Circular RNA circCSNK1G3 induces HOXA10 signaling and promotes the growth and metastasis of lung adenocarcinoma cells through hsa-miR-143-3p sponging

Abstract

Background

In the last decade, a relatively novel, ubiquitous and highly stable subclass of non-coding RNAs, called circular (circ)-RNAs, has increasingly been implicated in cancer development, and several of them have been shown to act as microRNA sponges. As yet, however, the role of circRNAs in lung adenocarcinoma (LUAD) development has largely remained unexplored.

Methods

Bioinformatics, microarray-based and qRT-PCR expression assays were used to assess circRNA, miRNA and mRNA expression in LUAD patient samples and cell lines. siRNA-mediated silencing was used to assess the effect of circCSNK1G3 on various LUAD-associated characteristics such as proliferation, migration, invasion and tumorigenesis, both in vitro and in vivo. Western blotting, immunohistochemistry, fluorescence in situ hybridization (FISH) and luciferase reporter activity assays were used to characterize relationships between circCSNK1G3, miR-143-3p and HOXA10 in LUAD cells.

Results

By screening for differentially expressed circRNAs, we found that circCSNK1G3 was aberrantly expressed in primary LUAD tissues and cell lines. An oncogenic role of circCSNK1G3 was deduced from its aberrant expression and associated enhancement of LUAD A549 and H1299 cell proliferation, migration and invasion. We also found that circCSNK1G3 can directly interact with and suppress miR-143-3p expression by serving as a ‘miR-143-3p sponge’. In addition, we found that circCSNK1G3 can modulate homeobox (HOX) A10 expression through miR-143-3p signaling and, thereby, affect LUAD tumorigenesis.

Conclusions

Our data indicate that circCSNK1G3 can induce HOXA10 expression and, thereby, promote the growth and metastasis of LUAD cells through hsa-miR-143-3p sponging. As such, our data highlight the targetability of the circCSNK1G3/miR-143-3p/HOXA10 signaling axis in patients with LUAD.

Graphical abstract

Introduction

With approximately 2.1 million new cases and 1.8 disease-related deaths in 2018 alone, lung cancer remains the most diagnosed malignancy and the leading cause of cancer-related mortality globally [1]. Of all subtypes, including small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC), NSCLC, comprising lung squamous cell carcinoma (LSCC) and lung adenocarcinoma (LUAD), accounts for 80% of all lung cancer cases [2]. Despite advances that have been made in diagnostics such as the use of serum biomarkers, i.e., carcinoembryonic antigen (CEA) and squamous cell carcinoma antigen (SCCA), or imaging techniques such as computed tomography (CT), and progress that has been made in therapeutic strategies during the last 30 years, the incidence of lung cancer continues to rise, especially among non-smokers and women [3–5]. This rise may be related to the intra-tumoral and/or inter-tumoral heterogeneity of lung cancer and, therefore, does necessitate the discovery of novel disease-specific molecular targets and/or new effective anti-lung cancer treatment modalities [6–9].

Following the initial discovery of circular RNAs (circRNAs) in the early 1980s, there has been an increased interest in the role of circRNAs in cancer initiation and progression, as well as putative therapeutic implications of variations in their expression or activity profiles [10, 11]. circRNAs constitute a novel subclass of ubiquitous, highly conserved endogenous noncoding RNAs, characterized by covalently looped transcripts lacking 5′-3′ polarity in eukaryotes [12, 13]. Although previously thought to be products of splicing errors and thus functionally insignificant, in recent years an increased understanding of the biological roles of circRNAs has emerged, including their function as ligands of RNA-binding proteins, as facilitators of protein translation and as microRNA sponges [14–16]. Additionally, circRNAs have been shown to play critical roles in the regulation of several biological processes, including modulation of gene expression, proliferation, apoptosis and migration in various cancer types. Consistent with this information, it has recently been shown that eukaryotic translation initiation factor 4A3 (EIF4A3)-activated circMMP9 facilitates the development of glioblastoma (GBM) by directly interacting with and sponging miR-124 to enhance the proliferation, invasion and metastasis of GBM cells [17]. Similarly, in gastric cancer, circular RNA Scm-like with Four Mbt domains 2 (circSFMBT2) has been implicated in the proliferation of gastric cancer cells by sponging miR-182-5p and enhancing the expression of cAMP responsive element binding protein 1 [18]. Other examples include circ_0005230 by modulation of miR-618/CBX8 signaling [19], circPVT1 by miR-125b sponging and E2F2 induction [20], and circCSNK1G3 by interacting with miR-181 [21], affecting the proliferation, invasion and migration of breast cancer, NSCLC and prostate cancer cells, respectively. Despite increasing information on the critical role of circRNAs in several malignancies, their role and putative underlying mechanisms in LUAD have remained underexplored.

In the present study, we investigated the putative oncogenic role of circular RNA casein kinase 1 gamma 3 (circCSNK1G3, also known as hsa_circ_0001522) in LUAD development using publicly available datasets (TCGA). We found that circCSNK1G3 was aberrantly expressed in patients with LUAD. In concordance with this observation, we found using qRT-PCR that circCSNK1G3 was significantly upregulated in tissue samples from our LUAD cohort (n = 96) compared to their normal counterparts. Similarly, we found that circCSNK1G3 was overexpressed in the human NSCLC cell lines A549, H1299, H460 and H2228, compared to the normal human bronchial epithelial cell line BEAS-2B, at both the protein and mRNA levels. In addition, we observed an inverse correlation between circCSNK1G3 and miR-143-3p expression, suggestive of an interaction between them, as well as of a miR-143-3p-sponging role of the former. We also observed an association between aberrant circCSNK1G3 expression, enhanced miR-143-3p sponging, HOXA10 overexpression and subsequent LUAD progression. Our data provide pre-clinical evidence for a role of circCSNK1G3 in LUAD initiation and progression as well as its poor prognosis, highlighting a possible use as therapeutic target and prognosticator in patients with LUAD.

Materials and methods

Clinical samples

Paired LUAD and adjacent normal lung tissues were obtained from therapy-naive patients with LUAD (n = 96) from the Department of Surgical Oncology during operation at the Affiliated Hospital of Qingdao University, Qingdao, China between month 1, 2017 and month 2, 2019. All specimens were immediately snap-frozen in liquid nitrogen after surgical resection and stored at −80 °C until use. Written informed consent was obtained from each patient involved in the study and the study protocol, including collection of specimens, was reviewed and approved by the Biomedical Ethics Committee of the Affiliated Hospital of Qingdao University (Approval no.: QYFYKY 2018-10-11-2; Approval date: 10/11/2018). All procedures were concordant with the World Medical Association Declaration of Helsinki guidelines. Patient clinicopathological staging and classification procedures were in conformity with the American Joint Committee on Cancer (AJCC) Classification Criteria. In addition, RNA expression profiles and matched clinical data in The Cancer Genome Atlas (TCGA)-LUAD cohort (n = 585) were downloaded from the National Cancer Institute Genomic Data Commons (GDC) data portal (https://portal.gdc.cancer.gov/projects/TCGA-LUAD), a public cancer database, and analyzed.

Cell lines and culture

The human LUAD cell lines A549, H1299, H460 and H2228, as well as the normal human bronchial epithelial cell line BEAS-2B were purchased from the Cell Bank of the Chinese Academy of Science (Shanghai, China). The cells were cultured in Roswell Park Memorial Institute (RPMI)-1640 medium (#11875093, Thermo Fisher Scientific, Inc., Bartlesville, OK, USA), supplemented with 10% Gibco® fetal bovine serum (FBS) (#16000044, Thermo Fisher Scientific, Inc., Bartlesville, OK, USA) and 1% penicillin/streptomycin (HyClone; GE Healthcare Life Sciences, Logan, UT, USA), except the BEAS-2B cells, which were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (#12430062, Thermo Fisher Scientific, Inc., Bartlesville, OK, USA). All cells were cultured at 37 °C in a 5% CO₂ atmosphere incubator, and regularly checked and confirmed to be mycoplasma-free. The cells were sub-cultured at 98% confluence every 48–72 h.

circRNA screening, identification and expression analysis

An Arraystar human circular RNA microarray (AS-S-CR-H-V2.0, ArrayStar Inc., Rockville, MD, USA), containing 13,617 circRNAs and > 5000 probes for circRNA specific splicing sites was used. After hybridization, 4 pairs of LUAD samples (paired tumor/non-tumor tissues mentioned above) with a diameter > 1 cm were chosen for the analyses. Differentially expressed circRNAs were identified using R package Limma, with |log₂FC | ≥ 1 and adj.p value < 0.05.

RNA extraction, qRT-PCR and agarose gel electrophoresis

Total RNA was extracted from matched LUAD and non-cancerous tissues, as well as from selected LUAD cell lines using a TRIzol Plus RNA purification kit (#12183555, Thermo Fisher Scientific, Inc., Bartlesville, OK, USA) according to the manufacturer’s protocol. A PrimeScript™ RT reagent kit (#RR037A v.0610, TaKaRa Biotechnology Inc., Dalian, China) was used for circRNA and mRNA reverse transcription. RNase R (#M1228–500, BioVision Inc., Milpitas, CA, USA) was used to degrade structured/linear RNA, after which circRNA was amplified by divergent circCSNK1G3 primers with the following sequences: 5′- AAGTTGGAGCCCATGAAATC -3′ (forward) and 5′- GTAGGTAGTCAATTAACTGGAGCAT -3′ (reverse). For miRNA assessment, RNase-free DNase I (#EN0521, Thermo Fisher Scientific, Inc., Bartlesville, OK, USA) was used to eliminate genomic DNA (gDNA), after which cDNA was obtained using a Mir-X™ miRNA First-Strand Synthesis Kit (#638313, TaKaRa Biotechnology Inc., Dalian, China). circRNA, miRNA and mRNA qRT-PCR analyses were performed using SYBR® Premix Ex Taq™ II (Tli RNaseH Plus) (#RR82LR, TaKaRa Biotechnology Inc., Dalian, China) and a PrimeScript™ RT reagent kit (#RR037A v.0610, TaKaRa Biotechnology Inc., Dalian, China). Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and U6 were used as internal controls. Relative RNA expression levels were calculated using the delta-delta comparative (2^-∆∆CT) method. The PCR products were separated by 2% agarose gel electrophoresis using a Tris-Acetate-Ethylenediaminetetraacetic (TAE) running buffer at 120 V for 30 min in a Bio-Rad horizontal electrophoresis system (Bio-Rad Laboratories, Inc., Hercules, CA, USA). DL1000 DNA Marker (#3591A, TaKaRa Biotechnology Inc., Dalian, China) was used for the visualization of DNA bands.

Targeted miRNA prediction

For the prediction of potential circCSNK1G3-sponged miRNAs, the open-source bioinformatics platform for bio-molecular interactions, starBasev3.0 (http://starbase.sysu.edu.cn/) was used as described by Li et al. [22]. For miRNA-ncRNA interaction analyses, the number of AGO CLIP-seq experiments (AgoExpNum) was ≥ 5, and the minimum free energy (MFE) was set at ≤ 0 kcal/mol. MiRNA expression profiles in the TCGA-LUAD cohort (n = 585) were also assessed using the National Cancer Institute GDC data portal (https://portal.gdc.cancer.gov/projects/ TCGA-LUAD). Differentially expressed miRNAs in the downloaded TCGA-LUAD digital gene expression data cohort was analyzed using the Bioconductor edgeR package (https://bioconductor.org/packages/release/bioc/html/edgeR.html), and Venn diagrams were employed for visualization of differentially expressed genes (DEGs).

Cell transfection assays

MiR-143-3p mimic, circCSNK1G3-siRNA and HOXA10-siRNA were designed and synthesized by Shanghai GenePharma Co. Ltd. (Shanghai, China). After human LUAD A549 or H1299 cells were seeded in tissue culture plates and cultured to 60–80% confluence, they were transfected with miR-143-3p mimic, circCSNK1G3-siRNA and HOXA10-siRNA using PolyPlus (Polyplus transfection® SA, New York, NY, USA) according to the manufacturer’s protocol. The negative controls were transfected under the same conditions. After transfection, the cells were harvested 24–48 h later for RNA detection and functional analyses. All transfection assays were repeated three times in triplicates.

Cell proliferation assay

Cell proliferation was evaluated using a 3-(4,5-dimethylthiazol-2-yl)-2,5- diphenyltetrazolium bromide (MTT) assay kit (ab211091, Abcam Plc., Aibo Trading Co., Ltd., Shanghai, China). Wild type, miR- or siRNA-transfected A549 or H1299 cells were seeded in triplicates at a density of 5 × 10⁶ per well in Corning®Costar® 96-well plates (Sigma-Adrich Inc., St. Louis, MO, USA) containing 100 μl culture medium and allowed to adhere in a 5% CO₂ incubator at 37 °C overnight. Thereafter, at 0 h, 24 h, 48 h and 96 h time-points, 20 μl MTT solution was added to each well and incubated in a 5% CO₂ incubator at 37 °C for 2 h. Next, optical densities (ODs) were measured at a wavelength of 490 nm using a SpectraMax i3x multi-mode microplate reader (Molecular Devices, LLC., San Jose, CA, USA).

Western blot analysis

Wild type, miR- or siRNA-transfected A549 or H1299 cells were collected and lysed using a Radioimmunoprecipitation assay (RIPA) buffer (#89900, Thermo Fisher Scientific, Inc., Bartlesville, OK, USA). After protein concentration determination using a Bicinchoninic Acid (BCA) protein concentration determination kit (#P0011, Beyotime Institute of Biotechnology, Nantong, China), equal amounts of protein samples were separated using 10% sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to polyvinylidene difluoride (PVDF) membranes (BioSHARP, Anhui, China). Next, the membranes were blocked with 5% non-fat milk in 10x Tris-buffered saline with 0.1% Tween 20 (TBST) for 1 h, followed by incubation with primary antibodies directed against HOXA10 (#sc-271,428, 1:1000 dilution, Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA) and β-actin (#sc-8432, 1:2000 dilution, Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA) overnight at 4 °C. The membranes were subsequently washed with TBST three times for 5 min each, before 1 h incubation with a horseradish peroxidase-conjugated (HRP) goat anti-mouse IgG secondary antibody (#21040, 1:20,000 dilution, Thermo Fisher Scientific, Inc., Bartlesville, OK, USA) at room temperature. Protein bands were detected using a Pierce™ enhanced chemiluminescence (ECL) western blotting substrate (#32209, Thermo Fisher Scientific, Inc., Bartlesville, OK, USA) according to manufacturer’s protocol, and protein band intensities were quantified using NIH ImageJ software (https://imagej.nih.gov/ij/).

Scratch wound healing assay

Wild type, miR- or siRNA-transfected A549 and H1299 cells were seeded in 6-well plates and cultured to 98–100% confluence. The cell monolayers formed were scratched along the median axes using sterile 200 μl pipette tips and carefully washed with 1× Gibco® Phosphate-buffered saline (PBS, Sigma-Adrich Inc., St. Louis, MO, USA) three times before serum-free medium was added and cells were incubated in a 5% CO₂ incubator at 37 °C allowing wound healing. Cell migration was monitored over time, and wound closure captured and recorded at 0 h and 48 h using an Olympus CX31 light microscope (Olympus Corporation, Tokyo, Japan). NIH ImageJ software (https://imagej.nih.gov/ij/) was used for wound closure estimation.

Apoptosis assay

For the assessment of apoptosis, we used a BD Pharmingen™ FITC Annexin V apoptosis detection kit 1 (#556547, BD Biosciences, San Diego, CA, USA) according to manufacturer’s protocol. Briefly, after cell trypsinization and careful washing with ice-cold PBS twice, 1 × 10⁶ cells were incubated with 5 μl FITC Annexin V/5 μl PI solution for 15 min at room temperature in the dark. Thereafter, 400 μl 1x binding buffer was added to each tube and apoptosis determined within an hour using a BD FACSCanto™ II flow cytometer (BD Biosciences, San Jose, CA, USA). Data were analyzed using FlowJo software v. 10.6 (https://www.flowjo.com/).

Transwell invasion assay

Transwell cell invasion assays were performed using 8 μm pore 24-well Corning® transwell chamber plates according to the manufacturer’s instructions. Transfected or wild type cells were seeded at a density of 5 × 10⁴ cells per ml into the upper chambers coated with 100 μl diluted matrigel and containing serum-free medium, while the lower chambers were filled with FBS-rich complete growth medium as chemoattractant. After 24 h incubation of the cells in a 5% CO₂ incubator at 37 °C, the un-invaded cells left on the upper surface of the membranes were carefully removed with sterile cotton buds, after which invaded cells on the lower surfaces of the membranes were fixed with 95% ethanol, washed twice with cold 1x PBS, and stained with 0.5% crystal violet for 10 min. Representative images were obtained using an optical microscope at ×100 and ×200 magnification. Thereafter, the dye was dissolved in 90 μl 10% acetic acid, and absorbance was measured at an OD of 590 nm using a SpectraMax i3x multi-mode microplate reader (Molecular Devices, LLC., San Jose, CA, USA).

Dual-luciferase reporter assay

PmirGLO dual-luciferase miRNA target expression vectors (GenePharma, Shanghai, China) were used to construct dual-luciferase reporter plasmids. A549 and H1299 cells were co-transfected with miR-143-3p mimics and the respective luciferase reporter plasmids and incubated for 48 h, after which luciferase activity was measured using a dual-luciferase reporter kit according to the manufacturer’s instruction. For comparisons, firefly luciferase activity was normalized to Renilla luciferase activity.

Fluorescence in situ hybridization (FISH)

FISH was performed using Cy3-labeled circCSNK1G3 sequence-specific probes, which were obtained from Shanghai Outdo Biotech (Shanghai, China). Briefly, after 4% paraformaldehyde (PFA) fixation for 10 min, cells were permeabilized for 5 min in 0.5% Triton X-100/PBS solution and then hybridized with labeled circCSNK1G3 sequence-specific FISH probes overnight at 37 °C. Next, 0.1% Tween 20 in 4x saline sodium citrate (SSC) solution and 1x SSC were used to wash the cells for 5 min each, after which the nuclei were counter-stained using 4,6-diamidino-2-phenylindole (DAPI). All the procedures were performed according to the manufacturer’s protocol (Genepharma, Shanghai, China).

Mouse xenograft tumor model

Four-week-old male BABL/c nude mice (n = 20, median weight [17.5 ± 0.5 g]) were purchased from Beijing Experimental Animal Center (Chinese Academy of Sciences, Beijing, China). The mice were randomly divided into two groups, i.e., experimental (n = 10) and control (n = 10) groups. The mice in the experimental group were injected subcutaneously into the right flank with 1.5 × 10⁶ A549 cells stably transfected with circCSNK1G3 siRNA, while the mice in the control group were injected with 1.5 × 10⁶ wildtype A549 cells. The cells were suspended in 300 μl RPMI-1640/Matrigel (1:1) colloid solution. Tumor volumes (V) were monitored for 30 days and measured using a caliper every 5 days, using the formula: V (mm³) = [l × w²]/2, where ‘l’ is the longest diameter and ‘w’ is the shortest diameter of each tumor. On day 30 post-tumor cell injection, the mice were sacrificed by cervical dislocation, and palpable tumors were resected and weighed. These studies were approved by the Institutional Animal Care and Use Committee (IACUC) of the Institute of Qingdao University (IACUC-2019-011).

Statistical analysis

Results are expressed as mean ± SD of experiments performed 3 times independently in triplicates. The two-tailed Student’s t test was used for comparisons between two groups, and Pearson’s correlation coefficient was used for correlation analysis. P values < 0.05 were considered statistically significant. All data were analyzed using IBM SPSS Statistics (IBM Corp. Released 2017. IBM SPSS Statistics for Windows, Version 25.0. Armonk, NY: IBM Corp.)

Results

circCSNK1G3 expression is upregulated in patients with LUAD

In order to initially identify differentially expressed circRNAs in LUAD, we performed microarray circRNA expression analysis using 4 paired LUAD and adjacent non-tumor tissues. Our results indicated that of the 54 circRNAs that were differentially expressed (fold change > 2) between the LUAD and normal tissues (p < 0.05), 41 were downregulated and 13 were upregulated in the LUAD tissues, including hsa_circ_0001522, also referred to as circCSNK1G3, which was most significantly upregulated (p < 0.05). In Table 1, the top 5 upregulated and downregulated circRNAs are listed. Our gene expression heatmap showed a similar expression profile, with the expression of circCSNK1G3 being significantly increased in LUAD tissues compared to adjacent non-tumor tissues (Fig. 1a). Based on comparative analyses of documented circCSNK1G3 sequences from the circBase (http://www.circbase.org/) and CSNK1G3 mRNA sequence database (NM_001031812.3) from the National Institutes of Health (NIH) NCBI website (https://www.ncbi.nlm.nih.gov/nuccore /NM_001031812.3), we conclude that the upregulated circCSNK1G3 is circularized and consists of back-spliced exons 2–4 of the CSNK1G3 gene located on chromosome 5, as confirmed by Sanger sequencing (Fig. 1b).

Table 1.

Top 5 upregulated and downregulated circRNAs according to the circRNA microarray-based analysis

CircRNA ID	Position	Spliced length	Best transcript	Gene symbol	Fold change	P value
Upregulated
hsa_circ_0004315	chr16:74491771–74,493,687	229	NR_027264	GLG1	2.75	0.00652
hsa_circ_0001522	chr5:12288111–122,893,258	536	NM_001044723	CSNK1G3	2.56	0.02602
hsa_circ_0000512	chr14:20811282–20,811,436	154	NR_002312	RPPH1	2.53	0.00201
hsa_circ_0032704	chr14:76173360–76,187,046	457	NM_015072	TTLL5	2.23	0.0173
hsa_circ_0007503	chr10:28872327–28,884,970	645	NR_024557	WAC	2.18	0.00347
Downregulated
hsa_circ_0043436	chr17:37580879–37,596,930	640	NM_004774	MED1	3.12	0.0408
hsa_circ_0005265	chr20:2928627–2,945,848	554	NM_080841	PTPRA	2.87	0.0109
hsa_circ_0004751	chr17:35310185–35,311,207	549	NM_012138	AATF	2.76	0.00467
hsa_circ_0006220	chr17:35800605–35,800,763	158	NM_001488	TADA2A	2.71	0.0354
hsa_circ_0072066	chr5:31508727–31,508,882	155	NM_013235	DROSHA	2.59	0.00841

Fig. 1.

circCSNK1G3 expression is upregulated in LUAD patients. a Heatmap showing differentially expressed circRNAs in LUAD vs. normal tissue samples. b Schematic representation of circCSNK1G3 formation from back-spliced CSNK1G3, and visualization of the Sanger-sequenced splice junction

circCSNK1G3 is primarily localized in the cytoplasm and stably overexpressed in clinical LUAD samples and cell lines

To assess whether circCSNK1G3 back-splicing could result from either genomic rearrangement or trans-splicing, two sets of primers, divergent and convergent, were designed. The divergent primer was used for circCSNK1G3 amplification, and the convergent primer for CSNK1G3 mRNA detection. After 2% agarose gel electrophoresis of cDNA and gDNA isolated from A549 cells, we found that circCSNK1G3 and CSNK1G3 were amplified from cDNA using the divergent and convergent primers, respectively. However, from the extracted gDNA only amplified CSNK1G3 products were obtained from convergent primers, but no circCSNK1G3 products (Fig. 2a). Additionally, using a RNase R degradation assay we found that upon exposure to RNase R, linear CSNK1G3 was significantly decreased (65.4%, p < 0.01), whereas no effect on circCSNK1G3 was apparent and statistically insignificant, thus confirming the stability of circCSNK1G3 (Fig. 2b). Moreover, using FISH, we found that circCSNK1G3 is localized predominantly (~ 90%) in the cytoplasm of A549 or H1299 cells (Fig. 2c). We also noted that the expression of circCSNK1G3 was significantly enhanced in the LUAD cell lines A549, H1299, H460 and H2228, compared to the normal lung epithelial cell line BEAS-2B (Fig. 2d). As a validation of our in vitro findings we found, using circRNA microarray analyses of our matched LUAD/non-tumor cohort (n = 96 pairs), that circCSNK1G3 was significantly overexpressed in the LUAD samples compared to their normal counterparts (p < 0.001) (Fig. 2e). Similarly, qRT-PCR analyses revealed that the median circCSNK1G3 expression level in the LUAD tissues was 42.8% (p < 0.001) higher than that in the corresponding normal tissues (Fig. 2f). Based on this upregulation, we next set out to test whether circCSNK1G3 expression is related to the prognosis of LUAD. Through the analysis of relationships between clinical data and circCSNK1G3 expression in LUAD samples, we found that the expression markedly related to pathologic stage (p = 0.028), tumor size (cm) (p = 0.001) and lymph node metastasis (p = 0.020) (Table 2).

Fig. 2.

circCSNK1G3 is primarily localized in the cytoplasm and stably overexpressed in LUAD clinical samples and cell lines. a Representative RT-PCR image showing that circCSNK1G3 is amplified by divergent primers from cDNA but not gDNA in A549 cells. b Graph showing circCSNK1G3 and parent CSNK1G3 mRNA expression levels in A549 cells in the presence or absence of RNase R, as detected by qRT-PCR. c Representative FISH images showing the subcellular localization of circCSNK1G3 in A549 and H1299 cells. The circCSNK1G3 probes were labeled with Alexa Fluor 555, while DAPI served to stain nuclei. Scale bars, 50 μm. d Graph showing the relative expression of circCSNK1G3 in BEAS-2B, H460, A549, H1299 and NCI-H2228 cells. e Line graph of the relative expression of circCSNK1G3 in LUAD compared to normal lung tissues form our matched LUAD/non-tumor cohort (n = 96 pairs) as detected by microarray analysis. f Graphical representation of the relative expression of circCSNK1G3 in LUAD compared to normal lung tissues form our matched LUAD/non-tumor cohort (n = 96 pairs) as detected by qRT-PCR. * p < 0.05, ** p < 0.01, *** p < 0.001

Table 2.

Correlation between circCSNK1G3 expression and clinical characteristics in LUAD

Characteristics	Total	circCSNK1G3 expression		P value
		low	high
Total	96	33	63
Age (years)				0.727
< 60	50	18	32
≥ 60	46	15	31
Sex				0.968
Male	41	14	27
Female	55	19	36
Pathologic stage				0.028**
I + II	62	20	42
III + IV	34	13	21
Tumor size (cm)				0.001***
< 1	44	25	19
≥ 1	52	8	44
Smoking				0.178
Yes	38	10	28
No	58	23	35
Lymph node metastasis				0.020**
Yes	59	15	44
No	37	18	19

*p < 0.05, **p < 0.01, ***p < 0.001

circCSNK1G3 silencing inhibits the proliferation and migration of LUAD cells

To better understand the role of circCSNK1G3 in LUAD, we employed a loss of function approach using short interfering (si)RNAs 1 and 2, designed according to the CSNK1G3 back-spliced region and specifically targeting circCSNK1G3 in A549 and H1299 cells. Using qRT-PCR, we observed 51% (p < 0.01) or 50% (p < 0.01) knockdown efficiencies using circCSNK1G3 siRNA1 and siRNA2, respectively, in A549 cells. In H1299 cells the knockdown efficiencies were 50% and 31.6% for circCSNK1G3 siRNA1 and siRNA2, respectively (Fig. 3a). Having stably silenced circCSNK1G3 in A549 and H1299 cells, we next evaluated its effect on the proliferation of LUAD cells. We found that compared to wild-type A549 cells, A549 cells transfected with circCSNK1G3 siRNA (si-circCSNK1G3) exhibited 22%, 49%, 41% and 43.1% reductions (p < 0.05) in numbers of viable cells after 24, 48, 72 and 96 h (Fig. 3b, upper panel), while 2% - 55% (p < 0.01) reduction in the proliferation of H1299 cells was observed over a 96 h time-span (Fig. 3b, lower panel). Furthermore, we found that compared with the wild-type control cells, si-circCSNK1G3 induced a 50.4% (p < 0.05) or 40.9% (p < 0.05) lag in wound closure (i.e., migration) by A549 or H1299 cells, respectively (Fig. 3c). Similarly, we found that si-circCSNK1G3 caused a significant decrease in LUAD cell invasion, i.e., 66.7% (p < 0.01) or 72.5% (p < 0.01) reductions in the number of invaded A549 and H1299 cells, respectively (Fig. 3d). In addition, using FITC-Annexin V/PI staining, we found that si-circCSNK1G3 increased the apoptotic fractions (Q2 + Q3) from 10.5% in control to 36.7% in transfected A549 cells (Fig. 3e).

Fig. 3.

circCSNK1G3 silencing inhibits the proliferation and migration of LUAD cells. a Graph of the knockdown efficacy of siRNA1 and siRNA2 targeting circCSNK1G3 expression in A549 and H1299 cells, as detected by qRT-PCR. b Line graphs showing the effect of si-circCSNK1G3 on the proliferation of A549 or H1299 cells, as detected by the MTT assay. Representative images and graphs showing the effect of si-circCSNK1G3 on A549 and H1299 (c) cell migration, as assessed by scratch wound healing migration assays, and (d) invasion, as determined by Transwell matrigel invasion assays. (e) FITC-Annexin V/PI staining data showing the effect of si-circSNK1G3 on the number of apoptotic A549 or H1299 cells. * p < 0.05, ** p < 0.01, *** p < 0.001

circCSNK1G3 interacts with and serves as a sponge for miR-143-3p in LUAD cells in vitro and in vivo

Based on the increasing implication of circRNAs in the ‘sponging’ and subsequent inhibition of the function of corresponding miRNAs [23], we next set out to investigate a putative role of circCSNK1G3 as a sponge of miRNAs in the TCGA-LUAD cohort using the starBase 3.0 platform (http://starbase.sysu.edu.cn/). First, we generated a Venn diagram showing that of the 4943 genes probed, 2193 were miRNAs and 2750 circRNAs, with 145 circRNAs/miRNAs overlapping in the TCGA-LUAD cohort (Fig. 4a). Furthermore, using a heatmap of differentially expressed miRNAs in the TCGA-LUAD cohort showing the top 50 enhanced and suppressed miRNAs, we identified hsa-miR-143-3p as one of the most differentially expressed miRNAs based on fold change in expression and statistical significance (Fig. 4b). Moreover, using miRanda-based bioinformatics analyses (http://www.microrna.org/), we found that circCSNK1G3 may directly interact with hsa-miR-143-3p at the chr5:122950141-122950161[+] locus with a very low miRNA support vector regression (mirSVR) score of −0.7777 and a high PhastCons score of 0.7200 (Fig. 4c), suggesting that circCSNK1G3 may serve as a sponge for hsa-miR-143-3p in LUAD cells. To confirm the circCSNK1G3/miR-143-3p interaction, we performed a dual-luciferase reporter assay and found that transfection with miR-143-3p mimic significantly suppressed the luciferase activity of wild-type (WT) circCSNK1G3 compared with the miR negative control (miR-NC) (p < 0.05) in both A549 and H1299 cells. Interestingly, the miR-143-3p mimic had no apparent effect on the luciferase activity of circCSNK1G3-Mut (Fig. 4d). In addition, we found that the expression of miR-143-3p was significantly upregulated (41.2%, p < 0.01) in A549 and H1299 cells transfected with si-circCSNK1G3 (Fig. 4e). We also found that miR-143-3p and circCSNK1G3 were inversely correlated (r = −0.34, p < 0.001) in our matched LUAD cohort (Fig. 4f). Based on our observation that circCSNK1G3 can sponge miR-143-3p, we next assessed differential expression of miR-143-3p in our matched LUAD cohort (n = 96 pairs) using qRT-PCR. By doing so, we found that compared to the LUAD samples, miR-143-3p was 60.8% (p < 0.01) more highly expressed in the normal lung samples (Fig. 4g). We also observed 49.5% (p < 0.05) and 61% (p < 0.01) reductions in miR-143-3p expression in A549 or H1299 cells, respectively, compared with non-tumor BEAS-2B cells (Fig. 4h).

Fig. 4.

circCSNK1G3 interacts with and serves as a sponge for miR-143-3p in LUAD, in vitro and in vivo. a Venn diagram showing differentially expressed genes and circRNAs/miRNAs overlapping in the TCGA-LUAD cohort, as predicted by starBase 3.0. b Clustered heatmap of markedly differentially expressed miRNAs in the LUAD vs. normal samples from the TCGA-LUAD cohort. c miRanda-predicted binding sites of miR-143-3p on circCSNK1G3. d Graphs showing luciferase reporter activity of circCSNK1G3 in A549 and H1299 cells co-transfected with mutant or wild-type circCSNK1G3 and miR-143-3p or miR-NC. e Graph of the effects of si-circCSKN1G3 on the relative expression of miR-143-3p in A549 and H1299 cells. f Correlation analysis of miR-143-3p and circCSNK1G3 expression in our LUAD cohort. g Graphical representation of the relative expression of miR-143-3p in LUAD compared to normal lung tissues form our matched LUAD/non-tumor cohort (n = 96 pairs). h Graph showing the relative expression of miR-143-3p in BEAS-2B, A549 and H1299 cell lines. * p < 0.05, ** p < 0.01, *** p < 0.001; DEG, differentially expressed genes; Mut, mutant; WT, wild-type; NC, negative control; r, Spearman correlation

circCSNK1G3/miR-143-3p interaction modulates HOXA10 signaling

Having shown that circCSNK1G3 interacts with and sponges miR-143-3p, we next sought to understand the putative role of such an interaction in LUAD progression. First, our pan-cancer analysis using the starBase 3.0 platform revealed that amongst the differentially expressed genes in the TCGA-LUAD cohort, the expression of HOXA10 was 17.75-fold increased (p = 1.2e-10) in LUAD compared to normal tissues, and was inversely correlated with hsa-miR-143-3p expression (r = −0.052, p = 0.24) (Fig. 5a). This observation was corroborated in our matched LUAD cohort (n = 96 pairs) showing a 1.87-fold increase in HOXA10 expression compared with the normal samples (p < 0.001) (Fig. 5b). Similarly, we found that HOXA10 was relatively overexpressed in A549 (4-fold, p < 0.01), H1299 (3.8-fold, p < 0.01), H460 (3.9-fold, p < 0.01) and H2228 (2.76-fold, p < 0.01) cells, compared to the normal bronchus epithelial BEAS-2B cells (Fig. 5c). Next, we transfected the LUAD cells with siRNAs, achieving knockdown efficiencies of 54% (p < 0.01) and 49.1% (p < 0.05) in A549 siRNA1 and siRNA2 cells, respectively, and 49% (p < 0.01) and 30.4% (p < 0.05) in H1299 siRNA1 and siRNA2 cells, respectively (Fig. 5d). As with si-circCSNK1G3, si-HOXA10 induced 60%, 61.5%, 26.5% and 24.2% reductions in the proliferation of A549 cells at the 24 h, 48 h, 72 h and 96 h time-points (p < 0.01), with similar suppressed proliferation trends in H1299 cells (Fig. 5e). As expected, si-HOXA10 also significantly suppressed the ability of the A549 and H1299 cells to migrate (Fig. 5f) and reduced the number of invading A549 or H1299 cells by 65.5% (p < 0.01) and 64.7% (p < 0.01), respectively (Fig. 5g). Furthermore, to confirm the modulatory role of circCSNK1G3/miR-143-3p signaling on HOXA10 expression and activity, we transfected the LUAD cells with miR-143-3p mimic and/or circCSNK1G3 mimic or miR-NC. We found that while miR-143-3p significantly downregulated the expression level of HOXA10, circCSNK1G3 upregulated this expression in the A549 and H1299 cells. Interestingly, transfection with circCSNK1G3 was found to be sufficient to abrogate the HOXA-suppressing activity of miR-143-3p (Fig. 5h). These findings, at least in part, show that circCSNK1G3 may serve as a ‘miR-143-3p sponge’ to induce and enhance HOXA10 oncogenic signals and, thereby, facilitate LUAD progression.

Fig. 5.

circCSNK1G3/miR-143-3p interaction modulates HOXA10 signaling. a Box and whiskers plot of the differential expression of HOXA10, and dot plot showing the correlation between HOXA10 and miR-143-3p expression in the TCGA-LUAD cohort as analyzed by starBase 3.0 platform. b Box and whiskers plot of the differential expression of HOXA10 in LUAD and normal tissue samples from our matched LUAD cohort. c Graph of the relative expression of HOXA10 in BEAS-2B, A549, H1299, H460 and NCI-H2228 cells. d Graphs showing the knockdown efficacy of HOXA10-targeting siRNA1 or siRNA2 in A549 and H1299 cells, as determined by qRT-PCR. e Line graph of the effect of siHOXA10 on the proliferation of A549 and H1299 cells at the indicated time-points. f Representative images and graphs showing the effect of siHOXA10 on the migration of A549 and H1299 cells at the indicated time-points. g Representative images and graphs showing the effect of siHOXA10 on the number of invaded A549 and H1299 cells. h Representative western blot images showing how transfection with miR-143-3p and/or circCSNK1G3 affects the expression of HOXA10 protein. β-actin served as internal control. * p < 0.05, ** p < 0.01, *** p < 0.001; siHOXA10, cells transfected with siRNA targeting HOXA10; OD, optical density; FDR, false discovery rate; r, Spearman correlation

circCSNK1G3 enhances LUAD tumor growth and proliferation in vivo

Finally, we examined the effect of circCSNK1G3/miR-143-3p/HOXA10 signaling on the tumorigenesis and growth of LUAD cells in vivo using mice (n = 20) subcutaneously injected with 1.5 × 10⁶ wild-type (n = 10) or si-circCSNK1G3-tranfected (n = 10) A549 cells. We found that over a 30-day experiment course, the si-circCSNK1G3 mice exhibited significantly fewer (data not shown), smaller (p < 0.01) and lighter (p < 0.01) tumors, compared with the control group (Fig. 6a and b). qRT-PCR analysis of miR-143-3p expression in the tumors resected from the mice on day 30 showed that mice bearing si-circCSNK1G3 tumors exhibited a significantly enhanced expression of miR-143-3p, compared to the control mice (p < 0.01) (Fig. 6c). In addition, using IHC, we observed a significant concurrent downregulation of the oncogene HOXA10 (p < 0.01) and the proliferation marker Ki67 (p < 0.01) in tumors resected from the si-circCSNK1G3 mice (Fig. 6d).

Fig. 6.

circCSNK1G3 enhances LUAD growth and cell proliferation in vivo. a Graph and representative image showing the effect of si-circCSNK1G3 on the volume of tumors formed as measured by caliper every 5 days. b Graph showing the effect of si-circCSNK1G3 on the weight of tumors at the end of the experiment on day 30 after tumor cell inoculation. c Graph of relative miR-143-3p expression in tumors from mice inoculated with control or si-circCSNK1G3-transfected A549 cells. d Representative IHC image and graph of the effect of si-circCSNK1G3 on Ki-67 and HOXA10 expression in tumor samples from the xenograft tumor groups. Scale bar: 50 μm. p < 0.01, *** p < 0.001

Discussion

During the last 3 decades the differential expression and vital role of non-coding RNAs, such as miRNAs and long non-coding (lnc)-RNAs, in the initiation and/or progression of various types of cancer has amply been documented [24–26]. Recently, circRNAs have been added to this list [27], including their implications in the initiation, proliferation, metastasis, progression and prognosis of several malignancies [28–30]. As yet, however, the implications of circRNAs in LUAD have remained relatively underexplored. In the present study, we investigated the hypothesis that circCSNK1G3 through the induction of downstream oncogenic signaling and miRNA sponging, may promote the growth and metastasis of LUAD cells. We found that the predominantly cytoplasmic circular RNA circCSNK1G3 is overexpressed in clinical LUAD samples and cell lines, and that circCSNK1G3 silencing inhibits the proliferation, migration and invasion of LUAD cells through upregulation of miR-143-3p, indicating a role for circCSNK1G3 as a sponge for miR-143-3p in LUAD, both in vitro and in vivo. Dysregulated circCSNK1G3/miR-143-3p signaling was also found to modulate HOXA10 signaling, and to enhance LUAD tumor cell proliferation in vivo. These findings are consistent with observations that circRNA perturbations are common in human cancers [31].

First, we assessed correlations between circCSNK1G3 expression and clinical data, showing that circCSNK1G3 expression was related to tumor size, pathological stage and lymph node metastasis. We also found that loss of circCSNK1G3 expression inhibits the proliferation, migration and invasion of LUAD cells. These observations are in line with results of recent studies showing that the circular RNA SWI/SNF-related Matrix-associated Actin-dependent Regulator of Chromatin, subfamily A, member 5 (circSMARCA5), which is aberrantly expressed in prostate cancer, enhances cell proliferation [32], similar to circular BTG3-associated nuclear protein (circBANP) and circular ATP-binding cassette subfamily B member 10 (circABCB), which promote the proliferation of colorectal and breast cancer cells, respectively [33, 34]. Furthermore, in concordance with emerging roles of circRNAs in biomolecule (microRNAs or proteins) sequestration, modulation of gene transcription and translation, as well as interference with splicing [35], we found that circCSNK1G3 interacts with and serves as a sponge for the miR-143-3p in LUAD cells and that this interaction modulates HOXA10 signaling. HOXA10 (also known as PL, HOX1, HOX1H, HOX1.8) has been identified as a mediator and/or effector that plays a vital role in the development of various cancers [36–39]. We found that HOXA10 is upregulated in LUAD, which is consistent with a previous report by Guo et al. [40]. Moreover, using loss of function experiments, we found that HOXA10 may act as an oncogene in LUAD. This is relevant in the context of the known tumor suppressor role of miR-143-3p in several cancer types, including esophageal squamous cell carcinoma [41], gastric cancer [42] and osteosarcoma [43], as well as with accumulating evidence implicating circRNAs as miRNA sponges, as e.g. circular matrix metalloproteinase 9 (circMMP9) which sponges miR-124 [16], circSFMBT which sponges miR-182-5p [18], circular plasmacytoma variant translocation member 1 (circPVT1) which sponges miR-125b [20], circular mitochondrial TRNA translation optimization 1 (circMTO1) which sponges miR-9 [23], or CDR1 which sponges miR-7 [32], in a broad range of malignancies.

Taken together, our data indicate that circCSNK1G3 may induce HOXA10 signaling and promote the growth and metastasis of LUAD cells by hsa-miR-143-3p sponging (Fig. 7). Our results highlight the targetability of the circCSNK1G3/miR-143-3p/HOXA10 signaling axis in patients with LUAD.

Fig. 7.

Diagram showing role of circCSNK1G3/miR-143-3p/HOXA10 in LUAD initiation and progression. HOXA10-mediated enhanced expression of circular RNA circCSNK1G3 promotes the growth and metastasis of lung adenocarcinoma cells through hsa-miR-143-3p sponging

Authors’ contributions

Study conception and experimental design: Tong Lu, Tong Qiu, Bin Han and Yuanyong Wang. Performed the experiments: Tong Lu, Xiao Sun and Yi Qin. Data collation and analysis: Ao Liu and Nan Ge. Manuscript writing: Tong Lu, Tong Qiu and Bin Han. Provided reagents, materials, and experimental infrastructure: Wenjie Jiao. All authors read and approved the final submitted version of the manuscript.

Funding

This study was funded by Key R & D programs in Shandong Province (grant number 2018GSF118119).

Data availability

The datasets used and analyzed in the current study are publicly accessible from TCGA (https://www.cancer.gov/tcga.)

Compliance with ethical standards

Conflict of interest

The authors declare that they have no potential conflicting interests.

All authors are working for either university or hospitals. We claim that we do not have any actual or potential conflict of interest, including any financial, personal, or other relationships with other people or organizations within three years of beginning the work submitted that could inappropriately influence our work.

Ethics approval and consent to participate

Written informed consent was obtained from each patient involved in the study, and the study was approved by the Institutional Review Board (IRB) of the Affiliated Hospital of Qingdao University (Approval number: Qingdao N201903047).