PLK1 and NOTCH Positively Correlate in Melanoma and Their Combined Inhibition Results in Synergistic Modulations of Key Melanoma Pathways

Melanoma is one of the most aggressive types of cancer and extremely difficult to treat when metastatic, with a median overall survival (OS) of less than 1 year. In the United States, 100,350 new melanoma cases and 6,850 melanoma-related deaths are predicted in 2020 (1). Current treatment options against metastatic melanoma have either failed to achieve significant response or the responses are short-lived because of the emergence of resistance to therapy in most patients (2). In melanoma, tumor progression and metastasis are complex processes involving multiple cellular events, including cell proliferation, enhanced survival, migration, and invasion (3), suggesting a need for simultaneous targeting of multiple signaling pathways to effectively manage this deadly neoplasm. Therefore, instead of monotherapeutic approaches with modest therapy responses, many researchers are now shifting their attention toward combinatorial approaches to overcome resistance, and there is growing interest in the development of combinations of molecularly targeted therapies to obtain superior antitumor response (4).

Mammalian polo-like kinase 1 (PLK1) is a serine/threonine kinase that plays key roles in cell proliferation with critical regulatory functions in mitosis and cytokinesis (5). We have previously shown that PLK1 is overexpressed in melanoma cells, as well as clinical melanoma tissues, and lentiviral short hairpin RNA or small molecule–targeted inhibition of PLK1 resulted in significant (i) decrease in growth and clonogenic survival, (ii) induction of G₂–M-phase cell-cycle arrest, (iii) increase in apoptosis, and (iv) altered metabolic regulation in melanoma cells (6–8). Another study from our laboratory demonstrated that numb, an antagonist of the NOTCH pathway, regulates stability and localization of PLK1 and is required for transit through mitosis (9). Thus, based on the oncogenic functions of PLK1 in melanoma and other cancer types, inhibition of PLK1 has been suggested as a potential strategy for cancer therapy. Several studies have evaluated PLK1 inhibitors in both preclinical and clinical studies (10). Volasertib (BI 6727) is one of the most specific, commercially available ATP-competitive PLK1 inhibitors, with an IC₅₀ of 0.87 nmol/L, and has been shown to induce mitotic arrest and apoptosis (10). In addition, a recent phase I clinical trial of BI 6727 in solid tumors resulted in a partial response in patients with melanoma, and found that BI 6727 was generally well-tolerated with a favorable pharmacokinetics profile (11). These studies suggest that further studies evaluating BI 6727 (or other PLK1 inhibitors) in conjunction with other agents may provide effective antimelanoma therapies.

The NOTCH pathway is a highly evolutionally conserved molecular pathway that plays important roles in cell fate determination, proliferation, differentiation, and survival (12). NOTCH signaling is deregulated in human hematologic malignancies and solid tumors (12). In humans, there are four NOTCH receptors (NOTCH1–4), and upon each receptors' activation, the NOTCH intracellular domain is cleaved and translocated into the nucleus where it functions as a transcriptional regulator. Interestingly, the NOTCH1 pathway is activated in human melanoma and has been shown to enable primary melanoma cells to gain malignant phenotype (13). Recently, NOTCH1 signaling in melanoma has been shown to facilitate tumor immune escape and promote melanoma progression via TGFβ1 secretion (14). These above-discussed studies provided us the scientific premise of this investigation, where we hypothesized that PLK1 and NOTCH pathways may correlate with each other and the simultaneous inhibition of PLK1 and NOTCH may provide a superior antiproliferative response against melanoma. Overall, our data provide a basis for potential use of concomitant inhibition of PLK1 and NOTCH for melanoma management.

We used a human melanoma tissue microarray (TMA) constructed in-house under the University of Wisconsin (Madison, WI; UW) Institutional Review Board–approved protocol (described previously; ref. 15) and processed and analyzed it as depicted in Fig. 1A. Immunostaining was performed at the UW Translational Research in Pathology Core Facility using a Lab Vision Autostainer 360 (Thermo Fisher Scientific). For immunostaining, the TMA slide was heated at 60°C for 30 minutes, and then deparaffinized and rehydrated. Slides were then pretreated for 20 minutes with citrate buffer (pH 6.0) in Labvision PT module, rinsed with TBS-T, and loaded into Autostainer 360. This was followed by rinsing the slides with TBS-T again and blocking with Sniper (BioCare Medical) for 10 minutes. The slides were then sequentially incubated with antibodies for NOTCH1, Alexa Fluor 555 secondary, PLK1, Alexa Fluor 647 secondary, and Zenon 488–labeled S100 with TBS-T rinses between incubations. Details of antibodies used in the study are provided in Supplementary Table S1. Stained slides were removed from autostainer and rinsed in deionized water before coverslipping with ProLong Gold with DAPI and allowing to cure for at least 24 hours before imaging.

Next, Vectra automated quantitative tissue imaging system coupled with an automated slide scanner and state-of-the-art software (Nuance and InForm; PerkinElmer) was used to image the TMA. This system takes advantage of the unique spectral characteristics (curve) of each chromogen via creation of a spectral library from single-chromogen slides to determine the relative contribution of each chromogen to the stained area on a slide. The stained TMA was imaged with the Vectra slide scanner, using a scanning protocol that was created on the basis of the core size and layout. Nuance software was used to develop the spectral library to be imported into the InForm software. For each slide, an 8-bit image cube from each of the TMA tissue cores was acquired and the InForm advanced image analysis software was used to segment tissues (melanoma vs. others) and cells (nucleus vs. cytoplasm) to analyze protein levels. Then, the target signals were quantified within the tissue and subcellular compartment(s) after unmixing the spectral curves with InForm software and removing signal noise and cross-talk. Continuous signal intensity data (mean optical density per pixel) were generated for each TMA core or cellular compartment and were used to assess IHC staining levels on a cell-by-cell and subcellular basis.

The statistical analysis of TMA data was performed in R, a publicly available programming language and environment, using freely available analysis “packages” of commands. Protein expression in the nucleus and cytoplasm was considered paired data, and thus, was compared using a nonparametric Wilcoxon signed-rank test using ggpubr package (v0.4.0) with P < 0.05 indicating a significant difference of such protein expression between nucleus and cytoplasm. Pairwise simple linear regression of protein expression in nucleus and cytoplasm was analyzed by PerformanceAnalytics package (v1.5.3). A correlation coefficient matrix was generated from the previous step, and plotted by pheatmap package (v1.0.12).

Relevant data from The Cancer Genome Atlas (TCGA) database containing the mRNA expression and clinical information of 479 patients with melanoma were downloaded from cBioPortal (TCGA, Firehose Legacy, accessed on 6 April 2020; refs. 16, 17). The following columns of TCGA melanoma database were extracted and loaded to survival analysis: PLK1 expression, NOTCH1 expression, OS, and disease-free survival (DFS) status, as well as OS and DFS months. Patients with melanoma were divided into two groups according to the expression levels of PLK1 and NOTCH1: (i) high PLK1 + high NOTCH1 (mRNA levels ≥ median) and (ii) others (mRNA levels < median) to study the cumulative effects of these two genes on survival outcome, as described by Amelio and colleagues (18). Using R, the OS and DFS between these two groups were analyzed by survival package (v2.43-3). The Kaplan–Meier curves between these two groups were plotted by survminer (v0.4.3) within R.

Human melanoma cell lines, A375, SK-MEL-2, and SK-MEL-28, were purchased from the ATCC. A375 cells were cultured in DMEM (Corning) with 10% FBS (Sigma-Aldrich). SK-MEL-2 and SK-MEL-28 cells were cultured in Eagle Minimum Essential Medium (Corning) supplemented with 10% FBS at standard cell culture conditions (37°C and 5% CO₂ in a humidified incubator). All cell lines were routinely authenticated (upon thawing or cultured for >3 months) using short tandem repeat testing. BI 6727, the small-molecule inhibitor used against PLK1, and MK-0572, the γ-secretase inhibitor used to inhibit NOTCH (19), were purchased from Selleck Chemicals and dissolved in DMSO (Thermo Fisher Scientific) before use.

For trypan blue exclusion assay, 48 hours following drug treatments, melanoma cells were trypsinized and an aliquot was stained with trypan blue dye (Bio-Rad). Viable, dead, and total cells were determined using the Bio-Rad TC10 Automated Cell Counter. Significance was analyzed by one-way ANOVA with Tukey multiple comparison test using the Prism Software package (GraphPad). The data are expressed as the mean ± SEM with statistical significance (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001). The combination index (CI) assessing drug synergism was calculated via CalcuSyn Software (Biosoft) employing the Chou–Talalay method (20).

Forty-eight hours following treatments, 1,000 viable cells from each treatment group were plated in separate wells of a 6-well plate in duplicates. Cells were maintained under standard tissue culture conditions for approximately 10 days. Cells were then stained at room temperature for 30 minutes with 0.5% crystal violet solution in 1:1 methanol:water, rinsed two times with PBS (pH 7.4), and air-dried followed by digital photography.

For immunoblot analyses, the protein lysates were prepared following treatment and harvesting of cells using standard protocol (9). The protein was quantified, and equal protein was subjected to SDS-PAGE followed by transfer to nitrocellulose membranes. The proteins were probed using primary antibodies (details in Supplementary Table S1) and appropriate secondary antibodies followed by chemiluminescence detection with Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific). Blots were imaged using a Li-COR Odyssey Fc.

For RNA sequencing (RNA-seq) analysis, total RNA was isolated from cells in four replicates in each treatment group. A total of 16 samples were submitted to the University of Wisconsin Biotechnology Center Gene Expression Center (Madison, WI) for quality control, library construction, and further analysis. The purity and integrity of total RNA were assessed on the NanoDrop One Spectrophotometer and Agilent 2100 BioAnalyzer, respectively. Samples that met the TruSeq Stranded mRNA Sample Preparation Guide (Rev. E) input guidelines were prepared using the Illumina TruSeq Stranded mRNA Sample Preparation Kit. The quality and quantity of prepared libraries were assessed by analysis on the Agilent 2100 Bioanalyzer and Qubit fluorimeter. An equimolar pool of all libraries with 30 × 10⁶ targeted reads per sample was sequenced on the Illumina NovaSeq S4 flow cell with 2 × 150 bp sequencing. Images were analyzed using the standard Illumina Pipeline (v1.8.2).

From the RNA-seq data, gene abundances were estimated using RSEM (v1.3.0; ref. 21), with read mapping performed using the STAR aligner (v2.5.3a; ref. 22). Reads were mapped to the GRCh38 assembly of the human genome with the GENCODE v31 comprehensive transcript annotation. The “–forward-prob 0” option was used for RSEM to reflect the fact that the RNA-seq libraries were strand-specific. RSEM-estimated gene counts and effective lengths were compiled via tximport (v1.10.1; ref. 23) and used for differential expression analyses with DESeq2 (v1.22.2; ref. 24) within R. Read counts were modeled using a negative binomial generalized linear model in DESeq2 with terms for each drug and an interaction term for effects resulting from the combination of the two drugs (Fig. 1B). Differentially expressed genes (DEG) for each contrast of interest were identified as those with an P_adj < 0.05 after applying a multiple test correction using an FDR of 0.05. RNA-Seq data were deposited into Gene Expression Omnibus data repository with accession no., GSE159095.

To analyze biological functions and pathways affected by drug treatments, DEGs with log₂ fold change ≥ 1 and log₂ fold change ≤ −1 were separately loaded to GOstats (v2.48; ref. 25) and KEGGprofile (v1.24; ref. 26) packages within R, respectively. The significantly up- and downregulated gene ontology (GO) terms and pathways with their P values were separately uploaded to the ClueGO application (v2.5.6) in Cytoscape (27). To reduce redundancy, the GO Term Fusion box was checked to fuse GO terms with shared genes.

Employing an in-house human melanoma TMA coupled with high-throughput, multispectral Vectra scanning and InForm analysis, we analyzed 110 melanoma tissues and 16 benign nevi (Supplementary Fig. S1A). The TMA was immunostained for PLK1, NOTCH1 (antibody used had a high affinity for the “activated” intracellular form of NOTCH1 over the full-length form), the melanoma biomarker, S100, and DAPI, followed by Vectra scanning and analysis with InForm software. A representative image with merged staining is shown in Fig. 2A. Using this system, we determined the levels of PLK1 and NOTCH1 in both the nucleus and cytoplasm of melanoma cells (as determined by positive S100 staining). We found that while PLK1 and NOTCH1 were detected in both the nucleus and cytoplasm, NOTCH1 levels were higher in the nucleus, and PLK1 levels were higher in the cytoplasm (Fig. 2B). To compare the correlative expression between PLK1 and NOTCH1, we generated a correlation matrix that demonstrated positive correlations between PLK1 and NOTCH1 in both the nucleus and cytoplasm (Fig. 2C and D).

This association between PLK1 and NOTCH1 in our TMA analysis led us to further explore the correlation of these proteins with melanoma clinical parameters. Therefore, we determined whether Breslow thickness, a measure of melanoma aggressiveness (28), is correlated with PLK1 and NOTCH1. Interestingly, we found a positive correlation between Breslow thickness and expression of both PLK1 and NOTCH1 (Supplementary Fig. S1B).

Furthermore, we determined the relationship between the mRNA expression of PLK1 and NOTCH1 and melanoma patient survival using TCGA database, where a total of 459 and 402 patients with available clinical information were selected for OS and DFS analysis, respectively. By comparing the survival probability between patients with highly expressed (≥median) PLK1 and NOTCH1 versus the rest of patients, we found that higher levels of PLK1 and NOTCH1 were associated with poor OS and DFS (Fig. 2E). In addition, we performed survival analysis to correlate with PLK1 and NOTCH1 expression individually using TCGA data available from cBioPortal and Protein Atlas platforms. Our results using Protein Atlas showed that high expression of both PLK1 and NOTCH1 correlate significantly with poor OS, however, with cBioPortal, only PLK1 expression was found to be significantly associated with poor OS, but not NOTCH1 expression, which may be due to the difference in sample size and populations used between the platforms (Supplementary Fig. S2). Taken together, these findings, combined with previously published studies, suggest that targeting PLK1 and NOTCH1 together may provide a novel means for melanoma management.

To determine the effects of combined inhibition of PLK1 and NOTCH, we used PLK1 inhibitor, BI 6727, and NOTCH pathway inhibitor, MK-0752 (a potent γ-secretase inhibitor), in A375 [BRAF mutant (V600E) with wild-type TP53], SK-MEL-2 (wild-type BRAF and mutant TP53), and SK-MEL-28 (both BRAF and TP53 mutant) human melanoma cell lines. These lines were used as they represent the genetic backgrounds and mutations in the most common genetically altered genes in melanoma (Supplementary Table S2). Employing trypan blue exclusion assays, we found that BI 6727 (10 and 20 nmol/L) and MK-0752 (50 and 100 μmol/L) treatment resulted in significant decreases (60%–80%) in viability and growth of human melanoma cells (Fig. 3A). The CI, a metric to determine whether drugs behave synergistically, was calculated using the Chou–Talalay theorem for the combination treatments. Our data demonstrated that the CI was less than 1 in all the tested combination doses of BI 6727 and MK-0752 in A375 and SK-MEL-2 cells, and at 10 nmol/L BI 6727 and 50 μmol/L MK-0752 in the SK-MEL-28 cells (Fig. 3A), indicating potential synergistic responses between these two drugs in melanoma. This was further confirmed using RealTime-Glo and MTT assays (Supplementary Fig. S3). Immunoblotting of protein lysates from the treated cells for HES1, a key downstream NOTCH pathway protein (29), showed that the MK-0752 treatment alone, as well as in combination with BI 6727, was associated with reduced HES1, confirming the efficient targeting of NOTCH pathway (Supplementary Fig. S4). To determine the effects of PLK1 and/or NOTCH inhibition on clonogenic survival of melanoma cells, we used a colony formation assay. Our data demonstrated that compared with single-drug treatments, the combination treatment resulted in a marked decrease in colony formation ability of the melanoma cells (Fig. 3B). Taken together, our data demonstrated that combined targeting of PLK1 and NOTCH signaling pathways imparts synergistic antiproliferative responses against melanoma.

To determine the possible mechanisms behind the observed synergistic antiproliferative response of combined PLK1-NOTCH inhibition, we used RNA-seq to measure the differential gene expression following treatment of SK-MEL-2 human metastatic melanoma cells with PLK1 inhibitor, BI 6727 (20 nmol/L), and/or NOTCH inhibitor, MK-0752 (100 μmol/L), with four replicates in each treatment (Supplementary Fig. S5A). While most samples generated around 30 × 10⁶ sequencing reads and approximately 79% of them were aligned to human genome, it appears that the MK1 sample was an outlier (Supplementary Fig. S5B), which was further confirmed by unsupervised clustering, and the sample was removed from further analysis. Correlation analysis and principal component analysis on the rest of the samples showed a favorable consistency between replicates within each treatment (Supplementary Fig. S5C and S5D).

After evaluating the consistency of the samples, we identified DEGs (|log₂ fold change| ≥ 1 and false-positive rate ≤ 0.05) when comparing the individual and combined treatments with vehicle (DMSO), as well as the interaction between BI 6727 and MK-0752. Specifically, we implemented a negative binomial regression model where the mean for each gene, after scaled by a sample-specific normalization factor, was modeled as:

where X_BI and X_MK are indicator variables that are 1 if the samples were treated with BI 6727 and MK-0752 and 0 otherwise; and X_BI:MK is the interaction effect for BI 6727 and MK-0752. Herein, we define that the term BI:MK represents the interaction effect between BI 6727 and MK-0752 (i.e., transcriptomic responses different than the sum of the responses to the monotherapies), while BI+MK represents the overall effects resulting from the comparison between the combined treatment and vehicle control (i.e., overall effects of the dual treatments; Fig. 1B).

Our data identified 909 DEGs from BI 6727 treatment, 675 DEGs from MK-0752 treatment, 2,142 DEGs from the combined treatment of BI 6727 and MK-0752, as well as 304 DEGs from the synergistic interaction of BI 6727 and MK-0752 (Fig. 4A). Interestingly, as seen in the Venn diagram (Fig. 4B), 1,245 unique genes were found solely in the combined treatment, suggesting that combined PLK1 and NOTCH inhibition may have unique and broad treatment response mechanisms compared with the individual agents. Figure 4C and Supplementary Fig. S6 illustrate the top 100 most significant DEGs with |log₂ fold change| ≥ 1 in the BI+MK combination and BI:MK interaction effects, along with their log₂ fold change in the individual treatments. The list also includes significant decrease in NOTCH target gene, HES1, further confirming the immunoblot analysis for HES1 protein presented in Supplementary Fig. S4. These exciting results led us to further explore the functionality of these DEGs in melanoma.

We conducted GO and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of the various DEGs. Specifically, we were interested in the overall transcriptomic responses (BI+MK), as well as the synergistic responses (BI:MK) of combined drug treatments. In GO analysis (counts ≥ 2; P ≤ 10⁻⁵), we identified 38 downregulated and four upregulated GO terms affected by BI+MK, including metabolism, cell proliferation, and migration (Fig. 5A). Furthermore, we identified 10 downregulated and six upregulated GO terms affected by the interaction effects (BI:MK) of combined drug treatment, including increased hypersensitivity and apoptosis, as well as decreased cell proliferation (Fig. 5B). In KEGG pathway analysis, the response of BI+MK was found to be associated with modulation of multiple pathways shared with single-drug treatments, such as PI3K–AKT, extracellular matrix–receptor interaction, and protein digestion and absorption, as well as pathways uniquely affected by combined treatment, such as MAPK, RAS, and RAP1 pathways (Fig. 5C). Furthermore, we found that BI:MK was associated with multiple affected pathways, including MAPK, RAS, proteoglycans, and phospholipase D (Fig. 5D).

To get a broader look at how the networks integrate with each other, we utilized the ClueGO application in Cytoscape to plot the dysregulated biological functions and pathways due to the combined inhibition of PLK1 and NOTCH. The hierarchical structures and the connections between nodes reflect the interplay among the altered biological function and pathways. The combined inhibition of PLK1 and NOTCH (BI+MK) resulted in a comprehensive modulation of melanoma-related networks, including PI3K-AKT, NIK/NF-κB, and MAPK signaling (Fig. 6A and C). These networks are related to cell proliferation and development, which were found to be synergistically modulated as well (Fig. 6B and D). Interestingly, our data demonstrated that the combined inhibition of PLK1 and NOTCH upregulated pathways associated with the hypersensitivity of melanoma cells (Fig. 6B).

Our cell culture data and RNA-seq results showed that the combined targeting of PLK1 and NOTCH led to enhanced apoptosis and decreased cell proliferation. To verify our findings, we analyzed several apoptosis-related proteins in treated A375 and SK-MEL-2 melanoma cells using immunoblotting. Our data demonstrated that BI 6727 and MK-0752 combination caused a marked increase in cleaved PARP and cleaved caspase-3 in both melanoma cell lines, as well as increased expression of tumor suppressor protein, TP53, in A375 melanoma cells, which contains wild-type TP53 (Fig. 7A). Interestingly, we also observed increased apoptosis in SK-MEL-2 melanoma cells, which carry mutant TP53 (Fig. 7B).

Because metastatic melanoma is notoriously resistant to existing therapeutic approaches, combinatorial therapeutic regimens may be a better strategy toward effective relapse-free melanoma management. Our study focuses on targeting two complementary and connected pathways (PLK1 and NOTCH pathways) in melanoma cells to obtain a superior antiproliferative response. These pathways have been shown either by our laboratory or other groups to be important in melanoma development and progression (6, 7, 13, 30). However, to our knowledge, combined inhibition of these key pathways has not been investigated in melanoma before. In this study, we used two small-molecule inhibitors that have been individually exploited for their anticancer effects in clinical trials, BI 6727 (11) and MK-0752 (31), and combined them and found a synergistic antiproliferative response in human melanoma cells.

The rationale for dual targeting was based on previous research discussed above and was further strengthened by our TMA and TCGA analysis (Fig. 2). In our TMA data, we observed a significant positive correlation between the expression of PLK1 and NOTCH1 in melanoma tissues. Our TMA results were supported by TCGA analysis of large melanoma cohorts, where we found that high mRNA expression of both PLK1 and NOTCH1 is associated with poorer OS and DFS. In our cell culture experiments with multiple melanoma cell lines with varying genetic makeup, we observed synergistic antiproliferative effects of small-molecule inhibition of PLK1 (BI 6727) in combination with NOTCH inhibition (MK-0752) on cell viability and clonogenic survival analysis (Fig. 3).

We also studied the potential mechanisms of combined inhibition utilizing RNA-seq. To fully address the mechanisms of additive and/or synergistic effects between two inhibitors, we implemented a negative binomial regression with interaction terms for differential gene expression analysis. In terms of comparing differential gene expression in combination treatments, many existing analytic approaches only focus on comparing the individual or combined drugs versus vehicle control, but have ignored the potential synergistic interactions between the drugs (32, 33). Thus, our statistical model provides us with two advantages: (i) it enables the evaluation of genes affected by the interaction responses of the two drugs (the BI:MK term) and (ii) it allows us to study the overall landscape of DEGs due to the combination of the two drugs (the BI+MK term). Although the interaction term has been used to evaluate drug combinations at the cell level (34), to our knowledge, thus far it has not been applied to differential gene expression analysis in RNA-seq.

Using this RNA-seq analysis, we have identified the top 100 most significant DEGs due to the combined treatment with BI 6727 and MK-0752 (Fig. 4C). The top five genes in this list are associated with cancer metastasis, including Uca1 (urothelial carcinoma-associated 1; ref. 35), Apobec3G (apolipoprotein B mRNA editing enzyme catalytic subunit 3G; ref. 36), Plau (plasminogen activator urokinase; ref. 37), Axl (AXL receptor tyrosine kinase; ref. 38), and Podxl (podocalyxin like; ref. 39). In addition, Axl has been implicated in the regulation of acquired resistance against BRAF inhibitors by activating the PI3K–AKT pathway (40). Together, the fact that the top five most-downregulated genes indicate reduced metastasis supports the findings of our study suggesting a superior response of combined inhibition of PLK1 and NOTCH against melanoma metastasis. Moreover, these data also open new avenues for investigations on Apobec3G, which has not been associated with metastatic melanoma previously.

Our RNA-seq analysis has also shown that the inhibition of PLK1 and NOTCH imparted significant downregulation of multiple melanoma-related pathways, especially the MAPK pathway (Figs. 5 and 6). Abnormities of MAPK pathway, such as BRAF and RAS mutations, account for more than 80% of melanoma cases (41), making it an important target for melanoma chemotherapy and immunotherapy (42). Our findings were supported by previous studies regarding the relationship between PLK1/NOTCH1 and MAPK. PLK1 has recently been shown to activate ERK1/2 signaling through phosphorylation of CRAF in prostate cancer, indicating a noncanonical function of PLK1 as a MAPK pathway stimulator (43). In addition, it has been shown that the activation of MAPK/PI3K–AKT pathways by NOTCH1 results in the progression of primary melanoma (44).

To understand the mechanism(s) of action of drug combination, we also determined top 100 most significant DEGs due to the BI:MK interaction effects (Supplementary Fig. S5). The result suggests that the interaction of these two drugs produces significant additional gene modulations in melanoma cells when compared with individual drugs (Figs. 5 and 6). Interestingly, the inhibition of PLK1 and NOTCH increased type I and type III hypersensitivity GO terms, defined as inflammatory responses by GO (45). The major contributors to these GO terms in our data were BTK (Bruton tyrosine kinase) and FCER1G (Fc fragment of IgE receptor Ig). Although these two genes have not been studied in melanoma, they have been shown to follow a context-dependent pattern in cancer progression. BTK, a Tec family kinase, has been shown to increase cell proliferation and survival signals or induce senescence and apoptosis depending on the tissue type (reviewed in ref. 46). The two available reports on FCER1G in cancer suggest its tumor promotor role in renal cell carcinoma (47) and tumor suppressor role in multiple myeloma (48). Despite the lack of research on type I and type III hypersensitivity in melanoma, there are a few studies regarding PLK1 or NOTCH1 and melanoma immune activity. For example, Li and colleagues have found a negative correlation between PLK1 and immune activities in many cancer types, including melanoma, and the inhibition of PLK1 led to the upregulation of MHC class I in melanoma cell lines (49). Moreover, NOTCH1 in melanoma was shown to create an immunosuppressive tumor microenvironment (50), while NOTCH1 inhibition improved the efficacy of immune checkpoint inhibitors against melanoma. This may provide us with a novel strategy of coupling this combined therapy with immune therapy.

Furthermore, we validated one of the antiproliferative pathways obtained from RNA-seq analysis in two melanoma cell lines with different mutation landscapes. We observed modulation of several apoptosis-related proteins after combined drug treatment (Fig. 7). In melanoma cells with mutant TP53, it is likely that the observed apoptosis following drug treatments is due to a TP53-independent process. However, in melanoma cells with wild-type TP53, a TP53-dependent phenomenon cannot be ruled out. Further experiments are needed to explore the role of TP53 in this process in greater detail. Overall, our molecular assays and RNA-seq data indicate that combined targeting of PLK1 and NOTCH pathways imparts synergistic antiproliferative responses against melanoma cells and should be explored further for melanoma treatment.

Although our in vitro study utilized multiple melanoma cell lines and RNA-seq after single or combined treatments to study the inhibition of PLK1 and/or NOTCH, the results need to be further validated in in vivo models of melanoma relevant to human disease. In addition, as the RNA-seq data were obtained in one mutant TP53 melanoma cell line, the results should be further verified in cell and mouse lines of varying genetic backgrounds. We have uncovered several genes and pathways affected by combined inhibition of PLK1 and NOTCH in TP53-mutated metastatic melanoma cells, which may or may not be linked to BRAF- or NRAS-mutated melanoma. Even with more work to be done, combined inhibition of the PLK1 and NOTCH pathways provides a potential novel therapeutic regimen to help in the fight against melanoma.

Taken together, our data suggest synergistic antiproliferative responses of combined inhibition of PLK1 and NOTCH in melanoma cells. We have found that combined targeting of PLK1 and NOTCH signaling imparts downregulation of multiple pathways related to melanoma progression and metastasis, as well as potentially increases sensitivity to other therapeutic targets. Our study also identified potential new targets, which have not been explored previously in detail for melanoma management. As the tested inhibitors, BI 6727 and MK-0752, have been used in clinical trials against certain cancer types, the combination studies with these two inhibitors have a high translational value that could be quickly developed into a successful strategy for melanoma management. Overall, our results suggest that combinatorial targeting of PLK1 and NOTCH shows promise and warrants additional studies in in vivo melanoma models, as well as clinical investigations.

N. Ahmad reports grants from the NIH and VA during the conduct of the study. No disclosures were reported by the other authors.

S. Su: Data curation, software, formal analysis, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. G. Chhabra: Conceptualization, supervision, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. M.A. Ndiaye: Project administration, writing-review and editing. C.K. Singh: Supervision, writing-review and editing. T. Ye: Formal analysis, writing-review and editing. W. Huang: Resources. C.N. Dewey: Data curation, software, formal analysis, supervision, writing-review and editing. V. Setaluri: Resources. N. Ahmad: Conceptualization, resources, supervision, funding acquisition, project administration, writing-review and editing.

The authors would like to thank the UW Department of Pathology & Laboratory Medicine's Translational Research Initiatives in Pathology (TRIP) laboratory for their TMA staining services, as well as the UW Biotechnology Center's Gene Expression Center for their RNA-sequencing services. We would also like to thank Dr. Cecile Ane for advice on data analysis and Lauren Benedict for experimental assistance. The results shown here are, in part, based upon data generated by the TCGA research network: https://www.cancer.gov/tcga. This work was partially supported by funding from the NIH (R01AR059130 and R01CA176748 to N. Ahmad and R21CA125091 to V. Setaluri) and the Department of Veterans Affairs (VA Merit Review awards I01BX001008 and I01CX001441 to N. Ahmad and 1I01BX004921 to V. Setaluri, as well as a Research Career Scientist Award IK6BX003780 to N. Ahmad). This work was also supported by the core facilities in the Skin Diseases Research Center Core grant P30AR066524 from NIH/NIAMS, the University of Wisconsin Translational Research Initiatives in Pathology laboratory (TRIP), supported by the UW Department of Pathology and Laboratory Medicine, the Office of The Director- NIH (S10OD023526), as well as the University of Wisconsin Carbone Cancer Center Support Grant P30CA014520.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.