MicroRNA-324-5p–CUEDC2 Axis Mediates Gain-of-Function Mutant p53-Driven Cancer Stemness

This article is featured in Highlights of This Issue, p. 1611

The “guardian of the genome,” TP53 is frequently mutated in human cancers (1). Majority of these mutations are missense point mutations in the core DNA binding domain (1). The six “hot spot” residues in the DNA-binding domain are categorized as “contact” (R248 and R273) mutants and “conformational” (R175, G245, R249, and R282) mutants depending on the function of the residue altered (2). Besides losing canonical tumor-suppressive functions, these hotspot mutants gain new oncogenic properties such as sustained proliferation, increased chemoresistance, invasion and metastasis, angiogenesis, and genomic instability, a phenomenon described as mutant p53 gain-of-function (GOF; ref. 3).

One of the major functions of p53 is to maintain a balance between genome stability and plasticity by regulating crucial processes such as cellular differentiation and self-renewal (4). TP53 GOF mutations perturb the regulated self-renewal of stem cells and promote de-differentiation of somatic cells to cancer stem cells (CSC; ref. 4). This is supported by the undifferentiated nature, enhanced chemoresistant and metastatic potential of GOF mutant p53 tumors (5). In a recent study, mutant p53 was found to regulate CSC proliferation in glioblastoma and breast cancers by downregulating WIP that controls YAP/TAZ stability (6). Another study reported that GOF mutant p53 induces expansion of CSC population by transcriptionally upregulating stem cell markers CD44, ALDH1, and Lgr5 in colorectal cancer (7). However, despite compelling evidence indicating a potential role of GOF mutant p53 in inducing stemness, the underlying molecular mechanism is poorly defined. Therefore, deciphering other pathways by which mutant p53 may alter CSC plasticity will provide further mechanistic insight into heightened stemness of mutant p53 bearing tumors.

GOF mutant p53 modulates the expression of various protein-coding genes and a large repertoire of miRNAs to promote cancer phenotypes (8). miRNAs have long been linked to the regulation of self-renewal and differentiation of embryonic stem cells (ESC; ref. 9). Recent reports also proposed miRNAs as critical regulators of CSC properties (10). Among the prominent miRNAs involved in regulation of CSCs, miR-7 (11), miR-34a (12), and miR-145 (13) exert tumor-suppressive roles whereas miR-21 (13) and miR-155 (14) have oncogenic functions. Thus, the understanding of the intricate network of miRNAs that specifically regulate CSCs in different cancer subtypes may have important diagnostic or prognostic value.

In this study, we investigated the stemness properties of GOF mutant p53 NSCLC cells and delineated its underlying molecular mechanism. Our experimental results suggest that GOF mutant p53 induces CSC phenotypes in NSCLC cells. We further identified miR-324–5p as a critical determinant of CSC properties in human cancer cells. We showed that increased c-Myc expression in GOF-mutant p53 cells transcriptionally upregulates miR-324–5p and underpins enhanced stemness properties of these cancer cells. In addition, we identified CUEDC2 as a downstream target gene of miR-324–5p in the context of increased CSC phenotypes. We showed that GOF mutant p53 cells maintain a low level of CUEDC2 that was substantially rescued upon miR-324–5p inhibition. The NF-κB signaling pathway has been implicated in maintaining functionality of CSCs (15). Our experimental results suggest that miR-324–5p contributes to activation of the NF-κB pathway in GOF mutant p53 cancer cells by downregulation of CUEDC2. Finally, we showed that TP53 mutations coupled with high levels of miR-324–5p expression predict poor prognosis in patients with TCGA (The Cancer Genome Atlas) lung adenocarcinoma and breast cancer. Collectively, our study provides mechanistic insight into GOF mutant p53-driven cancer cell stemness and presents the miR-324–5p–CUEDC2–NF-κB pathway as a novel regulator of CSC phenotype.

NSCLC cell lines H1299/EV, H1299/R175H and H1299/R273H were described previously (16) and cultured in RPMI-1640 (Gibco). SW480 and SkBr3 cells were obtained from the ATCC. The Human OSCC cell line UPCI: SCC131 was purchased from the University of Pittsburg. These cell lines were cultured in DMEM (Thermo Fisher Scientific). Mediums were supplemented with 10% of FBS (Thermo Fisher Scientific), 1% Pen Strep and 0.006% Gentamicin. To prevent Mycoplasma contamination, the cells were routinely cultured with Plasmocin (5 μg/mL) for 5 days. Cell lines were passaged for not more than 10 times and regularly checked for Mycoplasma infection using the MycoFluor kit (Invitrogen, #M7006). H1299 variants were authenticated by STR profiling from Cell repository, National Center for Cell Sciences, Pune and shows a percentage of match of 92% in the ATCC STR profile database. SW480, SkBr3, HT29, and SCC131 cells were not authenticated in recent times.

Mutant p53-expressing plasmids pCMV-Neo-Bam-p53-R175H and pCMV-Neo-Bam-p53-R273H were kindly provided by Dr. Bert Vogelstein (Johns Hopkins Kimmel Cancer Center, Baltimore, MD). Pre–miR-324–5p was cloned in pRNAU6.1/Neo vector provided by Dr. Nitai P. Bhattacharjee (Saha Institute of Nuclear Physics). pCMV3–N-FLAG–CUEDC2 construct was provided by Dr. Somsubhra Nath. pCMV6–XL5–RELA construct was obtained from Origene. MiR-324–5p overexpression cassette was sub-cloned from pRNAU6 into the pLKO.1 TRC vector (Addgene plasmid #10878). Packaging plasmids psPAX2 (Addgene plasmid # 12260) and pMD2.G (Didier Trono, Addgene plasmid# 12259) were used to generate the miR-324–5p overexpression lentiviral particles. Target cells were infected following the manufacturer's protocol and stable transduced cells were selected by puromycin (Gibco). Anti–miR-324–5p (AM10253) and control anti-miR (Cat no. 4464076) were obtained from Ambion, ON TARGETplus scrambled siRNA or sip53 were obtained from GE Dharmacon. All transfections were carried out using Lipofectamine 2000 (Invitrogen) and transfected cells were harvested after 48 or 72 hours for overexpression and knockdown studies, respectively.

Total RNA from cell lines was isolated using TRizol (Invitrogen). cDNA of total mRNA was synthesized from 1 μg of RNA using the Verso cDNA synthesis kit (Thermo Fisher Scientific) according to the manufacturer's protocol. For miRNA cDNA synthesis, 500 ng of total RNA was reverse transcribed using specific stem-loop primers for desired miRNAs (Supplementary Table S1). qRT-PCR was performed in StepOnePlus Real-Time PCR system (Applied Biosystems) using PowerUp SYBR Green Master mix (Thermo Fisher Scientific). β-Actin was used as the reference control. For real-time PCR of miRNAs, specific forward primers were used along with a universal reverse primer. U6snRNA was used as the endogenous reference control. The mean fold change values (2^−ΔΔCt) of three independent experiments were plotted using GraphPad Prism 8.2.0. The two-tailed t test was used to calculate the statistical significance. The qRT primers used for genes and miRNAs are listed in (Supplementary Table S1).

Cells were lysed using NP40 lysis buffer (Thermo Fisher Scientific) supplemented with protease and phosphatase inhibitor. Equal amount of protein was subjected to SDS polyacrylamide gel electrophoresis and probed with specific antibodies listed in Supplementary Table S2. Protein bands were developed on autoradiography films and in ChemiDoc (Bio-Rad) using SuperSignal West Pico chemiluminescent substrate (Thermo Fisher Scientific) and Luminata Forte (Millipore).

For subcellular protein fractionation, 3 × 10⁶ cells were incubated with 300 μL cytoplasmic buffer (10 mmol/L Tris HCl pH 7.9, 0.34 mol/L sucrose, 3.0 mmol/L CaCl2, 2 mmol/L MgCl2, 0.1 mmol/L EDTA, 1 mmol/L DTT, 0.1% IGEPAL CA630) at 4ºC for 10 minutes with gentle rocking. Cell lysates were centrifuged at 500 × g at 4ºC for 5 minutes and the supernatant was collected as the cytoplasmic fraction. To the pellet, 150 μL of nuclear extraction buffer (20 mmol/L HEPES pH 7.9, 3.0 mmol/L EDTA, 10% Glycerol, 150 mmol/L Potassium acetate, 1.5 mmol/L MgCl2, 1 mmol/L DTT, 0.5% IGEPAL CA630) was added, vortexed and incubated at 4ºC for 30 minutes with end-to-end rotation. The lysates were then centrifuged at 5,000 × g for 5 minutes and supernatants containing nuclear soluble fraction was collected.

Stable cells or transfected cells were cultured for 48 hours followed by trypsinization using 1X Trypsin-EDTA (Gibco) to obtain a single-cell suspension. Cells were counted using a hemocytometer and seeded at a concentration of 5,000 cells/well of an ultra-low attachment 6-well plate in DMEM-F12 serum-free medium supplemented with 20 ng/mL of EGF (Invitrogen), 20 ng/mL of basic FGF (Invitrogen) and 1% B27 supplement (Gibco). Medium was added at an interval of 2–3 days. After 5–7 days when the spheres were visible, photographs of the comparable groups were captured under phase contrast microscope Leica CTR4000 at ×20 magnifications or Olympus IX71 at ×10 magnifications. All experiments were performed in triplicate.

To investigate CD44 bright (CD44^Br) population, cells were trypsinized and resuspended in FACS staining buffer (1X PBS with 1%FBS and 0.02% sodium azide). The cell pellet was incubated with CD44-FITC–conjugated antibody diluted in 100 μL of FACS buffer for 20 minutes at 4ºC in dark. The cells were washed and subjected to Flow Cytometry in BD LSRFortessa and analyzed with BD FACSDiva 6.2 software. Isotype control was used to gate the cells with positive staining. The gate settings for CD44^Br population in control sample were adjusted at 1% and the same settings was applied to all other samples. Antibodies used are listed in Supplementary Table S2.

To determine cells with high ALDH activity, the ALDEFLUOR kit (Stem Cell Technologies) was used. The cells were trypsinized, resuspended in ALDEFLUOR assay buffer and incubated with activated ALDEFLUOR substrate in presence or absence of DEAB inhibitor according to the manufacturer's protocol. The ALDH-positive cells were detected in the green fluorescence channel of BD LSRFortessa. According to the manufacturer's protocol, the gate settings were done with the sample stained with ALDEFLUOR substrate and DEAB inhibitor.

Transfected cells were grown on coverslips for 48 hours and then fixed with 4% paraformaldehyde (Merck) followed by permeabilization with 1X PBS supplemented 0.025% Triton X 100. Blocking was done with 3% BSA in PBST (0.01% Tween 20). Coverslips were incubated with mouse anti-p65 (1:250 dilution, sc-8008) overnight at 4°C. Next day, cells were washed and incubated with goat anti-mouse Alexa Fluor 647 IgG (1:1,000 dilution, Invitrogen, A21235). Coverslips were mounted on slides using ProLong Gold Antifade Mountant, (Molecular Probes, P36934). Imaging was done with Leica SP8 confocal microscope using ×63 objective. To determine nuclear staining of p65, nuclei were marked in DAPI channel and the mean fluorescence intensity of p65 of more than 70 nuclei was recorded using LAS AF Lite software.

Chromatin immunoprecipitation (ChIP) assay was performed as described previously (16). Antibodies used are listed in Supplementary Table S2. MiR-324 promoter sequence (>hg18_dna range = chr17:7078262–7079262) identified on the basis of epigenetic marks (17) were retrieved from the UCSC genome browser (https://genome.ucsc.edu/). c-Myc–binding sites on miR-324 promoter were identified using TFBIND transcription factor predicting tool (http://tfbind.hgc.jp/). Primers used for ChIP PCR were designed using Primer-BLAST (https://www.ncbi.nlm.nih.gov/tools/primer-blast/) and are listed in Supplementary Table S1. The PCR amplified ChIP DNA were run on 2% agarose gel along with 100 bp ladder. Images were captured in GelDoc (Bio-Rad) system.

To analyze miR-324–5p expression in publicly available cancer cell dataset, we used the miRNA expression dataset of NCI-60 cell lines (GEO accession number GSE26375; ref. 18). On the basis of TP53 mutation status, we first classified the NCI-60 cell lines as wild-type and mutant p53 cells (http://p53.free.fr/Database/Cancer_cell_lines/NCI_60_cell_lines.html; ref. 19). Mutant p53 were further categorized as GOF and non-GOF (Supplementary Table S3; refs. 20, 21). Differential expression analysis of miR-324–5p between wild-type, non-GOF mutant p53 and GOF mutant p53 group of cell lines was carried out and statistical significance was calculated using Student t test. To validate the correlation between miR-324–5p expression and CD44 protein levels in NCI 60 cell lines, miR-324–5p expression values and protein expression intensities of CD44 were obtained from NCI-60 miRNA (GSE26375; ref. 18) and proteome (http://wzw.tum.de/proteomics/nci60; ref. 22) datasets, respectively (Supplementary Table S3) and the scatter plots were generated using GraphPad Prism 8.3.1.

To validate our cell line based observations in patient samples, we evaluated miR-324–5p and CUEDC2 expressions in 230 lung adenocarcinoma patient samples profiled by TCGA (23). On the basis of the TP53 mutation status of the TCGA patients with lung adenocarcinoma (23) we categorized them into patients with wild-type, null, missense (MS), and GOF mutant p53 and compared the median expression of miR-324–5p and CUEDC2 among these groups (Supplementary Table S4).

To investigate the clinical significance of TP53 mutations in patients with lung adenocarcinoma, we determined relative overall survival (OS) of 219 TCGA patients with lung adenocarcinoma harboring either wild-type or mutant p53 (Supplementary Table S5). Kaplan–Meier survival plot was generated using GraphPad Prism 8.3.1. Similarly, to analyze the clinical significance of miR-324–5p expression, the OS of patients with high miR-324–5p expression (top 20%, n = 44) was compared with those having low expression of the miRNA (bottom 20%, n = 44). The patients with high and low miR-324–5p expressions were further segregated on the basis of their TP53 mutation status, and the OS of these groups were compared with study the combined effect of TP53 mutation and miR-324–5p expression on prognosis of TCGA patients with lung adenocarcinoma. To analyze the clinical significance of miR-324–5p expression in TCGA breast cancer patient cohort (Supplementary Table S6), OS of patients with high miR-324–5p expression (>median expression, n = 145) was compared with patients with low miR-324–5p expression (<median expression, n = 146). These patients were then segregated on the basis of their TP53 mutation status and the combined effect of TP53 mutation and miR-324–5p expression was analyzed.

Given the undifferentiated and chemoresistant nature of mutant p53 tumors (5), we anticipated that TP53-mutated cancer cells may exhibit increased CSC characteristics. We, therefore, tested whether tumor-derived p53 GOF mutants modulate CSC marker expression in p53-deficient non–small cell lung carcinoma (NSCLC) cell line, H1299. CD44⁺ subpopulations in NSCLC cells have been previously characterized to be enriched for stem cell-like properties, including expression of stemness markers such as Bmi1 and Oct-4 (24). We compared the relative expression of CD44, Bmi1, Oct4, and Yamanaka factor c-Myc in H1299 cells either harboring an empty vector (H1299/EV) or stably expressing GOF mutant p53^R175H(H1299/R175H) and p53^R273H (H1299/R273H) (Fig. 1A). Compared with control cells, we observed increased level of CSC marker proteins in H1299/R175H and H1299/R273H cells, suggesting enriched CSC properties in GOF mutant p53 cells. Reduction in CSC marker levels upon mutant p53 knockdown in H1299/R175H cells further confirmed these observations (Fig. 1B). We also observed a dose-dependent increase in CD44, Bmi1, and Oct4 protein levels upon ectopic expression of mutant p53^R175H and p53^R273H in H1299 cells (Fig. 1C). These observations hold true for endogenous mutant p53 bearing cells as well, as silencing of GOF mutant p53^R273H in colorectal cancer cell lines SW480 and HT29 resulted in reduced CSC marker expression (Fig. 1D). We further investigated the role of GOF p53 mutants in promoting CSC phenotypes by determining their sphere forming ability. H1299/R175H and H1299/R273H cells exhibited an increased sphere forming ability as compared with control empty vector cells (Fig. 1E), demonstrating a greater CSC population in these cells that survived and formed spheres under serum-free and non-adherent condition. Consistent with these results, knockdown of endogenous mutant p53 resulted in decreased sphere formation in SW480 cells (Supplementary Fig. S1A). To directly measure the CSC population in mutant p53 cells, we determined percent CD44⁺ or CD44 bright (CD44^Br) populations (24) in H1299/EV, H1299/R175H, and H1299/R273H cells (Fig. 1F). Compared with control cells, p53 mutant cells exhibited increased CSC populations as evidenced by significantly higher %CD44^Br cells as well as increased CD44-FITC mean fluorescence intensity (Fig. 1F; Supplementary Fig. S1B). We further verified these results by measuring ALDH activity using ALDEFLUOR assay (https://www.stemcell.com/identification-aldh-expressing-cancer-stem-cells-using-aldefluor-lp.html) where we found a significantly higher percentage of ALDH bright (ALDH^Br) population in H1299/R175H cells as compared with control H1299/EV cells (Fig. 1G). Collectively, these results suggest that GOF mutant p53 cancer cells are enriched in CSC populations.

Role of cellular miRNAs in governing CSC plasticity is well documented. To identify the potential miRNA candidates involved in mediating mutant p53-driven CSC phenotypes, we investigated a previously published set of miRNAs differentially expressed between TP53^−/− and isogenic mutant p53^R273H-harboring NSCLC cells (25). Among the upregulated miRNAs, miR-324–5p was previously reported to be enriched in the side-population of gastric cancer cells, thereby indicating its potential role in cancer cell stemness (26). Compared with control empty vector carrying cells, we observed significantly increased miR-324–5p expression in H1299/mutant p53^R175H and H1299/mutant p53^R273H stable cell lines (Fig. 2A; Supplementary Fig. S2A) as well as upon transient expression of mutant p53 (Fig. 2B, Supplementary Fig. S2B). A significant reduction in miR-324–5p expression upon p53 knockdown in H1299/mutant p53^R175H cells further suggests mutant p53-dependent upregulation of miR-324–5p in these cells (Fig. 2C, Supplementary Fig. S2C). Consistent with these results, siRNA-mediated knockdown of endogenous mutant p53 in SKBr3, HT29, and SW480 cells significantly reduced miR-324–5p expression (Fig. 2D). We further validated these experimental results in publicly available cancer cell lines and patients' datasets. For this, we first analyzed NCI-60 miRNA expression dataset (GEO accession number GSE26375, (18). On the basis of the TP53 mutation status, we classified the NCI-60 cell lines as wild-type, GOF, and non-GOF mutant p53 cells (http://p53.free.fr/Database/Cancer_cell_lines/NCI_60_cell_lines.html; ref. 19–21; Supplementary Table S3). Differential expression analysis for miR-324–5p revealed significant (P < 0.05) upregulation of miR-324–5p expression in GOF mutant p53 group of cell lines (Fig. 2E). However, no significant difference in miR-324–5p expression was observed between wild-type and non-GOF mutant group of cell lines (Fig. 2E). These results suggest that increased miR-324–5p expression as a general characteristic of cancer cells carrying TP53 GOF mutations. Next, we analyzed miR-324–5p expression in The Cancer Genome Atlas (TCGA) miRNA sequencing dataset of 230 patients with lung adenocarcinoma (TCGA lung adenocarcinoma) characterized by their TP53 mutation status (refs. 2, 23; Supplementary Table S4). Differential expression analysis showed significantly (P < 0.05) higher expression of miR-324–5p in missense TP53-mutated patients as compared with patients harboring null TP53 mutations (Fig. 2F). Although not very statistically significant (P = 0.05), we observed increased miR-324–5p expression in patients harboring GOF mutant p53 as compared with the TP53 null patients (Fig. 2F). These observations, therefore, accord with the results obtained from our cell-based experiments, and suggest that miR-324–5p expression positively correlates with TP53 mutations in human NSCLCs.

Next, we investigated the underlying mechanism of increased miR-324–5p expression in mutant p53 cells. Because mutant p53 proteins lack direct DNA binding, they often piggyback to other transcription factors and regulate transcription of a large repertoire of cellular genes and miRNAs (2, 8). We, therefore tested mutant p53 recruitment to intragenic miR-324 promoter (>hg18_dna range = chr17:7078262–7079262; ref. 17) in H1299/R175H cells using ChIP assay. However, we observed binding of RNA polymerase II, but not mutant p53 to miR-324 promoter (Supplementary Fig. S2D). In contrast, recruitment of mutant p53 on CDC7 gene promoter was observed (Supplementary Fig. S2D) as reported previously (10). These results suggest that mutant p53 does not bind to the miR-324 promoter to transcriptionally activate miR-324 expression. To further investigate the molecular mechanism of mutant p53-driven upregulation of miR-324, we scanned miR-324–5p promoter for potential transcription factor binding sites using TFBIND transcription factor–predicting tool (http://tfbind.hgc.jp/; ref. 27). Interestingly, we identified multiple potential binding motifs of Myc, an oncogenic transcription factor known to be transcriptionally activated by GOF mutant p53 (Supplementary Fig. S2E; ref. 27). In accordance, we also observed increased c-Myc protein level in mutant p53 cancer cells (Fig. 1A). We hypothesized that increased miR-324 expression in mutant p53-bearing cells is attributed to elevated level of c-Myc proteins. Indeed, knockdown of c-Myc significantly abrogated increased miR-324–5p expression in H1299/R175H and H1299/R273H cells (Fig. 2G, Supplementary Fig. S2F). We further verified that these results in H1299 cells transiently transfected with mutant p53^R175H and mutant p53^R273H, where c-Myc depletion significantly compromised enhanced miR-324–5p expression (Fig. 2H; Supplementary Fig. S2G). Notably, c-Myc mRNA level was also found to be elevated in mutant p53-bearing stable lines as well as upon transient overexpression of mutant p53 (Supplementary Fig. S2F and S2G). To further confirm c-Myc–mediated transcriptional regulation of miR-324, we checked for c-Myc recruitment to miR-324 promoter in H1299/EV and H1299/R175H cells using ChIP assay. We observed prominent enrichment of miR-324 promoter region in anti–c-Myc immunoprecipitates, suggesting c-Myc binding to miR-324 promoter (Supplementary Fig. S2H). Taken together, these results suggest that GOF mutant p53 upregulates c-Myc that in turn binds to and transcriptionally activates miR-324–5p expression in cancer cells.

To investigate the correlation between miR-324–5p expression and cancer cell stemness, we compared the relative expression levels of miR-324–5p in sphere and adherent populations of H1299, SW480, and SCC131 cell lines. We observed profound increase in miR-324–5p expression in spheres as compared with their respective adherent counterparts, suggesting enrichment of miR-324–5p in the stem cell populations (Fig. 3A). Next, we cloned pre-miR sequence of hsa-miR-324–5p in a miRNA expression plasmid and subsequently determined CSC marker levels upon transient overexpression of hsa-miR-324–5p in H1299 and SW480 cells (Fig. 3B; Supplementary Fig. S3A). Notably, we observed elevated levels of CD44, Oct4, Sox2, and c-Myc proteins in both the cell lines, and Bmi1 in H1299 cells upon overexpression of miR-324–5p, suggesting a direct role of miR-324–5p in promoting stem characteristics in cancer cells (Fig. 3B). Consistent with these results, inhibition of endogenous miR-324–5p using a miRNA inhibitor led to reduced protein levels of stem cell markers in H1299/R175H, HT29, H1299 and SCC131 cell lines (Fig. 3C; Supplementary Fig. S3B). To further validate these results, we carried out an integrative correlation analysis between miR-324–5p expression and CD44 protein levels across NCI-60 cancer cell lines obtained from NCI-60 miRNA (GSE26375; ref. 18) and proteome (http://wzw.tum.de/proteomics/nci60; ref. 22) datasets, respectively (Supplementary Table S3). In agreement to our cell-based results, we found a significant positive correlation (Pearson r = 0.33, P = 0.02) between miR-324–5p expression and CD44 protein levels in NCI-60 cell lines (Supplementary Fig. S3C). Importantly, H1299 cells stably expressing miR-324–5p (H1299/miR-324–5p) exhibited greater sphere forming ability, increased CD44^Br and ALDH^Br population as compared with empty vector carrying cells (H1299/PLKO1; Fig. 3D–F). Furthermore, sphere-forming ability and ALDH^Br population were significantly reduced upon miR-324–5p inhibition in SW480 cells (Fig. 3G and H). These results collectively suggest miR-324–5p as a positive regulator of stemness in human cancer cells.

Because miR-324–5p is upregulated by GOF mutant p53 and promotes CSC phenotypes, we asked whether enhanced CSC characteristics of GOF mutant p53 cells are attributed to the increased miR-324–5p expression in them. Indeed, increase in CSC marker levels upon transient expression of mutant p53^R175H and p53^R2723H in H1299 cells was found to be markedly compromised upon miR-324–5p inhibition (Fig. 4A). Similarly, miR-324–5p inhibition abrogated elevated CSC marker expression in H1299/R175H and H1299/R273H stable cell lines (Fig. 4B and C). It should be noted, however, that the CSC marker levels in control p53-deficient H1299 cells were further reduced upon miR-324–5p inhibition, suggesting a fundamental role of miR-324–5p in driving cancer cell stemness (Fig. 4A–C). Importantly, the increased sphere-forming ability of H1299/R273H cells was significantly compromised upon miR-324–5p inhibition (Fig. 4D). These results suggest that increased miR-324–5p expression contributes to the heightened CSC characteristics of GOF mutant p53 cancer cells.

To get a mechanistic insight into miR-324–5p–mediated stemness, we looked into the putative as well as experimentally validated targets of miR-324–5p. CUEDC2, a novel CUE-domain–containing protein implicated in inflammatory response and tumorigenesis, has been experimentally shown as a direct downstream target of miR-324–5p in human cells (28, 29). To validate CUEDC2 as a downstream target of miR-324–5p, we ectopically expressed miR-324–5p in a dose-dependent manner in H1299 and SW480 cells and subsequently checked CUEDC2 expression. We observed a dose-dependent decrease in CUEDC2 protein levels and concomitant increase in CSC marker levels upon overexpression of miR-324–5p (Fig. 5A; Supplementary Fig. S4A). Consistent with these results, we observed reduced levels of CUEDC2 mRNA and protein in H1299 cells stably expressing mir-324–5p (H1299/miR-324–5p) as compared with control empty vector carrying cells (H1299/Plko1; Supplementary Fig. S4B and S4C). These observations confirmed CUEDC2 as a direct downstream target of miR-324–5p in cancer cells (28, 29). Further using in-built data analysis tool available at CellMiner CBD database (https://discover.nci.nih.gov/cellminercdb/; ref. 30), we performed an integrated miRNA–mRNA correlation analysis for CUEDC2 and miR-324–5p expression in a panel of human lung cancer cell lines available in CCLE (Cancer Cell Line Encyclopedia, Broad Institute, MIT) dataset (31). In agreement with our cell-based results, we observed a significant negative correlation between miR-324–5p and CUEDC2 expression [Pearson correlation (r) = −0.34; P = 0.0086] in CCLE NSCLC cell lines, substantiating the generality of our observation (Fig. 5B; Supplementary Fig. S4D). Because miR-324–5p is upregulated in patients with NSCLC with TP53 GOF mutations (Fig. 2F), we further asked whether TP53 mutation status influences the inverse correlation between miR-324–5p and CUEDC2 expression in those patients. For this, we analyzed miR-324–5p and CUEDC2 expression in TCGA patients with lung adenocarcinoma carrying either GOF or non-GOF mutations in TP53 gene (Supplementary Table S4). Although not statistically significant at the level of P = 0.05, we observed a marked negative correlation between CUEDC2 and miR-324–5p expression in patients harboring GOF TP53 mutations [Pearson (r) = −0.2496], but not in those bearing non-GOF [Pearson (r) = −0.0621] mutations in TP53 (Supplementary Fig. S4E). These results suggest that GOF-mutant p53-mediated upregulation of miR-324–5p may result in decreased CUEDC2 expression in human cancer cells.

Next, we investigated the possible roles of CUEDC2 in regulating stemness in human cancer cells. Notably, transient CUEDC2 overexpression in H1299 and SW480 cells resulted in a dose-dependent decrease in CSC marker levels (Fig. 5C), suggesting a regulatory role of CUEDC2 in cancer stemness. We further assessed whether miR-324–5p–driven stemness was attributed to miR-324–5p–mediated downregulation of CUEDC2. Interestingly, ectopic CUEDC2 expression alleviated increased CD44 and Oct4 protein levels in miR-324–5p–overexpressing H1299 and SW480 cells (Fig. 5D), suggesting reduction of stem characteristics in these cells upon CUEDC2 complementation. These results, therefore, suggest that miR-324–5p promotes stem cell phenotypes by targeting CUEDC2 in human cancer cells.

As GOF mutant p53 upregulates miR-324–5p that targets CUEDC2, we anticipated that GOF mutant p53 cells might exhibit reduced CUEDC2 levels. Indeed, along with increased CD44 levels, we observed a striking reduction in CUEDC2 protein as well as mRNA levels in H1299/R175H cells as compared with control H1299/EV cells (Fig. 5E; Supplementary Fig. S4F). Moreover, siRNA-mediated knockdown of endogenous mutant p53 in SW480 and HT29 cells led to elevated CUEDC2 protein level with a concomitant decrease in CSC markers (Fig. 5F). To check whether the reduced expression of CUEDC2 is a consequence of increased miR-324–5p expression in GOF mutant p53 cells, we inhibited miR-324–5p and tested relative protein levels of CUEDC2 in H1299/R175H stable line as well as H1299 cells exogenously expressing mutant p53^R175H (Fig. 5G and H). CUEDC2 protein level was found to be substantially rescued upon miR-324–5p inhibition in H1299/mutant p53^R175H cells accompanied by a striking reduction in CSC marker level (Fig. 5G and H). These observations led us to conclude that reduced CUEDC2 level in GOF mutant p53 cells is attributed to upregulated miR-324–5p, and that miR-324–5p–CUEDC2 axis is one of the potential pathways through which GOF mutant p53 promotes cancer cell stemness. To further validate these cell-based results, we analyzed CUEDC2 expression in a panel of human cancer cell lines from different cancer types, across different tissues available in the CCLE dataset. Correlation analysis of TP53 mutations and CUEDC2 expression in CCLE cell lines using CellMiner CBD correlation analysis tool (https://discover.nci.nih.gov/cellminercdb/) revealed a significant (P = 0.014) negative correlation between TP53 mutations and CUEDC2 expression [Pearson (r) = −0.08; Fig. 5I; ref. 31]. This result is consistent with our cell-based observations and strongly suggests that decreased CUEDC2 expression is a general phenotype of TP53-mutated human cancer cells. We further analyzed relative CUEDC2 expression among TCGA patients with lung adenocarcinoma harboring wild-type, null, missense or GOF mutant p53 (Fig. 5J; and Supplementary Table S4). Compared with wild-type patients, all three groups of TP53-mutated patients exhibited a downward trend in CUEDC2 expression with patients carrying TP53 GOF mutations showing significantly (P = 0.0378) reduced CUEDC2 expression levels (Fig. 5J). These results suggest that TP53 GOF mutations correlate with decreased CUEDC2 expression in patients with NSCLC.

CUEDC2 has been previously reported to suppress NF-κB signaling by recruiting protein phosphatase 1 (PP1) to the IKK complex (32). Because miR-324–5p downregulates CUEDC2, we perceived that miR-324–5p may positively regulate NF-κB signaling, one of the major signaling pathways implicated in maintaining self-renewal properties (32). As there are very few reports demonstrating NF-κB activation in lung cancer CSCs (33, 34), we tested whether NF-κB signaling was instrumental in regulating stemness in H1299 cells. Indeed, we observed a dose-dependent increase in CSC markers expression upon NF-κB p65 overexpression in H1299 cells (Supplementary Fig. S5A). We obtained similar results in oral squamous cell carcinoma cell line SCC131, suggesting that NF-κB positively regulates stemness in human cancer cells in general (Supplementary Fig. S5A). Next, to determine whether miR-324–5p modulates NF-κB signaling, we tested nuclear localization of NF-κB p65 and p50 subunit proteins upon miR-324–5p inhibition in H1299 and SW480 cells (Fig. 6A). Compared with the control anti-miR–treated cells, we observed decreased p65 and p50 proteins in nuclear fractions, but not in whole-cell extracts prepared from miR-324–5p–inhibited cells (Fig. 6A), suggesting compromised NF-κB pathway activation upon miR-324–5p inhibition. However, there is a concomitant increase in CUEDC2 level accompanied by a reduced level of CD44 in whole-cell extract upon miR-324–5p inhibition (Fig. 6A). As GOF mutant p53 augments NF-κB pathway activation in cancer cells (35), we anticipated that upregulation of miR-324–5p in mutant p53 cancer cells might contribute to hyperactivation of NF-κB signaling in these cells. Indeed, miR-324–5p inhibition abrogated increased nuclear accumulation of p65 and p50 in H1299/R175H cells (Fig. 6B). This was further accompanied by de-repression of CUEDC2 and reduced CD44 in total protein levels (Fig. 6B). We further verified these results by immunofluorescence microscopy, where enhanced nuclear localization of NF-κB p65 in H1299/R175H cells was significantly reduced upon miR-324–5p inhibition (Fig. 6C). Taken together, these observations suggest that miR-324–5p–mediated downregulation of CUEDC2 activates NF-κB signaling to promote CSC phenotypes in mutant p53-bearing cancer cells.

To investigate clinical significance of TP53 mutations in patients with lung adenocarcinoma, we determined relative OS of 219 TCGA patients with lung adenocarcinoma harboring either wild-type or mutant p53 (Fig. 7A). The association of TP53 mutation with clinical outcome in patients with lung cancer is highly debatable (36–42). A prospective study conducted by David Sidransky's laboratory revealed that TP53 mutations can independently predict poor survival in stage I patients with NSCLC, but not in stage II and II patients (43). Consistent with these observations, results from a meta-analysis study suggested that the potential survival benefit of wild-type p53 patients with NSCLC over those harboring mutated p53 is dependent on tumor stage (39). Our survival analysis in TCGA patients with lung adenocarcinoma did not reveal TP53 mutation as a significant (Log-rank P = 0.2683) predictor of survival (Fig. 7A). However, wild-type patients showed slightly better survival (median survival = 50.02 months) than patients bearing mutated TP53 (median survival = 41.33 months; Fig. 7A). Next, we aimed to determine the prognostic significance of miR-324–5p expression in TCGA lung adenocarcinoma cohort. Like TP53 mutation status, miR-324–5p expression was not found to be a significant predictor of survival in patients with NSCLC as we did not observe any significant (Log-rank P = 0.8632) difference in median survival of patients with high miR-324–5p expression (>20th percentile; median survival = 49.01 months) and those with low miR-324–5p expression (<20th percentile; median survival = 52.56 months; Fig. 7B). These results suggest that neither TP53 mutation status nor miR-324–5p expression alone can predict clinical outcome in patients with NSCLC.

We, therefore, sought to determine the combined effect of TP53 mutation and miR-324–5p expression on survival of patients with NSCLC. To this aim, based on their TP53 status, we further segregated the miR-324–5p high [wt p53 miR high (n) = 25, mtp53 miR high (n) = 19] and miR-324–5p low [wt p53 miR low (n) = 23, mtp53 miR low (n) = 21] patient groups. Interestingly, we observed a striking decrease in median survival time of TP53-mutated patients with either high or low miR-324–5p expression (36.63 and 39.32 months, respectively) as compared with wild-type patients with high or low miR-324–5p (88.07 and undefined, respectively; Fig. 7C), indicating that TP53 mutation predicts poor survival in these group of patients with NSCLC. Most importantly, among the four groups of patients, TP53-mutated patients with miR-324–5p high expression exhibited worst prognosis (median survival = 36.63 months, log rank P = 0.0520; HR, 2.763 in comparison with the WT p53/miR-324–5p low group). In contrast, wild-type patients with low miR-324–5p expression showed maximum survival as survival was beyond 50% at the time of last follow-up (median survival = undefined; Fig. 7C). We further extended our analysis in TCGA Breast Cancer patient cohort of 291 patients (44). Interestingly, patients with low miR-324–5p expression (median survival = undefined) showed significantly better (Log rank P = 0.0040) OS as compared with those with high miR-324–5p expression (median survival = 84.54 months; Fig. 7D). This suggests that high miR-324–5p expression can independently predict poor prognosis in patients with breast cancer. Furthermore, to determine the combined effect of TP53 mutation and miR-324–5p expression on survival of patients with breast cancer, we segregated the miR-324–5p high and miR-324–5p low patients based on their TP53 mutation status [wt p53 miR high (n) = 87, wt p53 miR low (n) = 109, mt p53 miR high (n) = 58 and mt p53 miR low (n) = 37]. Like patients with NSCLC (Fig. 7C), patients with breast cancer with mtp53 and high miR-324–5p expression showed worst prognosis (median survival = 83.81 months) whereas patients with wt p53 and low miR-324–5p expression exhibited maximum OS (median survival = undefined; Fig. 7E). Collectively, our survival analysis strongly suggests that TP53 mutation coupled with high miR-324–5p expression predicts poor clinical outcome in patients with cancer.

Wild-type p53 acts as the “guardian of the genome” and maintains genome stability in stem cells by regulating their differentiation and apoptotic pathways (4). In contrast, the undifferentiated and chemoresistant nature of TP53-mutated tumors suggest that TP53 mutations possibly drives CSC generation (5). The striking similarity of the transcriptional pattern of lung and breast tumors carrying TP53 mutations to that of stem cells further support this notion (45). Moreover, GOF mutant p53 knocked-in mouse embryonic cells are more efficiently reprogrammed to induced pluripotent stem cells (iPSC) as compared with those with TP53 knockout or wild-type TP53 (46). As CSCs possess enriched tumor-forming ability (47, 48), enhanced tumor formation by GOF mutant p53 cancer cells in xenograft mouse model further suggest increased CSC subpopulation in these cells. Although these experimental evidence link GOF mutant p53 to stemness phenotype, the underlying molecular mechanisms remain largely obscure. Here, we experimentally determined the CSC phenotype induced by GOF mutant p53 in human cancer cells, and subsequently investigated the underlying molecular mechanisms. Compared with TP53-deficient cells, GOF mutant p53 cells exhibited elevated CSC markers expression, greater sphere forming ability and are enriched in CD44^Br and ALDH^Br CSC populations. Furthermore, genetic depletion of endogenous GOF mutant p53 substantially abrogated increased CSC phenotypes, suggesting an important role of GOF mutant p53 in cancer cells stemness.

In addition to the protein coding genes, GOF mutant p53 regulates cellular miRNAs, thereby promoting different oncogenic properties (49). As self-renewal or de-differentiation of CSCs is mostly governed by epigenetic regulators, miRNAs play a major role in determining the fate of these cells (50). Several miRNAs have been implicated in regulating CSC phenotype where they modulate the expression of cell surface markers, and markers associated with stemness (Oct-3/4, Sox2, c-Myc, Klf4, and Nanog) or epithelia-to-mesenchymal transition (51, 52). We showed that upregulation of miR-324–5p underpins stemness in GOF mutant p53 cancer cells. We observed significant upregulation of miR-324–5p in mutant p53 carrying cancer cells. Importantly, increased miR-324–5p expression was successfully validated in NCI-60 cancer cell lines as well as in TCGA patients with lung adenocarcinoma harboring GOF mutant p53. Mechanistically, we showed that GOF mutant p53-induced c-Myc (53) transcriptionally activates miR-324–5p in human cancer cells. Unlike some of the reported mutant p53 target genes and miRNAs (8), mutant p53 does not directly bind to miR-324 promoter, rather modulates its expression by upregulating c-Myc. Thus, our study reports an indirect, however, a critical mode of miRNA regulation by GOF mutant p53.

In the present study, we report a fundamental role of miR-324–5p in cancer stemness regulation. We observed significant upregulation of miR-324–5p in spheres generated from cancer cell lines as compared with their adherent population. Furthermore, miR-324–5p overexpression led to enhanced CSC protein marker levels, greater sphere forming ability, and increased CD44^Br and ALDH^Br stem cells populations. Also, miR-324–5p inhibition abrogated the CSC properties, suggesting a critical role of miR-324–5p in maintaining stemness of cancer cells. Thus, we identified a previously uncharacterized biological function of miR-324–5p in promoting cancer cell stemness. Importantly, heightened stem characteristics of GOF mutant p53 cancer cells were found to be substantially compromised upon miR-324–5p inhibition, suggesting miR-324–5p as a critical determinant of mutant p53-driven cancer cell stemness. We further showed that miR-324–5p drives CSC phenotypes by targeting CUEDC2, a novel CUE-domain–containing protein implicated in cell-cycle regulation, inflammatory response, and tumorigenesis. Consistent with the previous findings, we detected significantly decreased CUEDC2 expression in miR-324–5p–overexpressing cells, thereby suggesting CUEDC2 as a direct downstream target of miR-324–5p in cancer cells (28, 29). Negative correlation between miR-324–5p expression and CUEDC2 levels in CCLE-LUAD cell lines as well as in TCGA patients with NSCLC with GOF TP53 mutations further corroborated our cell-based results. Downregulation of CUEDC2 has been shown to promote neuro-sphere formation of glioma cells whereas its upregulation abrogated the same (54). Here, we found that CUEDC2 overexpression led to a dose-dependent decrease in CSC marker proteins, suggesting that CUEDC2 suppresses CSC properties. Furthermore, attenuation of increased CSC marker phenotype in miR-324–5p–overexpressing cells upon CUEDC2 complementation ascertains that miR-324–5p promotes stemness by downregulating CUEDC2 in cancer cells. We also observed a striking reduction in CUEDC2 expression in mutant p53-bearing cancer cells. Consistent with our cell-based results, analysis of CCLE-Broad MIT dataset encompassing cell lines of nearly all cancer types revealed a strong negative correlation between CUEDC2 expression and TP53 mutation status. These findings were further corroborated by the results obtained from TCGA lung adenocarcinoma patient's dataset analysis that revealed patients with TP53-null mutations or those with missense mutations in TP53 gene have reduced CUEDC2 expression as compared with those bearing wild-type TP53. Most importantly, patients harboring missense TP53 mutations with reported GOF properties exhibited lowest CUEDC2 expression levels (16, 20). Thus, we identified CUEDC2 downregulation as a novel molecular signature of TP53-mutated human cancers. Of note, the reduced CUEDC2 expression in GOF mutant p53 cells was rescued upon inhibition of miR-324–5p. It would be interesting to further investigate how CUEDC2 downregulation is biologically favorable for growth and survival of TP53-mutated cancer cells.

The NF-κB signaling pathway has been implicated in regulating self-renewal and differentiation of stem cells (15). We observed a dose-dependent increase in CSC marker expression upon ectopic expression of p65, a NF-κB subunit, suggesting a role of NF-κB in promoting cancer cell stemness. CUEDC2 was previously shown to regulate inflammation by negatively modulating the NF-κB pathway (32). As miR-324–5p targets CUEDC2 to invoke CSC characteristics, we further explored whether miR-324–5p could modulate NF-κB signaling. We observed reduced nuclear translocation of NF-κB upon miR-324–5p inhibition, suggesting a positive regulatory role of miR-324–5p in NF-κB pathway activation. Furthermore, our experimental results suggest that miR-324–5p is a critical mediator of GOF mutant p53-driven NF-κB pathway activation in cancer cells.

Although TP53 is frequently mutated in human lung cancers, the mutation status solely does not affirm poor prognosis in patients. Although some retrospective studies and meta-analysis–based reports suggest TP53 mutations as independent predictor of poor clinical outcome in patients with NSCLC (36–39, 55), other independent studies report no significant association between TP53 mutations and lung cancer prognosis (40–42). In TCGA lung adenocarcinoma cohort, we observed that patients with TP53 mutations have a lower median survival time as compared with patients with wild-type p53; however, there was no significant difference in the OS. This may be due to the heterogeneity in clinical stage of the tumors selected in our analysis. Although we did not find a significant difference in OS between TCGA patients with NSCLC with high and low miR-324–5p expressions, high miR-324–5p expression was found to be an independent predictor of poor prognosis in TCGA breast cancer patient cohort. However, when we tested the combined effects of TP53 mutations and miR-324–5p expression on patients' survival, TP53-mutated patients with high miR-324–5p expression exhibited worst prognosis in patients with both lung adenocarcinoma and breast cancer.

In summary, our study provides novel mechanistic insight into the emerging role of GOF mutant p53 in regulating cancer cell stemness. We propose miR-324–5p as a novel regulator of stemness, and present miR-324–5p–CUEDC2–NF-κB axis operating downstream of GOF mutant p53 as critical signaling cascade required for enhanced CSC phenotype in cancer cells.

No disclosures were reported.

D. Ghatak: Conceptualization, data curation, formal analysis, investigation, visualization, writing–original draft. A. Datta: Conceptualization, data curation, software, formal analysis, validation, investigation, writing–original draft. T. Roychowdhury: Formal analysis, validation, investigation. S. Chattopadhyay: Resources, writing–review and editing. S. Roychoudhury: Conceptualization, supervision, funding acquisition, project administration, writing–review and editing.

We thank Varda Rotter (Weizmann Institute of Science, Rehovot, Israel) and Tanya Das (Bose Institute, Kolkata, India) for kindly providing us with the H1299/EV, H1299/R175H, and basic H1299 cell lines, respectively. We thank Bert Vogelstein (Johns Hopkins University) for kindly providing us with mutant p53 expression plasmids. We thank Nitai. P. Bhattacharjee (Saha Institute of Nuclear Physics) and Dr. Somsubhra Nath (Saroj Gupta Cancer Center and Research Institute) for providing us the necessary plasmids. We thank Tanmoy Dalui and Sounak Bhattacharya (CSIR-Indian Institute of Chemical Biology, India) for their assistance in flow cytometry and confocal microscopy, respectively. The work was supported by J.C. Bose National Fellowship grant, JCB/2017/000005 awarded to S. Roychoudhury.

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.