Research ArticleCancer

Regulation of autophagy, NF-κB signaling, and cell viability by miR-124 in KRAS mutant mesenchymal-like NSCLC cells

See allHide authors and affiliations

Sci. Signal.  12 Sep 2017:
Vol. 10, Issue 496, eaam6291
DOI: 10.1126/scisignal.aam6291

Subtype selectivity in cancer

Activating mutations in the protein KRAS are common in non–small cell lung cancers (NSCLCs). However, tumors of the mesenchymal subtype of KRAS mutant NSCLC are not responsive to RAS pathway inhibitors. By comparing tumor samples and cell lines of epithelial and mesenchymal NSCLCs, Mehta et al. found that loss of the microRNA miR-124 in mesenchymal tumor cells enables increased abundance of autophagy and related proteins, which promoted cell survival. Introducing miR-124 or inhibiting autophagy in the mesenchymal tumor cells triggered their apoptosis. These findings identify alternative molecular targets for potentially treating this subtype of NSCLC.


KRAS mutant non–small cell lung cancer (NSCLC) may be classified into epithelial or mesenchymal subtypes. Despite having the same “driver” mutation, mesenchymal NSCLCs are less responsive than are epithelial NSCLCs to inhibition of the RAS pathway. Identifying alternative networks that promote survival specifically in mesenchymal NSCLC may lead to more effective treatments for this subtype. Through their numerous targets in cellular signaling pathways, noncoding microRNAs (miRNAs) often function as tumor suppressors or oncogenes. In particular, some miRNAs regulate the epithelial-mesenchymal transition (EMT). We derived an EMT-related miRNA signature by profiling the abundance of miRNAs in a panel of epithelial (KE) or mesenchymal (KM) KRAS mutant NSCLC cell lines. This signature revealed a number of suppressed miRNAs in KM cell lines, including members of the miR-200 family, which can suppress tumor progression by inhibiting EMT. Reconstituting KM cells with one of these miRNAs, miR-124, disrupted autophagy and decreased cell survival by reducing the abundance of p62, which is both an adaptor for selective autophagy and a regulator of the transcription factor nuclear factor κB (NF-κB). Suppression of p62 by miR-124 correlated with reduced abundance of the autophagy activator beclin 1 (BECN1), the ubiquitin ligase TRAF6, and the NF-κB subunit RELA/p65. The abundance of miR-124 inversely correlated with the expression of BECN1 and TRAF6 in patient NSCLC samples. These findings reveal how the loss of miR-124 promotes cell survival networks in the aggressive mesenchymal subtype of KRAS mutant NSCLC, which might lead to improved subtype-selective therapeutic strategies for patients.


Non–small cell lung cancers (NSCLCs) are aggressive and difficult to treat if diagnosed at advanced stages. A critical barrier to identifying effective therapeutics for NSCLCs, as well as many other related tumor types, is tumor cell plasticity that can lead to inter- and intratumoral molecular heterogeneity (1). Global gene expression profiling of NSCLCs has revealed a number of heterogeneous and contrasting molecular subtypes that associate with distinct histopathological characteristics and therapeutic vulnerabilities (2, 3). Oncogenic KRAS mutations are found in 20 to 30% of NSCLC cases but do not stratify into a uniform molecular or histological subtype. KRAS mutant NSCLC cell lines display highly variable KRAS-dependent transcriptional networks and cell survival pathways (4). KRAS mutant NSCLC cells with epithelial morphology tend to be more dependent on KRAS for survival in contrast to KRAS mutant mesenchymal cells, which tend to have diminished KRAS dependency. These contrasting subtypes are subsequently designated KE (epithelial) and KM (mesenchymal). The underlying molecular mechanisms that contribute to reduced KRAS dependency in the KM subtype remain unclear. It is possible that co-occurring mutations in STK11/LKB1 and KEAP1 along with KRAS may confer reduced oncogenic KRAS dependency and provoke altered sensitivity to therapeutic agents (5, 6). Many KM cell lines harbor inactivating STK11 and/or KEAP1 mutations. Epithelial-mesenchymal transition (EMT) is also associated with reduced sensitivity to some targeted therapeutic agents in vitro and drug resistance in vivo (1, 7).

We previously compared microarray-based gene expression data for KE and KM cell lines and derived a KE/KM gene expression signature that associates strongly with oncogenic KRAS dependency and is enriched with EMT markers such as CDH1/E-cadherin (4). Preliminary association studies of this signature using the Oncomine Concepts Map (8) revealed a significant overlap with predicted miRNA target genes for miR-205 and miR-34. This provided a rationale to determine the functional significance of altered miRNA expression patterns in the context of epithelial/mesenchymal subtype of KRAS mutant NSCLC cell lines. The role of noncoding microRNAs (miRNAs) in the pathogenesis of lung cancer is highlighted by gene ablation studies of the processing enzyme DICER1 in a genetically engineered mouse model of mutant KRAS-driven lung cancer, which causes increased aggressiveness and size of tumors (9). In some cases, the function of miRNAs can be oncogenic, as highlighted by the role of miR-21 in KRAS-driven lung cancer progression (10).

In human cancers, miRNAs are globally down-regulated, suggesting general tumor suppressor functionality (11). A number of individual miRNAs have tumor suppressor properties, including the let-7 family, which suppress MYC and KRAS oncogene expression (12, 13). Also, the p53-regulated miR-34 family modulates some tumor suppressor functions including DNA damage response pathways (14). Last, the miR-200/205 family regulates EMT by targeting the ZEB1 transcriptional corepressor (15). Our goal was to identify miRNA gene regulatory networks that modulate epithelial differentiation and cell viability in KM subtype cells. Here, we derived a putative KE-KM subtype miRNA signature. Subsequently, we characterized the functional role of KE-correlated miRNAs in modulating EMT, autophagy, and cell death in KM subtype cells, with an overarching goal of identifying context-dependent, subtype-selective cell survival pathways that could be exploited for therapeutic benefit in the future.


KE versus KM subtypes are distinguished by a miRNA expression signature

Preliminary computational analysis of a previously derived KE/KM gene expression signature (4) using the Oncomine Concepts Map showed significant overlap with predicted target genes for miR-205 (odds ratio, 4.48; P = 0.002) and miR-34b/c (odds ratio, 3.65; P = 0.002). The role of miR-205 in modulating EMT and the role of miR-34 in the p53 pathway provided rationale to investigate altered miRNA expression and function in the context of KRAS dependency and EMT subtype–associated phenotypes. To that end, we performed quantitative reverse transcription polymerase chain reaction (qRT-PCR) TaqMan Low-Density Array (TLDA) assays to determine expression levels of 380 well-annotated mature miRNAs in a panel of six representative KRAS mutant NSCLC cell lines (table S1 and Fig. 1). First, to identify differentially expressed miRNAs in KE versus KM subtype cells, we performed Student’s two-sided t tests on normalized average ΔCt values (threshold cycle method) assuming unequal variances (16) for each respective miRNA in each subtype. We also calculated log2 fold differences comparing average ΔCt values for each subtype. We found that KE-correlated miRNAs were moderately higher in total number than KM-correlated miRNAs (160 versus 120, respectively) (Fig. 1A).

Fig. 1 KRAS oncogene dependency and epithelial-mesenchymal phenotypes are associated with a miRNA expression signature in NSCLC cell lines.

(A) Volcano plot of differential miRNA expression based on quantitative PCR (qPCR)–based profiling in six representative KRAS mutant NSCLC cell lines. Fold expression differences were calculated by comparing average expression levels in three KRAS-dependent/epithelial (KE-correlated) cell lines to three KRAS-independent/mesenchymal (KM-correlated) cell lines. Student’s t test–derived P values are plotted on a −log scale on the horizontal axis (two-sided test with unequal variances). (B) Heat map representation of differential miRNA expression levels (P < 0.1) in KE versus KM subtype cell lines, as used in the volcano plot in (A). Orange and purple indicate high and low median-centered expression levels, respectively. Cell lines and miRNAs are clustered by similarity as determined by Euclidean distance. (C) Western blotting for abundance of epithelial markers (P-cadherin, E-cadherin, and β-catenin) and vimentin as a mesenchymal marker in a panel of KM (red text) and KE (green text) NSCLC cell lines. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) serves as a gel loading control for (D) and (E). (D) Western blot showing effects of miR-200c and miR-205 reconstitution in H460 KM cells at two concentrations of synthetic mimic oligonucleotides (10 and 50 nM) on the abundance of epithelial and mesenchymal markers. Data are representative of two independent experiments. (E) Confocal immunofluorescence micrographs of H460 cells transfected with control or miR-200c mimics. Cells are shown costained with E-cadherin (green channel) and vimentin (red channel). Scale bars, 25 μM.

To select differentially expressed miRNAs, we set a slightly liberal threshold (t test–derived P < 0.1), given the low sample size (n = 6) and our overarching goal of identifying miRNAs for subsequent functional studies of tumor suppressor activity. This led to the identification of 30 miRNAs that met the threshold criteria, 9 of which are correlated with KM cell lines compared to 21 KE-correlated miRNAs. To visualize coclustered miRNA expression patterns, hierarchical clustering was performed to generate a heat map of differentially expressed miRNAs (Fig. 1B). Notably, miR-200c and miR-141, which are expressed as a tandem primary RNA transcript from a locus on chromosome 12p13.31, were coclustered, indicating concordant expression patterns. We identified miR-205 as KE-correlated, validating the initial Oncomine analysis. We then determined whether the expression of a select panel of KE-associated miRNAs was regulated by oncogenic KRAS. In H358 KE cells, small interfering RNA (siRNA)–mediated depletion of KRAS reduced the abundance of specific KE-correlated miRNAs (miR-124, miR-200c, miR-141, miR-199, and miR-625) (fig. S1A). However, in a different KE cell line (H2009), only miR-124 and miR-200c levels were reduced after KRAS depletion, whereas the others were up-regulated (fig. S1A). Therefore, one could surmise that KRAS promotes the proliferative phenotype in epithelial cells at least in part by supporting the expression of select miRNAs (such as miR-124 and miR-200c). How these miRNAs are suppressed in the KRAS mutant mesenchymal cells and which signaling mechanisms modulate the other EMT-associated miRNAs remain to be determined.

Members of the miR-200 family and miR-205 regulate EMT (15), and expectedly, their increased expression in KE cells and decreased expression in KM cells correlated with the abundance of epithelial and mesenchymal markers, respectively (Fig. 1C). To validate a causal role for miR-205 and miR-200c in regulating EMT in these KRAS mutant NSCLC lines, we performed synthetic miRNA oligonucleotide mimic transfection assays in representative KM cell lines. In H460 cells, reconstitution of miR-205 and miR-200c resulted in repression of an established target, Zeb1 (Fig. 1D). Zeb1 is a C2H2-type zinc-finger protein that functions as a transcriptional repressor of the CDH1 locus, resulting in reduced abundance of epithelial E-cadherin protein, mesenchymal differentiation, and cancer stem cell–like phenotypes (17). Suppression of Zeb1 abundance by miR-200c and miR-205 was associated with increased E-cadherin protein abundance. miR-200c reconstitution caused slightly higher E-cadherin protein induction compared to miR-205 reconstitution. Expression of the mesenchymal marker vimentin was reduced slightly upon reconstitution of high concentrations of miR-200c (Fig. 1D). Compared to H460 cells, A549 cells had a greater amount of E-cadherin under control conditions, which increased moderately with miR-200c or miR-205 reconstitution and was associated with decreased abundance of Zeb1 protein (fig. S1C). KE-correlated miRNAs did not broadly alter E-cadherin abundance in H460 KM cells; only miR-200c and miR-141 caused appreciable increases in E-cadherin protein abundance (fig. S1D). We also observed a change in E-cadherin subcellular localization from intracellular punctate structures in control cells to more pronounced localization at cell-cell contacts in miR-200c–reconstituted cells (Fig. 1E and fig. S1B). This phenotype is very similar to direct Zeb1 ablation by shRNA in KM cells (4). Finally, we validated a positive correlation (r = 0.476; P = 1.12 × 10−6) between expression levels of MIR200C and MIR124 primary RNA transcripts in a microarray data set derived from primary tumors in NSCLC patients [National Institutes of Health (NIH)/Gene Expression Omnibus (GEO) accession number GSE43458] (18). In summary, a comparison of KE and KM miRNA expression profiles reveals a distinct signature, strongly associated with EMT-related molecular characteristics.

A subset of KE-correlated miRNAs causes reduced cell viability in KM cell lines

We selected a panel of 22 differentially expressed KE-correlated miRNAs based on highest fold difference from the analyses described in Fig. 1. To determine potential tumor suppressor activities of identified miRNAs, we performed comprehensive functional cell viability assays after reconstitution of KE-correlated miRNA mimics in a test set of six KM cell lines (4). We performed synthetic miRNA mimic oligonucleotide transfections and assessed cell viability effects 72 hours later (19). We noted significantly reduced cell viability after reconstitution of a number of miRNAs compared to control transfection, which varied across the cell lines tested (Fig. 2A). From this analysis, we qualitatively determined that miR-124 had generally negative effects on cell viability in all the cell lines tested. SW900 and H460 cells were generally more sensitive to the cytotoxic effects of KE-correlated miRNA reconstitutions than the other cell lines tested.

Fig. 2 Reconstitution of specific KE-correlated miRNAs causes reduced cell viability in KM cell lines.

(A) Alamar Blue–based cell proliferation/viability assays measuring relative viability of indicated cell lines after transfection with 10 nM synthetic miRNA mimic oligonucleotides compared to NC-transfected cells 72 hours after transfection. Data are means of three replicates ± SEM. Viability effects of miR-124 are highlighted with a red asterisk. (B) Heat map representing normalized median-centered viability data from (A). Cell lines and miRNAs are clustered by a Euclidean distance similarity metric. miR-124 is highlighted in red text.

Next, we performed hierarchical clustering analysis of normalized cell viability data and generated a heat map (Fig. 2B). Cell lines could be coclustered by similarity of responses to miRNA reconstitution, with H1792 and SK-LU-1 cells showing the most concordant sensitivity patterns. H2030 and LU65 cells also showed a similar degree of concordance. We noted some cell line–specific effects, including sensitivity to miR-210 in LU65 and SW900 cells. Notably, we found that miR-205 caused minimal cell viability effects in all the cell lines tested, despite being the most significantly overexpressed miRNA in the KE subtype. In addition, miR-146b-5p and miR-125b failed to induce cell viability changes in the KM cell lines. We identified three miRNAs that suppressed cell viability in all cell lines tested: miR-124, miR-518d-3p, and miR-625. Of these, miR-124 had the most potent effects on reducing cell viability in all KM cell lines.

miR-124 reconstitution induces apoptosis and suppresses clonogenic growth of KM cell lines

To further elucidate cytotoxic mechanisms caused by reconstitution of KE-correlated miRNAs in KM cell lines, we used a luminescent caspase-3–Glo reporter assay to measure effects on apoptosis. Increased luminescence in this assay is proportional to increased effector caspase-3 activation. We tested three miRNAs (miR-124, miR-199a, and miR-625), which all caused increased caspase-3 activation in H460 KM cells (Fig. 3A). Of these, miR-124 caused the strongest caspase-3 activation compared to miR-199a and miR-625. In contrast, caspase-3 activity was weakly induced or absent in KRAS wild-type H322 NSCLC cells and MCF7 breast cancer cells (Fig. 3A).

Fig. 3 miR-124 reconstitution causes cytotoxic effects in KM cell lines.

(A) Luciferase-based caspase-3 activation assays after reconstitution of indicated miRNA mimics in H460 KM cells, H322 KRAS-WT NSCLC cells, and MCF7 KRAS-WT breast cancer cells. (B) Three-dimensional (3D) Matrigel assays of A549 KM cell growth after lentiviral expression of control luciferase or primary MIR124 (pri-MIR124). Fixed cells are shown stained with phalloidin to visualize actin filaments (red signal) and 4′,6-diamidino-2-phenylindole (DAPI) to visualize nuclear DNA (blue signal). Arrow indicates an apoptotic cell with fragmented DNA and low actin content. Scale bars, 25 μm. (C) Automated image-based quantitation of colony size and number from 3D Matrigel assays shown in (B). Colony size is in arbitrary pixel units. Bars represent population means ± SEM. (D) Bright-field microscopy images of NC- or miRNA-transfected cells. Arrows indicate large intracellular endosomal, possibly vacuolar structures. Scale bars, 25 μm. Data are representative of three independent experiments.

Given that, among the KE signature miRNAs tested, miR-124 had the greatest cytotoxic effects in KM cells, we focused our subsequent efforts on investigating miR-124. To determine whether the cytotoxic effects of miR-124 conferred reduced clonogenic growth, we transduced A549 KM cells with a lentiviral vector encoding the primary form of miR-124 (pri-MIR124). The MCF7 breast cancer cell line, which again have wild-type KRAS, was transduced alongside as a control. We then performed culture assays of stable cell lines in Matrigel. A549 cells expressing miR-124, when grown in Matrigel, displayed signs of cytotoxicity manifesting as reduced size and number of colonies (Fig. 3, B and C). Furthermore, nuclear DNA fragmentation indicative of cell death could be observed in many of the miR-124–expressing cells (Fig. 3B). In contrast to A549 cells, there were no changes in colony size and numbers in MCF7 cells expressing miR-124 (fig. S2, A and B). Reconstitution of miR-124 in H460 cells caused increased caspase-3 cleavage in 3D cultures, as determined by immunofluorescence microscopy (fig. S2C), and reduced clonogenic growth in monolayer cultures (fig. S2D). Finally, miR-124–mediated cytotoxicity was associated with distinct morphological features, including the appearance of dense intracellular vacuolar structures (Fig. 3D). This indicated a possible perturbation of lysosomal function, perhaps as a consequence of altered regulation of macroautophagy, referred to hereafter as autophagy. Constitutive activation of the cytoprotective autophagy pathway drives survival in subsets of KRAS mutant mesenchymal-like cancer cell lines (20). This prompted our interest in investigating a context-dependent role for miR-124 in perturbation of autophagy pathway activation in KM cells.

miR-124 reconstitution causes autolysosomal maturation defects in KM cells

A number of assays and guidelines have been developed and established to study activation of autophagy and so-called “flux” through the pathway (21). To monitor autophagy flux in cells reconstituted with miR-124, we used a common dual fluorescent LC3 reporter, which is expressed as an N-terminal fusion of mCherry and green fluorescent protein (GFP) to human LC3B (Fig. 4A) (22). Activation or inhibition of the autophagy pathway is determined by quantitating LC3 aggregates by live-cell fluorescence microscopy. Overlap of red and green signals from mCherry and GFP, respectively, will appear as yellow aggregates upon induction of phagophore formation and autophagosome (AP) accumulation. If the autophagy pathway is in flux, then these yellow aggregates will be transient as APs fuse with acidic lysosomes to form autolysosomes (AL) in which the low pH quenches GFP fluorescence, resulting in rapid accumulation of dense red aggregates (Fig. 4A).

Fig. 4 miR-124 reconstitution blocks autolysosome accumulation in KM cells.

(A) Schematic overview of the mCherry-GFP-LC3 reporter. (B) Live-cell imaging of H460 KM cells expressing the mCherry-GFP-LC3 reporter and transfected with miR-124 oligonucleotide dsRNA and/or treated with 50 μM chloroquine. Arrows indicate LC3 aggregates that are GFP and mCherry double-positive, indicative of phagophore or autophagosome accumulation. Time is shown in hours. Scale bars, 25 μm. (C) Quantitation of GFP/mCherry double-positive puncta, indicating phagophores/autophagosomes (AP) using image analysis software (BioTek Gen5). (D) Quantitation of mCherry-positive autolysosomes (AL). (E) Quantitation of AL/AP ratio. Data are representative of three independent experiments.

We introduced mCherry-GFP-LC3 (mCG-LC3) into H460 cells by stable retroviral transduction to monitor LC3 dynamics in live cells in real time. These cells were transfected with miR-124 mimics either alone or in combination with the lysosomotropic agent hydroxychloroquine (chloroquine), which causes rapid deacidification of lysosomes and a subsequent block in autophagosome/lysosome fusion. Live-cell imaging of mCG-LC3 aggregates over 72 hours after the addition of agents to H460 cells revealed dynamic alterations in autophagy flux induced by miR-124 and/or chloroquine compared to control cells (Fig. 4B and fig. S3A). In control cells, some green/yellow AP puncta were seen transiently. These AP puncta dissipated over several hours, and, as cell density increased over 48 to 50 hours, red AL aggregates began to form and accumulate (Fig. 4B, first row, fig. S3A, and movie S1). In miR-124–reconstituted cells, overall GFP fluorescence was weak, but yellow AP aggregates began to form around 22 hours after transfection (Fig. 4B, second row, and movie S2). Notably, the AP aggregates appeared as single large aggregates per cell. These aggregates persisted for an additional 8 to 20 hours and eventually dissipated. Cells with accumulated AP aggregates underwent cell death at 48 to 50 hours after transfection. Chloroquine treatment caused formation and accumulation of dense AP aggregates, as expected based on previous reports (Fig. 4B, third row, and movie S3) (23). Chloroquine-induced AP aggregates persisted for 50 hours, with some undergoing mild AL maturation. The AP aggregates in chloroquine-treated cells were more numerous and smaller per cell, in contrast to miR-124–reconstituted cells. Chloroquine treatment alone had mild cytotoxic effects in H460 cells. When combined with miR-124 reconstitution, chloroquine promoted synergistic cell death (Fig. 4B and movie S4).

We performed software-based image cytometry and set thresholds to identify and count LC3 aggregates based on size and fluorescence intensity. AP and AL puncta were identified by setting appropriate empirically derived thresholds for mCherry and GFP fluorescence (fig. S3B). To increase the accuracy of puncta identification and quantification, “gating” was performed to isolate high green and high red fluorescent AP puncta in contrast to low green and high red fluorescent AL puncta. In control cells, AP counts remained at a steady-state low level (Fig. 4C). miR-124 reconstitution caused a mild increase in AP counts, whereas chloroquine treatment caused a significant steady time-dependent increase in AP accumulation, as expected, because chloroquine blocks AP to AL maturation. Chloroquine-treated cells showed accumulation of multiple APs per cell. In contrast, in miR-124–reconstituted cells, chloroquine-induced AP accumulation was severely blunted about 10 hours after transfection. Control cells showed robust AL accumulation over time, consistent with constitutive basal activation of autophagic flux in these cells. Reconstitution of miR-124 caused a noticeable effect on blocking AL accumulation (Fig. 4D). This latter effect on blocking AL formation by miR-124 was comparable to the effects of chloroquine. This effect of miR-124 on blocking AL accumulation was additive when combined with chloroquine (Fig. 4D). Calculation of AL/AP ratios further demonstrated relatively high AL accumulation in control cells, which was severely impaired in chloroquine-treated cells and moderately impaired in miR-124–transfected cells (Fig. 4E). In summary, miR-124 causes defects in AP maturation and AL accumulation in H460 KM cells. mir-124 combined with chloroquine causes enhanced blockade of AP and AL accumulation, which is accompanied by additive cytotoxic effects.

miR-124 suppresses p62 and beclin 1 abundance selectively in KM cells

To determine mechanisms underlying miR-124–mediated regulation of the autophagy pathway, we analyzed the abundance of key autophagy pathway protein components after miR-124 reconstitution, including SQSTM1/p62 and LC3-I/II, which are well-established markers of pathway activation (24). First, we validated that reconstitution of miR-124 induced apoptosis in all KM cell lines tested, as assessed by increased levels of cleaved poly(ADP-ribose) polymerase (PARP). Abundance of beclin 1, which initiates phagophore formation and promotes autolysosome maturation (25, 26), was suppressed in all four KM cell lines by reconstitution of miR-124 (Fig. 5A). We used the TargetScan web portal to identify a high-probability predicted target site for miR-124 in the 3′ untranslated region (3′UTR) of beclin1/BECN1 mRNA (table S2). We confirmed that miR-124 directly targets the 3′UTR of BECN1 using an engineered luciferase reporter construct whose luminescence was reduced 40% by miR-124 reconstitution compared to control transfected cells, whereas miR-200c reconstitution failed to cause significant effects on reporter activity (fig. S4A). Furthermore, exogenous expression of beclin 1 in A549 cells caused a mild 10% rescue of miR-124–dependent cell proliferation defects (fig. S4B). Beclin 1 expression also partially rescued effects on PARP cleavage and LC3-II accumulation after miR-124 reconstitution (fig. S4, C and D). Finally, BECN1 and MIR124 expression were inversely correlated in primary NSCLC tumor samples from patients (r = −0.44; P = 4.35 × 10−5; fig. S4E).

Fig. 5 miR-124 reconstitution represses beclin 1 and p62 expression selectively in KM cells.

(A) Western blot showing protein expression levels of cleaved PARP, indicative of apoptosis, and autophagy pathway components beclin 1, ATG16L1 α/β splice isoforms, p62, and LC3-I/II after transfection with miR-124 in two KRAS wild-type (WT) cell lines (MCF7 and H322) and four KM cell lines. GAPDH is a gel loading control. (B) Western blot analysis of effects of miR-124 transfection minus/plus chloroquine treatment on expression levels of p62, NBR1, and LC3-II in H460 cells. (C) Western blot showing indicated protein levels after miR-124 reconstitution in H460 KM cells compared to H358 KE cells. Cleaved (Δ) PARP indicates apoptosis. E-cadherin and vimentin indicate epithelial versus mesenchymal differentiation. (D) Western blot analysis of cleaved PARP, E-cadherin, vimentin, and p62 in H358 KE cells compared to a stable mesenchymal derivative cell line H358M after miR-124 transfection. GAPDH serves as a gel loading control for (A) to (C) and (F). (E) Quantitation of endogenous mature miR-124 levels by qPCR in H358M cells relative to parental H358 cells. **P < 0.005. (F) Schematic representation of the human p62/SQSTM1 3′UTR showing the miR-124 binding site seed sequence and point mutants (1 and 2) that disrupt the miR-124/ SQSTM1 interaction. Reporter activity of WT or mutant SQSTM1 3′UTRs after miR-124 transfection. Data are means of three replicates + SEM. ***P < 0.0005; n.s., not significant. (G) Rescue of miR-124–induced cell viability defects with exogenous GFP-tagged p62 expression in A549 and SW900 KM cell lines. Data are means of three replicates ± SEM. **P < 0.005. All data are representative of three independent experiments.

Next, we found that miR-124 induced altered expression of ATG16L1, with increased abundance of the shorter α protein variant, which has previously been associated with impaired AP maturation (Fig. 5A) (27). In addition, miR-124 caused robust suppression of the selective autophagy adaptor protein p62. These effects of miR-124 were associated with increased expression of LC3-II, indicating a block in AL function, because LC3-II degradation and turnover are mediated by AL activity. Furthermore, immunofluorescence microscopy of p62 expression showed that miR-124 caused a significant reduction in the number and size of p62 aggregates in H460 cells (fig. S5, A and B). TargetScan analysis of the SQSTM1 3′UTR also showed a predicted miR-124 target seed sequence (table S2). The effects of miR-124 on the autophagy pathway correlated with reduced abundance of phosphorylated S6 at Ser235/236, indicating perturbation of mTOR/S6K or ERK (extracellular signal–regulated kinase)/RSK (ribosomal protein S6 kinase) activity (fig. S5C). Reduced SQSTM1 mRNA abundance upon miR-124 reconstitution was validated by qPCR (fig. S5D). Reduced p62 abundance is usually associated with an increased rather than a decreased autophagic flux. We suggest that miR-124 blocks AP formation and maturation via suppression of beclin 1, concomitant with suppression of p62 abundance.

To provide further supportive evidence for a role of miR-124 in blocking autolysosome formation and function, we analyzed expression levels of another selective autophagy adaptor protein, NBR1, which is involved in selective autophagic degradation of some ubiquitinated cargo proteins (28). Reconstitution of miR-124 caused increased NBR1 protein abundance in a panel of KM cell lines. This was associated with increased LC3-II levels, indicative of reduced AL function, consistent with the aforementioned data showing that miR-124 causes a block in AL accumulation (Fig. 4D). Subsequently, we investigated the effects of combining chloroquine with miR-124 reconstitution on NBR1 protein levels in H460 cells (Fig. 5B). Chloroquine treatment caused increased p62 protein levels, consistent with a block in AL function. This effect was abrogated by combined reconstitution of miR-124 with chloroquine treatment. Under the same conditions, NBR1 abundance was additively increased. Surprisingly, we found that chloroquine treatment alone did not promote an increase in the amount of NBR1. The reasons for this are unclear, given that p62 protein abundance was markedly greater in chloroquine-treated cells. Nonetheless, the findings indicate that miR-124 inhibits selective autophagy by blocking autolysosome formation and function.

Given the sensitivity of mesenchymal cell lines to miR-124 reconstitution, we investigated a causal role for EMT in modulating this sensitivity. Whereas miR-124 induces apoptosis in H460 KM cells, as assessed by increased abundance of cleaved PARP, there was no such apoptotic response in the KE H358 cell line (Fig. 5C). miR-124 caused suppression of vimentin abundance in both H460 and H358 cells. On the basis of TargetScan analysis, the VIM mRNA (encoding vimentin) harbors a predicted miR-124 target site in the 3′UTR (table S2). Notably, p62 abundance was decreased only in H460 KM cells but not in H358 KE cells (Fig. 5C). This could be due to additional p62 regulatory mechanisms that are active in KM cells.

To establish a causal role for EMT in mediating sensitivity to miR-124 reconstitution, we used a stable mesenchymal variant of the H358 cell line, designated as H358M (4). In contrast to H358 cells, H358M cells were more sensitive to miR-124 reconstitution as assessed by a more prominent PARP cleavage indicative of apoptosis. Gross morphological cytotoxicity could also be observed in miR-124–reconstituted H358M cells, which was not apparent with the parental H358 cells (fig. S5E). As with H460 KM cells, miR-124 suppressed vimentin and p62 levels in H358M cells (Fig. 5D). The sensitivity of H358M cells to miR-124 was associated with decreased levels of mature miR-124 RNA levels in H358M compared to parental H358 cells (Fig. 5E). Finally, to determine whether miR-124 had cross-tissue effects on p62, we investigated the effects in KRAS mutant mesenchymal pancreatic ductal adenocarcinoma (PDAC) cell lines, KP1N and KP4 (fig. S5F). As with NSCLC KM cells, miR-200c caused increased E-cadherin expression in PDAC KM cells. mir-124 reconstitution caused p62 suppression in PDAC cell lines, similar to the effects in NSCLC cell lines.

In addition to p62, we identified other putative miR-124 target genes, including beclin 1 (see fig. S4) and vimentin, lending uncertainty to which target genes mediate the phenotypic effects of miR-124 in KM cells (table S2). Therefore, we determined whether miR-124 directly suppresses SQSTM1/p62 expression by binding the 3′UTR of p62 mRNA and whether this suppression was responsible for the cytotoxic effects observed in KM cell lines. First, we noted that H460 cells were sensitive to siRNA-mediated depletion of p62 as assessed by PARP cleavage–associated cell death, which was not evident in the H358 KE cell line (fig. S5G). This indicates that p62 drives anti-apoptotic signaling in H460 and, possibly, other KM cell lines. Second, we performed a luciferase-based assay with a lentiviral reporter vector containing the entire 3′UTR of human SQSTM1 mRNA downstream of LUC/luciferase gene driven by an EF1α promoter (Fig. 5F). To functionally validate the predicted seed sequences for miR-124 in the SQSTM1 3′UTR, we tested the effects of miR-124 on reporter activity of either wild-type or mutated miR-124 target site seed sequences (Fig. 5F). Introduction of miR-124 resulted in a 50% reduction in the reporter gene activity of the SQSTM1 3′UTR containing wild-type miR-124–targeting sequences (Fig. 5F). In contrast, two different mutated SQSTM1 3UTR sequences did not respond significantly to miR-124–mediated luciferase suppression. Last, we performed rescue experiments using exogenous expression of GFP-tagged p62 that lacks the endogenous SQSTM1 3′UTR sequence (fig. S5H). Exogenous expression of GFP-p62, which is resistant to suppression by miR-124, rescued the cell viability defects caused by miR-124 reconstitution in SW900 and A549 KM cell lines (Fig. 5G). In summary, we conclude that cytotoxic effects of miR-124 are selective for KM subtype cells via coordinated effects on the autophagy pathway and apoptosis induction, with p62 as a central hub in the miR-124 regulatory network.

NF-κB signaling and cytokine expression are regulated by miR-124

In addition to a role in autophagy, p62 is critical for activation of the nuclear factor–κB (NF-κB) transcriptional complex and downstream effects on inflammatory cytokine signaling (29). We found that expression of the p65 subunit (NFKB1/RELA) of NF-κB was suppressed after miR-124 reconstitution. This was accompanied by reduced abundance of the E3 ubiquitin ligase TRAF6, which mediates IκB kinase (IKK) and subsequent NF-κB activation (Fig. 6A) (30). The TargetScan portal returns TRAF6 and p65/RELA as predicted miR-124 target genes (table S2). Furthermore, TRAF6 and MIR124 mRNA abundance were inversely correlated in primary NSCLC samples (r = −0.62; P = 8.45 × 10−10; fig. S6A). We verified that miR-124 caused reduced NF-κB signaling by using a κB binding element–firefly luciferase construct. In A549 and SW900 KM cells, we observed twofold to threefold decreases in NF-κB transcriptional activity after miR-124 reconstitution (Fig. 6B).

Fig. 6 miR-124 reconstitution induces an altered expression of NF-κB–regulated inflammatory cytokines.

(A) Western blots showing protein expression levels of the p65 (RELA) subunit of NF-κB and TRAF6 after miR-124 reconstitution in the indicated KM cell lines. GAPDH is a gel loading control. (B) NF-κB–dependent luciferase reporter assays showing arbitrary luminescence counts after control NC or miR-124 transfection in A549 and SW900 KM cells. ***P < 0.0005, **P < 0.005. (C) Antibody-based proteomics array showing expression levels of inflammatory cytokines in conditioned media from NC or miR-124–transfected H460 KM cells. Most significantly altered cytokines are highlighted in boxes. (D) Relative quantitation of pixel density from arrays shown in (C), comparing miR-124– to NC-transfected cells. Data are representative of two independent experiments.

Given the role of NF-κB in regulating inflammatory cytokine signaling, we hypothesized that cytokine expression levels might be altered in miR-124–reconstituted KM cells. Conditioned media from miR-124–reconstituted KM cells contained altered levels of key cytokines compared to nontarget control (NC)–transfected cells (Fig. 6C). Quantitation of cytokine levels revealed that proinflammatory cytokines, such as tumor necrosis factor–α (TNFα) and interleukin-1α (IL-1α), were increased upon miR-124 reconstitution in KM cells. Other induced factors included transforming growth factor–β, IL-12, RANTES, IL-11, interferon-γ, and IL-6 (Fig. 6C). Conversely, GMCSF (granulocyte-macrophage colony-stimulating factor), IL-4, IL-10, and soluble TNF decoy receptor (sTNFRI/II) were decreased upon miR-124 reconstitution. We verified that alterations in cytokine expression upon miR-124 reconstitution occurred because of changes in mRNA levels of key cytokines as assessed by qRT-PCR in the miR-124–reconstituted KM cells. We noted strong mRNA induction of IL1B, IL11, and TNF, but surprisingly saw reduced IL6 expression (fig. S6B). In summary, miR-124 reconstitution promotes suppression of TRAF6–NF-κB signaling and associated perturbations in pro- and anti-inflammatory cytokine expression levels.


Here, a KE/KM subtype miRNA signature was derived by differential miRNA expression profiling. The signature reveals EMT regulatory miRNAs, such as miR-205 and miR-200c, and a number of cytotoxic miRNAs such as miR-124, which are expressed at significantly lower levels in KM cell lines compared to KE cell lines. We demonstrate that miR-124 negatively regulates cell viability in several KM cell lines upon functional reconstitution. Induction of cell death upon miR-124 reconstitution is associated with the formation of dense vacuolar structures, suggesting perturbations in lysosomal turnover and possible defects in autophagy. Autophagy is constitutively activated in subsets of KRAS mutant cancers and cell lines in large part via transcriptional mechanisms (31). This is associated with maintenance of cell viability, suggesting a possible therapeutic vulnerability in KRAS mutant cancers (32). Our findings indicate that KRAS mutant cells with mesenchymal differentiation exhibit constitutive flux through the autophagy pathway and are particularly sensitive to perturbation of this flux, for example, via reconstitution of miR-124.

Selective autophagy is promoted by beclin 1, and a specific cargo is targeted for degradation by the adapter protein sequestosome1/p62 (33, 34). We conclude that suppression of beclin 1 and p62 by miR-124 is associated with defective AL maturation and accumulation. The defect in AL formation caused by miR-124 is supported by three independent experimental observations. First, miR-124 reconstitution prevents AL accumulation, as demonstrated by live-cell imaging of the fluorescent mCherry-GFP-LC3 reporter. Second, LC3-II protein levels accumulate upon miR-124 reconstitution, indicating a defect in AL function, which normally promotes LC3 degradation and turnover. Last, levels of another selective autophagy adaptor NBR1 accumulate after miR-124 reconstitution. NBR1 is critical for turnover of a number of ubiquitinated target proteins via autolysosomal degradation (28).

Concomitant with perturbation of AL accumulation, suppression of p62 leads to cell death in KM cell lines. This observation is concordant with previous studies that demonstrate a role for p62 in mutant RAS-driven tumor maintenance (35). The mechanisms by which p62 promotes the survival of KM cell lines is complex and may involve coordinated control of selective autophagy and activation of the NF-κB pathway. P62 binds to cargo targeted for degradation via the selective autophagy pathway, which includes ubiquitinated proteins and damaged organelles (24, 34, 36). We also implicate beclin 1 as a key miR-124 target. We hypothesize that miR-124–dependent suppression of beclin 1 causes defects in the formation of mature ALs, which partially contributes to cytotoxic effects. Our findings support the notion that autophagy activation in KRAS mutant mesenchymal cancer cells plays a protumorigenic role. These findings contrast with studies indicating tumor suppressor roles for the autophagy pathway, as demonstrated by a tumor predisposition phenotype in mice with Becn1 heterozygous loss (37). Context and molecular subtype dependence of oncogenic versus tumor suppressor roles for the autophagy pathway is a contentious issue and will need to be reconciled for the design and clinical use of autophagy inhibitors.

Our findings on the role of miR-124 in autophagy regulation are reinforced by previous studies demonstrating significant enrichment for miR-124–predicted target sites in a number of autophagy-lysosomal pathway genes (38). In addition to targeting of p62, as well as beclin 1, we found that miR-124 targets vimentin expression, which is an established mesenchymal cell marker protein. Studies show that vimentin can regulate autophagy by tethering an inhibitory AKT-phosphorylated beclin 1 in complex with 14-3-3 protein (39). Thus, miR-124 inhibits multiple steps in the autophagy pathway. Of all the putative miR-124 targets, our results strongly implicate SQSTM1/p62 as the most critical gene for the maintenance of KM cell survival, as demonstrated by rescue experiments.

Reduced p62 levels are commonly associated with increased autophagy flux, which promotes p62 autolysosomal degradation (33). We found that miR-124–dependent suppression of p62 is associated with increased NBR1 and LC3-II levels and a decrease in AL accumulation, indicative of a block in autophagy. How can these findings be reconciled? We hypothesize that miR-124 disrupts the formation and enclosure of autophagosomes, as occurs after beclin 1 or Vps34/class III PI3K gene ablation (40). When miR-124 is expressed, p62 levels are decreased perhaps via destabilization of SQSTM1/p62 mRNA. Beclin 1 also regulates AL maturation via sequestration of the negative autophagy regulator Rubicon (26). Therefore, miR-124–dependent suppression of beclin 1 could cause AL maturation defects, leading to increased LC3-II and NBR1 levels. The miR-124–dependent block in autophagy is seen selectively in KM cell lines, because miR-124 introduction into KRAS wild-type cells and H358 KE cells fails to induce increased LC3-II levels. The mechanistic basis for selective p62 suppression and dependency in KM subtype cells remains unclear and will be interesting to explore in future studies.

The effects of miR-124–mediated suppression of p62 on cell viability are likely to involve disruption of multiple pathways, including negative regulation of autophagic flux. A recent study indicates that loss of cell viability in A549 and H460 KM cell lines can be induced by combined disruption of autophagy and the oxidative pentose phosphate pathway. This latter pathway could be subject to perturbation by miR-124 as a mechanism underlying cell death induction (41). Furthermore, loss of p62 function has been linked to metabolic disruption in some contexts (42). Therefore, metabolic reprogramming could be a mechanism to explain the cytotoxic effects of miR-124 reconstitution in KM cells, which will be an interesting avenue to explore in future studies.

In addition to functioning as a selective autophagy adaptor, p62 regulates inflammatory signaling via IKK-dependent activation of the NF-κB pathway (35). We show that miR-124 reconstitution promotes NFKB1/p65 suppression, leading to reduced NF-κB transcriptional activity and altered expression of key cytokines. This is consistent with a previous study identifying miR-124 as the top negative regulator of NF-κB in a miRNA mimic functional screen (43). Whether this is a direct suppression of the p65 3′UTR by miR-124 remains unclear. Mechanistic associations between activation of autophagy and inflammatory responses have been documented. For example, pathogenic antigen phagocytosis in macrophages is associated with increased LC3 aggregation on APs, which is critical for innate immune responses (23). Furthermore, germ-line ATG16L1 mutations are causally associated with increased intestinal inflammation seen in Crohn’s disease (44). We found that miR-124 reconstitution induces differential expression of the ATG16L1α isoform, which could feed into regulation of inflammatory signaling.

KRAS/TP53 co-occurring mutations, reflective of the KE subtype, correlate with increased proinflammatory signaling, whereas KRAS/STK11 or KRAS/KEAP1 mutations associate with increased anti-inflammatory signaling, for example, via constitutive activation of the Nrf2 antioxidant pathway (5). Thus, activation of anti-inflammatory signaling is generally observed in a subset of KM cells that harbor STK11 and KEAP1 mutations such as A549 and H460 cells. Together, our studies have defined an NSCLC subtype–specific regulatory mechanism that controls the coordinated function of the autophagy and NF-κB signaling pathways, with p62 as a central and critical hub in the network (Fig. 7). From a therapeutic perspective, miRNA reconstitution is currently being tested in preclinical and early-phase clinical trial settings (45). Thus, reconstitution of miR-124 and other KE-correlated miRNAs may prove beneficial for a subtype of NSCLC that has molecular characteristics of KM cell lines.

Fig. 7 A model of miR-124 activity in KRAS mutant mesenchymal (KM) cells.

In KM cells, under basal conditions, the balance between cell survival and cell death is tightly regulated by autophagy and NF-κB signaling pathways. Upon miR-124 reconstitution in KM cells, abundance of SQSTM1/p62, BECN1, TRAF6, and RELA was suppressed, causing inhibition of autophagy flux and perturbation of NF-κB signaling. Ultimately, these coordinated miR-124 effects tip the balance in favor of cell death.


Quantitative profiling of miRNA expression in cell lines

Cells were grown to confluency in 60-mm dishes, trypsinized, and pelleted. Total RNA was extracted from cells using the Qiagen miRNeasy kit. miRNA expression was quantitated by qRT-PCR using TLDAs, as described previously (46). Raw TLDA data are shown in table S1 and are publicly available via the NIH GEO online portal (accession number GSE102298). Expression data were processed in MS Excel to generate a volcano plot and to identify differentially expressed miRNAs in KE versus KM cells. Data were further processed in “R” equipped with the “pheatmap” package to perform hierarchical clustering and to generate a heat map of differentially expressed miRNAs.

qPCR assays for miRNA and mRNA abundance

Abundance of mature miRNAs in cell lines was measured using the miScript qRT-PCR system (Qiagen Inc.). Briefly, cells were lysed in QIAzol reagent after transfection with siKRAS (48 hours after transfection) or under control conditions. Total RNA was extracted using the miRNeasy kit. Total RNA (2 μg) was reverse-transcribed using the miScript HiFlex buffer, to generate cDNA for small noncoding RNA and mRNA. PCR assays were performed on 1:50 diluted cDNA using sequence-specific primers and the Quantitect SYBR Green PCR mix. PCR products were detected using an Applied Biosystems Step One Plus instrument. Threshold cycle (Ct), ∂Ct, and ∂∂Ct were generated with the Step One Plus software. Data were normalized to abundance of endogenous SNORD6 RNA for miRNA quantitation or to GAPDH abundance for mRNA quantitation.

Cell lines, miRNA mimics, and siRNA transfections

Human cancer cell lines (H460, A549, SKLU1, H2030, SW900, H1792, LU65, SW1573, H2009, H322, H358, H358M, H441, and MCF7) were grown in RPMI supplemented with 5% fetal bovine serum (FBS) (heat-inactivated at 50°C), 1 mM sodium pyruvate, and penicillin/streptomycin, as described previously (4, 47). For initial small-scale screening, miRNA mimic and NC dsRNA oligonucleotides were purchased from Dharmacon Inc. in a 96-well arrayed format. Cells were transfected with RNAiMax reagent (Invitrogen) at 10 nM. For subsequent follow-up studies, NC and miRNA mimics were purchased from Qiagen and transfected at 10 or 50 nM for signaling pathway analyses by Western blotting. siRNA (dsRNA) oligonucleotides against SQSTM1 and KRAS were obtained from Integrated DNA Technologies and used at 10 nM final concentration in media.

Cloning of the pri-MIR124 lentiviral vector

Genomic DNA from 293T cells was used as a template to PCR-amplify the human primary MIR124-2 gene sequence (chr 8p12.3) using flanking forward/attB1 and reverse/attB2 Gateway cloning adapter primers (table S3). A two-step Gateway PCR protocol was used to clone the MIR124-2 605 nucleotide sequence into pDONR-223 and then into a Gateway-compatible pWPI lentiviral vector backbone (gift from D. Trono, Addgene plasmid #12254) using BP and LR clonases, respectively. Cloned sequences were verified by automated DNA sequencing. All viral vectors were packaged and generated in 293T cells using previously described methods (4, 48).

UTR reporter plasmid cloning and luciferase assay

A 1.2-kb DNA fragment corresponding to the 3′UTR of SQSTM1, with putative miR-124 binding sites, was PCR-amplified from human 293T cell genomic DNA by PCR using forward and reverse Gateway adapter primers (table S1). Mutant 3′UTR sequences lacking the critical base-pair complementarity within the 8-mer miR-124 seed sequence were generated by site-directed mutagenesis using the QuikChange kit (Agilent Inc.). SQSTM1 3′UTR DNA fragments were cloned into a modified pWPI lentiviral vector containing an upstream firefly luciferase gene driven by an EF-1α promoter. All clones were sequence-verified by DNA sequencing. For luciferase assays, cells were transduced with wild-type or mutant pWPI-SQSTM1 3′UTR lentiviral vectors. A similar approach was used for cloning of the BECN1 3′UTR. LUC-3′UTR–transduced cells were plated in 96-well plates at 5 × 103 cells per well. After 12 hours, these cells were transfected with miR-124 mimics or NCs at 50 nmol/μl with Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer’s recommendations. Two days after transfection, cells were treated with d-luciferin (100 μg/ml) and incubated at 37°C for 5 min, and “glow-based” luminescence was measured with Fluostar Optima plate reader (BMG Labtech) with a 5-s integration time. Relative luciferase activity was normalized to the vector control (pWPI-luciferase vector) and NC miRNA mimic control.

Cell viability and apoptosis assays

Cells were plated in 96-well plates at 5 × 103 cells per well. Cells were transfected with miRNA mimics or NC 18 to 24 hours after plating. Cell viability assays were performed by adding Alamar Blue (resazurin) reagent at 6.5 μg/ml and incubated for 30 min at 37°C. Converted resafurin fluorescence was measured at λExEm 590 nm using a Fluostar Optima plate reader. Cell viability was normalized to control transfected cells. For caspase-3 activity–based apoptosis assays, cells were plated in white 96-well plates and transfected with miRNA mimics, as described above. Caspase-3 activity was measured using the luminescence-based Caspase-Glo 3/7 assay (Promega Inc.) 48 hours after transfection, by directly adding 25 μl of the caspase assay mix to cells. Luminescence was measured using a Fluostar Optima plate reader, with a 10-s integration time. For p62 and beclin 1 rescue experiments, cells were transduced with GFP-p62 (pMXs-puro GFP-p62 was a gift from N. Mizushima, Addgene plasmid #38277) or pLEX-BECN1. Stable polyclonal cell populations were selected in puromycin (1 μg/ml) for 1 week. Stably transduced cells were plated in 96-well plates and transfected with miR-124 or NC at 50 nM. After 72 hours, cell viability was assessed using the Alamar Blue assay as described above.

Clonogenic and Matrigel assays

For clonogenic assays, 5 × 103 cells were plated in six-well plates and transduced with control or miR-124 lentiviral vectors. Transduced cells were plated at 500 cells per 60-mm dish. Clones were grown for 2 weeks, fixed in EM (electron microscopy)–grade 4% formaldehyde, and stained with Giemsa dye according to the manufacturer’s guidelines. For 3D clonogenic growth assays in Matrigel, control or miR-124–transduced cells were plated in eight-well chamber slides (NUNC Inc.) using a bottom layer “on-top” method, as described previously (49). Colonies were allowed to form for 2 weeks with media change every alternate day. Colonies were imaged by bright-field microscopy, and CellProfiler software (50) was used to determine colony counts and size. Colonies were fixed in EM-grade 4% formaldehyde and stained with DAPI and phalloidin. For cleaved caspase detection, colonies were permeabilized with 0.1% Triton X-100 and incubated with cleaved caspase-3 primary antibody (Cell Signaling Inc.) overnight. Anti-rabbit secondary antibody–conjugated to Alexa 594 was used for immunofluorescence detection. Images were acquired with a Cytation 3 imaging system (BioTek Inc.) equipped with 10× objective lens. Raw tif files were processed in Adobe Photoshop to adjust color balance and levels.

Immunofluorescence microscopy

Cells were fixed in 4% EM-grade formaldehyde and permeabilized with 0.1% Triton X-100. Cells were stained with primary antibodies against vimentin, E-cadherin, and p62 overnight at 4°C followed by fluorescent-labeled Alexa 488– or Alexa 594–conjugated goat anti-mouse/anti-rabbit secondary antibodies (Molecular Probes). Spheroids in Matrigel were fixed as described previously (49) and stained with Alexa 594–conjugated phalloidin to stain actin filaments (Thermo Fisher Scientific); DAPI or Hoechst 33256 dyes were used to stain nuclear DNA. Leica SP5 Laser Confocal Microscope equipped with a 63× oil objective (Fig. 3B and fig. S2A) was used for image acquisition. Micrographs shown in Fig. 1E and fig. S1B were captured on an Olympus IX81 Spinning Disk Deconvolution Microscope equipped with a 40× Plan-Apo oil objective lens. Digital images were processed with SlideBook. Images were compiled and further processed with NIH ImageJ, Adobe Illustrator, and Photoshop CS6.

Western blotting and antibodies

Protein lysates were prepared with 1× Laemmlli buffer followed by 15-s pulse sonication. Lysates were normalized for total protein content using bicinchoninic acid assay reagent (Pierce-Thermo). Proteins were separated by SDS–polyacrylamide gel electrophoresis followed by transfer to polyvinylidene difluoride membranes. Enhanced chemiluminescence (ECL) was performed with West-Pico or West-Dura reagents (Pierce-Thermo). The Syngene G-Box XT4 system and Genesys software were used for ECL imaging. The following antibodies were used for Western blotting: cleaved PARP (Cell Signaling, 1:1000), Zeb1 (Santa Cruz, 1:1000), E-cadherin (BD Biosciences, 1:1000), vimentin, phospho-S6, t-S6, beclin 1, Atg16L1, p62/Sqstm1, LC3-I/II, p65/NFkB1, NBR1 (Cell Signaling, 1:1000), and GAPDH (Santa Cruz, 1:5000). All antibodies were diluted in 2% BSA (bovine serum albumin)/TBS-T (tris-buffered saline with Tween 20) solution.

Live-cell imaging of autophagy

Cells stably expressing dual-reporter pBabe-puro mCherry-EGFP-LC3B (22) were plated at 5 × 103 density in a 12-well plate format in phenol red–free RPMI 1640/5% FBS. Twenty-four hours after seeding, cells were transfected with miR-124 or NC oligos. Six hours after transfection, these cells were treated with 50 μM chloroquine. Live cells maintained at 37°C and 5% CO2 were imaged in a Cytation 3 imaging system (BioTek Inc.) equipped with 10× objective lens for 72 hours. Nine fields per well were selected, and images were taken every 20 min using laser autofocusing. GFP or mCherry fluorescence was imaged using light-emitting diode filter cubes with λExEm 469/525 nm and 586/647 nm, respectively. Gen5 imaging software was used to identify and quantitate green and red LC3 puncta by setting empirically derived thresholds. Fluorescence intensity cutoffs were established to specifically count green/red puncta (autophagosomes/APs) or red puncta alone (autolysosomes/ALs).

miRNA isolation and measurement

Total RNA from cells was isolated using the miRNeasy mini kit (Qiagen) according to the manufacturer’s guidelines. Total RNA (1 μg) was transcribed to cDNA by using miScript II reverse transcription kit (Qiagen). Diluted cDNA (50×) was used in miScript primer assays (Qiagen). Relative expression levels of mature miRNA were analyzed by real-time qPCR in a StepOne Plus system (Applied Biosystems Inc.). U6 small nuclear RNA was used as an endogenous small RNA control for miRNA normalization, and relative expression levels of miRNAs were calculated using the 2−ΔΔCt method after normalization (16).

Cytokine expression array analyses

Cells (1 × 106) were plated in RPMI/0.5% FBS in 150-mm dishes. After 24 hours, cells were transfected with 50 nM control or miR-124 mimic. Forty-eight hours after transfection, conditioned media were collected and concentrated 20-fold with 3-kDa centrifugal filter units (Millipore). Concentrated media were applied to human inflammation antibody array C3 (Ray Biotech Inc.) to detect cytokines according to the manufacturer’s guidelines. Antibody binding was visualized and quantitated by ECL with the Syngene G-Box XT4 system/GeneSys software. Data analysis was performed using NIH ImageJ and a dot blot analyzer plug-in. Dot intensity data were normalized and processed in Microsoft Excel.

Statistical analyses

Statistical and computational analyses were performed using “R” or GraphPad Prism. Analyses were verified and validated by C.-T. Liu (Department of Biostatistics, Boston University School of Public Health). For parametric tests on normally distributed data, we performed Student’s two-sided t tests assuming unequal variances. Colony size data in Fig. 3C were compared using the nonparametric Mann-Whitney test. Where indicated, **P < 0.01 and ***P < 0.001. For correlation analyses of gene expression microarray data derived from primary tumors in NSCLC patients (NIH/GEO accession number GSE43458), RMA normalized data were generated in R, and individual gene-level data for MIR124, MIR200C (primary transcripts), BECN1, and TRAF6 were extracted. Pearson correlation coefficients and associated P values for pairwise gene expression comparisons were calculated in R using cor.test. Scatter plots were generated with the ggplot package in R.


Fig. S1. Regulation of mesenchymal-epithelial transition by KE-associated miRNAs.

Fig. S2. Effects of primary MIR-124 lentiviral expression on cellular growth and viability.

Fig. S3. Live-cell imaging of autophagic flux in H460 KM cells.

Fig. S4. BECN1 expression is suppressed by miR-124.

Fig. S5. Repression of p62 and downstream signaling by miR-124.

Fig. S6. Control of inflammatory cytokine expression by miR-124.

Table S1. TLDA raw data.

Table S2. TargetScan miR-124 predicted targets.

Table S3. Primer and oligo sequences.

Movie S1. Live-cell imaging of mCherry-GFP-LC3 in control H460 cells.

Movie S2. Live-cell imaging of mCherry-GFP-LC3 in miR-124–transfected H460 cells.

Movie S3. Live-cell imaging of mCherry-GFP-LC3 in chloroquine-treated H460 cells.

Movie S4. Live-cell imaging of mCherry-GFP-LC3 in miR-124 + chloroquine–treated H460 cells.


Acknowledgments: We thank our colleagues N. Ganem, H. Feng, and R. Flynn and their research teams for discussions and advice on the project. We thank R. Perera (University of California at San Francisco School of Medicine) for the critical evaluation of the manuscript. Funding: This work was supported by NIH/National Cancer Institute R00 CA149169, an American Lung Association Lung Cancer Discovery Award, and an American Cancer Society Institutional Research Grant Pilot Award to A.S. and NIH T32 GM008541 to K.H. A.K.M. and A.S. are partially supported by the Shamim and Ashraf Dahod Breast Cancer Research Center at Boston University School of Medicine. Author contributions: A.K.M., K.H., W.W., M.-T.N., J.W., and A.S. conceived, designed, and executed the experiments and analyzed the data. A.K.M., K.H., C.-T.L., A.S., J.H., J.W., and J.S wrote and reviewed the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Raw TLDA array data have been uploaded to the NIH GEO web portal (accession number GSE102298).
View Abstract

Navigate This Article