Targeting lipid droplet lysophosphatidylcholine for cisplatin chemotherapy

Abstract This study aims to explore lipidic mechanism towards low‐density lipoprotein receptor (LDLR)‐mediated platinum chemotherapy resistance. By using the lipid profiling technology, LDLR knockdown was found to increase lysosomal lipids and decrease membranous lipid levels in EOC cells. LDLR knockdown also down‐regulated ether‐linked phosphatidylethanolamine (PE‐O, lysosomes or peroxisomes) and up‐regulated lysophosphatidylcholine [LPC, lipid droplet (LD)]. This implies that the manner of using Lands cycle (conversion of lysophospholipids) for LDs might affect cisplatin sensitivity. The bioinformatics analyses illustrated that LDLR‐related lipid entry into LD, rather than an endogenous lipid resource (eg Kennedy pathway), controls the EOC prognosis of platinum chemotherapy patients. Moreover, LDLR knockdown increased the number of platinum‐DNA adducts and reduced the LD platinum amount. By using a manufactured LPC‐liposome‐cisplatin (LLC) drug, the number of platinum‐DNA adducts increased significantly in LLC‐treated insensitive cells. Moreover, the cisplatin content in LDs increased upon LLC treatment. Furthermore, lipid profiles of 22 carcinoma cells with differential cisplatin sensitivity (9 sensitive vs 13 insensitive) were acquired. These profiles revealed low storage lipid levels in insensitive cells. This result recommends that LD lipidome might be a common pathway in multiple cancers for platinum sensitivity in EOC. Finally, LLC suppressed both cisplatin‐insensitive human carcinoma cell training and testing sets. Thus, LDLR‐platinum insensitivity can be due to a defective Lands cycle that hinders LPC production in LDs. Using lipidome assessment with the newly formulated LLC can be a promising cancer chemotherapy method.


| INTRODUC TI ON
Platinum-based chemotherapy is the primary modality used for treating patients with solid tumours. Platinum is often considered a first-line chemotherapy drug. [1][2][3] Moreover, platinum has been combined with various chemoagents and used for treating cancers for which effective drugs are unavailable to date. 3 However, the clinical utility of platinum is limited because of the resistance of certain types of cancers to platinum and normal tissue toxicities, which are determined by the level of platinum accumulation in tissues. 1 Among the various reported mechanisms, the most acceptable hypothesis in this field was the expression of ATP-binding cassette (ABC) gene family in cancer cells. 4,5 The expression of the ABC gene family can pump the intracellular chemotherapy drug out of the cells to avoid the cytotoxic effect. 6 The use of an ABC gene blocker was tested for treating multiple drug resistance; however, the results were controversial. For instance, in a study, ABC blockers were used in multiple trials. 7 However, no evidence suggests ABC blockers to be an effective second-line chemotherapy drug.
Lipids, essential biological building blocks, can act as bioactive molecules, such as they can be constituents of cellular membranes or as a supplier of a sufficient amount of energy for the fast-growing nature of cancer cells. 8 One of the most important metabolic markers of cancer cells is the deregulation of lipid metabolism. 9 Recent studies have shown that lipid metabolism plays crucial roles for providing energy, macromolecules for membrane synthesis and lipid signals during cancer progression. 10 Moreover, lipid droplet (LD) accumulation in cancer cells is a pivotal adaptive response to deleterious conditions. 11 Lipid importation from circulation, other than endogenous lipid synthesis, affects cancer progression has been attracting tremendous attention in the field of cancer metabolism. 12 In a bioinformatics study, Li et al 13 found that low-density lipoprotein receptor (LDLR) expression is an important biomarker for renal cell carcinoma. Cholesterol importation occurs in pancreatic cancer potentially through LDLR expression in cancer cells. 14 Moreover, cholesterol importation 15 to steroidogenic enzymes except endogenous cholesterologenesis through LDLR and SR-D1 scavenger receptor is the prognostic biomarker of gastric cancer (GCa). 16 Targeting the steroidogenic enzyme cytochrome CYP450 19A1 (aromatase) with exemestane can be effective for patients with GCa. 17 Study shows that differential levels of LDLR expression in EOC cells determine the platinum sensitivity in an LDLR-dependent manner. 10 LDLR expression reprogrammes cellular transcriptome associated with lipid metabolism (Lands cycle in LD) to be the mechanism underlying cisplatin sensitivity. Moreover, lysophosphatidylcholine (LPC) acyltransferase 1/2 (LPCAT1/2), a Lands cycle enzyme, has been recognized as a key chemoresistance molecule in multiple cancers. 18,19 Abundant evidence has indicated that intracellular lipid resources, either endogenous or exogenous, are the key biochemical event indicating chemotherapy responsiveness of multiple cancers.
In the present study, we explored LDLR-mediated lipidome alteration for platinum therapy sensitivity by using lipidomics and bioinformatics approaches for illustrating cellular lipidome and testing the hypothesis that targeting lipidome, instead of gene expressions, is a useful therapeutic strategy.

| Reagents, cell culture and lentiviral-based gene delivery
Cells were maintained in various culture media depending on the culture requirements with 10% FCS (foetal calf serum; Invitrogen), 1% L-glutamine and 1% penicillin-streptomycin, as described previously. 15  and LPC (855675P, Avanti) were used.

| Lipid profiling for lipidome analysis
After the cells (1500 cells/µL × 300 µL) were washed with Ca 2+ -or Mg 2+ -free PBS, the lysates were subjected to lipid profiling executed by Lipotype GmbH. [20][21][22] Lipidomes were prepared from at least three replicates of each sample for all the experiments by using the subsequently described procedures.

| Nomenclature
The following lipid names and abbreviations were used: Cer, cera-

| Lipid extraction for mass spectrometry lipidomics
Lipids were extracted using a two-step chloroform-methanol pro- After extraction, the organic phase was transferred to an infusion plate and dried in a speed vacuum concentrator. Each first-step dry extract was resuspended a 1:2:4 (v/v/v) chloroform-methanolpropanol in 7.5 mmol/L ammonium acetate, and each second-step dry extract was resuspended in a 0.003:5:1 (v/v/v) methylaminechloroform-methanol in 33% ethanol. All liquid handling steps were conducted using the Hamilton Robotics STARlet robotic platform with the Anti-Droplet Control feature for organic solvent pipetting.

| Data analysis and post-processing
Lipid identification was performed on unprocessed mass spectra by using LipotypeXplorer (2). For the MS-only mode, lipid identification was based on the molecular masses of the intact molecules. The MS/MS mode involved the collision-induced fragmentation of lipid molecules. Moreover, lipid identification was based on both intact and fragmented masses. Before normalization and further statistical analysis, the lipid identifications were filtered, based on mass accuracy, occupation threshold, noise and background characteristics. The lists of the identified lipids and their intensities were stored in a database optimized for the particular structure inherent to lipidomic data sets. Lipid class-specific internal standards' intensity was used for lipid quantification. 24 The identified lipid molecules were quantified using normalization to a lipid class-specific internal standard. The amount of individual lipid molecules (species of subspecies) in p moles of a given lipid class was summed to yield the total amount of the lipid class. The lipid class amounts may be normalized to the total lipid value by yielding the mol.% for the total lipid amount.

| Lipid data processing
The lipid profiling data of each sample were scale-normalized by the total amount of lipid. The lipids with at least a twofold change between the LDLR knockdown cells and control cells were identified as lipids significantly regulated by LDLR. Then, Fisher's exact test was conducted to test the enrichment of the significantly regulated lipids for each lipid class, such as PC, PE and LPC.

| Lentiviral-based gene delivery
LDLR knockdown cells or overexpression clone cells were engineered by the stable transfection of human LDLR cDNA (pLenti-C-mGFP-LDLR, RC200006L2; OriGENE) or pLKO.1-shLDLR (targeting sequence: 5′-GGG CGA CAG ATG CGA AAG AAA) 10 and then selected after exposure to puromycin (10 μmol/L) for a month. [25][26][27] The pLKO-shLuciferase plasmids were obtained from the National RNAi Core Facility Platform (Institute of Molecular Biology or Genome Research Center, Academia Sinica, supported by the National Core Facility Program for Biotechnology; grant number: NSC107-2319-B-001-002). The lentiviral production and infection procedures used in this study followed those reported previously. 25 In brief, psPAX2 (packaging plasmid) and pMD2G (envelope plasmid) (Addgene) were cotransfected into the HEK293T cells. Then, the virus-containing media were harvested to infect the HCC cells.
The GFP + cell populations, as determined by the flow cytometry analysis (BD LSR II Flow Cytometry), were used to test the infection efficiencies.

| The Cancer Genome Atlas database DriverDB (version 2) and Kaplan-Meier plotter meta-analysis for cancer survival analysis and algorithm for hazard ratio scoring
DriverDB, 28,29 a database that incorporates >9500 cancer-related RNA-Seq data sets and >7000 exome-seq data sets from The Cancer Genome Atlas (TCGA), was used in this study. DriverDBv3 30 comprises 420 primary tumour data and 37 adjacent normal tissue data (including 34 normal-tumour pair data) in the EOC data set in TCGA.
For conducting the survival analysis of TCGA data, Kaplan-Meier (KM) survival curves were drawn. Moreover, a log-rank test was performed to assess the differences between the patient groups stratified by the median of gene expression. A P value of <.05 was considered statistically significant.
In the web-based KM plotter platform, the following previously established formula was used 15 to evaluate the hazard ratios (HRs) of the pathways (cluster of genes) with respect to patient survival: To evaluate the influence of each gene, the absolute HR for that gene minus 1 was calculated. To adjust for the gene effects, the HR of each gene was multiplied by a negative log 10 p for balancing the importance of the genes. The summed score was then divided by the number of genes and multiplied by 100 to obtain the HR or average HR of all the genes.

| Doubling time and 50% inhibition concentration measurement
For the cell viability assay, cells were seeded in 96-well plates (5 × 10 3 cells/well) and incubated overnight for attachment. These cells were then treated with the indicated drug doses in normal media for 48 hours. After the treatments, the media were replaced with MTT assay (0.5 mg/mL) at 37°C for at least 1 hour. After the removal of excess WST-1 (Sigma-Aldrich), the colorimetric absorbance of the cells at 490 nm was recorded. The measured values of 50% inhibition concentration (IC 50 ) 31 for each drug were determined using the CalcuSyn software 32 (BioSoft).
The calculation of the doubling time of cell lines in this study was the same as the procedure described on http://www.doubl ing-time.
com/compu te.php and calculated using the following equation: In brief, cells were seeded at 2 × 10 4 cells/6-well dishes, and the cells were collected after 24, 48, 72, 96 and 120 hours of seeding.
Then, the cells were placed on a cell counting chamber slide, and the cell concentration was recorded.

| LPC-liposome-cisplatin preparation and characterization
The liposome was prepared using a thin-layer hydration, 33  with 500 µL of purified water in a low-volume disposable sizing cuvette. The particle size and size distribution were measured in terms of the ZAve and polydispersity index, respectively.

| Cisplatin-DNA adduct measurement with flow cytometry
Platinum-DNA adduct measurement was described previously. 35 In brief, the cells were subjected to various treatments for 24 hours to form cisplatin-DNA adducts. Then, these cells were harvested using PBS containing 0.1% Triton X-100 and fixed with 70% ethanol. Subsequently, the cells were washed with PBS and stained with an anti-cisplatin-modified DNA [CP9/19] antibody (Abcam, 1:1000 dilution) overnight at 4°C. The cells were then stained using a goat (1) anti-rat FITC-conjugated secondary antibody for 2 hours. The signals of the cisplatin-DNA adducts were detected through flow cytometry (BD Biosciences). The data were analysed using FCS Express (version 3.0; De Novo Software).

| LD isolation
The LD isolation was conducted using a published protocol. 36,37 In summary, the cells were cultured with 20% FCS in a 150-mL flask.

| Quantitation of cisplatin uptake by cells
The platinum quantitation method for cells was modified from a previous study. 38 The platinum levels in cells were detected and

| Statistical analysis
Student's t and chi-square tests were conducted to identify the significant differences between groups and categorical variables.
A P of <.05 was considered significant. All data were presented as means ± standard errors of the means.

| LDLR-mediated LD lipidome alteration for cisplatin sensitivity
As We found that high LDLR uptake genes (eg LDLRAP and LPL) were associated with a poor PFS (Figure 2A,C). However, by calculating the effect of multiple genes on PFS, we found that LDLR uptake genes had a more severe influence on platinum therapy patients (evidenced by the differences in HRs calculated using

| Targeting LPC for manufacturing an efficacyboosting cisplatin-liposome drug
Notably, the discovery of the LDLR → LD lipidome → platinum therapy efficacy axis for targeting lipidome (eg LPC) could be of great significance in therapeutics. We conducted lipid profiling, defined as cisplatin cytotoxic at IC 50 , for testing this assumption and measured Liposome has been applied in pharmaceutics for decades as a drug carrier 48,49 and is particularly effective for delivering poor water-soluble or highly toxic small-molecule compounds. 50 The addition of drug efficacy-enhancing lipids while manufacturing liposomes has been proposed in pharmaceutics. 51 However, there have  to test its cytotoxic efficacy. As shown in Figure 4A, we successfully manufactured LLC for a nanoscale homogenous liposome particle. For comparing the conventional liposome to the LLC, we compared the cytotoxic efficacy of DOTAP-liposome-cisplatin (DLC) with that of LLC. As shown in Figure 4B body weight, we did not find obvious body weight change during treatments ( Figure 4F). After treatment term, we found the tumour was smaller in LLC-treated than in DLC-treated group ( Figure 4G,H).
The Tumor Suppression Index is 66.28%. These data implicated a potential of implementing LLC in the future clinical application.
As LPC can be targeted using LLC for cisplatin insensitivity, we examined whether this process was an LD-related event. In this study, we used SKOV3 LDLR low-expressed cisplatin-sensitive cells and MDAH-2774 LDLR high-expressed cisplatin-insensitive cells.
Then, the LDLR expression levels were manipulated ( Figure 5A) or LLC was used to measure the platinum amount in an LD and detect DNA adducts in the cells. As shown in Figure 5B,C, we compared the platinum amount in LDs isolated from MDAH-2774 cells by using LDLR parental (par) cells and LDLR knockdown (KD) cells. The results revealed that LDLR knockdown could significantly reduce the platinum amount in the LDs ( Figure 5B). By contrast, the amount of platinum in the non-LD part is comparable for both parental and knockdown cells ( Figure 5C). Based on the DNA adduct formation measurement ( Figure 5D, left panel), the formation was higher in the MDAH-2774 cells treated with LLC compared with the cells treated with cisplatin or DLC ( Figure 5D, right panel).  Figure 5D, right panel). The data in Figure 5E-G are the reverse proof that LDLR-->LPC-->LD metabolism is the important mechanism for cisplatin sensitivity. Finally, we observed the expression of perilipin-2 (an LD marker) in LDs from SKOV3 and MDAH-2774 cells. We found that LDLR overexpression increased perilipin-2 expressions in LD, but LDLR knockdown suppressed the expressions ( Figure 5H). The data presented in Figure 5 suggest that altering LDLR, indeed, reprogrammed LD metabolism and could carry platinum. Therefore, platinum was eventually pumped out of the cancer cells. Moreover, altering LD metabolism by using LLC could improve the therapeutic efficacy.

| Mechanism of LDLR-LPC axis in platinum insensitivity
We previously delineated the regulatory axis 'LDLR→LPC→FAM83B→FGFRs' for altering cisplatin sensitivity (not yet published manuscript). In the current study, we explored the LDLR-related lipidome alteration in organelles and particularly focused on the LD function in terms of the platinum therapy efficacy.
The major finding and significance of this study can be illustrated in An issue observed based on the data presented in Figure 5F that the use of LLC could reduce the platinum amount in both LD fractions should be addressed. These data suggest that LD could partially explain the LDLR-mediated cisplatin sensitivity. However, in Figure 5G, the platinum amount was also reduced in the non-LD fraction. These data clearly suggest that LPC might also alter other signals for improv- Targeting LPC to reprogramme LD metabolism could reduce the possibility of pumping platinum out of the cell. Thus, the possibility of platinum binding on DNA to form DNA adducts (green) was increased and promote stem cell proliferation, which is related to colorectal cancer tumorigenesis. 58 A recent study discovered that LD is a drug reservoir that is responsible for drug depletion in the macrophage of the antibiotics resistance state. 59 Moreover, although a few previous reports have discussed the roles of LDs in cancers, no study has discussed the potential of targeting LD for cancer therapy. The results of this study revealed that the experimental treatment of LPC with cisplatin did not alter insensitive cells. However, treatment with LLC exhibited excellent growth-suppressing efficacy. The possible mechanism is as follows: the LPC-liposome structure (which is a LPC-cholesterol mixture) can mimic LD structure and then fuse with the LDs to remodel their lipid composition. There is great interest in exploring this possibility further for developing new drug delivery methods. Moreover, this result suggests the great potential of targeting LDs for cancer therapy. The concept of nanodroplet 'adiposome' 60 was proposed as a tool for drug delivery, although this concept requires further validation.
In conclusion, here, we stated our novel findings explaining the cellular mechanism of LDLR-mediated cisplatin insensitivity from the lipid metabolism perspective. We found that LDLR expression confounds platinum chemosensitivity. Moreover, the novel LDLR→LPC→FAM83B→FGFRs regulatory axis revealed through transomics analysis may explain the discrepancies in platinum chemosensitivity. 10 Finally, LDLR-altered LD homeostasis contributed to platinum sensitivity, suggesting the potential value of targeting LDs with LLC treatment. Moreover, the lipidome profiling of cancer cells in association with drug sensitivity might be useful for manufacturing drug-specific liposomes for pharmaceuticals.

ACK N OWLED G EM ENTS
We acknowledge the instruction from Dr Keen Chung for the liposome preparation for drug delivery. We also acknowledge the technical support provided by Lipotype GmbH (Dresden, Germany) and Taiwan

CO N FLI C T O F I NTE R E S T
The authors of this study declare no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data sets can be provided by the corresponding author upon reasonable request.