated DNA, from white blood cells, was in vitro methylated with MssI DNA methyltransferase and used as a positive control. These regions were amplified from bisulphite-treated DNA using flanking primers lacking CpG sites (sequences available upon request) and analysed by direct sequencing. After identification of the critical site of methylation within the CpG island of CPEB1, 30 additional primary carcinoma cases were included in the analysis, making in total a series of 43 gastric primary carcinomas.
5-Aza-2′-deoxycytidine treatment
At 24 h after seeding KATOIII, MKN45 or GP202 cell lines on to six-well plates, 5-aza-2′-deoxycytidine (5-Aza-dC; Sigma-Aldrich, St Louis, Missouri, USA) dissolved in dimethyl sulphoxide (DMSO; Sigma) was added at a final concentration of 5 μM and replaced every 24 h. Control cells were treated only with DMSO. After 5 days of treatment, expression and methylation analysis was performed as described previously.
Cell transfection
KATOIII GC cells were transfected in six-well plates with plasmid DNA (CPEB1-GFP cloned in pEGFP-N1 or the pEGFP-N1 empty vector, Mock-GFP) mixed with lipofectamine 2000 reagent and used at a concentration of 2 μg/ml. After 12 h of incubation, Mock-GFP- or CPEB1-GFP-transfected KATOIII cells were resuspended to a concentration of 1×106 cells in a 10 μl suspension, for the in vivo chorioallantoic membrane (CAM) angiogenesis assay. For RNA and protein processing or immunocytochemistry, samples were collected at ∼72 h after transfection. For invasion assays, the protocol was started at 48 h after transfection.
Matrigel invasion assay
Invasion assays were performed using Biocoat Matrigel Invasion Chambers (BD Biosciences, Heidelberg, Germany). At 48 h after transfection, inserts containing a polyethylene terephthalate membrane with 8 μm pores and coated with matrigel basement membrane matrix were placed inside a 24-well plate. Then 500 μl of a culture medium suspension with 1×105 cells/ml was added to the top of the chamber, and 750 μl of complete medium was placed under the insert. After 24 h at 37°C (corresponding to 72 h after transfection), in a 5% CO2 atmosphere, cells on the upper side of the membrane (non-invasive) were removed by scraping with a cotton bud, and invasive cells were fixed for 10 min in ice-cold methanol (at −20°C). Inserts were then washed in phosphate-buffered saline. The membrane was removed and mounted on a slide in Vectashield Mounting Medium with DAPI (Vector Laboratories, Burlingame, CA, USA). Invasive cells were counted under fluorescence microscopy.
Angiogenesis assay
Fertilised chicken (Gallus gallus) eggs obtained from commercial sources (Pintobar, Braga, Portugal) were incubated at 38°C. At day 3 of incubation, a window was opened in the shell, and 2–2.5 ml albumen was removed. The window was sealed with adhesive tape, and the egg re-incubated. At day 10 of incubation, a 3 mm silicon ring was placed on the growing CAM under sterile conditions. The ring was filled with Mock-GFP (n=14) or CPEB1-GFP (n=16) KATOIII-transfected cells. The window was resealed, and 96 h after inoculation, the ring was removed, and the CAM was excised and photographed ex ovo under a stereoscope (Olympus; SZX16 coupled with a DP71 camera) at 20× magnification. The number of new vessels (<15 μm diameter) growing radially towards the ring area was counted. Means±1 average deviation (AD) were evaluated, and the statistical significance of the differences between the counts was determined using the Student t test (for samples with unequal variance).
Western blotting
Cells were lysed in cold catenin lysis buffer (1% Triton X-100/1% Nonidet P-40, both from Sigma, in phosphate-buffered saline) enriched with a protease and phosphatase inhibitor cocktail (Roche Diagnostics GmbH, Manheim, Germany and Sigma, respectively). Protein samples (25 μg) were separated by sodium dodecyl sulphate (SDS)/10% polyacrylamide gel electrophoresis, and electroblotted on to a Hybond enhanced chemiluminescence membrane (Amersham Biosciences–GE Healthcare, Amersham, Buckinghamshire, UK). Primary antibodies included 4A2 monoclonal anti-CPEB127 (1:800 dilution) and anti-α-tubulin (1:10000 dilution; Sigma). Sheep anti-mouse (1:3000 dilution; Amersham Biosciences) horseradish peroxidase-conjugated secondary antibody was used, followed by enhanced chemiluminescence detection (Amersham Biosciences). Bands were quantified using Quantity One 4.6.8 Software (Bio-Rad, Amadora, Portugal), and values were normalised to tubulin and to the protein expression levels of the Mock-GFP control.
Immunohistochemistry
Paraffin-embedded samples from the CAM were stained with H&E for histological examination and processed for GFP immunostaining. For immunohistochemistry we used a mouse monoclonal antibody against GFP (GFP B2: sc-9996; Santa Cruz Biotechnology, Santa Cruz, California, USA), and for detection we used the Envision detection system peroxidase/DAB (Dako, Carpinteria, California, USA) followed by H&E staining.
Statistical analysis
All statistical analyses were performed using the Student t test, except for the clinicopathological association study, for which we used a Fisher exact test, and the survival curves, which were estimated using the Kaplan–Meier method, considering death from cancer as the end point. Patients who died from other causes were considered as censored at the time of death. Differences between groups were determined using the log-rank test.
p<0.05 was considered significant.
Results
A Drosophila gain-of-function screen to identify GC-associated genes
The screen was designed on the basis of the previously developed Drosophila transgenic model.21 ,22 The rationale of our screen was that homologues of genes capable of modifying the phenotype caused by misexpression of GC-causing mutants may be part of molecular pathways relevant to GC. Specifically, we generated tester strains expressing, in the developing Drosophila retina, wild-type (WT) hEcad or either of two GC-causing germline hEcad mutants, previously found in patients with diffuse GC.8 ,28 Both A634V and V832M are pathogenically relevant GC hEcad mutations with functional consequences—that is, cells expressing these mutants invade and fail to aggregate, causing hereditary forms of diffuse gastric carcinoma.8 ,29 In Drosophila, misexpression of each of these forms resulted in reduction in eye size and roughening of the eye's surface (figure 1A). By combining these strains with a collection of GS lines, we identified those lines capable of modifying the eye phenotype.
We first determined which GS lines, when coexpressed with WT hEcad, resulted in significant enhancement or suppression of the phenotype produced by expression of either hEcad or GS lines alone. Then, interacting GS lines were screened against the two mutant hEcad genotypes to identify GS lines that showed differential interaction with WT and both the mutant forms, in order to sort putatively interesting GC-associated genes. We screened more than 200 individual P-GS lines, but only around 180 provided sufficient progeny for analysis. Independent P-GS insertions mapping in the proximity of the same gene or genes were used as internal controls and, indeed, consistently generated similar interaction phenotypes. After discarding these lines, we ended up evaluating insertions at 154 different loci. From these, four lines (2.6%) were selected as putative GC-related genes, since they differentially interacted with both GC-related mutations in hEcad, in comparison with a normal WT hEcad (figure 1B).
Using Ensembl, HomoloGene, InParanoid, as well as NCBI nucleotide and protein Blast, we determined the putative human orthologues of the fly candidate genes: Mef2 (FBgn0011656), CG12130 (FBgn0033466), CG15625 (FBgn0031644), CG3036 (FBgn0031645), dpp (FBgn0000490), orb (FBgn0004882) and cdc16 (FBgn0025781) (figure 1C). The six human genes selected as potentially GC associated were MEF2A, PAM, SLC17A5, BMP2, CPEB1 and CDC16.
In the case of the GS line targeting orb/cdc16, we additionally confirmed the interaction using an individual UAS line exclusively expressing orb (figure 2A,B), supporting orb as an interacting gene.
To analyse the potential of the human homologue of orb (CPEB1) to modulate E-cadherin mutant phenotype, we performed functional assays associated with E-cadherin loss-of-function. Thus, we carried out invasion assays, since both hEcad mutants are known to be very invasive in vitro.8 ,29
We transiently transfected KATOIII human gastric cells with CPEB1 or the empty vector and evaluated their invasive response. This cell model was selected because it was established from a signet-ring cell carcinoma harbouring E-cadherin loss-of-function, caused by loss of one CDH1 allele and a mutation in the remaining one.30 ,31 We observed that CPEB1-forced expression in KATOIII cells significantly (p=0.013) decreased cell invasion (figure 2C), demonstrating that CPEB1 was able to counteract E-cadherin loss-of-function.
CPEB1 expression in primary GC and cancer cell lines of different epithelial origin
We explored the mRNA expression levels of the six putative GC-associated candidates in nine GC cell lines, and compared them with the normal commercially available gastric mRNA. CPEB1 was the only gene that was significantly downregulated in all GC cell lines (p=1.8E-34) when compared with the normal gastric mRNA (figure 3A and online supplementary figure 1). In order to validate the in vivo importance of CPEB1 in GC, we performed qRT-PCR analysis in the primary GC cases (n=12) and paired normal mucosa for which RNA was available. CPEB1 was significantly and recurrently underexpressed in GC cases in comparison with matched normal mucosa (p=0.0026) (figures 3B and 4A). Eleven of the 12 (92%) cases analysed showed decreased levels of CPEB1 mRNA expression (figure 4A).
We also evaluated CPEB1 expression in cancer cell lines of different epithelial origin (seven from colon cancer and seven from breast cancer). Interestingly, CPEB1 expression was also significantly decreased in all breast and colon cancer cell lines tested (p=2.2E-23 and 4.6E-26, respectively) (figure 3A).
CPEB1 silencing through gene promoter hypermethylation
To obtain some insight into the molecular mechanisms that underlie the significant downregulation of CPEB1 in GC, we screened it for mutations and analysed methylation status in the CPEB1 promoter region.
None of the alterations found in the CPEB1 gene were tumour specific, and all sequence variants but one have been previously described as polymorphisms (online supplementary table 2), and therefore are not predicted to have functional relevance. One missense sequence variant, changing a C to a T at position 901 of exon 6 and generating the P301S amino acid change (online supplementary table 2), was identified in both normal gastric mucosa and neoplastic tissue of one patient. Although the putative functional effect of this alteration is not known and therefore cannot be discarded, it does not account for the reduced levels of expression of CPEB1 in the tumour compared with normal tissue, as the two bear the same mutation.
In contrast, we detected tumour-specific promoter hypermethylation of CPEB1 in 11 of the 12 GC cases for which RNA was available for expression studies. The presence of hypermethylation at the CPEB1 promoter was significantly associated with decreased expression of CPEB1 in both primary tumours and GC cell lines (p=0.048) (figure 4A). Interestingly, MKN28 was the only line that retained any CPEB1 expression (15% of control mRNA), and presented the lowest number of CpG methylated sites. The only primary GC sample (CT284) that retained high CPEB1 expression levels was also the only case presenting the lowest number of CpG methylated sites.
KATOIII cells were then treated with 5-Aza-dC, a well-characterised pharmacological inhibitor of DNA methylation.32 In fact, CPEB1 mRNA expression was recovered after treatment (figure 4B), along with a reduction in the methylation levels of the 79th CpG site of CPEB1 (figure 4B'). Subsequently, we verified that the 5-Aza-dC effect was cell line independent (online supplementary figure 2).
By analysing the methylation status of the most 3′ CpG sites of the CPEB1 promoter and the CPEB1 expression levels in tumours and GC cell lines, we verified that samples displaying higher levels of methylation at the 79th CpG site concomitantly showed significantly (p=0.041) lower levels of CPEB1 mRNA expression (figure 4A).
We further increased our number of samples to 43 to investigate the methylation status of the 79th CpG site of CPEB1 and study the clinicopathological features of these patients and tumours (online supplementary figure 3A). Interestingly, we observed a significant association of methylation at the 79th CpG site of the CPEB1 promoter with diffuse type GC (p=0.007), and, within gastric carcinomas, with lymph node metastases (p=0.042) (online supplementary figure 3B). In addition, cancer-related cumulative survival according to CPEB1 methylation status showed a worse prognosis in GC patients carrying a CPEB1 methylated phenotype (online supplementary figure 3C).
CPEB1 overexpression decreased the angiogenic potential of KATOIII cells in vivo
It is well known that, apart from cell invasion, haemangiogenesis and/or lymphangiogenesis are also required to produce metastases.33 In fact, a correlation between angiogenesis and the degree of lymph node metastasis has been shown in patients with GC.34 Accordingly, data from colon and prostate cancer (accessible at the NCBI GEO database,35 accession GDS75636 and GDS143937) demonstrate that CPEB1 downregulation is likely to be associated with increased metastatic capacity.
Therefore we decided to examine the functional role of CPEB1 in vivo, by evaluating its effect on new blood vessel formation. We inoculated KATOIII GC cells transfected with CPEB1-GFP or Mock-GFP in chick CAM (online supplementary figure 4) and verified that CPEB1 overexpression in KATOIII cells significantly decreased the angiogenic response in the CAM. After 96 h of cell inoculation, CAM with Mock-GFP-transfected cells presented highly disorganised vascular architecture, in contrast with that with CPEB1-GFP-transfected cells, where the normal branched shape was retained (figure 5A). A quantitative analysis of the number of neovessels confirmed that CPEB1-GFP-transfected cells induced a significantly (p=0.005) reduced angiogenic response (figure 5A'). This result suggests CPEB1 to be an angiogenic modulator.
To investigate the mechanism responsible for CPEB1 antiangiogenic potential, we analysed mRNA expression levels of MMP14 and VEGFA, two major angiogenic effectors.38 ,39 Not surprisingly, both were significantly (p=0.022 and p=0.0056, respectively) decreased in response to CPEB1 ectopic expression in KATOIII cells (figure 5B,C).
Discussion
In this study, we show the efficacy of our in vivo Drosophila model for identifying novel genes and molecular mechanisms associated with cancer, particularly GC. We screened for phenotypic modifiers of E-cadherin in a WT versus a GC-associated context, by using two hEcad mutations that cause GC, as the initial step. E-cadherin is a well-known protein involved in GC as a causative genetic alteration in hereditary diffuse GC (HDGC) as well as in its sporadic forms. The two hEcad-causing mutations used for the screening were found at the germline level in patients with hereditary diffuse GC.4–11 Our approach led us to identify six genes (MEF2A, PAM, SLC17A5, BMP2, CPEB1 and CDC16). From these candidates, the most striking result obtained was the strong downregulation of CPEB1 expression in every GC cell line analysed. This was further validated in GC primary tumours (only one of 12 samples behaved differently) meaning that altered expression of CPEB1 is not acquired during in vitro manipulation of cancer cells but occurs during the process of GC development in vivo. Two studies have shown CPEB1 silencing in ovarian cancer40 and multiple myelomas.41 Interestingly, we found CPEB1 underexpression in all colon and breast cancer cell lines studied. However, the role of CPEB1 in these malignancies is still unexplored, and this is the first time, to our knowledge, that underexpression of CPEB1 in GC has been reported.
CPEB1 codes for a cytoplasmic polyadenylation element (CPE) binding protein that controls the translational activation or repression of many proteins during development. The CPE uridine-rich sequence, to which this highly conserved protein binds, is found in the 3′ UTR of some mRNAs.42–44 When lost, CPEB has been shown to increase resistance to nutritional stress, possibly promoting the survival of rapidly dividing, hypermetabolic tumour cells as the levels of available nutrients decline.45 Cells knocked-down for CPEB show greater proliferation, have reduced mitochondrial respiration and, consequently, higher glycolytic rates.46 Although CPEB knock-out (KO) mouse embryonic fibroblasts do not originate tumours when injected into athymic mice, they form larger tumours than WT mouse embryonic fibroblasts when transformed with Ras.47 Probably, CPEB KO cells do not have transforming potential per se, but increase susceptibility to cancer development if these alterations occur in a context that harbours other genetic modifications. In accordance with this hypothesis, CPEB KO mice have a similar tendency to tumour formation to that of WT mice, but form papillomas at a significantly faster rate when exposed to chemical carcinogens.46
We have identified the molecular mechanism underlying CPEB1 silencing in tumours (in both GC cell lines and primary GC), by showing a cancer-associated hypermethylation phenotype at the most 3′-region of the CPEB1 gene promoter. We have also demonstrated that primary tumours when compared with normal gastric mucosa harbour both a higher number and higher levels of CpG site methylation. Furthermore, in the series under analysis, we excluded CPEB1 mutations as an underlying mechanism associated with GC. Heller et al described for the first time CPEB1 epigenetic silencing in cancer, but the CpG site that regulates its expression remained to be identified.41 Here, we demonstrate for the first time that CPEB1 mRNA expression levels are significantly associated with the methylation status of a specific CpG site at the CPEB1 promoter. Remarkably, we were able to reverse CPEB1 downregulation after treatment with 5-Aza-dC, which precisely decreased the methylation levels of the 79th CpG site of the CPEB1 promoter, highlighting its specificity. Indeed, only this particular C had a reversed methylation status, and this effect was associated with modifications of CPEB1 expression. Moreover, using several Bioinformatic Transcription Factor Prediction Tools (TFMatrix, TRANSFAC, Genomatix), we verified that this CPEB1-specific CpG site is a potential binding site for various transcription factors (TFs), probably indicating a key regulatory region at the CPEB1 promoter. USF1 and HIF1 were two of the TFs recurrently picked up in this bioinformatic screen and may represent putative CPEB1 transcriptional regulators relevant for future investigation (data not shown).
Our work also revealed new roles for CPEB1 related to cancer progression. Despite the number of studies describing the wide-ranging functions of CPEB1, its role as an invasion and angiogenesis regulator has never been described. First, we present orb (the CPEB1 fly homologue) as a modifier of the E-cadherin mutant phenotype in the fly. Then, we demonstrate that in vitro this modulation by CPEB1 probably occurs through invasion impairment, at least in KATOIII cells. Accordingly, we observed a highly significant correlation between the 79th CpG site methylation of CPEB1 and diffuse type GC, which is characterised by an infiltrative phenotype with cells invading the gastric wall.48 ,49 In addition, in our study we show that CPEB1 expression reduces the de novo formation of vessels in a chick CAM in vivo model. We propose that the antiangiogenic role of CPEB1 is mediated through downregulation of VEGFA and MMP14, two key players in angiogenesis.38 ,39 In accordance with this, CPEB1 had already been reported to interact with the 3′ UTR of HIF1α mRNA, the hypoxia master regulator which in turn controls angiogenesis by transcriptional regulation of VEGFA.50 In cancer development, it is well accepted that an increase in vasculature raises the risk of tumour cells giving rise to metastasis.51 Tumour cell invasion alone is not sufficient to induce metastasis; for that to occur, transport of malignant cells through the blood and/or the lymph vessels is required.33 This may explain why the 79th CpG site methylation of CPEB1 was significantly associated with the presence of lymph node metastases, revealing CPEB1 regulatory potential in terms of lymphatic diffusion. Therefore, in addition to all the previously described aspects in which CPEB1 has been implicated, its antiangiogenic role is likely to confer protection against tumour progression and metastasis. Finally, special attention should be given to the putative relationship verified between CPEB1 79th CpG site methylation and the poor rate of survival of patients with GC.
In conclusion, using the Drosophila model, we have identified a novel GC-associated gene and showed for the first time that CPEB1 hypermethylation is a very common event in GC. We show that its silencing possibly leads to an increase in tumour infiltration and metastasising capacity, as it interferes with invasion and angiogenesis. However, further investigations are required to elucidate CPEB1 transcriptional regulation by putative TFs, and to clarify the value of CPEB1 as either a tumour or prognostic marker, not only in GC, but in other types of cancer, particularly breast and colon. More studies focusing on CPEB1 interaction with E-cadherin should be conducted, as well as research on the potential of CPEB1 as a therapeutic target, since we have abrogated CPEB1 epigenetic inactivation with a known anticancer demethylating agent (5-Aza-dC).
Acknowledgments
We thank José Felix de Celis (CBMSO, Madrid, Spain) for providing access to the GS strain collection.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Download Supplementary Data (PDF) - Manuscript file of format pdf
- Download Supplementary Data (PDF) - Manuscript file of format pdf
- Download Supplementary Data (PDF) - Manuscript file of format pdf
- Download Supplementary Data (PDF) - Manuscript file of format pdf
- Download Supplementary Data (PDF) - Manuscript file of format pdf
- Download Supplementary Data (PDF) - Manuscript file of format pdf
Footnotes
Funding Supported by: The Portuguese Foundation for Science and Technology (FCT) (Projects: PTDC/SAU-ONC/110294/2009; PhD grants: SFRH/BD/29777/2006-JC, SFRH/BD/40090/2007-GC; Post-Doc grants: SFRH/BPD/48765/2008-JSC; Salary support from Program Ciência 2007 and 2008-CO, JP, MTP and PSP). IPATIMUP is an Associate Laboratory of the Portuguese Ministry of Science, Technology and Higher Education and is partially supported by FCT. Also supported by grants BFU2009-07044 and ‘From Genes to Shape’ (Consolider-Ingenio) from the Spanish MICINN to FC.investigation.
Competing interests None.
Patient consent All samples were obtained with informed consent and in compliance with the Helsinki Declaration (http://www.wma.net/e/policy/b3.htm). Written consent was obtained from every patient affected by primary GC. These consents are archived at the Department of General Surgery and Surgical Oncology (Hospital Santa Maria alle Scotte, Siena, Italy). In accord with the Italian D.L. n. 196 of the 30 June 2003, namely the Privacy Law, we cannot distribute personal information such as name, health condition, and specific information on patients. Therefore information has been sufficiently anonymised, meaning that neither the patient nor anyone else can identify the patient with certainty.
Ethics approval The study protocol was reviewed and approved by the appropriate ethics committees.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement We are happy to share our data without any patent protection or conflict of interest.