Metastatic melanoma cells WM164 (Rockland WM164-01-0001, RRID: CVCL_7928) and HT144 (ATCC HTB-63, RRID: CVCL_0318) were grown in tissue culture flasks containing Roswell Park Memorial Institute (RPMI) medium supplemented with 5% (v/v) fetal bovine serum (FBS; heat inactivated for 30 min at 56 °C) and 1% (v/v) penicillin/streptomycin, placed in an incubator set at 37 °C with 5% CO₂. Cells were subcultured by dilution upon reaching ~80% confluence, every 2-3 days using 0.25% Trypsin-ethylenediaminetetraacetic acid (EDTA) solution.

The authenticity of both cell lines was verified using short tandem repeat profiling performed at the QUT Genomic Research Centre before the experiments were undertaken. Cells were regularly tested for the presence of mycoplasma using Lonza’s MycoAlert® Mycoplasma Detection kit via the mycoplasma core facility at Translational Research Institute (TRI).

DTPs and PDRCs were generated from parental, drug-naïve melanoma cells as described before^[16,17]. Briefly, cells were treated with 100 nM dabrafenib (GSK2118436; Selleckchem) added to the RPMI medium of drug-naïve cells at 80%-90% confluency and the medium was replaced every 3-4 days with fresh dabrafenib. Cells treated with dabrafenib for 7-30 days were classified as DTPs, while DTPPs emerged after ~30 days as distinct colonies. PDRCs were established after ≥ 105 days of continuous treatment. Permanent drug resistance was confirmed by continued proliferation following a 2-week drug holiday and dabrafenib rechallenge^[16].

Structural lipids were extracted from drug-naïve cells, DTPs and PDRCs as before^[18] and previously detailed in our study^[16]. Extractions were performed in nine samples per condition: three cell pellets (technical replicates) from each of the three independent culture flasks (biological replicates).

The lipids in the outer leaflet of cell membranes were digested using a method adapted from Lorent et al. (2020)^[19]. Briefly, cells were trypsinised, washed twice with Dulbecco’s Phosphate-Buffered Saline (DPBS), and resuspended at 1.2 × 10⁶ cells/mL. Aliquots of 250 µL (3 × 10⁵ cells) were dispensed into 2 mL glass vials in triplicate for the following treatments: (1) undigested control; (2) outer leaflet digestion with phospholipase A₂ (PLA₂); and (3) total digestion with PLA₂. For the undigested control, cells were kept on ice. Outer leaflet digestions [treatment (2)] were performed by adding 2 µL of PLA₂ (100 µM stock from Apis mellifera) directly to intact cells and samples were incubated at 37 °C for ≤ 30 min to prevent PLA₂ from digesting the inner leaflet of the plasma membrane. For total digestion [treatment (3)], cells were lysed by 6-8 freeze-thaw cycles and 5 min of sonication. Subsequently, 2 µL of PLA₂ was added to samples, which were incubated at 37 °C for the same duration as treatment (2). A 10 µL aliquot from each treatment was collected to assess membrane integrity using a lactate dehydrogenase (LDH) assay, as described below. PLA₂ activity was inhibited to stop membrane digestion by adding 100 µL of stop buffer (1 M HCl in saturated NaCl, pH 1-2). Lipids were extracted immediately under ice-cold conditions using 110 µL methanol, 370 µL MTBE with 0.01% (v/v) BHT and 3 µL SPLASH LIPIDOMIX® Mass Spec Standard. To induce phase separation, samples were vortexed for 1 h and centrifuged at 2,000 g for 5 min at 4 °C; the organic phase was transferred to a fresh 2 mL glass vial and stored at -80 °C until analysis. All experiments were performed using at least nine samples per treatment: three cell pellets (technical replicates) from each of the three to four independent cell flasks (biological replicates).

To confirm the integrity of the plasma membrane inner leaflet after PLA₂ treatment, we measured the enzymatic activity of cytosolic LDH released into the supernatant. 10 µL aliquot from each sample was diluted with 80 µL DPBS in individual wells of a 96-well black microplate. The LDH enzymatic reaction was initiated by adding 30 µL of substrate buffer containing 600 µM reduced nicotinamide adenine dinucleotide (NADH) and 24 mM pyruvate, with a final concentration of 150 µM NADH and 6 mM pyruvate. The decrease in NADH concentration was monitored over time by measuring fluorescence intensity (λ_excitation = 340 nm, λ_emission = 460 nm) every 30 s for 30 min using a CLARIOstar microplate reader. Samples from undigested and partially digested cells with intact inner leaflets maintained a relatively high and stable fluorescence signal, indicating the absence of LDH in the medium and confirming that the membrane integrity was maintained. In contrast, sonicated, fully digested cell samples showed a rapid decline in fluorescence intensity, indicating conversion of NADH to NAD⁺ by cytosolic LDH released into the medium.

We used a triple quadrupole mass spectrometer (6500 QTRAP, SCIEX, ON, Canada) and methods adapted from Young et al. (2021)^[20] and detailed in our previous study^[16]. SCIEX data files obtained from the 6500 QTRAP were processed using Lipidview® (Version 1.3 beta; SCIEX) software, including referencing biological lipids to the lipid class-based, deuterated internal standards and correction of abundances to account for naturally occurring lipid isotopes. Graphical visualisation of the data and statistical analysis were conducted using Microsoft Excel.

Drug-naïve, DTP and PDR cell samples treated with PLA₂ showed an increased proportion of lysophospholipids and corresponding decrease in phospholipids compared to undigested samples [Supplementary Figure 1], confirming outer leaflet digestion.

A method adapted from Liebisch et al. (2006) was used to quantify free Chol from cell lipid extracts^[21] and previously detailed^[16]. Chol quantification was performed using three biological replicates, each comprising derivatised extracts from three technical triplicates.

Cells were harvested using 0.25% Trypsin-EDTA, washed twice with cold DPBS before being counted and resuspended at 2.5 × 10⁵ cells/mL in cold DPBS, and stored on ice until measurement. Disposable zeta cuvettes (Malvern, Folded capillary zeta cell) were rinsed with 100% ethanol, then ddH₂O and finally DPBS. Cuvettes were filled with cell suspension and placed in the Zetasizer (Malvern, Zetasizer Nano ZS), where they were equilibrated for 500 s at 37 °C. Twelve measurements were acquired at a voltage of 40 V, with 50 sub-runs and 90 s pause between each run^[17]. The experiments were performed using cell suspensions obtained from three independent cell passages or flasks.

Melanoma cells were seeded (3 × 10⁵ cells/well) in a 96-well black microplate with a clear bottom the day before the assay. On the day of the assay, cells were washed twice with DPBS and serum- and phenol red-free RPMI medium containing 2.5 µM Di-4-ANEPPDHQ was added. Cells were incubated for 15 min at 37 °C to allow incorporation of the dye into the cell membrane, and fluorescence measurements were performed at 37 °C using a CLARIOstar microplate reader. Excitation was set to 480 nm, and emission spectra were recorded from 520 to 780 nm to observe spectral shifts and calculate the generalised polarisation (GP) using:

$$ GP=\frac{I_{560}-I_{650}}{I_{560}+I_{650}} $$

where I₅₆₀ and I₆₅₀ are the emission intensities at 560 and 650 nm, respectively.

Large unilamellar vesicles (LUVs; diameter of 100 nm) were generated using synthetic lipids 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC), 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-l-serine (POPS), sheep wool Chol, and brain SM from porcine origin purchased from Avanti Polar Lipids or freshly extracted lipids from 2 × 10⁶ cells as described in section “Extraction of structural lipids from cells”. Synthetic lipids were solubilised in spectrophotometric grade chloroform and mixed at the appropriate molar ratios to produce defined model membranes at 500 µM. Fresh cell lipid extracts or model membrane lipid mixture were dried under a stream of nitrogen gas to form a lipid film in a round-bottom flask and subsequently kept in a desiccator overnight to remove residual solvent. LUVs were prepared by extrusion as previously described^[16] in HEPES buffer (10 mM HEPES containing 150 mM NaCl, pH 7.4) with 2.5 µM Di-4-ANEPPDHQ. Labelled LUVs were aliquoted in a 96-well black Optiplate and incubated at 37 °C for 15 min prior to performing fluorescence spectroscopy measurements, as described in the previous section “Fluorescence spectroscopy and generalised polarisation measurements with live cells”.

Melanoma cells (3 × 10⁵ cells/well) were seeded in an 8-well chambered cover glass (Nunc™ Lab-Tek™ II) and incubated overnight. On the day of the assay, serum-containing medium was removed and cells were gently washed twice with DPBS before adding phenol red- and serum-free medium containing 1 µM Flipper-TR®, a fluorescent dye to measure membrane tension in cells and tissues^[22]. Images were acquired with a 40×/1.1NA Water HC Plan Apochromat objective on a Leica DMi8 SP8 confocal microscope with full incubation control running LAS-X software. FLIM images were captured using FALCON software and FAST FLIM files were output separately for further processing. Sufficient counts were acquired by exciting the sample with a tunable white-light laser at 480 nm, and the signal was detected between 575-625 nm, at a 400 Hz scan speed, over 20-line repetitions. Image analysis and quantification was performed using Fiji^[23], and utilised the BIOP-Utilities wrapper^[24] to generate regions of interest (ROIs) with Cellpose-SAM^[25]. Mean intensity of each ROI was measured using a batch macro and statistical analysis was performed using GraphPad Prism (version 10.6.1).

Melanoma cells were seeded (3 × 10⁵ cells/well) the day before the assay in an 8-well chambered cover glass. On the day of the assay, cells were washed twice with DPBS before serum and phenol red-free RPMI medium containing 50 µM 6-dodecanoyl-2-dimethylaminonaphthalene (Laurdan) was added, and the microscope chamber was placed back in the incubator for 15 min at 37 °C before microscope measurements. Images were acquired with a 40×/1.3NA Oil immersion Plan Apochromat objective mounted to a Zeiss LSM710 confocal microscope and Spectra Physics MaiTai 2 photon laser running Zeiss Zen Black software. Samples were illuminated with 790 nm excitation light, and emission was detected at 400-560 and 474-532 nm. Additional brightfield images were acquired with a T-PMT using a 488 nm laser on a separate track. Images were processed and quantified as described above, with Cellpose used to generate ROIs, which were then measured for both detection windows. The ratio of the two detection windows was calculated, and statistical analysis was performed using GraphPad Prism (version 10.6.1).

Cells obtained from three biological replicates from individual flasks were lysed and digested for proteomics analysis as previously detailed^[16]. Digested proteins were separated using an online U3000 RSLCnano nanoHPLC system with a 50 cm Easy-Spray C18 analytical column (Thermo Fisher, catalogue 160454 and ES803A) and analysed on a Q Exactive™ Plus Orbitrap™ Mass spectrometer (Thermo Scientific, USA). Raw data were processed using Proteome Discoverer (version 2.4.1.15) with protein identification performed against the Homo sapiens proteome (SwissProt Taxonomy 9606, database release v2020 10 18) and an internally curated protein-contaminant database.

Data are shown as mean ± standard deviation (SD). Technical triplicates were averaged to generate a single value for each biological replicate. Biological replicates (typically n ≥ 3) consisted of independent flasks (DTP and PDR cells) or cell passages (drug-naïve cells). The biological replicate means (or mean proportions, where applicable) were used for statistical analyses using GraphPad Prism (version 10.6.1). To assess significant differences between variables, one-way analysis of variance (ANOVA) was used to compare more than two groups or 2-way ANOVA with Tukey’s, Dunnett’s, or Sidak’s multiple comparisons test to evaluate more than two groups with different treatments. When P-value ≤ 0.05, results were considered statistically significant. Lipid quantification is expressed as proportions (mol%) to provide evidence of the relative changes in membrane composition independent of cell size and total lipid content.

We collected WM164 and HT144 cells at multiple time points (0, 7 days, and every 15 days between 15 and 105 days) during continuous treatment with dabrafenib, extracted their lipids and quantified the relative abundance of the four major classes of GPL [i.e., phospholipids containing phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylinositol (PI), or phosphatidylserine (PS)-headgroups] and of the sphingolipid SM using a mass spectrometry-based shotgun lipidomic method as before^[17,20]. In parallel, we quantified the Chol content using the same lipid extracts via a mass spectrometry method adapted from Liebisch et al. (2006)^[21]. Untreated controls (“drug-naïve cells”) matching culture age and passage number have been previously investigated, and these major lipid classes showed no significant differences over a three-month period in culture^[16].

In drug-naïve WM164 and HT144 cells (Figure 2A, 0 d), PC was the predominant lipid class (57 and 59 mol%, respectively), followed by PE (21 and 15 mol%), PI (12 mol% for both cell lines), PS (7 and 9 mol%) and SM (3 and 5 mol%). After 30 days of treatment with dabrafenib, notable changes in the proportion of these lipids were observed in the persister cells (Figure 2A, 30 d). In WM164 DTPs, SM increased 3.7-fold (from 3 to 11 mol%, P ≤ 0.0001), negatively charged lipids (PS + PI) increased 1.2-fold (from 19 to 23 mol%), while the combined proportion of the zwitterionic GPL (PE and PC) decreased from 78 to 66 mol%^[17]. In HT144 DTPs, SM and PE increased 1.8-fold (from 5 to 9 mol%, P ≤ 0.0001) and 1.3-fold (from 15 to 20 mol%, P ≤ 0.0001), respectively, whereas PC proportion decreased from 59 to 51 mol%. At the PDR state (105 days), the overall GPL and SM composition of PDR WM164 and HT144 cells reverted to resemble their respective drug-naïve profiles (Figure 2A, 0 d and 105 d)^[17].

The proportion of Chol and its storage form cholesteryl esters (CE) also fluctuated during acquisition of resistance [Figure 2B and C, Supplementary Table 1]. In WM164 cells, Chol content increased in DTPs, but was similar in PDRCs, whereas in HT144 cells, Chol was elevated in both DTP and PDR states compared with drug-naïve cells (see Figure 2B and Supplementary Table 1, 0, 7 and 105 d). In addition, in WM164 cells, the proportion of CE relative to total lipid increased during the first 7 days of treatment before returning to drug-naïve levels, whereas in HT144 cells, CE proportion steadily decreased as the cells acquired drug resistance [Figure 2C]. These findings suggest that development of drug resistance involves shifts in Chol and CE homeostasis, supporting the hypothesis that altered Chol metabolism plays a crucial role in regulating membrane properties and in promoting resistant cell survival.

Because FA chain length and saturation influence membrane fluidity^[29], we assessed potential changes in FA chains by determining the total carbon number and degree of unsaturation (i.e., lipid sum composition) of the two FA chains in GPL and SM species during treatment of WM164 and HT144 with dabrafenib.

In WM164 cells, the acyl chain length and saturation changed within the first 30 days of treatment (Figure 2D, see 0, 15 and 30 d, Supplementary File 1). The proportion of shorter-chain FA species in all GPL classes decreased, while longer and more unsaturated species increased. For example, after 15-30 days of treatment with dabrafenib, SM 36:1 and PI 38:4 increased significantly by approximately 3-fold, whereas PI 36:1, 36:2, and 34:1 decreased (see P-values in Supplementary File 1). Overall, these results suggest elongation and desaturation of GPL and SM species during the DTP phase, while WM164 cells in PDR state regained a FA profile similar to drug-naïve cells^[17].

In HT144 cells, progressive changes in FA chain length and saturation were observed over time, becoming most evident and significantly different in PDRCs compared with drug-naïve cells [Figure 2D and Supplementary File 1]. Specifically, long polyunsaturated species (> 2 double bonds) within the PE, PI, PS classes decreased, while shorter-chain species increased proportionally. For example, PE 38:4, PI 38:4, and PS 40:5 were reduced by half, whereas PE 34:2, PE 36:2, PI 36:2 and PS 36:2 increased by 2 to 3-fold^[17] (see P-values in Supplementary File 1). Changes in PC and SM species were not statistically significant.

Overall lipidomics results show that DTPs from both cell lines have higher levels of SM, Chol, and polyunsaturated FA chains, compared with drug-naïve cells. While increases in SM and Chol are typically associated with greater membrane rigidity^[30,31], polyunsaturated FA chains have the opposite effect by increasing membrane fluidity^[29]. In WM164 cells, PDRCs had a similar lipid composition to that of drug-naïve cells. In contrast, relative to drug-naïve cells, HT144 PDRCs showed reduced polyunsaturated FA chains and increased total Chol, changes typically associated with increased membrane order (rigidity).

To identify which GPLs are exposed on the outer leaflet of the cell membrane in metastatic melanoma cells, we adapted a method described by Lorent et al. (2020)^[19]. Cells at different resistance states (i.e., drug-naïve, DTP and PDR) were incubated with PLA₂ for 30 min to selectively digest GPLs located on the outer leaflet of the cell membrane. PLA₂ hydrolyses the sn-2 ester bond of GPLs, releasing both a FA and a lysophospholipid. Intact lipids extracted after partial digestion with PLA₂, representing those from the inner leaflet and intracellular organelles, were then compared with lipids from whole-cell extracts. The GPLs exposed on the outer leaflet were identified by subtracting the profiles of partially digested samples from those of whole-cell extracts [Figure 3].

Overall, all major GPL classes (i.e., PC, PE, PI and PS) were detected on the outer leaflet of both melanoma cell lines across all resistance states [Figure 4A]. Notably, the PS class, typically localised exclusively to the inner leaflet of healthy cells^[32,33], was also present on the outer leaflet of both melanoma cell lines, consistent with previous reports of PS externalisation in cancer cells^[34,35]. The combined proportions of the four GPL classes were comparable between the outer leaflet and the pooled inner leaflet and intracellular compartments in DTPs and PDRCs, except for WM164 DTPs (3 mol% on the outer leaflet compared to 8 mol% on the inside of the cell). Although shifts in lipid class proportions were observed, these differences were not statistically significant [Supplementary Figure 4]. In contrast, drug-naïve cells displayed apparent distributions between the outer leaflet and the other membrane regions (Figure 4A, pie charts): WM164 cells have a higher proportion of PI and PE, and lower PS on the outer leaflet, relative to the inner leaflet and intracellular compartments, whereas HT144 cells have a higher proportion of PI and lower proportions of PE and PS.

Analysis of the FA sum composition for each GPL class in drug-naïve cells revealed differences primarily in the PS species in WM164 [Figure 4B]. In these cells, PS 36:2 comprised 39 mol% of PS species exposed on the outer leaflet, compared to 19 mol% within the inner leaflet and intracellular organelles. PS 36:1 (29 mol%) and PS 34:1 (18 mol%) were the other two major abundant outer-leaflet PS species. A modest increase in PI 36:1 was observed on the outer leaflet (32 mol%), relative to intracellular levels (21 mol%). Drug-naïve HT144 cells showed comparable proportions of lipid species detected on the outer leaflet and within the cell.

In DTP cells, changes in the FA molecular species within each lipid class were more pronounced in the outer leaflet of WM164 than in HT144 cells, which were still dividing but at a slower pace than drug-naïve cells. In WM164 DTPs, the PS and PI FA species present in the outer leaflet are distinct from those in the rest of the cell (i.e., inner leaflet and intracellular membranes). In fact, PS 38:2 was the most abundant PS species on the outer leaflet (28 mol%) but represented only 5 mol% of total cellular PS (P ≤ 0.0001). Conversely, PS 36:1 accounted for 24 mol% of PS on the outer leaflet yet comprised about half of the cellular PS (P ≤ 0.0001). PI 38:4 was the predominant PI species in the outer leaflet and on the other membranes (P ≤ 0.0001). Other lipid classes showed no significant variation (< 10%) between the outer and inner leaflets and intracellular compartments. In HT144 DTPs, the lipid species of the outer leaflet were similar to that of the inner leaflet and intracellular compartments, except within the PS class, where PS 36:1 made up 58% of all exposed PS compared to 37% inside the cell (P ≤ 0.0001). In the PDR state, the lipid species composition on the outer leaflet mirrored that of the inner leaflet and intracellular membranes, particularly for WM164. In WM164 PDRCs, the lipid profiles of the exoplasmic membrane and cell interior were similar, with comparable proportions across lipid species and only variations within the PI class. In contrast, HT144 PDRCs showed significant differences in PI (PI 38:3, 38:4, 38:5) and PS (PS 36:1, 40:5, 40:6) species, while PC and PE species were similar in the outer leaflet and the cell interior [Supplementary File 2].

Overall, the lipid profile of HT144 cells, unlike that of WM164 melanoma cells, showed no significant differences between the outer leaflet and total cellular lipid composition [Supplementary File 2]. This may reflect their distinct responses to dabrafenib: HT144 cells continued to proliferate slowly under treatment, whereas WM164 cells exhibited an even lower proliferation rate in the DTP state (7-30 days of treatment) before the emergence of highly proliferative colonies^[5]. Nevertheless, the most pronounced differences in WM164 DTPs were the PS species in the outer leaflet, compared to inner leaflet and intracellular PS species.

We investigated the cell surface charge of drug-naïve, DTP and PDR WM164 and HT144 cells by measuring their zeta potential (i.e., the electric potential at the cell interface) when treated with dabrafenib^[17] (Figure 4A, top of pie charts). Cancer cells are often characterised by a negative electric charge^[36]. However, it is not known if the surface charge is maintained when cancer cells become DTPs or PDRCs.

Both drug-naïve cell types had negatively charged cell surfaces (WM164: -20.19 ± 1.37 mV, HT144: -19.34 ± 0.42 mV), which became less negative in DTPs (WM164: -15.94 ± 0.42 mV, HT144: -15.94 ± 0.59 mV) [Figure 3A]. However, a more negative surface charge was regained by PDRCs (WM164: -20.18 ± 0.95 mV, HT144: -17.7 ± 0.30 mV)^[17]. Although the values of the surface charge were not statistically different, the zeta potential correlated with the lower amount of negatively charged lipids found on the outer leaflet of DTPs than on drug-naïve and drug-resistant cells, and the slight variations could be due to other charged molecules at the cell surface (e.g., glycosaminoglycans).

Studies have shown that cancer cells can modulate their plasma membrane and become more rigid to prevent drugs from entering cancer cells^[14,37,38]. Due to the increase in the overall amount of Chol and/or SM, we hypothesised that the WM164 DTPs and the HT144 PDRCs could be more rigid than their drug-naïve counterparts, even though the proportion of Chol and SM on the outer leaflet was not quantified. We measured the fluidity and tension of the plasma membranes of cells at the various resistance states via fluorescence spectroscopy and microscopy using several fluorescent membrane dyes such as Laurdan, Di-4-ANEPPDHQ, and FLIPPER-TR.

We used Laurdan to examine the lipid packing of the membrane of melanoma cells and measured the GP using 2-photon fluorescence microscopy [Figure 5A]. Laurdan is sensitive to the lipid membrane order by responding to the polarity of its surrounding environment within the lipid bilayer, where shifts in emission spectrum are caused by variations in membrane water content that are quantified by calculating the GP^[39]. In general, the GP is lower in more fluid membranes, and higher in more rigid membranes. Melanoma cells labelled with Laurdan showed no significant difference in their GP measurements indicating that cell membrane lipid packing was similar between both melanoma drug-naïve, DTP and PDR cells.

Di-4-ANEPPDHQ probe is also sensitive to cell membrane order, but its fluorescence emission is more strongly influenced by overall Chol content, charge distribution and membrane dipole potential than that of Laurdan (see model membranes in Figure 5B)^[40]. Di-4-ANEPPDHQ was therefore used here to detect potential differences in the membrane organisation that may not be captured by Laurdan. The results show no significant difference among all three states of resistance in WM164 and HT144 cells or liposomes generated from lipids extracted from these cells, suggesting no significant changes between the membrane order of drug-naïve, DTP and PDR cells; however, a trend consistent with Chol proportions [Figure 2B] was noticeable (GP of cells in Figure 5B). These results may also reflect differences in membrane asymmetry and membrane potential rather than differences in membrane fluidity.

Although membrane fluidity and membrane tension are interdependent, they represent distinct biophysical properties. Increased membrane tension generally reduces fluidity by restricting lipid mobility and promoting a more ordered membrane state. We used the live-cell fluorescent probe FLIPPER-TR to measure membrane tension of WM164 and HT144 cells in the drug-naïve, DTP and PDR states [Figure 5C]. HT144 cells showed no significant differences in membrane tension in the three different states. In contrast, WM164 DTP cells had a slightly shorter fluorescence lifetime compared to the drug-naïve and the PDR cells, suggesting a modest decrease in membrane tension, and a slightly more fluid membrane.

The results obtained using live-cell imaging and spectroscopy with multiple membrane dyes indicate that membrane fluidity in DTPs is comparable to that of drug-naïve cells. The combined increase of Chol and SM in the DTP [Figure 2B], which would typically increase membrane rigidity, appears to be counterbalanced by the increase in polyunsaturated FA chains [Figure 2B and C, Figure 4B]. Overall, these live-cell measurements suggest that metastatic melanoma cells do not undergo major changes in membrane fluidity during development of dabrafenib resistance. However, they may still differ in other membrane features, such as transmembrane potential or microdomain organisation.

To investigate how melanoma cells remodel their proteome during resistance development, we performed whole-cell proteomic analyses on drug-naïve, DTP and PDR WM164 and HT144 cells^[17]. Total protein extracts from dabrafenib-treated cells were compared with those from drug-naïve controls and processed following the same workflow.

Across both cell lines, DTPs exhibited a larger number of differentially expressed proteins relative to drug-naïve cells than PDRCs [Figure 6A]. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway [Figure 6B and Supplementary Figure 5]^[41-43] and gene ontology (GO) [Figure 6C and Supplementary Figure 6] analyses revealed enrichment of proteins implicated in drug-resistance mechanisms, including lysosomal proteins and enzymes involved in degradation of glycosaminoglycan [Figure 6B and C]. Relevant for this study, we also identified upregulation of proteins in pathways that could modulate cell membrane properties^[17]; specifically, glycosaminoglycan degradation, glycosphingolipids and PI biosynthesis, and sphingolipid degradation.

Proteomics data suggest that WM164 DTPs may undergo enhanced degradation of glycosaminoglycans and other glycans, supported by the enrichment of hydrolases acting on glycosidic bonds [Figure 6B and C]^[17]. Glycosaminoglycans play critical roles in cancer proliferation, invasion, metastasis formation, and in therapeutic resistance^[44,45]. Given their negative charge, their degradation might also contribute to the less negative surface charge observed in the zeta potential of DTPs.

Proteins involved in glycosphingolipids biosynthesis (KEGG: hsa00603 & hsa00604) and in sphingolipids degradation (e.g., HEXA, GLB1, ARSA, ASAH1 and GBA)^[46] were upregulated in DTPs, but not in PDRCs [Figure 6B and C]. Sphingolipid degradation increases ceramide levels, which can be converted into SM via SM synthase^[47]. These proteomic changes possibly contribute to the increased SM detected in the lipidome of DTPs^[17].

Proteins involved in FA beta-oxidation [Figure 6C and Supplementary File 3], a mitochondrial process that catabolises FA to generate energy, were also upregulated in WM164 DTPs but not in PDRCs. A similar pattern was observed in HT144 DTPs, except for the proteins ACAA2 and ACSF1 which remained upregulated in PDRCs, while THEM4 was downregulated compared to drug-naïve cells. Other proteins involved in FA metabolism, such as FASN^[48] and ACSL3/4, were mostly downregulated in both DTP and PDRCs, with ACSL3 being an exception^[49].

Several enzymes involved in PI biosynthesis, including CDIPT, CDS2, PI4KA and PIP4K2B, were upregulated in the DTP state, especially in WM164 cells [Figure 6C and Supplementary File 3]. The upregulation of CDIPT and CDS2 likely contributes to the relatively high amount of PI (PI 38:4) detected in the lipidome as both enzymes support de novo PI synthesis^[50,51]. PI 38:4 also serves as the predominant precursor for PI-derived polyphosphates such as phosphatidylinositol 4,5-bisphosphate (PIP2) and phosphatidylinositol 3,4,5-trisphosphate (PIP3), which activates protein kinase B (AKT)^[52]. In turn, AKT signalling regulates multiple pathways associated with multidrug resistance, including inhibition of apoptosis, stimulation of metabolism and cell growth^[53]. Furthermore, the overexpression of PI4KA has been linked with cancer cell metastasis and invasion^[54].

Most proteins associated with Chol biosynthetic pathways (e.g., HMGCS1, DHCR7, DHCR24, FDFT1) were downregulated in WM164 DTPs, whereas apolipoprotein E (APOE), which is involved in Chol transport and lipid metabolism^[55-57], was strongly overexpressed^[17] (5.6-fold; log₂FC = 2.48; Figure 6D and Supplementary File 3). In PDR WM164 cells, the expression level of Chol biosynthetic proteins was similar to that of drug-naïve cells (fold change 0.6-1.2). In HT144 cells, these proteins were generally downregulated in both DTPs and PDRCs (fold change 0.5-0.9), except for the very low-density lipoprotein receptor (VLDLR)^[17], which was upregulated [Figure 6D and Supplementary File 3]. These findings suggest that the increased Chol detected in WM164 and in HT144 DTPs is likely due to increased Chol uptake from the extracellular medium (mediated by APOE in WM164, and by VLDLR in HT144^[58]), rather than increased de novo synthesis.

Other proteins involved in Chol homeostasis, including those involved in the conversion of Chol to CE (e.g., ACAT1 & 2, SOAT) and in the hydrolysis of CE to obtain Chol (e.g., LIPA), were also differentially regulated. Several aggressive cancers (e.g., melanoma, prostate, breast) upregulate Chol esterification via ACAT1 (acyl-coenzyme A: Chol acyltransferase-1)^[59]. Consistent with this, ACAT1 was upregulated in both WM164 and HT144 DTPs (2.3 fold in WM164 and 1.4 fold in HT144). In WM164 PDRCs, ACAT1 expression was similar to that of drug-naïve cells, whereas in HT144 PDRCs it was downregulated. These patterns may explain the higher CE levels observed in WM164 DTPs and the reduced CE levels in HT144 PDRCs relative to drug-naïve cells [Figure 2C].

On the other hand, LIPA, an enzyme that breaks down CE to form Chol, was modestly upregulated (1.7 fold; see Supplementary File 3) in both WM164 DTPs and PDRCs compared to drug-naïve cells. It was similarly increased (2.3-fold) in HT144 DTPs but downregulated in HT144 PDRCs (0.7-fold). Upregulation of LIPA may contribute to elevated Chol levels detected in the WM164 DTPs. Overall, these data highlight the complexity of Chol metabolism and homeostasis in melanoma and suggest that their dynamic regulation plays a crucial role in the development of cancer resistance.