Background
Lung cancer is the leading cause of cancer-related death worldwide with a mean 5-year survival rate of less than 15% [
1]. Platinum-based therapy is the standard of care for patients with metastatic non-small cell lung cancer (NSCLC), the most common subtype of lung cancer [
2]. The introduction of targeted therapy using tyrosine kinase inhibitors (TKI) increased overall survival of patients with metastatic NSCLC harboring activating mutations in the epidermal growth factor receptor (EGFR) compared with standard cytotoxic therapy [
3]. Nevertheless, 25% of these patients respond poorly to the therapy and virtually all patients eventually relapse owing to acquisition of secondary EGFR mutations or reactivation of signaling pathways downstream of EGFR [
4,
5]. Thus, despite promising initial clinical responses in some patients, the 5-year survival rate of patients treated with TKI remains relatively low [
6]. A deeper understanding of underlying molecular processes of EGFR signaling may provide insights into improving the management of EGFR-mutant lung cancer patients.
The EGFR signaling pathway is among the most important drivers of lung tumorigenesis: mutations in
EGFR (10–15%) or mutations or translocations of downstream effectors including
KRAS (25–40%) and
ALK (5–7%) are frequently found in Caucasian NSCLC patients [
7]. This results in overactivation of effector pathways including the RAS/ERK, JAK/STAT AKT/mTOR pathway, and enhancement of five of six hallmarks of cancer including evasion of apoptosis, sustained angiogenesis, resistance to antigrowth signals, invasion and metastasis and self-sufficiency in growth signals [
4].
The activity of kinases in the EGFR signaling pathway is controlled by phosphatases, which remove the phosphate groups within minutes after phosphorylation [
8]. Thus, kinases and phosphatases are equally important in modulating the activity of signaling pathways, but the role of phosphatases is far less understood. Serine/threonine phosphatase PP2A is a heterotrimeric protein composed of a structural subunit A, a catalytic subunit C and a regulatory subunit B. Members of the regulatory B subunit exhibit tissue-specific expression profiles, and are implicated in diverse cellular functions by recruiting PP2A to specific substrates [
9]. PP2A is a critical regulator of ERK and AKT, and controls downstream effectors of EGFR including NF-κB, TP53 and Bcl2 [
9‐
11]. The importance of PP2A in EGFR signaling is also illustrated by the finding that administering SMAPs, small molecule activators of PP2A, results in substantial inhibition of KRAS-driven tumor growth [
12]. Conversely, procadherin 7, an endogenous inhibitor of PP2A, which acts through SET, potentiates ERK signaling through EGFR and KRAS, and promotes transformation of KRAS transduced bronchial epithelial cells [
13]. Consistent with these findings, PP2A is repressed in NSCLC by inactivating mutations, overexpression of PP2A inhibitory proteins or post-translational modifications [
14], but in most cases the underlying molecular mechanisms are unknown.
MicroRNAs (miRNAs), short regulatory RNA sequences, which control gene expression at the post-transcriptional level, are critical regulators of signaling pathways. They act as signal amplifiers or attenuators and promote the cross-talk between signaling pathways [
15]. In a previous study, we showed that miR-29b is a mediator of NF-κB signaling in KRAS-transduced NSCLC [
16]. In this study, we define miR-19b as a mediator of the PI3K/AKT signaling pathway. miR-19b is the major oncogenic miRNA of the miR-17-92 cluster, and plays a central role in tumorigenesis of B-cell lymphomas [
17‐
19]. miR-19b is also an oncogenic miRNA in NSCLC, and is implicated in proliferation [
20], attenuation of apoptosis and migration [
21]. Upregulation of miR-19b and its paralogue miR-19a in the tumor tissue as well as in the serum is associated with poor prognosis of patients with NSCLC [
22‐
24]. Here we report that miR-19b potentiates EGFR signaling by targeting PP2A B subunit PPP2R5E and confers apoptosis resistance by targeting BCL2L11 encoding the BH3 domain-containing protein BIM. Our results provide insight into oncogenic processes of miR-19b in NSCLC cells.
Methods
Cell lines and drug treatment
EGFR mutant NSCLC cell lines PC9 and PC9ER (kindly provided by PD Dr. A. Arcaro, Department of Clinical Research, University of Bern, Bern, Switzerland), HCC4011 (kindly provided by Prof. M.D. A. F. Gazdar and Prof. M.D. J. Minna, University of Texas Southwestern Medical Center, Dallas, TX, USA) and HCC827 (American Type Culture Collection, Manassas, VA, USA) were used in this study. All cell lines were cultured in complete Roswell Park Memorial Institute medium (cRPMI) (Sigma-Aldrich, Buchs, Switzerland), supplemented with 4 mmol/l L-alanyl-L-glutamine (Bioswisstec AG, Schaffhausen, Switzerland) 1% penicillin/streptomycin and 10% fetal bovine serum (Sigma-Aldrich) at 37 °C and 5–10% CO2. Cell lines were authenticated by STR profiling (Microsynth, Balgach, Switzerland) in March 2016.
EGFR inhibitors Gefitinib (Selleckchem, Munich, Germany) and Afatinib (Selleckchem), PI3K-inhibitor LY294002 (Selleckchem), and MEK-inhibitor U0126 (Selleckchem) were used at concentrations indicated in the text.
Constructs
Luciferase reporter constructs were obtained by cloning double-stranded oligonucleotides encompassing the wild type or mutated miR-19b target sites from PPP2R5E or BCL2L11, respectively, into the
XbaI and
XhoI sites of pmiRGLO Dual-Luciferase miRNA target expression vector (Promega, Dübendorf, Switzerland). Lentiviral expression vector hsa-miR-19b-NW was obtained by cloning a PCR product encompassing the pri-miRNA sequence of miR-19b into the
NotI and
EcoRI sites of PMIRH125b-1PA-1. Oligonucleotides used for cloning are indicated in Additional file
1: Table S1. Antisense hsa-miR-19b and antisense scrambled control (System Biosciences, San Francisco, CA) were used for attenuation of miR-19b-3p levels. Gene knockdown experiments were performed using shPPP2R5E, shBCL2L11 and shc002 constructs (Sigma-Alderich, Buchs, Switzerland).
Transfections and luciferase assays
NSCLC cells were transfected with 100 ng pmiRGLO vector using transfection reagent HiPerFect (Qiagen, Hombrechtikon, Switzerland) according to the fast-forward protocol provided by the supplier. Luciferase reporter assays were performed 48 h post transfection [
25].
Lentiviral transduction and cell-based assays
Lentiviral production was carried out as described [
26]. Transduction efficiency was assessed for GFP expression 3 days post transduction by FACS. Transduced cells were sorted by FACS or selected with 0.5 μg/mL puromycin (Sigma-Aldrich).
Apoptosis was induced by treating cells with 10 ng/ml TNFα (PeproTech, Rocky Hill, NJ, USA) in combination with 0.5 μg/mL actinomycin D (Sigma-Aldrich) for 6 h. Apoptosis and viability were assessed using the ApoTox-Glo Triplex assay (Promega) as described [
25]. Alternatively, apoptosis was assessed using the pacific blue annexin V apopotosis detection kit with PI (LucernaChem). Annexin V/propidim iodide-positive cells were analyzed using a LSR II Flow Cytometer (Becton Dickinson) and FlowJo software version9.8.2 (Tree Star).
Anchorage-dependent clonogenic assay was performed in six-well plates seeded with transduced cells and cultured for 10 days in cRPMI. Colonies were fixed with methanol and stained with 0.5% crystal violet solution (Sigma-Aldrich) for 30 min, washed with deionized water and lysed in 1 mL 1% (W/V) SDS. Clonogenic growth was assessed by measuring the absorption of the lysate at 505 nm using an Infinite 200 PRO plate reader (TECAN, Männedorf, Switzerland). At least three independent experiments were carried out for each experiment.
Cell proliferation was assessed by 5-bromo-2-deoxyuridine (BrdU) incorporation assay according to the manufacturer’s instructions (Roche Diagnostics). Four thousand cells were plated per well of a 96-well plate. BrdU incorporation was performed one day post-seeding for 5 h. At least three independent experiments were carried out for each experiment.
Wound healing assay was performed as described [
27]. Sixty thousand cells were allowed to adhere for 4–6 h in a 100 μL drop of cRPMI placed in the middle of a 6-well culture dish. The monolayer was artificially injured by scratching across the plate with a 200 μL pipette tip. Wells were washed twice with cRPMI to remove detached cells and wound healing was monitored over a period of 24 h using the imaging system Cell-IQ (Canibra, Bramsche, Germany) and the CellActivision software version R1.03.01 (Yokogawa Electric Corporation, Republic of Korea).
Phosphatase activity assay
Cell extracts were prepared as described [
28]. Following centrifugation for 10 min at 12000 g, the soluble fraction was passed through a NucAwayTM Spin column (Fisher Scientific, Reinach, Switzerland) equilibrated with storage buffer and the protein concentration in the eluate was determined using the Qubit protein assay (ThermoFisher). 15 ng of the eluate was analyzed using the Ser/Thr phosphatase assay (Promega) according to manufacturer’s instructions. Cell lysates were pre-incubated at 37 °C for 10 min and the reaction was continued in the presence of PP2A substrate for 2 h. Phosphatase activity was also assessed in the presence of 25 μM PP2A inhibitor LB-100 (Selleckchem). The reaction was stopped by the addition of molybdate dye and released P
i was quantified by absorption spectroscopy at 600 nm. Phosphatase activity in the presence of P
i depleted H
2O was used as a blank. The assay was linear for the indicated incubation period and the amount of protein extract.
Phospho-kinase array and western blot analysis
Phospho-kinase array analysis was performed using 800 μg total protein according to manufacturer’s instructions (R&D Systems, Zug, Switzerland). Briefly, cell lysates were mixed with biotinylated detection antibodies and phospho-proteins were captured using antibodies spotted in duplicate on nitrocellulose membranes and quantified by chemoluminescence. Following background subtraction, the average signal intensity of pair of duplicate spots was normalized to the overall signal intensity.
For Western blot analysis 20 μg total protein was loaded per lane on a 4–20% Mini-PROTEAN TGX Gel (Bio-Rad Laboratories AG, Reinach, Switzerland). Separated proteins were transferred to PVDF membranes using the transfer turbo system (Bio-Rad). Monoclonal antibodies used in this study were directed against AKT (40D4, 1:1000, Cell Signaling Technologies), phospho-AKT (D7F10, Ser473, 1:1000, CST), CCND1 (SP4, 1:100, Cell Marque), ERK1/2 (L34F12, 1:2000, CST), phospho-ERK1/2 (D13.14.4E, Thr202/Tyr204, 1:2000, CST), GSK3β (3D10, 1:1000, CST), phospho-GSK3β (D85E12, Ser9, 1:1000, CST), PPP2R5E (5A5-1F3, 1:1000, Millipore), BIM (C34C5, 1:1000, CST), PTEN (138G6, 1:1000, CST), S6 Ribosomal protein (54D2, 1:1000, CST), phospho-S6 Ribosomal protein (D57.2.2E, Ser235/236, 1:1000, CST), STAT3 (124H6, 1:1000, CST), phospho-STAT3 (D3A7, Tyr705, 1:1000, CST), α-tubulin (clone DM1A, 1:1000, CST), GAPDH (clone D16H11, 1:1000, CST). Secondary polyclonal- donkey anti-rabbit-HRP and donkey anti-mouse-HRP (Jackson Immuno Research, Suffolk, UK) were used at 1:5000. Protein levels were normalized to α-tubulin. Visualization and quantification of protein bands were performed using a luminescent image analyzer LAS-4000 (Fujifilm, Dielsdorf, Switzerland) and Multi Gauge software (Fujifilm v.3.0).
RNA isolation and real-time PCR
RNA extraction and real-time PCR were performed as described [
29]. miRNA levels were analyzed using TaqMan Assay (Applied Biosystems), and mRNA levels were analyzed using QuantiTec Primers (Qiagen). miRNA and mRNA levels were normalized to the levels obtained for RNU48 and GAPDH, respectively. Changes in expression were calculated using the ΔΔCT method.
High-throughput miRNA NanoString profiling
One hundred and fifty ng total RNA was analyzed using the nCounter Human miRNA Expression Assay Kit H_miRNA_V3 (NanoString, Seattle, WA, USA) according to manufacturer’s instruction. Each sample was scanned for 555 fields of view (FOV) using the nCounter Digital Analyzer. nCounter data imaging QC metrics revealed no significant discrepancy between the FOVs attempted, and the FOVs counted. The binding density for the samples ranged between 0.08 and 0.21 within the recommended range.
NanoString normalization
Positive control correction was used to confirm ligation of the miRNAs to the tags. The positive correction was performed by
$$ c\times \left(\frac{m}{s}\right) $$
In this equation c is count for a microRNA in a given sample, m is the mean of the sum of the positive controls across all samples, and s is the sum of all of the positive controls for that given sample. We modified NanoStriDE web application and implemented DESeq ANODEV (uses DESeq’s built in normalization methods) in an R script. Negative control (unique probes for which no target sequence is present in the human transcriptome) subtraction and normalization of positive control corrected data were performed using “NanoStringNorm” and “NanoStringDiff” R packages (available in CRAN). We used the mean of the negative controls summed with 2 standard deviations of the negative controls. mRNA sequences (ACTB, B2M, GAPDH, RPL19 and RPLP0) were used to confirm successful hybridization and to normalize variations in sample input.
Differential expression profiling of normalized microRNAs
The normalized count data were modelled over-dispersed Poisson data using a negative binomial model in the EdgeR Bioconductor package.
Hierarchical clustering and heatmap
Hierarchical clustering and the associated heatmap for miRNA profiling data was generated with the function heatmap2 in the R package gplots or GENE-E R package [
30]. We used pairwise correlation matrix between items based on Pearson correlation method. The correlation matrix was converted as a distance matrix. Finally, clustering was calculated on the resulting distance matrix. We used average linkage method to calculate the distance matrix.
Volcano graph
miRNA content in DMSO treated cells were compared to PI3K inhibitor treated cells.
-log10 adjusted p value was plot against log2 fold change of corresponding samples using a custom R function.
Prediction of altered canonical pathways based on differentially expressed microRNAs
The prediction of targets of differentially regulated microRNAs was done by TargetScan and the experimentally observed relationships was collected from TarBase. The significance values for the canonical pathways was calculated by Fisher’s exact test right-tailed. The significance indicates the probability of association of microRNA targets from our dataset with the canonical pathway by random chance alone. For Nanostring dataset, the intensity of alteration of mRNAs of each canonical pathway was calculated based on reverse regulation of microRNA fold changes. An ‘enrichment’ score [Fisher’s exact test (FET) P-value] that measures overlap of observed and predicted regulated gene sets was calululated.
Pathway analysis based on the dataset from the phosphatase array
In order to identify upstream regulators and causal network master regulators that can potentially create the changes in phosphorylation levels of the proteins in our phosphoproteomics dataset, Phosphorylation Core Analysis tool in IPA was used to predict the affected canonical pathways [
30].
Word cloud
To visualize gene enrichment data from a pathway analysis dataset, a cloud was created using
Wordle.net and Word cloud R package. The font size of a gene (tag) is determined by its incidence in the pathway analysis data set.
Prediction of biological function of canonical pathways
We used “BioFun” R Package (available upon request) tool that examines involvement of each IPA canonical pathway in the Biological Function Classification Database of IPA known as “Ingenuity canonical pathway” and counts the number of pathways involved in a specific biological function. The results are illustrated as radar graphs.
Statistical differences
Statistical differences were calculated using the unpaired two-tailed Student’s t-test in GraphPad Prism software (v.7.0a). Statistical significance was achieved at a probability of *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, not significant.
Discussion
MiRNAs participate in signaling pathways as signal amplifiers or attenuators and regulate the activity of downstream effector pathways, and allow crosstalk between these pathways (reviewed by [
15]). We show by microarrays and bioinformatics analysis that miRNAs that are regulated by the PI3K branch of the EGFR signaling pathway are also effectors of this pathway. In agreement with this finding, miR-100 [
39], miR-125b [
25,
40] and miR-9 [
41], which are induced by the PI3K branch of EGFR, are able to enhance NF-κB activity by targeting TRAF-7, TNFAIP3 and FoxO1, respectively. Likewise, miR-205 induces parallel signaling pathways by enhancing the expression of ERBB3 [
42]. The oncomiR-1 cluster, which includes miR-18a, miR-19a, miR-19b, miR-20a and miR-20b, is another prominent example of miRNAs involved in oncogenic processes in different cancer systems. Conversely, miR-181a, which is negatively correlated with PI3K activity, interferes with such processes by targeting oncogenic
KRAS [
43] and
Bcl2 [
44]. Thus, PI3K-regulated miRNAs act as downstream effectors of EGFR signaling. Interestingly, pathway analysis of the phosphoproteome dataset of miR-19b-attenuated cells and pathway analysis of the gene target dataset of the top 17 miRNAs that are dysregulated by the PI3K inhibitor revealed very similar biological function diagrams (Additional file
5: Figure S4c). This may suggest that the phenotype elicited by the combination of all PI3K-regulated miRNAs may be recapitulated by the phenotype elicited by miR-19b alone. In conclusion, our results are consistent with a model that PI3K-regulated miRNAs act in a concerted manner to modulate the activity of the EGFR signaling pathway.
Our results indicate that miR-19b and EGFR act together to control proliferation, migration and apoptosis of EGFR mutant NSCLC in a synergistic manner, forming part of the same signaling pathway. This was confirmed by Western blot analysis showing enhanced phosphorylation of the effectors of EGFR including ERK, STAT and AKT by miR-19b overexpression. Thus, although miR-19b is induced by the PI3K/AKT branch, it activates all three major branches of EGFR indicating that one role of miR-19b is to link these signaling pathways.
How is this achieved? Phosphoproteomic analysis of miR-19b-attenuated cells pinpoints PP2A as a common regulator of ERK, STAT and AKT signaling by miR-19b. PPP2R5E regulation by miR-19b was confirmed by luciferase reporter assays, RT-qPCR, Western blot analysis and PP2A phosphatase activity assays. Thus, PPP2R5E serves as a hub for miR-19b-mediated crosstalk between these pathways.
PPP2R5E is implicated in enhanced proliferation elicited by miR-19b evident from the observation that enhanced proliferation of NSCLC cells elicited by miR-19b was completely restored in the
PPP2R5E knockdown. In contrast, targeting
PPP2R5E proved to be dispensable for apoptosis resistance induced by miR-19b. Consistent with these findings, PPP2R5E inhibits proliferation by dephosphorylation of ERK rather than apoptosis [
9,
45]. Interestingly, the proapoptotic BH3-only protein BIM (encoded by
BCL2L11), which is a master regulator of cell death in cancer cells [
38], is a relevant target of miR-19b in spontaneous and TNFα/ActD-induced apoptosis. Enhanced apoptosis in miR-19b-attenuated cells is restored in the
BCL2L11 knockdown. In contrast, clonogenic growth is only partially restored by targeting either
PPP2R5E or
BCL2L11. One explanation for this finding may be that clonogenic growth is affected by both proliferation and apoptosis, and that only one of both processes is restored in a single
PPP2R5E or
BCL2L11 knockdown.
PTEN, a well-established target of miR-19b [
46], may potentially corroborate with PPP2R5E and BCL2L11 in miR-19b-induced processes. It remains to be shown if enhanced migration elicited by miR-19b is due to targeting
PTEN [
21],
PPP2R5E (our study) or a combination of both.
Enforced expression of miR-19b triggers epithelial-mesenchymal transition (EMT) [
21]. However, in contrast to our findings and findings obtained by others [
20], Li et al. reported that miR-19b overexpression was also responsible for reduced proliferation of the NSCLC cell line A549 [
21]. This could be due to off-target effects upon high level expression of miR-19b or cell-type-specific effects. Alternatively, EMT and reduced proliferation may appear during a later period following miR-19b induction. We found that miR-19b overexpressing cells lost their proliferation phenotype upon long term culture, but this was not associated with the appearance of EMT markers (data not shown).
Novel forms of therapies aiming at reactivating PP2A may become important for the treatment of lung cancer in the future. Activators of PP2A such as SMAPs (reviewed by [
47]) or inhibitors of negative regulators such as bortezomib or erlotinib, that restore PP2A activity by targeting CIP2A [
48], are currently tested in clinical phase I/II studies. These drugs could possibly be exploited for the therapy of EGFR or KRAS-driven NSCLC. One potential drawback may be that all PP2A holoenzymes are equally affected using these pharmacological approaches which may also have an impact on normal tissue. We found that PPP2R5E contributed to 30% PP2A activity in PC9 cells, but PP2A activity was significantly enhanced in miR-19b-attenuated cells which was associated with reduced clonogenic growth. In addition, we found that attenuation of miR-19b sensitized cells to gefitinib treatment. Thus, administering antagomiRs to block enhanced levels of miR-19b may be an interesting alternative therapeutic option as it specifically restores
PPP2R5E expression in the tumour tissue.
Acknowledgements
We thank C. Schlup and Jaison Phour for technical assistance, E. Stübi-Bondarenko for providing preliminary results, and A. F. Gazdar and J. Minna, for cell lines. Bernadette Nyfeler is thanked for introduction into flow cytometry, D. Krauer for help with lentiviruses, and S. Haemmig and S. Langsch for protocols and helpful discussions.