Metabolic enzyme ACSL3 is a prognostic biomarker and correlates with anticancer effectiveness of statins in non‐small cell lung cancer

Overexpression of the enzyme ACSL3 is an unfavorable prognostic marker in non‐small cell lung cancer (NSCLC). Moreover, overexpression of ACSL3 increases cell proliferation, migration, and invasion, altering the metabolic properties of cancer cells. Administration of statins causes antiproliferative effects and, importantly, patients with high expression levels of ACSL3 following statin treatment display better prognosis than those not treated with statins.

[Correction added on 28 November 2020, after first online publication: Peer review history is not available for this article, so the peer review history statement has been removed.] Lung cancer is one of the most common cancers, still characterized by high mortality rates. As lipid metabolism contributes to cancer metabolic reprogramming, several lipid metabolism genes are considered prognostic biomarkers of cancer. Statins are a class of lipid-lowering compounds used in treatment of cardiovascular disease that are currently studied for their antitumor effects. However, their exact mechanism of action and specific conditions in which they should be administered remains unclear. Here, we found that simvastatin treatment effectively promoted antiproliferative effects and modulated lipid metabolism-related pathways in non-small cell lung cancer (NSCLC) cells and that the antiproliferative effects of statins were potentiated by overexpression of acyl-CoA synthetase long-chain family member 3 (ACSL3). Moreover, ACSL3 overexpression was associated with worse clinical outcome in patients with high-grade NSCLC. Finally, we found that patients with high expression levels of ACSL3 displayed a clinical benefit of statins treatment. Therefore, our study highlights ACSL3 as a prognostic biomarker for NSCLC, useful to select patients who would obtain a clinical benefit from statin administration.

Introduction
Lung neoplasia is the leading cause of cancer death in both sexes, accounting for almost 20% of the total cancer deceases [1]. Non-small cell lung cancer (NSCLC) is the most frequent subtype and represents about 85% of all lung tumors [2]. Genetic and environmental factors have been associated with the development of NSCLC, with cigarette smoking being the most important modifiable risk factor. Accordingly, it is estimated that 70% of all NSCLC tumors arise in smokers [3].
Unfortunately, NSCLC is frequently diagnosed when advanced-stage disease is present [4,5]. The 5-year overall survival rate for NSCLC remains poor, from 68% in patients with stage IB disease to 0-10% in patients with stage IVA-IVB disease [6]. During the last decade, a significant improvement in the clinical outcome of patients has been reached, essentially due to the integration of targeted therapies and immunotherapy in genetically selected subpopulations of patients [7]. However, only few NSCLC patients will achieve long-term survival, and therefore, a deeper understanding of NSCLC molecular biology coupled with high-quality biomarkers of the disease is urgently needed.
Deregulation of cellular metabolism is a hallmark of cancer [8]. Tumor cells adapt their metabolic status to satisfy their novel requirements of growth, proliferation, and survival. Together with the Warburg effect, lipid metabolism contributes to cancer metabolic reprogramming [9]. Moreover, lipid metabolism genes have a promising role as prognostic biomarkers of cancer. Their deregulated expression has been associated with cancer recurrences and survival [10,11]. Nevertheless, the exact mechanism of the regulation and the biological consequences of lipid metabolism in NSCLC have not been elucidated yet.
Fatty acids (FAs) are indispensable components of cellular membranes, and they are essential for posttranslational protein modifications; moreover, they are energy generators and contribute to maintain redox homeostasis through b-oxidation [12][13][14]. The enzymatic family of long-chain acyl-CoA synthetases (ACSLs) transform free long-chain FAs into fatty acyl-CoA esters, which are used for both lipid synthesis and b-oxidation [15]. ACSL enzymes are ubiquitously expressed, although they differ in tissue expression and in their subcellular localization. For instance, ACSL3 is basically confined to endoplasmic reticulum (ER) and lipid droplets and ACSL4 is localized in peroxisomes and ER, while ACSL1, ACSL5, and ACSL6 are found in mitochondria, plasma membrane, and cytoplasm [16,17]. As a matter of fact, ACSLs enzymes are involved in the metabolic reprogramming of tumor cells. For example, ACSLs stimulate colon cancer progression increasing the metabolic activity as well as the invasive and migratory properties of malignant cells [18]. Besides, the pharmacologic inhibition of ACSLs promotes apoptosis in a TP53-deficient cancer context [19,20].
Other fundamental structural component of lipid membranes is cholesterol, that is also essential for cellular proliferation [21]. Tumor cells show a powerful fatty acid and cholesterol avidity, which can be satisfied by increasing the uptake of exogenous lipids and/ or lipoproteins or by activating de novo lipogenesis and cholesterogenesis [9]. Increased serum cholesterol levels have been associated with higher risk of cancer [22]. Statins are 3-hydroxy-3-methylglutaryl-coenzyme A reductase (3-HMGCR) inhibitors, and they are extensively used in the treatment of hypercholesterolemia [23]. Besides, in vitro and in vivo experimental data suggest that statins may also exert oncoprotective effects [24][25][26]. They have been implicated in the regulation of inflammation, immunomodulation, and angiogenesis [27] too. Particularly, in lung cancer cells, statins induce apoptosis [28], and in preclinical models, statins suppress tumor growth [29]. It has been reported that the long-term use of statins in patients with lung cancer and dyslipidemia decreases mortality [30]. However, not always consistent results are found in the epidemiological and clinical studies that explored the protective role of statins in lung cancer [30][31][32]. Notably, the effect of the use of statins in NSCLC patients, taking into account their lipid metabolic reprogramming, remains unexplored.
The prognostic value of lipid metabolism genes in high-grade NSCLC patients is evaluated for the first time in this work. Our goal is to define putative lipid metabolism-related biomarkers of survival and to analyze their functional consequences and mechanism of action. Moreover, we deepen into the effect of statins in specific metabolic subgroups of patients, providing novel precision medicine strategies for specific subgroups of NSCLC patients. 2015, were included in this analysis. All cases were of high stage (stages III-IV, and they were treated following a standardized protocol. All FFPE samples were studied by an anatomic pathologist confirming > 80% of tumoral cells in each case. Clinical and pathological data were collected from medical reports (see Table 1). This study was approved by the local ethical committee, and all patients signed the consent inform document. The study methodologies conformed to the standards set by the Declaration of Helsinki. Overall survival (OS) was defined from the date of diagnosis to the date of patient death. We also used population's data extracted from 984 patients from The Cancer Genome Atlas (TCGA) public dataset which characteristics were summarized in Table S1.

Gene expression analysis
We employed RNeasy Mini Kit or RNeasy FFPE Kit (Qiagen, Germantown, MD, USA) to get total RNA from NSCLC cancer cells or from FFPE tumor samples previously deparaffinated, respectively. One lg of total RNA was reverse-transcribed by High Capacity cDNA Archive Kit (Applied Biosystems, Waltham, MA, USA) for 2 h at 37°C. VeriQuest SYBR Green qPCR Master Mix (Affymetrix, Santa Clara, CA, USA) was applied for gene expression analysis of NSCLC cells as previously described [33,34] (see oligos used in Table S2). In clinical samples, a Taq-Man Low Density Array (Applied Biosystems) was specially designed for this experiment and was formed of 22 lipid metabolism-related genes ( Table 2). These 22 genes within all genes present in pathways related to lipid metabolism were selected due to their role as regulators of cell metabolism, key steps of interconnection among lipid pathways, or their reported role in biological processes associated with cancer [10]. Gene expression assays were performed in a HT-7900 Fast Real-time PCR. The B2M gene expression was used as endogenous control. RT-STATMINER software (Integromics Ò Inc., Madison, WI, USA) was used to detect and determine the quality control and differential expression analyses of data [10].

Cell culture, treatments, and stable cell lines generation
A panel of NSCLC cells (CRL 5803, CRL 5872, CRL 5875, CRL 5877, and CRL 5908) and HEK-293T cells was obtained from American Type Culture Collection (ATCC, Middlesex, UK). ATCC performs cell line authentication through STR profiling and mycoplasma contamination testing. Frozen aliquots were stored, and cells were passaged in the laboratory for fewer than 6 months after resuscitation [35]. All cell lines were cultured and maintained under manufacture's conditions. Generation of ACSL3 stable overexpression cellular models was performed using lentiviral systems as previously described [18]. HEK-293T cells were transfected using Lipofectamine 2000 (Life Technologies, Thermo Fisher Scientific, Waltham, MA, USA) with lentiviral vectors expressing ACSL3/No ORF empty vector (DNA 2.0, Newark, CA, USA) along with a set of packaging plasmids (Addgene, Cambridge, MA, USA) [18,35].

Proliferation, migration and invasion assays
Cell proliferation was measured in real time using the xCELLigence TM system (ACEA Biosciences, San Diego, CA, USA) according to manufacturer's protocols. Real-time monitoring of proliferation was performed for 5 days in 30-min intervals [35]. A density of 100 000 cells per well was seeded in serum-free medium into inserts of a BD migration and MatrigelTM invasion chamber (BD Biosciences, San Jose, CA, USA). Migrative and invasive cells were measured as previously described [18,35].

Extracellular flux analysis of the acidification rate (ECAR) and the oxygen consumption rate (OCR)
Aerobic glycolysis (Cell GlycoStress Test) and mitochondrial oxidative respiration (Cell MitoStress Test) were examined with the XFe96 Cell Bionalyzer (Seahorse Biosciences, North Billerica, MA, USA, XFe96). Optimal cell density and drugs titration were determined as previously described [36]. The requirements of the cells for aerobic glycolysis and oxidative phosphorylation were screened after the injection of several modulators of both bioenergetic pathways. We monitored mitochondrial respiration and glycolytic function using XF Cell Mito Stress Test kit and XF Glycolysis Stress kit according to manufacturer instructions (Seahorse Biosciences). Briefly, for GlycoStress assay, 20 000 cells were seeded in an XFe96 well plate and maintained in DMEM 10% FBS to allow the cells to attach [36]. Then, the culture medium was changed to 2 mM glutamine without glucose in XFe DMEM (5 mM Hepes) to starve the cells for 1 h. First basal extracellular acidification rate (ECAR) was determined (1-3 measurements). Glucose (10 mM final concentration) was injected (4-6 measurements) to measure glycolysis (i.e., the increased ECAR from basal starved situation after glucose addition). This indicates the competence of the cells to employ glucose. Next, maximal glycolytic capacity was assayed (7-9 measurements) after the injection of oligomycin, which blocks the ATP production from the oxidative mitochondrial respiration. Finally, a third injection with 50 mM DG final concentration was done to shut down aerobic glycolysis (10-12 measurements) [36]. For MitoStress assay, 40 000 cells were seeded in an XFe96 well plate, and cells were maintained in DMEM 10% FBS to allow the cells to attach. Then, the medium was Acyl-CoA synthetase longchain family member 1 ACSL3 Acyl-CoA synthetase longchain family member 3 ACSL4 Acyl-CoA synthetase longchain family member 4 ACSL5 Acyl-CoA synthetase longchain family member 5 ACSL6 Acyl-CoA synthetase longchain family member 6 AMACR Alpha-methylacyl-CoA racemase HSD17B4 Hydroxysteroid (17- changed to 10 mM glucose, 2 mM glutamine, and 1 mM pyruvate in XFe DMEM (5 mM Hepes), and cells were incubated for 1 h at 37°C without CO 2 . Three different modulators of the mitochondrial respiration were sequentially introduced. After basal oxygen consumption rate (OCR) determination (1-3 measurements), oligomycin (1 µM), which inhibits ATPase, was added to determine the amount of oxygen dedicated to ATP production by mitochondria (3-6 measurements) [36].
To determine the maximal respiration rate or spare respiratory capacity, FCCP (carbonyl cyanide-4-(trifluoromethoxy)phenylhydrazone) was injected (0.4 µM) to free the gradient of H + from the mitochondrial intermembrane space (7-9 measurements) and thus to activate maximal respiration. Finally, antimycin A and rotenone (0.5 µM) were added to completely inhibit the mitochondrial respiration (10-12 measurements) [36]. At least 6 replicates per condition were done for each experiment.

Statistical analysis
The primary endpoint of the study was the relationship between gene expression and overall survival (OS). Gene expression (Q) in tumor samples was quantified with the 2-DC t method as previously described [35]. Violin plots were used to display and analyze the distributions of gene expression. The Kaplan-Meier method was used to estimate overall survival curves, and univariate Cox regression analysis was used to test the association between overall survival and clinical variables. Multivariable Cox regression models were derived to estimate the hazard ratio of gene expression and adjusting for potential confounding factors: age, sex, stage, ECOG, and smoking status. In order to minimize the effect of putative outliers and influential points, instead of a continuous gene expression variable a binarized version of it was used by determining a cut-point based on maximally selected rank statistics and using Log-rank scores [37]. The absence of influential observations or outliers was afterward verified through the analysis of the residuals deviance and the changes in the standardized coefficients upon deletion of each observation [37]. In addition, proportional hazards assumption in the Cox models was verified both through tests as described by Grambsch et al. [38] and by inspection of the Schoenfeld scaled residuals against time. Hazard ratios (HR) and their 95% confidence intervals (95% CI) and P-values were calculated from the Cox regression modes. Statistical significance was defined as P-value < 0.05 with bilateral tests. Bonferroni method was applied for multiple test correction of the P-values. The statistical analyses were performed using the R statistical software version 3.6.1 (www.r-project.org). Student's t-test was employed to assay statistically significant differences between sets. All reported P-values were two-sided. Statistical significance was defined as P < 0.05.

Metabolic reprogramming in NSCLC tumors is a reliable prognostic marker of survival
Altered lipid metabolism has been found in many types of cancer, and consequently, lipid metabolism genes could constitute a prognostic and therapeutic tool in these tumors [10,11,39]. In order to explore the putative prognostic value of lipid metabolism genes in NSCLC patients, we performed survival analysis based on gene expression in a series of 90 high-grade NSCLC tumor samples. The final dataset included 25 and 65 samples of stages III and IV, respectively. Median age at diagnosis was 63.5 years (range 40-91), and median overall survival (OS) was 7 months (range 0-62). Only 13.3% of patients had never been smokers. Sample characteristics are described in Table 1 A transcriptomic panel of lipid metabolism genes was assayed in NSCLC samples. Gene selection was made according to our previous analysis in two other types of cancer: colorectal and epithelial ovarian cancer [10,11,35]. Genes were chosen due to their previous described functionality as well as according to their predicted expression in lung tissue (https://www.prote inatlas.org/). Finally, 22 lipid metabolism genes were included in the panel which is shown in Table 2.
To validate these results, we performed a parallel analysis based on ACSL3, NID1, and RETN expression in larger series of NSCLC samples from The Cancer Genome Atlas (TCGA) study, a publicly available gene expression dataset (https://www.cancer.gov/tcga). The analysis included a total number of 984 NSCLC patients. ACSL3 and RETN expression had an overall survival risk effect, and the adjusted HR was significantly higher than 1 (1.38 (95% CI: 1.07-1.79), P = 0.016 and 1.51 (95% CI: 1.18-1.92), P = 0.001, respectively) ( Fig 1B).
Then, we also checked the potential value of ACSL3 and RETN as NSCLC prognostic biomarkers using The Human Protein Atlas website together with The Pathology Atlas online tool (https://www.proteinatlas. org/humanproteome/pathology), that contains mRNA data from TCGA study and protein expression data from different forms of human cancer [40][41][42]. Lung cancer tissues showed moderate-to-strong ACSL3 cytoplasmic immunoreactivity. Moreover, ACSL3 mRNA expression scrutiny displayed analogous results to our previous analysis with a clear association between high ACSL3 expression and poor patient's survival (P score = 0.0008, n = 994) (https://www.pro teinatlas.org/ENSG00000123983-ACSL3/pathology). All these data reliably validate our findings that upregulation of ACSL3 is associated with low survival in NSCLC patients.  Finally, as expected, Kaplan-Meier plots for OS of ACSL3 showed an association between high gene expression and poor clinical outcome, both in the initial group (Fig. 1C) and in TCGA validation dataset (Fig. 1D). Consequently, ACSL3 overexpression could be considered as an unfavorable prognostic factor in NSCLC.

Overexpression of ACSL3 enzyme increases cell proliferation, migration, and invasion altering metabolic properties of NSCLC cells
In order to explore the role of ACSL3 as a putative NSCLC prognostic factor, we generated NSCLC cell lines overexpressing ACSL3. Firstly, we analyzed mRNA and protein expression levels of ACSL3 in a panel of several NSCLC cell lines. Results are shown in Fig S2. None of the NSCLC cell lines analyzed expressed detectable levels of ACSL3 protein. By contrast, we were able to distinguish various levels of ACSL3 mRNA expression among cell lines. We validated our results using data from the Cancer Cell Line Encyclopedia [43] https://portals.broadinstitute.org/cc le/about) and COSMIC (https://cancer.sanger.ac.uk/c osmic) databases. Taken all data together, for stably overexpression of ACSL3, we decided to use the two cell lines with the lowest ACSL3 endogenous expression levels: CRL 5803 and CRL 5908. Overexpression of ACSL3 in both cell lines was confirmed by western blot and by qPCR (mRNA) (Fig S2D,F).
The enhanced cell proliferation, migration, and invasion are three essential characteristics for cancer progression and metastasis formation. Additionally, cancer cells reprogram their metabolic properties to facilitate cell survival and proliferation. To elucidate how ACSL3 overexpression affects the carcinogenic properties of NSCLC cells, we analyzed cell growth, migration, and invasion.
We used the xCELLigence TM system to assay cell proliferation and cell viability in real time (Fig 2A,B). We found that ACSL3 overexpression increased cell growth in CRL 5803 cells (54.61% of increase in the growth curve slope (1/h), P-value = 0.03 compared to the control No ORF CRL 5803). These changes were not observed in CRL 5908 cells. Then, we analyzed the migration capacities of ACSL3 overexpressing cells. We quantified cell migration starting from a serum-free environment to exclude proliferation differences (Fig 2C,D). Significantly, up to 57.15% more cells were found to migrate in ACSL3 CRL 5803 compared to the No ORF CRL 5803 control cells, and 135.36% more cells migrated in the ACSL3 CRL 5908 condition compared to the No ORF CRL 5803 control cells (P-values = 0.0075 and 0.0076, respectively). We also measured the invasive properties of ACSL3 overexpressing cells compared to their controls (Fig 2E,F). Both ACSL3 CRL 5803 and ACSL3 CRL 5908 displayed increased invasion through Matrigel TM when compared to their control cells (increases of 77.66%, P = 0.04 and 145.44%, P = 0.03, respectively). Student's t-test was used to evaluate statistically significant differences (**P < 0.01). (E) Boyden chamber transwell analysis of NoORF and ACSL3 in CRL 5803 and CRL 5908 cells invasion through Matrigel. After 48 h, cells were fixed and stained with crystal violet and measured under an optical microscope. Pictures were taken employing an Olympus CKX41 microscope (Olympus), with a 20X LCAch objective and recorder using analysis GETIT software (Olympus). Scale bars 500 µm. (F) Percentage of invasive cells counted under an optical microscope. Data are expressed as mean AE SEM of three independent assays each performed in triplicate. Student's t-test was used to evaluate statistically significant differences (*P < 0.05). (G) Oxygen consumption rate (OCR) of NoORF and GCNT3 CRL 5803 cells. Bioenergetics parameters were acquired by injecting 2 µM Oligomycin to inhibit ATP-linked OCR, 0.2 µM FCCP to uncouple mitochondria for maximal OCR, and 0.5 µM Rotenone/Antimycin A (Rot/AA) to shut down mitochondrial respiration. We show OCR measurements over time for cells stably expressing NoORF or ACSL3. We illustrate representative experiments of 6 measures (n = 3). (E) Quantification of basal respiration (oxygen consumption used to meet cellular ATP demand, calculated by subtracting non-mitochondrial OCR obtained upon Rot/AA injection) and spare respiratory capacity (capability to respond to an energetic demand, calculated as the difference between maximal and basal OCR) of NoORF and GCNT3 CRL 5803 cells. We illustrate representative experiments of 6 measures (n = 3). We next analyzed the expression of a panel of epithelial-to-mesenchymal transition (EMT)-associated genes. We tested E-cadherin, K18, and Na+/ K + ATPase b1 as epithelial markers and b-catenin, N-cadherin, Slug, and Vimentin as mesenchymal markers (Fig S3). A decrease in the E-cadherin expression levels was found in both ACSL3 overexpressing cells compared to their controls, although this reduction was only statistically significant in the ACSL3 CRL 5908 (P = 0.004). Also in ACSL3 CRL 5908 cells, we detected lower levels of K18 than in control cells (P = 0.02). This is in agreement with a loss of epithelial characteristics and gain of an EMT phenotype.
As ACSL3 seems to enhance the carcinogenic properties of NSCLC cells, we next evaluated the role of ACSL3 on cell bioenergetics, by monitoring the effect on the two main bioenergetic pathways: aerobic glycolysis and the mitochondrial oxidative phosphorylation. By mean of the Extracellular Flux Bioanalyzer (Seahorse Bioscience), we monitored (a) the extracellular acidification rate (ECAR) as an indirect readout of the lactate production by the aerobic glycolysis, and the (b) the oxygen consumption rate (OCR) as a readout of the oxidative phosphorylation at mitochondria (Materials and methods). Highly proliferative tumor cells frequently upregulate aerobic glycolysis (Warburg effect); nevertheless, we did not observed differences in the glycolytic performance of ACSL3 CRL 5908 compared to their control cells (data not shown).
As ACSL3 is a metabolic enzyme implicated in the activation of FAs, which is required prior to FAs entering into the b-oxidation pathway at mitochondria, we monitored the oxidative phosphorylation at mitochondria by mean of the quantification of the OCR after injection of several modulators of the electron transport chain. Remarkably, ACSL3 CRL 5803 cells displayed increased basal and maximal respiratory capacities (increased spare respiratory capacity in stressed conditions) compared to the control No ORF CRL 5803 cells (Fig 2G-bioenergetic profile, 2F).These results indicate that ACSL3 augments the cell bioenergetic capacity of the cells mainly by the enhancing the mitochondrial oxidative phosphorylation pathway, which may be related to the role of ACSL3 on FA activation for b-oxidation at mitochondria.
All these data confirm the carcinogenic role of ACSL3 overexpression by increasing cell growth, migration, and invasion. Moreover, our data propose that metabolic reprogramming is associated with an ACSL3 increase in cancer cells.

Statins in non-small cell lung cancer treatment
Anticancer use of non-oncologic drugs is an emerging area of research with promising results. Statins, that are lipid-lowering drugs, have been previously reported to display several anticancer properties. However, their exact roles in cancer initiation and progression have Relative gene expression was calculated using 0 µM sample as control. Data correspond to mean AE SEM of three independent assays each performed in triplicate. Student's t-test was used to evaluate statistically significant differences (*P < 0.05, **P < 0.01, ***P < 0.001). (E) Acyl-CoA synthetase isoforms abundance in non-tumoral lung tissue. Expression data in RPKM (Reads per kilo base per million mapped reads) obtained from Genotype-Tissue Expression (GTEx) project. (F) Gene expression in fatty acids synthesis pathway after statins administration. Relative expression assessed by means of qRT-PCR of ACSL1 and COX2 genes of fatty acids synthesis pathway in NoORF CRL 5803 and CRL 5908 cells after 48 h of increasing concentrations of simvastatin treatments. Relative gene expression was calculated using 0 µM sample as control. Data correspond to mean AE SEM of three independent assays each performed in triplicate. Student's t-test was used to evaluate statistically significant differences (*P < 0.05, **P < 0.01, ***P < 0.001). (G) Gene expression in apoptosis after statins administration. Relative expression assessed by means of qRT-PCR of CASP9 in NoORF CRL 5803 and CRL 5908 cells after 48 h of increasing concentrations of simvastatin treatments. Relative gene expression was calculated using 0 µM sample as control. Data correspond to mean AE SEM of three independent assays each performed in triplicate. Student's t-test was used to evaluate statistically significant differences (*P < 0.05, **P < 0.01, ***P < 0.001). not been clearly elucidated, as well as the proper conditions to be used. We used a public resource (https://depmap.org/repur posing/), containing the growth-inhibitory activity of 4518 drugs tested across 578 human cancer cell lines [44], to explore putative anticancer effects of statins. The most frequently used statin types have predicted selective activity in cancer cells. Statins selectively inhibited groups of cancer cell lines depending on their molecular features. Results are shown in Fig 3A. With the aim of studying the anticancer effects of statins in human NSCLC cells, we determined the antiproliferative activity of simvastatin in our CRL 5803 and 5908 cell lines (Fig 3B). As a quantitative measure of this, we analyzed the IC 50 parameter (concentration corresponding to 50% of cell viability Next, we examined the regulation of a selected panel of metabolic genes in the presence of growing concentrations of simvastatin. Nontreated cells were kept as reference controls. We measured genes related to apoptosis (CASP9) as well as genes implicated in de novo synthesis of cholesterol and FAs, genes regulating the intracellular lipid homeostasis-FA-activating genes, lipid droplet formation, FA oxidation, extracellular lipid, and cholesterol uptake-and plasma membrane remodeling enzymes, among others ( Fig 3C).
As expected, simvastatin treatment induced several transcriptional changes in the cholesterol biosynthetic pathway. We detected a consistent upregulation of HMGCR gene, ABCG8 transporter, and apolipoprotein A1 (APOA1) together with a slight downregulation of the expression of the ABCA1 cholesterol reverse transporter. These changes were observed in both NSCLC cell lines (Fig 3D).
The acyl-CoA synthetase isoform 1 (ACSL1) is the most frequent isoform found in the nontumoral lung tissue, as it is shown in Fig 3E from Genotype-Tissue Expression (GTEx) project. Simvastatin significantly modulated genes related to lipid metabolism, increasing the expression of ACSL1 FA-activating enzyme and COX2 expression implicated in the synthesis of prostaglandins E2 (PGE2) (Fig 3F).
Finally, we observed a robust induction of CASP9 expression in both cell lines (Fig 3G), indicating that simvastatin treatment induces apoptosis activation, as previously reported.
These data point out that simvastatin exerts an antitumoral effect modulating lipid metabolism and apoptosis in lung cancer tissue (Fig 3C).

Fatty acids metabolism through ACSL3 enzyme potentiates the antiproliferative effects of statins in NSCLC cells
Since fatty acid-related ACSL3 determines poor clinical prognosis in NSCLC, we analyzed the effect of statins administration in an ACSL3 overexpressing scenario, which would correspond to 76.6% of the NSCLC patients.
As it is shown in Fig 4A, Table 1). Importantly, the risk of death associated with ACSL3 overexpression increased up to 3.47 (95% CI: 1.57-7.69, P = 5.36 9 10 À4 ) in the 66 patients not following statin treatment vs those patients following a standardized statin treatment (n = 24), in which the expression of ACSL3 was not a marker of risk (HR: 1.23 (95% CI: 0.35-4.3), P = 0.75) (Fig 4C). In our population, there are a 22.22% of patients that overexpress ACSL3 and are having benefits from statin treatment. By contrast, there are a 57.77% of patients that overexpress ACSL3 and would advantage from statin antiproliferative action. Thereby, these results suggest that ACSL3 expression determines statins antitumoral activity and could be useful to select a potential candidate population of NSCLC patients that would significantly benefit from statin administration.

Discussion
Non-small cell lung cancer remains among the most lethal cancers, despite the numerous scientific efforts made in last decade in diagnosis and treatment of the disease [45]. Global NSCLC mortality continues increasing, and molecular biomarkers of diagnosis and prognosis are still needed. Nowadays, promising evidences of putative benefits of precision medicine allow us an opportunity to fight against these types of tumors. Precision medicine will converge in the identification of specific groups of patients with particular characteristics and consequently with personalized treatments [46].
One of the hallmarks of cancer is metabolic reprogramming [8]. Cancer cells modify their metabolic capacities and reprogram their energy sources; subsequently, this reprogramming offers us a putative therapeutic opportunity. Furthermore, lipid metabolism has been raised as a potential target for cancer diagnosis, prognosis, and treatment [9]. It has been shown that lipid metabolism alterations are interesting biomarkers of cancer prognosis in several types of tumors [10,11].
In the present study, we show that the expression of a lipid metabolism enzyme, ACSL3, classifies a group of NSCLC patients with poor survival, probably related to increased invasive properties moreover than proliferation-related events, as previously shown of the involvement of lipid metabolism in cancer progression. Importantly, ACSL3 expression is also categorizing patients who could benefit of putative antitumor properties of statin administration.
ACSL3 belongs to the family of long-chain acyl-CoA synthetases that is composed by five human isoforms (ACSL1, ACSL3, ACSL4, ACSL5, and ACSL6). They are lipid metabolic enzymes which convert free fatty acid to fatty acid-CoA. Based on the sequence homology, ACSL isoforms are separated into two groups: One is composed of ACSL1, ACSL5, and ACSL6 and the other ACSL3 and ACSL4 [17]. They have different subcellular locations and substrate preferences but all of them converge in the fatty acid synthesis pathway. ACSL3 prefers substrates like myristate, palmitate, arachidonate, and eicosapentaenoate [16] and localizes to the periphery of the lipid droplets of lipogenic cells, and it is also present on the cytoplasmic face of the endoplasmatic reticulum [47].
Long-chain acyl-CoA synthetases have been proposed as putative prognostic biomarkers of cancer. ACSL1 and ACSL4 overexpression was associated with a poor clinical outcome in stage II colorectal cancer patients [10,11,18,48,49]. In addition, ACSL4 is considered a biomarker for liver and breast cancer [50,51]. By contrast, downregulation of ACSL5 in breast cancer was associated with a worse prognosis [49,52] and none study have been reported on the association of ACSL6 and cancer survival [53]. An in silico study also suggested that high ACSL1 expression was associated with worse survival in lung cancer patients, and ACSL3 overexpression was associated with worse survival in patients with melanoma [49].
ACSL3 was found to be overexpressed in prostate cancer [54] and estrogen receptor-negative breast cancer [55]. Besides, tumor tissue microarray (TMA) staining [56] showed that ACSL3 is upregulated in lung cancer compared to the healthy lung tissue. However for the first time, in the present study, we report an association between ACSL3 expression with the prognosis and survival in NSCLC. In our set of patients, high ACSL3 expression levels showed an increased risk of death with a HR of 2.93 (95% CI: 1.5-5.73), P = 0.011 (Fig 1A). Moreover, we were able to replicate our results in a large sample of 984 NSCLC patients (HR: 1.38 (95% CI: 1.07-1.79), P = 0.016, Fig 1B). Importantly, these results are independent of putative confounding factors such as age, sex, stage, ECOG, histology, and smoking status and P-values are corrected for multiple testing.
The poor survival of patients who exhibit high ACSL3 expression levels is probably due to an increase in the carcinogenic capacities of the ACSL3 overexpressing cells. In our study, ACSL3 overexpression increased proliferation, migration, and invasion of tumor cells, which favors malignancy. ACSL3 overexpression increased cell growth in CRL 5803 cells but not in CRL 5908 cells. We hypothesized that the specific cellular background probably explains these differences. Nevertheless, migration and invasion properties are significantly increased in both cellular types upon ACSL3 overexpression (Fig 2C-F). Moreover, high expression levels of ACSL3 also augmented the mitochondrial oxidative capacity (increased basal and maximal respiratory capacity) (Fig 2G,H). It is tempting to speculate that ACSL3, due to its function on the activation of FAs, may contribute to augment the FA b-oxidation capacity of the cells conferring an advantage to tumor cells.
These results are in line with the observations that the ACSLs are upregulated in cancer. It has been proposed that ACSL-mediated lipid anabolism may support cancer initiation. Fatty acids promote tumor cell proliferation by providing crucial biosynthetic and functional intermediates. Independent knockdown of ACSL1, ACSL3, or ACSL4 decreased cell proliferation and anchorage-independent growth in many cancer cells and xenograft tumor growth in nude mice [50,53,[56][57][58]. By contrast, overexpression of these ACSLs increased cell proliferation and tumor growth [51,[59][60][61][62]. Also, ACSL1 and ACSL4 overexpression increases migration and invasion in colon cancer cells [18] but not proliferation and contribute to a non-Warburg advantageous energetic status of invasive colon cancer cells [57].
All these data allow us to hypothesize that, in NSCLC, ACSL3 overexpression is defining a metabolic scenario, associated with increased fatty acids activation conducting to an aggressive and invasive tumor cell and consequently with poor patient survival.
Statins are able to modulate the metabolic context since they are cholesterol-lowering drugs. They act by competitively inhibiting 3-hydroxy-3-methylglutarylcoenzyme A (HMG-CoA) reductase, decreasing the endogenous cholesterol synthesis. Emerging evidences suggest that statins display antitumoral, apoptotic, and immunomodulatory properties [21,27].
Statins have been shown to potentially decrease cancer risk and to improve survival in patients. Their anticancer effects have been described in several types of cancer, and several epidemiological and clinical studies have investigated the role of statins in protecting against lung cancer. Although much debate has been focused in describing their effects, it seems that high stage NSCLC patients could benefit of the statin administration [63,64].
The precise molecular mechanism by which statins generate anticancer effects is unknown, although appear to include inhibition of the mevalonate/cholesterol synthesis pathway, thus blocking the synthesis of intermediates key for prenylation and activation of the Ras/mitogen-activated protein kinase 1 signaling pathway. Moreover, statins have been recognized to modulate the phosphoinositide 3-kinase/Akt serine/threonine kinase 1 and inflammation signaling pathways and to modify lipid metabolism gene expression [27,65].
In the present study, we demonstrated that simvastatin exerts an antiproliferative action in NSCLC cell lines followed by transcriptional changes in lipid metabolism genes: HMGCR, ABCG8, APOA1, ABCA1, ACSL1, and COX2 (Fig 3B,C,D,G,F). ACLS1 isoform is the most abundant one in the healthy lung tissue ( Fig 3E); consequently, it is probably that ACSL1 transcription was the most finely regulated among ACSLs isoforms. We also demonstrated that simvastatin induces apoptosis in NSCLC cell lines via induction of CASP9 (Fig 3G).
Recent studies are exploring the use of common drugs for cancer therapy [44]. Based on them, we proposed that statins could be use in selective tumor scenarios ( Fig 3A). Here, we are tagging one of these scenarios; our data indicate that an ACSL3-upregulated context enlarges the antiproliferative effect of statins treatment in NSCLC cells (Fig 4A,B). In addition to ACSL3, it could be some other genes implicated in the mechanisms of response of tumor cells to statins.
In our study, and accordingly to recent reported results, we detected a trend between statin administration and better prognosis in NSCLC patients (Table 1). However, we found that significantly ACSL3 overexpression increases risk of death with HR of 3.47 (95% CI: 1.57-7.69, P = 5.36 9 10 À4 ) in patients not following statin treatment, while in those patients following statin treatment, ACSL3 is not a marker of risk ( Fig 4C).
Thereby, though a prospective clinical study would be necessary to further establish the clinical benefit of ACSL3 determination as a biomarker of statins benefit, these results strongly suggest that those patients overexpressing ACSL3 will obtain a clinical benefit from statin administration. The attractiveness of statins as part of chemotherapeutic drug arsenal lies in their relatively safe and well-tolerated toxicity profile together with their low cost. Thus, a metabolic biomarker of worse clinical outcome of NSCLC patients, together with a targeted well-tolerated treatment to overcome its aggressive potential, is the first time reported here.

Conclusions
A crucial point on cancer management is to diagnose patients who would benefit from specific therapies. Here, we found that the overexpression of a fatty acidrelated ACSL3 enzyme could be considered as an unfavorable prognostic marker in non-small cell lung cancer (NSCLC). Moreover, the overexpression of ACSL3 increased cell proliferation, migration, and invasion altering metabolic properties of lung cancer cells.
We found that statin administration caused antiproliferative effects, and importantly, those patients with high expression levels of ACSL3 following statins treatment displayed significant better prognosis than those without statins treatment.
ACSL3 emerges as a biomarker for clinical prognosis in NSCLC for the identification of patients that would benefit from anticancer properties of statin treatments that are safe, well-tolerated, and low cost.