Research ArticleCancer

The RNA-editing enzyme ADAR promotes lung adenocarcinoma migration and invasion by stabilizing FAK

See allHide authors and affiliations

Sci. Signal.  19 Sep 2017:
Vol. 10, Issue 497, eaah3941
DOI: 10.1126/scisignal.aah3941

Metastatic edits to FAK

Drugs that inhibit cell migration pathways, such as that mediated by the kinase FAK, may prevent metastasis and improve long-term survival in cancer patients. Amin et al. found that the RNA-editing enzyme ADAR supports the activity of FAK. In lung adenocarcinoma cells, ADAR bound to and edited FAK mRNA in a manner that improved its stability, thus increasing the abundance of FAK protein and enhancing the migration of these cells. High ADAR expression was a poor prognostic indicator in patients. These findings suggest that inhibiting FAK may be therapeutic in patients with ADAR-positive lung adenocarcinoma.

Abstract

Large-scale, genome-wide studies report that RNA binding proteins are altered in cancers, but it is unclear how these proteins control tumor progression. We found that the RNA-editing protein ADAR (adenosine deaminase acting on double-stranded RNA) acted as a facilitator of lung adenocarcinoma (LUAD) progression through its ability to stabilize transcripts encoding focal adhesion kinase (FAK). In samples from 802 stage I LUAD patients, increased abundance of ADAR at both the mRNA and protein level correlated with tumor recurrence. Knocking down ADAR in LUAD cells suppressed their mesenchymal properties, migration, and invasion in culture. Analysis of gene expression patterns in LUAD cells identified ADAR-associated enrichment of a subset of genes involved in cell migration pathways; among these, FAK is the most notable gene whose expression was increased in the presence of ADAR. Molecular analyses revealed that ADAR posttranscriptionally increased FAK protein abundance by binding to the FAK transcript and editing a specific intronic site that resulted in the increased stabilization of FAK mRNA. Pharmacological inhibition of FAK blocked ADAR-induced invasiveness of LUAD cells, suggesting a potential therapeutic application for LUAD that has a high abundance of ADAR.

INTRODUCTION

Non–small cell lung cancer (NSCLC) is the leading cause of cancer-related deaths in the United States, with a dismal 5-year survival of 15%. The primary cause of death is the development of metastatic disease (1). Lung adenocarcinoma (LUAD), the most common NSCLC histologic subtype, is characterized by specific oncogenes and driver mutations or translocations that initiate and maintain tumorigenesis (2). One of the biggest challenges in LUAD is the identification and characterization of novel aberrant oncogenic drivers and mutations that are responsible for disease progression. This is a crucial step for the continued development of targeted therapies.

Cancer cells create tumor-specific transcripts through dysregulation of posttranscriptional processes such as alternative splicing, 3′ processing, and RNA editing (3). Pan-cancer studies have demonstrated the presence of widespread dysregulated RNA-editing patterns in cancer, of which adenosine-to-inosine (A-to-I) RNA editing plays a major role (46). Adenosine deaminase acting on double-stranded RNA (ADAR) catalyzes the deamination of A-to-I (7, 8), which results in the translational machinery reading inosine as guanosine, therefore effectively creating A-to-G changes in the RNA. There are three ADAR gene family members: ADAR1, ADAR2, and ADAR3. ADAR1 (ADAR) and ADAR2 (ADARB1) are ubiquitously expressed, whereas ADAR3 (ADARB2) is expressed only in the brain (9). The editing activity of ADAR affects gene expression and function by (i) changing codons and, thus, amino acid sequences of proteins; (ii) altering RNA sequences, which can lead to pre-mRNA splice site changes; (iii) altering the seed sequences of microRNAs (miRNAs) targets; and (iv) affecting the stability of the RNA (10, 11). Amplification of ADAR has been associated with poor outcomes in patients with NSCLC (12). However, the mechanism(s) of increased ADAR expression and their downstream effectors in the progression of lung cancer remains unclear.

Focal adhesion kinase (FAK) is overexpressed in solid tumors (13) and correlates with tumor progression (14). FAK is a cytosolic tyrosine kinase that is a crucial regulator of cell migration (15), invasion (16, 17), adhesion (18), and tumor metastasis (13, 14). Given the importance of FAK in tumor progression, pharmacological inhibitors of FAK are currently in phase 1/2 clinical trials (ClinicalTrials.gov).

We hypothesize that ADAR expression is increased in LUAD and that ADAR activates FAK, which facilitates tumor progression in human LUAD. Here, we reported that ADAR promotes migration and invasion of LUAD cells by stabilization of FAK transcript in an RNA editing–dependent manner. Moreover, small-molecule inhibition of FAK activity abrogated ADAR-mediated LUAD cell migration and invasion, suggesting a potential therapeutic strategy in LUAD with high ADAR expression.

RESULTS

High ADAR expression is associated with tumor recurrence in LUAD patients

We analyzed The Cancer Genome Atlas (TCGA) LUAD and squamous carcinoma (SQ) patient cohorts using the cBioPortal for Cancer Genomics (19). This revealed that ADAR is significantly amplified and overexpressed in LUAD compared with SQ (ADAR DNA copy number amplification, 14.3% for LUAD versus 1.7% for SQ; ADAR mRNA overexpression, 23% for LUAD versus 8.4% for SQ) (fig. S1). We next examined ADAR copy number and mRNA expression in LUAD cells and normal human bronchial epithelial cells (HBECs) by droplet digital polymerase chain reaction (PCR) and quantitative reverse-transcription PCR (qRT-PCR), respectively. Consistent with observations from the TCGA cohort, ADAR was amplified and overexpressed in most tested LUAD cells compared with HBECs (Fig. 1, A and B). Moreover, the abundance of ADAR protein was also substantially greater in all tested LUAD cells than in HBECs (Fig. 1C).

Fig. 1 ADAR is overexpressed in LUAD and correlates with tumor recurrence.

(A) ADAR DNA copy numbers were determined by droplet digital PCR in HBECs and the indicated LUAD cells. Data are in triplicate from three experiments. (B) ADAR mRNA expression in HBEC and the indicated LUAD cells were assessed by qRT-PCR. HPRT was amplified as a reference. Data are means ± SEM and in triplicate from three experiments. (C) Western blot of ADAR protein expression in HBECs and LUAD cells. n = 3 experiments. (D) Kaplan-Meier curve of progression-free survival based on ADAR mRNA expression in 162 stage I LUAD patients in the NCCRI cohort (log-rank test, P < 0.0001). (E) Immunohistochemical analysis showing low and high ADAR expression in two representative stage I LUAD tumors. Scale bars, 100 μm (top) and 50 μm (bottom). (F) Cumulative incidence of recurrence based on ADAR protein expression in 802 patients with stage I LUAD (Gray’s test, P = 0.016). GAPDH, glyceraldehyde-phosphate dehydrogenase.

To assess the clinical relevance of increased ADAR mRNA expression in LUAD specimens, we performed an unbiased analysis using a publicly available gene expression microarray data set including 162 patients with stage I LUAD [National Cancer Center Research Institute (NCCRI) cohort (www.abren.net/PrognoScan/)] (20). Patients with high ADAR mRNA expression had decreased progression-free survival (Fig. 1D). To confirm that ADAR overexpression correlates with the progression of LUAD in a larger cohort of patients with stage I LUAD, we examined ADAR expression in Memorial Sloan Kettering Cancer Center (MSK) LUAD tissue microarray of stage I LUAD specimens. Immunostaining showed that ADAR was primarily located in the nucleus (Fig. 1E). As expected, high intratumoral ADAR expression was associated with higher cumulative incidence of tumor recurrence (Fig. 1F). Collectively, our data show that high ADAR expression correlates with tumor recurrence and poor prognosis in patients with early-stage LUAD.

ADAR is involved in LUAD cell migration and invasion

To examine the functional mechanisms of ADAR in LUAD cells, we stably knocked down ADAR abundance using two short hairpin–mediated RNAs (shRNAs) targeted to different regions in ADAR coding sequences (ADAR A and ADAR B) in endogenously ADAR-amplified H1975 and H358 cells. The efficiency of ADAR knockdown (KD) was confirmed by qRT-PCR and Western blot (Fig. 2, A and B), further validated by decreased expression of two known downstream targets of ADAR, F11R and CCND1, in ADAR KD cells compared to scramble control cells (fig. S2, A and B). To confirm that deaminase activity was also reduced in ADAR KD cells, we performed RNA-editing site-specific quantitative PCR (RESSq-PCR) (21) to quantify ADAR-edited AZIN1 transcripts (5). Compared to scramble control cells, ADAR KD cells had decreased abundance of edited AZIN1 transcripts (fig. S2C).

Fig. 2 ADAR KD inhibits cell migration and invasion.

(A) qRT-PCR for ADAR mRNA relative to 18S in H358 and H1975 scramble and ADAR KD cells. *P < 0.05 and **P < 0.01 compared with scramble (Mann-Whitney test). Data are means ± SEM from three independent experiments. (B) Western blot of ADAR expression in H358 and H1975 scramble and ADAR KD cells. n = 3 independent experiments. (C) Immunofluorescent staining with phalloidin (red, F-actin) in H358 and H1975 scramble and ADAR KD cells. Scale bar, 25 μm. n = 3 independent experiments. (D) Cell migration of H358 and H1975 scramble control and ADAR KD cells. *P < 0.05 compared with scramble (Mann-Whitney test). Data are means ± SEM from three independent experiments. Scale bars, 100 μm. (E) Cell invasion of H358 and H1975 scramble and ADAR KD cells. **P < 0.01 compared with scramble (Mann-Whitney test). Data are means ± SEM from three experiments. Scale bars, 100 μm. (F) Soft agar colony formation of H358 and H1975 scramble and ADAR KD cells. *P < 0.05 and **P < 0.01 compared with scramble (Mann-Whitney test). Data are means ± SEM from three independent experiments. Scale bars, 200 μm.

ADAR KD induced a change in LUAD cell morphology. ADAR KD cells exhibited a cobblestone, epithelioid morphological appearance, whereas scramble control cells had an elongated, mesenchymal shape (Fig. 2C). Immunofluorescence with F-actin staining revealed reduced stress fiber formation and a more boundary-like staining pattern in ADAR KD cells compared to scramble control cells (Fig. 2C). Furthermore, ADAR KD significantly inhibited cell migration (Fig. 2D) and invasion (Fig. 2E) in LUAD cells compared to scramble control cells. To confirm that the ADAR KD–mediated inhibition of cell migration and invasion was not secondary to decreased cell viability, we performed cell proliferation assays. Consistent with previous studies (12), loss of ADAR does not affect cell growth within 3 days of plating (fig. S2D). However, scramble control cells exhibited increased growth rates over 5 and 7 days compared to ADAR KD cells (fig. S2D). Cell doubling time assays revealed that, at the low cell density, there was no difference of growth rate between control (scramble-transfected) and ADAR KD cells (fig. S2E). In contrast, increasing cell density robustly promoted the growth of control cells relative to ADAR KD cells (fig. S2E). Collectively, these data suggest that cell-cell contact affects the growth of cells with abundant ADAR.

The ability to grow in an anchorage-independent manner is a hallmark of cancer. ADAR KD leads to a reduction in size and number of colonies in soft agar assays compared to those of control cells (Fig. 2F). Collectively, our data show that the presence of ADAR correlates with an altered cellular epithelial phenotype, which is associated with enhanced LUAD cell migration and invasion.

ADAR affects the FAK signaling pathway

To investigate the putative pathway(s) through which ADAR promotes cell migration and invasion in LUAD, we performed Illumina microarray analysis using H358 ADAR KD and scramble-transfected (control) cells. We identified 2207 genes that were differentially expressed between ADAR KD and control cells with a fold change of −32 to +243 (adjusted P < 0.01). To functionally annotate our microarray data and investigate relevant pathways, we analyzed the top 20% most significant genes using the Ingenuity software. Cellular movement emerged as one of the pathways most affected by the loss of ADAR. Differential expression of genes between ADAR KD cells and control cells revealed FAK to be the most significantly differentially expressed and down-regulated by the loss of ADAR (Fig. 3A). FAK promotes the reorganization of cytoskeletal components and cell motility (22). Hence, we hypothesized that FAK plays a role in ADAR-mediated cell migration and invasion.

Fig. 3 ADAR expression increases FAK expression.

(A) Heat map of genes related to cellular movement differentially expressed in H358 scramble and ADAR KD cells (P = 0.0009). n = 2 biological replicates. (B) qRT-PCR of FAK mRNA relative to 18S in the indicated cells. *P < 0.05, **P < 0.01, and ***P < 0.001 compared with scramble (Mann-Whitney test). Data are means ± SEM from three independent experiments. (C) Western blot of FAK in the indicated cells. n = 3 independent experiments. (D) Flag-tagged ADAR was stably transfected into H1975 ADAR KD cells. Western blot for the indicated proteins. n = 3 independent experiments. (E) Correlation of ADAR and FAK mRNA in advanced-stage LUAD (TCGA cohort; N = 57; P = 0.031). (F) Western blot of the indicated proteins in selected patient LUAD samples. n = 3 independent experiments. (G) Immunofluorescence staining cortactin in the indicated cells. Scale bar, 25 μm. n = 3 independent experiments. (H) Phospho-cortactin in the indicated cells was detected by coimmunoprecipitation using an antibody against pan-phospho-tyrosine followed by immunoblotting with an antibody against cortactin. n = 3 independent experiments. IP, immunoprecipitation; IgG, immunoglobulin G. (I) Western blot for phospho-paxillin Tyr118 and paxillin in the indicated cells. n = 3 independent experiments.

To experimentally address our hypothesis, we validated FAK expression in ADAR KD and scramble cells by qRT-PCR and Western blot. Consistent with our microarray data, FAK mRNA and protein were both greatly reduced in ADAR KD cells compared to scramble control cells (Fig. 3, B and C). Moreover, reconstituting ADAR KD H1975 cells with ectopic Flag-tagged ADAR rescued FAK protein abundance (Fig. 3D), proving the specificity of ADAR on regulation of FAK. To investigate whether ADAR increases FAK mRNA in human LUAD, we analyzed RNA sequencing (RNA-seq) data in the TCGA LUAD cohort (cBioPortal.org). ADAR positively correlates with the mRNA expression of FAK in patients with advanced stage LUAD (stages III and IV) (Spearman r = 0.29; Fig. 3E). In addition, protein analysis of selected human LUAD tumors showed that tumors with a high amount of ADAR also have a high amount of FAK (Fig. 3F). These findings further strengthen our hypothesis that ADAR increases FAK abundance.

To determine whether loss of ADAR also affects FAK signaling, we conducted immunofluorescence analysis for cortactin. Cortactin is a known downstream target of the FAK signaling pathway (23), and its localization at the peripheral edges of cells is a marker for actin-rich motility protrusions such as lamellipodia and invadopodia (24). Moreover, phosphorylation of the tyrosine residues in cortactin by FAK (25) increases F-actin turnover and cell motility (26). We found that cortactin was distributed throughout the cytoplasm in ADAR KD cells, whereas in scramble cells, it accumulated in the periphery (Fig. 3G). To explore whether ADAR KD leads to a reduction of phosphocortactin, we simultaneously performed an immunoprecipitation assay using an antibody against phosphotyrosine to pull down all phosphotyrosine proteins. After detection with cortactin, we observed that ADAR KD decreased phosphocortactin but not total cortactin expression (Fig. 3H). Another downstream target of FAK is the focal adhesion protein paxillin, which is phosphorylated by FAK on tyrosine 118 (Tyr118) (27). To further investigate the consequences of ADAR KD–induced decrease in FAK, we examined paxillin and phospho-paxillin. Knockdown of ADAR did not change total paxillin expression but reduced phospho-paxillin (Fig. 3I). Collectively, these data confirm that ADAR increases FAK expression and activity and reveal an association between ADAR-mediated cell migration and invasion and FAK signaling.

ADAR stabilizes FAK transcript in an RNA editing–dependent manner

Alteration of RNA stability is one of the mechanisms through which ADAR regulates gene expression (28). To determine whether ADAR increases FAK transcript through modulation of FAK RNA stability, we performed actinomycin D chase experiments in H358 and H1975 ADAR KD and scramble cells. The percentages of remaining FAK mRNA were statistically lower in both H1975 and H358 ADAR KD cells compared with their scramble cells (Fig. 4A). For both LUAD cells, FAK mRNA had a shorter half-life in ADAR KD cells than in scramble cells (Fig. 4A).

Fig. 4 ADAR stabilizes FAK transcript.

(A) Percentage of remaining FAK mRNA in the indicated cells after treatment with actinomycin D. **P < 0.01 compared with scramble (Wilcoxon rank sum test of area under the curve). Data are means ± SEM from three independent experiments. (B) Diagram of ADAR protein. Blue, Zα and Zβ domains; purple, dsRBDs; red, deaminase domain. (C) qRT-PCR for FAK mRNA relative to 18S in HCC827 and H1299 cells transfected with ADAR wild type (WT), mutants, or empty vector as control. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 compared with controls (Mann-Whitney test). Data are means ± SEM from three independent experiments. (D) Western blot for the indicated proteins in the indicated HCC827 and H1299 cells. n = 3 independent experiments. (E) Phosphocortactin in the indicated HCC827 and H1299 cells was detected by coimmunoprecipitation using an antibody against pan-phospho-tyrosine, followed by immunoblotting with cortactin. n = 3 independent experiments. (F) Percentage of remaining FAK mRNA in the indicated HCC827 and H1299 cells after treatment with actinomycin D. **P < 0.01 compared with control. ##P < 0.01 compared with the deaminase mutant (mut.) (Wilcoxon rank sum test of area under the curve). Data are means ± SEM from three independent experiments.

Whereas ADAR regulates RNA stability in either an RNA editing–dependent or an RNA editing–independent manner, RNA binding is essential for its RNA-editing activity (28). ADAR contains two Z-DNA binding domains (blue), three double-stranded RNA binding domains (dsRBDs; purple), and the deaminase region (red) (Fig. 4B). To discern whether the ADAR-induced FAK stabilization depends on its deaminase activity, we created an ADAR RNA binding–deficient expression construct (dsRBD mutant) by mutating three lysine residues of the KKXXK motif to EAXXA in all three dsRBD regions and a deaminase mutant through point mutation (E912A) (29, 30). After stable expression of Flag-tagged ADAR in LUAD HCC827 and H1299 cells with low endogenous ADAR (Fig. 1), we observed an increase in FAK transcript and protein compared to vector control cells (Fig. 4, C and D). However, stable expression of either ADAR dsRBD mutant or the deaminase mutant failed to induce FAK expression (Fig. 4, C and D). In addition, phosphorylation of paxillin (Fig. 4D) and cortactin (Fig. 4E) was increased only in cells with wild-type ADAR overexpression but not in ADAR mutant–expressing cells (Fig. 4, D and E). To confirm that the deaminase function is required for ADAR-induced stabilization of FAK mRNA, we performed actinomycin D chase experiments using HCC827 and H1299 stable cells described above. Overexpression of ADAR significantly increased the percentages of remaining FAK mRNA and half-lives of FAK mRNA in both tested cell lines compared to cells with either control or the deaminase mutant (Fig. 4F). Overexpression of ADAR deaminase mutant failed to alter the stability of FAK mRNA compared to control (Fig. 4F). These findings confirm that the deaminase activity of ADAR is critical for FAK transcript stabilization.

ADAR directly binds to and edits FAK RNA

To explore the functional ADAR-editing site(s) in FAK, we analyzed FAK RNA-seq data in the TCGA LUAD cohort using the Integrated Genomics Viewer tool. We identified three putative functional ADAR-editing sites in FAK intronic regions by correlating with FAK mRNA expression (fig. S3A). Tumors with editing of each of the putative sites have a significantly higher FAK mRNA expression compared to unedited tumors (Fig. 5A and fig. S3, B and C). To examine whether ADAR directly edits these sites in LUAD cells, we individually detected site-specific editing of these three sites in HCC827 and H1299 cells that stably express ADAR wild type or mutants. Ectopic expression of wild-type ADAR significantly increased the editing FAK only on site chr8:141,702,274 (Fig. 5B and fig. S3D) but not on the other two sites (fig. S3, E and F). Neither ADAR dsRBD mutant nor deaminase mutant affected the editing FAK on all three sites (Fig. 5B and fig. S3, D to F). These data indicate that site chr8:141,702,274 in the intron 26 of FAK is a functional ADAR-editing site in both LUAD patient specimens and cell lines, and the editing level of this site correlates with the expression of FAK mRNA in human LUAD.

Fig. 5 ADAR stabilizes FAK through RNA binding and editing.

(A) FAK mRNA in tumors with (edited; n = 41) or without (nonedited; n = 189) A-to-I editing in the TCGA cohort. ***P < 0.001 compared to nonedited tumors (Mann-Whitney test). (B) Chromatograms of FAK transcripts in the indicated cells. Arrow, the site chr8:141,702,274. The percentage of edited FAK detected by Sanger sequencing. n = 3 biological replicates. (C) RIP analyses for region A in H1975 parental cells. Data are means ± SEM from three independent experiments. Ab, antibody. (D) RIP analyses for region A on the indicated H1299 stable cells. *P < 0.05 comparing groups (Mann-Whitney test). Data are means ± SEM from three independent experiments. (E) RNA-protein interaction of in vitro double-stranded RNA and ADARs. n = 3 independent experiments. (F) Percentage of remaining pcDNA-FAK mRNA in indicated H1975 cells after actinomycin D treatment. *P < 0.05 compared with FAKE+I WT in scramble (blue); #P < 0.05 compared with FAKE+I edited in ADAR KD (purple); and φP < 0.05 compared with FAKE+I edited in scramble (red) (Wilcoxon rank sum test of area under the curve). Data are means ± SEM from three independent experiments. (G) Schematic ADAR binding and editing FAK in the intron 26. Red lines, regions A and B used for RIP assays; arrow, the editing site.

Next, to determine whether ADAR physically binds to FAK RNA endogenously, we performed RNA immunoprecipitation (RIP) experiments using H1975 cells, which have high endogenous ADAR expression (Fig. 1). Two regions (A, chr8:141,702,900 to 141,703,155; B, chr8:141,702,218 to 141,702,681) around the functional ADAR-editing site chr8:141,702,274 in the intron 26 of FAK were amplified. Endogenous ADAR was found to bind FAK at region A (Fig. 5C) but not at region B (fig. S4). To determine whether the functional dsRBDs are crucial for ADAR-FAK RNA binding, we performed RIP assays using H1299 cells that express the ADAR wild type, the dsRBDs mutant, or the deaminase mutant. To control for the effect of FAK transcript amount on the enrichment of ADAR-RNA binding, we normalized the enrichment of ADAR-RNA binding by FAK transcript in each corresponding cell line. There was an increased ADAR binding to region A of FAK in cells expressing the ADAR wild type and the deaminase mutant but not in those expressing the dsRBD mutant (Fig. 5D). To determine whether ADAR binds FAK RNA directly, we performed in vitro RNA-protein interaction assay using in vitro–translated ADAR or its mutants and in vitro–transcribed double-stranded FAK RNA region A. Both ADAR and the deaminase mutant, but not the dsRBD binding mutant, of ADAR were able to bind the double-stranded FAK RNA (Fig. 5E). Collectively, these studies demonstrate that ADAR directly binds to FAK intronic region A endogenously through its functional dsRBDs.

To investigate whether ADAR-induced editing on site chr8:141,702,274 results in stabilization of FAK, we generated plasmids encoding FAK intronic regions A and B and the adjacent exon (FAKE+I wild type), and a T>C mutation was generated in plasmid FAK DNA on site chr8:141,702,274 (FAKE+I edited; fig. S5). We transfected either FAKE+I wild type or FAKE+I edited into H1975 ADAR KD and scramble cells and performed actinomycin D chase experiments. Consistent with our previous findings, the mRNA of FAKE+I wild type had reduced stability in ADAR KD cells compared to in scramble cells (Fig. 5F). The mRNA of FAKE+I edited was significantly more stable than FAKE+I wild type in scramble cells and ADAR KD cells, and ADAR KD failed to affect the mRNA stability of FAKE+I edited (Fig. 5F). Collectively, our data suggest that ADAR binds to region A in intron 26 of FAK and edits FAK on an intronic site chr8:141,702,274 (Fig. 5G), resulting in the stabilization and an increase of FAK mRNA.

ADAR-induced cell migration and invasion is FAK-dependent

To assess whether ADAR KD–induced repression of invasion is specifically mediated by the loss of FAK, we stably transfected Flag-tagged FAK in H358 and H1975 ADAR KD cells. Ectopic FAK partially rescued the invasion capability of ADAR KD cells in both LUAD cell lines (Fig. 6A). To confirm that ectopic FAK reactivates downstream FAK signaling in ADAR KD cells, we evaluated the activity of the FAK downstream signaling target paxillin. Ectopic FAK rescued phospho-paxillin in ADAR KD cells without affecting paxillin (Fig. 6B). Together, these findings suggest that FAK is a key mediator involved in ADAR-induced migration and invasion in LUAD.

Fig. 6 ADAR-induced cell migration and invasion are FAK-dependent.

(A) Invasion assays of H358 and H1975 scramble, ADAR KD, and ADAR KD with Flag-FAK cells. #P < 0.05 compared with corresponding ADAR KD cells. *P < 0.05 and **P < 0.01 compared with scramble (Mann-Whitney test). Data are means ± SEM from three independent experiments. Scale bar, 100 μm. (B) Western blot for the indicated proteins in H358 and H1975 scramble control, ADAR KD, and ADAR KD with Flag-FAK cells. n = 3 independent experiments. (C) Coimmunoprecipitation using antibody against pan-phospho-tyrosine in HCC827 and H1299 control or ADAR-expressing cells treated with PND1186 or VS-6063 (2.5 μM) or dimethyl sulfoxide (DMSO) for 72 hours. Western blot for phospho-FAK and phospho-paxillin with antibody against FAK or paxillin. n = 3 independent experiments. (D) Invasion assays of HCC827 and H1299 control and ADAR-expressing cells treated with FAK inhibitors PND1186 or VS-6063 (2.5 μM) or DMSO for 72 hours. #P < 0.05 compared with DMSO-treated control cells. *P < 0.05 and **P < 0.01 compared with DMSO with the same vector (Mann-Whitney test). Data are means ± SEM from three independent experiments.

We next sought to determine whether pharmacologic inhibition of FAK activity is able to affect ADAR-induced cell invasion. PND1186 and VS-6063 are specific FAK inhibitors that are currently in clinical trials for the treatment of solid tumors (ClinicalTrials.gov). Cell viability assays were conducted to establish the optimal doses of PND1186 and VS-6063. A concentration of 2.5 μM was chosen because it did not cause excessive cell death (fig. S6, A and B) while still inhibiting the phosphorylation of FAK and its downstream target paxillin (Fig. 6C). Consistent with our previous findings, stable ectopic expression of ADAR resulted in increased invasion of both HCC827 and H1299 cells (Fig. 6D). Treatment with either pharmacological FAK inhibitor reduced invasion in HCC827 and H1299 control cells. Moreover, treatment with either FAK inhibitor completely abrogated ectopic expression of ADAR-induced increases of cell invasion in both tested LUAD cells (Fig. 6D). These findings support that the pharmacological inhibition of FAK activity is able to block ADAR-induced increases in cell invasion.

DISCUSSION

Advances in next-generation sequencing technology have led to the discovery of specific A-to-I RNA-editing events across different cancers, and LUAD is a cancer that is hyperedited (5). The “net” proportion of these overediting events significantly correlates with the expression of ADAR1 (ADAR) but not ADAR2 or ADAR3 (5). ADAR is located at chromosome 1q21, and this region is highly amplified in LUAD (31). It has been proposed that ADAR amplification is associated with NSCLC recurrence (12). Extending those studies, we now show that ADAR is amplified and overexpressed in multiple lung cancer cells and human lung cancer samples and that ADAR overexpression correlates with tumor recurrence and worse progression-free survival in early-stage LUAD. We also demonstrate that ADAR promotes cell migration and invasion in LUAD through stabilization of the FAK transcript in an RNA editing–dependent manner.

The precise role that ADAR-mediated RNA editing plays in the pathogenesis of cancer is increasingly being investigated. Reduced ADAR expression contributes to tumor growth and metastasis in brain tumors (32) and in metastatic melanoma through hypoediting of Alu repetitive elements or miRNA-455-5p (33). However, several tumor-specific A-to-I RNA-editing events in the coding regions in esophageal squamous cell carcinoma, such as AZIN1S367G and COPAI64V (34), are associated with tumor progression and are related to increased ADAR expression. We found that ADAR KD significantly inhibited cell migration and invasion in our LUAD cells. These functional discrepancies for ADAR in cancer might be related to tumor and/or target gene specificity. An alternative hypothesis is that the function of ADAR is dependent on tumor progression. Under certain pressures (such as cancer therapies) and natural selection, cancer cells dynamically alter their genetic architecture to contribute to tumor progression (35, 36). Here, we observed that increased ADAR expression is associated with increased recurrence of LUAD and that reduction of ADAR by shRNA inhibits the mesenchymal formation of tumor cells and maintains an epithelial phenotype. Our data suggest that high ADAR expression in LUAD drives metastases in situ through induction of a mesenchymal phenotype. Therefore, in early-stage LUAD, amplification and overexpression of ADAR appear to be important oncogenic events that contribute to LUAD progression.

This study begins to unravel the mechanism(s) whereby ADAR enhances invasion and migration of LUAD cells, namely, through the stabilization of FAK. FAK overexpression is observed in more than 20% of solid tumors, including ovarian, breast, colorectal, and lung cancers, and is associated with poor prognosis (13). It is well-documented that FAK transcriptional activity is increased in cancer cells by nuclear factor κB, Nanog, and Ago2 directly binding to the FAK promoter (3739). Here, we have identified posttranscriptional ADAR-induced RNA stabilization as a mechanism through which FAK is increased in cancer cells. ADAR edits FAK RNA and increases its stability, resulting in an increase of FAK expression and activation of FAK signaling.

Observed patterns of dysregulated A-to-I RNA editing in cancers (46) have shifted the focus in understanding how ADAR functions in cancer cells. For instance, ADAR editing of AZIN1 enhances hepatocellular carcinoma tumorigenesis and progression (34), and reduction of ADAR leads to regression of chronic myelogenous leukemia and pediatric astrocytomas (40, 41). Moreover, ADAR-catalyzed A-to-G changes are the most commonly identified RNA-DNA differences in lung cancer (42).

ADAR modulates gene expression in either an RNA editing–dependent or an RNA editing–independent manner. Reports suggest that ADAR interacts with Dicer, a component of the RNA-induced silencing complex, or DCGR8 to mediate pre-miRNA processing, eventually leading to RNA destabilization (43, 44). In addition, ADAR stabilizes RNA through interaction with an RNA binding protein in B cells (HuR) (28). Here, we have shown that ADAR stabilizes FAK RNA in an RNA editing–dependent manner. FAK RNA contains more than a thousand ADAR-editing sites, with most of these located in Alu repetitive elements. The editing of Alu sequences has several implications, such as “exonization” of intronic Alu sequences, retention of Alu sequences in the paraspeckles, suppression of interferon response, and heterochromatin formation, which has been reviewed (45). Here, using the TCGA cohort and our experimental LUAD cell line model system, we identified an intronic site on chr8: 141,702,274 in FAK, which is a functional ADAR-editing site that is directly related to FAK RNA stability. Given the differences in signaling pathways, genomic profiles, and tumor microenvironment between cultured cells and tumors (46, 47), there are likely to be other functional ADAR-editing sites in FAK. Overall, the mechanism(s) of ADAR editing on gene expression is evolving, and more work is necessary to precisely determine how ADAR-induced intronic editing affects gene expression (48).

PND1186 and VS-6063 are orally bioavailable adenosine 5′-triphosphate–competitive small molecules that block FAK phosphorylation (49). PND1186 is currently in phase 1 clinical trials for nonhematologic tumors, whereas VS-6063 is in phase 2 clinical trials for KRAS-mutant NSCLC (ClinicalTrials.gov) (50). Recently, the PF562271 FAK inhibitor was used in a preclinical high-grade mutant KRAS;INK4A/ARF–deficient LUAD mouse model and selectively promoted cell death by shutting down the dysregulated ERK (extracellular signal–regulated kinase)/RHOA (Ras homolog gene family, member A)/FAK pathway (51). Here, we show that inhibition of FAK activation by either PND1186 or VS-6063 abrogated ADAR-induced increases in invasion of LUAD HCC827 and H1299 cells, both of which are KRAS and INK4A/ARF wild type. However, p-FAKY397 is also strongly or moderately expressed in 30% of KRASWT LUAD specimens (51), suggesting that alternative mechanisms to increase FAK activity (for example, ADAR overexpression, which occurs in ~30% of LUAD) may allow these tumors to be treated with FAK inhibitors. Other potential discrepancies between our work and previous studies include the use of different FAK inhibitors, different cells, and the biological function assays performed.

In addition to demonstrating that FAK plays an important role in mediating ADAR-induced increases in LUAD cell invasion, we found that the addition of FAK to ADAR KD cells was unable to completely rescue the invasion potential of the cells. Given that ADAR is involved in many cellular processes and FAK is only 1 of more than 27 genes related to cellular movement that are differentially expressed when ADAR is knocked down, it is likely that stabilization of FAK is not the only mechanism through which ADAR induces cell invasion. For example, RAC2 and RAC3 gene expression are also differentially expressed when ADAR is knocked down. Like FAK, these proteins are involved in cytoskeleton reorganization through promotion of actin assembly, resulting in lamellipodia and membrane ruffle formation (52).

In summary, we have highlighted the importance of ADAR amplification and overexpression in the pathogenesis of LUAD. By identifying FAK as a novel ADAR-editing target, our work establishes a potential therapeutic strategy of targeting FAK to prevent metastasis in early-stage LUAD with high ADAR expression.

MATERIALS AND METHODS

Cell culture, antibodies, and reagents

All cell lines used were obtained from American Type Culture Collection and tested for mycoplasma. Human LUAD cells were grown in RPMI 1640 supplemented with 10% fetal bovine serum (Invitrogen). Normal HBECs were grown in keratinocyte-SFM (Life Technologies) containing bovine pituitary extract (50 μg/ml) and epidermal growth factor (5 ng/ml). Human embryonic kidney (HEK) 293 cells were maintained in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum. The primary antibodies used were ADAR1 (ab88574, Abcam), cortactin (MAB6096, R&D Systems), GAPDH (MAB374, EMD Millipore), FAK (sc-558, Santa Cruz Biotechnology), phosphotyrosine antibody (PY20, #03-7700, Thermo Fisher Scientific), FLAG-epitope (PA1-984B, Thermo Fisher Scientific), paxillin (AB32084, Abcam), and phospho-paxillin Tyr118 (MAB6164, R&D System). Collagen type IV (9007-34-5) and actinomycin D (50-76-0) were purchased from Sigma-Aldrich. Puromycin (A1113803) and Geneticin (10131027) were purchased from Life Technologies.

Droplet digital PCR

Genomic DNA was extracted from LUAD cells using the Genomic DNA Purification kit (Thermo Fisher Scientific) in accordance with the manufacturer’s instructions. Droplet digital PCR was performed using the QX200 system (Bio-Rad Laboratories) to determine ADAR copy number variation in LUAD cells; AP3B1 was used as the reference gene. Results were analyzed using the QuantaSoft software (version 1.7, Bio-Rad Laboratories).

Analysis of existing microarray data

Microarray data, available on PrognoScan, from a study of progression-free survival in patients with stage I and II LUAD (NCCRI cohort; N = 204) (20). The array was performed on the Affymetrix platform with a probe (ID 201786_s_at). Progression-free survival was assessed using the Kaplan-Meier method and compared between high and low ADAR mRNA groups using the log-rank test. A cutoff of <8.8575 for low versus high ADAR mRNA was determined by maximally selected rank statistics (53) using the package MaxStat in R 3.1.1 (R Core Development Team).

Clinical specimens and immunohistochemical analysis

A clinically annotated tissue microarray containing 802 stage I LUAD specimens was created by the MSK Thoracic Surgery Service (54). Human LUAD specimens were obtained with written, informed consent and approval of the Human Investigations Committee at MSK. Immunohistochemical analysis was performed as previously described (55). The ADAR antibody was used at a dilution of 1:200. ADAR staining was scored by a pathologist on the basis of the average staining intensity of three tumor cores (0, no staining; 1, weak; 2, moderate; 3, strong), as previously described (56, 57). Cumulative incidence of recurrence (CIR) was defined as the time between surgery and locoregional recurrence or distant metastasis. A cutoff of ≤0.63 for low versus high ADAR immunoreactivity score was determined by maximally selected rank statistics (53) using the package MaxStat in R 3.1.1. The difference in CIR between ADAR low and high was determined by Gray’s test.

Plasmids and shRNA construction

ADAR pLKO.1 shRNA constructs TRN0000050788 (ADAR A) and TRNC0000050789 (ADAR B) were purchased from Sigma-Aldrich. The lentiviral plasmid encoding Flag-tagged ADAR (EX-Z3143-Lv101) was purchased from GeneCopoeia. To generate the ADAR RNA binding–deficient (dsRBD) mutant and the deaminase mutant (E912A), we used the site-directed mutagenesis kit (E0554S, New England Biolabs) in accordance with the manufacturer’s instructions. FAK intronic region chr8:141,702,243-chr8:141,703,155 (the intron 26) and the exon 27 was amplified from HBEC complementary DNA and cloned into pcDNA3.1+ (FAKE+I wild type, GenScript), and a T>C mutation at site chr8:141,702,274 (FAKE+I edited) was generated by site-directed mutagenesis. The primer pairs used for mutations are listed in table S1. pWZL-Neo-Myr-Flag-PTK2 was a gift from W. Hahn and J. Zhao (Addgene plasmid #20610).

Virus production and infection

To produce virus, 10 μg of lentiviral (pReceiver or pLKO) or retroviral (pWZL-Neo Myr-Flag-PTK2) plasmid and 3 μg each of the packaging plasmid DNA (psPAX2 and pMD2.G for lentivirus; VSVG and Gag-Pol for retrovirus) were cotransfected into HEK293 cells by using the PolyFect reagent (301107, Qiagen). NSCLC cell lines H358 and H1975 were plated into 100-mm culture dishes at 60% confluence the day before infection. The following day, 3 ml of lentivirus- or retrovirus-containing media was added to the cells, together with polybrene (8 μg/ml; sc-134220, Santa Cruz Biotechnology), for infection. For pLKO lentivirus infection, cells were selected by using puromycin (2 μg/ml) for 2 weeks and screened by qRT-PCR and Western blot for ADAR expression. To select cell lines that overexpress Flag-PTK2 or Flag-ADAR, Geneticin (600 μg/ml) was used for 2 weeks, and the expression of Flag-tagged protein was screened using Western blot.

Total RNA isolation, qRT-PCR, and Sanger sequencing analyses of editing of FAK

Total RNA was isolated using TRIzol reagent in accordance with the manufacturer’s protocol (Invitrogen). To detect edited transcripts, we used the RNA-editing site-specific primer design strategy that is compatible with SYBR green qRT-PCR protocols (RESSq-PCR), as previously described (21).

For Sanger sequencing analyses, semiquantitative PCRs were performed using the pair of primers FAK274C forward and FAK274T reverse. The PCR products were purified using QIAquick PCR Purification kit (28106, Qiagen) and sent for Sanger sequencing using the primer FAK274T reverse at Genewiz. The primers used are listed in table S1.

Microarray analysis

Total RNA was extracted from H358 shRNA scramble and shRNA ADAR B cells by using the RNeasy kit (Qiagen) in accordance with the manufacturer’s instructions. Human HT-12 v4 Expression BeadChip (Illumina) was performed in duplicate by the Integrated Genomics Operation Core Facility at MSKCC. The raw data were extracted and analyzed using Partek Genomics Suite 6.6. Gene signal values were logarithm-transformed and normalized using the quartile method (58). Array data were filtered using an adjusted P < 0.01. Comparative analysis between shRNA scramble and shRNA ADAR B cells was performed on the basis of fold change in expression. Gene ontologic analysis was performed using the Ingenuity Pathway Analysis software.

Western blot and coimmunoprecipitation

Western blotting was conducted as previously described (59). Primary antibodies were used at a dilution of 1:1000, and secondary antibodies (P/N 925-32210, P/N 925-32211, LI-COR Biosciences) were used at a dilution of 1:20,000.

Blots were scanned using the Odyssey CLx machine (LI-COR Biosciences). For immunoprecipitation, 500 μg of cell lysates were incubated with 1 μg of antibody against phosphotyrosine (PY20, Thermo Fisher Scientific) or control rabbit IgG. The presence of phosphocortactin was detected by immunoblot.

Immunofluorescence

Cells (2.5 × 104/100 ml) were plated in four-well chamber slides, grown for 60 hours, and fixed in 4% paraformaldehyde. After incubation with 1% bovine serum albumin in phosphate-buffered saline with Tween 20 for 30 min, rhodamine phalloidin (1:40; R415, Life Technologies) was added for F-actin detection for 30 min. For cortactin and FAK detection, antibodies were used at a dilution of 1:100 at 4°C overnight. Secondary antibodies anti-rabbit Alexa Fluor 594 (Z25307) dye or anti-mouse Alexa Fluor 488 (Z25002) were used at a dilution of 1:2000 at room temperature for 1 hour (Invitrogen). The slides were mounted with coverslips using UltraCruz Mounting Medium (Santa Cruz Biotechnology). Immunofluorescence images were taken on the Leica TCS SP5 II (upright stand) microscope with 40× oil objective lens (numerical aperture, 1.25).

Soft agar colony formation assays

Soft agar assay for anchorage-independent growth was performed using 0.3% agarose with 5 × 103 suspended cells, which was plated on the top of a layer of 0.8% agar in a six-well plate. Three weeks after initial plating, colonies were fixed, stained with 0.1% crystal violet, and counted.

Migration and invasion assays

Migration assays were performed using Boyden chambers (353097, Corning), with 2.5 × 104 cells plated in each chamber. Cells were incubated for 24 hours, fixed, and stained with 0.1% crystal violet. Invasion assays were performed by precoating the Boyden chambers with type IV collagen (0.25 mg/ml) in 0.25% acetic acid overnight. After incubation for 48 hours, cells were fixed and stained with 0.1% crystal violet. Images were taken using the Olympus 1X71 microscope, and cells were counted using the ImageJ software.

Cell viability assay and doubling time detection

Cells were seeded into 96-well plates at 2.5 × 103 cells per well. At days 0, 3, 5, and 7, cell viability was determined using the CellTiter-Glo Luminescent Cell Viability Assay kit (G7570, Promega) in accordance with the manufacturer’s instructions.

Cells were seeded into six-well plates at a density of 5 × 103, 5 × 104, and 1 × 105 cells per well. After plating for 48 hours, cells were detached, and the Trypan blue cell exclusion assay was used to calculate the number of cells. A countless automated cell counter (AMQAF1000, Thermo Fisher Scientific) was used for cell counting. The doubling time was calculated using an online doubling time calculator (www.doubling-time.com/compute.php).

RNA stability assay

Cells were plated 24 hours before treatment with actinomycin D (5 μg/ml). Treated cells were collected after treatment at 0, 2, 4, 8, and 16 hours, and RNA was extracted. qRT-PCR was performed to assess RNA expression. Primers used to detect endogenous and exogenous FAK are listed in table S1. The percentage of remaining mRNA was normalized by the mRNA expression at day 0 in each group. The half-life of FAK mRNA was assessed using a one-phase exponential decay model with 18S mRNA as normalization.

RNA immunoprecipitation

RIP was performed using a Magna RIP RNA-Binding Protein Immunoprecipitation kit (17-701, EMD Millipore). Before reverse transcription, deoxyribonuclease digestion was performed on the precipitated RNA to remove any residual DNA. On the basis of the position of the functional ADAR-editing site in FAK RNA in the TCGA cohort, we used two putative ADAR-binding regions in the intron 26 of FAK RNA arbitrarily named regions A and B. These regions are located on chromosome 8, and their positions are as follows: A, 141,702,900 to 141,703,155; and B, 141,702,218 to 141,702,681. Regions A and B were pulled down by antibodies against ADAR1, FLAG-epitope, or negative-control IgG and amplified by qRT-PCR after reverse transcription.

In vitro RNA-protein interaction

ADAR, ADAR dsRBD mutant, and ADAR deaminase mutant proteins were in vitro–translated using the 1-Step Human Coupled IVT kit (88882, Thermo Fisher Scientific) in accordance with the manufacturer’s instructions. T7 promoter sequence was added by PCR to the 5′ and 3′ ends of FAK region A. Double-stranded RNA was in vitro–transcribed using the Megascript T7 Kit (AM1333, Ambion) and purified using the phenol:chloroform extraction and isopropanol precipitation protocol. Double-stranded RNA was biotin-labeled using the RNA 3′ End Desthiobiotinylation kit (20163, Thermo Fisher Scientific). Next, 100 μg of in vitro–translated protein and 50 pmol labeled double-stranded RNA were mixed and put through the Magnetic RNA-Protein Pull-Down kit (20164, Thermo Fisher Scientific) in accordance with the manufacturer’s instructions. Western blot was performed on the resulting eluent to assess for the interaction of ADAR with the double-stranded RNA.

FAK inhibitor treatment

H1299 and HCC827 cells (1 × 104/100 μl) were placed in 96-well plates for 24 hours and then treated with the FAK inhibitors PND1186 (SR-2156, Selleckchem) and VS-6063 (S7654, Selleckchem) at varying concentrations (0.1, 1, 2.5, 5, and 10 μm) in 1% DMSO for 72 hours. Cell viability was determined by CellTiter-Glo (G7570, Promega). Invasion assays and Western blots were performed on cells treated with FAK inhibitors (2.5 μm) for 72 hours.

Identification of editing sites in FAK in human LUAD

FAK editing sites were identified through a manually curated process where we compared mRNA sequenced reads for TCGA LUAD samples (N = 230 patients) using the Integrated Genomics Viewer tool developed at the Broad Institute. Candidate editing locations were chosen when they exhibited a good concordance in T>C changes across the reference cell line and the patient samples. We then used the “mpileup” tool from “SAMtools” to interrogate the chosen positions across all the TCGA samples with available sequencing data, and we counted the number of reads with T>C changes at those chromosomal locations per sample.

Statistical analysis

The results of all experiments represent the mean ± SD of at least three separate experiments. All statistical tests were two-sided and performed with R 3.3.1 (R Core Development Team) using the clinfun and coin packages. Statistical significance is defined as P < 0.05. No formal multiple-testing adjustments were conducted.

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/10/497/eaah3941/DC1

Fig. S1. The genetic alteration of ADAR in TCGA cohorts.

Fig. S2. The impact of ADAR KD on the downstream targets and cell viability.

Fig. S3. The editing status of FAK in the TCGA cohort.

Fig. S4. ADAR does not bind to region B in FAK.

Fig. S5. Comparison of DNA sequences of wild-type and edited pcDNA-FAKE+I.

Fig. S6. Cell viability after treatment with FAK inhibitors.

Table S1. PCR primers.

REFERENCES AND NOTES

Acknowledgments: Funding: This work was supported by grants R01 CA136705 (to D.R.J.), R01 CA192399 (to M.W.M.), R01 CA132580 (to M.W.M.), and U54 CA137788 (to P.S.A.). This work was also supported, in part, by NIH/National Cancer Institute Cancer Center Support Grant P30 CA008748. Author contributions: E.M.A. and S.D. designed and performed the experiments, analyzed the data, and wrote the manuscript. Y.L. helped with the experimental designs, data analysis, and writing of the manuscript. P.S.A provided the LUAD tissue microarray samples and data. K.S.T. assisted with the statistical analyses. N.C., S.D., M.W.M., F.S.-V., and N.S. assisted with writing and preparation of the manuscript. D.R.J. provided the conceptual design of the study, supervision, analysis of the data, and writing of the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The Illumina microarray data have been submitted to the Gene Expression Omnibus (www.ncbi.nlm.nih.gov/geo) under accession no. GSE93035.
View Abstract

Navigate This Article