Silencing of Long Noncoding RNA MIR22HG Triggers Cell Survival/Death Signaling via Oncogenes YBX1, MET, and p21 in Lung Cancer

Approximately 80% of lung cancers are non–small cell lung cancers (NSCLC), which include adenocarcinoma and squamous cell carcinoma (SCC) and represent a leading cause of cancer-related deaths (1). The poor prognosis and high recurrence rate of lung cancer is largely due to the high rate of metastases and complex cellular, molecular, and tumor microenvironmental factors (2, 3). Despite the emergence of new adjuvant therapy regimens and targeted biologic agents, the 5-year survival rate of NSCLC remains dismal at 18% (1). It is of paramount importance to understand the underlying pathologic mechanisms contributing to NSCLC to develop novel diagnostic biomarkers and therapeutic strategies and improve survival (4, 5).

Recent studies suggest that long noncoding RNAs (lncRNA) are involved in the initiation and progression of cancer and are often highly deregulated in tumors (6–9). LncRNAs are operationally defined as transcripts that are larger than 200 nt without protein coding potential, and they have been shown to play vital roles in various aspects of cancer biology (10, 11). Mounting evidence has shown that lncRNAs influence various cellular processes, such as proliferation, cell-cycle progression, cell proliferation, and apoptosis (12, 13). LncRNAs have been found to act as tumor suppressors or oncogenes (14, 15). In this regard, identifying cancer-associated lncRNAs and the molecular mechanisms they affect is necessary for understanding progression and establishing better treatments for NSCLC.

High-throughput RNA sequencing (RNA-Seq) in human cancers has shown a remarkable potential for identifying both novel markers of disease and uncharacterized aspects of tumor biology, particularly new lncRNA species (16, 17). To define lncRNAs important for lung cancer, by analyzing three large sets of RNA-Seq data including The Cancer Genome Atlas (TCGA), Seo, and UM, we identified 281 lncRNAs with a receiver operating characteristic (ROC) >0.7 or <0.25, which were listed in our previous publication (18). MIR22HG was one of the most downregulated lncRNAs in lung cancer as compared with normal lung tissues; thus, its biological functions and underlying molecular mechanisms in this disease were examined in this study.

The lung cancer and paired nontumor lung tissues were obtained from patients undergoing curative cancer surgery during the period from 1991 to 2014 at the University of Michigan Health System (Ann Arbor, MI). None of the patients included in this study received any preoperative radiation or chemotherapy. Written informed consents were obtained from subjects or their authorized representatives. The study protocol was approved by the University of Michigan Institutional Review Board and Ethics Committee. Resected specimens were frozen in liquid nitrogen and then stored at −80°C until use. The median follow-up time was 8.12 years among the patients that remained alive. The clinical information of lung adenocarcinoma (LUAD) samples used in qRT-PCR validation (101 LUADs) are in Supplementary Table S1.

Lung cancer cell lines (PC-9, H1975, H1299, H838, and H2228) were obtained from the ATCC (Supplementary Table S2). All cell lines were genotyped for identity at the University of Michigan Sequencing Core and were tested routinely for mycoplasma contamination using MycoAlert Mycoplasma Detection Kit (Lonza). All cell lines were maintained in RPMI1640 or EMEM supplemented with 10% FBS and 1% antibiotic–antimycotic agents and cultured at 37°C in a 5% CO₂ cell culture incubator. Cells were grown for no more than 25 passages in total for any experiment.

Two published microarray datasets representing 668 primary lung adenocarcinoma tissues were downloaded from NCBI Gene Expression Omnibus (GSE31210 and GSE68465). These included Okayama and colleagues, 226 LUADs with stage I and II (19), and Shedden and colleagues, with 442 stage I to III LUADs (20). The CEL files of microarray data were normalized using robust multiarray average method (21). We also obtained Seo (22) and TCGA (23) RNA-Seq datasets consisting of a total of 394 adenocarcinomas, 212 SCCs, and 150 normal lung tissues (https://portal.gdc.cancer.gov). Expression levels of transcripts were represented as FPKM (24). RNA-Seq of UM cohort (24), including 113 lung cancer tissues (67 LUADs, 36 SCCs, 10 large-cell lung cancers, and 6 normal), was also included in this analysis. Our primary outcome was overall survival, censored at 5 years. The information concerning adjuvant chemotherapy or radiotherapy was provided in the original articles.

Cells were plated at a desired concentration and transfected with targeting siRNA oligonucleotides or nontargeting controls 2 to 24 hours after plating with final concentration of 10 nmol/L. Knockdown was performed with Lipofectamine RNAiMax Reagent (Invitrogen) in OptiMEM medium according to the manufacturer's instructions. Knockdown efficiency was determined by qRT-PCR. All siRNAs were purchased from Dharmacon, and the sequences used in this study are provided in Supplementary Table S3.

Cells were plated in 96-well plates at a desired concentration and transfected with 10 nmol/L experimental siRNA oligonucleotides or nontargeting controls at 2 to 24 hours after plating. At 12, 24, 48, 72, 96, and 120 hours after transfection with siRNA, the proliferation rates were measured by Cell Proliferation Reagent (WST-1; Roche) according to the manufacturer's instructions. The cell viability percentage was calculated by normalizing to the nontarget control siRNA.

For colony formation, lung cancer cells transfected with MIR22HG siRNAs and control siRNA were seeded in a 6-well plate at a density of 200 cells/well. After 10 to 14 days of incubation at 37°C, 0.1% crystal violet (Sigma-Aldrich) and 20% methanol were used as dye solution to fix and stain the colonies. The number of colonies was counted in each well. Clones containing more than 50 cells were counted using a grid. Three independent experiments were performed.

RNA immunoprecipitation (RIP) assays were performed as described previously. RIP assays were performed using a Millipore EZ-Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore) according to the manufacturer's instructions. Antibodies to be used for RIP are rabbit polyclonal IgG (Millipore) and YBX1 (Santa Cruz Biotechnology) using 5 μg of antibody per RIP reaction.

Data were analyzed using GraphPad Prism 6 (GraphPad software), Excel, and R software. ROC curve and the area under the curve (AUC) analysis was used to show the tradeoff between sensitivity and specificity for the different possible cut-off points of a diagnostic test. Kaplan–Meier survival curves with log-rank test was used for survival analysis. Heatmaps were used to visualize the difference of gene expression using Tree View software. The other data such as proliferation were evaluated by unpaired Student t test. A two-tailed P value <0.05 was considered statistically significant. DAVID pathway analysis (25) was used to determine the potential biological processes associated with MIR22HG correlated genes.

Other materials and methods including RNAs extraction and qRT-PCR, protein extraction, immunoblot analysis and tissue array, cell-cycle analysis by flow cytometry, and MIR22HG lentiviral expression construct transfection were detailed in Supplementary Methods.

In the current study, a comprehensive analysis of the gene expression data matrix from three cohorts of RNA-Seq data was performed to define differentially expressed lncRNAs in LUADs. The three cohorts were two large publicly available RNA-Seq data: the Korean cohort (Seo; ref. 22), including 85 LUADs and 77 normal lung tissues; TCGA lung data (23), including 309 LUADs, 212 SCCs, and 73 normal lung tissue samples; and the UM cohort, including 113 lung cancer tissues (67 LUADs, 36 SCCs, 10 large-cell lung cancers, and 6 normal; ref. 24). MIR22HG, one of the top dysregulated lncRNAs, was found to be significantly reduced in LUAD as compared with normal lung tissues in Seo, TCGA, and UM RNA-Seq datasets (Fig. 1A–C). Accordingly, ROC curve analysis was performed, and the AUC values were larger than 0.94 in all three cohorts (Fig. 1D–F), indicating that MIR22HG may be potentially used as a novel diagnostic marker for this type of lung cancer. MIR22HG expression was also significantly lower in SCC and large-cell lung cancer (Supplementary Fig. S1A and S1B). Next, where survival information was available, the association of MIR22HG expression and patient survival in two independent published LUAD microarray datasets, Okayama and colleagues (226 LUADs; ref. 19) and Shedden and colleagues (442 LUADs; ref. 20) was evaluated. Kaplan–Meier survival curves and log-rank tests showed that lower expression level of MIR22HG was significantly correlated with poor patient outcome in the Okayama dataset (P = 0.02) and Shedden dataset (P = 0.02; Fig. 1G and H). These data demonstrated that the downregulation of MIR22HG may play a tumor suppressor role in lung cancers and could be a novel diagnostic marker.

To further verify the MIR22HG expression pattern discovered from RNA-Seq and microarray datasets, qRT-PCR assays using mRNA from an independent UM cohort including 101 LUADs and 27 normal lung tissues were performed. The qRT-PCR results showed that MIR22HG expression levels were significantly lower in LUAD as compared with normal lung tissues (P < 0.001; Fig. 2A). The AUC was 0.79, indicating that MIR22HG expression level could separate the tumors from normal lung (Fig. 2B). Kaplan–Meier survival curve and log-rank test indicated that higher expression of MIR22HG was significantly associated with better patient survival (P = 0.003; Fig. 2C). The relationship between MIR22HG expression and other clinical variables from this validation cohort was also analyzed. These clinical variables included age, gender, smoking status, tumor stage, lymph node status, differentiation, and KRAS mutation status. It was found that MIR22HG level had a lower trend in stage III (vs. stage I) NSCLC and in patients with positive lymph node metastasis (vs. N0; Supplementary Table S1).

To determine whether MIR22HG expression is associated with other types of cancer, we analyzed RNA-Seq expression data from the MiTranscriptome database, which contains 6,220 cancers (17). MIR22HG was found to be decreased in many types of cancer, including bladder, breast, gastric, head-neck, kidney, and liver cancer (FPKM log₂ value, normal vs. cancer, t test, P < 0.001; Fig. 2D). Interestingly, we found metastatic cancers of prostate and thyroid have even lower levels of MIR22HG as compared with their own primary cancer (Fig. 2E). These results indicated that MIR22HG expression is generally lower in most types of cancer and may also be an indicator of metastatic tumor.

We also analyzed MIR22HG expression levels in 33 lung cancer cell lines from RNA-Seq data (24). We found that MIR22HG was expressed in all 30 NSCLC cell lines, with lower expression in the small-cell lung cancer cell lines, H146, H526, and H82 (Supplementary Fig. S2).

The biological function of MIR22HG in lung cancer is poorly understood. To explore the possible role of MIR22HG in lung cancer pathogenesis, we used siRNA technology to knock down MIR22HG expression in 5 lung cancer cell lines having different genomic backgrounds, including H1975 (EGFR T790 mutation), PC-9 (EGFR 19 exon deletion), H2228 (EML4-ALK fusion), H1299 (TP53 mutation), and H838 (TP53 mutation; Supplementary Table S2). Using both RNA-Seq and qRT-PCR, we confirmed MIR22HG was expressed in these cell lines (Supplementary Fig. S3A and S3B). To minimize the possibility of off-target effects, we used SMARTpool gene-specific siRNAs including 4 different siRNAs. The knockdown efficiency was confirmed by qRT-PCR in these cells (Fig. 3A).

We then measured effects on cell proliferation using WST-1 assays after silencing of MIR22HG by siRNAs in these lung cancer cell lines. We found that the cell proliferation was increased by 30% to 70% as compared with control siRNA at 96 to 120 hours (Fig. 3B). Colony formation assays further confirmed that cell proliferation was affected by MIR22HG knockdown (Fig. 3C; Supplementary Fig. S3C). Cell-cycle analysis by flow cytometry indicated that the G₁ phase was arrested in PC-9 cells, whereas G₂ phase was arrested in H838 cells after MIR22HG knockdown with siRNA at 48 hours (Supplementary Fig. S3D), indicating that the effects on cell-cycle regulation were different in these cells. The possible reasons for why cell proliferation was increased but cell cycle was arrested are discussed in Discussion part. Transwell assays were employed to evaluate whether cell invasion and migration were affected by MIR22HG in lung cancer cells. We found that cell invasion through Matrigel-coated membranes significantly increased with MIR22HG knockdown as compared with the cells treated with nontarget control siRNA (Fig. 3D; Supplementary Fig. S3E). Consistent with the invasion assay results, silencing of MIR22HG also significantly promoted cell migration (Fig. 3D; Supplementary Fig. S3E). These results suggested that MIR22HG is involved in lung cancer pathogenesis by affecting proliferation/cell-cycle regulation, colony formation, invasion, and migration.

To uncover the cellular localization of MIR22HG, we measured MIR22HG expression in both nuclear and cytoplasmic fractions from H1975, H838, PC-9, and H1299 cells by qRT-PCR. The differential enrichment of GAPDH for cytoplasmic and U1 for nuclear expressions were used as fractionation indicators. We observed that MIR22HG level was primarily in cytoplasm (61%–76%), and less in nuclear fractions (24%–39%; Fig. 3E; Supplementary Fig. S3F). This suggested that MIR22HG may play a regulatory function mainly occurring in the cytoplasm.

To provide mechanistic insight into the role of MIR22HG in regulating lung cancer cell growth, we first performed receptor tyrosine kinase phosphorylation antibody arrays, which include 49 different phosphorylated proteins, and confirmed by Western blot analysis for proteins involved in major lung cancer–related pathways. We found that the MET and YBX1 proteins were significantly decreased, whereas p21 protein was increased after MIR22HG siRNA knockdown (Fig. 4A–C). Furthermore, we found that these proteins were altered as early as 18 to 24 hours after MIR22HG knockdown (Fig. 4C). RT-PCR indicated that MET mRNA was significantly decreased in all tested cell lines (Fig. 4D) after MIR22HG knockdown, suggesting that MIR22HG regulation of MET expression may be at the transcription level. YBX1 mRNA, however, was not changed after MIR22HG knockdown, indicating that YBX1 was regulated by MIR22HG at the protein level (Fig. 4E). The mRNA of p21 was increased by 20% to 40% in H2228, H1975, and PC-9 cells, while not changed in H1299 and H838 cells (Fig. 4F), suggesting both transcriptional and translational regulation mechanisms for p21 by MIR22HG in these cells. Protein levels of AKT, ERK1/2, and STAT3 were not changed after MIR22HG knockdown (Supplementary Fig. S4A and S4B).

We next examined how MIR22HG regulates YBX1, MET, and p21. It is reported that YBX1 can bind to the promoter of MET and regulate MET expression at both the mRNA and protein levels, and p21 could be also a potential target of YBX1 (26). Because YBX1 is an RNA/DNA–binding protein, an important oncoprotein (27–29) and YBX1 mRNA was not changed after MIR22HG knockdown (Fig. 4E), we hypothesized that MIR22HG may regulate MET or p21 through affecting the YBX1 protein. To test the interaction between MIR22HG and YBX1 protein, we performed RIP assays in which the RNA–YBX1 complex was immunoprecipitated using an YBX1 antibody. The amount of MIR22HG RNA in the coprecipitate was measured by qRT-PCR. As compared with the immunoglobulin G (IgG)-bound sample, we found that the YBX1 antibody–bound complex had a significant increase in the amount of MIR22HG, indicating MIR22HG may directly interact with the YBX1 protein (Fig. 5A). YBX1 was degraded by the 20S proteasome in an ubiquitin- and ATP-independent manner and was abolished by the association of YBX1 with mRNA (30). To explore the role of MIR22HG in YBX1 protein regulation, we first treated PC-9 cells with the protein synthesis inhibitor, cycloheximide, followed by MIR22HG knockdown. We found that the half-life of YBX1 was much shorter in MIR22HG knockdown cells as compared with controls (Fig. 5B). Next, we treated with MG132, a proteasome inhibitor, to test whether decreased YBX1 is through proteasome degradation. We found that the endogenous YBX1 protein in MIR22HG knockdown cells was not decreased (Fig. 5C), indicating that YBX1 degradation may be through ubiquitin conjugation–mediated protein degradation. Together, these findings suggest that MIR22HG binds and stabilizes the YBX1 protein and through a MIR22HG–YBX1 complex to promote MET transcription and inhibit p21 expression (Fig. 5D).

To confirm YBX1 is regulated by MIR22HG, we performed YBX1 immunofluorescence staining on PC-9 cells after MIR22HG knockdown. We found YBX1 protein was present in both the cytoplasm and nucleus and accumulated in nucleus at mitosis in control cells (Supplementary Fig. S5A), which was consistent with previous reports (31, 32). Importantly, we found that the nuclear and cytoplasmic total YBX1 protein was significantly decreased upon MIR22HG knockdown, suggesting that MIR22HG may not only affect total YBX1 degradation, but also affect YBX1 nuclear translocation.

To explore whether MET, p21, or MIR22HG were affected by YBX1, we transfected YBX1 siRNA into PC-9, H1975, H1299, H838, and H2228 cells. We found that both MET protein and mRNA were decreased (Supplementary Fig. S5B, S5E, and S5F), consistent with MIR22HG knockdown shown in Fig. 4, whereas p21 protein increased after YBX1 knockdown (Fig. 5E), which was similar to MIR22HG knockdown shown in Fig. 4. The change in p21 mRNA (increased in PC-9 and H1975, not changed in other cells) after YBX1 knockdown is also similar to MIR22HG knockdown (Fig. 4F; Supplementary Fig. S5C). We found that MIR22HG level was decreased upon YBX1 knockdown (Fig. 5G), indicating that MIR22HG expression was also affected by YBX1, perhaps through a positive feedback mechanism or directly regulating MIR22HG through targeting its promoter (Fig. 5D). Again, knockdown of YBX1 did not affect the expression of AKT, ERK1/2, and STAT3 protein levels (Supplementary Fig. S5D).

We have performed silencing of both MIR22HG and YBX1 to examine changes to the downstream proteins MET and p21 measured by Western blot analysis. We found that MIR22HG has a stronger regulation on MET or p21 than YBX1 (Fig. 5H). Because YBX1 is involved in cancer progression (28), we tested whether YBX1 affects cell proliferation in lung cancer cells. We found that cell proliferation was significantly decreased in PC-9 and H1975 cells after YBX1 knockdown but produced less than a 20% decrease in H2228, H1299, and H838 cells (Supplementary Fig. S5E), suggesting that YBX1 may only affect a subset of cancers by itself.

MET is also a well-known oncogene that can promote cell proliferation. We used siRNA to knock down MET expression, with the efficiency of MET knockdown confirmed by qRT-PCR in H1975, PC-9, H1299, and H2228 cells (Supplementary Fig. S5F). Proliferation was decreased in the PC-9 and H1975, but was not changed significantly in H2228, H1299, and H838 cells (Supplementary Fig. S5G) after MET knockdown. We found that YBX1 protein/mRNA were increased slightly after MET knockdown in PC-9 and H1975 cells but not changed in H1299 and H2228 cells (Supplementary Fig. S5H and S5I), indicating that a negative feedback between MET and YBX1 may be present in some cells. Similarly, we found that p21 protein/mRNA was increased in PC-9, H1975, and H2228 cells, but not in H1299 cells (Supplementary Fig. S5H and S5J), suggesting that MET may inhibit p21 expression in a subset of tumor cells. We also found that MET knockdown did not change MIR22HG expression except in H1975 cells (Supplementary Fig. S5K). Proliferation was decreased or not changed by YBX1 or MET knockdown (decreased in PC-9 and H1975, but not significantly changed in H222, H1299, and H838 cells), which was the opposite to MIR22HG knockdown where cell proliferation was increased in all of these five cell lines. These results indicate that the increased cell growth by MIR22HG knockdown is not due to YBX1 or MET knockdown, and the role of YBX1 or MET in regulation of cell proliferation is limited. We speculate that another protein, p21, may actually potentially play an oncogenic role in lung cancer, although p21 has been reported as a tumor suppressor gene (33, 34).

The p21 protein affects many biological processes (35) and can promote carcinogenesis and tumor progression (36, 37). Cell proliferation was increased upon MIR22HG knockdown, but YBX1 and MET expression was decreased. Knocking down of YBX1 or MET decreased cell proliferation in PC-9 and H1975 cells, indicating that other protein, potentially increased p21, may be involved in processes regulated by MIR22HG. To test the effect of p21 on cell growth, we performed p21 knockdown with siRNAs against p21 in five lung cancer cell lines. The knockdown efficiency of p21 is large than 70% as compared with nontarget control (Supplementary Fig. S6A). Surprisingly, we found that cell proliferation was dramatically decreased in all of these five cell lines (PC-9, H1975, H1299, H838, and H2228) after silencing of p21 by siRNA (Fig. 6A), indicating that p21 may play an oncogenic role in lung cancer. This is an agreement with our previous observation that MIR22HG knockdown significantly increased p21 expression (Fig. 4B and C) and promoted cell proliferation (Fig. 3B). To test whether MIR22HG could rescue p21 in regulation of cell proliferation, we transfected MIR22HG siRNA, p21 siRNA, or both siRNAs, respectively, into PC-9 and H838 cell lines. We found that the cell proliferation affected by p21 knockdown could be partially rescued by MIR22HG knockdown (Fig. 6B and C). This finding suggests that increased cell proliferation by MIR22HG knockdown may be via p21 activation to decrease cell proliferation by decreased MET and YBX1 expression.

Cytoplasmic p21 has been found to be overexpressed in a variety of human cancers, including breast, cervical, prostate, renal, and testicular (38). Cytoplasmic localization of p21 was linked to antiapoptotic actions, enhanced cell growth and survival, as well as acting as a chaperone for cyclin E (39). We found that the more abundant changes in protein expression of p21 and MET were in cytoplasm after MIR22HG knockdown in lung cancer cell lines (Supplementary Fig. S6B). Using the marker of apoptosis cleaved PARP, we observed decreased cleavage (along with increased p21, while p27 and TP53 were not changed) upon MIR22HG knockdown (Supplementary Fig. S6C), indicating an antiapoptosis effect by MIR22HG knockdown perhaps through p21 activation. To confirm whether p21 may play an antiapoptosis role and whether MIR22HG can rescue this phenotype, we measured PARP cleavage after treatment with siRNAs specific to MIR22HG, p21, or both in PC-9, H1975, and H838 cells. We found that p21 knockdown led to a significant increase in cleaved PARP, and MIR22HG knockdown could rescue apoptosis (Fig. 6D). We also found that CCNE1 protein was decreased upon p21 knockdown, which is consistent with p21 acting as a chaperone for cyclin E (Fig. 6E; ref. 39). In agreement with these findings, CCNE1 protein was increased after MIR22HG knockdown (Fig. 6F), indicating that upregulation of CCNE1 by MIR22HG knockdown is through increased p21. This result demonstrates that increased cell growth by MIR22HG knockdown may be partially through p21 antiapoptosis effects and promoting cell proliferation in lung cancer.

Following knockdown of p21, we found MIR22HG expression decreased, indicating that a positive feedback was present between p21 and MIR22HG, especially in H1299 and H838 cells (Supplementary Fig. S6D). A positive feedback was also present between p21 and phosphor-MET in H2228, PC-9, and H838 cells (Supplementary Fig. S6E).

MYC is reported to be inhibited by p21 (33); thus, we tested whether MYC is involved in MIR22HG signaling. We found that MYC protein was slightly increased upon p21 knockdown (Supplementary Fig. S6E), which was consistent with MIR22HG knockdown (MYC was decreased; Supplementary Fig. S6F), indicating that MIR22HG regulation of MYC may also be via p21. MYC mRNA was increased after p21 knockdown, suggesting this regulation is at the transcriptional level especially in H1299, H1975, and H838 cells (Supplementary Fig. S6G). We did not find TP53 was changed after MIR22HG knockdown (Supplementary Figs. S6B and S6F), although p21 was significantly increased (Fig. 4B and C; Supplementary Fig. S6B), indicating that MIR22HG regulation of p21 is independent of TP53. Knockdown of p21 also did not affect TP53 expression (Supplementary Fig. S6E).

To confirm the relationship between MIR22HG and p21, we transfected an MIR22HG overexpression construct into PC-9 and H1299 cell lines and found that p21 was markedly inhibited in expression (Fig. 6G and H; Supplementary Fig. S6H).

A recent report indicates that chronic overexpression of p21 (>10 days) acts as a powerful selection process, promoting both the escape of a subpopulation of cells from a senescence-like state and replication stress due to continuous re-replication (40). The key positive effects and functions of p21 were found to correlate with both modulation of apoptosis and the induction of senescence (33). In fact, p21 was first identified as an overexpressed gene in normal human diploid senescent fibroblasts (41), and later found to induce premature senescence in human lung cancer H1299 cells (42). We performed two rescue experiments to confirm the relationship between MIR22HG and p21 by cell proliferation and apoptosis analysis shown previously (Fig. 6B–D). To examine whether the senescence phenotype induced by MIR22HG knockdown can be rescued by p21 knockdown, we transfected siRNAs specific to MIR22HG, p21, or both into PC-9 and H838 cells followed by measurement of β-galactosidase activity. We observed that the expression of MIR22HG siRNAs remarkably increased the number of cells with typical senescence morphology, whereas p21 knockdown decreased β-galactosidase activity (Supplementary Fig. S6I and S6J). Finally, knockdown of p21 expression rescued the senescence phenotype induced by MIR22HG siRNA. Together, these results indicate that MIR22HG exerts its function at least in part by regulating p21 expression.

To further understand the role of MIR22HG in human primary tumors, we performed Pearson correlation analysis between MIR22HG and 16,000 coding genes based on RNA-seq data from 3 cohorts, including a total 461 LUADs (Seo 85 LUADs, UM 67 LUADs, and TCGA 309 LUADs). There are 444 coding genes with negative correlation (r < −0.35, P < 0.001) and 168 with positively correlated coding genes (r > 0.35, P < 0.001; Supplementary Fig. S7A). Using DAVID gene annotation and pathway analysis, we found that the MIR22HG-negative correlated genes were significantly involved in cell-cycle regulation (Fig. 7A; Supplementary Fig. S7B and S7C). Among these MIR22HG-negative correlated genes, the cell proliferation marker, MKI67 (protein Ki-67 coding gene), was significantly negatively correlated with MIR22HG in tumors but not correlated in normal lung tissues (Supplementary Fig. S7D), indicating that MIR22HG may inhibit tumor growth/cell cycle/proliferation in primary tumors. Upon further analysis of genes in the list, we found that increased MKI67 and other cell cycle/proliferation–related genes, CHEK1, BIRC5, and BUB1, were significantly unfavorable for patient survival (Supplementary Figs. S7E and S7F).

We found MET mRNAs were decreased by 34% to 46% after MIR22HG knockdown in 5 lung cancer cell lines (Fig. 4D). p21 mRNAs were increased by 20% to 40% in H2228, PC9, and H1975 cells, while no change was observed in H1299 and H838 cells (Fig. 4F). This indicated that MIR22HG regulates MET at the transcription level in all cells, whereas p21 only in a subset of cells. To evaluate whether this status found in vitro is also presented in primary tissues, we analyzed the MIR22HG, CDKN1A (p21), YBX1, and MET mRNA expression (Supplementary Fig. S8A and S8B) and performed the Pearson correlation analysis between MIR22HG and MET or p21 mRNA in multiple lung RNA-Seq datasets (total 1,096 samples), including normal, tumor, as well as lung cancer cell lines. We did find a significantly positive correlation between MIR22HG and MET mRNA in two sets of lung cancer cell lines (Supplementary Fig. S8C and S8D), but we did not found a significant correlation between MIR22HG and MET mRNA in normal and tumor tissues (Supplementary Fig. S8C). A possible reason may be the tumor microenvironment affecting MIR22HG in regulating MET expression. Further investigations of the correlation between MIR22HG and MET protein on tissue array are warranted.

Regarding the correlation between MIR22HG and p21 mRNA, we did not find a significant correlation in both primary tumor and cell lines (Supplementary Fig. S8C), which is constant with our finding that MIR22HG affects p21 mRNA expression in a subset of tumor cells. To our surprise, MIR22HG is significantly correlated to p21 mRNA in two sets of normal lung tissues (Supplementary Fig. S8C and S8E), indicating MIR22HG may positively regulate p21 at the transcription level in normal lung tissues.

Clinical studies of the presence of p21 in tumors have yielded divergent findings regarding its roles in cancer. Clinically, elevated levels of p21 have been detected in many human tumors associated with poor survival (gliomas, prostate, cervical, ovarian, and esophageal), whereas in other tumors (breast, gastric, and ovarian), loss of p21 expression has been associated with poor prognosis and decreased overall survival (33). We evaluated the association of p21 mRNA and patient survival in the Shedden and colleagues dataset, which includes 442 LUADs. Kaplan–Meier survival curve and log-rank tests showed that higher expression levels of p21 mRNA were significantly correlated with poor patient outcome (P = 0.04; Supplementary Fig. S9A). We have previously found that the cell proliferation was decreased upon p21 knockdown in lung cancer cell lines. Takeshima and colleagues (43) reported that p21 protein expression was significantly positively correlated with Ki-67 expression but not with p53 expression in 91 LUADs using IHC staining, suggesting that p21 may be involved in tumor cell proliferation. To confirm the p21 expression and tumor proliferation in lung cancer tissues, we performed IHC of p21 and proliferation marker Ki-67 on tumor tissue array (TMA) containing 97 lung adenocarcinomas (Supplementary Fig. S9B–S9M). Approximately 25% of tumors show positive staining for both p21 and Ki-67 in the cytoplasm or nucleus. These results indicate that p21 may promote tumor growth in a subset of lung cancer.

Recently, a number of lncRNAs have been identified and characterized as important factors in fundamental physiologic and pathologic processes (10, 11, 44). LncRNAs play essential roles as either tumor suppressors or oncogenes, but only a small proportion of lncRNAs have been well characterized (45). In this study, we found that MIR22HG was significantly downregulated in lung cancer, and decreased expression was associated with poor patient survival. MIR22HG can bind and stabilize YBX1 protein. Silencing of MIR22HG triggered YBX1 degradation by the proteasome and decreased MET mRNA/protein expression through promoter regulation by YBX1, whereas p21 mRNA and/or protein expression may be regulated by MIR22HG partially via YBX1 or MET. MIR22HG may play a tumor-suppressive role through the regulation of cell cycle–related genes in primary tumors.

As a master regulator of cancer cell biology, YBX1 is involved in all of Hanahan's “hallmarks of cancer” (28, 46). The YBX1 protein performs its functions both in the cytoplasm (role as RNA-binding protein) and in the cell nucleus (role as transcription factor; refs. 27, 47). We found YBX1 could affect cell proliferation in a subset of tumor cells, including EGFR-mutated cells (PC-9 and H1975). Nuclear YBX1 has been reported to promote MET expression by directly binding to MET promoter in basal-like breast cancers (26). Consistent with MIR22HG, we found that both protein and mRNA of MET were decreased after YBX1 knockdown), indicating YBX1 regulates MET expression at the transcription level. Furthermore, promoter luciferase assay and ChIP-PCR experiment are warranted. YBX1 accumulates in the centrosome during the mitotic phase (31). YBX1 could transition from the cytoplasm to the nucleus in the following cases: at G₁–S phase interface, treatment with UV radiation, DNA-damaging agents, upon oxidative stress hyperthermia, interaction with SRp30c and TP53 (47). We found the YBX1 protein was present in both the cytoplasm and nucleus and accumulates in nucleus at mitosis in control cells (Supplementary Fig. S5A). The nuclear and cytoplasmic YBX1 proteins were significantly decreased upon MIR22HG knockdown, indicating that MIR22HG may not only affect YBX1 degradation, but also affect YBX1 nuclear translocation.

Several genes are reported to display two separate roles in cancer, acting as either tumor suppressors or oncogenes including Myc, E2F1, AMPK, SIRT1, TGFb1, p38a, Notch pathway, Wnt pathway, and p21 (40). The two-face role of p21 in regulating proliferation (cell cycle) and antiapoptosis during different cell stages, for example, cell-cycle arrest in early stages and promotion of cell growth with higher levels of genomic instability at later stages, especially in TP53-deficient environment (40, 48). We have found that p21 is a strong oncogene that could promote tumor growth through cell-cycle regulation, apoptosis inhibition, and senescence promotion. We also found that CCNE1 protein was increased after MIR22HG knockdown, whereas decreased upon p21 knockdown, indicating that upregulation of CCNE1 may also contribute to increased cell proliferation by MIR22HG knockdown. This result demonstrates that increased cell growth by MIR22HG knockdown may be partially through p21 antiapoptosis effects and promote cell proliferation in lung cancer. We have found that approximately 25% of primary tumors show positive staining for both p21 and Ki-67 in the cytoplasm or nucleus, indicating that p21 may promote tumor growth in a subset of lung cancers. Our results indicate that the oncogenic role of p21 in lncRNA-mediated regulatory mechanism may provide a new therapeutic strategy.

An interesting observation was that cell cycle was arrested at 48 hours after MIR22HG siRNA transfection, while cell proliferation was increased at 96 to 120 hours after MIR22HG knockdown, although we found that cell proliferation was not changed significantly at 48 hours. A potential reason may be related to the two roles of p21 in regulating proliferation (cell cycle) and antiapoptosis in different cell stages, for example, cell-cycle arrest at early stages and promotion of cell growth at late stages as reported by Georgakilas (48) and Galanos (40). We also found silencing of MIR22HG increased cell senescence at 48 hours (Supplementary Fig. S6I and S6J), and cell senescence usually occurs in cycle arrested status (49). Other possible reasons may be due to the increased oncogene CCNE1 expression after MIR22HG knockdown (Fig. 6F) as overexpressed CCNE1 could shorten the length of the G₁ cell-cycle phase, leading to aberrant origin firing and DNA replication stress and promotion of cell proliferation (50).

The evidence provided by this study suggests a model in which ultimate cell survival or death trigged by MIR22HG knockdown in lung cancer is determined by two levels of balance at the cellular and molecular levels. The cellular level reflects a balance of the phenotypes of cell proliferation/cell cycle, apoptosis, and senescence, and the molecular level reflects the balance of the expression levels and roles of several oncogenes, including YBX1, MET, p21, CCNE1, and Myc, etc., trigged by lncRNA MIR22HG (Fig. 7B). In primary tumors with a changing microenvironment, even more oncogene/cell cycle–related genes change the balance, causing either tumor growth or death (Supplementary Fig. S7A–S7F).

In conclusion, this study demonstrates that MIR22HG is a potential tumor suppressor, which acts through binding and stabilizing YBX1, thereby regulating multiple cell survival/death signals, including cell proliferation, apoptosis, and senescence by targeting oncogenes MET and p21. MIR22HG has potential as a new diagnostic/prognostic marker and a therapeutic target for lung cancer treatment.

D.G. Thomas is a consultant at Resonant Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: W. Su, D.G. Thomas, R.M. Reddy, Z. Yang, G. Chen

Development of methodology: W. Su, D.G. Thomas, A.C. Chang, G. Chen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Chen, C. Guo, Z. Wang, D.G. Thomas, J. Lin, R.M. Reddy, M.B. Orringer, A.C. Chang, D.G. Beer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): W. Su, S. Feng, X. Chen, A.C. Chang, D.G. Beer, G. Chen

Writing, review, and/or revision of the manuscript: W. Su, X. Chen, D.G. Thomas, J. Lin, R.M. Reddy, M.B. Orringer, A.C. Chang, D.G. Beer, G. Chen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Feng, X. Yang, R. Mao, A.C. Chang

Study supervision: R.M. Reddy, A.C. Chang, Z. Yang, G. Chen

This work was supported in part by the NIH (grant R01CA154365 to D.G. Beer and A.M. Chinnaiyan; grant U01CA157715 to D.G. Beer; grant R21CA205414 to G. Chen), the University of Michigan's Cancer Center Support Grant (P30 CA46592), University of Michigan's Cancer Center Thoracic Oncology Program Research Grant (G. Chen), a University of Michigan Department of Surgery RAC grant (G. Chen), and the National Natural Science Foundation of China (NSFC) (81660324 to S. Feng; 81702270 to W. Su).

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