miRomics and Proteomics Reveal a miR-296-3p/PRKCA/FAK/Ras/c-Myc Feedback Loop Modulated by HDGF/DDX5/β-catenin Complex in Lung Adenocarcinoma

miRNA dysregulation has been implicated in a variety of diseases, including cancer (1–3). One such miRNA is miR-296, which has two transcripts, miR-296-5p and miR-296-3p. In the literature, miR-296-5p is generally referred to as miR-296. miR-296 (miR-296-5p) is progressively lost during tumor progression promoting cell growth and metastasis (4). Conversely, increased expression of miR-296 has been observed in multiple cancer types and contributes to carcinogenesis by directly targeting p21 (5). The role of miR-296-3p in carcinogenesis is less clear. In glioblastoma, miR-296-3p suppresses cell growth and promotes chemosensitivity (6). Conversely, miR-296-3p can function as an oncogene in prostate cancer by reducing ICAM-1 inducing metastasis (7). In non–small cell lung cancer (NSCLC), miR-296-3p has been found to be upregulated on the basis of microarray analysis from Gene Expression Omnibus (8), suggesting miR-296-3p as an oncogene. However, inverse result of miR-296-3p was observed as a tumor suppressor, reducing drug resistance in the latest study (9). Which data were correct?

Lung cancer is the most prevalent malignancy and the leading cause of mortality worldwide (10). NSCLC is the most common form, accounting for approximately 85% of all lung cancers (11). Lung adenocarcinoma (LADC) recently replaced squamous cell carcinoma as the most common histologic subtype in various countries (12, 13). Emerging evidence has implicated the role of miRNAs in LADC development and progression (14, 15). However, the expression pattern and functional mechanisms by which miR-296-3p contributes to LADC have not been examined.

Hepatoma-derived growth factor (HDGF) is a heparin-binding protein that was originally purified from the conditioned media of HuH-7 hepatoma cells (16). HDGF overexpression has been observed in various tumor types (17), including a well-documented role in lung cancer pathogenesis. Upregulation of HDGF is a frequent event in NSCLC, and its expression level is an independent prognostic factor closely associated with poor overall, disease-specific, and disease-free survival (18, 19). Anti-HDGF antibody treatment has been shown to enhance the antitumor activities of gemcitabine, bevacizumab, and chemotherapy in NSCLC (20, 21). In addition, HDGF can serve as a direct target of anti-oncomirs to suppress the malignant phenotype of NSCLC cells (22, 23). The precise mechanism by which HDGF modulates miRNAs to contribute to LADC progression remains unclear.

DEAD-box helicase 5 (DDX5) is a member of the DEAD-box family of RNA helicases (24, 25) and is overexpressed in multiple malignancies, including lung cancer (26, 27). DDX5 is involved in miRNA generation by shuttling between the nucleus and cytoplasm (28, 29). The role by which DDX5 mediates miRNA function in LADC needs further investigation.

Here, we identified an atypical miR-296-3p–PRKCA–FAK/Ras–c-Myc feedback circuit that suppresses LADC cell proliferation and metastasis and regulates sensitivity to cisplatin (DDP). This pathway is modulated by HDGF/DDX5/β-catenin signaling. Together, these data provide detailed mechanisms by which miR-296-3p functions as a tumor suppressor to prevent LADC cell growth and metastasis.

A549, BEAS-2B, and SPCA-1 cells were purchased from Chinese Academy of Sciences Cell Bank (Shanghai, China); HEK293T cells were obtained from the Cancer Research Institute of Southern Medical University (Guangzhou, China). These cells were confirmed to be free of mycoplasma. A549 and HEK293T cells were cultured in DMEM (Invitrogen) supplemented with 10% FBS (ExCell); SPCA-1 cells were cultured in RPMI1640 medium (Invitrogen) with 10% FBS (ExCell); BEAS-2B cells were cultured in BEBM medium (Lonza) according to the ATCC website. All cells were incubated in a humidified chamber with 5% CO₂ at 37°C.

Twenty-six (26) surgical resected fresh primary LADC tissues and paired normal lung tissues were obtained from the Third Affiliated Hospital of Kunming Medical University (Yunnan, China). Clinical processes were approved by the Ethics Committees of the Third Affiliated Hospital of Kunming Medical University, and patients provided informed consent. The tissue array that included 123 paraffin-embedded primary LADC specimens and 64 normal specimens was obtained from Shanghai Outdo Biotech Co. Ltd. Patients with a diagnosed relapse as well as those that had received preoperative radiation, chemotherapy, or biotherapy were excluded. Demographic and clinical data were obtained from the patients' medical records.

Lentiviral particles carrying hsa-mir-296 precursor and human HDGF short hairpin RNA (shHDGF-1, 2, 3) vector and their flanking control (NC) were constructed by GeneChem (Supplementary Table S1). A549 and SPCA-1 cells were infected with lentiviral vector, and expression of miR-296-3p was confirmed by qRT-PCR while levels of HDGF were measured using qRT-PCR and Western blot analysis.

siRNA for DDX5, PRKCA, β-catenin, and c-Myc or hsa-miR-296-3p mimics and inhibitor (Supplementary Table S1) were designed and synthesized by Guangzhou RiboBio (RiboBio Inc.). Plasmids for PRKCA, c-Myc, β-catenin, and DDX5 were purchased from Vigene Biosciences. Twenty-four hours prior to transfection, cells were plated onto a 6-well plate or 96-well plate (Nest Biotech) at 30% to 50% confluence. Plasmid constructs were then transfected into cells using Lipofectamine TM 2000 (Invitrogen Biotechnology) according to the manufacturer's protocol. Cells were collected after 48 to 72 hours for further experiments.

Total RNA was extracted from cells and tissues using TRIzol (Takara). RNA (1 μg) was reverse transcribed into cDNA, and cDNA was used as a template to amplify with specific primers (Supplementary Table S2). U6 and ARF5 genes were used as miRNA and gene internal controls, respectively. Experiments were performed according to the manufacturer's instructions (Takara). Independent experiments were done in triplicate.

Western blots were performed as described previously (30) with primary antibodies, including anti-HDGF, PRKCA, K-Ras, CCND1, DDX5, c-Myc, CDK4, FAK, p-FAK (Tyr397), and β-catenin. β-Actin was used as a loading control for all blots. Dilutions and sources of antibodies are shown in Supplementary Table S3. Images were captured with ChemiDocTM CRS+ Molecular Imager (Bio-Rad).

Cell proliferation was assessed using MTT as described previously (31). For lentivirus-mediated HDGF suppression, cells were seeded in 96-well plates at a density of 1,000 cells/well for 7 days. For transient transfection with siRNAs, miR-296-3p mimics, inhibitor, and PRKCA plasmid, cells were seeded at a density of 3,000 cells/well for 4 days. For each experimental condition, five parallel wells were assigned to each group. Experiments were performed in triplicate.

Clone formation was measured according to our previous study (31). Cells were seeded in 6-well culture plates at 200 cells/well. After incubation for 14 days, cells were washed twice with Hank's solution and stained with hematoxylin solution. The number of colonies containing ≥50 cells was counted under a microscope. All experiments were repeated at least three times.

EdU incorporation was performed using Apollo567 In Vitro Imaging Kit (RiboBio Inc.) according to the manufacturer's protocol. Briefly, cells were incubated with 10 μmol/L EdU for 2 hours before fixation with 4% paraformaldehyde, permeabilization by 0.3% Triton X-100, and staining with Apollo fluorescent dyes. Cell nuclei were stained with 5 μg/mL DAPI (4′,6-diamidino-2-phenylindole) for 10 minutes. The number of EdU-positive cells was counted under a microscope in five random fields. All assays were independently performed in triplicate.

In vitro cell migration and invasion assays were performed according to a previous description (32). All assays were independently repeated at least three times.

In vivo tumorigenesis in nude mice was performed on the basis of our previous study (32). A total of 5 × 10⁶ logarithmically growing cells in 0.1 mL Hank's solution were subcutaneously inoculated into the left-right symmetric flank of 4- to 6-week-old female BALB/c-nu/nu mice (n = 6 per group). Mice were maintained in a barrier facility on HEPA-filtered racks and fed an autoclaved laboratory rodent diet. All animal studies were conducted in accordance with the principles and procedures outlined in the Southern Medical University Guide for the Care and Use of Animals under assurance number SCXK (Guangdong) 2008-0002. Mice were sacrificed after 15 days and tumors were excised, weighed, and processed for histology.

To establish an LADC mouse model, 6 × 10⁵ miR-296-3p–overexpressing A549 cells or their controls were intraperitoneally injected into 14 g female nu/nu mice (n = 24 each). Tumors were allowed to grow for 3 days, and then, animals were randomized into two treatment groups consisting of either normal saline (NS) or DDP intraperitoneally injected every 3 days (NC + NS, NC + DDP, miR-296-3p + NS and miR-296-3p + DDP, n = 12/group). Survival curves were analyzed using Kaplan–Meier analysis.

In vivo metastasis assays were performed according to our previous study (33). A total of 5 × 10⁶ cells were injected under the liver capsule (n = 5 for each group). The optical fluorescence images were visualized to monitor primary tumor growth and formation of metastatic lesions. Forty days later, all mice were killed, livers were removed, and metastatic tissues were analyzed by H&E staining.

Paraffin sections (4 μm) from samples were deparaffinized and antigen retrieval was performed in citrate buffer for 3 minutes at 100°C. Endogenous peroxidase activity and nonspecific antigens were blocked with peroxidase blocking reagent followed by incubation with antibody overnight at 4°C. Antibody dilutions and sources are shown in Supplementary Table S3. After washing, sections were incubated with biotin-labeled secondary antibody and subsequently were incubated with streptavidin-conjugated horseradish peroxidase. The peroxidase reaction was developed using 3,3-diaminobenzidine (DAB) chromogen solution in DAB buffer substrate (Maixin). Sections were visualized with DAB and counterstained with hematoxylin, mounted in neutral gum, and analyzed using a brightfield microscope.

PRKCA was predicted to be a direct target of miR-296-3p by RNAhybrid and TargetScan software. A fragment of the PRKCA 3′-UTR (wild-type 3′-UTR) was amplified. Site-directed mutagenesis (mut) of the miR-296-3p–binding site was performed using GeneTailor Site-Directed Mutagenesis System (Invitrogen). The wt 3′-UTR or mut 3′-UTR were cloned into psiCHECK-2 vectors. For luciferase reporter assays, wt or mut 3′-UTR vector was cotransfected with miR-296-3p mimics/inhibitor or nonspecific control into HEK293T cells. Luciferase activity was measured 48 hours after transfection using the Dual-Luciferase Reporter Assay System (Promega Corporation). To examine the effect of c-Myc on miR-296-3p promoter activity, a 308-bp fragment containing the two c-Myc–binding sites was cloned into the pGL3-Basic luciferase reporter vector, and the c-Myc–binding site mutation vectors were constructed. These vectors and c-Myc plasmid were cotransfected into A549 and SPCA-1 cells. Luciferase activity of miR-296-3p promoter was measured 48 hours after transfection.

miRNA microarray was performed by Gene Co., Ltd. Affymetrix Gene Chip Micro 2.0 Array (Affymetrix, Inc.), which provides for 100% miRBase v17 coverage (www.mirbase.org) by a one-color approach, was employed for universal miRNA coverage (34). Total RNA (6–7 mg) was isolated from HDGF suppressed or control SPCA-1 cells. Statistical analysis was carried out using the open source R-software (http://www.r-project.org) as previously reported (35). Expression profile clustering and visualization was performed with Cluster and Treeview software (Ernest Orlando Lawrence Berkeley National Laboratory).

JASPAR (http://jaspar.genereg.net) and ALGGEN (http://alggen.lsi.upc.es) database analyses predicted two putative c-Myc–binding sites on miR-296-3p promoter region. DNA–protein complexes were immunoprecipitated from A549 cells with the Chromatin Immunoprecipitation Kit (Thermo Fisher Scientific) according to the manufacturer's protocol using anti-c-Myc or IgG (Cell Signaling Technology) antibodies. IgG served as a control for nonspecific DNA binding. qRT-PCR and PCR analysis were used to measure the enrichment of miR-296-3p promoter region.

c-Myc binding to miR-296-3p promoter was detected using EMSA Kit (Roche) according to the manufacturer's instructions. The probe sequences are shown in Supplementary Table S2. The 3′-end of the wild-type probe was labeled with biotin. Samples without nucleoprotein were used as negative controls. The binding mixture included 5 μg nuclear extracts and 1 μg poly (dI:dC) incubated in the presence or absence of 100-fold specific oligonucleotide competitor (unlabeled wild-type or mutant c-Myc probes) for 15 minutes at room temperature before the addition of the biotin-labeled wild-type probe. Signals were recorded using a BioSens Gel Imaging System (BIOTOP). Electrophoretic mobility shift assay (EMSA) analysis was performed at Biosense Bioscience Co. Ltd.

Coimmunoprecipitation (Co-IP) was carried out using a Pierce Co-Immunoprecipitation kit (Thermo Fisher Scientific) according to the manufacturer's instructions. Briefly, total proteins were extracted and quantified. A total of 1,000 μg protein in 400 μL supernatant was incubated with 10 μg anti-HDGF, anti-DDX5, or anti-IgG antibodies for 12 hours at 4°C. Beads were washed, eluted in sample buffer, and boiled for 10 minutes at 100°C. Immune complexes were subjected to Coomassie brilliant blue staining, mass spectrometry, and Western blot analysis. Anti-IgG was used as a negative control.

Immunofluorescent staining was performed as described previously (32). Briefly, cells were seeded on coverslips in 6-well plate and cultured overnight before fixation with 3.5% paraformaldehyde and permeabilization by 0.2% Triton X-100 at room temperature. Cells were incubated with mouse anti-HDGF and rabbit anti-DDX5 antibody for 30 to 45 minutes at 37°C. After incubating for 30 to 45 minutes at 37°C with secondary antibody, coverslips were mounted onto slides with mounting solution containing 0.2 mg/mL DAPI and sealed with nail polish. Slides were stored in a dark box and observed under a fluorescent microscope.

GST pull-down was performed at Biosense Bioscience Co. Ltd. Fusion proteins GST-DDX5 and His-HDGF were constructed by inserting the coding region of human DDX5 and HDGF into pGEX-6P-1 and pET-28a, respectively. GST-DDX5 was transformed into E. Coli BL21, and soluble lysates were prepared with protease and phosphatase inhibitors. Soluble fusion proteins were purified using glutathione sepharose. His-HDGF fusion protein was prepared. For GST pull-down, approximately 500 μg of total protein lysate was incubated with GST-fusion protein/GST-control at 4°C with gentle shaking overnight before incubation with glutathione-sepharose beads for 2 hours. Beads were washed three times with binding buffer and boiled in SDS loading buffer at 95°C for 10 minutes. Immunoblotting was used to detect bound proteins.

Drug sensitivity test was determined by MTT. Cells were seeded in 96-well plates at a density of 5 × 10³ cells/well and treated with 0, 1, 2, 4, 8, or 16 μmol/L cisplatin (Qilu Pharmo Co. Ltd.) for 48 hours. Subsequently, 20 μL of MTT (5 mg/mL; Sigma) was added to each well and incubated at 37°C for 4 hours. Then, supernatants were removed and 150 μL of DMSO (Sigma) was added to measure absorbance value (OD) of each well at 490 nm. Calculated rates were used for curve fitting and IC₅₀ calculations. Experiments were carried out three times.

In situ hybridization was performed on paraffin-embedded specimens (4-μm thickness). After processing with 3% H₂O₂, sections were treated with proteinase K (2 μg/mL) at 37°C for 30 minutes, washed, and prehybridized for 2 hours at 37°C. Hybridization with digoxygenin (DIG)-labeled miRCURY LNA 15 probes (probe sense: BIO-PR228; Biosense Bioscience Co., Ltd.) was performed overnight at 37°C. Slides were then washed and incubated with alkaline phosphatase–conjugated sheep anti-DIG Fab fragments for 1 hour at room temperature. Staining was visualized by adding BM purple AP substrate (Roche) according to the manufacturer's instructions.

Statistical analyses were performed with the SPSS 16.0 statistical software packages (SPSS Inc.). Data are expressed as mean± SD from at least three independent experiments. Differences were considered to be statistically significant at values of P < 0.05 by Student t test for two groups, one-way ANOVA analysis for multiple groups, and parametric generalized linear model with random effects for tumor growth and MTT assay. Analysis of miR-296-3p, PRKCA, HDGF, and DDX5 expression in 26 fresh primary LADC tissues was performed using paired-sample t test. Relationships were analyzed with Spearman correlation analysis. Survival analysis was performed using the Kaplan–Meier method. All statistical tests were two-sided, and asterisks indicate statistical significance.

To assess its biological function, we overexpressed miR-296-3p in LADC cells using mimics and precursor (miR-296) lentivirus particles (Supplementary Fig. S1A–S1C). Cell proliferation (Supplementary Fig. S1D), G₁–S cell-cycle phase transition (Supplementary Fig. S1E), and colony-forming ability (Supplementary Fig. S1F) were notably inhibited after ectopic expression. These inhibitory effects were reversed by introducing miR-296-3p–specific inhibitors into these overexpressing cells. Next, we inoculated miR-296-3p–overexpressing LADC cells into nude mice to examine the effect on tumorigenesis. Consistent with in vitro data, tumor weights and growth rates were significantly reduced after miR-296-3p overexpression (Supplementary Fig. S1G). RT-qPCR verified elevated miR-296-3p expression compared with NC tumors (Supplementary Fig. S1H). IHC staining revealed that Ki67 and PCNA protein were markedly decreased in miR-296-3p–overexpressing xenografts (Supplementary Fig. S1I). These results demonstrated miR-296-3p functions as a tumor suppressor in LADC.

Subsequently, the ability of cell migration and invasion was observed to be reduced in the miR-296-3p–overexpressed group compared with control cells (Supplementary Fig. S2A). Furthermore, LADC cell metastasis was measured in vivo by inoculating miR-296-3p–overexpressed and control cells under liver capsule of mice. Notably, intrahepatic metastasis was decreased in the miR-296-3p–overexpressed group compared with control cells (Supplementary Fig. S2B).

DDP has been widely used for treatment of various solid tumor types, including LADC; thus, we examined how ectopic miR-296-3p affected inhibition rates. The IC₅₀ of DDP was significantly reduced after miR-296-3p overexpression (Supplementary Fig. S1J). Furthermore, the LADC cells with or without miR-296-3p transfection were treated with NS, 2, or 4 μmol/L DDP for 1, 2, 3 days; the cell growth of miR-296-3p groups significantly slowed compared with the NC groups (Supplementary Fig. S1K). This effect also translated in vivo using miR-296-3p–overexpressing A549 xenografts. Survival times calculated by Kaplan–Meier analysis revealed that DDP treatment (NC + DDP) or miR-296-3p overexpression (miR-296-3p + NS) alone prolonged survival compared with untreated normal controls (NC + NS). However, miR-296-3p overexpression coupled with DDP treatment (miR-296-3p + DDP) significantly improved survival group beyond the other three groups (Supplementary Fig. S1L). Notably, there was no significant difference between the NC + DDP group or miR-296-3p + NS groups.

To determine the mechanisms by which miR-296-3p inhibited LADC cell proliferation, metastasis, and chemoresistance, the key regulators of cell cycle and epithelial–mesenchymal transition (EMT) were analyzed by Western blot analysis. We observed that the introduction of miR-296-3p into LADC cells downregulated levels of c-Myc, CCND1, CDK4, N-cadherin, vimentin, and Snail and upregulated E-cadherin (Fig. 1A; Supplementary Fig. S2C). We also observed that miR-296-3p mimics inactivated FAK/K-Ras signaling (Fig. 1B), an important pathway for cell proliferation, metastasis, and chemosensitivity (36).

To further elucidate the mechanism by which miR-296-3p regulates LADC cell proliferation, we used TargetScan and RNAhybird algorithms to identify direct target genes. The PRKCA 3′UTR was among the predicted targets that matched the miR-296-3p seed sequence (Fig. 1C). Overexpression of miR-296-3p strikingly downregulated PRKCA in LADC cells and immortalized cell line BEAS-2B cells by Western blot analysis (Fig. 1D), whereas suppressed miR-296-3p with inhibitor notably increased PRKCA expression in LADC cells and BEAS-2B cells (Fig. 1E). In addition, a dramatic decrease in PRKCA levels was detected in miR-296-3p–overexpressing xenografts by IHC (Fig. 1F). Interestingly, miR-296-3p overexpression had little effect on PRKCA transcript levels by qRT-PCR (Fig. 1G), implying that miR-296-3p inhibits PRKCA through a posttranscriptional mechanism. Next, we cloned wild-type and mutant PRKCA 3′UTR into the psicheck-2 dual-luciferase reporter vector and cotransfected HEK293T cells with these reporters along with miR-296-3p mimics or inhibitors. We found that the luciferase activity of the wild-type PRKCA 3′UTR but not the mutant 3′UTR was significantly reduced by miR-296-3p mimics. Conversely, miR-296-3p inhibitors caused robust activation of luciferase activity of the wild-type PRKCA 3′UTR instead of the mutant 3′UTR. This effect was not observed in the psicheck-2 vector control groups (Fig. 1H). These data demonstrated that PRKCA is a direct target of miR-296-3p.

To further observe the effect of PRKCA on miR-296-3p–modulated pathways, PRKCA plasmid was transiently transfected into miR-296-3p–overexpressing LADC cells. PRKCA transfection increased proliferation and metastasis (Fig. 1I and J; Supplementary Fig. S3A). Overexpression of PRKCA also decreased DDP sensitivity of miR-296-3p–overexpressing LADC cells (Fig. 1K and L). Furthermore, levels of p-FAK, K-Ras, c-Myc, CCND1, CDK4, N-cadherin, vimentin, and Snail were notably increased; E-cadherin expression was decreased (Fig. 1M; Supplementary Fig. S3B). LADC cells were then transfected with two PRKCA-specific siRNAs to evaluate downstream effects, and it was found that PRKCA downregulation markedly reduced levels of p-FAK, K-Ras, c-Myc, CCND1, and CDK4 (Fig. 1N). Collectively, these results indicate that miR-296-3p inactivates FAK–Ras signaling to suppress cell cycle and EMT transition by directly targeting PRKCA.

JASPAR (http://jaspar.genereg.net) and ALGGEN (http://alggen.lsi.upc.es) database analysis predicted that the human miR-296 promoter region contains two putative c-Myc–binding sites (site P1 from −147 to −152, site P2 from −316 to −321; Fig. 2A). To examine whether c-Myc reversibly regulates miR-296-3p via a negative feedback loop, we used two specific c-Myc siRNAs in A549 and SPCA-1 cells (Fig. 2F). c-Myc suppression greatly increased the expression of miR-296-3p and its precursor, and no significant doses or time-dependent responses were observed once the c-Myc siRNA worked (Fig. 2B; Supplementary Fig. S4), implying that c-Myc is an upstream regulator of miR-296-3p.

Chromatin immunoprecipitation (ChIP) was subsequently used to determine whether endogenous c-Myc binds to the miR-296-3p promoter in LADC cells. Indeed, there was a significant enrichment of this specific region from immunoprecipitated chromatin DNA compared with negative control (IgG) pull-down (Fig. 2C). EMSA assay was also used to detect c-Myc bound to miR-296-3p as shown in Fig. 2D. A migrating complex appeared when LADC cell nuclear extracts were incubated with the biotin-labeled wild-type probe (lane 5), whereas complex formation was inhibited by unlabeled wild-type competitor (lane 2). The complex was not significantly affected by mutated P1 or P2 competitors (lane 3 and 4) but were supershifted by a c-Myc antibody (lane 6). These results suggest c-Myc specifically binds to the miR-296-3p promoter.

Next, we cloned the wild type as well as mutant P1, P2, and P1 + P2 miR-296-3p promoters into the pGL3-Basic vector to examine the effect of c-Myc on promoter activity. These reporters were cotransfected into HEK293T cells along with c-Myc plasmid. Luciferase activities of wild-type, Mut P1 and Mut P2 promoter, were all significantly reduced by c-Myc cotransfection (Fig. 2E). However, Mut P1 + P2 promoter activity was not affected, implying that c-Myc inhibits the transcriptional activity of miR-296-3p promoter. In addition, we found that c-Myc knockdown markedly suppressed levels of PRKCA, p-FAK, K-Ras, CCND1, and CDK4 (Fig. 2F).

To identify pathways modulated by HDGF, we stably downregulated its expression using two shRNA lentiviral particles (Supplementary Fig. S5A and S5B). Suppression of HDGF reduced cell proliferation, migration, and invasion (Supplementary Figs. S5C and S6A and S6B), clone formation (Supplementary Fig. S5D), cell cycle G₁–S transition (Supplementary Fig. S5E), and xenograft growth (Supplementary Fig. S5F). Ki67 and PCNA protein levels were markedly decreased in HDGF-suppressed xenograft tumors compared with NC xenograft tumors (Supplementary Fig. S5G).

We next examined the mechanism by which HDGF modulates proliferation and metastasis and found that stably silenced HDGF suppressed β-catenin cell-cycle signaling, including downregulation of c-Myc, CCND1, CDK4, N-cadherin, vimentin, and Snail (Fig. 3A; Supplementary Fig. S6C). In addition, HDGF suppression inhibited the PRKCA/p-FAK/K-Ras pathway. Total FAK protein was not affected by decreased HDGF (Fig. 3B).

We next examined global miRNA expression changes in SPCA-1 cells after HDGF knockdown. Compared with the control group, 33 significantly dysregulated (fold change > 2) human miRNAs, including miR-296-3p, were identified (Fig. 3C; Supplementary Table S4). Upregulation of miR-296-3p was validated by qRT-PCR in A549 and SPCA-1 cells after HDGF knockdown (Fig. 3D). These data suggest that miR-296-3p is negatively modulated by HDGF.

To further explore the relationship between miR-296-3p and HDGF in controlling LADC cell proliferation and metastasis, we transfected miR-296-3p inhibitor into LADC cells with HDGF knockdown and observed that cell proliferation, G₁–S cell-cycle transition, migration, and invasion were significantly accelerated (Fig. 3E and F; Supplementary Fig. S7A). In addition, miR-296-3p inhibitor increased expression of PRKCA, p-FAK, K-Ras, c-Myc, CCND1, and EMT regulators (Fig. 3G; Supplementary Fig. S7B), indicating that miR-296-3p is involved in HDGF-mediated LADC cell growth and metastasis.

We next used ChIP analysis to examine how HDGF silencing affects c-Myc targeting of the miR-296-3p promoter. HDGF suppression markedly decreased c-Myc binding to the miR-296-3p promoter in A549 cells (Fig. 3H), suggesting that HDGF exerts its effects on miR-296-3p–mediated cell growth through the β-catenin/c-Myc pathway. To confirm this notion, β-catenin plasmid was transiently transfected into HDGF-suppressed LADC cells. Indeed, expression of miR-296-3p target protein PRKCA and the downstream factors, including p-FAK, K-RAS, c-Myc, and CCND1, were significantly increased by ectopic β-catenin (Fig. 3I).

To explore the precise molecular mechanisms and relevant interactions with HDGF, Co-IP combined with mass spectrometry was used in A549 cells. This analysis yielded 712 potential HDGF-interacting proteins (Supplementary Table S5), including DDX5 (69 kDa band) and β-catenin (92 kDa band; Fig. 4A). Exogenous and endogenous Co-IP demonstrated that HDGF and DDX5 interact in A549 cells, while β-catenin associated with DDX5, but not HDGF (Fig. 4B and C). However, nuclear colocalization of HDGF and DDX5 proteins was observed by immunofluorescence (Fig. 4D). GST pull-down further validated the interaction between HDGF and DDX5 (Fig. 4E). Together, these results suggest that HDGF and DDX5 directly associate in LADC.

In subsequent investigations, we observed that stable HDGF suppression resulted in DDX5 downregulation (Fig. 4F). Interestingly, DDX5 knockdown did not affect HDGF protein levels (Fig. 4G). Decreased DDX5 expression after HDGF silencing was confirmed at the mRNA level (Fig. 6H), implying that HDGF stimulates DDX5 expression at the transcriptional level. Consistent with previous studies that observed β-catenin direct control of DDX5 transcription (26, 37), we observed that β-catenin suppression decreased DDX5 expression (Fig. 4I).

Suppression of DDX5 by siRNAs elevated the expression of miR-296-3p (Fig. 5A). Loss of DDX5 also resulted in decreased levels of β-catenin, c-Myc, CDK4, and CCND1, similar to a previous report in NSCLC (27). In addition, DDX5 downregulation also markedly reduced PRKCA, p-FAK, and K-Ras levels (Fig. 5B), implying that DDX5 controls HDGF-modulated miR-296-3p signaling.

To further explore the involvement of miR-296-3p in DDX5-mediated cell proliferation and metastasis, miR-296-3p inhibitor was transfected into DDX5-suppressed LADC cells. Functional analysis revealed that miR-296-3p inhibitor enhanced cell growth, G₁–S cell-cycle transition, migration, and invasion of DDX5-suppressed LADC cells (Fig. 5C and D; Supplementary Fig. S8A). miR-296-3p inhibitor increased the protein expression of p-FAK, PRKCA, K-Ras, c-Myc, CCND1, N-cadherin, vimentin, and Snail and decreased the E-cadherin expression (Fig. 5E; Supplementary Fig. S8B). DDX5 knockdown also significantly reduced the c-Myc binding to miR-296-3p promoter by ChIP analysis (Fig. 5F). These data provide evidence that DDX5 upstream of HDGF modulates miR-296-3p in LADC. Consistent with this model, DDX5 overexpression elevated expression of miR-296-3p target protein PRKCA and downstream factors, including p-FAK, K-Ras, c-Myc, and CCND1 (Fig. 5G).

Compared with noncancerous lung tissues, significant downregulation of miR-296-3p was observed in LADC tissues (paired-sample t test, P < 0.05; Fig. 6A). Low expression of miR-296-3p in LADC specimens was also confirmed by in situ hybridization (Fig. 6B; Supplementary Table S6). Clinical characteristics associated with miR-296-3p are summarized in Supplementary Table S7. Low expression of miR-296-3p was negatively correlated with AJCC clinical stage (I–II vs. III–IV; χ² test, P = 0.012), T classification (T1–T2 vs. T3–T4; χ² test, P = 0.041), and N classification (N0–N1 vs. N2–N3; χ² test, P = 0.017), but not other clinical features. Patients with low expression of miR-296-3p had poorer overall survival rates by Kaplan–Meier survival analysis (log-rank test, P = 0.025; Fig. 6C).

To further analyze the relationship between miR-296-3p, PRKCA, HDGF, and DDX5, we used qRT-PCR to examine the expression of PRKCA, HDGF, and DDX5 in LADC tissues (T) and the corresponding noncancerous lung tissues (N). Expression of PRKCA, HDGF, and DDX5 were notably upregulated in LADC tissues compared with the normal lung tissues (paired-sample t test, P < 0.05, respectively; Fig. 6D). miR-296-3p expression was inversely correlated with the expression of HDGF and DDX5 by Spearman correlation analysis, but had no obvious correlation with PRKCA levels (Fig. 6E).

miR-296-3p appears to serve dual roles in carcinogenesis (6, 7). It was recently reported to be conflicting in NSCLC (8, 9). Moreover, the precise functions and molecular mechanisms of miR-296-3p in the context of NSCLC have not been reported. Therefore, a complete understanding of the role of miR-296-3p may help design novel therapeutic strategies.

In this study, we first found that miR-296-3p not only significantly inhibited LADC cell growth and metastasis in vitro, but also in vivo tumorigenicity and metastasis. Subsequently, we found that miR-296-3p sensitized LADC cells to DDP in vitro, as well as prolonged the survival time of tumor-bearing mice undergoing DDP treatment. Our data support miR-296-3p as a tumor suppressor in LADC.

Cell-cycle and EMT transition are the key factors that regulate tumor cell growth, metastasis, and chemoresistance (38 –41). The biological effects of miR-296-3p identified in this investigation provide a mechanism for its role in carcinogenesis. We observed that miR-296-3p forms a negative feedback loop via key oncogenic genes and signals, including PRKCA, FAK, Ras, and c-Myc, which suppresses cell-cycle and EMT transition, ultimately inhibiting cell growth, metastasis, and sensitizing cells to DDP.

PRKCA, a member of the protein kinase C (PKC) family, regulates a variety of cellular functions, including cell proliferation, survival, and metastasis (42), through activation of FAK (43) and downstream signals such as Ras and PI3K/AKT (32, 36). We observed that miR-296-3p directly targets PRKCA and suppresses Ras-induced cell-cycle and EMT signaling, consistent with the expression pattern in PRKCA-silenced LADC cells. Overexpression of PRKCA reversed the inhibitory growth, migration, and invasion effects of miR-296-3p. These results demonstrate that miR-296-3p directly targets PRKCA to suppress Ras-stimulated cell-cycle and EMT signaling.

Increasing evidence indicates that aberrant transcriptional regulation of miRNAs can underlie pathogenesis of various diseases (44, 45). We identified putative c-Myc–binding sites in the miR-296-3p promoter region and confirmed this interaction serves to negatively modulate miR-296-3p expression. Together, these results imply that miR-296-3p can induce its own expression though a complex miR-296-3p–PRKCA–Ras–c-Myc loop during LADC pathogenesis.

Although HDGF overexpression has been reported to play a pivotal role in NSCLC progression (18), the exact molecular mechanisms are not well known. We found that HDGF upregulation promoted cell proliferation, migration, and invasion as well as DDP chemoresistance, similar previous studies (21, 46). HDGF induced expression of β-catenin, PRKCA, and Ras-induced cell cycle and EMT signals, an effect opposite of that from ectopic miR-296-3p. To our knowledge, HDGF regulation of PRKCA–Ras signaling during carcinogenesis has not been previously documented.

To investigate the effect of HDGF on miRNAs, we used an miRNA chip following HDGF knockdown in LADC cells. Interestingly, we observed a marked upregulation of miR-296-3p. Inhibition of miR-296-3p partially reversed the proliferative and metastatic effects of HDGF. These effects were achieved by activation of PRKCA-Ras–stimulated cell-cycle and EMT signaling. We observed that miR-296-3p was negatively modulated by c-Myc, a downstream positive regulator of the β-catenin pathway (26). Interestingly, β-catenin is induced by HDGF in some tumors (47, 48); thus, we speculated that miR-296-3p antagonizes HDGF/β-catenin/c-Myc signaling. Indeed, we found that HDGF knockdown reduced β-catenin expression and further decreased c-Myc binding to the miR-296-3p promoter and ultimately stimulated miR-296-3p–mediated inhibition of PRKCA/FAK/Ras pathway.

To better understand its molecular mechanisms, we searched for proteins that directly interact with HDGF in LADC and identified DDX5. DDX5 is a member of DEAD-box proteins and participates in tumor pathogenesis (26, 49, 50). We confirmed that DDX5 is positively modulated by HDGF by inducing β-catenin–mediated transcription (26, 47). Interestingly, a recent study found that DDX5 overexpression promotes NSCLC cell proliferation and tumorigenesis by increasing β-catenin nuclear accumulation and coactivating its downstream effector c-Myc (27). We confirmed that DDX5 lies upstream of the miR-296–PRKCA–Ras–c-Myc feedback loop, significantly inhibiting these downstream targets of HDGF.

In LADC patient tissues, miR-296-3p levels were decreased compared with normal lung by in situ hybridization. Low expression of miR-296-3p was negatively correlated with clinical stage, T classification, and N classification, suggesting miR-296-3p was involved in the progression of LADC. Patients with low miR-296-3p expression had poorer overall survival by Kaplan–Meier analysis. In addition, we found that HDGF, DDX5, and PRKCA were increased in LADC specimens, consistent with the inverse relationship with miR-296-3p observed in vitro. At the transcript level, miR-296-3p was negatively correlated with HDGF and DDX5 expression, but not with PRKCA mRNA levels.

Together, our findings elucidate a complex molecular feedback circuit that involves miR-296-3p, PRKCA, c-Myc, and HDGF/DDX5 in LADC (Fig. 7). The first component of the loop links miR-296-3p to cell proliferation and metastasis by suppressing c-Myc and PRKCA, which activates the downstream FAK–Ras pathway. The second component of the pathway connects HDGF/DDX5 activity to miR-296-3p expression via transcriptional regulation of c-Myc activity, which binds to the miR-296-3p promoter. In summary, these mechanistic findings highlight miRNA targeting as a useful therapeutic option for the treatment of LADC.

No potential conflicts of interest were disclosed.

Conception and design: X. Song, W. Fang

Development of methodology: Q. Fu, R. Li, Y. Chen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Q. Fu, X. Song, R. Luo, C. Ge, R. Li, Z. Li, M. Zhao

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Q. Fu, X. Song, Z. Liu, X. Deng, R. Luo, C. Ge, R. Li, Z. Li, Y. Chen, X. Lin

Writing, review, and/or revision of the manuscript: Q. Fu, X. Song, C. Ge, W. Fang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Z. Liu, Q. Zhang

Study supervision: W. Fang

This work was supported by Guangdong Province Education Foundation (no. 2014KTSCX102 and 2014KTSCX107), Nature science key fund of Guangdong Province (no. 2015A030311005), National Natural Science Foundation of China and Yunnan Joint Foundation (no. U1502222), National Funds of Developing Local Colleges and Universities (no. B16056001), and The Supporting plan for Special Talents in Guangdong Province (no. 2016TQ03R466).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.