Joint action of miR‐126 and MAPK/PI3K inhibitors against metastatic melanoma

Emerging data support the rationale of combined therapies in advanced melanoma. Specifically, the combined use of drugs with different mechanisms of action can reduce the probability of selecting resistant clones. To identify agents active against melanoma cells, we screened a library of 349 anti‐cancer compounds, currently in clinical use or trials, and selected PIK‐75, an inhibitor of the phosphatidylinositol 3‐kinase/protein kinase B (PI3K/AKT) pathway, as the ‘top active’ drug. PIK‐75 was then used alone or in combination with vemurafenib, the first BRAF inhibitor approved for patients with melanoma harboring BRAF mutations. We identified a combined dose of PIK‐75 and vemurafenib that inhibited both the PI3K/AKT and mitogen‐activated protein kinase pathways, thereby overcoming any compensatory activation. In view of the important tumor suppressor function induced by restoring expression of microRNA (miR)‐126 in metastatic melanoma cells, we examined whether miR‐126 has a synergistic role when included in a triple combination alongside PIK‐75 and vemurafenib. We found that enforced expression of miR‐126 (which alone can reduce tumorigenicity) significantly increased PIK‐75 activity when used as either a single agent or in combination with vemurafenib. Interestingly, PIK‐75 proved to be effective against early passage cell lines derived from patients’ biopsies and on melanoma cell lines resistant to either vemurafenib or dabrafenib, thus suggesting that it potentially has the capability to overcome drug resistance. Finally, the synergistic role played by miR‐126 in combination with vemurafenib and/or PIK‐75 was demonstrated in vivo in mouse xenograft models, in which tumor growth inhibition was associated with increased apoptosis. These results not only show the efficacy of PIK‐75 and vemurafenib co‐treatment but also indicate that restoration of miR‐126 expression in advanced melanoma can enhance their antitumor activity, which may possibly allow dose reduction to decrease adverse events without reducing the therapeutic benefits.

Emerging data support the rationale of combined therapies in advanced melanoma. Specifically, the combined use of drugs with different mechanisms of action can reduce the probability of selecting resistant clones. To identify agents active against melanoma cells, we screened a library of 349 anti-cancer compounds, currently in clinical use or trials, and selected PIK-75, an inhibitor of the phosphatidylinositol 3-kinase/protein kinase B (PI3K/AKT) pathway, as the 'top active' drug. PIK-75 was then used alone or in combination with vemurafenib, the first BRAF inhibitor approved for patients with melanoma harboring BRAF mutations. We identified a combined dose of PIK-75 and vemurafenib that inhibited both the PI3K/AKT and mitogen-activated protein kinase pathways, thereby overcoming any compensatory activation. In view of the important tumor suppressor function induced by restoring expression of microRNA (miR)-126 in metastatic melanoma cells, we examined whether miR-126 has a synergistic role when included in a triple combination alongside PIK-75 and vemurafenib. We found that enforced expression of miR-126 (which alone can reduce tumorigenicity) significantly increased PIK-75 activity when used as either a single agent or in combination with vemurafenib. Interestingly, PIK-75 proved to be effective against early passage cell lines derived from patients' biopsies and on melanoma cell lines resistant to either vemurafenib or dabrafenib, thus suggesting that it potentially has the capability to overcome drug resistance. Finally, the synergistic role played by miR-126 in combination with vemurafenib and/or PIK-75 was demonstrated in vivo in mouse xenograft models, in which tumor growth inhibition was associated with increased apoptosis. These results not only show the efficacy of PIK-75 and vemurafenib co-treatment but also indicate that restoration of miR-126 expression in advanced melanoma can Abbreviations BRAFi, BRAF inhibitors; EOB, excess over bliss; IC 50 , half maximal inhibitory concentration; IRS-1, insulin receptor substrate 1; LC-MS/MS, liquid chromatography-tandem mass spectrometry; MAPK/ERK, mitogen-activated protein kinase/extracellular signal-regulated kinase; MAPK, mitogen-activated protein kinase; miR, microRNA; MTV, median tumor volume; PARP, poly (ADP-ribose) polymerase; PI3K/AKT, phosphatidylinositol 3-kinase/protein kinase B; PI3Ki, PI3K inhibitors; Pt1, Pt2, Pt3 and Pt4, cell lines obtained from bioptic patient samples; qRT-PCR, quantitative reverse transcription PCR; SD, standard deviation; TGI, tumor growth inhibition. enhance their antitumor activity, which may possibly allow dose reduction to decrease adverse events without reducing the therapeutic benefits.

Introduction
Cutaneous melanoma is a highly aggressive form of cancer whose incidence is constantly increasing worldwide (Apalla et al., 2017;Matthews et al., 2017). Although, it is surgically curable in its early phase, after dissemination, metastatic melanomas are characterized by poor prognosis and short survival of patients (Eggermont et al., 2014). In the last few years, a number of novel promising therapeutic agents have been developed, specifically including molecular targeting drugs and immune checkpoint inhibitors. Despite the impressive clinical responses, only a small fraction of patients obtain a durable response, and the emergence of acquired resistance appears a major problem for both treatments (Matthews et al., 2017;Nazarian et al., 2010). To overcome this failure, the emerging therapies are increasingly considering combinatorial treatments based on different mechanisms of action (Hepner et al., 2017;Shtivelman et al., 2014;Volpe et al., 2017).
One of the most relevant targeted proteins is BRAF, a Ser/Thr protein kinase mutated in 60-70% of melanoma patients, mostly as a V600E substitution (Davies et al., 2002). Mutated BRAF represents a constitutively active isoform which maintains the mitogen-activated protein kinase (MAPK) pathway always activated (Cantwell-Dorris et al., 2011). Vemurafenib, the first BRAF inhibitor (BRAFi) approved for BRAF-mutated melanoma, induces a clinical response in a large proportion of patients with an overall response rate (RR) of up to 69%. Unfortunately, 90% of these patients develop resistance and undergo tumor progression within 6-8 months (Chan et al., 2017;George et al., 2017), in most cases because of the presence of additional mutations in the same (MAPK) or in other signaling pathways. Phosphatidylinositol 3-kinase/protein kinase B (PI3K/AKT) is the second most frequently deregulated pathway in melanoma (Lim et al., 2017). Several inhibitors targeting the PI3K/ AKT pathway (PI3Ki) have been developed and are now at different stages of clinical trials (Massacesi et al., 2016). Although some clinical activity has been observed with PI3Ki as a single agent, clinically significant results have not been observed. Conversely, several publications showed improved efficacy of PI3Ki used in combination with other agents (Tolcher et al., 2018). Indeed, possibly because of the complexity derived from molecule interactions and feedback loops, single-agent therapies did not yield satisfactory results, with failures at least partly due to the significant functional cross-talk existing between the MAPK and PI3K/AKT key pathways. In fact, inhibition of just one of these two signaling pathways often leads to a compensatory activation of the other, eventually interfering with adequate therapeutic responses (Pappalardo et al., 2016). Hence, there is a strong scientific rationale to explore the simultaneous targeting of more than one signaling pathway, including MAPK and PI3K/AKT (Lim et al., 2017;Smalley and Sondak, 2013). Moreover, regulation of gene expression at the post-transcriptional level, such as that mediated by microRNA (miR), may interfere with the melanoma malignant phenotype. We previously showed the marked decline of miR-126 and its complement miR-126* (miR-126-3p and miR-126-5p, respectively, hereafter referred as to miR-126) in metastatic cells, demonstrating their tumor suppressor function and proposing the restoration of their levels as a potential therapeutic option for the treatment of advanced melanoma (Felli et al., 2013(Felli et al., , 2016. Interestingly, miR-126 down-modulation also plays a role in other neoplasms, where its restored expression significantly inhibits growth and dissemination as well as angiogenesis (Alhasan, 2019;Han et al., 2018;Liu et al., 2019). Indeed, miR-126 expression was shown to be significantly associated with outcomes of patients with metastatic colorectal cancer receiving bevacizumab (Fiala et al., 2017) and reversal of resistance to TRAIL in cervical cancer . In addition, miR-126 serum amount has been demonstrated to represent a prognostic biomarker (Feng et al., 2018). Recently, a number of long noncoding RNA have been reported to favor cancer progression via miR-126 down-regulation (Huang et al., 2018;Sun et al., 2019;Wang et al., 2019).
Here, we have investigated the possibility of using miR-126 as a sensitizing agent to increase the effects of targeted therapies against key signaling pathways such as MAPK/extracellular signal-regulated kinase (ERK) and PI3K/AKT. Specifically, as an MAPK inhibitor, we selected vemurafenib, since it represents a standard melanoma first-line treatment in BRAF-mutated melanoma (Savoia et al., 2019). Conversely, PIK-75 was identified among the more active drugs by screening an anti-cancer drug library on metastatic melanoma cells.

Cell culture and transduction
The malignant human melanoma cell line A375M was kindly provided by R. Giavazzi (Istituto Mario Negri, Bergamo, Italy). The human melanoma cell lines A375 and SK-Mel28 and the sublines A375-R and SK-Mel28-R with acquired resistance to dabrafenib, and the early passage cell lines obtained from bioptic samples (reported as Pt1, Pt2, Pt3, Pt4) were obtained from S. D'Atri (IDI IRCCS, Rome, Italy) (Caporali et al., 2016(Caporali et al., , 2017. All biological materials were obtained with the informed written consent of patients and the study was conducted according to the Declaration of Helsinki Principles. The study methodologies were approved by the local ethics committee. Vemurafenib-resistant A375M (A375M-VR) cell lines (TripZ and miR-126) were generated by culturing parental cells in increasing concentrations of vemurafenib (from 0.5 to 10 lM; Selleckchem, Munich, Germany) for at least 2 months and subsequently maintained in full growth medium containing 5 lM vemurafenib (Su et al., 2012). Overexpression of miR-126 was obtained in melanoma cells using the doxycycline-inducible lentiviral vector TripZ (Dharmacon, Lafayette, CO, USA), as already reported (Felli et al., 2013). After infection, transduced cells were selected with puromycin and red fluorescent protein level was evaluated by flow cytometry (FACSCanto; BD Biosciences, Milan, Italy) upon doxycycline induction (Sigma-Aldrich). When indicated, the A375M-VR melanoma cell line was treated with the proteasome inhibitor MG-132 for 4 h at a final concentration of 20 lM (Sigma-Aldrich). MiR-126-3p mimic and negative control expressions were obtained using 40 nM of miRIDIAN miR mimic (Product numbers #300626-07 and #CN-001000-01; Dharmacon). MiR-126-3p silencing was performed using 200 nM of miRCURY locked-nucleic-acid (anti-miR-126-3p). Power Inhibitor knockdown probe for miR-126-3p and negative control were obtained from Exiqon (Vedbaek, Denmark; Product numbers #426717-00 and #199006-00). Transfections were performed using Lipofectamine 2000 (Life Technologies, by Thermo Fisher, Waltham, MA, USA) according to the manufacturer's instructions. All melanoma cell lines were cultured in Dulbecco's modified Eagle's medium (Gibco by Life Technologies, Paisley, UK) supplemented with 10% FBS (Gibco). Cells were incubated at 37°C with 5% CO 2 in a humidified chamber. Cell lines were authenticated by standard short tandem repeat-based genotyping (Banca Biologica e Cell Factory, IRCCS Azienda Ospedaliera Universitaria San Martino-IST Istituto Nazionale per la Ricerca sul Cancro, Genoa, Italy).

Drug cytotoxicity experiments
For cytotoxicity assays, melanoma cell lines were plated in 96-well plates at a density of 2 9 10 4 cellsÁmL À1 in triplicate. Anti-cancer compound library (349 bioactive compounds), vemurafenib and PIK-75 were purchased from Selleckchem. A complete list of the compounds used for the screening is available in Table S1. Compounds were dissolved in DMSO and added 48 h after cell plating. ATP levels were measured 24 and 48 h later, using CellTiter-Glo TM (Promega Corporation, Madison, WI, USA) and quantified using an ELISA plate reader (VICTOR2; Perkin Elmer, Norwalk, CT, USA). During the experiments, cells were grown in the presence of doxycycline.
The excess over bliss (EOB) independence model was used to quantify the synergy for each compound pair. EOB evaluates the difference between the observed and expected inhibition of different dose drug combinations. Specifically, red squares indicate strong antagonistic effect, orange and light orange squares indicate slight antagonistic effect, dark blue indicates additive effect, and light green and dark green squares indicate slight and strong synergistic effects, respectively.

Western blot
Total cell lysates were prepared by using Nonidet-P40 lysate buffer (1% NP40, 200 nM NaCl, 50 mM Tris pH 7.4) plus protease and phosphatase inhibitors, maintained on ice for 30 min, vortexed and then centrifuged at 800 g for 10 min at 4°C. Protein concentration was measured by BioRad protein-assay (Hercules, CA, USA). Western blot was performed according to standard procedures. Total cell lysates were separated by the precast NuPAGE polyacrylamide gel system (Life Technologies by Thermo Fisher).

Apoptosis detection
Apoptosis was measured by flow cytometry. Cells showing a sub-G0 DNA content were identified as apoptotic. For DNA staining, 2 9 10 4 cells were plated in 24-well microtiter plates. After 48 h, cells were treated with vemurafenib (500 nM) and PIK-75 (40-60-80 nM) alone or in combination. After 24 h, cells were resuspended in Nicoletti's buffer containing propidium iodide 50 lgÁmL À1 . Samples were analyzed with FACSCanto flow cytometer (BD Biosciences) and data were analyzed with FACS DIVA software (Becton Dickinson, BD Biosciences, Milano, Italy).

Xenograft mouse models
Empty vector or miR-126/126*-transduced A375M cells were subcutaneously injected into the flanks of 5-weekold female nude mice (minimum n = 5 mice per group; Harlan, Udine, Italy) at a dose of 10 6 cells mixed 1 : 1 with Matrigel (BD Biosciences) in 200 lL PBS. Treatments were initiated when xenografted tumors reached 30-60 mm 3 and were administered daily intraperitoneally. Mice were randomly allocated to four treatment arms: PIK-75 (2 mgÁkg À1 DMSO : H 2 O), vemurafenib (20 mgÁkg À1 DMSO : PBS), the combination (2 mgÁkg À1 PIK-75 plus 20 mgÁkg À1 vemurafenib), and DMSO : H 2 O : PBS (control). Since PIK-75 and vemurafenib are reversible inhibitors, they were administered five times per week to maintain stable inhibition. Tumor growth was monitored twice a week and diameters measured with calipers. Volumes were calculated using the following formula: volume (mm 3 ) = (width) 2 9 length/2, where length represents the largest tumor diameter and width the perpendicular. Percentage of tumor growth inhibition (TGI) was calculated as the difference between the median tumor volume (MTV) of a treated group and the control group, using the following formula: %TGI = [(MTV control À MTVtreated )/MTV control ] 9 100. During the experiments, doxycycline was administered to the mice by daily oral gavage (10 mgÁmL À1 ; Sigma-Aldrich). Relative tumor growths were calculated by normalizing over initial tumor volume. Animal experiments were performed at the Istituto Superiore di Sanit a (Rome, Italy) according to Italian law and institutional guidelines.

Immunofluorescence and apoptosis analysis on formalin-fixed paraffin-embedded tissue
Serial sections from mice tumor nodules embedded in paraffin were dewaxed and rehydrated. For immunolocalization studies, slides were subjected to heat-mediated antigenic retrieval (10 mM sodium citrate buffer pH 6.0) and subsequently permeabilized (0.1% Triton X-100 for 10 min) and saturated (3% BSA for at least 2 h) at room temperature. After incubation overnight at 4°C in a humidified chamber with the primary antibodies, slides were incubated with specific fluorophore conjugated secondary antibodies (Alexa Fluor; Molecular Probes, Eugene, OR, USA) for 45 min at room temperature. Negative controls were performed by omission of the primary antibody in each experiment.
Finally, slides were mounted with SlowFade anti-fade reagent containing DAPI (Molecular Probes) and analyzed by Olympus F1000 laser-scanning confocal microscopy (Olympus, Tokyo, Japan).
For apoptosis analysis, slides were subjected to TUNEL assay using In situ Cell Death Detection Kit, POD (Roche Italia, Monza, Italy). Very briefly, after digestion with Proteinase K (18 lgÁmL À1 for 30 min at room temperature), permeabilization and endogenous peroxidase blocking, specimens were incubated with TUNEL reaction mixture for 1 h at 60°C. Apoptotic positive nuclei were detected by incubation with Converter-POD, followed by the colorimetric development of the 3,3 0 -diaminobenzidine chromogenic substrate (Pierce by Thermo Fisher Scientific, Rockford, IL, USA). Finally, sections were counterstained with Hematoxylin & Eosin. Negative controls were performed by omission of TdT enzyme solution in the TUNEL mixture. Slides were evaluated using a Nikon Eclipse E1000 equipped with a Nikon DXM 1200 digital camera with dedicated acquisition software (Nikon ACT-1 v. 2.1; all from Nikon Instruments, Campi Bisenzio, Florence, Italy).

Proteomic analysis and data processing
Proteins extracted from mice tumor were resolved by SDS/PAGE on precast 4-12% Bis-Tris minigels (Life Technologies). Following staining with NuPage Colloidal Coomassie (Life Technologies), gel lanes were cut into 10-12 slices, and in-gel tryptic digestion and liquid chromatography-tandem mass spectrometry (LC-MS/MS) were performed as already described (Boussadia et al., 2018). Data acquisition was performed in data-dependent Top5. Spectra files were analyzed by Sequest HT search engine with PROTEOME DISCOVERER 1.4 (Thermo Fisher) using the Human Uniprot-Swissprot database (released on June 2016) and containing also decoy database. The carbamidomethylation of cysteines was specified as a fixed modification, and the oxidation of methionine and phosphorylation of serine, threonine and tyrosine were set as variable modifications; mass tolerance was set to 1 Da for precursor ion and 0.4 Da for fragment ions; a maximum of two missed cleavages was allowed. The Percolator tool was used for peptide validation based on the q-value and high confidence was chosen, corresponding to a false discovery rate ≤ 1% at peptide level. Proteins were identified with a minimum of one peptide rank = 1. Protein abundance was determined according to the label-free Top3 method as already described (Fratini et al., 2017).

Statistical analysis
Statistical analysis was performed using the t-test with P < 0.05 deemed statistically significant. Unless otherwise stated, results were representative of at least three independent experiments. Two-way ANOVA analysis was performed using GRAPHPAD version 6.0 for Windows (San Diego, CA, USA) followed by Bonferroni post hoc test when appropriate.

Screening of anti-cancer compounds reveals high effectiveness PIK-75 in melanoma growth inhibition in vitro
In view of the significant clinical effects obtained with BRAFi in melanoma patients carrying the BRAF V600 mutation, but bearing in mind that nearly all of them eventually progress, we were searching for additional drugs that, combined with vemurafenib, might be able to overcome drug resistance.
On these bases, we screened a library of 349 anticancer compounds, including agents under evaluation in clinical trials or already approved for clinical use, on the BRAF V600E A375M metastatic melanoma cell line. The initial screening strategy was based on single in vitro treatment of melanoma cells at high drug concentration (5 lM) and on the subsequent evaluation of cell vitality and proliferation. The primary end-point was represented by the percentage of live cells, as evaluated through ATP quantification at 24 and 48 h after treatment in comparison with untreated controls. Despite the high dosages used,~60% of the drugs (230 of 349 at 24 h and 195 at 48 h) did not produce significant effects, leaving alive more than 80% of the treated cells (Fig. 1A, top panel, Table S2). Based on these results, we selected the 19 drugs able to reduce the cell number below À2.0 9 standard deviations (SD) from the overall mean viability at 24 h compared with the untreated counterparts (Fig. 1A, bottom panel). As our previous data showed the tumor suppressor role of miR-126 in melanoma (Felli et al., 2013), we evaluated the possibility that miR-126 forced expression might sensitize melanoma cells, potentiating the effects of chemical agents, thus indicating miR-126 restoration as a potential therapeutic option for advanced melanoma. For this reason, a concentration-response analysis was performed with the 19 selected chemical agents on both A375M control and miR-126-transduced cells. As first step, qRT-PCR confirmed miR-126 overexpression in miR-vs TripZ-transduced melanoma cells. Specifically, we obtained~14-fold increase of miR-126-3p and~4fold increase of miR-126-5p after 48 h of induction with doxycycline (Fig. S1). Different starting doses were then chosen for each drug based on the vitality measures obtained in the initial screening, finding in most cases the expected dose-response curves (see Fig. 1B for PIK-75 and Fig. S2 for other drugs). Among the compounds analyzed, PIK-75 was shown as the most effective drug and, even more important, the only displaying a synergistic effect with miR-126. In fact, although similar levels of vitality were displayed by 'drug' and 'drug plus miR-126' treated cells, being the PIK-75 half maximal inhibitory concentration (IC 50 ) 55.05 and 50.68 nM, respectively, the EOB values related to lower doses of PIK-75 (20 and 40 nM) indicated a significant synergistic effect (Fig. 1B, bottom panel).

Identification of miR-126, PIK-75 and vemurafenib synergistic combination for BRAFmutated melanoma cell line treatment
Mitogen-activated protein kinase and PI3K/AKT/ mTOR pathways represent the key signaling cascades involved in melanoma development and progression (Fig. 1C). As already reported (Xiao et al., 2016), miR-126-3p directly targets and represses the PI3KR2 (p85b) subunit of PI3K. Moreover, bioinformaticbased approaches (TargetScan, PicTar and RNA hybrid) indicated PI3KCD (p110d) as an additional putative target of miR-126. To demonstrate this hypothesis, we performed a reporter luciferase assay by cloning the p110d 3 0 UTR downstream of the Renilla open reading frame in the pSiCheck vector. The reporter construct was transfected into the A375M melanoma cell line stably transduced to overexpress miR-126 or in control cells transduced with the empty vector. Results indicated a significant miR-126-dependent inhibition of the luciferase activity, demonstrating its targeting of p110d. Western blot analysis confirmed the down-regulation of both p85b and p110d associated with miR-126 overexpression in the A375M cell line (Fig. S3). Interestingly, PIK-75 plays a regulatory role on the same pathway through the direct repression of the p110a (Zheng et al., 2011). Conversely, vemurafenib is a small molecule, which inhibits the constitutively active MAPK cascade by targeting mutant BRAF. The potentially convergent action of these agents, acting either on the same or on interconnected pathways, prompted us to focus our attention on their possible combination to obtain a strong and not easy to circumvent therapeutic effect.
After proving a PIK-75 and miR-126 synergistic effect (Fig. 1B), we demonstrated in in vitro experiments that the anti-cancer activity played by vemurafenib was enhanced by the overexpression of the oncosuppressor miR-126, with an IC 50 of 730.86 nM for vemurafenib treatment of A375M/miR-126 compared with 1095.65 nM for control A375M cells. Again, at the lower doses of vemurafenib, the EOB results showed the effectiveness of the 'drug plus miR' treatment ( Fig. 2A). Proceeding to evaluate the efficacy of the combinatorial approach of anti-cancer drugs and looking for a treatment capable of overcoming the acquired resistance associated with BRAFi in the majority of melanoma patients, we focused on the possible combination of PIK-75 with vemurafenib.
The possible therapeutic approaches were tested on A375M viability at multiple concentrations, according to a model in which each drug was supplemented either as single agent or in combination. The inhibitory effects were evaluated on cell proliferation and viability according to the Bliss independence model. On these bases, we obtained high synergistic scores with low doses of both drugs: 100 and 250 nM for vemurafenib combined with doses of PIK-75 ranging between 20 and 60 nM (Fig. 2B).
To better dissect the synergistic results obtained by overexpressing the miR-126 precursor, the sole miR-126-3p mimic sequence was transfected in the A375M cell line, looking for the specificity of its role in increasing the drug-dependent actions. Indeed, significant enhancement of PIK-75 and vemurafenib effects, demonstrated by reduced cell viability, was observed in single or combined treatments on A375M cells transiently transduced with miR-126-3p mimic (Fig. S4A, B, right graph). The functionality of miR-126-3p in mediating the combined action of PIK-75 and vemurafenib was confirmed by the sensible reduction of drug effects observed after its silencing (Fig. S4B, left graph). The A375M cell line, selected for its low but detectable level of miR-126 (Felli et al., 2013), represented a good cellular system for both overexpression or silencing experiments.

Effect of single drug treatments on the expression levels of PI3K/AKT and MAPK key proteins
To evaluate the action of PIK-75 and vemurafenib, administered as single drugs, we performed a series of short time in vitro treatments on both A375M control (A375M/TripZ empty vector) and miR-126 overexpressing cells, using increasing doses of either PIK-75 (0.125, 0.25 and 0.5 lM) or vemurafenib (2.5, 5 and 10 lM). After 5 h, cells were harvested and analyzed by western blot for both AKT and ERK phosphorylation levels. Results showed that PIK-75 inhibited AKT phosphorylation in a concentration-dependent manner and that this effect was significantly more marked in A375M/miR-126 than in control cells (Fig. 3A), in line with a direct action of miR-126 on the PI3K/AKT pathway. PIK-75 treatment also caused the cleavage of PARP, a well-known marker of the apoptotic process and the contextual induction of ERK phosphorylation (Fig. 3A). In parallel experiments, vemurafenib induced a concentration-dependent inhibition of ERK phosphorylation and a consequent cyclin D1 downmodulation with the contemporary induction of AKT phosphorylation, without clear differences associated with miR-126 enforced expression (Fig. 3B). These experiments confirmed in our cellular system the expected effects on these key signaling pathways, showing the crosstalk between PI3K/AKT and MAPK pathways and driving the strong increase of pERK when the drug treatment inhibited pAKT and vice versa (Fig. 3A,B). Quantitation of western blot images is reported in Fig. S5A,B.

Effective conditions for dual targeting of PI3K/AKT and MAPK pathways
In view of such a crosstalk between PI3K/AKT and MAPK and looking for the possibility of successfully inhibiting both pathways, we investigated several different PIK-75 and vemurafenib dose combinations. Considering the results obtained by single administration of 0.25 lM PIK-75 and 2.5 lM vemurafenib, the former able to decrease pAKT in the A375M/miR-126 cells and the latter pERK in both A375M/TripZ and A375M/miR-126 cell lines, we initially selected this combination. However, at these concentrations, the co-administration of PIK-75 and vemurafenib was not able concurrently to inhibit AKT and ERK phosphorylation (Fig. S6A,B). PIK-75 and vemurafenib were then used at 0.5 and 5 lM, either as a single drug or in combination. The single agent-based treatments with 0.5 lM PIK-75 or 5 lM vemurafenib confirmed their inhibitory effects on AKT or ERK phosphorylation, respectively. Particularly evident was the reduced phosphorylated fraction of AKT in miR-126 overexpressing cells treated with PIK-75. More important, a significant synergistic result was obtained by the combination as, at these concentrations, the two drugs were capable of inhibiting both the PI3K/AKT and MAPK pathways, simultaneously preventing the compensatory effect usually observed (Fig. 3C). The capacity of these treatments to induce apoptosis was assessed by evaluating the amounts of cleaved PARP and caspase 3. Indeed, PIK-75 administration was confirmed to induce apoptosis strongly and this effect was potentiated by the combined treatment. As expected (see Welsh et al., 2016), cyclin D1 was down-modulated by vemurafenib, in line with the reduction of ERK phosphorylation, whereas it was increased by PIK-75 (Fig. 3D). Quantitation of western blot images is reported in Fig. S5C,D.
Of note, the main effects obtained by PIK-75 treatment, i.e. the reduced AKT phosphorylation and the induction of apoptosis, were more pronounced in miR-126 overexpressing cells. Flow cytometry-based apoptosis quantification in A375M/miR-126 cells treated by PIK-75 or PIK-75 plus vemurafenib indicated the presence of a significantly higher percentage of apoptotic cells when compared with the A375M/TripZ cells treated with PIK-75 (P < 0.01) or PIK-75 plus vemurafenib (P < 0.01; Fig. 3E, left panel). According to our previous results (Felli et al., 2013), we further substantiated the capability of miR-126 enforced expression to favor the apoptotic process of the A375M metastatic melanoma cell line at 0.1% serum culture conditions, even without any drug treatment. As shown in a representative analysis (Fig 3E, right  panel), the percentage of apoptosis in A375M/miR-126 cells kept in low serum was 28.7% compared with 4.7% of control cells. This difference was visible, although less marked, in more permissive culture conditions associated with 10% serum (13.7% vs 3.7% of apoptotic cells in miR-126 and TripZ A375M cells, respectively).

Vemurafenib, PIK-75 and miR-126 triple combination showed a strong synergy on a vemurafenib-resistant cell line
The suppression of MAPK signaling by BRAFi, like vemurafenib, is a therapeutic strategy widely employed in BRAF V600 -mutant melanomas; although very high RRs are obtained, up to 90% of vemurafenib-responding patients develop resistance. Trying to overcome such resistance development, several dual therapies have been approved (Tolcher et al., 2018) and now represent the standard of care.
To evaluate the efficacy of the triple combination vemurafenib/PIK-75/miR-126 in cells resistant to vemurafenib, we established a resistant sub-line (A375M-VR) by continuous exposure of the parental A375M cell line to increasing concentrations of vemurafenib up to the final 10 lM. After selection, the A375M-VR cells were lentivirally transduced to overexpress miR-126. qRT-PCR-confirmed miR-126 overexpression in miR-vs empty vector-transduced melanoma cells. Specifically, we obtained~13-fold increase of miR-126-3p and~3-fold increase of miR-126-5p after 48 h of doxycycline induction (Fig. S7). The persistence of BRAF V600E activating mutation in A375M-VR was verified by sequencing.
We then analyzed how the acquisition of resistance to vemurafenib affected PI3K/AKT and MAPK pathways with or without miR-126 enforced expression. The first unexpected result was the detection of the very low level of total AKT in A375M-VR compared with the A375M counterpart cells either untreated or treated with PIK-75 (Fig. 4B, left panel). To assess whether this result might derive from a different AKT protein stability, we performed the PIK-75 treatment in the presence of the proteasome inhibitor MG-132. As shown in Fig. 4C, we obtained a complete MG132dependent rescue of AKT protein in both control and miR-126 cells treated with PIK-75, suggesting a significant proteasome-dependent degradation of AKT in this condition.
In PIK-75-treated A375M-VR cells, expressing or not expressing miR-126, we observed caspase 3 activation and PARP cleavage at levels significantly higher than those observed in control A375M cells, suggesting that the apoptotic process induced by PIK-75 was stronger in A375M-VR. Finally, probably due to the onset of escape mechanisms of adaptation to chronic vemurafenib exposure, the fraction of phosphorylated ERK was originally very high and was not further increased by PIK-75. Unexpectedly, the high level of cyclin D1 observed in A375M contextual to pERK-increase cells was not visible in A375M-VR cells (Fig. 4B, right panel). Quantitation of western blot images is reported in Fig. S8A-C.

PIK-75 inhibited cell viability in dabrafenibresistant melanoma cell lines
To confirm the wider effectiveness of PIK-75 on cell growth viability, we evaluated its action on melanoma cell lines resistant to dabrafenib compared with their parental counterparts. Dabrafenib, like vemurafenib, inhibits BRAF and is used for metastatic melanoma therapy. Two metastatic melanoma cell lines, A375 and SK-Mel28, were grown at increasing concentrations of dabrafenib up to their final selection (Caporali et al., 2016(Caporali et al., , 2017. Cell viability was evaluated on both parental (A375 and SK-Mel28) and dabrafenib-resistant sublines (A375-R and SK-Mel28-R). As shown in Fig. 4D, PIK-75 was able to inhibit A375 and SK-Mel28 cell growth (IC 50 96.25 and 145.47 nM, respectively) as well as A375-R and SK-Mel28-R sublines (IC 50 135.64 and 225.88 nM, respectively), thus further suggesting this drug as a good candidate for a dual therapeutic application.

Combination of PIK-75 and vemurafenib showed a strong synergy in metastatic melanoma cells derived from patients
As a last step, to strengthen our results, we selected four primitive cell lines derived from metastatic melanoma patients, three (Pt1, Pt2, Pt3) harboring the commonest BRAF mutation (V600E) and one (Pt4) the V600R BRAF mutation. The effects of vemurafenib and PIK-75, as either single or combined treatments, were evaluated on cell viability. As reported in Fig. 5A, whereas the sensitivity to vemurafenib as single agent was extremely diversified (IC 50 from 372.41 to 6124.18 nM), PIK-75 produced a strong effect regardless of the type of BRAF mutation. Of note, vemurafenib was synergistic with PIK-75, as it considerably reduced the PIK-75 IC 50 in all patients' derived cell lines as shown by the values of EOB, particularly for Pt3 and Pt4 (Fig. 5B). Considering miR-126 replacement as a possible therapeutic option, we evaluated the expression level of miR-126-3p in these patients' derived metastatic cell lines (Pt1, Pt2, Pt3, Pt4) by comparing its amount with that expressed by the primary tumor counterpart of Pt4 and the A375M cell line. As expected, the expression level of miR-126-3p in metastatic cells was markedly lower than in primary tumor cells and similar to that of A375M (Fig. 5C).
We then selected Pt1 and Pt2 for further analysis because, showing a very different sensitivity to vemurafenib (IC 50 6124.18 and 399.41 nM, respectively), they could represent the paradigmatic condition of melanoma patients: initial disease regression after BRAFi treatment (Pt2) and subsequent appearance of relapse due to development of resistance (Pt1). After 5 h of drug treatment, we used western blot to analyze modulation of PI3K/AKT and ERK pathways, PARP cleavage and cyclin D1 regulation in these two primitive cell lines (Fig. 5D). Cells from patient Pt1 (less sensitive to vemurafenib) showed a high level of pERK at basic condition, as already observed in A375M-VR cells. Vemurafenib inducing down-modulation of pERK was detected in Pt1 and cyclin D1 in both cell lines (Pt1 and Pt2). As expected, PIK-75 caused a strong cleavage of PARP with induction of ERK phosphorylation, and only the combined treatment, vemurafenib + PIK-75, was able to maintain low levels of both pAKT and pERK and the efficient cleavage of PARP protein (Fig. 5D). Densitometric quantitation of western blot images is reported in Fig. S9A,B.

In vivo studies
The possible synergistic role played by miR-126 in combination with vemurafenib and/or PIK-75 therapeutic administrations was evaluated in vivo in mouse xenograft models. Athymic nude mice were injected with A375M/miR-126 cells or A375M/TripZ control cells and the growth of tumor nodules followed over~4 weeks. When xenograft tumors became palpable, mice were randomly allocated to four arms of treatment consisting in: daily intraperitoneal injection of PIK-75 (2 mgÁkg À1 ); vemurafenib (20 mgÁkg À1 ); their combination; or the vehicle. The effects of the different conditions were evaluated as TGI% compared with mice inoculated with A375M/TripZ. After 3 weeks, in the vehicle group, tumor volumes derived from A375M/miR-126 cells were significantly smaller than those from A375M/TripZ, with 32% of TGI, thus confirming the tumor suppressor function of this miR. An important synergistic effect was observed between miR-126 enforced expression and PIK-75 administration, used either as single agent (TGI: 64% in A375M/miR- EOB at each drug concentration relative to drug effects on patients (Pt1, Pt2, Pt3 and Pt4). Red squares indicate strong antagonistic effect, orange squares strong slight antagonistic effect, light orange squares slight antagonistic effect, blue square additive effect, light blue squares slight synergistic effect, light green squares strong slight synergistic effect, and green squares strong synergistic effect. (C) qRT-PCR analysis illustrating the expression levels of miR-126 on metastatic melanoma patients (Pt1, Pt2, Pt3 and Pt4). Primary Pt4 cell line and A375M were used as controls. Columns, means AE SD of at least three independent experiments. Differences were statistically evaluated by Student's ttest, ***P < 0.001. (D) Representative western blots illustrate the evolution of AKT and ERK phosphorylation and PARP and cyclin D1 expression in two cell lines derived from metastatic melanoma patients (Pt1 and Pt2) exposed to vemurafenib (Vem; 5 lM) or PIK-75 (PIK; 0.5 lM) as a single agent or in combination for 5 h. b-Actin was used as protein loading control. V + P: vemurafenib + PIK-75.
126-derived tumors vs 30% in A375M/TripZ-derived tumors, P < 0.01) or combined with vemurafenib (TGI: 73% in A375M/miR-126-derived tumors vs 53% in A375M/TripZ-derived tumors, P < 0.05; Fig. 6A). Conversely, no significant differences were determined by vemurafenib monotherapy in tumors derived from A375M/TripZ control cells or from A375M/miR-126 cells. During the experiments, no mice in any group lost bodyweight and no major abnormalities were observed in the sacrificed mice, indicating the absence of general toxicity due to the treatments. To confirm the inhibitory effects observed in vitro on PI3K/AKT and MAPK pathways, we analyzed by western blot the key molecules associated with these pathways in the excised xenograft tumors (Fig. 6B). Results confirmed that overexpression of miR-126 was able to reduce the level of AKT phosphorylation in all the observed conditions. In agreement with in vitro results, we also observed a strong reduction of the uncleaved PARP protein and an increase of the activated caspase 3, both markers of the apoptotic process, only in PIK-75 treated mice, which corresponds well to the overall TGI results (Fig. 6A,B). No reduction of ERK phosphorylation was observed (Fig. 6B). Quantitation of western blot images is reported in Fig. S10A.
In line with all these findings, immunofluorescence analysis on serial tumor slides highlighted that the amount of activated caspase 3 clearly correlated with PIK-75 administration and miR-126 overexpression. Caspase 3 activation was even more evident in A375M/miR-126-derived tumors of mice cotreated with PIK-75 and vemurafenib (Fig. 6C, left panel). These findings were confirmed by TUNEL analysis, which showed a significantly enhanced apoptosis in the triple treatment, i.e. miR-126 enforced expression plus vemurafenib-PIK-75 co-administration (Figs 6D and  S10B). An important observation was that, although the effects observed in A375M/miR-126-derived tumors treated with vemurafenib seemed poorer in terms of growth inhibition with respect to control cells (Fig. 6A), we detected higher amounts of apoptosis associated with miR-126 overexpression, as indicated by caspase 3 activation (Fig. 6C, left panel). The evaluation of Ki67, as a marker of cell proliferation, showed a significant reduction of cell growth in the xenograft tumors of all treated groups compared with controls (Figs 6C, right panel, and S10B).
Finally, as an insulin feedback was reported to limit the efficacy of PI3K inhibition in several tumor models (Hopkins et al., 2018), we checked for the effects of the combined PIK-75 plus vemurafenib therapy on the expression of the IRS-1. IRS-1, a key molecule in signal transduction between the insulin receptor and PI3K, was demonstrated to be a target gene of miR-126 in different cancers (Fang et al., 2017). Expression studies, besides confirming IRS-1 down-regulation in miR-126-transduced A375M melanoma compared with TripZ control cells, showed the strong induction of IRS-1 in PIK-75 + vemurafenib cotreated A375M cells, a result suggestive of the association of its activation with drug resistance. Indeed, no up-regulation of IRS-1 was detected in miR-126-overexpressing A375M cell line, thus indicating that the miR-126-dependent inhibition might represent an alternative therapeutic strategy to interfere with the activation of the IRS-1/ PI3K/AKT pathway (Figs 6E and S10C).

Proteome profiling of A375M/TripZ-and A375M/miR-126-derived xenograft tumors
We then sought to elucidate the significance of miR-126 restored expression on the metastatic melanoma cell proteome, identifying sensitive candidate pathways. To this purpose, we analyzed LC-MS/MS, A375M/miR-126-vs A375M/TripZ-derived tumor proteome by reversed phase, with technical and biological reproducibility higher than 80% for both samples. We identified a total of 1230 proteins: 795 proteins were shared by both cell lines, and 294 were exclusive of TripZ and 141 of miR-126 samples. Among those proteins common to TripZ and miR-126 cells, 69 proteins were down-regulated and 34 were up-regulated by miR-126 overexpression (Fig. 7A, Tables S3 and S4).
Looking for the enrichment pathways possibly associated with the inhibitory action of miR-126, we analyzed with PANTHER software the annotations of 363 proteins, whose expression was restricted to A375M/ TripZ or resulted in down-regulation of miR-126 cells, thus leading to a TripZ/miR-126 ratio ≥ 2. The pattern of protein expression repressed or down-regulated by miR-126 overexpression was enriched in cellular and metabolic process clusters of genes (Fig. 7B).
Further analyses were then run using DAVID software. Interestingly, as shown in Fig. 7C, the modulation by miR-126 of several proteins involved in cancer and particularly in melanoma progression (Ernst et al., 2018;Laudisi et al., 2018) is consistent with tumorsuppressing effects of this miR. Interestingly, the KEGG annotation analysis highlighted the mTOR signaling pathway, which represents one of the main signaling pathways activated in cancer through important metabolic changes (Mossmann et al., 2018). This result strengthens the tumor suppressor role of miR-126 and its synergistic effect when associated with AKT and MAPK inhibitors (Fig. S11).

Discussion
An important challenge in clinical oncology is to overcome the acquired resistance often occurring after firstline treatments, a field where malignant melanoma represents a paradigmatic pathology (Johannessen et al., 2013;Welsh et al., 2016). Vemurafenib treatment of patients with BRAF V600 -mutated melanoma (≥ 50%) resulted in a significant prolongation of progression-free survival (PFS), with an RR of 53%. Unfortunately, responses are not durable and most patients invariably relapse within 6-8 months with drug-resistant disease (Savoia et al., 2019). A recent comparative analysis of survival data from selected trials representative for MAP kinase inhibition and immune checkpoint blockade (anti-PD-1 or CTLA-4 monoclonal antibodies) showed the superiority of combined BRAF and MEK inhibitors in the first 6 months after initiation of treatment, whereas immune check point inhibitors were superior at 24 months. This observation, although preliminary, is consistent with the finding that as many as 60% of patients display primary resistance to immunotherapy but that a subpopulation of patients obtains a long-term PFS. However, both therapeutic strategies unfortunately show a common phenomenon of acquired resistance and disease progression in a short period of time in a large proportion of patients (Najem et al., 2017;O'Donnell et al., 2016;Ugurel et al., 2017).
Our recent studies have identified miR-126, already known to play an inhibitory role on the PI3K/AKT pathway (Xiao et al., 2016;Zhang et al., 2013), as a tumor-suppressor in melanoma (Felli et al., 2013), demonstrating the existence of an auto-regulatory loop that, connecting the tumor suppressor miR-126 with the onco-miR-221/222, functions in melanoma development and progression (Felli et al., 2016). Also, significant modulations of miR have been associated with therapeutic treatments, indicating these short non-coding RNA as additional factors for patient classification and clinical choices. Interestingly, miR-126 was shown to be among those miR modulated by a combined treatment with temsirolimus (anti-mTOR) and bevacizumab (anti-VEGF) (Wagenseller et al., 2013).
Based on these data and seeking an effective therapeutic approach for metastatic melanoma, we demonstrated the potential role of miR-126 as a sensitizing agent able to increase the efficacy of more traditional drugs. Combining drugs operating through different mechanisms of action is a therapeutic approach used to avoid the occurrence of resistance in oncology, but also against infectious agents easily developing drug resistance, such as Mycobacterium tuberculosis (Fischbach, 2011;Worthington and Melander, 2013). To find candidates for effective combinations, we screened a library of 349 anti-cancer compounds, including several target agents already in clinical use or trials, looking for antiproliferative effects on the A375M metastatic melanoma cells. Among many others, we selected PIK-75 for in-depth evaluations in view of its high potency, but also for its inhibitory role on the p110a subunit of PI3K (Han and Zhang, 2010;Zheng et al., 2011), which in turn binds p85b, a second subunit directly targeted by miR-126 (Xiao et al., 2016) (Fig. 1C). Indeed, this PIK-75 selective inhibition could possibly induce less adverse side effects compared with pan-PI3K inhibitors (PI3Ki) (Ciraolo et al., 2011). Actually, PIK-75 is a potent inducer of apoptosis in primary human AML cells, enhancing the survival of mice with established AML without significant toxicity (Thomas et al., 2013). Besides PIK-75, in combination therapy we utilized vemurafenib, currently used in patients with BRAF-mutated advanced melanoma (Kim and Cohen, 2016). The treatment of metastatic melanoma cells with PIK-75 or vemurafenib as single agents resulted in growth inhibition, but it was the combination of the two that produced a stronger synergistic effect, as indicated by statistical models (Fig. 2B). The molecular analysis after single-agent treatment revealed a strong cross-talk between PI3K/ AKT and MAPK pathways, as shown by the inverse balance between pAKT and pERK levels (Fig. 3A,B). Indeed, we found the correct conditions for PIK-75 and vemurafenib combined administration able to inhibit AKT phosphorylation, avoiding simultaneous MAPK activation. The main effect of this treatment was a strong induction of the apoptotic process. It is important to point out that the enforced expression of miR-126 increased PIK-75 activity in both single and combined treatments (Fig. 3C,D).
We then extended the evaluation of PIK-75 supplementation to a vemurafenib-resistant cell line (A375M-VR), confirming its effectiveness in impairing cell viability through the induction of the apoptotic process, especially when associated with miR-126 overexpression. PIK-75 effectiveness was confirmed on two additional metastatic melanoma cell lines harboring the BRAF V600E mutation, either in their standard conditions or selected for dabrafenib resistance (A375 vs A375-R and SK-Mel28 vs SK-Mel28-R). The same synergistic effects were obtained by treating four early passage cell lines obtained from metastatic melanoma patients (Pt1; Pt2; Pt3 and Pt4) with PIK-75 and vemurafenib. Again, the analysis of the key molecular pathways confirmed that the combined treatment inhibits PI3K pathway and prevents MAPK reactivation (Fig. 5). Finally, the validity of this approach, which is able simultaneously to target PI3K and MAPK in the presence of miR-126, was confirmed in a mouse xenogeneic model of human melanoma. In tumor samples, we showed reduced proliferation and increased apoptosis as indicated by Ki67 and caspase 3 modulations. According to recently published data (Hopkins et al., 2018), miR-126 also appeared able to disrupt the insulin regulatory loop by reducing the expression of the key factor IRS-1, reported as a target gene of miR-126 (Fang et al., 2017). In addition, proteomic analysis showed the effectiveness of miR-126 action on the mTOR pathway, one of the key controllers of cancer metabolism (Fig. 7). Overall, our results showed the capability of miR-126 to enhance the efficacy of PI3K and/or MAPK inhibitors in line with what observed by reducing the systemic insulin response through a ketogenic diet (Hopkins et al., 2018). Since, at present, anti-cancer drugs targeting IRS-1 protein are not available (Reuveni et al., 2013), our data may represent preclinical proof of concept for miR-126-based direct targeting of IRS-1, increasing the possible role of this miR in cancer therapy, including drug-resistant melanoma.
There is growing evidence supporting the efficacy of combination therapies and the reduction of paradoxical activation of the MAPK pathway in melanoma Nazarian et al., 2010;Tolcher et al., 2018). Nonetheless, one of the major problems is the severe toxicity associated with dual targeting therapeutics, particularly when based on a horizontal action, such as the use of MAPK and PI3K/AKT inhibitors (Shimizu et al., 2012). In this clinical setting, the restored expression of miR-126 might have additional value. Indeed, novel more efficient and specific transfer technologies will favor miR-126 administration.

Conclusions
Besides showing the higher efficacy of vemurafenib and PIK-75 cotreatment, our results support the central role of miR-126 in melanoma biology. Indeed, the restored expression of this miR in advanced melanoma could act as a sensitizer agent allowing reduced doses of the combined therapeutics, in turn leading to decreased frequency and severity of adverse events without limiting the benefits. In addition, miR-126 direct targeting of the IRS-1 may represent an efficient method to impair the insulin feedback loop reported to limit the efficacy of PI3K inhibition. Considering the excellent opportunity to use the PI3K/AKT inhibitors, miR-126 might represent a novel promising tool for combination therapies.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. qRT-PCR analysis of miR-126-3p and miR-126-5p on A375M melanoma cell line. Fig. S2. Dose-response effects on A375M TripZ-and miR-126-transduced cells. Fig. S3. Evaluation of p85b and p110d direct targeting by miR-126. Fig. S4. Specific synergistic role of miR-126-3p in combination with vemurafenib and/or PIK-75.      .  Table S1. List of the 349 compounds of the Selleckchem anti-cancer library, including names, targets and a brief description of the drugs. Table S2. Cell viability at 24 and 48 h of treatment with 349 anti-cancer compounds. Table S3. Proteome profiling of A375M/TripZ-and A375M/miR-126-derived xenograft tumors: complete list of the identified proteins. Table S4. Quantitative profile of differentially expressed proteins by comparing TripZ and miR-126.