DNAJB6 governs a novel regulatory loop determining Wnt/β-catenin signaling activity

Abstract

Dickkopf 1 (DKK1) is a secreted inhibitor of the Wnt signaling pathway and a critical modulator of tumor promotion and the tumor microenvironment. However, mechanisms regulating DKK1 expression are understudied. DNAJB6 is an HSP40 family member whose expression is compromised during progression of breast cancer and melanoma. Inhibition of Wnt/β-catenin signaling pathway, by up-regulation of DKK1, is one of the key mechanisms by which DNAJB6 suppresses tumor, metastasis and epithelial-mesenchymal transition (EMT). Analysis of the DKK1 promoter to define the cis-site responsible for its up regulation by DNAJB6 revealed the presence of two binding sites for a transcriptional repressor, MSX1 (muscle segment homeobox gene). Our investigations showed that MSX1 binds the DKK1 promoter and inhibits DKK1 transcription. Interestingly, silencing DNAJB6 resulted in up-regulation of MSX1 concomitant with increased stabilization of β-catenin. ChIP studies revealed that β-catenin binds MSX1 promoter and stabilization of β-catenin elevates MSX1 transcription, indicating that β-catenin works as a transcription co-activator for MSX1. Functionally, exogenous expression of MSX1 in DNAJB6 expressing cells promotes the mesenchymal phenotype by suppression of DKK1. Thus we have identified a novel regulatory mechanism of DNAJB6 mediated DKK1 transcription up-regulation that can influence epithelial-mesenchymal transition. DKK1 is a feedback regulator of β-catenin levels. Thus our studies also define an additional negative control of this β-catenin-DKK1 feedback loop by MSX1, which may potentially contribute to excessive stabilization of β-catenin.

Introduction

Wnt/β-catenin signaling regulates events central to development and cellular homeostasis. Due to its critical involvement in complex signaling that regulates cell proliferation, migration, differentiation and stem-ness, perturbations in β-catenin levels can manifest as diverse cellular phenotypes. Hence carefully controlled degradation of β-catenin is essential for optimal cellular function. Accidental stabilization of β-catenin can promote malignant progression. This happens either by overexpression of pathway activators such as Wnt1 overexpression in non-small cell lung cancer [1] or mutations that cause truncations in adenomatous polyposis coli (APC) (C terminal loss of function mutations) or degradation resistant forms of β-catenin that cause stabilization of the β-catenin protein such as in colon cancer or hepatocellular carcinoma [2, 3]. However in other cancers, such as breast cancer, mutations in the players of this pathway are rarely observed. Notably, inactivation of antagonists of the Wnt signaling pathway can misbalance the controlled degradation of β-catenin leading to aberrant promotion of malignant progression [4, 5]. Thus understanding the mechanisms of regulation of antagonists of the Wnt/β-catenin pathway is of great importance.

Dickkopf 1 (DKK1) is one of the key secreted antagonists of Wnt/β-catenin pathway which signals the cells for degradation of β-catenin. There are several reports of DKK1 upregulation or over-expression causing suppressed tumor growth [6-11]. Conversely, downregulation or loss of DKK1 expression has been documented in a number of clinical studies involving breast cancer [12], melanoma [13] and colon cancer [14]. Interestingly DKK1 transcription is upregulated by β-catenin and this is thought to be a negative feedback loop [15]. However other significant details of DKK1 regulation are yet to be revealed.

Our work on understanding the role of HSP40 family member, DNAJB6, in cancer biology has revealed that it inhibits tumor growth, epithelial-mesenchymal-transition (EMT) and metastasis by suppressing Wnt/β-catenin signaling [6, 16]. We observed that DNAJB6 up-regulates DKK1, hence we undertook a detailed investigation to understand the underlying molecular mechanisms and functional significance of DKK1 up-regulation. Our studies reveal that DNAJB6 down-regulates, MSX1 (muscle segment homeobox 1, a.k.a. HOX7); a transcription repressor that down-regulates DKK1 expression. We also show that MSX1 is a downstream target of β-catenin. Thus we describe a novel regulatory loop of β-catenin, MSX1 and DKK1 that is regulated by DNAJB6, which may be aberrantly activated upon loss of DNAJB6 and may lead to tumor growth, EMT and metastasis.

Materials and Methods

Cell Culture

MCF10A and MCF10CA1d.cl.1 (obtained from Karmanos Cancer Institute) were cultured in DMEM/F-12 with 5% horse serum (Invitrogen, Carlsbad, CA), 100 ng/ml cholera toxin (Calbiochem, Merck KGaA, Darmstadt, Germany), 10 μg/ml insulin (Sigma-Aldrich, St. Louis, MO), 500 ng/ml hydrocortisone (Sigma), and 25 ng/ml epidermal growth factor (Sigma). MCF7 cells were cultured in DMEM/F-12 containing 5% horse serum and insulin (10 μg/ml). SUM159 cells were cultured in Ham's F-12 with 5% fetal bovine serum supplemented with insulin (5 mg/ml) and hydrocortisone (1 mg/ml) and was purchased from Asterand plc, Detroit, MI. MDA-MB-435 cells were cultured in DMEM/F-12 with 5% fetal bovine serum (Atlanta Biologicals, Atlanta, GA), 1% nonessential amino acids, 1mM sodium pyruvate. MDA-MB-435 cells stably transfected with DNAJB6 and the corresponding vector control cells were cultured as above in medium supplemented with 500 μg/ml G418 (Invitrogen). All cells were maintained at 37°C with 5% CO₂ in a humidified atmosphere.

Plasmid constructs

pDKK1-535 and pDKK1-122, are luciferase constructs encompassing regions −535 to +43 base pairs and −122 to +43 base pairs (respectively) of human DKK1 promoter respectively and were a gift from the Varmus laboratory [15]. 3×MSX1 reporter vectors are luciferase constructs containing the two MSX1 binding sites (Sites A and B) from the DKK1 promoter respectively. Each site is cloned independently in tandem triplets into pGL3-Promoter vector (Promega BioSciences, Inc., Madison, WI) between KpnI and XhoI. Constructs with Site A deletion, site B deletion and deletion of both sites A and B, were generated using inverse PCR from pDKK1-535. pCMV6-XL4 vector and MSX1 were obtained from OriGene Technologies, Inc. (Rockville, MD).Various primers used for constructs are listed in Supplemental Table 1.

Luciferase Reporter Assay

Cells were transfected using Lipofectamine 2000 (Invitrogen) as per the manufacturer's instructions. Total protein was harvested and luciferase activity measured using a Turner 20/20 luminometer (Turner Biosystems, Sunnyvale, CA). The luciferase reading was normalized to the total protein concentration as reported previously [6, 7].

RNA Isolation and Real Time Quantitative PCR (qRT-PCR)

RNeasy mini kit (Qiagen, Valencia, CA) was used to isolate total RNA from cells. cDNA synthesis was carried out using a cDNA synthesis kit (Applied Biosystems Inc., Foster City, CA) with 1μg of total RNA as the template and random primers. qRT-PCR analysis was performed on the experimental mRNAs. The PCR primers and probes for DKK1, DNAJB6, MSX1 and normalization control gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were purchased from Applied Biosystems Inc. qRT-PCR was performed using BioRad iQ5 Real-Time Detection System (Bio-Rad, Hercules, CA). The gene expression ΔCt values of mRNAs from each sample were calculated by normalizing with GAPDH and relative quantitation values were plotted using GraphPad Prism® (La Jolla, CA).

Immunoblotting

Cells were washed twice with CMF-PBS and lysed with cold lysis buffer (150 mM NaCl, 50 mM Tris, 1% Nonidet P-40, and protease and phosphatase inhibitors). The lysates were kept on ice for 1 hour and centrifuged at 10,000 rpm for 30 min at 4°C. Protein concentration was measured using Precision Red protein assay reagent (Cytoskeleton, Denver, CO). Lysates corresponding to equal amount of protein were subjected to SDS-PAGE and transferred to PVDF membrane (0.2 μm). The membrane was blocked with 5% skimmed milk in TBST (1 M Tris, pH 7.5, 9% NaCl, 0.5% Tween 20) and incubated with primary antibodies overnight at 4 °C. The membrane was then washed three times with TBST, incubated with respective secondary antibodies for 1 hour at room temperature, washed three times with TBST and then developed using SuperSignal™ (Pierce) following washes.

DNAJB6 rabbit polyclonal (MO1) antibody (Abnova Corp., Taipei City, Taiwan) was used (1:5000) with 5% milk in PBS containing 0.2% Tween 20. β-catenin antibody (Cell Signaling Technologies (Danvers, MA) was used at (1:1000). Horseradish peroxidase-β-actin (Sigma) was used at (1:50,000). DKK1 antibody (Santa Cruz Biotechnology, Inc.) was used at (1:500). MSX1 antibody (Sigma) was used at (1:1000).

siRNA Knockdown

Cells were transiently transfected with MSX1 siRNA (sc-43945), designed to target human MSX1 or nontargeting siRNA control (sc-36869) (Santa Cruz Biotechnology, Inc.) using Lipofectamine 2000 (Invitrogen) following the manufacturer's protocol. RNA was extracted 30 hours after transfection and assessed for MSX1 knockdown.

Chromatin Immunoprecipitation (ChIP)

ChIPs were performed using the ChIP-IT Enzymatic Express kit (Active Motif, Carlsbad, CA) as per the manufacturer's protocol. Primers were designed to the two MSX1 binding sites on the DKK1 promoter, a non-specific region downstream from the MSX binding sites on the DKK1 promoter, the two β-catenin/TCF binding sites on the MSX1 promoter and a non-specific region upstream of the MSX1 promoter. The set of primers targeted to the first MSX1 binding site (Site A) encompass the region −608 to −407 of the human DKK1 promoter and generates a 201 base pair fragment. The set of primers targeted to the second MSX1 binding site (Site B) encompass the region −287 to −86 of the human DKK1 promoter and generates a 201 base pair fragment. The set of non-specific primers amplifies region +541 to +739 of the human DKK1 promoter (region does not bear a predicted MSX1 site) and generates a 198 base pair fragment. These primers are listed in Supplemental Table 2. To verify β-catenin binding sites on the MSX1 promoter, primers specific to the first β-catenin binding site (Site 1) encompass the region −5487 to −5306 of the human MSX1 promoter were designed (PCR yields a 181 base pair produce). The set of primers that targeted to the second β-catenin binding site (Site 2) amplifies region −5571 to −4682 of the human MSX1 promoter and generates a 184 base pair fragment. The set of non-specific primers amplifies region −1407 to −1222 of the human MSX1 promoter (region does not bear a predicted β-catenin site) and generates a 185 base pair fragment. These primers are listed in Supplemental Table 3.

3D culture

8-well chambered glass slides (Nunc, Rochester, NY) were coated with 3D Culture Matrix™ Basement Membrane Extract Reduced Growth Factor (Phenol red free) (Trevigen, Gaithersburg, MD). 5000 cells/well were seeded into the wells in complete media containing 2% 3D Matrix. Where indicated, recombinant DKK1 (750ng, R&D Systems, Inc., Minneapolis, MN) was added to the appropriate wells along with the cells and replenished every 2 days. The slides were incubated at 37°C, humidified with 5% CO₂ atmosphere. Media was replaced every 4 days.

Invasion assay

The inserts from BioCoat™ Matrigel Invasion Chambers (BD Biosciences, San Jose, CA) were hydrated with serum free media for 1.5 hr. 10μg/ml fibronectin in serum free media was used as an attractant in the lower chamber. 20,000 cells in serum free media were introduced into the top chamber (per well). Chambers were incubated at 37°C, 5% CO₂ atmosphere for 18 hr. One experimental set was treated with recombinant DKK1 (750ng). Upon termination of the assays, cells were fixed using 4% paraformaldehyde, stained using crystal violet and enumerated.

Soft agar colonization assay

Tissue culture wells were coated with a layer of hard agar (0.75% bactoagar in complete media containing 10% serum). Cells (2×10³) were mixed with soft agar (0.35% bactoagar in complete media containing 10% serum) and overlaid on the hard agar. The agar was always maintained under a layer of complete media. Visible colonies (> 50cells) were stained with crystal violet and counted with the aid of a dissecting microscope.

Foci formation assay

Cells were transfected using Lipofectamine 2000 (Invitrogen). Serum was added to the cells 6 hours after transfection. After 36 hours of transfection, cells were detached and plated into three 10 cm plates (100 cells/plate) with selective media and grown for 14-18 days. Foci were fixed to the plate using ice cold 100% methanol. Crystal violet staining was used for visualization of foci.

In vivo Tumor growth

DKK1 cDNA was cloned in pIRES2-EGFP. SUM 159 cells were transfected with pIRES2-EGFP-DKK1 plasmid and SUM159 cells overexpressing DKK1 were generated by selecting the stable mixed population that was resistant to G418. The in vivo tumor growth assays were performed as described before [7]. Briefly, cells at approximately 80% confluence were washed with chilled CMF-DPBS, and re-suspended in ice-cold Hank's Balanced Salt Solution (Invitrogen) to a final concentration of 107 cells/ml. Cells (1×106/100 μl) were injected into the third mammary fat pad of 6 week old, female athymic mice (Harlan Sprague-Dawley, Indianapolis, Indiana). Tumor growth was monitored twice weekly and mean tumor diameter calculated by taking the square root of the product of orthogonal measurements. Experiment was terminated after the mean tumor diameters reached 1 cm. Animals were maintained under the guidelines of the National Institute of Health and under evaluation and approval of IACUC (University of South Alabama). Food and water were provided ad-libitum.

Results

DKK1 suppresses tumor growth

DKK1 is a secreted inhibitor of Wnt/β-catenin pathway and hence will be expected to have tumor growth suppressive effects. Certainly this is observed in breast cancers under experimental conditions [8-10, 17, 18]. However some reports suggest a negative etiologic role for this molecule [19-21]. Hence to gain a better understanding of the role of DKK1 in breast cancer, we established SUM159 cell line constitutively expressing DKK1. We observed a dramatic reduction (P<0.0001) in the growth of xenografts of these cells in nude mice as compared to their vector controls. Tumor incidence as well as growth rate was notably decreased in DKK1 expressing SUM159 cells (P<0.0001) (Figure 1A & B). These observations emphasized the need to investigate molecular mechanisms that regulate DKK1 expression.

Figure 1. DKK1 suppresses tumor growth.

SUM159 cells overexpressing DKK1 were compared with the corresponding vector controls for tumor growth rates in athymic mice. Cells (1×10⁶ in 0.1 ml) were injected in the mammary fat pad. 8 mice were used per group and experiment was repeated once.

A. Tumor measurements were taken twice a week. The data represents the mean tumor diameter +/− standard error of mean. The inset shows immunoblot analysis for the serum free conditioned medium from equal number of cells that was performed to verify the level of secreted DKK1 prior to injection.

B. Percent tumor incidence was recorded at respective time point.

The region of DKK1 promoter responsive to DNAJB6 has two putative MSX1 binding sites

Secreted proteome analysis of DNAJB6 expressing cell lines showed increased DKK1 levels [6]. DNAJB6 itself has no known DNA-binding domain. Hence we hypothesized that it could mediate the up-regulation of DKK1 via modulation of some other transcription factor(s). In order to gain insight into the molecular mechanism by which DNAJB6 up-regulates DKK1, we performed luciferase reporter assays using DKK1 promoter-luciferase reporter constructs pDKK1-535 and a deletion pDKK1-122. pDKK1-535 was activated about 3-fold by DNAJB6; this activation was lost in pDKK1-122 (Figure 2A [I]). A survey of transcription factor binding sites in the 400 bp deleted region revealed two putative MSX1 binding sites. We will refer to these two MSX1 binding sites as ‘Site A’ and ‘Site B’ (Figure 2A [II]). We also analyzed various breast cancer cells for their steady state levels of DNAJB6, MSX1 and DKK1 and found an interesting correlation. While the transcript levels of endogenous DNAJB6 and DKK1 decreased, levels of MSX1 concomitantly increased in breast cancer cell lines as compared to the immortalized mammary epithelial cell line, MCF10A (Figure 2B). Furthermore, DNAJB6 expressing cells show significantly elevated DKK1 levels in serum free conditioned medium, concomitant with significant reduction in MSX1 protein expression (Supplementary Figure 1). This suggested that MSX1 could be a candidate negative regulator (repressor) of DKK1.

Figure 2. MSX1 binds to DKK1 promoter and represses its transcription.

A. [I] In MDA-MB-435 cells, luciferase activity of DKK1 promoter p535 was compared, in presence or absence of co-transfected DNAJB6, with deletion p122 or deletions of Site A or Site B or simultaneous deletion of both sites. The assay was performed in triplicate and experiment was repeated twice.

[II] Diagrammatic representation of location of MSX1 binding sites in DKK1 promoter. The table shows comparison of Site A and Site B with the consensus MSX1 binding site.

B. mRNA levels of DKK1, DNAJB6 and MSX1 from various breast cancer cell lines were compared with immortalized mammary epithelial line MCF10A. Assay was performed in triplicate and repeated once for confirmation. (* indicates P<0.05)

C. Luciferase reporter vector with three tandem repeats of MSX1 Site A or B from DKK1 promoter were assessed for their activity in presence of exogenous MSX1 and compared with corresponding empty vector transfectants in MCF10A cells.

D. The same reporters were assessed MDA-MB-435 cells in presence of exogenous DNAJB6 and compared with corresponding empty vector transfectants.

E. Chromatin immunoprecipitation was performed in DNAJB6 expressing and vector only transfected MDA-MB-435 cells, to test the binding of MSX1 to Site A and Site B. The immunoprecipitated DNA was queried using primers for a negative control, a non-specific primer set and primers specific to Site A or Site B. PCR was performed using DNA immunoprecipitated using Isogenic IgG control and water only as negative controls. The PCR from input DNA served as positive control. The products were analyzed by agarose gel electrophoresis.

Predicted MSX1 binding sites on the DKK1 promoter are responsive to MSX1 and DNAJB6

Each putative MSX1 binding ‘Site A’ or ‘Site B’, from DKK1 promoter, was cloned in tandem triplicate to obtain 3×MSX1-luciferase reporters for the respective sites and assessed for responsiveness to MSX1. We found that MSX1 repressed the luciferase activity of the 3× constructs of both the putative MSX1 binding sites on the DKK1 promoter (Figure 2C). This strongly suggested that the two putative sites are indeed responsive to MSX1.

When evaluated for their responsiveness to the presence of DNAJB6 (Figure 2D), we noted that constitutive expression of DNAJB6 caused about 2.5 to 3 fold increase in luciferase reporter activity of the respective 3×MSX1 binding site constructs. This suggests that the activation of DKK1 by DNAJB6 is potentially mediated through the same candidate cis-sites that are (potentially) used by MSX1 to repress DKK1 activity.

Furthermore, deletion of each of the sites, ‘Site A’ and ‘Site B’ individually and together from pDKK1-535, rendered DNAJB6 incapable of activating the DKK1 promoter (Figure 2A). This suggested potential importance of both the sites in repression of DKK1 by MSX1. We were able to amplify specific products encompassing individual sites (Site A or B) from the chromatin precipitate using anti-MSX1 antibody only in absence of exogenous DNAJB6 expression, confirming that MSX1 physically binds to both Sites A and B on the DKK1 promoter (Figure 2E and Supplementary Figure 2).

MSX1 is known to homodimerize [22]. It is possible that MSX1 binds to the DKK1 promoter as a homodimer. This may explain how deletion of one MSX1 binding site still results in an inhibition in transcription.

Silencing DNAJB6 up-regulates MSX1, which in turn, down-regulates DKK1

Based on our observations described thus far, we hypothesized that increased DNAJB6 levels could down-regulate the MSX1 (transcription repressor), leading to up-regulation (relief from suppression) of DKK1. Since MSX1 levels are practically undetectable in cell lines that express DNAJB6 (data not shown), to test our hypothesis, we analyzed the effects of silencing DNAJB6 from breast epithelial line MCF10A. As seen in Figure 3A [I] we observed an increase in MSX1 expression with a concomitant decrease in the expression of DKK1. Independently, we over expressed MSX1 in MCF10A cells and found that expression of DKK1 was suppressed (Figure 3B). Moreover, these results are supported by our independent observation that expression of MSX1 caused a marked decrease in the reporter activity of pDKK1-535 in MCF10A cells (Figure 3C). Cumulatively this established a direct link of MSX1 up-regulation with DKK1 suppression. Our next step was to understand the mechanism of down-regulation of MSX1 by DNAJB6.

Figure 3. MSX1 is a downstream target of β-catenin.

A. [I] DNAJB6 expression was silenced from MCF10A cells using shRNA, sh_807. mRNA levels of DNAJB6, DKK1 and MSX1 were quantitated using qRT-PCR. The results are expressed in fold change compared to vector control. The results are graphed in logarithmic scale.

[II] Lysate of DNAJB6 silenced from MCF10A (sh_807 treated) was analysed for DNAJB6 expression as well as β-catenin expression. β-actin was used to verify equal protein loading.

B. MCF10A cells were transiently transfected using MSX1 expressor vector. Change in DKK1 mRNA level was determined using qRT-PCR. The measurements were performed in triplicates and repeated once.

C. DKK1 Luciferase reporter activity (using p535) was measured in MCF10A cells transiently transfected with MSX1 expressor vector and compared with corresponding empty vector control. The measurements were performed in duplicate and repeated thrice.

D. (I) Diagrammatic representation of location of β-catenin binding sites in MSX1 promoter. The table shows comparison of Site 1 and Site 2 with the consensus β-catenin/TCF binding site (II) Chromatin immunoprecipitation was performed in MDA-MB-435 cells, to test the binding of β-catenin to Sites 1 and 2. The immunoprecipitated DNA was amplified using PCR performed using primers for a negative control, a non-specific primer set and primers specific to Site 1 or Site 2. PCR performed using DNA immunoprecipitated using isogenic IgG control and water only control served as negative controls. The PCR from input DNA served as positive control. The products were analyzed by agarose gel electrophoresis.

E. [I] TCF/LEF reporter activity was measured in MCF7 cells treated with 15mM LiCl. Activity was also measured without any treatment (No Rx) or NaCl treatment as a control.

[II] Under the same treatment conditions, MSX1 mRNA level in MCF7 cells was determined. The measurements were performed in triplicate and repeated once. Expression levels determined without any treatment or with just NaCl (15mM) served as controls.

MSX1 is a downstream target of β-catenin

We recently showed that DNAJB6 enhances the proteosomal degradation of β-catenin [6]. Silencing DNAJB6 from breast epithelial line MCF10A increases β-catenin levels (Figure 3 A [II]). Also, overexpression of DNAJB6 in MDA-MB-435 decreases β-catenin levels concomitant with noticeable reduction in MSX1 (Supplementary Figure 1). Murine MSX1 was shown to be a target of β-catenin/TCF transcription [23]. Hence, we hypothesized that the down-regulation of β-catenin/TCF transcription by DNAJB6 will contribute to the down-regulation of MSX1. Human MSX1 promoter region shows two potential β-catenin/TCF binding sites at positions −5370 and −4774 (Figure 3D [I]). ChIP assay results show that both the predicted sites were occupied by β-catenin. Absence of PCR product using independent primers (1407 base pair upstream, in a region that does not bear any predicted β-catenin sites) helped ensure specificity of our finding (Figure 3D [II]).

MCF7 is a Wnt responsive cell line (has β-catenin localized predominantly to membrane but capable of translocating to the nucleus in response to Wnt ligands) [24]. As seen in Figure 2B, MCF7 has a relatively high DNAJB6 expression as compared to metastatic breast cancer cells. Treatment of these cells with LiCl (activator of β-catenin signaling by inhibition of GSK3β) resulted in about 3 fold increase in TopFlash activity indicating enhanced β-catenin driven TCF/LEF activity (Figure 3E [I]). Concomitantly we noticed a 100% increase in MSX1 expression (Figure 3E [II]). These observations further demonstrated that up-regulation of MSX1 is a downstream response of β-catenin signaling.

MSX1 promotes mesenchymal phenotype by down regulating DKK1

To obtain an insight about functional relevance of MSX1, we evaluated the impact of restoring the expression of MSX1 in DNAJB6 expressors. As presented in Figure 4 we observed that upon co-expression of MSX1, DNAJB6 expressors gained a significant ability to grow in an anchorage independent culture (Figure 4A) and form foci (Figure 4B). Further more, MSX1 co-expressing cells had noticeable increase in invasive capabilities (Figure 4C). However, exogenous DKK1 expression could override the malignant characteristics imparted by restoring MSX1 and impede the invasion, foci formation and soft agar colony formation ability (Figure 4 A,B,C).

Figure 4. Effect of expression of MSX1 on attributes of malignancy.

A. MSX1 was ectopically expressed in 435-DNAJB6 cells and the soft agar colonization of these cells was compared to empty vector control. Independently DKK1 was co-expressed with MSX1 and colony formation of those cells was compared to the MSX1 only cells with empty vector corresponding to DKK1 expression construct. The experiment was performed in triplicate and repeated once.

B. The same experimental groups were analyzed for foci formation ability. The experiment was also performed in triplicate and repeated once.

C. The same experimental groups were analyzed for their ability to invade using BioCoat™ Matrigel Invasion chamber. The experiment was also performed in triplicate and repeated once.

D. MDA-MB-435 and SUM159 were analyzed for their 3-D morphology in 3-D Culture Matrix™ Basement Membrane Extract. Following groups were analyzed (I) DNAJB6-expressors with empty vector (II) DNAJB6-MSX1 expressors (III) DNAJB6-MSX1 expressors treated with recombinant human DKK1 (rhDKK1).

Independently, 3-D morphology of MDA-MB-435 or SUM159 cells silenced for MSX1 by treatment of siRNA (siMSX1) was compared with the corresponding control siRNA control (IV & V).

All experiments were performed in triplicate. Scale bars correspond to 100μM.

EMT phenotype is correlatively linked to invasive capability. Hence we compared DNAJB6 expressors to the corresponding constitutive MSX1 expressors. DNAJB6 expressors in highly mesenchymal cells viz. MDA-MB-435 and SUM159 show a compact spherical, circumscribed, epithelial-like growth in 3D culture. Upon constitutive overexpression of MSX1, the 3D growth pattern acquired irregular, invasive growth characteristics. However, addition of exogenous DKK1 showed a dramatic reversal to a compact, more epithelial like ascinar growth (Figure 4D I-III). Conversely, silencing MSX1 from the parental MDA-MB-435 and SUM159 cells caused a significant reversal to epithelial like morphology in 3D culture compared to the corresponding non-targeting siRNA controls (Figure 4D IV-V). Further more, immunoblotting performed to determine the effects of modulating MSX1 on molecular markers of EMT revealed that MSX1 up-regulated the levels of mesenchymal markers N-cadherin, TWIST and Zeb1 and their levels were decreased upon addition of recombinant DKK1. Keratin18, an epithelial marker showed opposite trend Supplementary Figure 3. We did not notice any specific change in Slug. Overall these observations indicate ability of MSX1 to promote a mesenchymal-like phenotype.

Discussion

A growing body of literature is suggesting that HSP40 family members are instrumental in fine-tuning critical cellular signaling that is also relevant to suppression of malignant progression and metastasis [25-30]. DNAJB6 normally shows constitutive expression (heat shock cognate protein) and thus it potentially contributes to the essential regulation of rapid turnover of β-catenin. We find that DNAJB6 reduced expression of MSX1 by negatively regulating β-catenin. MSX1 is involved in regeneration of amphibian limbs and supports the undifferentiated and proliferative state of embryonic cells [31] and has been demonstrated to be a transcriptional repressor [32]. Besides cell division/proliferation, development of an embryo involves fate determination and cell differentiation. All these processes require precise regulation of various growth factors and inhibitors in a coordinated fashion. DNAJB6, MSX1 and DKK1 are all involved in embryonic or mature organ development [15, 33-40]. Hence it is plausible that aberrant expression of these genes can bear catastrophic consequences. Mutations or abnormal expression of MSX1 gene may be associated with increased risk of cancers [41-43]. During development, MSX1 inhibits terminal differentiation of progenitor mesenchymal cells and its expression is associated with mesenchymal cells and tissue, proliferation and transformed phenotype [32, 44]. EMT and its reversal, mesenchymal–epithelial transition (MET) play important roles during embryonic development [45]. The EMT phenomenon is recognized as a contributor to invasive phenotype and tumor promotion [46-49]. We show that MSX1 imparts a mesenchymal-like phenotype to tumor cells. One of the contributors to this MSX1-driven phenotype is possibly the ability of MSX1 to repress the Wnt antagonist DKK1, which is consistent with the contributions of DKK1 in maintaining an epithelial like state as reported by us and others [6, 8, 50]. Our observations are in apparent contrast with study published by Revet et al. that describes MSX1 as inducer of Wnt antagonists including DKK1 [51]. We would like to note that this study was conducted in neuroblastoma cells. It may be possible that MSX1 could play different roles in different cell types. Interestingly Revet et al have also published that MSX1 activates delta-Notch pathway in neuroblastoma [52]. Based on certain other reports such as one by Katoh et al.[53], one may speculate that activated Notch signaling may potentially be able to activate DKK1 transcription by yet another mechanism that may be specific to neuroblastoma.

Besides being a well accepted secreted antagonist of the Wnt/β-catenin signaling, DKK1 is a downstream target of β-catenin transcription [15]. This implies a feed-back inhibitory regulation of β-catenin signaling, as the secreted DKK1 will in-turn compromise the stability of β-catenin. This however presented the following paradox: when any cell needs to activate β-catenin dependent transcription by stabilizing β-catenin, it will need to have an alternative regulator to keep the transcription of DKK1 insensitive to the potential upregulation by increased β-catenin levels.

The novelty of our finding is that we have unraveled MSX1 as the ‘alternative regulator’ of DKK1 transcription. We demonstrate that MSX1 is a downstream target of β-catenin transcription activity and acts as a transcription repressor of DKK1. This provides evidence for another level of regulation of DKK1 transcription by β-catenin. We have shown that DNAJB6 chaperones a multiprotein complex that maintains GSK3β in a de-phosphorylated active state [54]. Thus, in absence of DNAJB6 (during cancer progression), GSK3β is predominantly phosphorylated and hence is incapable of targeting β-catenin for degradation. The stabilized β-catenin in turn will up-regulate MSX1 and this MSX1 will be able to block the simultaneous up-regulation of DKK1 by β-catenin. This ensures the increased stability of β-catenin which contributes to the tumor progression, invasive-mesenchymal phenotype and metastasis. Overall, our findings have revealed a novel regulatory mechanism of fine-tuning Wnt signaling (Supplemental Figure 4).

Supplementary Material

Supplementary Figure 1: DNAJB6 expressors have lower levels of MSX1.

DNAJB6 expressing MDA-MB-435 cells were analyzed for levels of MSX1 protein. Protein from total cell lysate (30μg) was compared with same amount of protein from the vector control cells for the levels of MSX1 and β-catenin. Similarly serum free conditioned medium (SFM) from equal number of cells was analyzed for levels of DKK1.

The results indicate that DNAJB6 expression leads to increased DKK1 levels with concomitant decrease in β-catenin and MSX1.

Supplementary Figure 2: Silencing DNAJB6 allows MSX1 to bind to DKK1 promoter.

Chromatin immunoprecipitation was performed in MCF10A cells or MCF10A cells silenced for DNAJB6 expression using shRNA sh_807, to test the binding of MSX1 to Site A and Site B.

The immunoprecipitated DNA was queried using primers for a negative control, a non-specific primer set and primers specific to Site A or Site B. PCR was performed using DNA immunoprecipitated using Isogenic IgG control and water only as negative controls. The PCR from input DNA served as positive control. The products were analyzed by agarose gel electrophoresis.

Supplementary Figure 3: Constitutive MSX1 expression up-regulates mesenchymal markers in DNAJB6 expressors.

MSX1 cDNA was expressed transiently in MDA-MB-435-DNAJB6 expressors. Total protein (40μg) isolated from these cells was compared total protein from identical set of transfectant of MSX1 treated with recombinant DKK1 and cells transfected with corresponding empty vector control. The numbers below each panel are relative intensity of the bands compared to empty vector control (arbitrarily set at 1).

Supplemental Figure 4: Diagrammatic representation of effect of DNAJB6 on DKK1 expression.

In absence of DNAJB6 (during cancer progression), GSK3β is predominantly phosphorylated and hence incapable of targeting β-catenin for degradation. The stabilized β-catenin in turn will upregulate MSX1 and this MSX1 will be able to block the simultaneous upregulation of DKK1 by β-catenin. This ensures the increased stability of β-catenin which contributes to the tumor progression, invasive-mesenchymal phenotype and metastasis. We propose that in normal cells (cells with DNAJB6), most of the β-catenin is either membrane bound or has a very short half life (due to active GSK3β). These low levels of β-catenin are insufficient to upregulate MSX1. Thus in absence of MSX1 (one of the repression switches being removed), DKK1 up-regulation is feasible by other known transcription activators such as SP1. This will further ensure the tight regulation needed on β-catenin, a protein we need at the membrane to maintain the epithelial phenotype but too much of which tips the balance towards malignant progression.

Table 1: Oligomers used for generating reporter constructs

Table 2: Primers used for ChIP experiments to test the binding of MSX1 on DKK1 promoter

Table 3: Primers used for ChIP experiments to test the binding of β-catenin on MSX1 promoter

Abbreviations used

cDNA

complementary deoxyribonucleic acid

ChIP

chromatin immunoprecipitation

DKK1

dickkopf homolog 1

DMEM

Dulbecco's modified Eagle's medium

HSPs

heat shock proteins

HSP40

heat shock protein 40

DNAJB

sub-type B of the heat shock protein 40 family

DNAJB6

mammalian relative of DnaJ (MRJ)

FBS

fetal bovine serum

G418

geneticin

GAPDH

glyceraldehyde-3-phosphate dehydrogenase

GSK3β

glycogen synthase kinase 3β

HRP

horseradish peroxidase

IHC

immunohistochemistry

IgG

immunoglobulin G

Fzd

frizzled receptor

LRP5/6

low density lipoprotein receptor related protein 5/6

MSX1

muscle segment homeobox 1

PBS

phosphate buffered saline

PCR

polymerase chain reaction

PMSF

phenyl methyl sulfonyl fluoride

PVDF

polyvinylidene fluoride

qRT-PCR

quantitative real time – polymerase chain reaction

shRNA

short hairpin ribonucleic acid

siRNA

small interfering ribonucleic acid

SFM

serum-free media

TBST

tris-Buffered Saline Tween-20

TCF/LEF

T-cell factor/lymphoid enhancer factor

TOPFlash

Super8XTOPFlash luciferase reporter construct