p38 MAPK Down-regulates Fibulin 3 Expression through Methylation of Gene Regulatory Sequences

Background: p38α MAPK regulates migration/invasion. Results: p38α induces hypermethylation of Fibulin 3 gene regulatory sequences leading to Fibulin 3 down-regulation. This contributes to regulate migration and invasion in MEFs and HCT116 cells. Conclusion: p38α down-regulates fibulin 3 expression through promoter methylation to control p38α-mediated migration and invasion. Significance: To understand the function of new p38α targets in migration/invasion and tumorigenesis. p38 MAPKs regulate migration and invasion. However, the mechanisms involved are only partially known. We had previously identified fibulin 3, which plays a role in migration, invasion, and tumorigenesis, as a gene regulated by p38α. We have characterized in detail how p38 MAPK regulates fibulin 3 expression and its role. We describe here for the first time that p38α, p38γ, and p38δ down-regulate fibulin 3 expression. p38α has a stronger effect, and it does so through hypermethylation of CpG sites in the regulatory sequences of the gene. This would be mediated by the DNA methylase, DNMT3A, which is down-regulated in cells lacking p38α, but once re-introduced represses Fibulin 3 expression. p38α through HuR stabilizes dnmt3a mRNA leading to an increase in DNMT3A protein levels. Moreover, by knocking-down fibulin 3, we have found that Fibulin 3 inhibits migration and invasion in MEFs by mechanisms involving p38α/β inhibition. Hence, p38α pro-migratory/invasive effect might be, at least in part, mediated by fibulin 3 down-regulation in MEFs. In contrast, in HCT116 cells, Fibulin 3 promotes migration and invasion through a mechanism dependent on p38α and/or p38β activation. Furthermore, Fibulin 3 promotes in vitro and in vivo tumor growth of HCT116 cells through a mechanism dependent on p38α, which surprisingly acts as a potent inducer of tumor growth. At the same time, p38α limits fibulin 3 expression, which might represent a negative feed-back loop.

p38 MAPKs regulate migration and invasion. However, the mechanisms involved are only partially known. We had previously identified fibulin 3, which plays a role in migration, invasion, and tumorigenesis, as a gene regulated by p38␣. We have characterized in detail how p38 MAPK regulates fibulin 3 expression and its role. We describe here for the first time that p38␣, p38␥, and p38␦ down-regulate fibulin 3 expression. p38␣ has a stronger effect, and it does so through hypermethylation of CpG sites in the regulatory sequences of the gene. This would be mediated by the DNA methylase, DNMT3A, which is down-regulated in cells lacking p38␣, but once re-introduced represses Fibulin 3 expression. p38␣ through HuR stabilizes dnmt3a mRNA leading to an increase in DNMT3A protein levels. Moreover, by knocking-down fibulin 3, we have found that Fibulin 3 inhibits migration and invasion in MEFs by mechanisms involving p38␣/␤ inhibition. Hence, p38␣ pro-migratory/invasive effect might be, at least in part, mediated by fibulin 3 downregulation in MEFs. In contrast, in HCT116 cells, Fibulin 3 promotes migration and invasion through a mechanism dependent on p38␣ and/or p38␤ activation. Furthermore, Fibulin 3 promotes in vitro and in vivo tumor growth of HCT116 cells through a mechanism dependent on p38␣, which surprisingly acts as a potent inducer of tumor growth. At the same time, p38␣ limits fibulin 3 expression, which might represent a negative feed-back loop.
p38 MAPKs 5 are a subfamily of MAPKs activated by several stimuli, which are involved in the regulation of the main cellular functions, including migration and invasion (1)(2). There are four isoforms of p38 MAPKs: p38␣, p38␤, p38␥, and p38␦, which can have both overlapping and specific functions (1). p38␣ and p38␤ show a high grade of homology and are ubiquitously expressed, while p38␥ and p38␦ have more restricted expression patterns and some specialized functions (2). p38␣ is essential for embryonic development (3)(4), being expressed at high levels (5). It regulates different cellular functions. For example, it can inhibit proliferation (6 -8) and adhesion (9) and promote differentiation (7,10), apoptosis (11)(12)(13), migration (14), and invasion (15)(16). In addition, it can also activate proliferation (17)(18) or survival (19 -21). In fact, it is now clear that p38␣ can play dual roles depending on the stimulus, cellular context, or other additional factors (8). This is also true for cancer, where p38␣ behaves as either a tumor suppressor or promoter depending on the type of cancer and the tumor stage (8). In a number of tumors, p38␣ inhibits tumor initiation because of its role in cell cycle arrest and in the induction of apoptosis (13). However, at later stages, it can promote survival (19 -20), migration, and invasion leading to metastasis (8,16).
According to this, in several tumor cells lines p38␣ mediates migration and invasion through regulation of cell motility, MMP expression, and/or activity (16,(22)(23).
The function of p38␥ and p38␦ in cancer has not been well characterized. It has been recently shown that they can play a tumor suppressor role inhibiting cell migration, MMP2 secretion and tumor growth in MEFs (24). In contrast, p38␦-null mice are more resistant to skin tumor development (25) and colon cancer development is impaired in p38␥/␦-deficient mice (26).
Fibulins are a family of extracellular matrix (ECM) proteins (27)(28). They are secreted glycoproteins characterized by the presence of a shared globular domain at the C terminus called "fibulin-like" domain (27)(28)(29). This domain is preceded by a series of epidermal growth factor (EGF)-like domains. These proteins play relevant roles in the assembly and stabilization of supramolecular ECM complexes and as a consequence they regulate essential cellular functions such as adhesion, migration, or proliferation, being involved in tissue organogenesis, vasculogenesis, fibrogenesis, and tumorigenesis (27)(28)30). In particular, Fibulin 3 (also referred to as EFEMP-1) is expressed in different tissues and it can play a dual role in cancer (30). In some tumors, such as glioma, it is overexpressed, promoting cell migration and invasion (31)(32). Similarly, in pancreatic adenocarcinoma Fibulin 3 is up-regulated, which is associated with metastatic tumor growth (33). In contrast, fibulin 3 was shown to be down-regulated in non-small lung carcinoma, breast cancer, or coloncarcinoma, where it might behave as a tumor suppressor (34 -37). In particular, in non-small lung carcinoma it has been recently demonstrated that Fibulin 3 inhibits epithelial to mesenchymal transition (EMT) and self-renewal of lung cancer stem cells (35).
In some of the tumors, fibulin 3 expression is down-regulated as a consequence of promoter methylation (34,37). In fact, transcriptional regulation of fibulin 3 gene is a relevant mechanism controlling its expression through different regulatory sequences present at the 5Ј-end (38). However, it remains unknown the mechanism responsible for this epigenetic regulation of Fibulin 3 expression.
DNA methyltransferases (DNMTs) are responsible for DNA methylation (39). There are three enzymatically active mammalian DNMTs, DNMT1, DNMT3A, and DNMT3B. DNMT1induced DNA methylation is associated with DNA replication (40), while DNMT3A/3B are thought to function as de novo DNA methyltransferases, and their levels can be regulated, being of relevance its post-transcriptional regulation (39). In particular, binding of HuR protein to the 3Ј-UTR of DNMT3B mRNA enhances its stability, increasing its protein levels (41). Microarrays analyses revealed that Fibulin 3 mRNA levels were up-regulated in p38␣ Ϫ/Ϫ MEFs. 6 Based on that, together with the above described functions of Fibulin 3 and p38␣ in the control of migration and invasion, it could be hypothesized that p38␣ could act through Fibulin 3 to regulate these processes. Therefore, we explored in detail if p38␣ MAPK and other p38 isoforms were able to regulate Fibulin 3 expression in non-tumor cells (MEFs), the mechanisms involved, and its function. We also determined whether p38␣ was also able to regulate Fibulin 3 expression in the HCT116 colon carcinoma cell line and its impact on migration, invasion, and tumorigenesis in these cells.
Pyrosequencing-Genomic DNA was extracted from 24h serum-deprived MEFs using the alkaline lysis method and modified by sodium bisulfite using BisulFlash DNA Modification Kit (EPIGENTEK, P-1026). The DNA region Ϫ28853253/ Ϫ28853452 was amplified by PCR using the PyroMark PCR kit (Qiagen 978703) using the following primers: forward: 5Ј-CCTC-CTGTGGCTGCTGCTGCAG-3Ј; reverse (biotinylated): 5Ј-CACTTTGACATGTCTCTTCTACCTCCA-3Ј. PCR cycles were as follows: 30 sec at 95°C, 30 sec at 52°C, and 30 sec at 72°C (45 cycles). PCR products were converted into singlestranded DNA. One strand was isolated using streptavidin-Sepharose beads and was used as a template in the pyrosequen-cing PCR reaction using two different primers: 5Ј-GCTGCCC-TCCCCTACGCACTCCTT-3Ј for the analysis of the methylation status of five CpG sites and 5Ј-CCCGCAGGTAGGAGC-CCAAAGC-3Ј for the analysis of seven additional CpG sites.
Fibulin 3 and HuR Knock-down-Permanent fibulin 3 knockdown in MEFs was performed by infection with mouse fibulin 3 shRNAs Lentiviral Particles (75000 infectious units) containing a mixture of different shRNAs (Santa Cruz Biotechnology sc-44625-V) in the presence of 10 g/ml Polybrene (Santa Cruz Biotechnology sc-134220) or a control shRNA for non-silenced cells. Similarly, permanent HuR knock-down in wt MEFs was performed by transfection of a mixture of plasmids containing different mouse HuR shRNAs (Santa Cruz Biotechnology sc-35620-sh) using Metafectene-Pro (Biontex T040 -0.2) as previously described (21). Then, cells were selected with 1 g/ml puromycin.
Permanent fibulin 3 knock-down in HCT116 cells was carried out using a human fibulin 3 shRNA (OriGene Technologies, TR30018) inserted in the pGFP-B-RS vector. Cells were transfected using Metafectene-Pro (Biontex T040-0.2). Different clones were selected with blasticidin (2 g/ml, Invitrogen R210-01) and used for the experiments. As a control, cells transfected with the empty vector were also generated. Transient fibulin 3 knock-down was also performed using a second human fibulin 3 shRNA (Sigma NM-004105). Cell assays were initiated 48 h after transfection.
Wound Healing Assays-Confluent cells were pre-treated with mitomycin C (25 g/ml, Sigma-Aldrich M0503) for 30 min to inhibit cell growth. Then, a straight scratch was performed and the medium replaced by a fresh one without serum (for MEFs) or with 2% FBS (ϮHGF for HCT116 cells). Cells were maintained for 12-72 h at 37°C and 5% CO 2 . Migration was followed by a phase-contrast microscope (Eclipse TE300 Nikon coupled to a digital camera) at different time points. Photographs were taken to quantify (using TScratch program) the percentage of wound healing closure at the different times.
Invasion Assays-Invasion through matrigel was assayed using matrigel (444 g/cm 2 ) (BD Biosciences, 356234)-coated transwells (8 m filter, BD 353097). Cells (50,000) were seeded in the upper chamber in a serum-free medium. In the lower chamber, FBS (10%) or HGF (40 ng/ml) was added to the medium to act as a chemoattractant. Then, cells were left in the incubator for 24 h at 37°C. Medium and matrigel from the upper chamber were removed, and cells present in the lower chamber were fixed with 4% paraformaldehyde and stained with crystal violet 0.2% p/v (Sigma-Aldrich C-0775). Cells were counted using a phase-contrast microscope.
Zymography-To determine MMP-2 and MMP-9 activities, 80% confluent cells were serum-deprived for 24 -48 h and the culture medium was used for an electrophoresis in 8% SDSpolyacrylamide gels polymerized in the presence of 0.1% gelatin under non-reducing conditions. Gels were washed with 2.5% Triton X-100 (30 min) to remove SDS, rinsed with substrate buffer (0.2 M NaCl, 5 mM CaCl 2 , 1% Triton X-100, 0.02% NaN3, 50 mM Tris pH 7.5) and incubated in this buffer at 37°C overnight to allow protein renaturation and MMP activation. To visualize gelatin degradation, the gel was stained with Coomassie Brilliant Blue (Bio-Rad, 161-0400).
Focus Formation Assays-To measure anchorage-dependent growth, 300 cells (MEFs or HCT116 cells) were seeded in a 10-cm dish. After 10 -13 days for MEFs or 8 -10 days for HCT116 cells, foci were stained with a 0.2% crystal violet solution. The total number of foci was quantified using Image J program and their size using OpenCFU program. The size of colonies was measured as volume applying the equation 4/3r3, where r is the radius of foci.
Growth in Soft Agar-To measure anchorage-independent growth, cells were cultured in 24-well dishes containing two agar layers. Cells (3 ϫ 10 3 ) were resuspended in 0.7% agar (BD, 214530) diluted in complete medium (2ϫ) and poured onto a 0.5% layer of agar (diluted in medium). Fresh medium was added to the top layer every 3 days. After 2 weeks, colonies were stained with 0.005% crystal violet and counted using a dissecting microscope.
Xenografts Assays-HCT116 cells (10 6 cells/100 l) were resuspended in McCoy's medium and injected subcutaneously into the flank of eight-week-old male nude mice (Harlan Laboratories). Tumor growth was monitored twice a week for 6 weeks. Tumor volume was calculated by the formula ((L/2) ϫ (W/2)) ϫ , where L and W are the longest and the shortest diameter in millimeters, respectively. All animal experiments were carried out in compliance with the institutions guidelines.
Statistical Analysis-Data are represented as the mean values Ϯ S.E. The comparisons were done between two experimental groups. An unpaired Student's t test was used.

p38 MAPKs Down-regulate Fibulin 3 Expression-Using
Affymetrix gene Chips, we identified several mRNAs that were up-regulated in p38␣ knock-out MEFs growing with serum as compared with wt (data not shown). Among them, fibulin 3 mRNA levels were highly up-regulated. This was validated and quantified by RT-qPCR ( Fig. 1A) using different MEFs cell lines (wt and p38␣ Ϫ/Ϫ ). To further demonstrate that these changes in fibulin 3 expression were dependent on p38␣ MAPK, the effect of the selective p38␣/␤ inhibitor SB203580 was determined. As shown in Fig. 1B, treatment of wt MEFs with SB203580 led to an up-regulation of fibulin 3 mRNA, while it had no effect on p38␣ Ϫ/Ϫ cells, even when using a higher SB203580 concentration (data not shown). This indicates that p38␣, but not p38␤, is responsible for fibulin 3 regulation. We also analyzed if other p38MAPKs isoforms could also regulate fibulin 3 expression. Results from Fig. 1A show increased levels of fibulin 3 mRNA in MEFs lacking p38␥, p38␦ or both as compared with wt cells, although lower than in p38␣ Ϫ/Ϫ MEFs. Therefore, we decided to focus our studies on the analysis of the function of p38␣ in the regulation of Fibulin 3 expression. According to mRNA data, we found increased levels of the secreted and intracellular Fibulin 3 protein in MEFs deficient in p38␣ (Fig. 1, C and D). Similarly, treatment of wt MEFs with SB203580 led to the up-regulation of secreted Fibulin 3 protein levels (Fig. 1D). Moreover, transient re-expression of p38␣ into p38␣ Ϫ/Ϫ MEFs abolished Fibulin 3 up-regulation (Fig. 1E). All these data indicate that p38 MAPKs, mainly p38␣, are negative regulators of fibulin 3 expression.
As fibulin 3 expression is deregulated in different tumors, including colorectal cancer (30,37), we tested whether p38␣ could down-regulate fibulin 3 expression in the colon-carcinoma HCT116 cell line. Fig. 1F shows an increase in fibulin 3 mRNA levels in HCT116 cells with permanent p38␣ knockdown. Similarly, the levels of the secreted and intracellular Fibulin 3 protein were higher in p38␣ knock-down cells (Fig.  1G). Moreover, inhibition of p38␣ with SB203580, as an alternative experimental approach, also increased Fibulin 3 secretion in cells expressing p38␣ (Fig. 1H). Therefore, p38␣ is also a negative regulator of fibulin 3 in HCT116 cells.
p38␣ MAPK Induces the Hyper-methylation of Regulatory Sequences of Fibulin 3 Gene-Fibulin 3 expression can be repressed through promoter methylation in some tumors (34, 37) leading to changes in their invasive capacity (34). So, we wondered whether p38␣ might down-regulate fibulin 3 through promoter methylation. To analyze it, we first studied the effect of the DNA methylation inhibitor, 5-aza-2Ј-deoxicytidine (5A2dC). As shown in Fig. 2A, fibulin 3 mRNA levels highly increased in wt MEFs treated with 5A2dC, while they remained unchanged in p38␣ Ϫ/Ϫ MEFs. Accordingly, 5A2dC treatment induced an increase in the levels of the secreted Fibulin 3 protein both, in wt MEFs (Fig. 2B) and in non-silenced HCT116 cells (Fig. 2C), while it had no effect, either in p38␣ knock-out MEFs (Fig. 2B) or knock-down HCT116 cells (Fig.  2C). These results suggest that p38␣ might down-regulate fibulin 3 expression through a mechanism dependent on DNA methylation. To demonstrate it, we analyzed methylation status of fibulin 3 gene regulatory sequences in wt and p38␣ knock-out MEFs. In particular, the methylation levels of twelve CpG islands sites present in the 5Ј-untranslated region (UTR) of fibulin 3 gene (1210bp upstream of the ATG translation start site) were determined by pyrosequencing. The average methylation percentage of 10 out of 12 CpG sites was higher in wt as compared with p38␣ Ϫ/Ϫ MEFs (Fig. 2, D and E). These results strongly indicate that p38␣ represses fibulin 3 expression through hypermethylation of 5Ј-UTR regulatory sequences.
DNA methylation results from the activity of DNMTs (39). DNMT3A/3B protein levels can be regulated, being relevant its post-transcriptional regulation (39). In particular, binding of HuR protein to the 3Ј-UTR of DNMT3B mRNA enhances its stability, increasing its protein levels in colorectal RKO cells (41). p38␣ MAPK can phosphorylate HuR, which enhances its binding to certain mRNAs such as p21mRNA, increasing its protein levels (42). This raised the possibility that p38␣ MAPK could be regulating DNMT3A/3B protein levels through a HuR-dependent mechanism. So, we analyzed DNMT3A protein levels in wt and p38␣ Ϫ/Ϫ MEFs and in non-silenced and p38␣ knock-down HCT116 cells. We found a significant decrease in DNMT3A protein levels in both, p38␣ Ϫ/Ϫ MEFs and p38␣ knock-down HCT116 cells, as compared with cells expressing p38␣ (Fig. 2F). These results indicate that p38␣ MAPK positively regulates DNMT3A protein levels, which inversely correlates with Fibulin 3 levels, suggesting that DNMT3A would be responsible for the hypermethylation of fibulin 3 gene regulatory sequences and the subsequent downregulation of Fibulin 3. To prove it, DNMT3A was re-introduced in p38␣ deficient MEFs and in p38␣ knock-down HCT116 cells by transfection of a dnmt3a construct, which led to a strong decrease in Fibulin 3 levels (Fig. 2G).
Additionally, we analyzed how p38␣ controls DNMT3A protein levels. We found higher levels of dnmt3a mRNA in wt than in p38␣ Ϫ/Ϫ MEFs (Fig. 3A). Upon inhibition of transcription with actinomycin D, those levels highly decreased in p38␣ Ϫ/Ϫ MEFs at 2h and to a significant less extent in wt cells (Fig. 3B), which suggests that dnmt3a mRNA is stabilized by p38␣. As HuR could be involved in dnmt3a mRNA stabilization, we first measured total HuR protein levels and found them down-regulated in p38␣ Ϫ/Ϫ MEFs and p38␣ knock-down HCT116 cells (Fig. 3C). Moreover, HuR knock-down decreased DNMT3A protein levels in wt MEFs (Fig. 3D). Therefore, our results demonstrate that p38␣ acting through HuR stabilizes dnmt3a mRNA, leading to increased DNMT3A protein levels, which in turn would down-regulate Fibulin 3 via hypermethylation of regulatory sequences of the gene.
Fibulin-3 Knock-down Increases Migration and Invasion of MEFs-Fibulin 3 has been shown to play a role in migration and invasion in several tumors, either promoting or inhibiting cell invasiveness depending on the tumor type (30). p38␣ can mediate cell migration and invasion (8,16), so it would be possible that p38␣, through down-regulation of fibulin 3, could favor cell migration and invasion. To analyze it, fibulin 3 was permanently knocked-down in MEFs (Fig. 4A). As expected, wound healing assays revealed a faster migration of wt MEFs as compared with p38␣ Ϫ/Ϫ cells (Fig. 4B). Fibulin 3 knock-down highly increased migration of p38␣ Ϫ/Ϫ MEFs up to the levels of wt cells, while no significant changes were observed in wt MEFs (Fig. 4B). Moreover, fibulin 3 knock-down promoted invasion through matrigel of p38␣ knock-out MEFs, so that the number of invading cells was similar to the one found in wt MEFs (Fig.  4C). Although wound healing assays did not reveal any significant change in cell migration in wt MEFs upon fibulin 3 knockdown, invasion through matrigel was slightly increased (Fig.  4C). These results indicate that fibulin 3 acts as an inhibitor of cell migration and invasion in MEFs as it happens in certain tumors, such as non-small lung carcinoma (34 -35). However, this effect was stronger in p38␣ Ϫ/Ϫ MEFs, where the levels of secreted Fibulin 3 are much more higher. Moreover, the p38␣ pro-migratory/invasive effect might be, at least in part, mediated by Fibulin 3 down-regulation in MEFs.
As MMPs are relevant for extracellular matrix degradation during cell migration/invasion (43-44), we evaluated whether the increased invasion observed in fibulin 3 knock-down MEFs was due to changes in MMPs levels and/or in their activities. We did not find significant changes in the levels of MMP2, -7, -9, -10, -11, and -13 mRNAs that could explain invasion results (data not shown). Thus, we next evaluated MMP2 and MMP9 activities. As observed in Fig. 4D, MMP2 and -9 activities were lower in p38␣ Ϫ/Ϫ than in wt MEFs and MMP9 activity increased in p38␣ knock-out MEFs upon fibulin 3 knock-down, which could be responsible for their enhanced invasion. However, MMP2/9 activities were decreased in fibulin 3 knock-down wt MEFs (Figs. 4D and 5D), which did not correlate with its enhanced invasive capacity. Therefore, we determined the effect on cell invasion of a broad spectrum MMP inhibitor, marimastat. We found that it impaired invasion in all cell lines (Fig. 4E). This suggests that other MMPs, different from MMP2 and -9, might be responsible for the increased invasion of fibulin 3 knockdown wt MEFs. We also evaluated the impact of fibulin 3 knock-down in some of the signaling pathways regulating cell migration and invasion such as p38 MAPKs, PI3K/Akt and ERKs. As shown in Fig. 5A, an increase in P-ERKs and P-p38 MAPKs levels was induced by fibulin 3 knock-down, mainly in cells stimulated with serum. Moreover, upon fibulin 3 knock-down p38␣ MAPK phosphorylation was enhanced in wt MEFs, while in p38␣ Ϫ/Ϫ cells there was a strong increase in the phosphorylation of another p38 MAPK isoform (potentially, p38␤) with a lower mobility (Fig. 5A). To determine the relevance of the hyper-activation of these p38 MAPKs, we evaluated the effect of the treatment with SB203580 on migration and invasion. As shown in Fig. 5, B and C, inhibition of p38␣/␤ impaired migration and invasion of fibulin 3 knock-down cells and wt MEFs. This partially correlates with the decrease in MMP9 activity upon treatment with SB203580 (Fig. 5D). Moreover, transient knock-down of p38␤ in fibulin 3 knock-down p38␣ Ϫ/Ϫ MEFs abolished p38 hyperactivation (Fig. 5E), as well as migration in wound healing assays (Fig. 5F). All these data indicate that the enhanced activation of p38␣ in wt and p38␤ in p38␣ Ϫ/Ϫ cells induced by fibulin 3 knock-down is necessary for migration and invasion of these cells.
As fibulin 3 knock-down increases the invasive capacity of MEFs, mainly that of p38␣ Ϫ/Ϫ cells, we further analyzed the behavior of these cells. They grew faster than wt and p38␣ Ϫ/Ϫ MEFs (data not shown). In addition, anchorage-dependent growth assays revealed an enhanced foci formation upon fibulin 3 knock-down (wt and p38␣ Ϫ/Ϫ MEFs) (Fig. 6, A and B). In contrast, as shown in Fig. 6C, foci size was only increased in p38␣ knock-out cells (with or without fibulin 3 knock-down). All these data indicate that contact inhibition is lost in fibulin 3 knock-down cells. This suggests that these cells could have suffered a process of transformation as impaired contact inhibition is considered a hallmark of cell transformation (45). However, fibulin 3 knock-down MEFs were unable to grow in soft agar or to induce tumors in xenograft assays (data not shown), which suggests that fibulin 3 knock-down is not sufficient to induce transformation, but it may collaborate with other genes as it happens in lung carcinoma (34 -35).
Fibulin 3 Knock-down Inhibits Migration and Invasion of HCT116 Cells-In colorectal cancer, a down-regulation of fibulin 3 gene expression by promoter methylation was shown to occur in advanced stages, which correlated with the induction of metastasis (37). We have shown here that fibulin 3 expression is repressed in non-silenced HCT116 cells as compared with p38␣ knock-down cells and this can be prevented by DNA demethylation. So, we evaluated the function of Fibulin 3 in migration and invasion in HCT116 cells through gene silencing (Fig. 7A). As shown in Fig. 7B, non-silenced HCT116 cells migrated faster than p38␣ knock-down cells, either in the absence or presence of HGF. Fibulin 3 knock-down impaired migration of non-silenced cells and slightly reduced that of p38␣ knock-down cells. This was confirmed using another shRNA against fibulin 3 (sh2Fib3 in Fig. 7, C and D). Similarly, fibulin 3 silencing blocked basal and HGF-induced invasion through matrigel (Fig. 8A). Accordingly, MMP2 and 9 activities were lower in fibulin 3 knock-down cells (Fig. 8B), which correlated with its reduced invasive capacity. To evaluate the relevance of MMPs in the invasion capacity of fibulin 3 knockdown HCT116 cells, the effect of the MMP inhibitor, marimastat, was assessed. As shown in Fig. 8C, HGF-induced invasion was impaired by marimastat treatment. Therefore, the changes in the activity of MMPs might mediate the pro-invasive effect of fibulin 3 in HCT116 cells.
It should be noticed that although Fibulin 3 levels were lower in non-silenced cells, as compared with p38␣ knock-down cells, its knock-down inhibited both migration and invasion. In fact, Fibulin 3 appears to be a positive regulator of cell migration and invasion in non-silenced and p38␣ knock-down HCT116 cells, but this effect is more prominent when p38␣ is expressed. This correlates with the levels of p38␣/␤ phosphorylation (Fig. 8D). Thus, in non-silenced HCT116 cells, HGF induced the activation of p38␣ MAPK and another isoform with a lower mobility (probably p38␤), which was the only one activated in p38␣ knock-down cells (Fig. 8D). This p38 MAPKs activation was highly reduced in fibulin 3 knock-down cells, which might account for the decreased migration. In fact, inhibition of p38␣/␤ with SB203580 had a similar effect to that of fibulin 3 knock-down (data not shown). ERKs and Akt activation was also decreased in fibulin 3 knock-down cells (data not shown), but its relevance in the migration of these cells appears to be unclear.
Together these results indicate that Fibulin 3 promotes migration and invasion of HCT116 cells through a mechanism that requires p38␣ and/or p38␤ activation. At the same time, p38␣ limits Fibulin 3 expression, which could represent a negative feed-back loop. So, we next wanted to determine the function of Fibulin 3 in the regulation of the tumorigenic capacity of these cells. As shown in Fig. 9A (left panel), fibulin 3 knockdown significantly reduced the number of foci in both nonsilenced and p38␣ knock-down cells, although the effect was more prominent in non-silenced HCT116 cells. In addition, the foci size was smaller (Fig. 9A, right panel). The number of foci was also reduced by p38␣ knock-down, but differences were not statistically significant. This effect of fibulin 3 knock-down was confirmed using another shRNA (sh2Fib3, Fig. 9B). To further understand the function of Fibulin 3 in the tumorigenic capacity of HCT116 cells, xenografts assays in nude mice were performed. As shown in Fig. 9C, non-silenced HCT116 cells led to tumor formation 14 days after injection and the tumor size progressively increased over time. Tumors sizes were significantly decreased by fibulin 3 knock-down in non-silenced cells, but not in p38␣ knock-down cells, where the size of the tumors was highly reduced, independently of fibulin 3 silencing (Fig.  9C). All this indicates that Fibulin 3 promotes tumor growth in HCT116 cells through a mechanism dependent on p38␣.
Moreover, in these cells, p38␣ is a potent promoter of tumor growth. Nevertheless, to further confirm the role of Fibulin 3 and p38␣ in tumorigenesis, soft agar assays were performed using also SB203580 to inhibit p38␣/␤. As shown in Fig. 9D, SB203580 treatment and, either fibulin 3 or p38␣ knock-down, highly reduced anchorage independent cell growth, but only in cells expressing p38␣. Accordingly, SB203580 had no further effect when fibulin 3 or p38␣ were knocked-down. These data are in agreement with data derived from xenografts assays, although the effect of p38␣ knock-down is lower, probably due to differences in the microenviroment and/or the influence of the angiogenic processes.

DISCUSSION
Data presented here uncover Fibulin 3 as a new target of p38 MAPK, which participates in the regulation of migration and invasion in MEFs and HCT116 cells. p38␣, p38␥, and p38␦ regulate Fibulin 3 expression, but the effect of p38␣ is more dramatic, so it has been characterized. We described for the first time that p38␣ down-regulates Fibulin 3 expression through hyper-methylation of fibulin 3 gene regulatory sequences, leading to changes in migration and invasion. Moreover, p38␣ would do so through the up-regulation of DNMT3A protein levels. According to this, re-introduction of DNMT3A in p38␣ Ϫ/Ϫ MEFs and p38␣ knock-down HCT116 cells downregulates Fibulin 3. The up-regulation of DNMT3A by p38 is in agreement with the previously shown p38-mediated increase in methyltransferase activity in response to anandamide (46).
As previously mentioned, it is known that dnmt3b mRNA, highly homologous to dnmt3a mRNA, is stabilized by the binding of HuR protein to its 3Ј-UTR (41). Therefore, we hypothesized that p38␣ would stabilize dnmt3a mRNA through HuR phosphorylation, as it happens with p21 mRNA (42). In that case, p38␣ MAPK would phosphorylate HuR, leading to cytoplasmic accumulation of HuR and enhancement of its binding to the 3Ј-UTR of the mRNA. Similarly, p38 MAPK mediated cytoplasmic accumulation of HuR stabilizes survival motor neuron mRNA (47). Other studies also support this hypothesis, but they involve the participation of additional proteins. For example, p38 MAPK via MK2 regulates the stability of other mRNAs such as TNF mRNA through regulation of HuR and tristetraprolin (TTP) (48). MK2 phosphorylates TTP, decreasing its affinity to the AU-rich element and its ability to replace FIGURE 9. Fibulin 3 and p38␣ knock-down inhibit tumorigenesis of HCT116 cells. In vitro and in vivo growth of HCT116 cells (non-silenced (Ϫ) and p38␣ knock-down (shp38␣), with (shFib3 (in A) and sh2Fib3 (in B)) or without fibulin 3 knock-down) was analyzed. HCT116 cells with a permanent (shFib3) or transient (sh2Fib3) fibulin 3 knock-down using two different human fibulin 3 shRNAs were used. A and B, anchorage dependent growth of HCT116 cells at 10 days (A) or at 3 days (B). Histograms show the mean Ϯ S.E. of foci number (expressed as the percentage of that of non-silenced cells at 14 d) and the percentage of those with a size Ն50 pixels. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. C, xenograft assay. Immunodeficient nude mice were injected subcutaneously with HCT116 cells. Histograms show the mean value Ϯ S.E. of tumor volume at the indicated time points (n ϭ 6). **, p Ͻ 0.01; ***, p Ͻ 0.001. D, anchorage independent growth of HCT116 cells at 14 days, in the absence or presence of SB203580 (5 M) as indicated. Histograms show the mean value Ϯ S.E. of the foci number expressed as the percentage of that from non-silenced cells. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001 as compared with non-silenced cells.
HuR, which allows HuR-mediated initiation of TNF mRNA translation. Hyperphosphorylation of TTP via p38 MAPK is also involved in the up-regulation of IL-8 and VEGF in malignant gliomas (49). In this line, our results support the involvement of HuR in the p38␣-mediated up-regulation of DNMT3A through dnmt3a mRNA stabilization. Accordingly, HuR knock-down highly decreases DNMT3A protein levels in wt MEFs.
It is important to highlight the relevance of fibulin 3 gene silencing induced by hypermethylation of its regulatory sequences in cancer (34,37). This down-regulation of fibulin 3 is associated with poor prognosis in some tumors such as nonsmall cell lung cancinoma (34 -35). However, the mechanisms controlling fibulin 3 gene hypermethylation remain unknown. Therefore, the finding of p38␣ MAPK as a novel regulator of this process in normal (MEFs) and tumor cells opens new perspectives to fully characterize how fibulin 3 expression is controlled under physiological conditions and in cancer.
Although p38␣ MAPK down-regulates Fibulin 3 in both MEFs and HCT116 cells, the role played by Fibulin 3 in the control of migration and invasion is different in the two cell types. This agrees with previous data from the literature. For example, in glioma and pancreatic adenocarcinoma, Fibulin 3 is overexpressed, promoting migration and invasion (31)(32)(33). In contrast, in non-small cell lung cancer cell lines, Fibulin 3 is a negative regulator of invasiveness, so in those cell lines where fibulin 3 is silenced by promoter methylation, cells became highly invasive and expressed higher levels of MMP2 and 7 (34). Wnt/␤-catenin pathway activation also contributes to invasion (50). Fibulin 3 down-regulation also promotes epithelial to mesenchymal transition and self-renewal of lung cancer stem cells (35). On the other hand, although Fibulin 3 down-regula-tion in colorectal cancer was previously correlated with lymph node metastasis and poor survival (37), in the HCT116 colon carcinoma cell line we have demonstrated that fibulin 3 silencing decreases migration, invasion, and tumor growth.
MMPs appear to play a role in both, MEFs and HCT116 cells. In particular, there is a good correlation between MMP2/9 activities and the invasive capacity of HCT116 cells, suggesting their involvement, which was supported by the impairment of invasion by a broad spectrum MMP inhibitor. Moreover, p38␣ is a positive regulator of these MMPs, as described in other tumor cell lines (16).
Curiously, in the two cell models that we have studied, Fibulin 3 regulates p38␣ and/or p38␤ activity, but in an opposite way, and their effects appear to be dependent on this regulation. In MEFs, Fibulin 3 down-regulates p38␣ and p38␤ activation, which would limit migration and invasion. In contrast, in HCT116 cells, Fibulin 3 enhances p38␣/␤ activation, favoring cell migration/invasion. So, the low level of Fibulin 3 produced by HCT116 cells expressing p38␣ is enough to promote migration and invasion through p38␣. In the A549 lung carcinoma cell line, p38 is also activated by Fibulin 3, but its function has not been characterized (51).
Although there is not a straightforward explanation to the distinct regulation of p38 activation by Fibulin 3 in MEFs and HCT116 cells, there are a great number of differences between these two cell models that might account for this discrepancy. MEFs are non-tumoral embryonic mesenchymal cells from a murine origin, while HCT116 cells are epithelial tumor cells from a human origin. These differences can support an opposite response mainly based on the following reasons: (i) epithelial and mesenchymal cells express different proteins (i.e. E-cadherin and N-cadherin, respectively) and (ii) HCT116, as FIGURE 10. p38 MAPK down-regulates fibulin 3 expression leading to regulation of migration/invasion and tumor growth. Model showing that p38␣, ␥ and ␦ decrease fibulin 3 transcription, leading to low levels of secreted fibulin 3. p38␣-HuR-mediated DNMT3A up-regulation might be responsible for hypermethylation of regulatory sequences of fibulin 3 gene and its silencing. In MEFs, Fibulin 3 negatively regulates migration and invasion through p38␣/ p38␤ inhibition. In HCT116 cells, Fibulin 3 promotes migration/invasion and tumor growth through p38␣ activation, although p38␤ could contribute to it. At the same time, p38␣ limits fibulin 3 expression as a negative feed-back loop.
other tumor cells, present several genetic alterations. Thus, it should be highlighted that the colon carcinoma HCT116 cell line bears a mutation in codon 13 of the ras gene, which leads to the up-regulation of a number of signaling pathways such as Ras/ERKs, PI3K/Akt or even p38 MAPKs. In addition, TGF-␤1 and ␤2 are expressed by these cells, which would dysregulate additional pathways and gene expression.
It is also noticeable the role played by p38␣ promoting tumor growth of HCT116 cells in vitro and in vivo. This is in contrast with its pro-apoptotic function, previously identified in HCT116 cells treated with cisplatin (12) and with its tumor suppressor role in other tumor cell lines (8). However, in agreement with our results, increased levels of phosphorylated p38␣ have been also correlated with malignancy in various cancers (8) such as head and neck carcinoma (52), where p38␣ promotes tumor growth in vitro and in vivo. In addition, in a mouse model of colitis-associated tumor induction, p38␣ deficiency decreases cell proliferation and survival of colon tumors (53). Furthermore, the role of Fibulin 3 favoring tumor growth of non-silenced HCT116 cells also supports this pro-tumorigenic function of p38␣ and it correlates with p38␣ activation levels.
In conclusion, we have described for the first time that p38␣ down-regulates fibulin 3 expression through hypermethylation of regulatory sequences of the gene. p38␣ might do so through the p38␣-HuR-mediated up-regulation of DNMT3A. Depending on the cellular context, Fibulin 3 acts as either a positive or a negative regulator of migration and invasion (Fig. 10) through mechanisms involving p38␣/␤. In addition, Fibulin 3 also promotes tumor growth of HCT116 cells through a mechanism dependent on p38␣, which acts as a potent promoter of tumor growth. At the same time, p38␣ limits fibulin 3 expression, which could represent a negative feed-back loop (Fig. 10).