LC-MS/MS and high-spatial single-cell MSI reveal selective downregulation of PC(16:0_16:0) and PC(18:0_18:1) due to mPGES-1 knockdown in hormone-resistant prostate cancer cell line

Abstract

Prostate cancer is one of the most common cancers in men, and 10–20% of cases progress to a hormone-resistant form, posing a major therapeutic challenge. The microsomal prostaglandin E2 synthase-1 (mPGES-1), a key enzyme in prostaglandin E2 biosynthesis from arachidonic acid, has been predicted as a therapeutic target in suppressing prostate cancer. However, the impact of mPGES-1 on the cellular lipidome has not been systematically investigated. Here, this study applied LC-MS/MS analysis and high-spatial single-cell mass spectrometry imaging (MSI) analysis using atmospheric pressure matrix-assisted laser desorption/ionization-MSI (AP-MALDI-MSI) to investigate the role of the mPGES-1 gene in altering the lipidome of hormone-resistant DU145 prostate cancer cells. LC-MS/MS results revealed significant downregulation of phosphatidylcholine (PC) and triglyceride (TG) species, and upregulation of ceramide species in KD (stable knockdown of the mPGES-1 gene) compared to controls. Upregulation of PCs is the signature of the tumor microenvironment of prostate cancer cells, and the upregulated ceramides are associated with programmed cell death. Consistent findings were observed in high-spatial single-cell MSI analysis for PC(16:0_16:0) and PC(18:0_18:1), which were significantly (p < 0.001) downregulated in KD compared to controls, and also revealed a striking loss of metabolic heterogeneity of these PC molecules. The loss of metabolic heterogeneity due to mPGES-1 knockdown may reduce cancer cells’ adaptive capacity, suggesting mPGES-1 inhibition as a strategy to overcome metabolic plasticity in hormone-resistant prostate cancer. Thus, this study highlights that suppressing mPGES-1 can be a future therapeutic target in preventing the progression of hormone-resistant prostate cancer development.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s00216-025-06296-y.

Introduction

Prostate cancer is the second most frequently diagnosed malignancy in men over 65 years and remains a major public health concern worldwide [1]. Prostate cancer is commonly diagnosed at localized or regional stages, for which curative-intent treatment with radical prostatectomy or definitive radiotherapy (external-beam or brachytherapy), often combined with androgen deprivation therapy in higher-risk disease [2], achieves excellent outcomes; population data show ≥ 99% 5-year relative survival for localized/regional disease [3]. By contrast, advanced or metastatic disease is managed with androgen deprivation therapy as the systemic backbone, with early treatment intensification improving survival but rarely achieving a cure [4]. Despite initial disease control, approximately 10–20% of patients ultimately develop castration-resistant prostate cancer (CRPC) within about 5 years [5, 6]. CRPC represents an aggressive, ultimately fatal stage of the disease with limited treatment options. Therefore, understanding the molecular mechanisms underlying CRPC development and identifying novel therapeutic targets remains a critical unmet need in prostate cancer management. Chronic inflammation is increasingly recognized as a key driver of prostate cancer progression, contributing to the remodeling of the tumor microenvironment and promoting metastatic dissemination [7, 8]. Indeed, among the mediators of inflammation, prostaglandins, lipid-derived molecules synthesized from arachidonic acid (AA) through a cyclooxygenase (COX)-dependent pathway, play pivotal roles in tumorigenesis [9]. Among them, prostaglandin E2 (PGE2) produced by microsomal prostaglandin E synthase-1 (mPGES-1) from prostaglandin H2 (PGH2) [10] has been implicated in immune evasion, angiogenesis, and proliferation in various cancers, including prostate cancer [11, 12]. Consistent with these observations, non-steroidal anti-inflammatory drugs (NSAIDs) targeting COX-2 have been investigated for prostate cancer prevention, with some studies suggesting a protective effect, particularly for aspirin [13, 14]. Aberrant expression of mPGES-1 in combination with COX-2 can contribute to tumorigenesis [15].

mPGES-1 is a terminal enzyme in the PGE2 biosynthetic pathway from arachidonic acid, catalyzing the isomerization of PGH2 into PGE2 [16]. Elevated expression of mPGES-1 has been reported in aggressive prostate cancer phenotypes, and pharmacological inhibition of mPGES-1, such as with the selective compound KH176m, the active metabolite of sonlicromanol, which has been shown to suppress tumor growth in vitro and in vivo [17]. mPGES-1 is a component of arachidonic acid metabolism, linking inflammatory signaling to lipid pathways. Altered lipid metabolism is now recognized as a hallmark of prostate cancer progression, influencing tumor growth, survival, and therapeutic resistance [18, 19]. However, the impact of mPGES-1 knockdown on the broader lipidome of prostate cancer cells remains largely unexplored. DU145 cells, derived from a brain metastasis of prostate cancer, represent an androgen-independent model of advanced disease, thus offering a relevant system for studying castration-resistant prostate cancer biology [20]. Previous studies have shown that mPGES-1 is highly expressed in DU145 cells compared to benign hyperplasia and that its knockdown attenuates tumor progression and induces necrotic changes in xenograft models [21]. Understanding how mPGES-1 modulates the cellular lipid alterations could provide new insights into prostate cancer progression and potential therapeutic targets.

This study applied an integrated approach combining untargeted liquid chromatography-tandem mass spectrometry (LC-MS/MS) with spatially resolved single-cell mass spectrometry imaging (MSI) using atmospheric pressure matrix-assisted laser desorption/ionization-MSI (AP-MALDI-MSI) to investigate the lipidomic alterations induced by mPGES-1 knockdown in hormone-resistant DU145 prostate cancer cells. LC-MS/MS enables comprehensive and sensitive profiling of lipid species. On the other hand, the MSI technique is an advanced emerging tool that has been applied to the spatial analysis of lipids [22–24], proteins [25, 26], glycans [27], drugs [28], metabolites [29, 30], as well as quantifications [31] in biological tissues. It provides localization information on the biomolecules, drugs, and metabolites in tissues. In recent advancements, the instrumental capability for MALDI-MSI analysis has been substantially improved to attain a high spatial resolution of ~ 5 µm pixel size, allowing the visualization of the single cell [32, 33]. Some recent studies have reported lipidomics analysis at the single-cell level with high spatial resolution using trapped ion mobility MALDI-MSI and AP-MALDI-MSI techniques with machine learning algorithms to identify cellular heterogeneity [34, 35]. In addition, the AP-MALDI mass microscope has the advantage of being an integrated microscope and has proven sensitivity for data acquisition with high spatial resolution [28, 36], which leads to its application in spatial analysis at the single-cell level.

Therefore, this study applied LC-MS/MS and AP-MALDI-MSI modalities to explore the knockdown effect of the mPGES-1 gene on lipidomic changes in hormone-resistant prostate cancer cells, and we found a significant differential distribution of phosphatidylcholine (PC), triglycerides (TG), and ceramide species in the knockdown cells compared to the controls.

Materials and methods

Chemicals

Hematoxylin, eosin, xylene, pathomount, paraformaldehyde (PFA), LC-MS grade methanol, chloroform, ethanol, 2-propanol, ultrapure water, acetonitrile, and formic acid were purchased from Wako Pure Chemical Industries (Osaka, Japan). 2,5-Dihydroxybenzoic acid (DHB) was purchased from Bruker Daltonics (Bremen, Germany). Fetal bovine serum and 9-aminoacridine (9AA) were purchased from Sigma-Aldrich (St. Louis, MO, USA). RPMI-1640 medium was purchased from Thermo Fisher Scientific (USA). DLPC [PC (12:0_12:0)] was purchased from Sigma-Aldrich (USA).

Cell culture

This study was conducted using the DU145 cell line. DU145 WT (ATCC HTB-81, certified by STRA) is a prostate cancer cell line with high constitutive expression of the mPGES-1 gene. The study was designed to include three groups: mPGES-1 knockdown (KD) cells, which were transfected with an shRNA-expressing vector specifically targeting mPGES-1 to achieve stable gene silencing; wild-type (WT) cells, which were unmodified and not exposed to any vector; and negative control (NC) cells, which were transfected with non-target shRNA sequences containing vector [21]. All DU145 cells were cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum at 37 °C in a humidified atmosphere containing 5% CO₂.

For spatial single-cell MSI analysis, each group of cells was cultured on the conductive indium tin oxide (ITO) coated glass slide placed inside a chamber (Fig. S1), where an equal number of cells was seeded and incubated under the conditions as mentioned above. After 48 h of incubation, when the cell confluency reached around 50%, the slides were collected and prepared for the MALDI-MSI measurement.

Lipid extraction from cultured cells for LC-MS/MS analysis

Whole cell lipids were extracted by the Bligh and Dyer method with slight modifications [37] for LC-MS/MS analysis. A similar number of DU145 cells (KD, WT, and NC) were seeded on 6-cm cell culture dishes and cultured until the cell confluency reached approximately > 90% (for each cell type). The cells were then collected for lipid extraction. First, the dishes were placed on wet ice to slow down metabolic activity, and the cells were harvested by scraping and transferred into 15-mL glass tubes. After one wash with ice-cold PBS, 800 µL of ultrapure water was added to each tube. Subsequently, 2 mL of cold methanol was added, and the mixture was vortexed. The mixture was then sonicated using a bioruptor for 5 min. One milliliter of chloroform was added, vortexed to mix well, and incubated for 10 min at room temperature. Again, 1 mL of chloroform containing the internal standard (PC(12:0_12:0)) at a concentration of 10 ng/µL was added and vortexed to mix thoroughly. Then, 1 mL of 0.28 M acetate was added and mixed well. Finally, it was centrifuged for 10 min at 4 °C at 4000 rpm, and the lower phase contained the lipids and fatty acids. The lower organic phase was carefully collected and dried under vacuum at 4 °C using a centrifugal concentrator. Dried lipid extracts were either stored at −80 °C. Before performing the LC-MS/MS analysis, the dried lipid sample was reconstituted in 100 µL of pure methanol. Reconstituted lipids were purified by centrifugal filtration (0.22-µm PTFE membrane filter) at 1000 rpm for 2 min and then transferred into LC-MS insert vials (Fig. 1a).

Fig. 1.

Schematic illustration of LC-MS/MS and high-spatial single-cell MSI analysis. a Similar numbers of all groups (KD, WT, and NC) of cells were cultured for 48 h and collected for LC-MS/MS sample preparation using the Bligh-Dyer method and measured by Q-Exactive. Data were analyzed using LipidSearch software (version 5.1). b Three groups of cells (KD, WT, and NC) were seeded on an ITO-coated chambered slide, and after 48 h of incubation, the slide was prepared for single-cell MALDI-MSI measurement. Matrix deposition was performed using iMLayer, and MALDI-MSI was conducted using the iMScope mass microscope. H&E staining of the measured slide. MALDI-MSI data analysis using IMAGEREVEAL software

Sample preparation for high-spatial single-cell MSI analysis

The ITO-coated glass slides containing each group (KD, WT, and NC) of cells were gently washed in situ. The culture medium was first removed by pipetting, followed by sequential washes using (i) culture medium without FBS and (ii) 0.1 M phosphate-buffered saline (PBS), with each solution applied and removed after a few seconds. Cells were then fixed with 4% PFA for 10 min at room temperature. Following fixation, the samples were washed sequentially with PBS, 100 mM ammonium formate, and ultrapure water to remove residual salts and fixative. After washing, the chamber walls were detached, and the ITO-coated slides were placed in a fume hood to air-dry for 10 min, followed by desiccation in a vacuum chamber for 3 h. Dried slides were stored at −80 °C until MALDI-MSI analysis. Two matrices were used for MALDI-MSI to acquire the data in positive and negative ion modes. DHB was used for positive ion mode, and 9-AA for negative ion mode. Both matrices were deposited by  sublimation method (iMLayerTM, Shimadzu, Japan). DHB was sublimated at 180 °C to achieve a coating thickness of ~ 0.8 µm, while 9-AA was deposited at 220 °C with a target thickness of ~ 1.0 µm. Matrix deposition was performed under vacuum to ensure homogeneous crystallization. Prepared slides were subjected to MALDI-MSI analysis (Fig. 1b).

LC-MS/MS analysis of lipids

The LC-MS/MS analysis was conducted using a Q-Exactive Hybrid Quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific, USA) coupled to an ultra-high-performance liquid chromatography system (Ultimate 3000, Thermo Fisher Scientific). Ions of the lipid species were generated using an electrospray ionization (ESI) source. The LC-MS/MS analysis condition was described in [38] autosampler injected a 5-µL sample, and the lipids were separated in a hydrophobic column (Acclaim 120 C18, 150 mm × 2.1 mm, 3 µm, Thermo Scientific, USA). Throughout the analysis, the LC column and the autosampler chamber were kept at 50 °C and 10 °C, respectively. The mobile phase (A and B) was delivered at a flow rate of 300 µL/min for elution. Mobile phase A was prepared as follows: water-acetonitrile-methanol (2:1:1 v/v/v), 0.1% formic acid, and 5 mM ammonium formate. Mobile phase B consisted of acetonitrile-isopropanol (1:9 v/v), 0.1% formic acid, and 5 mM ammonium formate. The elution conditions were programmed as follows: gradients started at 20% B, increased linearly to 100% B in 50 min, and were then held isocratically at 100% B for 10 min. At 60.1 min, the gradient returned to the initial condition (20% B) and was maintained at 20% B until 70 min, allowing the column to equilibrate. Blank (pure methanol) was run before each sample injection.

The MS conditions were optimized as follows: capillary temperature, 250 °C; probe heater temperature, 350 °C; spray voltage, 3.5 kV (for positive mode) and 2.5 kV (for negative mode); S-lens RF level, 50; auxiliary gas flow, 15 (au); sweep gas flow, 0 (au); and sheath gas flow, 50 (au). Data were acquired in both full MS and data-dependent MS/MS (dd-MS2) modes using Xcalibur v3.0 software (Thermo Scientific). The settings for full MS mode were as follows: automatic gain control (AGC) target, 1 × 10⁵; m/z range, 220–2000; mass resolving power, 70,000 (FWHM, at m/z 200); and maximum sample injection time (IT), 100 ms. The following conditions were used for dd-MS2: mass resolving power of 17,500 (FWHM, at m/z 200); AGC target of 1 × 10⁵; maximum IT of 80 ms; loop count of 5; TopN of 5; isolation window of 2.0 m/z; normalized collision energy (NCE) of 30.0 eV; dynamic exclusion of 15.0 s; and stepped NCE of 15.0% for positive mode and 35.0% for negative mode.

Single-cell MALDI-MSI analysis by AP-MALDI mass microscope

MSI data were acquired using an AP-MALDI mass microscope (a prototype of the iMScope, Shimadzu, Japan), which integrated a high-resolution optical microscope system for high-resolution imaging of cells and tissue samples. In the iMScope mass microscope, for the positive ion mode experiment, the following parameters were used to acquire the data: polarity, positive; laser shot, 100; laser power, 14; laser diameter, 5 µm; mass range (m/z), 600–900; sample voltage, 3.5 kV; detector voltage: 2.1 kV; and pixel size (x × y), 5 µm × 5 µm. For the negative ion mode experiment, the parameters were as follows: polarity, negative; laser shot, 100; laser power, 13; laser diameter, 5 µm; m/z range, 600–900; sample voltage, 3.5 kV; detector voltage, 2.1 (kV); and pixel size (x × y), 5 µm × 5 µm. The Imaging MS Solution (Shimadzu, Japan, version 2.00.00A) software was used to acquire the MSI data.

Hematoxylin and eosin staining

Hematoxylin and eosin (H&E) staining was performed on the cultured cell slides after MSI measurement to facilitate histological examination. The cell slide was placed into the 80% acetone for 10–30 s to remove the matrix. After that, the slide was placed into hematoxylin for 3–5 min and then gently washed using the opposite side of the slide with running tap water for 1 min. Then, the slide was put into 80% ethanol for 1 min. After that, it was immersed in eosin for 30 s, washed using the opposite side of the slide with running tap water for 1 min, and immersed again in the sequential wash with 80%, 90%, 100%, and 100% ethanol for 1 min, respectively. Finally, the slide was placed into xylin for 3 min to clear the debris and then air-dried. After that, a soft mount was dropped onto the tissue sections, and a coverslip was placed on top. Then, the stained slide was imaged using a digital slide scanner (Hamamatsu nanoZoomer S60, Hamamatsu Photonics, Hamamatsu, Japan).

Data analysis

LC-MS lipidomics data of DU145 cells were analyzed using LipidSearch Software version 5.1, with the parameters listed (Table S1). For further processing, the LC-MS data were normalized by internal standard PC (12:0/12:0) to ensure comparability across samples. The normalized data were subsequently used for statistical analysis. Statistical analysis was performed by RStudio (version 4.5.1). Pathway analysis using LION/web: lipid ontology software.

The Imaging MS Solution software was used to acquire AP-MALDI-MSI data by iMScope. After the acquisition, the raw data in the kbd file format were converted to the imdx file format using the imdx converter software. The converted data were then analyzed by IMAGEREVEAL software (Shimadzu, Japan, version 1.20.0.10960). For the annotation of lipid molecules obtained from AP-MALDI-MSI, we used the Xcalibur Qual Browser software (Thermo Fisher, USA). Statistical analysis was performed by RStudio (version 4.5.1).

Results

Untargeted LC-MS lipidomic analysis revealed signature lipid molecules

Untargeted LC-MS profiling of whole cell lipids among the KD, WT, and NC groups identified 819 lipid species across major lipid classes, including phosphatidylcholines (PC), phosphatidylethanolamines (PE), phosphatidylserines (PS), phosphatidylinositols (PI), phosphatidylglycerols (PG), ceramides (Cer), diacylglycerols (DG), and triacylglycerols (TG). Principal component analysis (PCA) of biological triplicates revealed a clear separation of KD from both WT and NC groups, indicating a distinct lipidomic signature, whereas the WT and NC profiles clustered closely (Fig. 2a and b). Differential abundance analysis revealed extensive lipid remodeling in KD cells compared to WT cells; 43 lipids were significantly upregulated, while 111 were downregulated (Fig. 2c). Similarly, comparison with NC cells yielded 49 upregulated and 151 downregulated species (Fig. 2d). Hierarchical clustering analysis of the top 50 significantly differentiated lipid molecules showed pronounced downregulation of PC and TG species and upregulation of ceramides in KD compared to WT and NC. In contrast, WT and NC showed similarity (Fig. 2e).

Fig. 2.

Lipidomic analysis in KD, WT, and NC groups in DU145 cells measured by LC-MS. a, b PCA plot of triplicate LC-MS results of KD, WT, and NC cells, where the KD group was distinguished from the WT and NC groups; on the other hand, the WT and NC groups showed similarity. The volcano plot showed differential lipid abundance in KD versus WT cells. A total of 111 lipid species were downregulated and 43 lipid species were upregulated in KD compared to WT. d Volcano plot comparing KD versus NC cells, where 151 lipid species were downregulated and 49 lipid species were upregulated. Upregulated lipids were shown in red color, and downregulated lipids were shown in blue color. e Heatmap showed the visualization of the top 50 lipid molecules that have a significant difference in KD, WT, and NC groups, where ceramide species were significantly upregulated and TGs and PCs were downregulated in KD compared to WT and NC. Statistical evaluation was conducted using Welch’s t-test, followed by the Benjamini-Hochberg method. T-test significant, P < 0.05, ǀlog2FCǀ > 0.5

Pathway analysis of the signature lipid molecules observed in LC-MS analysis

In the significantly differentiated lipid molecules among the groups, 109 lipid species were commonly altered across both comparisons, where 85 were consistently downregulated and 24 upregulated in the KD cells (Fig. 3a, b, and Tables S2 to S4). Pathway enrichment analysis of downregulated lipids identified glycerophospholipid metabolism and diacylglycerol pathways as the most significantly affected (Fig. 3c), while upregulated species were enriched in bioactive signaling lipids, particularly ceramides (Fig. 3d).

Fig. 3.

Pathway enrichment analysis of significantly differentiated lipid species in the KD group compared to the WT and NC. a The Venn diagram highlights the 85 downregulated lipid species shared across KD vs WT and KD vs NC comparisons. b Venn diagram of 24 lipid species consistently upregulated in the KD group compared to the WT and NC groups. c Pathway enrichment analysis of the downregulated lipids in the KD groups showed higher abundance of glycerophospholipid and diacylglycerol groups. d Upregulated lipid in the KD group showed enrichment in the bioactive signaling lipids pathway

High-spatial MSI analysis of lipids at the single-cell level

We acquired high-spatial MSI (pixel size ~ 5 µm) of lipids of KD, WT, and NC cells at the single-cell level. We obtained ion images of several PC species in positive ion mode and PI, PE, and phosphatidic acid (PA) species in negative ion mode (Table S5). Among them, some abundantly detected lipid species, including PC(16:0_18:1), PC(16:0_16:1), PC(16:0_16:0), PC(18:0_18:1), and PC(42:8) (Fig. 4a, c, d, f, and g), and some lower abundant lipid species (Fig. 4b, e, h, and i) with a representative ion image in positive ion mode. All of the lipid molecules observed in MSI analysis were annotated by LC-MS/MS analysis, confirming their fatty acid moieties (Supplementary File 2), except for some candidate species (without fatty acid moieties) that were tentatively annotated by the Lipid Maps database. H&E image and overlaid ion image with the H&E image of the MSI measured area of KD, WT, and NC groups of cells confirmed the distribution of the lipid molecules at the single-cell level (Fig. S2). We also observed several lipid molecules measured both in positive and negative ion mode (Fig. S3 to Fig. S6).

Fig. 4.

Ion images of the higher and lower abundant lipid molecules in the KD, WT, and NC cells measured by high-spatial MSI analysis in positive ion mode at the single-cell level. a–i Representative ion images of the lipid species PC(16:0_18:1), PC(15:0_16:1), PC(16:0_16:1), PC(16:0_16:0), PC(18:1_18:1), PC(18:0_18:1), PC(42:8), PC(17:0_18:1), and PC(16:0_17:1) at m/z 760.58 [M+H]⁺, 718.54 [M+H]⁺, 732.56 [M+H]⁺, 734.56 [M+H]⁺, 786.60 [M+H]⁺, 788.61 [M+H]⁺, 896.58 [M+H]⁺, 774.60 [M+H]⁺, and 746.57 [M+H]⁺, respectively

mPGES-1 knockdown showed no significant impact on arachidonic acid–containing phospholipid species distribution

Since mPGES-1 is involved in PGE2 synthesis from arachidonic acid (AA), we examined the knockdown effect of the gene on the distribution of AA-containing lipid species. LC-MS/MS analysis identified 28 AA-containing PC species across all samples. Surprisingly, there was no significant difference in their abundance among the KD, WT, and NC groups except for PI(16:0_20:4) and PI(18:0_20:4), which showed modest upregulation in KD versus WT (fold change < 2). However, there was no significant upregulation difference between KD and NC cells (Table S6). Additionally, in MSI analysis, we found a significant downregulation of PC(16:0_20:4) in the KD group compared to the NC groups, and upregulation in KD compared to WT. However, the WT group also showed a significant difference in the downregulation compared to the NC group (Fig. S7a). We found no significant upregulation or downregulation of PI(18:0_20:4) in the KD group compared to the WT and NC groups (Fig. S7b). Overall, AA-containing lipid species showed no significant changes following mPGES-1 knockdown.

High-spatial MSI analysis showed downregulation of specific PC species in KD cells

To examine the knockdown effect of the mPGES-1 gene on the distribution of the lipid molecules, we performed statistical analysis of the spatial distribution of the lipid molecules by means of intensity at the single-cell level using a similar number of single cells in each group (cell size of the different groups may vary according to their properties). From the microscopic image, ion image, and H&E image, the KD and WT cells were almost of the same size, but the NC cells showed a slightly larger size than the KD and WT cells.

Interestingly, PC(16:0_16:0) and PC (18:0_18:1) distribution were significantly (p < 0.001) downregulated in the KD compared to WT and NC at the single-cell level. Notably, there was no significant difference between WT and NC (Fig. 5). These findings were also similar to the LC-MS/MS results (Table S4).

Fig. 5.

Statistical analysis of the differential spatial distribution of lipid molecules in the KD groups compared to the WT and NC groups. a(i) H&E image and ion image, and a(ii) overlaid intensity distribution with significance test of the PC(16:0_16:0) of KD, WT, and NC cells. b(i) H&E image and ion image, and b(ii) overlaid intensity distribution with significance test of the PC(18:0_18:1) of KD, WT, and NC groups. Kruskal-Wallis’s significance test. P. adj ≤ 0.0001 ~ ****, P. adj ≤ 0.001 ~ ***, P. adj ≤ 0.01 ~ **, P. adj ≤ 0.05 ~ *

Reduced cellular heterogeneity of PC content due to mPGES-1 knockdown

Later, we checked the individual intensity of PC(16:0_16:0) and PC (18:0_18:1) of each cell across all groups and found interesting results. The average intensity distribution of the cells was divided into four groups based on the actual intensity range, each represented with different colors (very low—yellow, low—light yellow, medium—light pink, and high—dark pink). Each cell (approximately 60 cells of each group) was visualized with a color that indicated the intensity variation. The spread and diversity of colors within KD, WT, and NC groups directly illustrate the underlying heterogeneity in lipid content among individual cells for PC(16:0_16:0) and PC(18:0_18:1). Interestingly, we found that the KD group showed strong clustering at the low-intensity range; almost all the cells displayed similarly low PC levels, indicating reduced heterogeneity after knockdown, whereas the WT and NC groups showed broader distribution ranges and higher counts across various levels of intensity diversity, indicating increased cellular heterogeneity in PC content. We also observed that KD was the most consistent group, having the lowest coefficient of variation percentage (CV%) among all groups (Fig. 6).

Fig. 6.

Cellular heterogeneity effects of PC(16:0_16:0) and PC(18:0_18:1) in KD, WT, and NC cells. a(i) and b(i) Yellow color cells had very low intensity, light yellow cells had low, light pink cells had medium, and dark pink cells had high intensity across all groups. H&E images confirming the single cell. a(ii) and b(ii) Cell counts, depending on the absolute intensity of each cell, were categorized into very low (intensity≦0.5 × 10⁵), low (intensity≦1 × 10⁵), medium (intensity≦1.5 × 10⁵), and high (intensity≦2 × 10⁵) intensity groups, shown in the bar chart and the coefficient of variation (CV%) of each group. KD groups showed strong clustering in the very low intensity range and CV% with reduced heterogeneity, while WT and NC groups displayed broader distribution ranges across various intensity levels with higher CV%, indicating increased cellular heterogeneity in PC content

Discussion

This study employed LC-MS/MS and single-cell MALDI-MSI to characterize lipidomic alterations induced by mPGES-1 knockdown in hormone-resistant DU145 prostate cancer cells. In LC-MS/MS analysis, 819 lipid species were identified and revealed extensive metabolic reprogramming, with clear segregation of KD cells from both WT and NC controls in PCA analysis. The identification of 109 commonly altered lipid species, where 85 were downregulated and 24 were upregulated (Fig. 2), demonstrates that mPGES-1 plays a broader role in cellular lipid homeostasis beyond its canonical function in prostaglandin synthesis.

Pathway enrichment analysis revealed that glycerophospholipid metabolism and diacylglycerol pathways were most significantly affected among downregulated lipids, while ceramide pathways were enriched among upregulated species. The concurrent upregulation of ceramides and downregulation of TG and PC species is particularly noteworthy (Tables S2, S3, and S4). Ceramides are well-established mediators of programmed cell death, and our previous study performed a functional analysis to assess apoptosis using adriamycin, where KD cells showed sensitivity but the WT cells were resistant [21, 39]. Therefore, the present findings of the upregulation of ceramide species in the KD cells indicate the programmed cell death that causes suppression of cancer cell growth, which is consistent with the previous report. In addition, elevated serum TG levels associate with increased prostate cancer development risk [40, 41]. These reciprocal changes suggest a shift from pro-survival to pro-apoptotic lipid signaling due to mPGES-1 knockdown. However, this study was conducted in a single-cell line (DU145), and validation in additional hormone-resistant prostate cancer models would strengthen these observations.

Advanced MSI-based single-cell molecular analysis provides crucial spatial information on molecular diversity and cellular heterogeneity in bioscience research. Our study revealed that the application of an AP-MALDI mass microscope has the advantage of an embedded microscope, which can be a valuable tool for the single-cell MSI analysis of biomolecules. This integrated optical system enables precise targeting and real-time visualization during data acquisition, significantly improving the accuracy of single-cell analysis. Successfully, we achieved high spatial resolution (pitch size 5 µm) and laser ablation within a cell, which was confirmed after the H&E image (Fig. S8). This improved spatial precision is essential to capture the metabolic heterogeneity of tumor cell populations and to localize lipidomic changes associated with mPGES-1 knockdown at the cellular level.

Among the PC species, PC(16:0_16:0) and PC(18:0_18:1) emerged as the most consistently affected species, validated by both LC-MS/MS and high-spatial single-cell MSI analysis (Fig. 5 and Table S4). Previous studies demonstrated that elevated PC(16:0_16:0) levels correlate with worse outcomes in metastatic hormone-resistant prostate cancer [42], while PC(18:0_18:1) is highly associated with drug-resistant CRPC [43]. The observed downregulation aligns with previous reports showing that mPGES-1 inhibition suppresses tumor progression in xenograft models [21].

Single-cell MSI analysis revealed a striking loss of cellular heterogeneity in PC content following mPGES-1 knockdown (Fig. 6). The coefficients of variation for both PC(16:0_16:0) and PC(18:0_18:1) were markedly reduced in KD cells compared to controls, indicating uniform suppression across the entire cell population. This homogeneous metabolic reprogramming contrasts with the substantial heterogeneity observed in WT and NC cells, where individual cells displayed broad intensity distributions spanning four distinct categories. Single-cell mass spectrometry provides molecular information on a cellular level and enables the analysis of cellular heterogeneity [44], which is critical for understanding tumor cell populations’ metabolic diversity. Metabolic heterogeneity within tumor populations is recognized as a key factor in therapeutic resistance and disease progression [45]. The uniform metabolic state induced by mPGES-1 knockdown may reduce the adaptive capacity of cancer cells, potentially enhancing their vulnerability to therapeutic interventions. This finding suggests that mPGES-1 inhibition could be an effective strategy to overcome the metabolic plasticity that often leads to treatment resistance in CRPC.

Notably, while LC-MS/MS provided a comprehensive overview of global lipid remodeling due to mPGES-1 knockdown, single-cell MALDI-MSI uniquely uncovered the loss of cellular heterogeneity in PC species, highlighting the power of spatial single-cell lipidomics in cancer research. This dual-platform strategy illustrates how MSI complements conventional lipidomics and positions single-cell MSI as a powerful tool to predict tumor metabolic reprogramming. In the LC-MS analysis, several lipid classes were obtained, but in the MSI analysis, very few lipid classes were obtained (Table S5). However, in this study, the amine-containing lipids (ceramides, SMs, and PE) detected by LC-MS/MS were not observed by MSI, likely due to PFA fixation effects, which create cross-linking between the amine-containing lipid molecules, resulting in their inability to ionize [26, 46]. In addition, the instrumental sensitivity and matrix selection to ionize the lipid molecules might also be considered. Future studies might benefit from alternative fixation methods or matrix-free imaging approaches to capture these species spatially.

Interestingly, among the 28 AA-containing phospholipids detected in LC-MS/MS, none showed significant changes in KD compared to WT and NC (Table S6). Additionally, MSI analysis showed that PC(16:0_20:4) was significantly altered (upregulated in KD compared to WT and downregulated in KD compared to NC downregulated in WT compared to NC), across KD, WT, and NC groups. This alteration might be the effect of cellular heterogeneity, not the impact of the mPGES-1 gene knockdown (Fig. S7). This finding is consistent with a previous study of metabolic shunting, where blocking PGE2 synthesis redirects arachidonic acid toward alternative prostanoid and leukotriene pathways. Therefore, after blocking mPGES-1 in the conversion of PGH2 to PGE2, PGD2 was produced from PGH2, but it might differ in various cell types [47]. Thus, these findings indicate that the mPGES-1 does not affect the AA-containing lipid molecules.

Now it is crucial to explain how mPGES-1 causes the alteration of the lipid molecules that we observed in our LC-MS/MS and spatial analysis. Based on our findings and previous reports [48–51], we proposed a model (Fig. S9) of how PGE2 exerts its role in PC synthesis and its consequences in prostate cancer. PGE2 transmits its signal through G-protein-coupled cell surface receptors (EP1 to EP4). After PGE2 binding to the EP receptors, various intracellular oncogenic pathways are stimulated, including the PI3K/Akt pathway, which enhances the activity of choline kinase (CK), resulting in the upregulation of PC biosynthesis. It has been reported that the uptake of choline and the enhancement of CK enzyme activity for biosynthesizing PCs are increased in various types of cancers, including prostate cancer, creating a conducive tissue microenvironment for tumor progression and metastasis.

Notably, this study used only the DU145 cell line; more studies should be required using other prostate cancer cell lines like PC3, an androgen-independent cell line, and LNCaP, an androgen-sensitive cell line, and also in a tumor model sample to consolidate these findings.

Further study is required at the molecular level to elucidate the exact pathways by which the mPGES-1 gene affects the abundance of the specific lipid molecules. The direct causal relationships between mPGES-1 activity, PGE2 levels, and selective PC alterations require further mechanistic investigation, potentially through EP receptor subtype–specific analysis. Additional studies are needed on downstream functional outcomes such as cell proliferation assays (live cell imaging, cell cycle analysis, and DNA synthesis assay), apoptosis assays to detect and quantify programmed cell death (western blotting, mitochondrial membrane potential assay), and, finally, the therapeutic response needed to achieve desired outcomes.

Conclusion

In this study, we employed LC-MS/MS and spatial single-cell MALDI-MSI modalities on the DU145 hormone-resistant prostate cancer cell line to explore the role of the mPGES-1 gene in the lipidomic alteration. We found ceramide species upregulation and TGs and PCs downregulation due to the knockdown effect of the gene in LC-MS analysis. Interestingly, we also observed similar results for PC(16:0_16:0) and PC(18:0_18:1) species in spatial single-cell MSI analysis that showed the uniform suppression of the molecules across all KD cells, with the loss of metabolic heterogeneity induced by the knockdown effect of the gene. These results indicate metabolic reprogramming by reducing the tumor microenvironment and augmenting programmed cell death of the cancer cells. Therefore, this study emphasized that the inhibition of mPGES-1 can be a therapeutic target for suppressing the progression of CRPC.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contribution

Mst. Sayela Afroz: methodology, investigation, formal analysis, and manuscript writing: original draft. Shuhei Aramaki: supervision, writing: review and editing. Md. Muedur Rahman: investigation, formal analysis, and manuscript revision. Maxime Dubail: formal analysis and manuscript revision. Zhang Chi: manuscript revision. Kyoka Fujii: manuscript revision. Mariko Kurosawa: manuscript revision. Keita Tamura: manuscript revision. Tomoaki Kahyo: manuscript revision. Katsumasa Nakamura: manuscript revision. Teruo Inamoto: manuscript revision. Olof Rådmark: manuscript revision. Mitsutoshi Setou: conceptualization, resource, funding, and manuscript revision. Hiromi Hanaka: methodology, manuscript revision, and supervision.

Funding

This work was supported by the MEXT Project for promoting public utilization of advanced research infrastructure (Imaging Platform) under Grant Number JPMXS0450200225. This work was also supported by AMED under Grant Number 21ak0101179, and by JSPS KAKENHI under Grant Numbers JP22H02793, JP23K18197, and JSPS J-PEAKS.

Data availability

All the data are available in the manuscript and supplementary data files.

Declarations

Conflict of interest

The authors declare no competing interests.