MiR-16 regulates the pro-tumorigenic potential of lung fibroblasts through the inhibition of HGF production in an FGFR-1- and MEK1-dependent manner

Primary lung fibroblasts (“cancer-associated” (CAFs), obtained from the tumor site, and "adjacent" (AF) or “normal” (NFs), obtained from normal tissue proximal or at least 3 cm from the neoplastic lesion, respectively) were derived from surgical specimens and cultured as already described [8, 30]. All available samples were analyzed, and miRNA expression profiles were deposited in the Gene Expression Omnibus (GEO) Repository (accession number: GSE97545). Human NSCLC cell lines A549 (adenocarcinoma, CCL-185) and Calu-1 (squamous cell carcinoma, HTB-54) cells were purchased from American Type Culture Collection (ATCC, LGC Standards). LT73 primary cell line was derived in our laboratory from a surgical specimen of a male patient with lung adenocarcinoma [8]. Cancer cells were cultured in RPMI 1640 (Lonza) supplemented with 10% fetal calf serum (FCS) [30]; in vivo experiments were performed as described by Bertolini and colleagues [31]. HEK293FT (R700-07) cells were purchased from Invitrogen (Thermo Fisher Scientific, Waltham, MA, USA). All cell lines were routinely tested to exclude the presence of mycoplasma contamination, grown as adherent monolayer, and harvested at controlled density.

For immortalization, primary fibroblasts were transduced with retroviral particles produced as described [32] using the pLX-SP-hTERT vector and cultured in medium plus 1 μg/ml puromycin. The expression levels of the transgene were determined by real-time PCR using TERT TaqMan assay (Thermo Fisher Scientific) and normalized relative to HPRT. Fibroblasts expressing ectopic hTERT were characterized both in vitro (Additional file 1: Figure S1) and in vivo (Additional file 2: Figure S2) to exclude that the immortalization process affected their pro-tumorigenic features. Senescent cells were detected using the Senescence Cells Histochemical Staining Kit (Sigma-Aldrich), according to the manufacturer’s instruction.

Ectopic expression of FGFR-1 and mitogen-activated protein kinase/ERK kinase 1 (MEK1) was obtained by lentiviral particles prepared in HEK293FT cells as described [33] using the pWZL Neo Myr FLAG FGFR1, a gift from William Hahn & Jean Zhao (Addgene #20486) [34], and human Lenti-ORF pLenti-C-Myc-DDK clone of human mitogen-activated protein kinase kinase 1 (MAP2K1, OriGene #RC218460L1), respectively.

For fibroblast transfection, the negative control miRNA #1 (miR-C, #4464058) and hsa-miR-16-5p (miR-16, MC10339) miRNA mimics, the negative control miRNA #1 (miR-C inh, #4464076) and hsa-miR-16-5p (miR-16 inh, MC10339) miRNA inhibitor (Thermo Fisher Scientific), control siRNA (siCtr; 5′-CGUACGCGGAAUACUUCGATT-3′, Eurofins), siFGFR-1 #SI02224677 and #SI02224684, siHGF #SI03046946, siMEK1#6 #SI00300699, and siMEK1#7 #SI02222955 (Qiagen, Hilden, Germany) were used. Transfection of fibroblasts was performed by reverse transfection in 12-well plates, and small RNAs (50 nM final concentration for mimics, 100 nM for miRNA inhibitors) were mixed with Lipofectamine RNAiMAX Transfection Reagent (Thermo Fisher Scientific) to each well. After incubation, 1.5 × 10⁵ fibroblasts were seeded, left to adhere, and incubated for 72 h. Conditioned medium (CM) was then collected for analysis and cancer cell stimulation, and cells were harvested for western blots. Small RNAs targeting essential genes were employed in parallel as internal transfection controls. Transfection efficiency was evaluated by real-time PCR using an hsa-miR-16 probe (391 assay ID, Thermo Fisher). Before stimulation with FGF-2 (ATCC, LGC Standards), fibroblasts were serum-starved for 48 h. Viability was evaluated using Cell TiterGlo reagent (Promega) according to the manufacturer’s instructions. The FGFR-1 inhibitor SU5402 was purchased from Calbiochem.

The selection of cohorts analyzed was approved by the Fondazione IRCCS Istituto Nazionale dei Tumori Ethics Review Board. Written informed consent was obtained from each patient and healthy individual for blood collection. Whole blood samples (5–10 ml) were collected as first blood with spray-coated K₂EDTA (BD-Becton, Dickinson and Company, Plymouth, UK). Within 2 h, plasma was separated by a first centrifugation step at 1258g at 4 °C for 10 min. The supernatant containing plasma was carefully collected avoiding the fraction closest to the lymphocytic ring. Plasma was then centrifuged a second time at 1258g at 4 °C for 10 min and collected for further analysis [35].

In silico prediction of miRNA targets was obtained combining six different algorithms (DIANA microT-CDS [22285563], microrna.​org database [18158296], mirDB [18048393], PITA [17893677], RNA22 [16990141], and TargetScan v6.2 [15652477]). Putative mRNA targets predicted by at least five out of six algorithms were selected. The Jaccard Index was calculated for each pair of miRNAs as a measure of similarity between the lists of predicted targets. To identify clusters of miRNAs sharing common targets, we applied hierarchical clustering to the Jaccard Index matrix, with Euclidean distance and average linkage as clustering parameters. Graphs and statistical analysis were performed with GraphPad Prism 5.02.

The analysis of HGF plasmatic levels was performed on a set of 90 healthy heavy smokers enrolled in a lung cancer screening program [36]. Circulating HGF was measured by using commercially available ELISA kits (R&D) according to the manufacturer’s instruction. Duplicate measures were performed for each sample. Protein levels were expressed in OD value measured by Microplate Reader Tecan Infinite® M1000. Raw absorbance values were corrected by exploiting the values of the ELISA standards. Boxplots and Wilcoxon test were used to evaluate the association between HGF and categorical variables; the association between HGF and continuous variables was assessed by means of scatter plots and the calculation of Spearman correlation coefficient. Analyses were carried out using R software, version 3.2.0 (http://​www.​r-project.​org/​). The test results were considered statistically significant whenever a two-sided P value below 0.05 was achieved.

As the large-scale screening experiment was not feasible with primary CAFs due to the limited cell number, fibroblasts (CAFs, AFs, and NFs) from different patients were transduced with retroviral particles to stably express human TERT (hTERT) and immortalize cells (see above). For the high-throughput screening, at day 1, CAF154-hTERT fibroblasts were reverse transfected with a library of human miRNA mimics composed of 988 mature miRNAs arrayed on 96-well plates (875 unique sequences, miRBase v.13.0, miRIDIAN technology, Dharmacon). Briefly, 15 μl miRNA (500 nM) was spotted per well, and a mix of 35 μl Opti-MEM containing RNAiMAX (Thermo Fisher Scientific) was added. After 30 min of incubation, 100 μl of medium containing 8000 fibroblasts was added. A549-green fluorescent protein (GFP) cells (3500 cells/well) were seeded at day 3 (48 h after fibroblast transfection) and co-cultured for further 48 h. The experiment was stopped by fixing the cells with 4% paraformaldehyde, and nuclei were counterstained with Hoechst 33342. Image acquisition was performed using an ImageXpress Micro automated high-content screening fluorescence microscope (Molecular Devices) at a × 4 magnification; a total of four images were acquired per wavelength, well, and replicate, corresponding to ca. 10,000–15,000 cells analyzed per experimental condition and replicate. Image analysis to determine the number of fibroblasts (GFP-negative) and A549 cells (GFP-positive) was performed using the “Multi-Wavelength Cell Scoring” application module implemented in MetaXpress software (Molecular Devices). To score the effect of each miRNA in stimulating or decreasing fibroblast and A549 cell growth, results were normalized per plate, relative to the median of the samples. Screening was performed in duplicate, at the ICGEB High-Throughput Screening Facility (Trieste, Italy). In validation co-culture experiments, the same conditions of the HTS were applied, but plates were incubated in a Cell-IQ SLF instrument (CM Technology Oy, Tampere, Finland) for image acquisition and automated analysis.

For secretome analysis, 1.5 × 10⁵ CAF154-hTERT cells were reverse transfected as described above in 12-well plates and CM was collected 72 h later. Biological triplicates were prepared for each condition, i.e., transfection with miR-C and miR-16. In this study, we quantified simultaneously 64 cytokine/chemokine/growth factor biomarkers by using the Human Cytokine Array/Chemokine Array 64-Plex Discovery Assay, 17 angiogenesis/growth factor biomarkers of the Human Angiogenesis Array and Growth Factor Array, and 13 matrix metalloproteinases/tissue inhibitors of metalloproteinases of the Human MMP and TIMP Panel (Eve Technologies Corp, Calgary, AB, Canada). To quantify HGF in different fibroblast cell lines, serum-free media conditioned for 24 h by 1 × 10⁶ cells were collected and analyzed by ELISA (Human HGF Instant ELISA #BMS2069INST; eBioscience) according to the manufacturer’s instructions.

Western blots were performed as already described [33]. The following primary antibodies were employed: α-cMet #sc-10 (Santacruz); α-Actin #A1978 and α-pERK1/2 #M8159 (Sigma-Aldrich); α-p-cMet #44888G (Life Technologies); α-FGFR-1 #9740, α-MEK1/2 #4694, α-MEK1 #12671, and α-pAKT #9275 (Cell Signaling Technologies); and α-HGF #10679 (Abcam).

For miRNA target validation, the firefly luciferase-expressing pMirTarget plasmids containing the 3′ untranslated region (UTR) of HGF #SC206339, of MEK1 #SC211301, and of FGFR-1 #SC211347 were employed (OriGene). The pMirTarget 3′UTR HGF was mutagenized using QuikChange II XL Site-Directed Mutagenesis Kit (Agilent, Santa Clara, CA) to delete the putative binding site for miR-16. Target sequences are shown in Additional file 3: Figure S3. The HEK293T cells were seeded (5 × 10⁵ cells/well) in 96-well plates and left to adhere overnight. Then, miRNA transfection was performed using Lipofectamine 2000 reagent (100 nM final miRNA concentration) and pMirTarget plasmid (0.1 μg/well), together with the Renilla luciferase-expressing pGL4.74 plasmid (0.01 μg/well). After 24 h, both the firefly and Renilla luciferase activities were measured by dual-luciferase reporter assay (Promega) according to the manufacturer’s instructions.

Cell motility was studied in wound-healing experiments by seeding 6 × 10⁴ A549 cells in each chamber of a culture insert (Ibidi) and left to adhere overnight in medium with 0.1% serum. The day after, medium was replaced with the CM collected from fibroblasts diluted 1:1 in fresh medium and left for 6 h before removing the inserts and starting the image acquisition using a Cell-IQ SLF instrument as already described [37].

In vivo studies to characterize the immortalized fibroblasts were performed as already described [8, 30]. Experiments were approved by Ethics Committee for Animal Experimentation of the Fondazione IRCCS Istituto Nazionale dei Tumori (Milan, Italy) according to EU directive 2010/63/EU. CD1-Nude female mice, 5–10 weeks old, were divided in uniform groups on the basis of their weight. No engrafted animal was excluded from the analysis.

For tumorigenic assays, viable A549 tumor cells (5 × 10⁵) were exposed to CM of fibroblasts transfected with miR-C or miR-16 or CM supplemented with HGF-neutralizing antibody. After 2 days, cancer cells were injected subcutaneously (s.c.) into both flanks of nude mice with Matrigel 1:1 v/v (BD Biosciences) in a final volume of 200 μl. For co-injection experiments, A549 cells (1 × 10³) were injected s.c. with fibroblasts transfected with miR-C or miR-16. At the end of the observation period or when the tumor volume reached at least 300 mm³, lungs were removed and dissociated for the analysis of disseminated cancer cells as previously described [8].

To study the effect of individual miRNAs on fibroblast pro-tumorigenic features, we performed a high-throughput screening employing the patient-derived CAF154-hTERT cell line (see the “Methods” section for a detailed description). Fibroblasts were transfected with a library of 875 unique mature miRNAs, and after 48 h, GFP-expressing A549 lung cancer cells were added (Fig. 1a). MiRNAs were ranked on the basis of their capability to inhibit or stimulate the growth of co-cultured A549-GFP cells (Fig. 1b, c). Although the correlation was not very marked (Spearman r = 0.35), a significant number of miRNAs had the same effect on both A549 and CAF154-hTERT cell growth (Fig. 1d). However, a number of miRNAs produced opposite outcomes on the growth of these cell lines (Fig. 1d).

The 60 miRNA candidates displaying the strongest stimulatory effect on co-cultured A549 cells and the top 60 miRNAs which resulted in an inhibitory effect were then clustered according to their predicted ability to interact with the same mRNAs. Using this approach, these miRNAs could be grouped into six pro-stimulatory (Table 1) and five pro-inhibitory (Table 2) clusters, characterized by common seed regions.

In an attempt to identify clinically relevant miRNAs, we focused on cluster #1 (Table 2) of the inhibitory miRNAs, which contains miR-16. In fact, by analyzing the levels of this miRNA in 47 primary cell lines of lung fibroblasts established from fresh tumor biopsies obtained from lung cancer patients (including 21 matched samples from both cancer tissue—cancer-associated fibroblasts (CAFs)—and non-involved lung parenchyma—normal fibroblasts (NFs)—for a total of 26 CAF lines and 21 NF lines), we explored the correlations between miR-16 expression levels and clinical parameters (Table 3). In detail, miR-16 levels in fibroblasts were not statistically associated with sex or age of patients, tissue of origin (cancer vs. normal), histology, and stage or grade of disease, but lower levels of miR-16 were detected in fibroblasts from smoke-exposed lungs compared to fibroblasts from non-exposed lungs [median intensity value 860 (IQ range 310–1437) vs. 1470 (IQ range 733–2120); p = 0.215], and in NF, this difference was statistically significant [324 (223–972) vs. 1589 (1494–2237); p = 0.006]. Interestingly, strongly reduced levels of miR-16 were also detected in fibroblasts from patients with chronic obstructive pulmonary disease (p = 0.048), in particular in NF (p = 0.004).

We therefore speculated that inflammation and smoke-related miR-16 reduction found in patients could contribute to generation of a microenvironment conductive to cancer proliferation and aggressiveness, while its restoration in fibroblasts could reduce the growth of cancer cell. Accordingly, transfection of CAF154-hTERT cells with miR-16 resulted in the inhibition of both fibroblast and adjacent A549 cell proliferation (Fig. 1e, upper panel) in our HTS, and the effect on A549 cells co-cultured with transfected fibroblasts was further confirmed in independent experiments (Fig. 1e (lower panel), f). Importantly, also, miR-497 and miR-15b, which target the same seed region of miR-16, were found among the top candidates (cluster #1, Table 2), and the other miRNAs belonging to the same family (miR-195, miR-15a, and miR-424), even if less efficiently, all showed an inhibitory effect on A549 cells (0.902, 0.837, and 0.834 normalized A549 cell number, respectively).

To investigate the mechanisms underlying the inhibitory effect of miR-16 on cancer cells, we first considered the possibility that miR-16 could be released from fibroblasts and taken up by A549 cells. However, with experiments on transfected fibroblasts conditioned medium (CM), we estimated a concentration of released miR-16 ranging between 0.1 and 0.05 nM and found that similar amounts of miRNA did not influence A549 cell proliferation (data not shown).

We next investigated therefore whether miR-16 ectopic expression in CAF154-hTERT fibroblasts could affect the secretome profile of the transfected cells, thereby influencing the growth of A549 cells. CAF154-hTERT fibroblasts were transfected with control miRNA (miR-C) or miR-16, and CM analyzed after 72 h by multiplex array to quantify 91 unique soluble factors, belonging to three human protein sets: (i) cytokines, (ii) MMPs and TIMPs, and (iii) angiogenetic factors. We observed marked differences in the CM obtained from fibroblasts transfected with miR-16, finding 26 out of 91 soluble factors with concentrations < 75 or > 125% compared to miR-C transfected fibroblasts (Fig. 2a).

Among the perturbed factors, HGF was the most affected by miR-16 transfection and was almost completely depleted in the CM (Fig. 2a, b). Moreover, HGF expression was reduced also at the cellular level (Fig. 2b, right), suggesting a reduction of its expression rather than an inhibition of its release. Conversely, the levels of its cognate receptor cMet increased, compatibly with a stalled turnover due to reduced paracrine or autocrine stimulation (Fig. 2b, right). Importantly, inhibition of miR-16 resulted in an opposite effect and caused the accumulation of HGF both in the CM (Fig. 2c, left) and intracellularly (Fig. 2c, right). To rule out the possibility that miR-16 regulates HGF levels in a cell-type-specific manner, we transfected a number of patient-derived primary fibroblasts with miR-16 and consistently found a striking reduction of HGF levels in the CM (Fig. 2d), while miR-16 inhibition reproducibly increased HGF levels (Additional file 4: Figure S4). We checked whether miR-16 directly targets the HGF 3′UTR by luciferase assay and found a partial, yet significant, reduction of the luciferase activity when cells were transfected with miR-16 (Fig. 2e). Importantly, when the sequence corresponding to the seed region of miR-16 was mutated in the HGF 3′UTR (Additional file 3: Figure S3), the effect of the miRNA was almost completely abrogated (Fig. 2e), indicating a direct targeting of miR-16 on HGF mRNA.

Finally, we reasoned that the microenvironmental changes related to miR-16 reduction, which we previously observed in lung-derived fibroblasts (Table 3), could also potentially influence systemic levels of HGF. Interestingly, analysis of plasma samples from healthy heavy smokers enrolled in a CT-screening program for early lung cancer detection revealed a strong correlation between smoke exposure or COPD and circulating levels of HGF (p = 0.001 and p = 0.041 respectively, Fig. 2f). No correlation was found with age (data not shown).

Taking into account the strong effect of miR-16 on HGF secretion (Fig. 2a, b) and the only partial reduction of the luciferase activity in the HGF 3′UTR reporter assay (Fig. 2e), we also considered additional inhibitory mechanisms and focused on the FGFR-1 receptor. Stimulation of this receptor by means of its cognate ligand FGF-2 resulted in the accumulation of HGF in the CM (Fig. 3a) while miR-16 transfection strongly reduced the levels of FGFR-1 (Fig. 3b), but not of its cognate ligands FGF-1 and FGF-2 (Additional file 5: Figure S5), and it hindered the activation of the FGFR-1 downstream mediators ERK1/2 (Fig. 3b). Of note, miR-16-induced reduction of FGFR-1 apparently did not stem from a direct effect of this miRNA on FGFR-1 mRNA, as judged by the lack of effect of miR-16 on FGFR-1 3′UTR in a luciferase reporter assay (Fig. 3c). Nonetheless, the transfection of CAF154-hTERT cells with miR-16 or an FGFR-1-specific siRNA both resulted in HGF reduction (Fig. 3d). Moreover, treatment with a specific inhibitor of FGFR-1 caused a time-dependent reduction of HGF in the CM (Fig. 3e).

The stimulation of A549 cells with CM collected from control CAF154-hTERT cells caused a clear and rapid activation of the cMet receptor (Fig. 3b), while CM collected from the same cells transfected with miR-16 and with siRNAs directed against FGFR-1 and HGF failed to trigger the phosphorylation of cMet (Fig. 3b). Importantly, we also observed a direct correlation in patient-derived fibroblasts between HGF and FGFR-1 expression, which was however evident only in HGF high-expressing fibroblasts (HGF > 1, Fig. 3f).

We then asked whether the restoration of FGFR-1 levels was sufficient to prevent the effect of miR-16 on HGF and therefore transduced CAF154-hTERT cells to ectopically express FGFR-1 (Fig. 3g). Despite high levels of FGFR-1, miR-16 still markedly reduced the secretion of HGF (Fig. 3h), suggesting the presence of other miR-16 targets which control HGF expression downstream FGFR-1. Interestingly, miR-16 transfection abrogated ERK1/2 activation also in the presence of FGFR-1 over-expression and this correlated with the downregulation of the MEK1/2 kinases (Fig. 3g), which are responsible for ERK1/2 phosphorylation and could therefore represent additional targets of miR-16.

To confirm the involvement of MEK1 in the miR-16-dependent regulation of HGF, we transfected CAF154-hTERT cells with this miRNA and found that MEK1 was indeed reduced (Fig. 4a). Moreover, silencing of MEK1 (but not MEK2) prevented HGF accumulation in the CM (Fig. 4b, c and data not shown). By using a luciferase 3′UTR reporter assay, we confirmed that MEK1 is a direct target of miR-16 (Fig. 4d) and it represents a determinant of HGF secretion in the CM (Fig. 4b). Similar to what was observed for FGFR-1, ectopic expression of MEK1 is not sufficient to prevent the miR-16-dependent downregulation of HGF (Fig. 4e, f). Altogether, our data show that miR-16 acts as potent inhibitor of HGF by targeting directly and indirectly several mediators of the FGFR-1 pathway which in turn controls HGF expression.

The stimulation of A549 cells with CM collected from CAF154-hTERT cells had a pleiotropic effect, activating different pathways known to enhance survival and aggressiveness of cancer cells. In fact, in time-course experiments, we observed a rapid activation of cMet, AKT, and ERK1/2 (Figs. 5a and 3b). Therefore, we investigated whether the stimulation with CM could promote proliferation and motility of A549 cancer cells in a miR-16-dependent manner. CM collected from CAF154-hTERT cells transfected with miR-16 displayed a reduced capacity to promote proliferation of A549 cancer cells (Fig. 5b, upper panel) and also induced a small, but significant, difference in proliferation in another cell line previously shown to be responsive to microenvironment cues [30] (LT73, Fig. 5b, bottom panel). Additionally, A549 cells stimulated with the same CM showed reduced migratory capacity in wound-healing experiments (Fig. 5c). Of note, the pro-migratory property of HGF was assessed in our settings also by using CM collected from CAF154-hTERT cells transfected with a siRNA targeting HGF. This caused a marked delay of A549 cell migration compared to controls (Fig. 5d), confirming a central role for HGF in this setting. A panel of CM collected from primary patient-derived fibroblasts was then characterized for the concentration of HGF (Fig. 5e). Conditioned media collected from fibroblasts containing low (green arrows) and high (black arrows) levels of HGF (Fig. 5e) were employed to stimulate A549 cells in wound-healing experiments in the presence or absence of an HGF-neutralizing antibody (Fig. 5f). A549 cells stimulated with CM containing high levels of HGF migrated significantly more rapidly compared to those stimulated with low-HGF CM (Fig. 5f, g), and the HGF-neutralizing antibody partially abrogated this effect (Fig. 5f, g), thus confirming the role of fibroblast-derived HGF in promoting motility of cancer cells. Moreover, the migration speed (expressed as time point at which the gap is reduced to half of the original area, T1/2) was inversely correlated to the concentration of HGF in the CM (Fig. 5h). To verify whether the HGF-dependent pro-migratory effect of fibroblast CM was specific for A549 cells or generally applicable to other cancer cell lines, we repeated these experiments with squamous lung cancer Calu-1 cells. Migration experiments confirmed that cell motility depends on the HGF concentration present in CM (Fig. 5i–j).

Having shown that fibroblast miR-16 levels slightly affect the proliferation capacity of adjacent cancer cells and can strongly influence their migration, we investigated the relevance of the miR-16/HGF axis in vivo. Two experiments were performed by culturing the A549 cells with the CM collected from CAF154-hTERT fibroblasts transfected with miR-16 and control miR-C (Fig. 6a, b) before mouse engrafting (Fig. 6c) or by directly co-injecting the A549 cells with the transfected fibroblasts in nude mice (Fig. 6d).

In both cases, transfection with miR-16 of CAF154-hTERT fibroblasts delayed the ability of the A549 cancer cells to form nodules. Moreover, inhibition of HGF in CM by means of a neutralizing antibody showed an effect comparable to miR-16 transfection (Fig. 6c). Importantly, mouse lung analysis revealed that HGF inhibition also reduced the capacity of A549 cells to metastasize to lungs (Fig. 6e) in a significant manner while ectopic expression of miR-16 in CAF154-hTERT fibroblasts reduced the metastasizing potential of cancer cells, albeit not significantly (Fig. 6e, f).