The GEPIA website (http://gepia.cancer-pku.cn/) was used to screen for potential oncogenes in LUAD and to analyze the correlation between gene expression and clinical characteristics. The differentially expressed genes in LUAD were obtained by searching the ‘Differential Expression Analysis’ section of GEPIA website (http://gepia2.cancer-pku.cn/#degenes) with the following terms: Cancer name, LUAD; |log2 fold-change (FC)|≥1; q-value ≤0.01 and differential methods, ANOVA. The top 500 differential survival genes in LUAD were obtained by searching the ‘Most Differential Survival Genes’ section of GEPIA website (http://gepia2.cancer-pku.cn/#survival) with the following terms: Cancer name, LUAD and group cut-off, median. The Kaplan-Meier curves were downloaded from the ‘Survival Analysis’ section of GEPIA website (http://gepia2.cancer-pku.cn/#survival) with the following terms: Gene, FAM83A or FAM83A-AS1; group cut-off, median and cancer name, LUAD. The staging boxplots were downloaded from the ‘Expression DIY-Stage Plot’ section of GEPIA website (http://gepia2.cancer-pku.cn/#analysis) with the following terms: Gene, FAM83A or FAM83A-AS1 and cancer name, LUAD. The median expression (transcripts per million) of genes in normal and LUAD tissue was obtained from ‘Differential Expression Analysis’ section of GEPIA website (http://gepia2.cancer-pku.cn/#degenes). The hierarchical clustering map in Fig. 1B was drafted with these data using GraphPad Prism version 6 (GraphPad Software). The Kaplan-Meier curves were downloaded from the ‘Survival Analysis’ section of GEPIA website (http://gepia2.cancer-pku.cn/#survival) with the following terms: Gene, FAM83A or FAM83A-AS1; group cut-off, median and cancer name, LUAD. The violin plots in Fig. 2B and E were downloaded from the ‘Expression DIY-Stage Plot’ section of GEPIA website (http://gepia2.cancer-pku.cn/#analysis) with the following terms: Gene, FAM83A or FAM83A-AS1 and cancer name, LUAD. The boxplots in Figs. 3A and 4A were downloaded from the CCLE dataset (https://portals.broadinstitute.org/ccle) by searching for FAM83A or FAM83A-AS1. The Ensembl genome browser (http://asia.ensembl.org/index.html) was scanned to observe the genomic location of FAM83A and FAM83A-AS1. Gene clusters highly correlated with FAM83A and FAM83A-AS1 (Pearson score >0.3) were submitted to the database for Annotation, Visualization and Integrated Discovery Bioinformatics resources version 6.8 (http://david.abcc.ncifcrf.gov/) (12) for Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis, and GO terms and KEGG pathways with a gene count ≥3 were included in subsequent analysis. The mRNA expression data of FAM83A and FAM83A-AS1 in different types of cancer cells was downloaded from the CCLE dataset (https://portals.broadinstitute.org/ccle). The histograms in Figs. 3B and 4B were constructed using these data and GraphPad. The secondary structure of FAM83A-AS1 was downloaded from the HUGO Gene Nomenclature Committee (http://www.genenames.org/) (13). The cBioPortal website (http://www.cbioportal.org/) (14) was utilized to analyze the correlation between FAM83A and FAM83A-AS1. LncBase version 2 (http://carolina.imis.athena-innovation.gr/diana_tools/web/index.php) (15) and StarBase version 2.0 (http://starbase.sysu.edu.cn/starbase2/browseNcRNA.php) (16) were browsed to identify the miRNAs that bind FAM83A and FAM83A-AS1.

NCIH1650 and A549 cells was purchased from the Shanghai Institutes for Biological Science and were cultured in RPMI1640 (Nanjing KeyGen Biotech Co., Ltd.) medium containing 10% fetal bovine serum (Invitrogen; Thermo Fisher Scientific, Inc.) in a humidified incubator containing 5% CO₂ at 37°C. LUAD cells were seeded in 6-well plates with a density of 5×10⁵ cells per well. Two small interfering (si)RNAs targeting FAM83A/FAM83A-AS1, miR-495-3p mimics and miR-495-3p inhibitors were constructed and purchased from Guangzhou RiboBio Co., Ltd, and si-NC or miR-NC (non-targeting) was used as a control. Transfection of siRNA and miRNA mimics was performed in room temperature for 6 h according to the Lipofectamine^® 3000 reagent (Invitrogen; Thermo Fisher Scientific, Inc.) protocol. For 6-well plates, 5 µl siRNAs (20 nM) or miRNAs mimics (20 nM) were transfected in each well. Subsequent experiments were performed after 48 h of transfection. The rescue experiments were performed by transfecting siRNA of FAM83A/FAM83A-AS1 and miR-495-3p inhibitor. The sequence of siRNAs and miRNA mimics were as follows: si-NC, forward: 5′-UUCUCCGAACGUGUCACGUTT-3′and reverse: 5′-ACGUGACACGUUCGGAGAATT-3′; si1-FAM83A, forward: 5′-UUCUCCGAACGUGUCACGUTT-3′ and reverse: 5′-ACGUGACACGUUCGGAGAATT-3′; si2-FAM83A, forward: 5′-UGAACUUCUCCCGGAUUUGTT-3′ and reverse: 5′-CAAAUCCGGGAGAAGUUCATT-3′; si1-FAM83A-AS1, forward: 5′-GCUGCCACCUACAAGAUAATT-3′ and reverse: 5′-UUAUCUUGUAGGUGGCAGCTT-3′; si2-FAM83A-AS1, forward: 5′-GGCCCUGGGCUGAAUAAUUTT-3′ and reverse: 5′-AAUUAUUCAGCCCAGGGCCTT-3′; miR-NC, 5′-UUCUCCGAACGUUCACGUTT-3′; miR-495-3p mimics, 5′-AGGAUGUCUAAAUGUUUGUUA-3′ and miR-495-3p inhibitor, 5′-AAGAAGUGCACCAUGUUUGUUU-3′.

TRIzol^® reagent (Invitrogen; Thermo Fisher Scientific, Inc.) was used to extract the total RNAs from cells transfected as aforementioned. cDNAs were obtained by reverse transcription using a Reverse Transcription kit (cat. no. RR036A; Takara Bio, Inc.) at 37°C for 15 min and 85°C for 5 min. RT-qPCR was conducted with SYBR Select Master mix (cat. no. 4472908; Applied Biosystems; Thermo Fisher Scientific, Inc.) and the relative expression was calculated using the 2^−ΔΔCq method (17), with ACTB as a reference gene. The thermocycling conditions for RT-qPCR were as follows: Initial denaturation, 95°C for 30 sec; denaturation, 95°C for 5 sec, 60°C for 35 sec, for a total of 40 cycles; annealing and extension, 95°C for 15 sec, 60°C for 1 min and 95°C for 15 sec. The primers used are shown in Table SI.

For protein extraction, cells transfected as aforementioned were lysed in RIPA lysis buffer with freshly added protease inhibitor PMSF (1:100; Nanjing KeyGen Biotech Co., Ltd.). Protein concentration was determined using a BCA kit (Nanjing KeyGen Biotech Co., Ltd.) and 30 µg of lysate protein was loaded per lane for western blotting. Briefly, protein extracts were loaded on 10% SDS-PAGE gels for electrophoresis and then were electrotransferred to a PVDF membrane. The membrane was blocked at room temperature with non-fat milk for 2 h and incubated overnight with respective primary antibodies at 4°C. The following primary antibodies were used: Anti-β-actin (1:1,000; cat. no. 3700; Cell Signaling Technology, Inc.) and anti-FAM83A (1:1,000; cat. no. ab214014; Abcam). Then secondary antibodies (goat anti-rabbit antibody, 1:10,000; cat. no. ab6721; Abcam, or goat anti-mouse antibody, 1:10,000; cat. no. ab6789; Abcam) were used to incubate the membrane for 2 h. Blots were visualized using ECL detection with an ECL Substrate kit (cat. no. ab133406; Abcam) and ImageJ (version 1.48; National Institutes of Health) software was used to quantify protein expression.

An EdU Apollo^® 488 In Vitro Imaging kit (Guangzhou RiboBio Co., Ltd.) was used to detect proliferation rate of NCIH1650 and A549 cells. After transfection of 24 h, 8,000 cells/100 µl were plated in 96-well plates overnight and then incubated with 100 µl 50 µM EdU solution for 2 h. The cells were then fixed with 50 µl 4% paraformaldehyde at room temperature for 10 min and neutralized using 50 µl 2 mg/ml glycine. After permeabilization with 0.5% Triton X-100 for 20 min, the cells were stained for DAPI with 100 µl 1× Apollo solution and stained for Edu with 100 µl 1× Hoechst33342 solution, both at room temperature for 30 min. Images were captured using fluorescence microscopy with 40 times magnification and proliferation cells ratios were counted from three random fields of view.

The CCK-8 assay was performed to evaluate the relative number of NCIH1650 and A549 cells. After transfection of 24 h, 2,000 cells/100 µl were plated in 96-well plates and total of 20 µl CCK-8 reagent (Guangzhou RiboBio Co., Ltd.) was added to each test well and cells were then incubated for 2 h at 37°C. The absorbance at 450 nm was measured every 24 h.

Transwell assay inserts (8 µM PET, 24-well Millicell) and Matrigel-coated membranes (BD Biosciences) were used to detect the migration and invasion of cells, respectively. In total, 200 µl serum-free medium was added to the upper chamber and 800 µl 10% DMEM serum-containing medium to the lower chamber. Matrigel pre-coating was performed at 37°C for 30 min. The NCIH1650 and A549 cells were harvested 24 h after transfection and added to the upper chamber (50,000 cells). After incubation at 37 C for 24 h (migration assay) or 48 h (invasion assay), the migrated or invaded cells were fixed with 4% polyformaldehyde and stained with crystal violet, both at room temperature for 30 min. Images were captured with light microscopy and cells were counted by ImageJ software.

A Dual Luciferase Reporter Assay system (Promega Corporation) was used for luciferase assays. The binding sites of FAM83A-AS1 with FAM83A 3′UTR were identified using StarBase as aforementioned. A fragment of the FAM83A-AS1 or FAM83A 3′UTR sequence containing a putative target site or mutated target site for miR-495-3p was amplified and cloned into the pGL3 reporter vector (Guangzhou RiboBio Co., Ltd.). The luciferase reporter plasmids were co-transfected miR-495-3p mimics or miR control (Guangzhou RiboBio Co., Ltd.) as aforementioned. After transfection for 48 h, Renilla and Firefly luciferase activities were detected in the dual luciferase reporter gene kit (Promega Corporation). Firefly/Renilla value was used to measure relative luciferase activity and results were normalized to the activity of the control. The sequences of miRNA mimic/inhibitor were as follows: miR-NC, 5′-UUCUCCGAACGUUCACGUTT-3′; miR-495-3p mimics, 5′-AGGAUGUCUAAAUGUUUGUUA-3′ and miR-495-3p inhibitor, 5′-AAGAAGUGCACCAUGUUUGUUU-3′

Generally experiments were repeated three times and data are presented as the mean ± standard deviation (unless otherwise shown). Unpaired student's t-tests were used to determine statistical significance between two groups. ANOVA test and Dunnett's post hoc were used for the comparison of multiple groups. Spearman's test was used for the correlation analysis of two genes. Graphs were made using the GraphPad Prism 6.0 software package. Statistical analysis was performed using SPSS version 20.0 (IBM Corp) and P<0.05 was considered to indicate a statistically significant difference.

To screen for genes abnormally expressed and associated with survival of patients with LUAD, the following filter criterion was used: Cancer name, LUAD; gene expression, |log2 FC|≥1, q-value ≤0.01 (tumor vs. normal); differential survival genes with overall survival (OS) rates (top 500) and disease-free survival (top 500) rates. Using this method, 42 genes were obtained (Fig. 1A). To sort these genes by differences in expression, it was demonstrated that FAM83A was the most differentially expressed gene and its antisense lncRNA, FAM83A-AS1, was also among the list (Fig. 1B). As shown in Ensembl genome browser, FAM83A-AS1 is the antisense of FAM83A and both have several different transcripts (Fig. 1C).

Further analysis of GEPIA indicated that FAM83A was significantly upregulated in LUAD (tumor, n=483 vs. normal, n=347) (Fig. 2A) and its expression was relatively higher in patients with LUAD with more advanced stages (Fig. 2B). Survival analysis showed that patients of high FAM83A (n=239) had a worse OS (P<0.001) compared paired with patients with low FAM83A expression (n=239) (Fig. 2C). Analogously, FAM83A-AS1 was also overexpressed in LUAD and associated with tumor stage and survival of patients (Fig. 2D-F). Moreover, to predict the potential biological function of FAM83A and FAM83A-AS1, genes highly co-expressed with these were submitted for GO and KEGG pathway analysis, respectively. Results suggested that FAM83A and FAM83A-AS1 bear strong similarities in biological processes, mainly including ‘cell-cell adhesion’ and ‘central carbon metabolism in cancer’ (Fig. 2G-J). FAM83A and FAM83A-AS1 both associated with clinical characteristics of LUAD and it is important to deeply study their roles in LUAD.

The CCLE dataset was searched to evaluate the expression of FAM83A in different cancer cells and the results indicated that the expression of FAM83A in NSCLC (lung_NSC) and multiple other types of cancer cell was elevated (Fig. 3A). In the list of lung_NSC cell lines with high expression of FAM83A, the common LUAD cell line, NCIH1650, has been used for research in a previous study (18) and was selected for subsequent experiments (Fig. 3B). To explore the biological role of FAM83A, two siRNAs targeting FAM83A were transfected into NCIH1650 and both siRNAs effectively silenced the expression of FAM83A, confirmed by RTq-PCR and western blotting (Fig. 3C). CCK-8 and EdU assays validated that knockdown of FAM83A inhibited proliferation of NCIH1650 cells (Fig. 3D and E). Transwell and Matrigel assays revealed that silencing FAM83A suppressed the ability of NCIH1650 cells to migrate and invade (Fig. 3F and G). Additionally, knockdown of FAM83A inhibited proliferation, migration and invasion of A549 cells (Fig. S1). These results suggested that FAM83A played a vital role in the carcinogenesis of LUAD.

The average expression of FAM83A-AS1 in lung_NSC cells was high, ranking 5th in all types of cancer (Fig. 4A). The expression of FAM83A-AS1 in NCIH1650 was also elevated compared with other common LUAD cell lines (Fig. 4B). FAM83A-AS1 is an antisense transcript and with limited ability to encode proteins (19,20). Two siRNAs were transfected to knock down the expression of FAM83A-AS1 in NCIH1650 cell and to further study its function (Fig. 4D). Consistently, silencing FAM83A-AS1 inhibited the proliferation, migration and invasion of NCIH1650 cells (Fig. 4E-H). Moreover, knockdown of FAM83A-AS1 also suppressed the proliferation, migration and invasion of A549 cells (Fig. S2). Based on these results, it was concluded that both FAM83A and FAM83A-AS1 played oncogenic roles in LUAD.

Previous studies have shown that numerous antisense lncRNAs exert biological roles via regulating its natural antisense transcript (21–23). Thus, it was hypothesized that FAM83A-AS1 might promote the progression of LUAD by influencing the expression of FAM83A. Using the cBioPortal website, it was reported that the expression of FAM83A-AS1 positively correlated with FAM83A (Spearman score=0.90; Fig. 5A). After knockdown of FAM83A-AS1, the expression of FAM83A was decreased (Fig. 5B). Accumulating evidence has highlighted the important roles of lncRNAs acting as ceRNAs in cancer development (24–26), and FAM83A-AS1 might regulate FAM83A via sponging specific miRNA. Using LncBase and StarBase, miR-495-3p was predicted to interact with both FAM83A-AS1 and FAM83A (Fig. 5C). Consistent with the assumption that FAM83A-AS1 might regulate FAM83A via sponging specific miRNA, treatment of miR-495-3p mimics significantly decreased mRNA levels of FAM83A-AS1 and FAM83A in NCIH1650 cells (P<0.001; Fig. 5D and E). To clarify their potential interactions, luciferase vectors (wild-type and mutant type) were constructed according to the putative binding sites (Fig. 5F and G). The luciferase reporter assay showed that ectopic expression of miR-495-3p decreased the luciferase activity of FAM83A/FAM83A-AS1 wild-type reporters, but not mutant type (Fig. 5H and I). Furthermore, the rescue experiment showed that miR-495-3p counteracted the effect of FAM83A and FAM83A-AS1 in NCIH1650 cells. The influence of FAM83A/FAM83A-AS1-knockdown on proliferation, migration and invasion in LUAD cells could be restored by miR-495-3p inhibitor (Fig. S3). These results indicated that FAM83A-AS1 functioned as a ceRNA through binding miR-495-3p, thereby decreasing the levels of FAM83A mRNA transcripts.