Roles of FXR in Lipid Biochemistry within Hepatocellular Carcinoma Using MALDI MSI

Abstract

Hepatocellular carcinoma (HCC) is a common type of cancer with a high mortality rate, which is affected by the farnesoid X receptor (FXR) in the liver. FXR is a ligand-activated nuclear receptor involved in lipid metabolism and homeostasis. Matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI MSI) enables spatially resolved molecular analysis and is particularly suited to profiling lipid distributions within tissues. In this study, MALDI MSI was employed to investigate alterations in liver lipid composition associated with HCC and FXR expression with a focus on phosphatidylcholine (PC) species. Distinct spatial and compositional differences in PC lipids were observed between HCC and nontumorous regions. Furthermore, we hypothesize a role of PC(38:4) in HCC based on cell culture experiments in which Hep3B cells were incubated in the presence of PC(38:4).

Introduction

Hepatocellular carcinoma (HCC) has the third-highest mortality rate, and is the sixth most predominant type of cancer, with prevalence continuing to rise. The risks of HCC include steatosis caused by obesity or drinking alcohol. , In human and mouse models of HCC, the bile acid and lipid profiles are altered, which is likely related to the reduced expression of the farnesoid X receptor (FXR) gene. ,, FXR is a ligand-activated nuclear receptor predominantly expressed in the liver and intestine and is expressed in hepatocytes and enterocytes. FXR has numerous implications when genetically knocked out, such as a smaller body size, the accumulation of bile acids, inflammation, and the spontaneous development of HCC. , Whole body, knock out (KO) mice spontaneously developed steatosis, metabolic dysfunction-associated steatohepatitis, and progressed into HCC. − Moreover, Western diet intake facilitates the progression of these metabolic diseases in a male-dominant manner, like humans. −

FXR has more than 80 potential ligands that can be agonists or antagonists. Those include endogenous bile acids, as well as polyunsaturated fatty acids (PUFAs), like docosahexaenoic acid (DA, 22:6), arachidonic acid (AA, 20:4), and linolenic acid (LA, 18:3). Many of these are involved in lipid biosynthesis. Previous studies have demonstrated a connection between FXR and HCC, ,,,, as well as cholestasis and liver fibrosis. , Notably, FXR deactivation leads to BA accumulation and liver carcinogenesis. ,

Lipids are a diverse group of amphipathic or hydrophobic metabolites that are involved in many biological processes and compose the majority of the cell membrane. , In mammals, lipids are produced by the liver that act through multiple receptors, including FXR. The most abundant phospholipid class is phosphatidylcholine (PC), which also makes up 45–55% of the cell membrane in mammals and has been linked to the progression of liver disease. , Phospholipids are of interest in cancer research because they are related to nutrition, signaling, and membrane fluidity. The last of which is critical for metastasis, including epithelial-mesenchymal transition (EMT). , We therefore studied the link between FXR and HCC through spatial analysis of phospholipids to uncover FXR-affected phospholipid metabolism linked with HCC.

PC lipids are made up of a glycerol backbone, two fatty acid acyl chains at the sn-1 and sn-2 positions, and a phosphate moiety with a choline headgroup at the sn-3 position. PC lipids are primarily synthesized by either the cytidine diphosphate-choline (CDP-choline) pathway using dietary choline or via the phosphatidylethanolamine N-methyltransferase pathway in the liver using phosphatidylethanolamine (PE) lipids. The mammalian cell membrane has diverse functions and diverse lipid species, leading to lipid profiles that differ between cell types and most cancers results in changed lipid metabolism. − For example, phosphatidylethanolamine N-methyltransferase activity is affected in HCC, resulting in a decreased PC:PE ratio. , Additionally, an increase in monounsaturated fatty acid-containing PC lipids is associated with hepatocyte proliferation and HCC carcinogenesis. Moreover, sphingomyelin (SM) lipids are involved in HCC, , and are noteworthy in tumor-related studies because mammalian cells lacking these lipids require their exogenous addition for proliferation. ,

Mass spectrometry is one of the only methods capable of surveying discrete lipids, and matrix-assisted laser desorption/ionization (MALDI) mass spectrometry imaging (MSI) obtains this information spatially. MALDI MSI is a label-free and untargeted analytical technique that can measure hundreds to thousands of molecules in their native spatial context. In short, a tissue section of approximately 10–20 μm is thaw-mounted onto a conductive glass slide, which is subsequently coated with an organic small molecule matrix. This matrix absorbs light from the UV laser to aid in the desorption and ionization of endogenous biomolecules within the tissue. Mass spectra are then acquired from designated locations on the tissue section as the UV laser is rastered across the tissue in a grid-like pattern, which is used to construct a heatmap of the relative abundance for each detected ion. MALDI MSI is commonly used to analyze tissue regions of mm² size at cellular resolutions for many types of samples, including tumors. − By maintaining the spatial integrity of the sample using a tissue section and MALDI MSI, the tumor microenvironment can be observed. Spatial metabolomic analysis of HCC is particularly of interest because spatial analysis bypasses the difficulty of physically separating tumor from nontumor tissue for separate bulk analysis. Here, we highlight the changes in phospholipid metabolism by investigating phospholipid distributions within HCC as a result of knocking out FXR. Using MALDI MSI, we determined that PC(38:4) was likely involved in tumor pathogenesis, which was supported by cell culture experiments.

Materials and Methods

Chemicals

All chemicals were used without further purification unless otherwise specified.

Murine Model

Male 8-week-old C57B6/L wildtype and FXR KO mice were hydrodynamically injected with plasmids pT3-EF1α-HA-myr-AKT, pT/pT/Caggs-Nras-v12, and pCMV-SB11, as previously described. − Briefly, oncogene plasmids and transposons were diluted to 2 mL (10% body weight) in saline and injected into the mouse tail vein within 5∼7 s. Mice were euthanized one month after the injection to harvest HCC. Sex and age-matched mice without HCC were used for comparison. The presence of HCC in injected mice was confirmed by H&E staining.

MALDI Mass Spectrometry Imaging

Fresh-frozen liver samples were sectioned to 10 μm thickness and thaw-mounted onto an indium–tin-oxide-coated glass slide (Delta Technologies, Ltd., Loveland, CO). Tissues were not washed prior to matrix application to maintain salt gradients and reduce molecular diffusion. α-Cyano-4-hydroxycinnamic acid matrix (Sigma-Aldrich, Co., St. Louis, MO, USA, 5 mg/mL in 70% methanol) was applied using an HTX M3+ sprayer (HTX Technologies, LLC, Chapel Hill, NC). All chemicals were used without further purification. Spray parameters include an 80 °C nozzle, gas pressure of 15 nitrogen psi, 120 μL/min solvent flow rate, and a 2 mm track spacing for 8 passes. MALDI MSI experiments were performed by using a timsTOF fleX dual source MALDI mass spectrometer (Bruker Scientific, Billerica, MA). Each experiment was performed in positive ion mode with a 50 μm × 50 μm raster width. The raster width is defined here as the scanned area per laser ablation. A detailed description of the method is available in Supporting Information. In short, the global attenuator value was set to 5%, the local laser power was set to 88%, 150 shots were fired in a single burst, and the mass range acquired was a mass-to-charge ratio (m/z) 350–1200. The MALDI MSI data sets are available as .imzml and .ibd files at https://metaspace2020.org/project/bender-2026. MALDI MS/MS of featured lipid peaks was performed on subsequent tissue sections by CID with nitrogen gas (collision energy of 40 eV), all on the same timsTOF fleX (Spectra available in Supporting Information).

Cell Culture

The Hep3B cell line was seeded in 24-well plates at 5 × 10⁴ cells per well. Cells were seeded in Dulbecco’s Modified Eagle Medium (Gibco, Grand Island, NY) containing 4.5 g/L d-Glucose, l-Glutamine, and 100 mg/L sodium pyruvate. Fetal bovine serum (Gibco, Grand Island, NY), nonessential amino acids (Gibco, Grand Island, NY), and antibiotic-antimycotic (Gibco, Grand Island, NY) were added to DMEM for seeding cells. After 24 h, the medium was replaced with serum-free DMEM containing antibiotic-antimycotic. Once cells reached 50% confluency, the medium was replaced with either lipid-dosed medium containing 0.5% DMSO or 0.5% ethanol or an equivalent medium without lipids for controls. Cells were incubated with lipid-dosed medium for 48 h and received fresh lipid-dosed medium after 24 h. Each cell growth condition was done in three replicates for the MTT assay (Sigma-Aldrich, St. Louis, MO) and three replicates for electrospray ionization. Cells were all incubated at 37 °C with 5% CO₂.

Electrospray Ionization

For ESI experiments, 1 × 10⁵ cells were collected from each well of 24-well plates for Huh-7 and Hep3B cell lines in three replicates. Cells were homogenized via the Bligh and Dyer method to extract lipids. Extracted lipids were dried under a stream of nitrogen gas and then diluted in HPLC-grade methanol (Sigma-Aldrich, St. Louis, MO) according to the lipid extraction dried mass for electrospray ionization (final concentrations were matched to 10 mg/mL before being diluted to 50 μg/mL). No external standard was included; therefore, we are not absolutely quantitative. Each replicate was directly injected into a timsTOF fleX dual source MALDI mass spectrometer (Bruker Scientific, Billerica, MA). The average mass spectrum for each replicate was base peak normalized, then relative intensities of the ion of interest was compared by a student’s t test. ,

Data Analysis

An internal, quadratic mass calibration was completed within DataAnalysis (Bruker Scientific) using common lipid features ([PC(26:0)+H]⁺, [PC(32:0)+H]⁺, [PC(32:0)+Na]⁺, [PE(36:1)+H]⁺, [PC(34:2)+H]⁺, and [PC(36:1)+Na]⁺). LIPID MAPS database was used for putative lipid identification (<5 ppm), excluding lipids containing odd-chained fatty acids (SI Table S1). Only H⁺, Na⁺, and K⁺ adducts were considered during the LIPID MAPS database search. MS/MS of featured lipid peaks were annotated according to literature values associated with lipid head groups and adducts. − Area under the receiver operating characteristic curve (ROC AUC), box plots, and MALDI MSI ion images were performed using SCiLS Lab version 2024b Pro (Bruker Scientific) with total ion count normalization and a 15 ppm ion window. All data were analyzed in the same SCiLS iteration. Segmentation of tumors from nontumor tissue was performed with weak denoising, peak area interval processing mode, correlation distance metric, and total ion count normalization, using all spectra and the bisecting k-means algorithm via the segmentation tool included in the SCiLS Lab software. Depiction of subsequent segmented regions surrounding the tumor for spatial analysis can be found in Supporting Information (SI Figure S1). PCA was performed using MetaboAnalyst 6.0 online analysis software using ion windows (SI Figure S2).

Results and Discussion

Hepatocellular Carcinoma Affects Hepatic Lipids in WT and FXR KO Mice

MALDI MSI was performed on WT (n = 4), WT/HCC (n = 4), FXR KO (n = 3), and FXR KO/HCC mouse livers (n = 3). 87 phospholipid peaks were detected, and 74 phospholipids were putatively identified. The putatively identified phospholipids include 36 phosphatidylcholine (PC), 18 phosphatidic acid (PA), 9 phosphatidylethanolamine (PE), and 11 sphingomyelins (SM). However, it is important to note the possibility of PC, PE, or PS phospholipid in-source decay resulting in PA-like artifacts. The mass spectral features putatively identified to be PA lipids could be endogenous PA phospholipids or PA-like artifacts (Figure B–F, light gray).

1.

Liver phospholipid profiles change in the presence of HCC. Lipids of the phosphatidylcholine (PC), sphingomyelin (SM), phosphatidylethanolamine (PE), and phosphatidic acid (PA) lipid classes were detected. Phospholipids were detected with hydrogen, sodium, or potassium adducts because these elements are involved in ionizing the lipids for detection by mass spectrometry. Semicircle radius is linearly scaled for each intensity category (A–F) to represent the relative abundance of a phospholipid. Each intensity category (A–F) is scaled separately for visual purposes due to highly variable intensities across lipid species. The maximum intensity in each intensity category is represented by 0.5 in. circles, as shown in the figure key. Very high intensity lipids (A) have a maximum intensity of 5200, high intensity lipids (B) have a maximum intensity of 2500, medium-high intensity lipids (C) have a maximum intensity of 1100, medium-low intensity lipids (D) have a maximum intensity of 500, low intensity lipids (E) have a maximum intensity of 300, and very low intensity lipids (F) have a maximum intensity of 150. The color of the semicircle represents the comparison, with darker colors to represent higher ROC scores (see color scale in figure key, top left). Higher ROC scores indicate a stronger differentiation between the two groups according to that phospholipid. ROC scores comparing sample groups in sets of two: WT versus WT/HCC (pink), FXR KO versus FXR KO/HCC (green). Of the 74 putatively identified phospholipids, 47 were discriminating (ROC score >0.800) in at least one of the above comparisons.

ROC scores are employed in this context to quantitatively assess the ability of a phospholipid species to distinguish between two groups. ROC scores range from 0.5 to 1.0, where 0.5 represents no discrimination, and 1.0 represents a perfect discrimination between two groups. The goal of these comparisons is to understand how liver lipids change due to HCC and how FXR expression affects lipid profiles in the context of HCC. The development of HCC causes a metabolic change that affects the phospholipid composition (Figure ). Of the lipids that differentiated WT livers versus WT/HCC (ROC score of 0.800 or higher), most lipids were more abundant in the presence of HCC (Figure , pink). In healthy livers, 67% of the detected PC lipids had an ROC score of 0.800 or higher and 44% had an ROC score of 0.900 or higher when compared to WT/HCC (Figure , pink). Interestingly, all of the PC lipids that had an ROC score of 0.900 or higher in the WT liver comparison were more abundant in WT/HCC than WT. The highly discriminatory ROC score for these PC lipids suggests an upregulation of these lipid species due to HCC. In both the WT versus WT/HCC and FXR KO versus FXR KO/HCC comparisons, there was a distinguishable increase in monounsaturated lipids [PC(34:1)+K]⁺, [PC(34:1)+H]⁺, [PC(36:1)+K]⁺, [PC(32:1)+K]⁺, [PC(36:1)+H]⁺ in HCC (Figure A–E, pink, green). In general, our results appear consistent with prior studies. For instance, Lu et al. found significant alterations of m/z 758.57 [PC(34:2)+H]⁺, m/z 782.57, m/z 806.57, m/z 808.58, m/z 820.53 [PC(36:4)+K]⁺, m/z 832.58 [PC(38:4)+Na]⁺, and m/z 856.59 some of these lipids in human serum, human tissue, and mouse tissue when liver cancer was present. ,

Although many lipid species were more abundant in HCC, some lipids were less abundant or showed insignificant changes, such as SM lipids (Figure , pink). In WT mice, there were no SM lipids with an ROC score above 0.800, signifying a lack of change in SM lipids due to HCC (Figure , pink). Contrary to the WT and WT/HCC comparisons (Figure , pink), all ROC discriminating SM lipids were more abundant in FXR KO/HCC than FXR KO (Figure , green). [SM­(34:1;O2)+K]⁺ and [SM­(34:1;O2)+Na]⁺ both distinguish HCC from non-HCC in FXR KO (Figure C,E, green), but not in WT (Figure C,E, pink). Alternatively, [SM­(42:2;O2)+K]⁺ is more abundant in HCC in both WT/HCC and FXR KO/HCC (Figure D). An elevated sphingomyelin content has been linked to tumor promotion, , so we hypothesize SM­(34:1;O2) to be involved in tumor promotion in the absence of FXR. [PA(36:3)+Na]⁺ (Figure C, pink), [PA(38:3)+Na]⁺ (Figure C, pink), and [PA(34:1)+Na]⁺ (Figure E, pink), were less abundant in WT/HCC than WT. We hypothesize the downregulation of [PA(36:3)+Na]⁺, [PA(38:3)+Na]⁺, and [PA(34:1)+Na]⁺ to be a downstream effect of the metabolic changes occurring with the progression of HCC. For example, PA lipids are converted to diacylglycerol, which is subsequently combined with CDP-choline to form PC lipids via the Kennedy pathway. The Kennedy pathway is expected to be upregulated in HCC. Therefore, [PA(36:3)+Na]⁺, [PA(38:3)+Na]⁺, and [PA(34:1)+Na]⁺ may be less abundant in WT/HCC due to conversion to PC(36:3), PC(38:3), and PC(34:1), respectively. Alternatively, the phospholipase D activity to convert PC(36:3), PC(38:3), and PC(34:1) to PA(36:3), PA(38:3), and PA(34:1) may be downregulated for PC(36:3), PC(38:3), and PC(34:1) in HCC.

The effect of HCC on liver phospholipids appears to be different in FXR KO. For example, [PA(36:3)+Na]⁺ and [PA(38:3)+Na]⁺ were HCC specific, as they were less abundant in WT/HCC and FXR KO/HCC compared with their respective controls without HCC (Figure C). However, [PA(34:1)+Na]⁺ was discriminatory between HCC and non-HCC WT livers (Figure E, pink), but not in the FXR KO livers (Figure E, green). Unlike the WT comparison, many lipids were less abundant in FXR KO liver than in FXR KO/HCC (Figure , green). Interestingly, the discriminating lipids (ROC score 0.800 or higher) that were less abundant in FXR KO/HCC liver were all putatively identified as polyunsaturated fatty acid-containing lipids. Unsaturated membrane lipids result in looser lipid packing compared to saturated lipids, which affects membrane fluidity. Therefore, the lower abundance of poly unsaturated fatty acid-containing lipids in FXR KO/HCC liver may contribute to the liver rigidity observed in HCC. Overall, there were many lipids that discriminated HCC liver from non-HCC liver. However, many of the discriminating lipids did not discriminate HCC from non-HCC in the same direction in the absence of FXR. This finding highlights the influence of FXR on HCC.

FXR KO Affects Lipids Profiles in HCC

Phospholipids change in the liver due to the presence of FXR (Figure ). Lipids that distinguish WT from FXR KO liver could indicate lipid species that FXR regulates for lipid homeostasis. In the absence of FXR, there were no discriminating changes (ROC score >0.800) in phospholipid metabolism in SM and PE lipid classes (Figure , brown). However, there are changes seen in PA and PC lipid classes when FXR is knocked out. Most lipids that distinguished FXR KO liver from WT liver (ROC score >0.800) were more abundant in FXR KO liver (Figure , brown). However, [PC(34:1)+H]⁺ (Figure B, brown) and [PC(32:0)+H]⁺ (Figure D, brown) were less abundant in the FXR KO liver. The lipids that most strongly distinguished FXR KO livers from WT livers (ROC score >0.900) were [PC(34:2)+K]⁺, [PC(36:4)+K]⁺, [PC(36:2)+K]⁺, [PC(38:4)+K]⁺, [PA(36:3)+K]⁺, [PA(38:3)+K]⁺, and [PA(40:5)+K]⁺ (Figure A–C, brown). Interestingly, these strongly distinguishing lipids were all more abundant in FXR KO livers than in WT livers. This inspires questions regarding the potassium adduct lipids mentioned above distinguishing FXR KO and WT livers more strongly than the nonpotassium adduct lipids. Since FXR is a bile acid receptor and previous studies show that bile acids activate BK­(Ca) potassium channels, , these data may reflect a trend in potassium regulation. While the connection between FXR and potassium in the liver is important to investigate, a salt-focused analysis was not pursued within the scope of this experiment.

2.

Phospholipid profiles of the liver are affected by FXR. Lipids of phosphatidylcholine (PC), sphingomyelin (SM), phosphatidylethanolamine (PE), and phosphatidic acid (PA) lipid classes were detected. Phospholipids were detected with hydrogen, sodium, or potassium adducts because these elements are involved in ionizing the lipids for detection by mass spectrometry. Semicircle radius is linearly scaled for each intensity category (A–F) to represent the relative abundance of a phospholipid. Each intensity category (A–F) is scaled separately for visual purposes due to highly variable intensities across lipid species. The maximum intensity in each intensity category is represented by 0.5 in. circles, as shown in the figure key. Very high intensity lipids (A) have a maximum intensity of 5200, high intensity lipids (B) have a maximum intensity of 2500, medium-high intensity lipids (C) have a maximum intensity of 1100, medium-low intensity lipids (D) have a maximum intensity of 500, and low intensity lipids (E) have a maximum intensity of 300. The color of the semicircle represents the comparison, with darker colors to represent higher ROC scores (see color scale in figure key, top left). Higher ROC scores indicate a stronger differentiation between the two groups according to that phospholipid. ROC scores comparing sample groups in sets of two: WT versus FXR KO (brown), WT/HCC versus FXR KO/HCC (purple). Of the 74 putatively identified phospholipids, 35 were discriminating (ROC score >0.800) in at least one of the above comparisons.

Although [PC(34:2)+K]⁺, [PC(36:4)+K]⁺, [PC(36:2)+K]⁺, [PC(38:4)+K]⁺, [PA(36:3)+K]⁺, [PA(38:3)+K]⁺, and [PA(40:5)+K]⁺ distinguished FXR KO livers from WT (Figure , brown), these lipids were not discriminating for FXR KO/HCC from WT/HCC livers (Figure , purple). This highlights a difference in the metabolic role of FXR in HCC. The PC, PE, and PA lipids that discriminated WT/HCC from FXR KO/HCC (ROC score >0.800) were all less abundant in FXR KO/HCC (Figure , purple). Conversely, all discriminating SM lipids were more abundant in FXR KO/HCC (Figure , purple). SM lipids are highly saturated and have many biological roles. It has been established that SM lipids colocalize with cholesterol in the cellular membrane, and affect cholesterol homeostasis. , Additionally, tumors are known to require cholesterol to maintain proliferation, and promote tumorigenesis in HCC. Since an increase in SM lipid is connected to tumor promotion, , the upregulation of [SM­(34:1;O2)+K]⁺, [SM­(42:2;O2)+K]⁺ and [SM­(34:1;O2)+Na]⁺ in FXR KO/HCC potentially attests to the role of FXR in HCC (Figure C–E, purple).

Lipid Abundances Differ between HCC and the Tissue

Principal component analysis (PCA) showed clustering of WT (dark blue), FXR KO (red), WT/HCC (light blue), or FXR KO/HCC (green) mouse liver tissue (SI entry Figure S2). Principal component 1 (PC1) represented 32.1% of the variance in the data and separated WT liver samples with and without HCC. Conversely, WT/HCC samples primarily clustered in the negative direction of PC1. The lipids associated with the separation across PC1 were [PA(36:3)+Na]⁺, [PC(34:2)+H]⁺, [PA(38:3)+Na]⁺ in the positive direction and [PC(34:1)+K]⁺, [PC(34:2)+K]⁺, and [PC(36:4)+K]⁺ in the negative direction. There was a significantly lower presence of [PA(36:3)+Na]⁺ (p = 0.038) and [PA(38:3)+Na]⁺ (p = 0.048) in WT/HCC compared with WT liver. Conversely, there was a higher abundance of [PC(34:2)+H]⁺ (p = 0.080) in the WT/HCC liver. Additionally, PC2 explained 23.7% of the variance and primarily separated WT/HCC from WT, FXR KO, and FXR KO/HCC. PC2 was most influenced in the positive direction by [PC(34:2)+H]⁺, [PC(34:1)+H]⁺, and [PC(34:1)+Na]⁺. Finally, PC3 represented 19.4% of the variance and primarily separated FXR KO from the FXR KO/HCC liver. PC3 was influenced in the positive direction by [PC(34:1)+K]⁺ (p = 0.013), [PC(34:1)+H]⁺ (p = 0.018), and [PA(36:2)+K]⁺ (p = 0.035), all of which were significantly more abundant in FXR KO/HCC liver. In the negative direction, [PC(34:2)+K]⁺, [PC(34:2)+Na]⁺, and [PA(36:3)+K]⁺ caused the most separation in PC3. FXR KO/HCC liver had a lower abundance of [PC(34:2)+K]⁺ (p = 0.011), [PC(34:2)+Na]⁺ (p = 0.081), and [PA(36:3)+K]⁺ (p = 0.025) compared to FXR KO livers. According to our spatial analysis, lipids that were more abundant in tumors than surrounding tissue (Figure A-D.1, and A-D.2) influence PC1 and PC3 in the same direction (SI Figure S2.D). PCA suggested that liver phospholipids were different in the presence of HCC. Additionally, the clustering suggests the phospholipids were affected by HCC differently in the absence of FXR. The livers with HCC clustered closely together, as did the non-HCC livers.

3.

Phospholipids are expressed differentially between tumor vs tumor environment. MALDI MSI was used to detect 74 lipid features in positive ion mode for both WT/HCC (A-I.1) and FXR KO/HCC tissues (A-I.2). Of these lipids, [PC(32:1)+K]⁺, [PA(36:2)+K]⁺, m/z 774.5975 (unknown lipid), and [PC(34:1)+K]⁺ localize within the tumor regions for both WT/HCC and FXR KO/HCC tissues (A.1-D.1 and A.2-D.2). Alternatively, [PC(38:6)+K]⁺, [PC(38:4)+K]⁺, [PA(40:5)+K]⁺, and [PC(40:6)+K]⁺ distinguish nontumor regions within both WT/HCC and FXR KO/HCC tissues (E.1-H.1 and E.2-H.2). H&E stain of WT/HCC (I.1) and FXR KO/HCC (I.2) reveal the location of visible tumors (black dotted line). Segmentation was used in combination with H&E stained tissue to determine tumor boundaries and highlight smaller, less apparent tumors (SI Figure S1). ROC AUC score was used to determine which lipids were discriminative of tumor and nontumor tissue.

Phospholipids detected by MALDI MSI showed a difference in expression between the tumors and the surrounding tissue (Figure ). Using ROC AUC analysis, [PC(32:1)+K]⁺, [PA(36:2)+K]⁺, m/z 774.5975 (unknown lipid), and [PC(34:1)+K]⁺ were determined to characterize the tumor (AUC > 0.750) (Figure A-D.1, A-D.2). The lipid feature m/z 774.5975 has an exact mass similar to a few lipids found in the LIPID MAPS database. However, tandem MS revealed fragmentation inconsistent with database suggestions (SI entry Figure S6). Meanwhile, some lipids were more abundant in the tumor environment than in the tumors. [PC(38:6)+K]⁺, [PC(38:4)+K]⁺, [PA(40:5)+K]⁺, and [PC(40:6)+K]⁺ strongly characterized the tumor environment (AUC > 0.800) (Figure E-H.1, E-H.2). Phospholipids that discriminated the tumor against the tumor environment were consistent between both WT/HCC (Figure I.1) and FXR KO/HCC groups (Figure I.2), with the exception of [PC(38:6)+K]⁺ (Figure E.1–2). [PC(38:6)+K]⁺ was the most discriminatory phospholipid between tumor and tumor environment within WT/HCC samples (Figure E.1) but was less discriminating and lowly abundant in FXR KO/HCC (Figure E.2). This suggests that [PC(38:6)+K]⁺ may be affected by knocking out the FXR gene.

Tumors in HCC samples had a higher abundance of monounsaturated phospholipids [PC(32:1)+K]⁺ (Figure A.1–2) and [PC(34:1)+K]⁺ (Figure C.1–2) compared to the surrounding tissue. The increase in monounsaturated phospholipids has been previously correlated to cellular proliferation and hepatic carcinogenesis in HCC by Hall et al. Additionally, the increase of PC(32:1) and PC(34:1) has been previously reported in human plasma of patients with HCC by Cotte et al. Here, the monounsaturated phospholipids are observed within tumorous regions, which are highly proliferative. Of the phospholipids that highly characterize the tumors (Figure A-D.1, A-D.2), [PA(36:2)+K]⁺ was only phospholipid that was not monounsaturated (Figure B.1–2).

Phospholipids with a higher relative abundance outside the tumor generally contained more points of unsaturation (Figure E-H.1, E-H.2). [PC(38:6)+K]⁺ and [PC(40:6)+K]⁺ (Figure E.1–2 and H.1–2, respectively) can occur in several isomeric forms, but often contain a docosahexaenoic acid fatty acid tail. Docosahexaenoic acid is an omega-3 fatty acid containing a 22-carbon chain with six cis points of unsaturation, known to be beneficial to human health, including cancer. , Separately, [PC(38:4)+K]⁺ (Figure F.1–2) and [PA(40:5)+K]⁺ (Figure G.1–2) commonly contain arachidonic acid as one of their fatty acid tails, which is known to be involved in inflammation.

Segmentation and clustering was performed in SCiLS lab software using bisecting k-means criterion to enable unsupervised clustering of the mass spectrometry data into tumor and nontumor tissue. This data-driven separation of data allows for a consistent classification of tumor location across samples and validated with hematoxylin and eosin (H&E) stain of the tissue section (SI Figure S1). Segmentation and H&E results were used to establish the molecular environments surrounding tissue at multiple distances. In data analysis, 50 μm wide concentric boundaries surrounding tumors were established to compare the lipid profiles of tissue regions at increasing distances from the tumors (Figure ). For both WT/HCC and FXR KO/HCC livers, [PC(32:1)+K]⁺ most strongly distinguished the tumors from nontumor tissue by ROC AUC score. [PC(32:1)+K]⁺ is 2.5-times as abundant in WT/HCC tumors and 1.8-times as abundant in FXR KO/HCC tumors, compared to nontumor tissue 50 μm from the tumor (Figure A). At 150 μm from the tumor, WT/HCC and FXR KO/HCC tumors have 3.4- and 4.1-times the abundance of [PC(32:1)+K]⁺ in nontumor tissue, respectively. This demonstrates the decrease in [PC(32:1)+K]⁺ as a function of the distance from the tumor. [PA(36:2)+K]⁺, m/z 774.5975, and [PC(34:1)+K]⁺ also distinguish the tumors from the nontumor tissue and decrease in average abundance as a function of distance from the tumor (Figure B-D). Conversely, the phospholipids that most represent the nontumor tissue are [PC(38:6)+K]⁺, [PC(38:4)+K]⁺, [PA(40:5)+K]⁺, and [PC(40:6)+K]⁺ (Figure E–H). In WT/HCC livers, [PC(38:6)+K]⁺ was a distinctive phospholipid in nontumor tissue compared to tumors by ROC AUC analysis. [PC(38:6)+K]⁺ appeared 2.3-times more abundant in nontumor tissue than in tumors (Figure E). Subsequently, [PC(38:6)+K]⁺ is significantly more abundant in the WT/HCC tumor and tumor microenvironment compared to FXR KO/HCC (Figure E). In FXR KO/HCC livers, the phospholipid that is most strongly characterizes nontumor tissue by ROC AUC analysis is [PC(38:4)+K]⁺ (Figure F). [PC­(38:4)+K]⁺ is 2.9-times more abundant in the FXR KO/HCC tumor microenvironment than in the tumor (Figure F). [PC(38:4)+K]⁺ is significantly less abundant in the microenvironment of WT/HCC tumors compared to FXR KO/HCC (Figure F). [PC(38:6)+K]⁺ is expected to contain a docosahexaenoic acid tail and [PC(38:4)+K]⁺ is expected to contain an arachidonic acid tail. The difference in expression of [PC(38:6)+K]⁺ and [PC(38:4)+K]⁺ may indicate a different abundance or uptake of these fatty acid tails during phospholipid synthesis due to FXR KO.

4.

MALDI MSI indicates that the tumor is molecularly different from the microenvironment in the WT/HCC and FXR KO/HCC tissues. The phospholipids that are differentially expressed between WT/HCC and FXR KO/HCC tissue display different localizations within the surrounding tissue as well. Lipids that most define the tumor (A–D) lessen in relative abundance as a function of distance from the tumor. These lipids are nearly uniform in relative abundance further than 100 μm from the tumor in WT/HCC (n = 4). However, the same lipids continue to decrease in relative abundance at least 150 μm from the tumor in FXR KO/HCC (n = 3). Conversely, lipids that most define the tumor microenvironment (E–H) increase in relative abundance as a function of distance from the tumor. Further concentric borders around the tumor were not considered due to proximity of tumors. Significance indicated as * = p < 0.1, ** = p < 0.05, *** = p < 0.01.

Some lipids that distinguished the tumor environment were more abundant in WT/HCC than FXR KO/HCC. For example, there was a significantly higher abundance of [PC(38:6)+K]⁺ and [PC(40:6)+K]⁺ in WT/HCC tumors compared to FXR KO/HCC tumors (Figure E,H, respectively), suggesting the FXR gene may be involved in the regulation of [PC(38:6)+K]⁺ and [PC(40:6)+K]⁺ in HCC. Membrane fluidity is affected by the presence of tighter hydrophobic packing of saturated lipids and looser packing of unsaturated lipids. The highly unsaturated lipids (Figure E–H) localizing outside the tumor suggest a more fluid cell membrane in the tumor environment compared to that in the tumor. This pattern attests to the rigidity observed in tumors.

Nonuniform changes of phospholipid expression were observed in the tissue adjacent to the tumor, exemplifying the microenvironment around the tumor. This observation demonstrates the complexity of the tumor microenvironment and supports the difficulty of identifying the outer boundary of the tumor in a clinical setting. For example, tissue residing within 50 μm from the tumor has similarities to both tumor and nontumor tissue. This is evident in m/z 774.5975 and [PC(40:6)+K]⁺ (Figure C,H, respectively). In the 50 μm border around the FXR KO/HCC tumors, m/z 774.5975 was not significantly different in abundance than in the tumor (Figure C). In contrast, [PC(40:6)+K]⁺ was significantly different from the tumor in the 50 μm border surrounding the tumor (Figure H). One explanation for these nonuniform phospholipid changes may be the accumulation of different cell types surrounding the tumor. Phospholipids are intrinsic parts of a cell membrane and can be indicative of cell type. , Instead of indicating tumor growth, epithelial-mesenchymal transition, or metastasis, some phospholipids may denote the presence of other cell types, such as immune cells that have been known to reside in the tumor microenvironment. To explore this further, future studies could benefit from a staining method alternative to the H&E stain used here, such as antibody-based immunofluorescence to reveal cell-types present in the same tissue section. ,

Hypothesis for the Role of PC(38:4) in Hepatocellular Carcinoma

The information gained from MALDI MSI is integral to establishing testable hypotheses. In this case, we observed that PC(38:4) was present in cells surrounding the tumor, and we aimed to test this using cell culture. We grew Hep3B cells in various concentrations of PC(38:4) to understand its effect on proliferation (Figure ). Overall, we used an MTT assay and showed that Hep3B cells showed significantly decreased proliferation when grown in serum-free growth media containing 5 × 10³ μg/L of PC(38:4) phospholipid standard (Figure A). Initially, we believed that this lipid facilitated cell growth, but eventually, the overabundance of cell growth may cause cell death. We then assessed other lipid changes that occurred using a lipid extract from the cells to see if this offered additional insight into these cellular mechanisms. Indeed, CE(20:4) increased significantly as a function of PC(38:4) concentration when many other lipids remained unchanged (Figure B). Interestingly, PC­(18:0/20:4; the standard we used) can be converted to CE(20:4) by lecithin:cholesterol acyltransferase (LCAT) through the addition of cholesterol [PC­(18:0/20:4) + Cholesterol --- LCAT→ CE(20:4) + LPC (18:0)] offering additional insight. Indeed, tumor cells require cholesterol to maintain proliferation, and LCAT is protective against HCC. This connection may indicate that cells surrounding the tumor may deplete cholesterol, limiting tumor growth as one potential hypothesis that can be explored further on. However, another potential hypothesis is that the higher concentration of the PC(38:4) lipid standard-caused cell death in cell culture experiments. Future experimentation to test these hypotheses will be needed to form conclusions about the mechanisms behind these observations.

5.

Hep3B cells grown in serum-free growth media containing 5 × 10³ μg/L PC(38:4) phospholipid standard resulted in (A) significantly decreased proliferation and (B) significantly increased abundance of CE(20:4). Each condition was conducted in triplicate. Proliferation was measured via MTT assay, in which the absorbance at 570 nm represents proliferation. Average absorbance values are represented as bar graphs with error bars representing standard deviation of absorbance values for each condition. Significance is indicated above each graph (* = p < 0.05; ** = p < 0.01; *** = p < 0.005).

Conclusions

In conclusion, the present study has demonstrated a considerable change in HCC phospholipid profiles when FXR, a suppressor of HCC and regulator of lipid homeostasis, is knocked out. Through MALDI MSI, we observed that the presence of HCC alters phospholipid composition in the liver, and these changes differ markedly in the absence of FXR. Notably, monounsaturated phospholipids such as [PC(32:1)+K]⁺ and [PC(34:1)+K]⁺ were enriched in tumor regions, while poly unsaturated lipids including [PC(38:4)+K]⁺ were more abundant in the tumor microenvironment, particularly in FXR KO models. Functional assays further supported the hypothesis that PC(38:4) may influence cell proliferation and cholesterol metabolism. These findings highlight the role of FXR not only in lipid regulation but also in shaping the molecular landscape of the tumor and its surrounding tissue. Overall, our results underscore the utility of spatial lipidomics for uncovering metabolic signatures associated with tumor progression and suggest that FXR loss drives distinct lipidomic remodeling in HCC. Future studies will involve further assessing these mechanisms, including the use of transcriptomics and proteomic analyses, as well as assessment in other types of cancers.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jasms.5c00365.

• MALDI MSI Method (ZIP)

• The WT/HCC and FXR KO/HCC tissue stains and segmentation are shown in SI Figure S1; PCA of lipid features detected in HCC vs non-HCC mouse liver tissue by MALDI MSI is shown in SI Figure S2; the experimental workflow is shown in SI Figure S3; a table of annotated mass spectral features is available in SI Table S1; MS/MS of mass spectral features are shown in SI Figure S4–S13 (PDF)

The authors declare no competing financial interest.

Published as part of Journal of the American Society for Mass Spectrometry special issue “Honoring Richard M. Caprioli.”