Inhibition of ERK5 Elicits Cellular Senescence in Melanoma via the Cyclin-Dependent Kinase Inhibitor p21

Malignant melanoma is one of the most aggressive cancer and its incidence is increasing worldwide. Early-stage disease can be cured in the majority of cases, while late-stage melanoma is a highly lethal disease (1). Genomic sequencing studies of melanomas have found mutations in BRAF, NRAS, and NF1 that hyperactivate the mitogen-activated protein kinases (MAPK) extracellular signal-regulated kinase 1 and 2 (ERK1/2), thus supporting cell proliferation (2, 3). These findings allowed the development of BRAF- and MEK1/2-targeting drugs that, together with immunotherapy, increased the overall survival of melanoma patients (4). However, either the lack of responsiveness to immunotherapy or the presence of intrinsic and acquired resistance to BRAF-MEK1/2 inhibitors limit the benefits of available therapies (5, 6).

ERK5 is involved in cell survival, antiapoptotic signaling, proliferation and differentiation of several cell types, as well as angiogenesis (7), and plays a relevant role in the biology of cancer (8). ERK5 activation is achieved through MEK5-dependent phosphorylation, that stimulates ERK5 nuclear translocation, a key event for cell proliferation (9, 10).

Cellular senescence, a permanent state of cell-cycle arrest, is widely recognized as a potent tumor-suppressive mechanism (11, 12), so that induction of senescence is included among the possible strategies to fight cancer (12, 13). Melanoma is frequently characterized by the loss of the pathways supporting cellular senescence that would prevent tumor growth and progression (14). We previously showed that ERK5 inhibition reduces the growth of melanoma, determining a block of the cell cycle rather than inducing apoptosis (15). Because MAPK are frequently involved in cellular senescence (16, 17), we investigated the effects of ERK5 targeting on senescence with the aim to identify a novel therapeutic strategy for melanoma.

A375 melanoma cells (18) were obtained from ATCC; SK-Mel-5 melanoma cells (19) were kindly provided by Dr. Laura Poliseno (CRL-ISPRO, Pisa, Italy). SSM2c melanoma cells were already described (Supplementary Table S1; ref. 20). Cells were maintained in DMEM supplemented with 10% heat-inactivated fetal bovine serum (FBS), 2 mmol/L glutamine, 50 U/mL penicillin, and 50 mg/mL streptomycin (EuroClone). Cell lines were yearly authenticated by cell profiling (Promega PowerPlex Fusion System Kit; BMR Genomics s.r.l). Presence of Mycoplasma was periodically tested by PCR.

ERK5 inhibitor XMD8–92 (21) and MEK5 inhibitor BIX02189 (22) were from MedChemExpress LLC.

Total cell lysates and nuclear–cytoplasmic fractions were obtained using Laemmli buffer or hypotonic buffer, respectively, as reported previously (23, 24). Proteins were separated by SDS-PAGE and transferred onto Hybond PVDF (GE Healthcare) by electroblotting. Infrared imaging (Odyssey, Li-Cor Bioscience) was performed. Images were quantified with ImageJ software. Antibodies are in Supplementary Table S2.

TRC1.5-pLKO.1-puro lentiviral vectors (Supplementary Table S3) were produced in HEK293T cells as reported previously (25). Transduced cells were selected with 2 μg/mL puromycin for at least 72 hours.

Total RNA was isolated using RNeasy Mini Kit (Qiagen), and mRNA expression evaluated with Affymetrix Clariom-S Human Genechip following the manufacturer's instructions. Transcriptome analysis console (TAC) software was used (fold change >1.5/<−1.5 and P ≤ 0.05) to identify differentially expressed genes (DEG). Most enriched pathways were identified by Meta-analysis of DEG using WebMeV (Multiple Experiment Viewer).

Total RNA was isolated using TRIzol (Life Technologies). cDNA synthesis was carried out using ImProm-II Reverse Transcription System, while quantitative PCR (qPCR) was performed using GoTaq qPCR Master Mix (Promega Corporation). For primer sequences see Supplementary Material. QPCR was performed using CFX96 Touch Real-Time PCR Detection System (Bio-Rad). mRNA expression was normalized to GAPDH and 18S mRNA.

Cell-cycle phase distribution (propidium iodide staining) was determined as previously reported using a FACSCanto (Beckton Dickinson; ref. 26).

Fourteen days after lentiviral transduction, medium was replaced with DMEM/10% FBS. Conditioned media (CM) were harvested after 72 hours. Chemokine expression was measured in CM using RayBiotech Quantibody array (RayBiotech) following the manufacturer's instructions. Each antigen was measured in quadruplicate and relative fluorescence intensity determined using ImageJ software. Results were normalized for protein content in lysates.

Cell viability was measured by 3-(4,5-dimethylthiazol-2-yl)-3,5-diphenyltetrazolium bromide (MTT) assay. Cells were seeded in 96-well plate in DMEM/10% FBS. After 24 hours, medium was replaced with CM and cells further incubated for 72 hours. MTT (0.5 mg/mL) was added during the last 4 hours. Plates were read at 595 nm using a Microplate reader-550 (Bio-Rad). For neutralization experiments, control isotype IgG or neutralizing antibodies (Supplementary Table S2) were added to CM prior to administration to cells.

Cells were incubated for 72 hours and then fixed (2% formaldehyde, 10 minutes, room temperature). After washing, senescence-associated (SA) βGal staining solution (X-gal 1 mg/mL, 40 mmol/L citric acid, 5 mmol/L C₆FeK₄N₆, 5 mmol/L C₆N₆FeK₃, 150 mmol/L NaCl and 2 mmol/L MgCl₂; pH 6.1 -SK-Mel-5- or 5.9 -A375, SSM2c; ref. 27) was added (16 hours, 37°C). Senescent cell quantification was performed by counting SA-βGal–positive (blue) cells in 10 random images/well taken using a brightfield microscope. An average of 800 cells/condition were counted. In other experiments, cell area was evaluated with ImageJ software.

Formalin-fixed paraffin-embedded (FFPE) sections from archival xenografts established with shNT or shERK5–2 A375 or SSM2c cells, or with A375 cells from XMD8–92 (25 mg/kg)- or vehicle (2-hidroxypropyl-β-cyclodextrin 30%)-treated mice were used (15). Experiments had been approved by the Italian Ministry of Health (authorization no. 213/2015-PR) and were in accordance with the Italian ethic guidelines and regulations. Sections (3 μmol/L thick) were deparaffinized and stained with Sudan Black Blue (SBB; Bio-Optica) to reveal lipofuscin (28, 29), and counterstained with Nuclear Fast Red (NFR, Bio-Optica). Immunohistochemistry (IHC): After citrate buffer antigen retrieval, staining was performed with the UltraVision LP Detection System HRP Polymer Kit (Thermo Fisher Scientific) following the manufacturer's instructions. Sections were incubated overnight at 4°C with primary antibodies (Supplementary Table S2) and (3,3′-diaminobenzidine; Thermo Fisher Scientific; DAB) used as a chromogen. Sections were counterstained with hematoxylin.

Cells were treated for 72 hours and then seeded in p60 dishes in DMEM 10% FBS. Colonies (i.e., more than 50 cells) were counted following crystal violet staining after 7 (A375) or 10 (SSM2c) days.

Cells were plated on glass coverslips and incubated for 72 hours in DMEM/2,5%FBS, and fixed with 4% paraformaldehyde (10 minutes, room temperature). Cells were permeabilized (0.2% Triton X-100) and saturated with 10% horse serum in PBS/1% BSA for 45 minutes. Incubation with primary antibody (overnight, 4°C) and with Cy2- or Cy3-labeled secondary antibodies was performed. Cell nuclei were labelled with DAPI (Invitrogen). Images were analyzed with a confocal inverted microscope equipped with a Nikon S Fluor 60× immersion oil objective (Nikon Instruments).

Data represent mean or ± SD values calculated on at least three independent experiments. P values were calculated using Student t test (two groups) or one-way ANOVA (more than two groups). P < 0.05 was considered statistically significant.

We previously showed that ERK5 inhibition reduces proliferation of melanoma cells in vitro and growth of melanoma xenografts in vivo (15). To deepen the role of ERK5 in melanoma growth, we performed transcriptomic experiments in BRAFV600E A375 and SK-Mel-5 melanoma cells after ERK5 KD with two different shRNAs (Fig. 1A). DEG analysis showed that the most impacted pathways dealt with cell-cycle regulation. Interestingly, cellular senescence was the tenth most impacted pathway (0.0032 degree of enrichment; Fig. 1B; Supplementary Table S4). We also identified several senescence-related genes, whose expression was significantly changed upon ERK5 KD (Table 1).

We then verified whether ERK5 inhibition elicits cellular senescence. Both SK-Mel-5 and A375 ERK5-KD cells showed a marked increase (20%–30%) of SA-βGal–positive cells, when compared with naïve or to control nontargeting shRNA-transduced cells (Fig. 1C and D). The same effects were observed in triple wild-type SSM2c cells (Fig. 1E). In all three cell lines, ERK5 KD determined an increase of the cell area (Fig. 1F), a morphologic change typical of senescent cells (16). Importantly, the ERK5 inhibitors XMD8–92 (Fig. 2A–C) and JWG-045 (Supplementary Fig. S1A and S1B), used at concentrations with negligible off-target effects (15), as well as JWG-071 (Supplementary Fig. S1A), a more specific inhibitor of ERK5 over bromodomain-containing proteins (30, 31), were able to induce cellular senescence. Interestingly, the MEK5 inhibitor BIX02189 did not induce cellular senescence (Supplementary Fig. S1A–S1C). Pharmacologic inhibition of ERK5, which lasted for the entire duration of the experiments (Supplementary Fig. S1D), determined an increase of cell area (Fig. 2D), and of senescence-associated heterochromatin foci (SAHF)-positive cells (Fig. 2E and F). The latter effect was also observed after ERK5 KD (Supplementary Fig. S1E). The BRAFV600E inhibitor vemurafenib, but not the MEK1/2 inhibitor trametinib, determined an increase of senescent cells similarly to XMD8–92, and in combination with the latter further increased the percentage of senescent cells (Supplementary Fig. S1F).

The occurrence of cellular senescence in vivo was then investigated performing lipofuscin staining on archival xenografts of A375 and SSM2c cells transduced with shNT- or shERK5-encoding lentiviral vectors, the latter of which drastically reduced tumor growth (Supplementary Fig. S2A and S2B; ref. 15). In A375 and SSM2c xenografts, the amount of lipofuscin and the percentage of lipofuscin-stained area were markedly higher in ERK5-KD than in control xenografts (Fig. 3A and B). Similarly, systemic administration of XMD8–92, which induced a marked reduction of tumor growth (Supplementary Fig. S2C; ref. 15), determined a robust increase of lipofuscin in A375 xenografts with respect to vehicle-treated mice (Fig. 3C).

We then deepened the effects of ERK5 inhibition on cell-cycle progression that is impaired in senescent cells characterized by an irreversible block of proliferation. ERK5 KD determined an increase of cells in G₀–G₁ or G₂–M phase in A375 or SSM2c cells, respectively (Fig. 4A and B). This increase is robust when taking into consideration that 20% to 30% of the population undergoes senescence upon ERK5 KD (Fig. 1C–E). ERK5 KD determined a reduction of PCNA protein, in keeping with a reduced proliferation (Supplementary Fig. S3). Moreover, exposure to mitogens (i.e., 10% FBS), that increased the percentage of vehicle-treated cells in S phase, did not affect cell-cycle phase distribution in XMD8-92–treated cells (Fig. 4C and D). Of note, XMD8–92 blocked cell-cycle progression with a significant increase of cells in G₀–G₁ or G₂–M phase in A375 and SSM2c cells, respectively, in line with ERK5-KD cell results (Fig 4A and B) and with our previous data (15). Colony assay experiments using 72 hours XMD8-92–treated cells demonstrated their irreversible inability to proliferate (Fig. 4E and F).

To deepen the impact of ERK5 inhibition on cell-cycle regulators involved in senescence, we first revealed that ERK5 KD determined a reduction of RB phosphorylation and an increase of the cyclin-dependent kinase inhibitor (CDKi) p21 protein level in A375 and SSM2c cells (Fig. 5A), both representing the most used markers of cellular senescence. Furthermore, we found a reduced expression of cyclin E, D1, and A2 (Fig. 5A), in keeping with a reduced proliferation. Interestingly, either genetic (Fig. 5A and B) or pharmacologic (Fig. 5C and D) ERK5 inhibition increased the amount of the CDKi p14 protein. The increased protein levels of p21 (Fig. 5E and F) and of p14 (Fig. 5G and H) upon ERK5 KD were confirmed in vivo using A375 and SSM2c xenografts. Of note, the increase in the percentage of p14 positive cells in ERK5-KD A375 xenografts, although significant, was relatively low (Fig. 5G), and, accordingly, the increased amount of p14 protein was appreciable in nuclear extracts only (Fig. 5A–C). Consistently, we observed an increase of p21 and a decrease of PCNA-positive A375 cells after treatment with XMD8–92 in vivo (Supplementary Fig. S4A and S4B). Upon ERK5 KD, we also confirmed the increase of p27 and p53 protein levels, and the reduction of RBL-1 mRNA, identified by the transcriptomic analysis (Table 1). Moreover, the increased phosphorylation of CDK1 at Y15 seems to support the observed reduction of phosphatase CDC25a expression (Supplementary Fig. S4C–S4E).

Among the cell-cycle regulators, whose expression is markedly deregulated upon ERK5 inhibition, we focused on p21 in order to deepen the mechanism of ERK5 KD-associated senescence. Of note, p53 was found increased in the transcriptomic data (and confirmed at the protein level), and p21 is the effector of the majority of p53-induced biological effects including cellular senescence. Genetic inhibition of p21 using two different shRNAs (Fig. 6A) halved the percentage of A375 cells undergoing cellular senescence following treatment with XMD8-92 or H₂O₂ (Fig. 6B), revealing a key role of p21 in ERK5-dependent cellular senescence. The observed reduction of the percentage of cells undergoing senescence upon p21 KD was robust, taking into consideration that XMD8-92 induced a slight but significant increase of p21 protein level in p21 KD cells (Fig. 6C; compare lanes 5 and 6 vs. 2 and 3, respectively). In this respect, the lack of effect of BIX02189 on cellular senescence (Supplementary Fig. S1A–S1C) is in keeping with the ineffectiveness of this compound in increasing p21 protein level (15), while XMD8-92 induced a dose dependent increase of cellular senescence and p21 protein (Supplementary Fig. S5A–S5C). The above data indicate that cellular senescence induced by ERK5 inhibition in melanoma cells relies, at least in part, on p21.

One of the key features of senescent cells is the senescence-associated secretory phenotype (SASP; ref. 32). Transcriptomic data showed that SASP (0.0020 degree of enrichment, Supplementary Table S4) and SASP-related genes (Table 1) were significantly modulated upon ERK5 KD. Among the latter, transcriptomic analysis indicated that CXCL1/GROalpha and CXCL8/IL8, two of the most prominent SASP products (33), and CCL20/MIP3alpha mRNA were significantly upregulated in both A375 and SK-Mel-5 ERK5-KD cells (Fig. 7A). IL6, another key component of the SASP, was not altered upon ERK5 KD (Fig. 7A). QPCR experiments confirmed that lack of increase of IL6 mRNA in the same experimental settings in which CXCL8 mRNA was significantly increased in A375, SK-Mel-5, and SSM2c ERK5-KD cells (Fig. 7B). Importantly, protein array in CM (Fig. 7C) confirmed that CXCL1, CXCL8, and CCL20 protein levels were significantly increased in ERK5-KD A375 cells. Interestingly, publicly available datasets from the cBioPortal for Cancer Genomics (34, 35) confirmed the coexpression of the above-mentioned SASP products in patients with melanoma (Supplementary Fig. S6A and S6B). Indeed, the amount of CXCL1/GROalpha mRNA showed a robust positive correlation with that of CXCL8 (0.58 Spearman correlation; P = 6.68e^–41), and with that of CCL20 (0.49 Spearman correlation; P = 1.34e^–27; Skin Cutaneous Melanoma TCGA, PanCancer Atlas; Supplementary Fig. S6A). The positive correlation between the mRNA level of CXCL1/GROalpha and of CXCL8 was confirmed (0.35 Spearman correlation; P = 0.028) in another data set (Metastatic Melanoma, DFCI, Science 201; Supplementary Fig. S6B; ref. 36).

SASP products may elicit either pro- or antiproliferative effects on tumor cells (29). To discriminate between these possible opposite outcomes, we tested CM from A375 and SK-Mel-5 cells and found that CM recovered from ERK5-KD cells markedly reduced the viability of either A375 or SK-MEL-5 cells with respect to CM from shNT cells (Fig. 7D). The proof that CXCL1, CXCL8, and CCL20 are among the SASP products responsible for the reduced proliferation was obtained using neutralizing antibodies (Fig. 7E). Indeed, each blocking antibody was able to partially rescue melanoma cell proliferation.

Cellular senescence is considered a potent suppressive mechanism of tumorigenesis, and therapies that enhance senescence, besides promoting a stable cell growth arrest, may stimulate the activation of the antitumor immune response. Based on that, prosenescence molecules are actively sought after in view of their potential use as antineoplastic treatments (12, 13). In this study, we demonstrated a prominent role of ERK5 in cellular senescence in human melanoma. In BRAF-mutated and wild-type melanoma cells and xenografts, ERK5 inhibition induced indeed marked cellular senescence and production of several soluble mediators involved in the SASP. Mechanistically, we demonstrated that ERK5-dependent senescence is mediated by the CDKi p21.

The induction of cellular senescence upon ERK5 inhibition was demonstrated using different approaches. Senescence emerged first, in transcriptomic experiments, as one of the most impacted pathways upon ERK5 KD in BRAFV600E-mutated melanoma cells. Besides cellular senescence, ERK5 inhibition significantly impacted oxidative stress induced senescence and the SASP. In this respect, ERK5 had been previously reported to be activated in colorectal adenocarcinoma cells undergoing methotrexate-induced senescence (37). We demonstrated the occurrence of cellular senescence by showing that both genetic and pharmacologic inhibition of ERK5 determined a marked increase of SA-βGal positivity in melanoma cells harboring wild-type or oncogenic BRAF, as well as of cellular area and of SAHF. In addition, we found that the expression of a number of genes known to regulate cell cycle and cellular senescence, including CDC25a, RBL-1, p27, and p53, was altered upon ERK5 KD.

Cellular senescence upon ERK5 inhibition was also demonstrated in vivo. Indeed, in A375 and SSM2c xenografts, ERK5 KD determined a robust increase of lipofuscin, an indicator of cellular senescence in vivo (28). More importantly in view of a possible translation to the clinics, similar effects were observed in melanoma xenografts of XMD8-92–treated mice. The occurrence of a block of cell-cycle progression, an event invariably linked to cellular senescence (16), was supported by the fact that ERK5-KD cells underwent an accumulation in G₀–G₁ or G₂–M phase, depending on the cell line, and cells treated with XMD8-92 showed reduced ability to respond to mitogens and did not form colonies upon replating. Consistently, ERK5 inhibition determined a reduction of RB phosphorylation and cyclin protein levels, as well as an increase of p21 protein. This CDKi was identified as a prominent player of ERK5-dependent cellular senescence. Indeed, the amount of p21 protein was increased following ERK5 pharmacologic inhibition, whereas p21 genetic inhibition reverted the prosenescence effect of XMD8-92 treatment. While the involvement of p21 in ERK5-mediated biological effects has been widely reported (21, 38), the role of p21 in ERK5-mediated cellular senescence is a novel finding.

Cellular senescence may be accompanied by the SASP, which consists of secretion of a number of soluble factors in the surrounding microenvironment. SASP determines both beneficial and deleterious biological outcomes, which makes of senescence a double-edged sword with respect to cancer onset and development (32, 39). In this work, we identified a number of SASP-related chemokines, including CXCL1, CXCL8, and CCL20, that are markedly increased in CM of ERK5-KD cells undergoing cellular senescence. Our data showed that these CM markedly reduced the viability of melanoma cells, and that the treatment with neutralizing antibodies for each of the above chemokines rescued the proliferation of melanoma cells. In this respect, despite the growth-promoting functions of CXCL8, CXCL1, and CCL20 reported so far (40–44) may appear in contrast with the antiproliferative effects elicited by CM from ERK5-KD cells, it has been demonstrated that senescent cells may activate a self-amplifying secretory network in which CXCR2-binding chemokines (i.e., CXCL1 and CXCL8) reinforce growth arrest (45). As regard the link between ERK5 inhibition and SASP occurrence, multiple different nuclear and cytoplasmic factors have been shown to trigger SASP, including chromatin remodeling and SAHF formation. In this respect, it is worth pointing out that ERK5 KD determined a profound remodeling of the mRNA levels of a number of histone variants, as well as SAHF increase. However, the molecular mechanism linking ERK5 KD to chemokine regulation remains to be investigated.

Another relevant finding of our study is the identification of prosenescence drugs that may be exploited for cancer treatment. Indeed, XMD8-92, JWG-045, and JWG-071 elicited robust senescence in melanoma cells. The fact that XMD8-92, extensively showed to reduce tumor growth in vitro and in vivo (8), has off-target effects (46) does not reduce the possibility to exploit this small-molecule inhibitor as a prosenescence drug. Interestingly, the prosenescence effect of XMD8-92 was increased when used in combination with vemurafenib, that we found able to induce senescence in melanoma cells, in keeping with a previous report (47). Thus, combined targeting of BRAF and ERK5, besides reducing tumor growth more efficiently than single treatments (15), supports prosenescent signals. These effects are in line with the well-established fact that, either chemo and radio, as well as targeted therapies engage a senescence response as part of the outcome (13). However, if this combination is more effective in preventing resistance mechanisms with respect to BRAF/MEK inhibitors alone remains to be investigated.

No disclosures were reported.

A. Tubita: Data curation, formal analysis, investigation, writing–original draft, writing–review and editing. Z. Lombardi: Data curation, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. I. Tusa: Data curation, formal analysis, supervision, writing–original draft, writing–review and editing. A. Lazzeretti: Formal analysis, methodology. G. Sgrignani: Formal analysis. D. Papini: Formal analysis. A. Menconi: Formal analysis. S. Gagliardi: Investigation. M. Lulli: Data curation, investigation. P. Dello Sbarba: Resources, writing–review and editing. A. Esparís-Ogando: Resources, writing–review and editing. A. Pandiella: Resources, writing–review and editing. B. Stecca: Investigation, writing–review and editing. E. Rovida: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, writing–original draft, project administration, writing–review and editing.

The work in E. Rovida's lab was supported by grants from Associazione Italiana per la Ricerca sul Cancro (AIRC, IG-15282 and IG-21349), by Ente Fondazione Cassa di Risparmio di Firenze (ECRF), and Università degli Studi di Firenze (Fondo di Ateneo ex-60%). A. Tubita was supported by a “Carlo Zanotti” Fondazione Italiana per la Ricerca sul Cancro (FIRC)-AIRC fellowship (ID-23847).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.