FOXP3 promotes tumor growth and metastasis by activating Wnt/β-catenin signaling pathway and EMT in non-small cell lung cancer

Human NSCLC cell lines, A549 and NCI-H460, were obtained from the American Type Culture Collection (ATCC, Manassas, VA), and cultured in RPMI-1640 medium. Human embryonic kidney epithelial cell line 293T cell line was purchased from Invitrogen, and cultured in DMEM medium. Lipofectamine 2000 (Invitrogen, Carlsbad, CA) was used to transfect cells with plasmid DNA. Total RNA and proteins were extracted from 80% confluent cells in culture dishes.

One hundred and six patients with histologically-confirmed NSCLC who underwent surgery at Prince of Wales Hospital between 2001 and 2007 with complete clinicopathologic characteristics (Additional file 1: Table S1) and follow-up data were enrolled in the study. Histologic type was determined according to the 2015 WHO classification [17]. Immunohistochemistry was performed on formalin-fixed paraffin sections according to standard protocols using two types of primary antibodies from Santa Cruz and Abcam, The tumor FOXP3 staining intensities were scored using the immunoreactive score (IRS) method [4] by a pathologist and an investigator. Tumor infiltrating FOXP3⁺ regulatory T-cells (Treg cells) were also counted at the same time. Human studies were approved by the Joint CUHK-NTEC Clinical Research Ethics committee.

The anti-E-Cadherin (CellSignaling, Danvers, MA, 1:2000), anti-N-Cadherin (Cellsignaling, 1:2000), anti-Vimentin (Abcam, 1:1000), anti-Snail (Cellsignaling, 1:2000), anti-Slug (Abcam, 1:2000), anti-MMP9 (Cellsignaling, 1:2000), anti-β-actin (Santa Cruz, 1:3000), anti-c-Myc (Santa Cruz, 1:1000), anti-Cyclin D1 (Santa Cruz, 1:1000), anti-TCF4 (Santa Cruz, 1:1000; Cellsignaling, 1:2000)), anti-LaminB1 (Cellsignaling, 1:2000), anti-FOXP3 (Abcam 1:3000, Cellsignaling 1:2000), anti-β-catenin (Cayman, Michigan, 1:5000), anti-GFP (Santa Cruz, 1:4000) were used as the primary antibodies. A 1:3000–5000 dilution of the HRP-linked anti-IgG (Santa Cruz) was used as the secondary antibody.

Genes of interest were detected using SYBR® Premix Ex Taq ™ (Takara, Dalian, China) on Quantstudio™ 12k Flex Real-time PCR system (Applied Biosystems, Foster City, CA). The primer sequences are listed in Additional file 1: Table S2.

The full-length ORF (open reading sequence) of human FOXP3 gene (NM_014009.3) was PCR amplified from pcDNA3.1-FOXP3 and subcloned into the self-inactivating lentiviral vector PHIV-EGFP (Addgene) with an EF1-alpha promoter, an internal ribosome entry site (IRES) and EGFP. Lentiviral preparations were generated by transient transfection of HEK-293T cells by using pHIV-EGFP-FOXP3 (10 μg), pRSV-Rev. (2.2 μg), pMDLg/pRRE (4.72 μg), pMD2.G (3.08 μg).

FOXP3 shRNA was selected from the RNAi Consortium library (www.​broadinstitute.​org/​rnai/​public), which contains shRNAs against 15,000 human genes. We selected 5 highly scored target sequence against FOXP3 (NM_014009.3) mRNA sequence from the database: CCTCCACAACATGGACTACTT; CACACGCATGTTTGCCTTCTT; CTGAGTCTGCACAAGTGCTTT; TCCTACCCACTGCTGGCAAAT; TGTCCCTCACTCAACACAAAC. The corresponding oligoes generated by Invitrogen were subcloned into pLKO.1 (Addgene). Lentiviral preparations were generated as above except for that the transfection cocktail was replaced with pLKO.1 (6 μg), psPAX2 (4.5 μg) and pMD2.G (1.5 μg).

Cell viability was determined by the MTT assay (ThermoFisher, Waltham, MA), and the absorbance was measured at 570 nm.

One thousand cells were seeded and cultured for 10–14 days. Colonies (⩾50 cells/colony) were counted.

Difco™ Agar (BD Biosciences, Erembodegem, Belgium) was used in the assay, 1000 cells per well were added in top agar and were cultured for 2 weeks. Colonies (⩾50 cells/colony) were counted.

Scratch was made by a 200 μl pipette tip after cells were grown to 80% confluence. Cells were then incubated in medium containing 5% FBS. Gap size was measured 24 h later.

2.5 × 10⁴ cells were seeded in upper chamber with 300 μl medium containing 5% FBS. The lower chamber was filled with 1200 μl medium containing 10% FBS acting as chemotactic factors. After the incubation of 24–48 h, the migrated cells to the underside of the membrane were fixed and stained with 0.1% crystal violet.

The anti-E-Cadherin (Cellsignaling, 1:200) and anti-FOXP3 (Abcam, 1:100) were used as the primary antibodies. The cells were incubated with Alexa Fluor® 594 dye conjugated secondary antibody (ThermoFisher). The nucleus was stained by DAPI (ThermoFisher).

HEK293T cells were plated in 24-well plates and co-transfected with various plasmids as indicated in the figures. Cells were collected 24 h after transfection, and luciferase activities were analyzed by the dual-luciferase reporter assay kit (Promega, Madison, WI). Reporter activity was normalized to the control Renilla.

Total RNA was extracted from A549-FOXP3 and A549-Control using TRIzol reagent (Invitrogen). The marked cRNAs were hybridized with the Agilent human whole genome gene expression Microarray (Agilent Technologies, Santa Clara, CA). Gene expression levels were standardized by the level of GAPDH. Differentially expressed genes were screened by the threshold of 2.0 fold-change and p value that was more than 0.05. Student’s t test was adopted for statistical analysis. Pathway analysis and Gene Ontology (GO) analysis were applied to determine the functions of those differentially expressed mRNAs by GO (www.​geneontology.​gov) [18] and the KEGG (Kyto Encyclopedia of Genes and Genomes) pathway database (http://​www.​genome.​jp/​kegg/​pathway.​html).

Cells were resuspended in 600 μl ice-cold Buffer I (1.5 mM MgCl2, 10 mM HEPES, 10 mM KCl, and protease inhibitor cocktail, pH 8.0), incubated on ice for 15 min and rotated once every 5 min. Then 10% Nonidet P-40 was added to a final 1% concentration. After a 10-s slight vortex, cells were centrifuged at 14,000 rpm for 3 min. Then the supernatants were collected as the cytoplasmic protein. The pellets were resuspended in 220 μl ice-cold Buffer II (420 mM NaCl, 20 mM HEPES, 0.2 mM EDTA, 1.5 mM MgCl2, 25% glycerol, and protease inhibitor cocktail, pH 8.0) and incubated on ice for 30 min. Then samples were centrifuged and the supernatants were transferred to new tubes as the nuclear fraction which was stored at −80 °C for later use.

HEK-293T cells were co-transfected with the indicated plasmids with lipofectamine 2000 (Invitrogen), and the nuclear and cytoplasmic proteins were extracted as previously described [19, 20]. Three kinds of beads were used in this study for Co-IP assay: anti-FLAG M2 Magnetic Beads (Sigma-Aldrich, St Louis, MO); Pierce Anti-c-Myc Magnetic Beads (ThermoFisher); Protein A/G PLUS-Agarose (Santa Cruz). Briefly, the protein extracts were incubated with the equilibrated beads at 4 °C overnight with gentle mixing to capture the FLAG fusion proteins or Myc fusion proteins or specific antibody captured proteins. The magnetic beads or agarose beads were collected by placing the tube in the appropriate magnetic separator or by centrifuging. The beads were washed with TBS buffer to remove all of the non-specifically bounded proteins. The bounded fusion proteins were eluted from the beads with corresponding elution buffer for western blot analysis.

2 × 10⁶ A549-FOXP3 and A549-Control cells were separately subcutaneously inoculated into the left and right flank in the dorsal of the nude mice for in vivo xenograft assay. Tumor size was measured every 3 days for 18 days. The tumor volume (V) was calculated by the formula (length × width × width)/2. The tumors were excised and embedded in paraffin. For lung metastasis formation, 5 × 10⁵ A549-FOXP3 and A549-Control cells were injected into the lateral tail vein of the nude mice. Mice were euthanized 9 weeks after injection, and the lung, liver and spleen of each mice were subjected to formaldehyde fixation and followed by H&E staining. All experimental procedures were approved by the Animal Ethics Committee of the Chinese University of Hong Kong.

To determine whether FOXP3 was expressed in NSCLC and correlated with patient prognosis, we performed immunohistochemistry on NSCLC tissues using two types of monoclonal anti-FOXP3 antibodies. FOXP3 immunostaining was seen in NSCLC cells and Treg cells (Additional file 2: Figure S1). Tumor FOXP3 expression mainly exhibited a mixture of diffuse staining in the cytoplasm, the nucleus or in both of them. The FOXP3 expression levels in NSCLC were scored 1 to 12 according to the IRS method. Specimens scored 1–4 were classified as ‘Tumor FOXP3-Low’ and 6–12 as ‘Tumor FOXP3-High’. In 106 NSCLC specimens, 41 cancer tissues were detected high FOXP3 expression and 65 cancer tissues showed low FOXP3 expression. Statitically, the overall expression of FOXP3 was much higher in cancer tissues than in normal adjacent tissues (p < 0.001) (Fig. 1a). To confirm this finding, western blotting was performed to detect their protein levels in 5 randomly selected paired NSCLC specimens. As shown in Fig. 1b, 3 out of 5 showed higher level of FOXP3 protein in tumor tissues than that in adjacent normal tissues.

We analyzed the correlation between clinicopathologic characteristics and tumor FOXP3 expression, the results showed that there was no significant association of FOXP3 expression levels with clinicopathologic characteristics. However, we found a correlation between tumor FOXP3 expression and Treg cell counts (Additional file 1: Table S3).

The correlation of FOXP3 expression with the long-term survival rate after surgery operation was analyzed. Kaplan–Meier analysis showed that patients with high levels of FOXP3 expression in tumor had shorter overall survival time than those with low levels of FOXP3 expression (P < 0.05, Fig. 1c). The mean overall survival time of patients with high FOXP3 expression in tumor tissues was 44.3 ± 6.54 months, while that with low FOXP3 expression level was 78.1 ± 7.35 months (P = 0.007).

Kaplan–Meier analysis also showed that patients with high levels of FOXP3 expression in tumor had shorter recurrence-free survival time than those with low levels of FOXP3 expression (P < 0.05, Fig. 1c). The mean recurrence-free survival time of the patients with high FOXP3 expression in tumor tissues was 30.5 ± 5.47 months, while those with low FOXP3 expression was 63.8 ± 8.16 months (P = 0.019).

To further explore whther FOXP3 had any prognostic value, we conducted Cox proportional Hazard regression analysis of patients’ overall survival and recurrence-free survival. In multivariate analysis of overall survival, when other risk factors such as histology grade, pathology stage, pathology tuberculosis and nodal status were taken in consideration, the HR of high FOXP3 expression was improved from 1.86 (95% CI: 1.18–1.92, P = 0.01) to 2.09 (95% CI: 1.27–3.44, P = 0.00) (Additional file 1: Table S4). In multivariate analysis of recurrence-free survival, when other risk factors were taken in consideration, the HR of high FOXP3 expression was improved from 1.71, (95% CI: 1.09–2.70, P = 0.02) to 1.83 (95% CI: 1.11–3.02, P = 0.02) (Additional file 1: Table S5). Both univariate analyses and multivariate analyses indicate that FOXP3 is an independent predictor for overall survival and recurrence-free survival of NSCLC patients.We also analyzed the Treg cells in NSCLC. Patients were divided into two groups according Treg cell numbers, we found that Treg cells alone did not have a prognostic value in NSCLC (P > 0.05, Fig. 1d).

We examined the effect of ectopic expression of FOXP3 on the several tumorigentic features of cancer cells using NSCLC cell lines (A549 and H460). FOXP3 lentivirus was used to improve transfection efficiency, and the expression of FOXP3 was validated by western blot (Fig. 2a). Ectopic expression of FOXP3 caused a significant increase in cell viability in both A549 (p < 0.001) and H460 (p < 0.001) cells (Fig. 2a). This promoting effect was further confirmed by the colony formation assay, in which FOXP3 significantly increased the colony formation in A549 (p < 0.001) and H460 (p < 0.001) when compared with control (Fig. 2b).

Since FOXP3 promotes cell growth in vitro, we next tested whether FOXP3 could promote the growth of NSCLC cells in nude mice. 18 days after inoculation we found that the mean volume of tumors infected with FOXP3 was 2.6-fold larger than that of infected with control virus, which formed either no tumor or much smaller tumors. The tumor growth curves of A549-FOXP3 and A549-Control in nude mice were shown in Fig. 2c and Additional file 2: Figure S2. In contrast to A549-control, A549-FOXP3 exhibited markedly increased growth rate in the xenograft model (p < 0.05). Taken together, these findings have demonstrated that FOXP3 can exhibit obvious tumorigenic functions in NSCLC.

To further validate the oncogenic effect of FOXP3 on the growth of NSCLC, shFOXP3 lentivirus was generated and the silence effect was validated by qPCR (Fig. 2d). MTT assay showed that the knockdown of FOXP3 in A549 could inhibit cell growth (p < 0.05) (Fig. 2d). The colony formed by shFOXP3 lentivirus-infected cells was fewer than the control lentivirus-infected cells (Fig. 2d). These results further support that FOXP3 is an oncogenic molecule in NSCLC.

When we first infected A549 cells by FOXP3 lentivirus, we found a significant spindle shape change in A549 cells (Fig. 3a) compared to control cells, which is further confirmed in H460 cells infected with FOXP3 lentivirus (Fig. 3a). When we seeded A549-FOXP3 cells in 6-well plate to form colony for about 1–2 weeks, it was found that cells lost cell-cell contact and formed a scattered phenomenon (Fig. 3a). A similar finding was also observed in H460-FOXP3 cells after two-week growth (Fig. 3a). Spindle shape change and lose of cell-cell contact are the characteristics of cell morphology change during Epithelial-Mesenchymal Transition [21]. E-Cadherin is the most remarkable markers of EMT, immunofluorescent staining demonstrated that E-Cadherin was markedly reduced in A549-FOXP3 cells compared with control cells (Additional file 2: Figure S3), which is consistent with the cell morphology change that FOXP3 induced the EMT. Western blot was used to analyze EMT markers induced by FOXP3 in A549 and H460 cells (Fig. 3b). E-Cadherin was downregulated in both A549-FOXP3 and H460-FOXP3 cells compared to control cells, which is consistent with our immunofluorescent staining assays. An upregulation of N-Cadherin, vimentin, Snail, Slug and MMP9 was also induced by FOXP3 in both cells. All these changes in biomarkers suggest a transition of epithelial cells to mesenchymal cells.

Then, we knocked down FOXP3 to check whether it could reverse the change of EMT markers induced. A549-shFOXP3 cells which was generated early was used for analysis. Western blot results (Fig. 3c) showed that E-Cadherin was upregulated, N-Cadherin, Vimentin, snail, slug and MMP9 were downregulated compared to control cells. These changes were opposite to FOXP3 overexpression experiment, and further validated the role FOXP3 on EMT induction.

First, the result of the wound healing assay showed that cells infected with FOXP3 lentivirus had a significant healing ability than control cells in wound formation, which is about 1.35 fold faster than control cells (p < 0.01) (Fig. 4a, Additional file 2: Figure S4A), suggesting A549-FOXP3 has an stronger ability in migration than A549-control. Second, we did Matrigel invasion assays to test the ability of invasion. Photomicrographs and histograms (Fig. 4b) indicated that the number of A549-FOXP3 cells migrated to lower chambers through Matrigel is more than that of control cells migrated to the lower chambers (2.4 fold, p < 0.01). Third, soft agar assays showed that A549-FOX3 and H460-FOXP3 cells formed more and larger colonies (p < 0.01) compared to control cells (Fig. 4c). Finally, to test whether FOXP3 could promote metastasis in vivo, A549-FOXP3 and control cells were inoculated into nude mice via tail vein injection. The mice were sacrificed 2 months later to examine metastatic nodules in lung, liver and spleen. The results of H&E staining showed that ectopic expression of FOXP3 in A549 cells led to a significant increase in the total number of metastatic nodules in the lung of nude mice (Fig. 4b-d), although we didn’t find metastatic nodules in the liver and spleen of nude mice (data not shown). All these findings suggest that FOXP3 is able to promote metastasis in NSCLC cells.

To gain insights into the mechanism of FOXP3-mediated tumorigenicity and EMT, microarray gene expression profiling was conducted. Hierarchical clustering analysis showed that FOXP3 overexpression led to a clear and consistent difference in the gene expression profile (Additional file 2: Figure S5). We identified 1811 upregulated genes, and 2292 downregulated genes (data not shown). We conducted GO analysis to reveal general functional features implemented by FOXP3 in A549 cells using upregulated genes. The results of both Biological process analysis (Additional file 2: Figure S6A) and cellular component analysis (Additional file 2: Figure S6B) coincided with the characteristics of EMT process.

Pathway analysis suggests multiple pathways might be involved in A549 cells after FOXP3 stimulation (Additional file 2: Figure S7). Wnt/β–catenin pathway was selected for further studies as it is broadly involved in the process of promoting tumorigenicity, cell stemness and EMT induction [21, 22]. Data analysis of Wnt/β–catenin pathway (data not shown) suggested that 4 out of 6 direct downstream genes including c-Myc, Cyclin D1, c-Jun, fra-1 were upregulated, the finding of which is another evidence to support an activated Wnt/β–catenin pathway in A549-FOXP3 cells. To determine whether Wnt/β–catenin pathway was activated by FOXP3 in NSCLC cells, we first performed a luciferase experiment with Top-Luc Flash reporter and pRL-TK (as an internal control). Our results showed that FOXP3 increased luciferase activity dramatically in both A549 and H460 cells (Fig. 5a). And we observed a significant reduction of the luciferase activity in A549 cells when FOXP3 was silenced by lentivirus (Fig. 5b). These results suggest that the overexpression of FOXP3 stimulates the Wnt/β–catenin pathway in NSCLC cells.

We also analyzed the expression of c-Myc and Cylin D1, two well-known target genes of the Wnt/β–catenin pathway [23‐25], in A549-FOXP3 and H460-FOXP3 cells. Both qPCR (Fig. 5d) and western blot (Fig. 5c) results showed the elevated expression of c-Myc and Cylin D1.

In most cases, the elevated expression of β-catenin is the cause of activating canonical pathway. However, our western blot results did not show the significant changes in β-catenin levels in A549-FOXP3 and H460-FOXP3 cells compared with corresponding control cells. Since our immunofluorescent staining results (Additional file 2: Figure S8) which showed the localization of overexpressed FOXP3 protein was mainly in the nucleus of A549-FOXP3 and H460-FOXP3 cells, it is likely that the oncogenic function of FOXP3 is correlated with the LEF/TCF complex, which is the potent Wnt/β-catenin transcriptional factor in the nucleus [23, 26]. We thus examined whether FOXP3 interacted with the nuclear components of the Wnt/β-catenin signaling pathway. Reciprocal immunoprecipitation experiments indicated that FOXP3 and TCF4 interacted with each other in HEK-293T cells (Fig. 6a), suggesting that FOXP3 could form a complex with TCF4 in the nucleus. In the canonical pathway, β-catenin combined with LEF/TCF4 complex to activate Wnt signaling [19, 20, 27, 28]. We thus questioned whether FOXP3 could also form a complex with β-catenin. The Co-IP test demonstrated that β-catenin and FOXP3 could precipitate down each other in the nucleus (Fig. 6b), suggesting that FOXP3 could also associate with β-catenin to form a complex. To this end, we speculated FOXP3 might enhance the formation of β-catenin·TCF4 complex and work as a co-activator of Wnt/β-catenin pathway. Immunoprecipitation results showed that the interaction of FLAG-β-catenin with Myc-TCF4 increased 1.45 fold in the nucleus when FOXP3 was co-expressed (Fig. 6c). Furthermore, an immunoprecipitation experiment using endogenous proteins demonstrated that the interaction of endogenous β-catenin and endogenous TCF4 was significantly increased 1.79 fold in the nucleus by FOXP3 overexpression (Fig. 6d). Taken together, these results suggest that FOXP3 is a co-activator to enhance the formation of the β-catenin and TCF4 complex in the nucleus of NSCLC cells.