Antitumor effect of Melaleuca alternifolia essential oil and its main component terpinen-4-ol in combination with target therapy in melanoma models

Essential oils (EOs) have been recently emerging for their promising biological activities in preventing tumorigenesis or progression of different tumor histotypes, including melanoma. In this study, we investigated the antitumor activity of a panel of EOs in different tumor models. The ability of Melaleuca alternifolia (tea tree oil) and its main component, terpinen-4-ol, to sensitize the target therapy currently used for melanoma treatment was also assessed. Our results demonstrated that EOs differently affect the viability of human cancer cells and led us to select six EOs effective in melanoma and lung cancer cells, without toxic effects in human fibroblasts. When combined with dabrafenib and/or trametinib, Melaleuca alternifolia synergistically reduced the viability of melanoma cells by activating apoptosis. Through machine learning classification modeling, α-terpineol, tepinolene, and terpinen-4-ol, three components of Melaleuca alternifolia, were identified as the most likely relevant components responsible for the EO’s antitumor effect. Among them, terpinen-4-ol was recognized as the Melaleuca alternifolia component responsible for its antitumor and proapoptotic activity. Overall, our study holds promise for further analysis of EOs as new anticancer agents and supports the rationale for their use to improve target therapy response in melanoma.


Introduction
Cutaneous melanoma is the most aggressive type of skin cancer. BRAF represents the most common driver mutation present in~50% of patients and predicting a more aggressive behavior 1 . Although target therapy and immunotherapy represent a great opportunity for melanoma treatment, patients often face lack of clinical response, the emergence of resistance to treatment, and invalidating side effects 2 . Consequently, innovative and combined therapies are still urgent to treat and eventually eradicate advanced melanoma. In light of this consideration, a large number of preclinical and clinical trials are ongoing to identify new therapeutic approaches.
Over the past decades, compounds extracted from plants have demonstrated their effectiveness in different diseases, including melanoma 3 . Examples include vinblastine 4 , vincristine 5 , paclitaxel 6 , and camptothecin 7 . Scientific evidences have demonstrated that, among natural compounds, essential oils (EOs) showed great potential for the management of a number of diseases including cardiovascular 8 , diabetes 9 , and Alzheimer 10 . EOs also represent a valid source to prevent the invasion of SARS-CoV-2 into the human body 11 , or to downregulate angiotensinconverting enzyme 2 expression in epithelial cells 12 .
Due to their minimal cytotoxicity 13,14 , EOs are considered pharmaceutically safe and could represent a good alternative natural source of anticancer agents, thus deserving further investigations to ascertain their mechanism of action and to validate their possible clinical uses as alternative/complementary antitumor agents. In the last 20 years, preclinical studies demonstrated anticancer activity of either some EOs or their main components 15,16 and led to case-control studies 17 and clinical trials [18][19][20] . At present, EOs are used to ameliorate cancer patients' quality of life and clinical trials are ongoing to evaluate their efficacy or the efficacy of their components in cancer patients (NCT02336087, NCT03449 303, NCT04560114, NCT04449315, NCT00003219, NCT00 003238, NCT01459172, NCT01046929, NCT04296266). From the hundreds of studies published in the last years, it is evident that, in addition to their chemopreventive effects, several EOs and their constituents show antioxidant, antiproliferative, proapoptotic, antiangiogenic, and antimetastatic activity in melanoma models [21][22][23] . Synergistic effect of EO components such as geraniol 24,25 , β-elemene 26,27 , β-caryophyllene 28 , limonene 29 , eugenol 30 , and thymoquinone 31,32 with cancer therapy has been also reported.
To shed light on the use of EOs as possible anticancer agents, in this investigation we reported the in vitro anticancer effect of a panel of EOs and investigated the possible use of Melaleuca alternifolia (TTO, EO05 in this investigation) as a sensitizer of targeted therapy in melanoma models. Furthermore, machine learning (ML) classification models were developed and used to investigate the possible efficacy of the more important EOs' single components.

A panel of EOs differently affects the viability of melanoma cells
The antitumor activity of 61 EOs (Table S1) was firstly assessed for their ability to affect the proliferation/viability of M14 melanoma cell line (50 μg/ml, 72 h). As reported in Fig. 1a, 18 EOs significantly reduced the proliferation/ viability of M14 cells, and 12 of them inhibited at least 50% of cell proliferation. Among the 12 EOs, EO14 and EO40 were excluded from further investigations owing to their low solubility. M14 cells were treated with the remaining 10 EOs (10-50 μg/ml, 24-72 h). After 24 h treatment, a dose-dependent reduction of cell proliferation/viability was observed for seven EOs, whereas between 48 h and 72 h no significant differences in terms of IC 50 were observed (Fig. 1b, c, Table S2). EO22, EO32, and EO52 were the less effective in reducing the M14 proliferation/viability and showing the highest deviation from the median IC 50 for each time point (Fig. 1b, c).
As reported in Fig. S1a, the six most effective EOs (EO05, EO12, EO18, EO20, EO29, EO49), but not EO39, showed no significant effect on the proliferation/viability of normal human fibroblasts (50 μg/ml, 72 h), therefore EO39 was not further investigated. The antitumor activity of the final selected EOs was then explored on cell lines with three different tumor histotypes: lung (H1299, A549), colon (HCT116), and breast (MDA-MB-231) carcinoma. As shown in Fig. S1b, lung cancer cells treated with each EO (50 μg/ml, 48 h) were as sensitive as M14 cells, with cell proliferation/viability inhibition ranging from 67% to 82% for both cell lines used. On the contrary, the proliferation/viability of MDA-MB-231 cells was significantly reduced only by EO12, whereas HCT116 cells were resistant to the six EOs.
Even though at different extend, increasing concentrations of each EO displayed a similar ability in significantly reducing the viability of both BRAF wild type/NRAS mutant (Sbcl1, ME4405), BRAF wild type/NRAS wild type (ME1007), and BRAF mutant/NRAS wild type (M14, A375, LOX IMVI) melanoma cells ( Fig. 2a-f, Fig. S1c), thus indicating the absence of relevance of BRAF or NRAS status in the sensitivity to EOs.

ML binary classification algorithms identify the most likely relevant components of EOs
To identify the most important chemical components likely responsible for viability inhibition of M14 cells, ML models were developed as reported in supplementary methods. At 50% proliferation/viability inhibition threshold, Matthews correlation coefficient and area under the curve value were 0.604 and 0.537, respectively (Fig. S2a). Inspection of the weighted feature importance values revealed α-terpineol, terpinolene, and terpinen-4-ol as those components mainly responsible for proliferation/ viability inhibition of M14 cell line (Fig. S2a). The chemical composition of the EOs with the higher efficacy is reported in Table 1 and Tables S4-S8. All the three components identified through ML analysis were evidenced only in EO05 and EO49, even if at different concentrations (Table S3).

EO05 sensitizes melanoma cells to target therapy
We next combined EO05, a very well characterized EO from Melaleuca alternifolia 33 containing all the three components identified through the ML approach, with the targeted therapy currently used for the treatment of advanced melanoma patients harboring BRAF mutations 34 . Growth inhibitory curve and relative analysis of drug interaction demonstrated that 24 h EO05 followed by 48 h dabrafenib (BRAF inhibitor) resulted in a synergistic effect on M14 proliferation/viability reduction with combination index (CI) = 0.6 ( Fig. 3a). Accordingly, this combination produced a synergistic effect also in A375 cells (Fig. S3a).
A mean of 18.5% and 16.8% of subG1 peak, indicative of dead cells, was detected after treatment with dabrafenib or p-values were calculated between control (Ctrl) and EO-treated cells. *p < 0.05; **p < 0.01; # p < 0.001 after applying Student's t test. Dotted columns represent the six EOs further investigated in this study. b MTT assay of M14 cells treated with the indicated EOs (10-50 μg/ml, 24-72 h). a, b Results are reported as "cell proliferation-viability of treated cells/cell proliferation-viability of control cells × 100" and represent the average±standard deviation of at least three independent experiments. c Quantification of 50% inhibition of cell proliferation/ viability (IC 50 ) of the indicated EOs calculated for M14 cells treated as reported in b. The median of IC 50 is shown. ***p < 0.001; ****p < 0.0001 after applying one-way ANOVA test. EO05, respectively. Interestingly, in cells treated with EO05 followed by dabrafenib, the subG1 population significantly increased up to 40.2% (Fig. 3b, c). In addition, treatment with the caspase inhibitor zVAD-FMK (zVAD) significantly reduced the subG1 peak in cells treated with EO05 alone (4.9%) or in combination with dabrafenib (8%), thus demonstrating apoptotic cell death. Apoptosis induction was also confirmed by the increase of active caspase 3 and cleaved PARP in cells treated with the combination when compared to single treatments (Fig. 3d, e). Similar to what observed for dabrafenib, administration of 24 h EO05 followed by 48 h trametinib (MEK inhibitor) showed a synergistic effect strongly reducing M14 cell proliferation/viability (CI = 0.5) (Fig. 4a). Accordingly, treatment of EO05 followed by trametinib increased the percentage of subG1 peak, caspase 3, and PARP cleavage ( Fig. 4b-e) when compared with trametinib or EO05 alone. Moreover, the addition of zVAD significantly decreased the subG1 peak in cells treated with EO05 alone or in combination (Fig. 4b, c). A synergistic effect of EO05 followed by trametinib was also obtained in the BRAF wild type melanoma cells, ME4405 (CI = 0.6) (Fig. S3b). Next, the effect of EO05 in combination with dabrafenib and trametinib, the current standard treatment for BRAF mutant melanoma patients, was also assessed. Interestingly, 24 h EO05 followed by 48 h of dabrafenib/ trametinib treatment strongly reduced the proliferation/ viability of M14 cells compared with exposure to EO05 alone or to dabrafenib/trametinib (Fig. 4f).

Discussion
In this study, we provided evidence about the antiproliferative effect of a panel of EOs in melanoma and lung carcinoma cells. More importantly, we determined the ability of TTO to synergize with target therapy in melanoma models. In particular, an initial screening of 61 EOs led to select six of them (TTO, Pinus Sylvestris, Lavandula Angustifolia, Citrus Paradisi, Pinus Sibirica, Cupressus Sempervirens) as the most efficacious in terms of reduction of tumor cell proliferation/viability, without affecting normal fibroblasts viability. We also found that the efficacy of EOs depends on the tumor histotype examined. In fact, the treatment with the six EOs reduced cell proliferation of melanoma and lung carcinoma cells in a dose-dependent manner, whereas they were ineffective in breast and colon carcinoma cells. The mechanism that renders the different histotypes differently sensitive to the six EOs is not yet clear. No reports have been yet published about the six EOs used in colon cancer models. Nevertheless, TTO has been reported to induce apoptosis in breast cancer cells at concentration six times higher than those we used in our study 35 , whereas Pinus Sylvestris EO (EO29) exhibited some potential as an antiproliferative agent in the same cellular model (i.e., MDA-MB-231) 36 , thus suggesting a different composition of EO used. In fact, we and other authors previously reported that multiple factors affect EO composition [37][38][39][40] .
A panel of melanoma cell lines, harboring wild type or mutant BRAF and NRAS, showed sensitivity to the six EOs, even if at a different extend, thus indicating that the effect of EOs was not related to BRAF or NRAS status.
All the six selected EOs, except for Pinus Sibirica (EO20), were investigated for their effect on cancer [41][42][43][44][45] but only TTO (EO05) showed antitumor efficacy in preclinical  46,48 , and interfere with in vitro invasive/migratory capability 49 of melanoma models. Moreover, a topical formulation of TTO retarded the in vivo growth of subcutaneous melanoma and evidenced immune effector cell recruitment on the treated region 50 . Considering all these effects, the EOs lipophilicity, the fact that EOs are well absorbed through the skin 51 , as well as the fact that chemoprevention is an essential approach for cancer control 52 , TTO has been suggested as a possible chemopreventive candidate to be used in topical formulations against melanoma and other types of skin cancer 48,53 .
Despite the great interest in TTO reported in the last years 54,55 , the contribution of TTO as a sensitizer of cancer, and in particular, of melanoma therapy 56,57 , is unknown. We demonstrated that TTO synergized with dabrafenib and trametinib, when administered either as single agents or in combination, in terms of apoptosis induction, when TTO treatment was followed by exposure to one of the two drugs. However, we cannot exclude  that TTO alone and in combination with targeted therapy may activate other forms of cell death.
In agreement with studies demonstrating that among TTO components, terpinen-4-ol is responsible of TTO efficacy 46,47,49 , we demonstrated the relevance of terpinen-4-ol the main component present in TTO (37.5%), in the antiproliferative effect and in the sensitization to target therapy. In mouse or human melanoma cells, TTO and terpinen-4-ol elicited G1 cell cycle arrest, showed an antiproliferative effect, antimigratory/antiinvasive ability against cells resistant to chemotherapy, and induced necrotic and apoptotic cell death 46,47,49 . We and other authors also reported terpinen-4-ol ability to affect in vitro and in vivo growth of tumors with different origin [58][59][60][61][62] , and to enhance the effect of several chemotherapeutic or biological agents in cancers not including melanoma 61 . Results from ML analysis performed on the M14 screening were in good agreement with experimental data effectively indicating terpinen-4-ol as one of the components mainly responsible for viability inhibition of melanoma cells. Indeed, among the final selected six EOs, EO05 did contain terpinen-4-ol at the highest percentage. The antiproliferative effect of EO12, EO18, EO20, EO29, and EO49, showing low or nondetectable levels of terpinen-4-ol could be due to other components present in their composition and reported to affect proliferation of melanoma cells, such as linalool 63 , limonene 64 , camphene 65 , α-, and β-pinene 66 .
In agreement with studies demonstrating (i) the nature of terpenes as lipophilic molecules able to disrupt normal structure and function of cell membranes 46 , and (ii) the ability of TTO and terpinen-4-ol to interact with the lipid bilayer of cellular membranes and to inhibit the intracellular signaling induced by p170 glycoprotein 49,67 , we  The results are reported as "cell proliferation-viability of treated cells/cell proliferation-viability of control cells × 100". a, e The results represent the average±standard deviation of three independent experiments. Experiments with zVAD were repeated twice. p-values were calculated between control (Ctrl) and treated cells, cells treated in combination and cells treated with single drugs, or between cells treated or not treated with zVAD. *p < 0.05; **p < 0.01, ***p < 0.001 after applying Student's t test.
suggest that the synergistic effect of TTO or terpinen-4-ol with target therapy could be related to their effect on plasma membrane, i.e., reorganization of lipid architecture, thus favoring the entrance of drug in the cell.
Our data are in agreement with previous studies reporting the ability of EOs such as Cymbopogon citratus, or EO components, such as β-elemene and thymoquinone, to increase the efficacy of radiation in melanoma models 68,69 , or curcumol, β-caryophyllene, citral, or valencene to enhance the sensitivity of tumors from different origin to antineoplastic treatment [70][71][72] .
To the best of our knowledge, this is the first study examining the ability of TTO, and in particular, terpinen-4ol, to potentiate the targeted therapy of melanoma, highlighting the importance of our investigation. The efficacy of the combination TTO/target therapy could be of relevant importance as it can lead to the use of a lower concentration of drugs commonly used for the management of melanoma patients and consequently lower toxic treatments in terms of side-effect and more efficacious. The potential use of TTO is further supported by its non-toxicity in normal cells 35 and by its penetrability in the skin 73 .
Supported by low toxicity and side-effect of EOs, as well as their good tolerance by patients, our study hold promise for further analysis of EOs as new anticancer drugs and/or as a source of potential anticancer supplement against melanoma. The effect of TTO on melanoma cells and the analysis of its main components are worthy of further investigation.

Western blot and flow cytometric analyses
Western blot analyses were performed as previously reported 77 using primary antibodies directed to PARP (cod. 51-6639GR, BD Bioscience, San Jose, CA) or HSP72/73 (cod. D00175805, Calbiochem, Saint Diego, CA, USA,) as control of loading and transfer. Anti-mouse immunoglobulin G-horseradish peroxidase-conjugated antibody (cod. 1858413, Amersham Biosciences, Freiburg, Germany) was used as a secondary antibody.
Cell cycle distribution by propidium iodide staining was performed as previously described 78 . Caspase 3 activation was evaluated using an active caspase 3-PE antibody (cat. 559565, BD Bioscience, San Jose, CA), following the manufacturer's instructions. All the cytofluorimetric analyses were performed using BD Accuri TM C6 flow cytometer.

GC-MS analysis
GC-MS analyses were carried out using a Perkin Elmer Clarus 500 GC equipped with a flame ionization detector and coupled with a Clarus 500 mass spectrometer. A Stabilwax capillary column (Restek, Bellefonte, PA, USA) was used with helium as carrier gas (1.0 mL/min). GC oven temperature was kept at 60°C for 5 min and programmed to 220°C at a rate of 5°C/min, and kept constant at 220°C for 30 min. Mass spectra were acquired over 40-500 amu with ionizing electron energy 70 eV. In all, 1 μL of the EO was diluted in 1 mL of methanol and 1 μL of the solution was injected into the GC injector at 280°C. The identification of compounds of EOs was performed by comparing mass spectra with those reported in Nist and Wiley libraries. Linear retention indices were calculated after injection of C8-C30 aliphatic hydrocarbons mixture under the same conditions described above and compared with available linear retention indices data in the literature.

ML binary classification
All calculations were performed using the Python programming language (version 3.7, https://www.python.org/) by executing in-house code in the Jupyter Notebook platform, as previously reported 79,80 . For details see supplementary material and Table S9, 10.

Statistics
Unless otherwise indicated, at least three independent experiments have been performed. Six technical points for each experimental group were used for MTT assay. The data were expressed as mean ± standard deviation or ± standard error of the mean. For continuous variables, differences between two groups were analyzed with Student's t test (unpaired, two-sided). One-way ANOVA test was used to analyze differences between the three groups. P < 0.05 was considered statistically significant. All statistical tests and the estimation of variation between groups were performed with GraphPad Prism 6 (GraphPad Software, Inc., La Jolla, CA, USA). All data were included in the analyses. Based on the variation shown in our preliminary results, we determined the sample sizes by using power analysis.