Post-Transcriptional Regulation of Cadherin-11 Expression by GSK-3 and β-Catenin in Prostate and Breast Cancer Cells

Background The cell-cell adhesion molecule cadherin-11 is important in embryogenesis and bone morphogenesis, invasion of cancer cells, lymphangiogenesis, homing of cancer cells to bone, and rheumatoid arthritis. However, very little is known about the regulation of cadherin-11 expression. Methodology/Principal Findings Here we show that cell density and GSK-3β regulate cadherin-11 levels in cancer cells. Inactivation of GSK3β with lithium chloride or the GSK3 inhibitor BIO and GSK3β knockdown with siRNA repressed cadherin-11 mRNA and protein levels. RNA Polymerase II chromatin immunoprecipitation experiments showed that inhibition of GSK3 does not affect cadherin-11 gene transcription. Although the cadherin-11 3′UTR contains putative microRNA target sites and is regulated by Dicer, its stability is not regulated by GSK3 inhibition or density. Our data show that GSK3β regulates cadherin-11 expression in two ways: first a β-catenin-independent regulation of cadherin-11 steady state mRNA levels, and second a β-catenin-dependent effect on cadherin-11 3′UTR stability and protein translation. Conclusions Cadherin-11 mRNA and protein levels are regulated by the activity of GSK3β and a significant degree of this regulation is exerted by the GSK3 target, β-catenin, at the level of the cadherin-11 3′UTR


Introduction
Cadherin-11, also known as OB-cadherin was first identified in mouse osteoblasts and is normally expressed in cells with a mesenchymal phenotype and of mesodermal origin, including the mesenchyme of the kidney and brain during development [1]. Cadherin-11 is also expressed in cartilage synoviocytes and is an important mediator of the synoviocyte reaction that characterizes rheumatoid arthritis [2,3]. The cadherin-11 knockout mouse exhibits bone and behavioral abnormalities and is resistant to induction of rheumatoid arthritis [2,3]. Additionally, cadherin-11 is expressed in several types of cancer including breast and prostate cancers, osteosarcoma, and colon cancer [4][5][6][7][8]. In these cancers cadherin-11 expression is associated with the most aggressive and most metastatic cancer cells.
The principle pathway associated with canonical Wnt signaling involves inhibition of GSK3. GSK3 is intricately involved in many pathways essential to cellular function. Other mechanisms that also negatively regulate GSK3b include GSK3b phosphorylation on Ser9 by Akt and inhibition by LiCl treatment and 6bromoindirubin-39-oxime (BIO) treatment [18][19][20]. In the absence of canonical Wnts, GSK3 binds and phosphorylates bcatenin, in a multi-protein complex including axin, adenomatous polyposis coli (APC), and protein phosphatase 2A. Phosphorylation by GSK3b results in b-catenin protein degradation [13][14][15].
In this study we use breast cancer cells with low levels of endogenous activated GSK-3 and b-catenin and prostate cancer cells with constitutively activated GSK-3 and b-catenin to show that, GSK3b regulates cadherin-11 expression in two ways. The first mechanism is through b-catenin-independent regulation of steady state cadherin-11 mRNA levels. The second mechanism involves a b-catenin-dependent effect on cadherin-11 39UTR stability and protein translation.

Results
Cell density and inhibition of GSK3b regulate cadherin-11 mRNA and protein expression MDA-MB-231 cells, a mesenchymal-like breast cancer cell line that expresses cadherin-11, have undetectable levels of other cadherins and low levels of activated b-catenin. In pilot experiments we noticed that cadherin-11 expression varied noticeably depending on the number of cells plated. Figure 1 shows that cells grown at higher cell densities had markedly increased cadherin-11 mRNA ( Figure 1A) and protein ( Figure 1B) expression on a per cell basis compared to those grown at low density. In Xenopus laevis, cadherin-11 is regulated by xWnt8, although it is not known if the cadherin-11 gene is a direct transcriptional target or if either the canonical or non-canonical pathways are involved [11]. To test if the canonical Wnt pathway regulates cadherin-11 in MDA-MB-231 cells, we treated cells with lithium chloride (LiCl), an inhibitor of GSK3b. LiCl treatment repressed cadherin-11 mRNA and protein levels and blocked the effects of increased cell density ( Figure 1A, B). Cadherin-11 protein levels were also inhibited by LiCl in three other cell-lines (BT549, Hs578t and PC-3 Figure 1C). Remarkably, N-cadherin, which is endogenously expressed by these cell-lines and not by MDA-MB-231 cells, was unaffected by LiCl, strongly indicating that inhibition of GSK3b specifically regulated cadherin-11 expression. If the effects of LiCl are due to transient inhibition of GSK3b we anticipated that removal of LiCl should reverse its effects on cadherin 11. Figure 1D shows that this is indeed the case as cadherin-11 levels returned to control values 12 hours following LiCl washout. To confirm that the effects of LiCl were related to inhibition of GSK3b, we treated cells with 6-bromoindirubin-39oxime (BIO), a small molecule kinase inhibitor selective for GSK3a/b ( Figure 1E, F). Inhibition of GSK3b with BIO stabilized b-catenin, and like LiCl, reduced cadherin-11 mRNA and protein accumulation. Furthermore, siRNA directed against GSK3b mRNA decreased cadherin-11 protein expression in the absence of either LiCl or BIO in MDA-MB-231 but not in PC-3 cells ( Figure 1G). Therefore, GSK3b knockdown had little effect on cadherin-11 levels in PC-3 cells ( Figure 1G). Unlike MDA-231 cells, PC-3 cells have high levels of nuclear activated b-catenin as a result of a PTEN mutation and constitutive inactivation of GSK3b [21]. Consequently, removal of GSK3b in these cells may have less effect as it is already inactivated.

Recruitment of RNA polymerase II to the cadherin-11 gene is unaffected by
To test if GSK3b regulates cadherin-11 expression at the level of transcription, MDA-MB-231 cells were treated with the RNA polymerase II inhibitor, actinomycin D and were compared to the results of cells treated with LiCl. Figure 2A shows that cadherin-11 mRNA levels declined following either LiCl or actinomycin D treatment. However, in this experiment, and on other occasions, we noticed that LiCl treatment repressed cadherin-11 mRNA levels more rapidly than actinomycin D suggesting an additional non-transcriptional level of regulation. If the effects of GSK3b inhibition on cadherin-11 mRNA have a transcriptional component, we would anticipate that recruitment of RNA polymerase II to the cadherin-11 gene to decrease. To test this hypothesis, we carried out RNA polymerase II chromatin immunoprecipitation assays (Pol II ChIP) using cadherin-11 specific primers which amplify a portion of intron 1 (see Materials and Methods). Figure 2B shows that, as expected, actinomycin D markedly reduced the amount of Pol II recruited to both GAPDH and the cadherin-11 gene. However, even though cadherin-11 mRNA levels are markedly reduced by treatment of cells with BIO ( Figure 1) it did not affect Pol II recruitment to the cadherin-11 gene ( Figure 2B). In other experiments we found that members of the Snail/Slug family, transcriptional repressors known to be regulated by GSK3b and by b-catenin are not involved in cadherin-11 repression ( Figure S1) [22]. Taken together these experiments indicate that inhibition of GSK3b by LiCl or BIO influences cadherin-11 mRNA and protein levels by mechanisms largely independent of transcriptional repression.
The cadherin-11 39untranslated region (UTR) is highly conserved and destabilizing in cadherin-11 expressing cells In many situations regulation of mRNA stability and its capacity for translation are mediated by the 39-untranslated region (UTR). If the cadherin-11 39-UTR is important in the normal regulation of cadherin-11 expression we might expect it to be conserved across species. Table 1 shows the percent identity of the human cadherin-11 and E-cadherin transcripts and 39-UTRs compared with other species. Remarkably for a region that is not translated the cadherin-11 39-UTR was almost completely conserved across species (94% human with mouse). In contrast the E-cadherin-39UTR is poorly conserved (17% human with mouse). Analysis of the secondary structure of this region of both the cadherin-11 and E-cadherin 39UTRs revealed a high degree of stem loop structures ( Figure 3B and Figure 3C). These observations strongly suggest that the cadherin-11 39UTR has a conserved regulatory function.
The NCBI database identifies a region of the cadherin 39UTR which spans approximately 1 kb and includes a traditional poly-A signal and site. However, the Ensembl database identifies a cadherin-11 39UTR of roughly 3 kb which includes a second poly A signal 2363 bp from the stop codon ( Figure 3A). The poly-A signal sequences (AAUAAA) within the 39UTR are bolded in Figure 3A. Blue highlighted sequences indicate Shaw-Kamens sequences, also known as AU-rich elements (AREs [Shaw and Kamen, 1986]). Many microRNA binding elements also exist in the cadherin-11 39UTR (too numerous to indicate) as determined by ARGONAUTE 2 and miRanda. To determine the actual length of the cadherin-11 39UTR, total RNA was isolated from PC3 metastatic prostate cancer cells ( Figure 3D), MDA-MB-231 cells (data not shown), and MRC5 lung fibroblast cells (data not shown) for analysis. Primers were designed approximately every 500 bp of the Ensembl sequence to make progressively shorter PCR products. In all three cell lines primers designed to produce a PCR product of 1451 bp amplify a product, but the primers designed to make a product of 1952 bp do not ( Figure 3D). Although, this 501bp does not include a poly-A signal or addition site one does exist at position 2363 indicating that this longer transcript may be an alternate 39-UTR.
To investigate the function of the cadherin-11 39-UTR, it was cloned from MDA-MB-231 genomic DNA. A series of stop codons were included at the 59-end of the construct followed by the DNA sequence directly 39 to exon 13, the 39-UTR. Both the NCBI and Ensembl 39UTRs were cloned using the red highlighted primer sequences in Figure 3A into a luciferase reporter vector and named luc-CDH11 39UTR NCBI and luc-CDH11 39UTR Ensembl, respectively.
In cadherin-11 positive PC-3 and MDA-231 cells both luc-CDH11 39UTR NCBI and luc-CDH11 39UTR Ensembl were significantly less active than pGL3, which has the SV40 poly A addition signal and site ( Figure 4A). In HEK293 cells which do not express cadherin-11, luc-CDH11 39UTR Ensembl was more active than either control vector or luc-CDH11 39UTR NCBI ( Figure 4A). These data indicate that the both cadherin-11 39UTRs confer destabilization properties when expressed in cells which normally make cadherin-11. In contrast, in HEK293 cells, which do not normally make cadherin-11, luc-CDH11 39UTR Ensembl but not luc-CDH11 39UTR NCBI is stabilizing.

The activity of the cadherin-11 (UTR) is regulated by Dicer
To investigate if microRNA influenced the activity of luc-CDH11 39-UTR we knocked down Dicer, which is essential for the production of microRNAs. Removal of Dicer significantly increased the activity of luc-CDH11 39-UTR indicating that microRNAs regulate the endogenous activity of the cadherin-11 39-UTR ( Figure 4B). Consistent with a role for miRNA, several putative miRNA binding sites exist in this region, however, none of these predicted miRNAs was regulated by inhibition of GSK3b ( Figure S2). These data demonstrate that although miRNAs can regulate luc-CDH11 39-UTR this is not the mechanism whereby GSK3b inhibition regulated cadherin-11 mRNA levels. Unlike siRNAs, which always direct mRNA degradation, miRNAs often regulate protein translation without affecting mRNA levels [23][24][25]. To test this we examined the effects of Dicer knockdown on cadherin-11 protein and mRNA levels. Figures 4C and 4D show that even in the face of Dicer knockdown, endogenous cadherin-11 mRNA and protein levels were unchanged and the repressive Figure 1. Density increases cadherin-11 and GSK3b inhibitors decrease cadherin-11 expression. A and B: MDA-MB-231 cells were plated at two densities, 50% and 90% confluency, called low and high, respectively. The cells were allowed to grow overnight in serum-free medium. 24 hours after plating cells were treated with 20 mM NaCl (N) or 20 mM LiCl (L). 24 hours after treatment, RNA and protein were collected for realtime PCR (A) and Western (B) analysis. C: Cancer cells BT549, Hs578T, PC-3, and Kato-III were plated at a medium density and allowed to adhere. Cells were then treated with 20 mM LiCl or NaCl control. 48 hours after transfection protein was collected for Western blot analysis. D: MDA-MB-231 cells were plated at a medium density, serum starved overnight, and treated with 20 mM NaCl or 20 mM LiCl. 24 hours after treatment, the cells were washed once with PBS and maintained in serum-free medium containing only 20 mM NaCl. RNA was collected at the times indicated for real-time PCR analysis. E and F: MDA-MB-231 cells were plated at a medium density and serum starved overnight. 16 hours after plating, the cells were treated with 1 mM meBIO (control), 1 mM BIO, or 20 mM LiCl. RNA (E) or protein (F) was collected at the designated times and analyzed using real-time PCR (E) and Western blot analysis (F). G: MDA-MB-231 and PC-3 cells were transfected with either non-specific scrambled siRNA (siScramble), or siRNA directed against GSK3b (si GSK3). 48 hours after transfection protein was isolated for Western blot analysis. doi:10.1371/journal.pone.0004797.g001 effects of LiCl were not reversed. These data indicate that although the basal activity of the cadherin-11 39UTR is affected by Dicer knockdown, cadherin-11 mRNA and protein levels are not.
b-catenin regulates the activity of the cadherin-11 39UTR, steady state mRNA levels and cadherin-11 protein translation in PC-3 cells Perhaps the best known substrate of GSK3b is b-catenin and many studies have shown that Wnt-mediated inhibition of GSK3b stabilizes and activates cytoplasmic b-catenin [17,26]. Although the effects of inhibition of GSK3b on cadherin-11 are not mediated transcriptionally it is possible that stabilization of bcatenin may repress cadherin-11 mRNA and/or protein levels using an alternative mechanism. To test this PC-3 cells, which normally express high levels of activated b-catenin, were transfected with luc-CDH11 39UTR Ensembl or luc-CDH11 39UTR NCBI and treated with b-catenin siRNA [21]. Like Dicer knockdown, knockdown of b-catenin significantly increased the activity/stability of both cadherin-11 39UTR reporters ( Figure 5A). However, in contrast to Dicer knockdown, removal of b-catenin significantly increased cadherin-11 protein and steady state mRNA levels in PC-3 cells ( Figure 5C).

Discussion
The majority of patients dying from breast or prostate cancer have metastases to the skeleton. Although bone metastases are incurable, patients survive several years suffering serious morbidity, including fractures, spinal cord compression, severe bone pain, and hypercalcemia [5,27]. In the adult, cadherin-11 is strongly expressed in bone as well as certain metastatic cancers, particularly those inclined to metastasize to bone [4,6,27]. Once in the bone, it is possible that these cadherin-11-expressing tumor cells activate either osteoclasts or osteoblasts, depending on the type of cancer metastasis, leading to bone remodeling. As the bone is remodeled, growth factors are released into the matrix further stimulating the tumor cells by creating a fertile environment for growth. These changes are reminiscent of those that occur during rheumatoid arthritis, a disease that is also characterized by an important role for cadherin-11 in the activated synoviocyte [2,3]. Given its role in these important diseases we sought to determine how cadherin-11 expression is regulated. In the present study we show that cadherin-11 mRNA and protein levels are increased by high cell density and are markedly regulated by the activity of the ubiquitous serine/threonine kinase GSK3b. Remarkably, a significant degree of regulation is exerted by the GSK3b target, b-catenin, at the level of the cadherin-11 39UTR.
A role for GSK3b and b-catenin in cadherin-11 mRNA and protein homeostasis The best studied mechanism by which GSK3b is regulated is Akt-dependent serine phosphorylation. Growth factors, such as insulin, activate the phosphatidylinositol 3-kinase (PI3K)/Akt pathway. PI3K converts PIP 2 to PIP 3, a critical step necessary for the activation of PKB/Akt. PTEN is a phosphatase commonly mutated in cancers that is necessary for the dephosphorylation of PIP 3 thereby inhibiting Akt activation. Once activated Akt acts on many cellular proteins including mTOR, caspase 9, Bad, IKK, and GSK3b. Akt phosphorylates GSK3b on serine 9 and GSK3a on serine 21 to inhibit their kinase activity. Understanding GSK3b regulation by the Akt pathway is important to the present study because a PTEN mutation in PC-3 prostate cancer cells results in activated Akt, GSK3b inhibition, and activation of endogenous bcatenin [28]. Importantly, GSK3b is central to the promotion of inflammation in various inflammatory diseases including colitis and arthritis [29,30]. The second well-defined mechanism of  GSK3b regulation is the canonical Wnt pathway in which inhibition of GSK3b and stabilization of b-catenin results in marked changes in gene expression. Although many studies have observed that Wnt-GSK3b-mediated activation of b-catenin results in accumulation of nuclear and cytoplasmic b-catenin, most studies concentrate on the function of b-catenin as a transcriptional co-activator of TCF, and more recently nuclear receptors, in the nucleus [31][32][33]. However, b-catenin has significant homology to the RNA binding protein pumilio and several recent studies have pointed to an important role for bcatenin in post-transcriptional regulation [34]. For example, bcatenin regulates VEGF-D, cyclooxygenase-2 and cyclin D1 mRNA stability potentially by interaction with their respective 39-UTRs [35][36][37][38]. b-catenin can also regulate alternative splicing of estrogen receptor b (ERb) in colon cancer cells [35][36][37][38]. The ribosome-associated factor eIF6 associates with b-catenin and the complex may have a role in translational regulation in concert with miRNAs [39,40]. However, although the endogenous activity of the cadherin-11 39-UTR is regulated by Dicer, cadherin-11 mRNA and protein regulation by GSK3b are not, ruling out a role for miRNAs in this process. Our data show that the activity of the cadherin-11 39-UTR is regulated by b-catenin in cells which normally express activated b-catenin such as PC-3 prostate cancer cells. Taken together these data point to an important role for GSK3b and b-catenin in cadherin-11 mRNA and protein homeostasis and may lead to the development of therapeutics for diseases such as, metastatic prostate and breast cancer and rheumatoid arthritis, that are characterized by elevated cadherin-11. Indeed inhibition of GSK3b or ablation of the GSK3b gene ameliorates inflammation dependent arthritis [29].

Expression Vectors
Plasmid DNA encoding wild type (WT) b-catenin was used as previously described and DNA encoding wild type Snail was a kind gift from Dr. Mien-Chie Hung [15,22]. Dominant negative (DN) Snail was generated by PCR from the wild type Snail construct using primers described in Yamasaki et al.  Real time quantitative PCR mRNA. Relative quantitation was used to evaluate the raw data obtained from real-time PCR (7900 HT real time PCR system, Applied Biosystems, Foster City, CA). Single-stranded cDNA was prepared using TaqMan Reverse Transcription Reagents (N808-0234, Applied Biosystems) following the manufacturer's protocol. TaqMan Universal PCR Master Mix (4304437, Applied Biosystems) was used for all reactions. All primer/probe mixes (CDH11 Hs00156438_m1, GAPDH Hs99999905_m1) were obtained from Applied Biosystems and performed in triplicate. The samples were analyzed using the delta-delta Ct method of analysis [43]. The final value obtained was a measure of the fold change in gene expression for the particular gene of interest between the treated sample and the untreated sample. Experiments were run in triplicate, and for all analyses a p-value of ,0.05 was considered to be statistically significant.
microRNA. cDNA was generated by TaqMan MicroRNA Reverse Transcription kit (4366596, Applied Bioscience) following manufacturer's protocol. Primer/probe mixes specific for each microRNA (RNU6B, 4373381; hsa-miR-19a, 4373099; hsa-miR-27a, 4373287; hsa-miR-33, 4373048; hsa-miR-101, 4373159; hsa-miR-133b, 4373172; hsa-miR-337, 4373044; hsa-miR-424, 4373201) were obtained from Applied Biosystems (Foster City, CA). TaqMan Universal Master PCR Mix (4304437, Applied Biosystems) was used for all reactions. All experiments were performed in triplicate. The average value of the triplicate readings for each unknown was normalized to the corresponding value for U6 RNA. The samples were analyzed using the deltadelta Ct method of analysis [43]. The final value obtained was a measure of the fold change in gene expression for the particular gene of interest between the treated sample and the untreated sample. For all analyses a P-value of ,0.05 was considered to be statistically significant.

Luciferase Assay Analysis
Cells were plated at 2610 5 in a 12-well plate in growth medium. Cells were transfected in triplicate with 1.8 ug of luciferase plasmid DNA and 0.2 ug of pCMV-Renilla (Promega). For cells treated with anti-Dicer siRNA, 100 nM siRNA was transfected together with Luciferase and Renilla using Amaxa nucleofection and collected and analyzed 48 hours later. Cells at varying densities were transfected using Amaxa nuclefection and 1.67610 5 cells (low) or 6.25610 5 cells (high) were plated in triplicate in a 12-well plate. Cells treated with meBIO or BIO (1 mM) were transfected in triplicate using Fugene6 (Roche). 24 hours after transfection, cells were treated with either BIO or meBIO. 48 hours after treatment cells were collected. In all cases, cells were lysed using passive lysis buffer provided with the Dual-Luciferase Reporter Assay System (Promega). 20 ml of each sample was loaded in duplicate in a 96well plate and analyzed using the Dual-Luciferase Reporter Assay System reagents (E1960, Promega). The plate was read on a Berthold MicroLumat Plus LB 96V (Germany) using WinGlow software (Berthold).