Delineating lipidomic landscapes in human and mouse ovaries: Spatial signatures and chemically-induced alterations via MALDI mass spectrometry imaging: Spatial ovarian lipidomics.

To establish a comprehensive characterization of the mouse ovarian lipidome, ovaries from 182 mice were collected and subjected to lipid extraction. Reversed-phase chromatography was coupled to a high-resolution mass spectrometer to separate out lipids based on their hydrophobicity, which can resolve isomeric lipids and reduce sample complexity before mass spectrometry detection in both positive and negative ion modes. Identification was conducted by matching the precursor mass, isotope pattern, and the tandem mass spectrum (MS/MS) of a lipid against the in silico database within the MS-DIAL software. In addition to the collection of MS1-only data from the pooled sample, iterative MS/MS was utilized during data acquisition for the comprehensive profiling of the mouse ovarian lipidome, which took advantage of a rolling exclusion list over seven LC injections of the pooled sample. This workflow increased the number of unique MS/MS spectra acquired, thus increasing the overall number of lipid annotations. Lipids identified by their MS/MS were kept, and features with a coefficient of variation greater than 20 % between three injections of the pooled sample were parsed out. Additionally, after manually parsing for duplicate lipid identification between MS polarity modes, 853 lipids were identified by MS/MS in the characterization, and up to level 2 annotation was made based on the guidelines proposed by Sumner et al. ( Sumner et al. 2007 ). The detected lipids covered four major lipid categories (fatty acyls, glycerolipids, glycerophospholipids, and sphingolipids). The exact number of lipids identified by lipid class and subclass is listed in Table S1 . The lipid class and subclass were identified, including up to their fatty acyl composition, when the MS/MS data permitted proper annotation. Otherwise, the sum composition of the total carbon and double bond numbers is reported. Figure 2 shows the lipid identification map where all lipid classes/and subclasses are plotted according to their m/z values and retention time. Class-specific lipid plots (by retention time and total number of carbons) are provided in the Supplementary Figures S1 – S6 . MALDI MSI analysis targeted ovarian sections to characterize the lipid distribution in a spatial manner. Approximately 100 lipid ions were detected using DAN matrix, across both positive and negative ion modes. Representative mass spectra for each mode are shown in Figure S7 . Using the lipid library generated from the LC-MS lipidomics analysis, 78 peaks were putatively assigned an identity based on their closest match by accurate mass. Peaks with variable intensity were found mainly in the m/z range of 300 to 980. In addition, the resulting ion density maps revealed that a fraction of those lipid species showcased a spatial distribution that was highly enriched (almost exclusive) to a particular ovarian compartment ( Figure 3A ). Using the LC-MS data, lipids were putatively assigned based on the closest match by mass. For instance, the negative lipid adducts putatively annotated as phosphatidyl inositol (PI) 38:4 ( m/z 885.6), phosphatidylglycerol (PG) 36:3 ( m/z 771.6), (PS) 36:6 ( m/z 778.5), and phosphatidylethanolamine PE 40:1 ( m/z 800.5), displayed a preferential localization for the ovarian cortex, stroma, follicles, and oocytes, respectively. MSI data visualization was used to generate composite ion images where the abundance of selected m/z peaks was represented by different colors ( Figure 3B ). The accuracy of these enrichments was validated using the ROIs selected using histological images of the matching ovarian sections. TBT exposure induced a dose-dependent accumulation of neutral lipids in exposed ovaries compared to control ovaries, evidenced by a significantly higher BODIPY intensity, corresponding to more neutral lipids, primarily localized in the stroma and outer layers of the follicular wall ( Figure 4 ). To further evaluate if TBT exposure affects lipid composition (beyond neutral lipids) in a spatially resolved manner, lipid signatures from control and TBT-exposed mouse ovaries (n = 5 / group) were compared using MALDI MSI. When comparing the lipid abundance between the control and TBT-exposed ovaries across entire ovarian sections, three fatty acids - palmitic acid (FA 16:0), stearic acid (FA 18:0), and docosahexaenoic acid (DHA; FA 22:6) - were significantly downregulated, while phosphatidylcholine PC 32:0 was upregulated upon TBT exposure. Coupling MALDI MSI with H/E staining allowed co-registration of the lipid data to ovarian tissue histology. Based on this histology-guided MSI workflow, whether lipid signatures varied across ovarian regions (namely stroma, cortex, growing follicles, and oocytes) and how the regional abundance of lipid species differed between control and TBT-exposed ovaries were evaluated. In the stroma of TBT-exposed ovaries, the downregulated lipid species ( Table S2 ) were palmitic acid (FA 16:0), stearic acid (FA 18:0) and DHA (FA 22:6). In the cortex of TBT-exposed ovaries, the dysregulated lipids were mostly phospholipids, including four phosphatidylcholines (PC) and nine phosphatidylethanolamines (PE; Table S3 ). Regarding ROIs containing follicles, 34 different lipid species were dysregulated upon TBT exposure ( Figure 5 , Table S4 ). Of note, six fatty acids were down-regulated (FA 16:0, FA 18:0, FA 19:0, FA 20:1, FA 22:5, and FA 22:6), while free cholesterol was upregulated, along with several phospholipids. The majority of these phospholipids were PEs and ether-linked PEs. In addition, ROIs from TBT-exposed oocytes displayed 27 dysregulated lipids, including the downregulation of six fatty acids (FA 16:0, FA 18:0, FA 19:0, FA 20:1, FA 22:5, and FA 22:6) and the upregulation of sphingomyelins, PEs, and ether-linked PEs ( Figure 5B , Table S5 ). Next, an enrichment analysis was performed using the LION/web lipid ontology platform. Lipid identifiers were ranked by logarithmic transformation, reflecting which lipid properties were overrepresented in the dysregulated lipid subset of each ovarian compartment. Among the four ovarian compartments studied, significant enrichments were only observed in the dysregulated lipids from follicles and oocytes exposed to TBT. Seven LION terms were significantly enriched in TBT-exposed follicles compared to non-exposed ones ( Figure 6A ), including: ‘ glycerophosphoethanolamines ’, ‘ mitochondrion’ , ‘ fatty acids and conjugates’ , and ‘ contains ether bond’ . For TBT-exposed oocytes, the six significantly enriched LION terms included: ‘ fatty acids and conjugates’ , ‘ glycerophosphoethanolamines ’, and ‘ above average bilayer thickness ’. To gain additional insights as to the pathways in which TBT dysregulated lipids are involved, we conducted a series of pathway analyses using the BioPan platform. In oocytes and follicles, the conversion of certain lysophosphatidylethanolamines (LPEs) to phosphatidylethanolamines (PEs) is favored upon TBT exposure, along with the elongation of FA 16:0 into 18:0, in both compartments. MALDI MSI analysis of human ovarian tissue was performed in negative ion mode, which enabled visualization of approximately 150 negatively charged lipid ions (mainly [M-H] - ). Spectra were recorded for more than 60,000 data points throughout an entire ovarian section. The skyline projection spectra generated from all detected ion signals in the 200 – 1,500 mass range within the tissue section displayed numerous peaks with variable intensity mainly over the 300 – 900 m/z range, which is the mass range that provides the highest sensitivity in mass spectrometry (Zemski Berry et al. 2011). Consistently with our observations in mouse ovaries, the molecular images of certain lipid ions reflect that the relative abundance of those lipids was highly aligned to functional compartments of the human ovary ( Figures 8 and 9 ). For instance, the ion map corresponding to m/z 615.3 was almost exclusively detected in corpora lutea, whereas m/z 1050.8 was observed to be highly aligned with the theca cell layer ( Figures 8A and 8D , respectively). Additionally, multi-ion images of selected m/z values depicting lipid ions with distinct spatial distribution were generated in three-color overlay images ( Figure 9 ). Each composite ion map was matched with its corresponding H/E image to validate the accuracy of each regional lipid enrichment. Notably, the theca cell layer was identifiable across different follicles using an m/z value of 1050.7, which appeared to be almost exclusive to this cell type. Conversely, the granulosa cell layer can be distinctly visualized using m/z 279.3 ( Figure 9C ) consistently across follicles of all stages (from primordial to antral). The pool of primordial follicles along the cortical ridge can be visualized using a combination of m/z values (279.3 and 835.7; Figure 9D ). Additionally, spatial distribution of the m/z 426.2 peak was highly co-localized with blood vessels of various diameters ( Figure 9E ). While an abundance comparison between the mouse and human lipid profiles would not be appropriate (due to the difference in sample preparation and lack of lipid assignments for the human ovary, among others), it is possible to compare the presence or absence of negatively charged lipid ions ( Figure S8A ). Most of the lipid ions detected in mouse ovaries (87%) were also detected in human ovaries. However, only 17% of the human lipid ions were also detected in mouse ovaries. Two lipid ions (putatively assigned in the mouse lipidome as PI 36:2 and PI 36:1) were enriched in similar compartments in both mouse and human (namely, oocytes and primordial follicle cells) ( Figure S8B – C ). Lastly, an unsupervised spatial segmentation in SciLS was performed using the MALDI MSI data from Figure 10 . This clustering approach of mass spectrometry imaging data creates a visual representation of a dataset by assigning pseudo-colors to different clusters of pixels containing similar molecular features without any prior histological knowledge. This segmentation map allowed the visualization of ovarian compartments based only on molecular (mass spectral) and spatial information. The exploration of the spatial segmentation map was limited to 16 clusters to simplify visualization. The spatial segmentation clusters generated from the molecular features (mainly lipid ions) ( Figure 10 ) were then compared to the ovarian compartments (ROIs) defined from histological observation in H/E sections ( Figures 9A2 and 9B2 ). This comparison revealed that compartments determined by histology had a high percentage of overlap with several molecular clusters of the spatial segmentation maps ( Figure 10 B – D insets).

All experimental procedures performed on animals were in accordance with the Animal Care Committee at the University of Illinois at Chicago (#20–232). A total of 182 mouse ovaries from 91 mice were used to develop an in-house ovarian lipid library (See Figure 1A for study design) to enable accurate lipid annotations from the MALDI-derived analyses (described below). Specifically, 76 ovaries from CD1 female mice and 106 from C57BL/6 female mice with ages ranging between 21 – 220 days of age were included. After sacrifice, ovaries were flash frozen and kept at −80 °C until processing for untargeted lipid analyses. Age- and strain-matched ovaries were pooled to create 37 pools containing at least 10 g of ovarian tissue each for untargeted lipidomics. Tissue homogenization was performed by sonicating samples at 4 °C using a Bioruptor 200-UCD sonicator bath (Diagenode, Liège, Belgium) in tubes with 400 μL of 1x PBS for 5 cycles x 1 min (30 s on/off at maximum amplitude). Aliquots of these homogenates were used to measure protein concentration with a BCA assay as previously described ( Pu et al. 2022 ). Aliquot homogenates were used for lipid extraction. Lipid extraction was conducted on ice on aliquots of sample equivalent to 200 μg of protein using a modified protocol by Matyash et al. ( Matyash et al. 2008 ). Briefly, 5 μL of the equiSPLASH LIPIDOMIX (Avanti Polar Lipids, Alabaster, AL, USA) were added to each sample, followed by 300 μL of cold methanol. Then, 1,000 μL of cold methyl tert-butyl ether (MTBE) were added followed by vortexing and bath sonication. Liquid chromatography-mass spectrometry (LC-MS) grade water was then added to induce phase separation, followed by another round of vortexing and bath sonication. Each sample was then centrifuged at 14,000 rpm for 2 min. The upper organic phase of each sample was collected for lipidomics using a Hamilton syringe and rinsing with MTBE between every sample. MTBE was added a second time, and this process was repeated to collect a second extraction of lipids from each sample. Lipid fractions were dried using a nitrogen gas stream and stored at −80 ⁰C until LC-MS analysis. Dried lipid extracts were resuspended using a solution of methanol/toluene (9:1, v/v) to obtain a protein equivalent concentration of 2 μg/μl for each sample. A representative pooled sample was subsequently made from equal portions of all the mouse samples. This pooled sample was injected periodically throughout the whole LC-MS analysis to monitor system suitability. In addition, the pooled sample was also used to generate an in-house spectral library using Auto MS/MS mode, utilizing the Iterative MS/MS function over seven injections. This type of data-dependent acquisition (DDA) enabled the use of a rolling exclusion list of retention time and mass to charge ( m/z ) pairs throughout the seven injections to obtain a more comprehensive lipidomics profiling of the pooled sample. Biological samples were analyzed in MS-Only mode. All samples were injected on a 1290 Infinity II UHPLC System (Agilent Technologies Inc., Santa Clara, California, USA) onto an Acquity Premier C18 column (1.9 μm, 150 × 2.1 mm) (Waters Corporation, Milford, Massachusetts, USA) for reversed-phase chromatography which was maintained at 50 °C with a constant flow rate at 0.200 ml/min, using a gradient of mobile phase A (9:1 water/methanol, 10 mM ammonium acetate, 0.2 mM ammonium fluoride) and mobile phase B (2:3:5 acetonitrile/methanol/isopropanol, 10 mM ammonium acetate, 0.2 mM ammonium fluoride). The gradient program was as follows: 0 – 1 min, 70 %B; 1 – 2.5 min, 70 – 86 %B; 2.5 – 8 min, 86 – 96 %B; 8 – 11 min, 96 – 100 %B, 11 – 17 min, hold 100 %B; 17 – 17.10 min, 100 – 70 %B; 17.10 – 20 min, hold 70 % %B. “MS-Only” positive and negative ion mode acquisitions were conducted on an Agilent 6545 quadrupole time-of-flight mass spectrometer equipped with a JetStream ionization source. The source conditions were as follows: gas temperature, 200 °C; drying gas flow, 10 l/min; nebulizer, 50 psi; sheath gas temperature, 300 °C; sheath gas flow, 12 l/min; Vcap, 3,500 V (3,000 V for negative mode); fragmentor, 150 V; skimmer, 65 V; and Oct 1 RF, 750 V. The acquisition rate in MS-Only mode was 3 spectra/second, utilizing m/z 121.050873 and m/z 922.009798 as positive ion reference masses, or m/z 112.985587 and m/z 1033.988109 as negative ion reference masses as reported by the instrument vendor. The acquisition parameters for Iterative Auto MS/MS are similar to MS-Only mode, with the addition of collision energy, 25; isolation width, narrow (1.3 m/z ); top three precursors per cycle; precursor threshold, 4,000 counts and 0.001 %; active exclusion was enabled for one spectrum and released after 0.05 min. Raw LC-MS data (Agilent .d files) were directly imported to MS-DIAL (v.4.9.221218) ( Tsugawa et al. 2020 ). The MS/MS data from the pooled injection files were used to identify lipids against the MS-DIAL lipidomics in silico library. The abundance of each identified lipid generated by MS-DIAL was obtained by the area of the respective chromatographic peak. Each lipid was normalized to the appropriate lipid internal standard for the class or subclass. If a standard was not directly available, the closest eluting LIPIDOMIX internal standard was used as a normalization. The relative abundance of lipids was normalized against the assigned internal standard and protein amount. The negative ion and positive ion data were manually merged and further curated to remove (if any) redundant compounds found in both polarities, and more importantly, remove false positive identifications. The R packages SCOPE and lipidR were used to visualize the lipidomics data ( Mohamed et al. 2020 ). To note, different sets of ovaries were used for chemical exposure followed by MALDI MSI and for creating the mouse lipid library, as MALDI MSI requires embedding and cryosectioning, whereas LC-MS untargeted lipidomics needs a larger amount of non-embedded tissue for reliable lipid extraction. Prepubertal female mice (3 weeks old, C57BL/6J, n = 8/group) were humanely euthanized using CO 2 . Their ovaries were collected and cultured in Transwell inserts (Cat#: 3460, Corning, Kennebunk, ME, USA) in α-MEM (Cat#: 11095–080, Life, Grand Island, NY, USA) supplemented with 10% fetal bovine serum (Cat#: 35–010-CV, Corning, Manassas, VA, USA), 0.25 mM sodium pyruvate (Cat#:11360–070, Sigma, Grand Island, NY, USA), 1% insulin transferrin selenium (Cat#: I1884, Sigma, Saint Louis, MO, USA), and 1% penicillin/ streptomycin (Cat#: 30–002-Cl, Corning, Manassas, VA, USA) in 5% CO 2 at 37 °C. The ovaries were treated with vehicle (0.1% DMSO) or tributyltin chloride (TBT; Cat#: T50202 , Sigma, Saint Louis, MO, USA; 50 ng/ml) for 11 days, and half of the medium volume was replaced daily. At the end of the culture, the ovaries were embedded in either optimal cutting temperature compound (OCT) or gelatin-casted (see details below) and stored at −80 °C until cryosectioning. The exposure dose of 50 ng/ml of TBT was based on human exposure levels as previously reported ( Pascuali et al. 2024a ). The U.S. Environmental Protection Agency oral reference dose is 300 ng/kg BW/day ( Agency 1997 ), while human circulating concentrations ranges between 0.049 and 10.5 ng/ml and reaches up to 85 ng/ml ( Kannan et al. 1999 ). Thus, the dose used is within relevant human exposure levels. Human ovarian tissue was obtained with written informed consent after Institutional Review Board (IRB#: 2020–1278) approval from the University of Illinois at Chicago and consistent with relevant guidelines and regulations. A human specimen from a 26-year-old female was harvested from a healthy non-pregnant donor subjected to ovariectomy at the University of Illinois Hospital (Chicago, IL, USA). Exclusion criteria included polycystic ovarian syndrome, ovarian cancer, menopause, drug addiction, and/or diagnosed with HIV or hepatitis B or C. Ovarian tissue was placed at 4 °C in transport medium (DMEM/F12 supplemented with 2 % penicillin / streptomycin and 2% antibiotic-antimycotic) and processed within 1 h of the surgery. The tissue was dry blotted, frozen on a flat dry ice surface to protect tissue structural integrity and stored at −80 °C until cryosectioning ( Figure 1C ). A total of 8 mouse ovaries were used per experimental group in this study. Mouse ovaries (n = 3 / treatment) were placed in OCT-filled (Cat#: 4583, Sakura, Torrance, CA, USA) disposable cassettes ( Figure 1B ). The OCT blocks were obtained by controlled snap freezing in isopentane (2-methylbutane), pre-chilled in liquid nitrogen. Samples were then transferred to an −80 °C freezer for storage. A 3-dimensional (3D) Petri dish mold (Cat#: 12–36TO, Microtissues, Inc., Providence, RI, USA) was used to make gelatin micromolds. Briefly, 5 g commercially available unflavored gelatin powder (Cat#: 252510, Walmart Inc, Bentonville, AR, USA) was dissolved in distilled water and warmed in a pressure cooker. To form the gelatin base, 500 μL of warm gelatin were carefully pipetted into the 3D molds and stored at −20 °C to harden. Upon collection, ovaries (n = 5 / group) were transferred onto the gelatin micromold and capped by carefully pipetting 200 μL of warm gelatin. Gelatin blocks were then stored at −80 °C until cryosectioning. Due to their large size, no tissue embedding was required for the human ovaries. Cryosectioning of human and mouse ovaries was performed on a Leica CM 1950 cryostat (Leica Biosystems, Buffalo Grove, Illinois, USA). For neutral lipid stain analyses (only mouse ovaries), 5-μm sections were mounted onto microscope slides (Cat#: 12–550-15, Thermo Fisher Scientific, Waltham, MA, USA). For MALDI analyses (mouse and human ovaries), ovaries were cryosectioned into 10-μm sections and thaw-mounted onto indium tin oxide (ITO)-coated slides (Cat#: CG-90IN-S115, Delta Technologies, Loveland, CO, USA). All slides were stored at −80 °C until further use. Neutral lipid stain was performed on mouse ovary OCT-embedded frozen sections. Sections were first thawed at room temperature for 10 min, then rehydrated in 1x Dulbecco’s Phosphate-Buffered Saline (DPBS). Next, slides were fixed in 10% neutral buffered formalin solution (Cat#: HT501128, Sigma, Saint Louis, MO, USA) for 10 min, then incubated with 1 μg/ml BODIPY493/503 working solution for 20 min. After five washes with 1x DPBS, the slides were incubated with DAPI working solution (5 μg/ml) for 10 min then washed again five times with 1x DPBS. After adding a drop of Fluoromount, a glass cover slide was placed on top of the sections and were sealed with nail polish. Whole ovary sections were imaged on a 10x objective lens using a Nikon Eclipse Ti-U fluorescence microscope (Tokyo, Japan) coupled with a high-performance EMCCD & CCD camera (Tucson, AZ, USA). Green and blue channels using 461 nm and 519 nm filters were used to capture BODIPY-stained lipid droplets and DAPI-stained nuclei, respectively, using the NIS-Elements imaging software (AR 4.40.00, Nikon, Melville, NY, USA). Lipid droplets were quantified using a streamlined pipeline in the open-access software CellProfiler (version 4.2.5), as previously described ( Adomshick et al. 2020 ). The integrated density corresponding to BODIPY fluorescence (density/section area) was quantified for each image taken (n = 3 – 4 sections / animal; n = 3 animals/group). MALDI MSI was performed in human and mouse ovaries ( Figure 1D ) at the Integrated Molecular Structure Education and Research Center (IMSERC) at Northwestern University. Frozen glass slides were defrosted under vacuum conditions using a vacuum chamber. MALDI matrices 1,5-diaminonaphthalene (DAN) or 9-aminoacridine (9AA) were used for negative mode imaging, while 2,5-dihydroxybenzoic acid (DHB) was used for positive mode imaging. Using an HTX TM-Sprayer (HTX Technologies, LLC, Chapel Hill, NC, USA), a solution of 5 mg/ml of DAN, dissolved in (90/10) (v/v) acetonitrile and water was applied in four passes using the following sprayer settings: nozzle temperature, 70 °C; flow rate, 0.1 ml/min; velocity, 800 mm/min; gas flow rate, 2 l/min; and drying time, 10 sec. In the case of 9AA, a saturated solution (10 mg/ml), dissolved in (75/25) (v/v) acetonitrile and water was filtered using a 0.22 μm-pore membrane and applied in 8 passes. The sprayer settings for 9AA were: nozzle temperature, 60 °C; flow rate, 0.12 ml/min; velocity, 1200 mm/min; gas flow rate, 3 l/min; and drying time, 10 sec. Alternatively, a solution of 10 mg/ml of DHB dissolved in (50/50) (v/v) acetonitrile and water was applied in eight passes using the following sprayer settings: nozzle temperature, 60 °C; flow rate, 0.12 ml/min; velocity, 1,200 mm/min; gas flow rate, 3 l/min; and drying time, 2 sec. All matrices were applied with a crisscross pattern with spacing of 2 mm, 10 psi pressure, and a nozzle height of 40 mm. MALDI MSI acquisition was performed using a rapifleX MALDI Tissuetyper (Bruker Daltonik GmbH, Bremen, Germany). Data was acquired in positive- and negative-ion reflectron modes and processed using the software SciLS Lab 2023 Pro (Bruker Daltonik GmbH, Bremen, Germany). For the mouse ovary imaging, the mass range collected was m/z 200 – 2000, collecting 500 shots per pixel, using a laser power at 75% - 80%, a sampling rate of 1.25 GS/s and the raster and laser focus were matched to obtain a 20 μm image resolution. Putative lipid assignments were made by comparing accurate mass to their closest match in the in-house lipid library generated from the LC-MS lipidomics analysis. All annotations of the MALDI data were based on the lipid summed annotation format as MS1-only data can not determine the double bond chemistry and stereochemistry. Replicate analysis was carried out using averaged region-of-interest mass spectrum data by normalizing the TBT-exposed mass spectra to the controls. After MSI, slides were retrieved and subjected to H/E staining (adapted from ( Kaya et al. 2017 )). Briefly, the matrix was washed away by submerging in 100% ethanol (2 × 2 min). Tissue sections were rehydrated in a decreasing ethanol concentration gradient (2 min; 90%, 70%, and 50%) and washed in milliQ water for 2 min. Slides were then stained in Harris hematoxylin (Cat #:HHS16, Sigma Aldrich, Saint Louis, MO, USA) for 5 mins, washed in running tap water for 1 min and counterstained using eosin Y 1% in water (Cat#: 0109, Amresco, Solon, OH, USA) for 3 min. Sections were dehydrated with 10 sec ethanol washes (70%, 90%, and 100%) and submerged in 100% xylene. Tissues were then mounted using Permount mounting media (Cat#: SP15–100, Thermo-Fisher, Rockford, IL, USA) and coverslips. Whole slide scanning was performed using a Aperio AT2 Scanner (Leica Biosystems, Wetzlar, Germany) at the University of Illinois Research Histology and Tissue Imaging Core Facility. H/E scans were uploaded to SciLS and carefully annotated using a combination of selection tools in SciLS and in the freely available software QuPath v0.4.1 ( Bankhead et al. 2017 ), integrating histological information with MSI data, as reported in other tissues ( Buerger et al. 2021 ; Neumann et al. 2021 ). Specifically, QuPath annotations were imported into SciLS as regions of interest (ROI). For mouse ovaries, the following morphological compartments were selected: whole ovarian tissue, cortex, stroma, growing follicles, and oocytes. These ROIs were used for spectral comparisons. Small follicles and atretic follicles were not taken into consideration for the analysis due to the variability in their presence across the samples. In the case of large follicles (> 0.1 mm) and oocytes, each compartment included 10 – 30 ROIs (individual follicles or oocytes, respectively). For statistical analysis of ion intensities and spatial distribution between control and TBT-exposed compartments, the hypothesis test within SciLS was utilized (Student’s t-test). This step determined if certain m/z values can discriminate differences between control or TBT-exposed ovaries. The output list was manually curated to validate spatial distribution differences. Finally, the average intensity for a particular ROI per ovary was calculated, exported to Microsoft Excel and confirmed for normal distribution and statistical significance. For human ovaries, spatial segmentation was conducted in SciLS as previously reported ( Trede et al. 2012 ). The bisecting k-means algorithm with the correlation distance approach was used for spatial segmentation analysis. The resulting spatial segmentation map and its corresponding hierarchical dendrograms are composed of pixel clusters grouped as pseudo-colors, where each cluster shares similar mass spectral characteristics. Lipid ontology (LION) enrichment analysis was conducted using LION/web (version v. 2023.04.14), an online tool performing network analysis within lipidomic datasets ( Molenaar et al. 2019 ). Based on a library of > 50,000 lipid species, LION/web does not rely on lipid nomenclature alone but rather reports whether specific physicochemical properties, biological functions, or cellular localization are enriched in a condition of interest. Four independent analyses were performed, one per ovarian region (follicles, oocytes, stroma, and cortex). For each analysis, the putative identities of dysregulated lipids within each region were compared with the list of all the lipid species detected in mouse ovaries. This allowed to identify lipid ontology enrichment within each region. In these analyses, a given term was considered to be enriched if the lipids belonging to said term were overrepresented compared to that to be expected by chance. Lipid terms were sorted by enrichment using -log (false discovery rates (FDR) q-values) and a threshold was set for each analysis to determine those significantly enriched within each TBT-exposed region. LC-MS/MS data were analyzed using BioPan ( Gaud et al. 2021 ), an openly available lipidomic pathway analysis tool that is part of the LIPID MAPS Lipidomics Gateway ( https://lipidmaps.org/biopan ). First, the datasets (including putative lipid identity and relative abundance per sample) were converted into the LipidLynx nomenclature. Then, these were entered into the BioPan lipidomics analysis tool, which allows visualization of lipidomics data in the context of their enzymatic reactions.

Despite ever-growing evidence that dyslipidemia is associated with infertility, the ovarian lipidome remains vastly unexplored. To address this gap, our study first evaluated the mouse ovarian lipidome employing a robust analytical workflow involving LC-MS/MS and identified and putatively annotated 853 lipid features, offering a comprehensive view of the ovarian lipid landscape. Leveraging this mouse ovarian lipid library, we next investigated the impact of an environmental exposure on the ovarian lipidome. Using MALDI MSI, we interrogated this effect from a spatial context perspective. We have not only discovered that the ovary holds specific lipid enrichment in regions that are biologically relevant, but also documented that chemical exposures impact the ovarian lipidome in a region-specific manner. This regional lipid enrichment in biologically relevant regions of the ovary was also detectable in human ovaries. These compartmentalized lipidome observations are to our knowledge the first of their kind in a human healthy ovarian sample and open the door to understanding ovarian-specific dyslipidemia in the context of environmental exposures. The identification and annotation of 853 lipids offered one of the most complete outlooks of the ovarian lipid composition in the mouse ovary thus far. The 33 lipid subclasses detected illustrate the rich diversity of lipids. In comparison with other mouse organs, in which typically a range of 150 to 900 lipid features have been detected using similar technical approaches ( Jain et al. 2022 ; Surma et al. 2021 ), the ovary falls in the upper limit of the range next to the kidney (855 unique lipid features; ( Jain et al. 2022 )) and the lung (793 unique lipid features; ( Jain et al. 2022 )). Few lipidome analyses are available in vertebrate ovaries or ovarian cell types, with most reproductive-related studies in females reporting plasma ( Jové et al. 2018 ; Li et al. 2016 ; Pradas et al. 2019 ), follicular fluid ( Ban et al. 2021b ). Studies that have evaluated the human ovarian lipidome in a spatial context ( Arafah et al. 2014 ; Dória et al. 2016 ; Meriaux et al. 2010 ; Tzelepi et al. 2023 ),do not provide information on healthy ovarian tissue with majority of evaluated tissues neither presenting functional structures (follicles, corpora lutea), nor being representative of healthy ovarian stroma. Among the few reports of the lipid composition of ovarian cells or tissues, a lower diversity of lipid species has been reported. For instance, a study found 439 lipids across 21 subclasses in bovine follicular cells ( Bertevello et al. 2018 ), while other studies reported 111 lipids across 8 subclasses in cattle oocytes ( Chen et al. 2020 ) and 662 lipids across 17 subclasses in goose granulosa cells ( Yuan et al. 2021 ). Furthermore, a recent study that quantified the mouse embryo lipidome during development, detected approximately 300 lipid species from 23 classes in oocytes from superovulated females ( Zhang et al. 2024 ). The identification of over 30 lipid subclasses with a wide range of m/z values, from fatty acids (as low as m/z 250) to cardiolipins (as high as m/z 1,500), is indicative of the robustness of this MS/MS approach. To our knowledge, this is one of two studies in the mouse ovary that have reported the presence of cardiolipins ( Oemer et al. 2020 ), mitochondria-specific phospholipids exclusively located in the inner mitochondrial membrane and essential for mitochondrial respiration and dynamics ( Dudek 2017 ; Lu and Claypool 2015 ) and reported to hold significant differences in tissue-specific abundance and phospholipid composition ( Oemer et al. 2020 ). MSI datasets are structured in such a way that each pixel of a digitalized tissue image is associated to a mass spectrum, which records the presence and relative abundance of each analyte detected as a particular peak ( Galli et al. 2016 ). Leveraging this capability, MALDI-MSI was utilized to identify the spatial location of each detected lipid peak. The current approach yielded a high percentage of annotation (~80%), especially when compared with other MALDI MSI studies in reproductive tissues (~50%; ( Bertevello et al. 2020 )), which is likely due to the extensive mouse in-house lipid library generated via LC-MS/MS. Our findings support the notion that the ovary holds marked lipid enrichment within biologically relevant ovarian regions, including stroma and follicles ( Uzbekova et al. 2015 ). Our results also demonstrate that compartments as small as mouse oocytes, which are ~70 μm in diameter, also have a unique lipid signature. Interestingly, the subset of lipids that exhibit the most evident spatial distribution are phospholipids, a characteristic that has also been reported in brain tissue ( Kaya et al. 2023 ). Whether unique phospholipid species tend to accumulate in specific ovarian regions likely relates to the enrichment of specific metabolic pathways within that region, as for other organs ( Kaya et al. 2023 ) remains to be investigated. To interrogate lipid-mediated mechanisms driving ovarian dysfunction, we used the obesogenic chemical TBT ( Veiga-Lopez et al. 2018 ) as a ‘model chemical’ exposure in an organotypic culture. Our findings demonstrated that TBT induces neutral lipid accumulation in mouse ovaries in a dose-dependent manner, which was particularly exacerbated in the stroma and outer follicular layer of large antral follicles. This is consistent with our previous reports, in which TBT induced 1) dysregulation in cholesterol trafficking ( Pu et al. 2019 ) and 2) a dose-dependent lipid droplet accumulation in ovine and human primary theca cells, along with dysregulation of other lipid species and lipid-associated enzymes ( Pascuali et al. 2024b ). While a higher abundance in neutral lipids suggests that TBT-induced lipid dysregulation, lipid droplets are not only composed of a neutral lipid core often containing cholesteryl esters, but also surrounded by a phospholipid monolayer that can contain hundreds of different phospholipid molecular species ( Penno et al. 2013 ). A similar shift in lipid metabolism to that induced by TBT has been reported in cancer cells, where cells upregulate de novo lipogenesis and lipid uptake, resulting in the active incorporation of fatty acids into complex lipids coupled with accumulating lipid droplets ( Matsushita et al. 2021 ). It is worth noting that triglycerides and cholesterol esters were not detected in MALDI MSI due to the challenges in ionizing non-polar molecules. However, the lower abundance of cholesterol observed in TBT-exposed ovaries could be driven by increased esterification and accumulation of these esters in lipid droplets. When interrogating biologically relevant ovarian regions (stroma, follicles, oocytes), distinct region-specific lipid dysregulations were detected upon TBT exposure. Specifically, in antral follicles and oocytes, TBT leads to a higher abundance of numerous phospholipids, mostly PEs. Of note, PEs constitute the most abundant glycerophospholipid and second most abundant phospholipid in eukaryotic cells and are of high abundance in the inner mitochondrial membrane ( Vance 2015 ). These results indicate that TBT likely activates one of the four PE synthetic pathways at the endoplasmic reticulum-mitochondria interface ( Calzada et al. 2016 ) resulting in a higher abundance of PEs in antral follicles. Whether this occurs directly by affecting a specific step in one of the four PE synthetic pathways (CDP-ethanolamine pathway, acylation of LPE, head group base exchange reactions, and phosphatidylserine decarboxylase (Psd) pathway) or indirectly by triggering an intracellular response that, in turn, activates PE synthesis remains to be investigated. These insights are validated by the results obtained using pathway analyses in follicle- and oocyte-dysregulated lipids, which point to an increase of PEs over LPEs. As for the impact of these dysregulations on ovarian cells, one may speculate that the shift from LPEs to PEs upon TBT exposure could indicate a shift in lipid metabolism, affecting overall lipid composition within oocytes and follicles. Since PEs are key in maintaining membrane fluidity and structure ( Dawaliby et al. 2016 ), which in turn is vital for cellular signaling and overall membrane dynamics, an increase in PE levels may lead to altered membrane properties and potentially affect key processes like meiosis. This shift in lipid balance may also interfere with signaling cascades that regulate cell growth, differentiation, and survival. Additionally, changes in this conversion process may influence metabolic pathways and the availability of fatty acids for energy production. Altogether, this could impact both, oocyte development and follicular maturation. Regarding the potential activation of the fatty acid elongation pathway from 16:0 (palmitic acid) to 18:0 (stearic acid), this would represent a shift toward longer-chain fatty acids and more complex lipid structures. Altered fatty acid balance could disrupt metabolic homeostasis, which is essential for follicular growth and oocyte quality ( Shi and Sirard 2022 ). Moreover, the reprogramming of lipid elongation pathways and remodeling could induce oxidative stress, which may be part of an adaptive response to chemical exposure. Such stress has the potential to increase the risk of oxidative damage in oocytes and surrounding follicular cells, negatively impacting ovarian function. Independently of the mechanism, the high content of PE in cellular membranes is likely the reflection of changes in mitochondrial bilayer membrane and protein topology ( Bogdanov et al. 2002 ; Osman et al. 2011 ). This is further supported by the enrichment analyses where ‘ mitochondrion’ was among the top lipid ontology terms in this study, as well as in our prior work where primary human theca cells were exposed to TBT ( Pascuali et al. 2024b ). The cause of the regional specificity (antral follicles) upon TBT exposure remains to be elucidated, but we hypothesize that changes in the mitochondrial membrane of antral follicles — high energy-demanding biological units ( Makanji et al. 2014 ; Sugiura et al. 2005 ; Sugiura et al. 2007 ) — may be more susceptible to mitochondrial dysregulation than less energy-demanding ovarian structures ( e.g., smaller follicles or stroma). To note, the fatty acyl chains of most upregulated PEs were 16 and 18 carbon chains. We hypothesize that the reduction in the abundance of several saturated fatty acids (16:0, 18:0, and 19:0) in the antral follicle region is possibly the direct consequence of the increased abundance of 16 and 18 fatty acyl chain PEs. This hypothesis is supported by our prior work that demonstrated that TBT drives the upregulation of several genes involved in fatty acid synthesis, including FASN ( Pascuali et al. 2024b ), the last step before palmitate final synthesis. In addition, a lower PC/PE ratio has been associated with several cellular disruptions, including mitochondrial deficits (reduced ATP production) and increased lipid droplets ( van der Veen et al. 2017 ). In the absence of TBT-induced changes in PCs, we expect that the PC/PE ratio to be lower upon TBT exposure. Our current work confirms this hypothesis (higher neutral lipid droplets) and higher abundance of neutral lipid droplets in TBT exposed human primary theca cells and mouse oocytes exposed to TBT during in vitro maturation ( Pascuali et al. 2024b ). Of significance in the context of environmental exposures, other chemicals have also been reported to induce neutral lipid accumulation in ovarian cells (flame retardants ( Wang et al. 2021 ) and perfluorinated chemicals ( Hallberg et al. 2019 )). Being more accessible than other available lipid markers, the neutral lipid quantification approach has become the default when assessing in situ tissue or cell dyslipidemia in most environmental and toxicological studies. Our results with TBT suggest that when neutral lipid dysregulation is observed, this finding may just represent “the tip of the iceberg” of a broad dyslipidemic environment occurring within ovarian cells. To our knowledge, the exploration of lipid ions in healthy human ovarian tissue is the first of its kind and allowed the visualization of particular m/z values enriched in biologically relevant ovarian regions. Several studies have described the lipid changes in human follicular fluid under different conditions, such as the polycystic ovarian syndrome (PCOS) ( Ban et al. 2021a ; Ding et al. 2022 ), endometriosis ( Cordeiro et al. 2015 ), and ovarian aging ( Cordeiro et al. 2018 ). In addition, previous works have reported spatial lipidomic information in human ovarian tissue in the context of cancer ( Arafah et al. 2014 ; Dória et al. 2016 ; Meriaux et al. 2010 ). However, because these studies were conducted on the tumorigenic portion of the tissue (which may not even stem from a steroidogenic cell type), healthy, functional structures (follicles, corpora lutea, etc.) were not studied ( Arafah et al. 2014 ; Dória et al. 2016 ; Meriaux et al. 2010 ), and no study had yet provided a visualization of the human lipidome on ovarian tissue. While our initial findings are based on a single subject (a 26-year-old female), this study has revealed that several lipid ions are abundantly enriched in biologically relevant regions, including the primordial follicle ridge, antral follicles, corpora lutea, ovarian stroma, and blood vessels. The presence of these structures has been unequivocally corroborated by combining these images with histological information. A surprising aspect of the spatial lipidomic approach was the detection of different lipid ions that were highly enriched for each of the antral follicle layers — the theca and the granulosa layers. Differentiation of the theca cell layer has proven to be extremely challenging, even using the most recent approaches, such as single-cell sequencing ( Isola et al. 2024 ; Morris et al. 2022 ). The outlining of these two distinct layers through lipid enrichment suggests a set of metabolically active pathways within each layer that could aid in developing markers for these cellular layers. Finally, as for prior studies in other tissues ( Balluff et al. 2015 ; Scott et al. 2019 ), unbiased segmentation analyses were incorporated to interrogate if lipid abundance can segregate distinct ovarian regions. The use of spatial segmentation analysis, rather than the selection of single ion peaks, helped validate the hypothesis that spectral data can identify lipid signatures by generating segmentation maps to inform functionally relevant areas of the human ovarian tissue. The major limitation of our human lipidome analysis is that it was conducted on a single human sample due to limited access to ovarian tissue from healthy reproductive-age donors without fertility-related diseases and of sufficient size to analyze histological structures and compartments, and not just ovarian cortical samples. As such, our results cannot currently be generalized to the broader population. Further investigation across a range of human samples is needed to validate these findings. In conclusion, our work has unveiled regional lipid enrichment that aligns with functionally distinct regions in both mouse and human ovaries. It has also provided the first evidence that chemical exposures can result in regional-specific ovarian dyslipidemia. Altogether, our research underscores the vulnerability of the ovarian lipidome to environmental factors and lays the groundwork for unraveling the molecular pathways underlying ovarian toxicity mediated through lipid dysregulation.

Infertility affects over 50 million couples worldwide ( Arafah et al. 2014 ; Dória et al. 2016 ; Mascarenhas et al. 2012 ; Meriaux et al. 2010 ), with half of the cases attributed to female-related factors and the majority of those of unknown etiology. Higher risk of reproductive disorders ( Brewer and Balen 2010 ; Broughton and Moley 2017 ; Saucedo et al. 2021 ; Seif et al. 2015 ; Silvestris et al. 2018 ) and infertility ( Bakeer et al. 2018 ; Cai et al. 2021 ; Kumar et al. 2017 ; Li et al. 2018 ; Liu et al. 2021 ; Schisterman et al. 2014 ; Wang et al. 2020 ; Zhang et al. 2018 ) is associated with obesity and dyslipidemia, both highly prevalent in U.S. women (39.7% ( CDC 2020 ) and 20% ( Carroll and Fryar 2020 ), respectively). Prevalence of dyslipidemia is also high in ovarian disorders ( Kim and Choi 2013 ; Kumar et al. 2017 ; Ollila et al. 2016 ; Zhang et al. 2018 ), and is negatively associated with embryo quality ( Liu et al. 2021 ; Schisterman et al. 2014 ; Wang et al. 2020 ), implantation rate, and lower live birth rates ( Cai et al. 2021 ; Li et al. 2018 ). Supportive of dyslipidemia, aberrant lipid accumulation in steroidogenic cells is linked to poor pregnancy outcomes ( Raviv et al. 2020 ) supportive of dyslipidemia as a player in ovarian dysfunction ( Liu et al. 2020 ; Lolicato et al. 2015 ; Marei et al. 2019 ; O’Reilly et al. 2017 ; Raviv et al. 2020 ; Wang et al. 2020 ; Yang et al. 2012 ). On the other hand, environmental chemical exposures have been reported to increase adipogenesis and adipose tissue accumulation ( Amato et al. 2021 ; Chamorro-Garcia and Veiga-Lopez 2021 ; Egusquiza and Blumberg 2020 ; Ren et al. 2020 ; Ribeiro et al. 2020 ; Veiga-Lopez et al. 2018 ) and have been associated with dyslipidemia in humans ( Averina et al. 2021 ; Kang et al. 2021 ; Lee et al. 2011 ; van der Meer et al. 2020 ; Zhou et al. 2016 ) and animal models ( Afolabi et al. 2015 ; Goudarzi et al. 2015 ; Schlezinger et al. 2021 ). Notably, these chemicals can also interfere with pathways controlling lipid metabolism ( Amato et al. 2021 ; Egusquiza and Blumberg 2020 ; Ren et al. 2020 ; Veiga-Lopez et al. 2018 ) and induce lipid accumulation in non-adipose tissues ( Bertuloso et al. 2015 ; Brulport et al. 2017 ; Carnevali et al. 2017 ; Martella et al. 2016 ; Santangeli et al. 2018 ; Wang et al. 2022 ). More recently, we and others have also demonstrated that environmental exposures can result in ovarian steroidogenic cells (granulosa and theca) dyslipidemia ( Pascuali et al. 2024b ; Wang et al. 2023 ). Despite the crucial role of lipids on ovarian function ( Kim and Choi 2013 ; Kumar et al. 2017 ; Liu et al. 2021 ; Ollila et al. 2016 ; Schisterman et al. 2014 ; Wang et al. 2020 ; Zhang et al. 2018 ), knowledge regarding the ovarian lipidome, including abundance, distribution, function, or factors that can destabilize it, remains in its infancy. Lipids account for ~40% of cells and ~50% of cell membranes ( Ingolfsson et al. 2014 ). Yet, we still do not have a fundamental understanding of the biology associated with lipid abundance within ovarian regions, such as the cortex (where the follicular reserve resides), the hilum (access of ovarian vessels to the ovary), the follicles (nurturing homes for germ cells), or the female germ cells. With recent advances in single-cell transcriptomics and spatial transcriptomic technologies, knowledge of the ovarian transcriptional landscape is making significant headway in our understanding of gene and protein regulation. However, fundamental understanding of ovarian lipidomics (with or without a spatial context), is extremely scarce due to lack of techniques available to visualize them compared to proteins or genes ( Ryan et al. 2019 ; Vathiotis et al. 2021 ). Matrix-assisted laser desorption ionization mass spectrometry imaging (MALDI MSI) has become an increasingly powerful tool over the last few years ( Alexandrov 2023 ; Tobias and Hummon 2020 ). Its label-free detection of diverse biomolecules, including metabolites, proteins, glycans, and lipids, in a tissue sample allows for detecting of both the composition and spatial distribution of different targets, providing invaluable information into biological systems. A handful of studies on mammalian species (mouse, cow, pig) has provided evidence that lipid species have distinct regional patterns in the ovary ( Bertevello et al. 2020 ; Bertevello et al. 2018 ; Campbell et al. 2012 ; Cordeiro et al. 2020 ; Uzbekova et al. 2015 ). However, thus far, low image resolution and lack of accurate lipid assignments have prevented the understanding of their specific location or role within each ovarian region ( e.g., follicles in mouse ovaries). In addition, information regarding human ovarian lipidome in a human context is only available in the context of pathology (e.g.: ovarian cancer) ( Arafah et al. 2014 ; Dória et al. 2016 ; Meriaux et al. 2010 ). In this work, we investigated the ovarian lipidome in both mice and humans. First, using > 180 ovaries, we characterized the mouse ovarian lipidome by developing an in-house lipid library via untargeted lipidomics. Next, we leveraged this lipid library (> 800 lipid species) to inform assignments from MALDI MSI spatial lipidomics of mouse organotypic ovarian cultures exposed to an environmental exposure paradigm, namely the obesogenic chemical tributyltin (TBT). TBT is one of the most commonly used organometallic compounds worldwide (> 100 million lbs/year) ( Cole et al. 2015 ; WHO 2006 ), and its ability to induce adipogenesis in vitro and adipose tissue accumulation in vivo ( Kassotis et al. 2021 ; Ticiani et al. 2023 ; Veiga-Lopez et al. 2018 ). Exposure to TBT leads to reproductive dysfunction in non-mammalian species mammalian ( a Marca Pereira et al. 2014 ; Cangialosi et al. 2010 ; Horiguchi 2006 ; Li et al. 1997 ; Peranandam et al. 2014 ; Rossato et al. 2016 ; Scheider et al. 2018 ; Shi et al. 2014 ; Thibaut and Porte 2004 ; Titley-O’Neal et al. 2013 ; Xiao et al. 2018 ; Zhang et al. 2013 ; Zheng et al. 2005 ), as well as in rats and mice ( de Araujo et al. 2018 ; Podratz et al. 2012 ; Si et al. 2012 ; Yang et al. 2019 ). Postnatal TBT exposure (doses ranging 10 – 500 ng/kg body weight (BW) for 10 – 16 days) has been shown to disrupt reproductive cyclicity, ovarian hormones (lower estradiol, higher testosterone), and follicular development (higher number of atretic and cystic follicles, fewer corpora lutea) ( de Araujo et al. 2018 ; Podratz et al. 2012 ; Si et al. 2012 ; Yang et al. 2019 ). TBT has also been reported to interfere with the hypothalamic-pituitary axis ( Merlo et al. 2016 ; Sena et al. 2017 ), which could indirectly affect ovarian function. Coupling MALDI MSI with this exposure paradigm has allowed us to define discrete functional ovarian regions segmented by lipid signatures and discover regions most sensitive to lipid dysregulation induced by environmental exposure. To evaluate if human ovaries also have lipid enrichment and distribution that would match functional regions of the ovary and thus potentially be susceptible to environmental exposures, we next conducted MALDI MSI spatial lipidomics in human ovarian tissue and demonstrated that lipid signatures conform to spatial regionality corresponding with ovarian structures (e.g., stroma, follicles, and oocytes). To our knowledge, this is the first study to report spatial lipidomics in a healthy human ovary and that functional regions of the ovary (e.g., stroma, granulosa layer, theca layer) are enriched for specific lipid species. In addition, our work is the first to document that ovarian lipids are susceptible to dysregulation upon environmental exposures.