Progesterone induces steroidome and lipidome remodeling in a BRCA-resembling mouse model of high-grade serous ovarian cancer.

Elisabeth M. Schwiebert: Writing – review & editing, Writing – original draft, Validation, Software, Methodology, Investigation. Albert Vega-Herrera: Investigation, Formal analysis, Data curation. Andro Botros: Investigation, Formal analysis, Data curation, Conceptualization. Samuel G. Moore: Investigation, Formal analysis, Conceptualization. David A. Gaul: Writing – review & editing, Visualization, Validation, Formal analysis, Data curation. Olga Kim: Methodology, Formal analysis, Conceptualization. Jaeyeon Kim: Writing – review & editing, Project administration, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Facundo M. Fernández: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization.

The animal study was approved by the IACUC at Indiana University School of Medicine (Indianapolis, IN, United States). The study was conducted in accordance with the local legislation and institutional requirements.

To establish lipidome phenotypes of treated BRCA-resembling DKO mice, all sixty serum samples from the four treatment groups, P4, P4 placebo, mifepristone, and mifepristone placebo, individually underwent LC-MS profiling. An initial examination of the dataset, including PCA and PLS-DA, was conducted on the entire dataset, including both annotated and unannotated features from both positive and negative ion modes to visualize potential differences among the treatment groups. The results of the PCA and PLS-DA models are shown in Fig. 3 . Fig. 3 Partial least squares discriminant analysis of multiple treatments on DKO mice. (A) The 2-D scores plot from principal component analysis (PCA) shows the four treatment groups: P4, P4 placebo, mifepristone (abbreviated RU486 in the figure), and mifepristone placebo, when applied to the BRCA-resembling DKO mice ( n = 15). (B) A subsequent partial least squares discriminant analysis (PLS-DA) of the same samples was also performed. (C) The PLS-DA variable importance in projection (VIP) plot highlights the twelve most distinguishing lipids in the PLS-DA model across the treatment groups. The data were median-normalized and Pareto-scaled before building the model, which was cross-validated (CV) using 5-fold CV with respect to the Q2 performance metric, with a maximum value of 0.4 for the fourth component. This Figure. was created using the ‘Statistical Analysis [one factor]’ module from MetaboAnalyst v6.0[24]. Fig 3 dummy alt text Partial least squares discriminant analysis of multiple treatments on DKO mice. (A) The 2-D scores plot from principal component analysis (PCA) shows the four treatment groups: P4, P4 placebo, mifepristone (abbreviated RU486 in the figure), and mifepristone placebo, when applied to the BRCA-resembling DKO mice ( n = 15). (B) A subsequent partial least squares discriminant analysis (PLS-DA) of the same samples was also performed. (C) The PLS-DA variable importance in projection (VIP) plot highlights the twelve most distinguishing lipids in the PLS-DA model across the treatment groups. The data were median-normalized and Pareto-scaled before building the model, which was cross-validated (CV) using 5-fold CV with respect to the Q2 performance metric, with a maximum value of 0.4 for the fourth component. This Figure. was created using the ‘Statistical Analysis [one factor]’ module from MetaboAnalyst v6.0[24]. Substantial overlap was observed among all treatment groups in the PCA model ( Fig. 3 A), whereas more distinct clustering emerged in the PLS-DA model ( Fig. 3 B). The more notable trends in the PLS-DA results include the distinct clustering of serum samples from P4-treated mice and the broader overlap between DKO mice receiving mifepristone and those receiving the mifepristone placebo. These trends were considered biologically relevant, as P4 treatment accelerated the development of HGSC, resulting in a more differentiated lipidome. Additionally, both mifepristone- and mifepristone placebo-treated mice had intact reproductive tracts, resulting in greater metabolic similarity. Variable importance in projection (VIP) scores ( Fig. 3 C) were used to identify which metabolites or lipid species contributed to differences among the four treatment groups. The first twelve lipids with the highest VIP scores were annotated and included five sphingomyelins (SM), five phosphatidylcholines (PC), one lysophosphatidylcholine (LPC), each with different fatty acid chain lengths and degrees of unsaturation, and one steroid. To further investigate the temporal trends in treatment-dependent metabolomic alterations, additional PLS-DA models were examined at the one-week, three-week, and three-month time points, as shown in Figure. S1 . When temporally separated, the P4-treated mice have progressively distinct clustering patterns correlating with accelerated HGSC development, while the three remaining treatment groups, P4 placebo, mifepristone, and mifepristone placebo, consistently have some degree of overlap. The twelve most differentiating lipids in each VIP plot also agree with those shown in Fig. 3 C, with many PCs, SMs, and LPCs represented along with cholesterol. Their weights in the PLS-DA model prompted us to conduct a more detailed investigation of the lipidomic profiles of the four treatment groups to identify relationships among lipid classes and trends across classes. Most of the twelve differentiating lipids in Figs. 3 C and S1I exhibited higher, circulating concentrations in P4-treated DKO mice compared to those treated with mifepristone. Anatomical examination of reproductive tissues ( Fig. 4 A and B) showed that three-month treatments with P4 and mifepristone, respectively, accelerated or inhibited the growth of HGSC. These images provided substantial evidence of significant treatment-dependent differences in HGSC growth, particularly during early tumor development. Interestingly, the differences at the anatomical level showed concurrent trends at the lipidome level ( Fig. 4 C–E), with differences between P4 and mifepristone treatments becoming more pronounced as the disease progressed. Additional volcano plots comparing P4-treated to P4 placebo mice and mifepristone to mifepristone placebo mice are presented in Figure. S2. Lipids noted with colored symbols are grouped by lipid subclasses, with their level 2 annotations given in Table S7 and respective abbreviations listed in Table S8 . Unannotated lipids were plotted alongside annotated lipids using grey symbols. The large number of unannotated lipids suggests that considerable information remains to be uncovered as lipidomics technology advances, enabling examination of this fraction of the lipidome. Fig. 4 Acceleration and inhibition of HGSC tumor phenotype in the BRCA-resembling DKO mouse model. (A) Following a bilateral ovariectomy at 5–5.3 weeks of age (premalignant stage), DKO mice implanted subcutaneously with a pellet of P4 (25 mg/3 months) developed ET HGSC from the fallopian tube at 4.2 months of age. (B) Premalignant DKO mice aged 4.4–5.4 weeks (no ovariectomy), implanted with a pellet of mifepristone (abbreviated RU in the figure) (9 mg/3 months), formed cystic fallopian tubes with intact ovaries at 4.2 months of age. (C-E) Changes in metabolites/lipids observed between P4- and mifepristone-treated mice at 1 W (C), 3 W (D), and 3 M (E). The lipid subclass name corresponding to each abbreviation can be found in Table S8. An additional image depicting the effects of the mifepristone placebo treatment is available in Figure. S3. Ut, uterus. Ov, ovary. Ft, fallopian tube or Ft cyst. RU, RU486. Fig 4 dummy alt text Acceleration and inhibition of HGSC tumor phenotype in the BRCA-resembling DKO mouse model. (A) Following a bilateral ovariectomy at 5–5.3 weeks of age (premalignant stage), DKO mice implanted subcutaneously with a pellet of P4 (25 mg/3 months) developed ET HGSC from the fallopian tube at 4.2 months of age. (B) Premalignant DKO mice aged 4.4–5.4 weeks (no ovariectomy), implanted with a pellet of mifepristone (abbreviated RU in the figure) (9 mg/3 months), formed cystic fallopian tubes with intact ovaries at 4.2 months of age. (C-E) Changes in metabolites/lipids observed between P4- and mifepristone-treated mice at 1 W (C), 3 W (D), and 3 M (E). The lipid subclass name corresponding to each abbreviation can be found in Table S8. An additional image depicting the effects of the mifepristone placebo treatment is available in Figure. S3. Ut, uterus. Ov, ovary. Ft, fallopian tube or Ft cyst. RU, RU486. Two lipids of interest, previously linked to OC development, tetrahydrocortisone ( p -value: 0.0240, log2FC value: 2.127) and LPE(18:1) ( p -value: 0.0577, log2FC value: 2.273), were found to be statistically or close to statistically significantly elevated in P4-treated mice at the three-week mark. Three lipid subclasses, PC, SM, and TG, showed distinctly elevated levels at the premalignant time point of three weeks and further increased at three months during early-stage HGSC. These lipid subclasses were especially noteworthy because of their known dysregulation during tumor development and the prominence of multiple PC and SM in the PLS-DA VIP scores plot ( Fig. 3 C). Since lipid function can vary with fatty acid chain length and degree of unsaturation, these three lipid subclasses were further examined in more detail. After analyzing the PLS-DA model results ( Fig. 3 B and C) alongside the biological confirmation of accelerated and suppressed HGSC development resulting from P4 and mifepristone treatments, respectively ( Fig. 4 ), the trends of three lipid subclasses, PC, SM, and TG, were analyzed to further evaluate their potential associations with treatment-driven differences in tumor progression. Individual analyses of these three subclasses were performed using volcano plots, where lipids with positive FC indicate higher abundance in P4-treated mice and correspond to the observed acceleration of HGSC development. Notably, PC lipids were elevated during the pre-malignant and early stages of HGSC development at the three-week and three-month intervals, respectively. Among the 119 annotated PC species, there was a wide range of fatty acid chain lengths and degrees of unsaturation. Considering these characteristics can influence lipid biological function, the FC in individual PC lipids between the P4 and mifepristone treatment groups were examined, as shown in Figure. S4 . As shown in the volcano plots in Figure. S4 (A-C) , the increase in FC among PC lipids aligned with extended P4 treatment and the subsequent appearance of early-stage HGSC development. Several individual PC lipids were statistically significant ( p -value < 0.05) with substantial FC (either log2FC 1, demonstrating significant negative and positive FC, respectively), and the number of significant PC grew as the cancer progressed. At the premalignant, one-week mark, two PC species were significant: PC(O-38:2) ( p -value: 0.0144, log2FC value:1.862) and PC(42:9) ( p -value: 0.0273, log2FC value: 1.163). At the premalignant stage of three weeks, three PC lipids were significant, all showing positive FC: PC(38:4) ( p -value: 0.00227, log2FC value: 1.098), PC(40:7) ( p -value: 0.00380, log2FC value: 1.182), and PC(44:12) ( p -value: 0.00564, log2FC value: 1.158). During early HGSC development at three months, fourteen PC lipids were significant, all with positive FC: PC(30:1) ( p -value: 0.0267, log2FC value: 1.052), PC(32:2) ( p -value: 0.0165, log2FC value: 1.150), PC(32:3) ( p -value: 0.0284, log2FC value: 1.688), PC(34:4) ( p -value: 0.0477, log2FC value: 1.012), PC(35:0) ( p -value: 0.0117, log2FC value: 1.207), PC(35:2) ( p -value: 0.00999, log2FC value: 1.384), PC(36:0) ( p -value: 0.0143, log2FC value: 1.189), PC(36:6) ( p -value: 0.00.345, log2FC value: 1.391), PC(38:0) ( p -value: 0.0311, log2FC value: 1.013), PC(38:4), PC(38:7) ( p -value: 0.00345, log2FC value: 1.202), PC(41:3) ( p -value: 0.0415, log2FC value: 1.042), and PC(44:12) ( p -value: 0.0286, log2FC value: 1.179). The FC for most of the annotated PC species appeared to increase with prolonged P4 treatment and cancer development. Additionally, a moderate rise in PC lipids with fatty acid chains longer than 38 carbons containing several double bonds was observed at the time point before cancer development ( Figure. S4E ). Comparisons across DKO treatment groups and time points for fatty acid chain length and degree of unsaturation in various PC lipids are shown in Figure. S5 . Overall, PC lipids showed a progressive, stage-dependent rise with P4 treatment, with only a few changes emerging at premalignant time points, but a broad and robust elevation of diverse PC species appeared by three months, consistent with a transition from early metabolic dysregulation to overt HGSC development. Among the annotated SM lipids, a similar trend of correlated elevation with P4 treatment was noted starting at the one-week mark. Although only 35 SM species were annotated, their consistent association with accelerated HGSC development was particularly noteworthy. The FCin these lipids between the P4 and mifepristone treatments were examined using volcano plot analyses, as shown in Figure. S6 . Similar to previous observations among the PC subclass, several SM species showed statistical significance and substantial FC at each time point, with the number of significant lipids increasing after HGSC tumors have developed. At the premalignant, one-week mark, three SM lipids were significant, all showing positive FC: SM(d30:1) ( p -value: 0.0162, log2FC value: 1.114), SM(d32:2) ( p -value: 0.00176, log2FC value: 1.017), and SM(d40:1-OH) ( p -value: 0.00383, log2FC value: 1.032). At the premalignant stage of three weeks, two SM species were significant, both with positive FC: SM(d42:0) ( p -value: 0.00458, log2FC value: 1.155) and SM(d42:2) ( p -value: 0.0117, log2FC value: 1.002). By early HGSC development at three months, twelve SM lipids were significant, all exhibiting positive FC: SM(d30:1) ( p -value: 0.0121, log2FC value: 1.195), SM(d32:1) ( p -value: 0.0152, log2FC value: 1.259), SM(d32:2) ( p -value: 0.0267, log2FC value: 1.108), SM(d35:2) ( p -value: 0.0129, log2FC value: 1.315), SM(d34:2-OH) ( p -value: 0.0278, log2FC value: 1.101), SM(d35:1) ( p -value: 0.0261, log2FC value: 1.056), SM(d36:3) ( p -value: 0.0148, log2FC value: 1.122), SM(d36:1) ( p -value: 0.0299, log2FC value: 1.034), SM(d41:1) ( p -value: 0.0268, log2FC value: 1.110), SM(d42:0) ( p -value: 0.0477, log2FC value: 1.138), SM(d42:1) ( p -value: 0.0338, log2FC value: 1.045), SM(d43:1) ( p -value: 0.0353, log2FC value: 1.045), SM(d43:2) ( p -value: 0.0288, log2FC value: 1.026), and SM(d43:3) ( p -value: 0.0254, log2FC value: 1.037). A moderate increase in FC was noted among SM species with fatty acid chain lengths exceeding thirty-nine carbons and possessing only one or two points of unsaturation at the three-week time point, as illustrated in Figure. S6E . The lengths of fatty acid chains and the degrees of unsaturation of all detected SM lipids were also evaluated across all DKO treatment groups and time points, and are shown in Figure. S7 . Similar to the PC lipid subclass, the SM lipids exhibited modest early increases at one and three weeks but expanded dramatically in number and magnitude of significant species by three months, indicating that SM accumulation becomes increasingly pronounced as HGSC advances from premalignant stages to early-stage malignancy. Since elevated TG are typically a hallmark of energy dysregulation during the development of cancerous tumors [31] , the observation of elevated TG lipids at only the three-month time point was unexpected. The distribution of the fifty-five annotated TG species remained relatively consistent at both the one-week and three-week time points, with significant differentiation evident only at the early stages of HGSC development, as shown in Figure. S8 . Although no TG lipids were of significance at the one-week or three-week intervals, several specific TG species showed statistical significance at the early stages of tumor development. At the early-stage HGSC development three-month mark, three TG lipids were significantly noteworthy: TG(50:6) ( p -value: 0.0451, log2FC value: 1.040), TG(56:10) ( p -value: 0.0230, log2FC value: 1.204), and TG(58:11) ( p -value: 0.0192, log2FC value: 1.114) all exhibited positive FC. However, as depicted in Figure. S8F , TG levels with fatty acid chains totaling greater than fifty-five carbons in length and degrees of unsaturation of greater than or equal to nine were markedly elevated at the three-month point, correlating with the development of HGSC. All detected TG lipids, sorted by fatty acid chain length and degree of unsaturation, were examined across all DKO treatment groups and time points and are shown in Figure. S9 . In contrast to PC and SM subclasses, the TG lipids showed little to no alteration during the premalignant one- and three-week stages but became distinctly elevated, and highly unsaturated only at the three-month early HGSC time point, suggesting that TG dysregulation emerges later and aligns more specifically with established tumor development rather than early premalignant conditions. Among the most differentiating annotated lipids between the P4- and mifepristone-treated mice, a direct comparison of metabolite intensities was performed to evaluate the extent to which P4-associated changes were reversed, partially reversed, or unaffected by treatment with the competitive antagonist mifepristone. Several species were found to be reversed or partially reversed at the one-week, three-week, or three-month time points, as shown in Figure. S10 . To enable this comparison, raw metabolite intensities were median-normalized within each sample and subsequently log2-transformed. Group-level means were then calculated across the four treatment conditions (P4, P4 placebo, mifepristone, and mifepristone placebo), and differential abundance between P4 and its placebo was assessed using Welch’s two-sample t -test. Metabolites exceeding predefined thresholds for effect size (log2FC 〈 -1 or log2FC 〉 1) and statistical significance ( p -value < 0.05) were retained for further analysis. A reversal index was subsequently computed to quantify the extent to which mifepristone counteracted P4-induced changes, defined as the proportional shift of the mifepristone-treated mean toward the P4 placebo baseline related to the P4-induced deviation. Based on this metric, metabolites were classified as reversed, partially reversed, or unaffected, with trajectory plots of group means enabling the direct visualization and interpretation of these reversal patterns. While the SQUAD analysis included the concurrent collection of targeted, steroidomic, and non-targeted lipidomic datasets, the nine targeted steroids were not consistently and reliably detected across the serum samples from all sixty DKO mice. This inconsistency likely reflects the limited sensitivity of the mass spectrometric platform used, as the shorter duty cycle and acquisition settings were optimized to detect a broad range of lipid species rather than the low endogenous concentrations of circulating steroids within the complex matrix. An alternative steroidomic platform was therefore developed in parallel, using a more sensitive triple-quadrupole mass spectrometer, yielding excellent results. The levels of fifteen steroid hormones related to P4 and the steroid biosynthetic pathway were quantitatively measured in DKO mouse serum. The individual levels of the detected steroids across different treatments and time points (one week, three weeks, and three months) are shown in Figure. S11 . Moreover, the metabolic changes observed among the detected steroids were further investigated under P4 and mifepristone treatments. Pathway analysis, followed by enrichment analysis, was carried out between one week and three weeks, corresponding to the premalignant HGSC microenvironment, and between three weeks and three months, corresponding to the early stages of HSGC development. The results of the applied DID linear logistic regression model for steroid concentrations demonstrate the causal impact of both P4- and mifepristone treatments over time by comparing them with their corresponding placebos ( Tables S9 & S10 ). Such pathway enrichment analyses have the potential to reveal broader metabolic reprogramming during cancer development and patterns that cannot be detected by single-metabolite comparisons alone, helping pinpoint actionable biological pathways. Fig. 5 , Fig. 6 summarize the metabolic pathways that were found statistically significantly enriched among the P4- and mifepristone-treated mice. These pathways reveal early activation of E1/E2 and androgen-related pathways under P4 exposure and, conversely, suppression of these pathways with mifepristone treatment, thereby highlighting coordinated pathway-level shifts that accompanied changes in individual metabolite abundances between one week and three weeks, and between three weeks and three months. Fig. 5 Steroid biosynthetic pathway enrichment analysis related to P4 treatment. Steroid metabolic changes were profiled for P4-treated DKO mice during the premalignancy (blue arrows, one week to three weeks) and early stages of HGSC development (yellow arrows, three weeks to three months). The changes were identified using DID linear logistic regression. (A) An integrated network of metabolic pathways and enrichment analyses was conducted for (B) one week to three weeks and (C) three weeks to three months through MetaboAnalyst v6.0 [ 24 ]. Fig 5 dummy alt text Fig. 6 Steroid biosynthetic pathway enrichment analysis related to mifepristone treatment. Steroid metabolic changes were profiled for mifepristone (abbreviated RU486 in the figure)-treated DKO mice during the premalignancy (blue arrows, one week to three weeks) and early stages of HGSC development (yellow arrows, three weeks to three months). The changes were identified using DID linear logistic regression. (A) An integrated network of metabolic pathways and enrichment analyses was conducted for (B) one week to three weeks and (C) three weeks to three months through MetaboAnalyst v6.0 [ 24 ]. Fig 6 dummy alt text Steroid biosynthetic pathway enrichment analysis related to P4 treatment. Steroid metabolic changes were profiled for P4-treated DKO mice during the premalignancy (blue arrows, one week to three weeks) and early stages of HGSC development (yellow arrows, three weeks to three months). The changes were identified using DID linear logistic regression. (A) An integrated network of metabolic pathways and enrichment analyses was conducted for (B) one week to three weeks and (C) three weeks to three months through MetaboAnalyst v6.0 [ 24 ]. Steroid biosynthetic pathway enrichment analysis related to mifepristone treatment. Steroid metabolic changes were profiled for mifepristone (abbreviated RU486 in the figure)-treated DKO mice during the premalignancy (blue arrows, one week to three weeks) and early stages of HGSC development (yellow arrows, three weeks to three months). The changes were identified using DID linear logistic regression. (A) An integrated network of metabolic pathways and enrichment analyses was conducted for (B) one week to three weeks and (C) three weeks to three months through MetaboAnalyst v6.0 [ 24 ].

This study identifies early metabolic signatures that precede high-grade serous ovarian cancer (HGSC) development in a double knockout mouse model resembling BRCA 1/2 mutations, with direct implications for cancer risk assessment and prevention. Progestin-driven remodeling of key lipid subclasses, including phosphatidylcholines, sphingomyelins, and triacylglycerols, emerges as an early and consistent feature of premalignant progression, supporting their potential as circulating biomarkers of metabolic reprogramming in BRCA1/2 mutation carriers. Simultaneously, steroidomic profiling reveals coordinated changes in progesterone, glucocorticoids, androgens, and estrogens, highlighting the conversion of progesterone to corticosterone as a key factor in early-stage ovarian cancer risk. Notably, anti-progestin treatment reduces pro-oncogenic steroid signaling while promoting aromatase-mediated estrogen production, which is linked to a decreased risk of early HGSC. Overall, these findings provide a mechanistic framework for integrating lipidomic and steroidomic biomarkers into early detection efforts and support the development of hormone-modulating interventions to personalize ovarian cancer prevention and management in genetically high-risk populations.

Liquid chromatography-mass spectrometry (LC-MS)-grade 2-isopropanol (IPA) was purchased from Fisher Chemical (Fisher Scientific, Pittsburgh, PA, USA) and used for sample preparation. LC-MS-grade methanol, water, acetonitrile (ACN), formic acid, ammonium formate (99.9% purity or higher), and ammonium hydroxide (99.5% purity or higher) were purchased from Fisher Chemical and used to prepare chromatographic mobile phases. Isotopically labeled steroid standards for P4, corticosterone, hydrocortisone, and estrone (E1) were purchased from Cayman Chemical (Ann Arbor, MI, USA). Labeled standards for cortisone and 17α-hydroxyprogesterone were purchased from Sigma Aldrich (St. Louis, MO, USA). Labeled standards for testosterone and androstenedione were purchased from Cerilliant-Certified Reference Materials (Round Rock, TX, USA). The labeled standard for E2 was purchased from MedChemExpress (Monmouth Junction, NJ, USA). All purchased standards were used to prepare the internal standard mixture. Unlabeled steroid standards for P4, hydrocortisone, cortisone, corticosterone, 11-deoxycortisol, dehyroepiandrosterone (DHEA), 17α-hydroxypregnenolone, 17α,20β-dihydroxy-4-prenen-3-one, 11β-hydroxyprogesterone, 5β-tetrahydrocortisol, 5α-dihydroprogesterone, E2, and E1 were acquired from Cayman Chemical (Ann Arbor, MI, USA). Unlabeled standards for testosterone and androstenedione were purchased from Cerilliant-Certified Reference Materials (Round Rock, TX, USA). For this study, a BRCA-resembling Dicer1 flox/flox Pten flox/flox Amhr2 cre/+ DKO mouse model that develops HGSC tumors with histological, molecular, and phenotypical similarity to human HGSC [ 22 , 23 ] underwent multiple treatments with cross-sectional sera collections at the one-week, three-week, and three-month time points as detailed in Table S1 . Serum samples from these mice were collected with the approval of the Indiana University School of Medicine's Institutional Animal Care and Use Committee (IACUC). The study involved sixty DKO mice, divided into four treatment groups: P4, a P4 placebo, mifepristone, and a mifepristone placebo. Each group, consisting of fifteen mice, underwent an internal time-course study designed to simulate two critical stages of cancer development by varying treatment durations. These durations were 1 week, 3 weeks, and 3 months, corresponding to two time points prior to tumor development (Pre-T) and one time point at the early stages of tumor development (ET), respectively. The stages of HGSC development at these time points were determined by the confirmed presence or absence of HGSC tumors at the time of sacrifice, as verified by surgical investigation and comparative analysis to the same types of developed tumors using the same DKO model in a previous study [20] . Fig. 1 provides an overview of the DKO mouse treatment groups, sample preparation steps, and both the workflows for the lipidomics and steroidomics analyses employed in this study. Fig. 1 Workflow for sample collection and mouse serum steroidomics and lipidomics. (A) Sixty DKO mice were divided into four treatment groups: P4 (ovariectomized), a P4 placebo (ovariectomized), mifepristone, and a mifepristone placebo. These groups were further categorized based on a time-course study of one week, three weeks, and three months, which represent two Pre-T time points and one ET time point. (B) Overview of the IPA-based liquid extraction non-targeted lipidomic sample preparation process and SQUAD instrumental analysis. (C) Overview of the solid phase extraction (SPE) sample preparation process and triple quadrupole analysis. (D) Collection of software platforms used for the non-targeted data processing, including both Xcalibur™ v4.3 and Compound Discoverer™ v3.3 from ThermoFisher Scientific. (E) Both Waters MassLynx™ software and MetaboAnalyst v6.0[24] were used during the targeted data processing workflow. Additional details on the Orbitrap SQUAD analysis and triple quadrupole are shown in Fig. 2 . This Figure. was created with Biorender.com. Fig 1 dummy alt text Workflow for sample collection and mouse serum steroidomics and lipidomics. (A) Sixty DKO mice were divided into four treatment groups: P4 (ovariectomized), a P4 placebo (ovariectomized), mifepristone, and a mifepristone placebo. These groups were further categorized based on a time-course study of one week, three weeks, and three months, which represent two Pre-T time points and one ET time point. (B) Overview of the IPA-based liquid extraction non-targeted lipidomic sample preparation process and SQUAD instrumental analysis. (C) Overview of the solid phase extraction (SPE) sample preparation process and triple quadrupole analysis. (D) Collection of software platforms used for the non-targeted data processing, including both Xcalibur™ v4.3 and Compound Discoverer™ v3.3 from ThermoFisher Scientific. (E) Both Waters MassLynx™ software and MetaboAnalyst v6.0[24] were used during the targeted data processing workflow. Additional details on the Orbitrap SQUAD analysis and triple quadrupole are shown in Fig. 2 . This Figure. was created with Biorender.com. The mice in the P4 treatment group were subcutaneously implanted with a P4 pellet (25mg/90 days/mouse) following ovariectomy at 5–6 weeks of age. As a control, the P4 placebo treatment group of DKO ovariectomized mice received a placebo pellet (one without hormone). To investigate whether mifepristone inhibits HGSC development in DKO mice by blocking P4 receptors (PR), DKO mice with intact ovaries were implanted once with a mifepristone pellet (9 mg/90 days). For the mifepristone control group, DKO mice with intact ovaries received a placebo pellet. In all the treatments mentioned above, the respective pellets were inserted subcutaneously after a single skin incision (0.5–1.0 cm) on the back near the ovary. The skin incision was closed using either monofilament suture or wound clips. The pellets were sourced from Innovative Research of America (Sarasota, FL, USA). The doses and procedures for all four treatments of the sixty DKO mice followed those of a previous study by Kim et al. [20] . Blood samples were collected in serum separation tubes (BD Microtainer® Tube; 365,967; BD Biosciences, San Jose, CA, USA) from DKO mice or ovariectomized DKO mice, with or without hormone treatment. Serum was isolated by centrifuging the blood at 19,960 × g (14,000 rpm) for 5 min at room temperature. After serum collection, to confirm the relevant stage of HGSC development within the DKO mice at the time points of one week, three weeks, and three months, the mice were humanely sacrificed in accordance with animal protocol #21,124 approved by the IACUC at Indiana University School of Medicine (Indianapolis, IN, USA). The serum samples were then shipped on dry ice and, upon arrival, stored at -80 °C until LC-MS analysis. For SQUAD targeted and non-targeted UHPLC-MS, serum samples were thawed on ice before extraction of the non-polar (lipid) metabolome. In 1.5-mL microcentrifuge tubes, 25 µL of serum were extracted with 75 µL of an IPA stock solution containing all nine labeled internal standards. The samples were vortexed for 15 s and then centrifuged at 21,100 × g (14,800 rpm) for 5 min. The resulting supernatants were transferred into LC vials for LC-MS analysis. Samples were stored at –80 °C until analysis. For targeted triple quadrupole UHPLC-MS/MS steroid quantification, serum samples were thawed, and 100 µL of each sample was collected. For each sample aliquot, 400 µL of cold ACN (- 20 °C) was added, vortexed for 10 s, incubated on ice for 10 min, and then centrifuged at 13,174 × g (12,000 rpm) for 10 min to solubilize steroid hormones and precipitate serum proteins. The resulting supernatants were transferred to 1.5 mL plastic Eppendorf tubes containing a 1:1 ACN:water solution, and solid-phase extraction (SPE) was performed under vacuum using reverse-phase OASIS HLB 1 mL 30 mg SPE cartridges purchased from Waters Corporation (Milford, MA, USA). SPE extracts of ACN were collected in LC vials, evaporated under a N 2 stream until dry, and reconstituted in 0.5 mL of 1:1 ACN:water. Final extracts were stored at -20 °C until analysis. For SQUAD steroid targeted quantitation, stock solutions of each of the nine labeled steroid internal standards were prepared in IPA before thawing the DKO serum samples for analysis. A combined stock solution was prepared, containing all nine internal standards at a final concentration of 0.075 nM in IPA, as detailed in Table S2 . These stock solutions were then used to prepare a series of calibrators with concentrations ranging from 0.01 nM to 1000 nM. The complete set of calibrators was injected at the beginning, middle, and end of the batch of LC-MS analyses in both positive and negative ion modes. A blank sample was also prepared using LC-MS-grade IPA and underwent the same preparation process as the other samples. A pooled quality control (QC) sample was created by mixing 3 µL aliquots from each sample extract. This pooled QC was analyzed every 10 LC-MS injections to monitor instrument stability and account for any drift that may have occurred during the experiment. Samples were randomized prior to LC-MS analysis to account for any potential instrument errors. For targeted triple quadrupole LC-MS/MS steroid quantitation, three procedural blanks were processed in parallel with DKO mouse serum samples to assess potential contamination sources throughout the analytical pipeline. Blank subtraction was applied to the samples’ raw data when necessary. Individual stock standard solutions of 1000 mg·mL −1 for all unlabeled steroid standards described in section 2.1 were prepared and subsequently used to perform a mixed calibration curve over two different ranges: 2 to 1000 ng·L −1 and 1000 to 500,000 ng·L −1 , with R 2 > 0.990 in all cases. The method details and corresponding mass-to-charge ( m/z ) values of the monitored precursor ions for each targeted steroid are listed in Table S3 . Intraday repeatability, matrix effects, and recoveries at the 1000 ng·L −1 ( n = 3) were also assessed. Table S4 lists the quality assurance (QA)/QC validation parameters for each steroid hormone, along with their corresponding limits of detection (LoD) and limits of quantification (LoQ). Before conducting triple quadrupole LC-MS/MS experiments, samples from different treatment groups were randomized to minimize memory effects. The entire mixed calibration curve was run in triplicate at the beginning, middle, and end of the sequence to measure and correct for potential changes in ionization efficiency. For SQUAD lipidomics, reverse phase (RP) LC-MS was performed using a Waters ACQUITY UPLC BEH C18 column measuring 2.1 × 100 mm with 1.7 µm particles, mounted on a Vanquish™ Horizon LC system (ThermoFisher Scientific, Waltham, MA, USA). This chromatography setup was interfaced to an Orbitrap ID-X™ Tribrid™ mass spectrometer (ThermoFisher Scientific, Waltham, MA, USA). The mobile phases used were as follows: for positive ion mode, mobile phase A was water with 0.1% formic acid, and mobile phase B was a mixture of IPA/ACN (80:20 v/v ) with 0.1% formic acid. For negative ion mode, mobile phase A was water with 0.05% ammonium hydroxide, and mobile phase B was IPA/ACN (80:20 v/v ) with 0.05% ammonium hydroxide. Additional details of the LC gradient for the RP SQUAD analysis are provided in Table S5 . During SQUAD experiments, targeted MS data were obtained in parallel to non-targeted data using the dual-pressure linear ion trap analyzer of the Orbitrap ID-X™ Tribrid™ mass spectrometer, employing eighteen m/z isolation windows in both positive and negative ion modes, corresponding to each targeted steroid and its respective labeled standard ( Table S2 ). A heated electrospray ionization (H-ESI) source was used with static voltages of 3800 V and 4000 V for positive and negative ion modes, respectively. For annotation purposes, MS/MS data for these analytes were collected by fragmenting precursor ions via high collision dissociation (HCD) at a normalized collision energy of 40%. Non-targeted MS lipidomics data were acquired via the Orbitrap mass analyzer in both positive and negative ion modes, covering the 150–2000 m/z range. For annotation purposes, MS/MS experiments were performed on the pooled QC sample using the ThermoFisher Scientific AcquireX data acquisition workflow. Precursor ions were fragmented in the HCD cell with a normalized collision energy of 40% and were sequentially fragmented with a collision-induced dissociation of 40%. A visual representation of both targeted and non-targeted analyses conducted on the Orbitrap ID-X™ Tribrid™ mass spectrometer is shown in Fig. 2 A-C. Fig. 2 Targeted and non-targeted data collection via SQUAD Orbitrap and triple quadrupole MS and MS/MS analysis. (A) A schematic of the Orbitrap ID-X™ Tribrid™ mass spectrometer, sourced from Thermo Fisher Scientific, visually depicts the analyzers used for targeted and non-targeted data collection. (B) Targeted SQUAD analysis was conducted in the dual-pressure linear ion trap, incorporating 9 m/z isolation windows. (C) Non-targeted SQUAD analysis was performed in the Orbitrap analyzer. (D) A schematic of the XEVO TQ Absolute mass spectrometer, adapted from a similar Figure. provided by Waters Corporation, was used to collect targeted MS/MS data. This Figure. was created using Biorender.com. Fig 2 dummy alt text Targeted and non-targeted data collection via SQUAD Orbitrap and triple quadrupole MS and MS/MS analysis. (A) A schematic of the Orbitrap ID-X™ Tribrid™ mass spectrometer, sourced from Thermo Fisher Scientific, visually depicts the analyzers used for targeted and non-targeted data collection. (B) Targeted SQUAD analysis was conducted in the dual-pressure linear ion trap, incorporating 9 m/z isolation windows. (C) Non-targeted SQUAD analysis was performed in the Orbitrap analyzer. (D) A schematic of the XEVO TQ Absolute mass spectrometer, adapted from a similar Figure. provided by Waters Corporation, was used to collect targeted MS/MS data. This Figure. was created using Biorender.com. Targeted steroid analysis was carried out by means of ultra-high performance liquid chromatography mass spectrometry in tandem (UHPLC-MS/MS). The chromatographic separation was achieved using an Acquity BSM LC system (Waters Corporation, Milford, MA, USA), equipped with a RP column (Acquity Premier BEH C18, measuring 2.1 × 50 mm with a 1.7 µm particle size) working at room temperature and at a flow rate of 0.35 mL·min −1 . For each analytical run, 5 µL of 1:1 ACN:water final extracts were injected, and the chromatographic separation was performed using a binary mobile phase of (A) 100% ACN and (B) 100% water with 0.05% formic acid and 5 mM ammonium formate in positive ion mode and (A) 100% ACN and (B) 100% water in negative ion mode. Additional details of the LC gradient for the RP LC-MS/MS analysis are provided in Table S6 . The LC system was coupled to a XEVO TQ Absolute triple quadrupole mass spectrometer (Waters Corporation, Milford, MA, USA), equipped with an electrospray ionization (ESI) source, operating in either positive (3.33 kV) or negative (2.87 kV) ionization modes as shown in Fig. 2 D The desolvation temperature was set at 500 °C while desolvation gas flow and collision gas flow were set to 1000 L·hr −1 and 0.18 mL·min −1 , respectively. Quantitative data acquisition was performed in multiple reaction monitoring (MRM) mode, establishing two specific precursor-product ion transitions for each targeted analyte. Table S3 lists each targeted steroid and its corresponding quantitation and confirmation transitions. The raw, targeted SQUAD Orbitrap MS data were processed using Thermo Fisher Scientific Xcalibur™ v4.3 via the Qual Browser, Processing Setup, and Quan Browser modules. This involved creating an Xcalibur™ processing method by identifying the analytes' retention times and m/z transitions as listed in Table S2 . Peak areas were determined by integrating the areas under each curve of either the targeted fragment or precursor ion, as listed in Table S2 . These peak areas were then converted to absolute concentrations for the P4 and corticosterone analytes by calculating the ratios of the corresponding 0.075 nM internal standard to the endogenous analyte. The raw, untargeted Orbitrap SQUAD LC-MS data were processed using Thermo Fisher Scientific Compound Discoverer™ v3.3. This process included blank feature filtering, retention time alignment, peak picking, peak area integration, isotopic feature grouping, adduct grouping, and compound area drift correction using the QC-based regression curve. The same Compound Discoverer™ v3.3 software was used to assign annotations referencing several spectral libraries, including mzCloud from Thermo Fisher Scientific and an in-house mzVault library developed for the Molecular Transducers of Physical Activity Consortium [25] . Tandem MS data collection was carried out using Thermo Fisher Scientific’s AcquireX workflow, enabling the annotation of 336 lipids in both positive and negative ion modes. Information on each annotated lipid species is provided in Table S7 . These lipids were annotated using several databases, including the aforementioned mzCloud from Thermo Fisher Scientific and an in-house mzVault library, as well as online reference libraries, such as the Human Metabolome Database (HMDB) [26] , LipidMaps [27] , METLIN [28] , and Metabolomics Workbench [29] . Each annotated metabolite received an assigned annotation level as adapted from the Metabolomics Standards Initiative [30] . The annotation levels are defined as follows: (1) metabolite identification based on m/z or retention time match to an internal standard; (2) metabolite annotation based on characteristic MS/MS fragmentation data; (3) metabolite annotation based on accurate mass match; (4) unknown metabolite. Triple quadrupole targeted UHPLC-MS/MS raw data were auto processed using the TargetLynx module in the MassLynx™ v4.2 software (Waters Corporation, Milford, MA, USA). Two TargetLynx methods were built for positive and negative steroid analytes, respectively. For each analyte, quantification and confirmation transitions, retention time, the type of sample (blank, quality control, standard, or real sample), as well as smoothing and peak integration parameters, were set among others. Then, the raw data was automatically processed and displayed in TargetLynx Browser. The processed dataset was exported and finally analyzed through enrichment and pathway analyses performed in MetaboAnalyst v6.0 [24] . The complementary data analysis of both the targeted steroidomics and non-targeted lipidomics datasets reveals a coordinated metabolic response to P4 signaling, where changes in steroid hormone levels are accompanied by broader lipid remodeling. These datasets provide a more comprehensive view of the biochemical pathways underlying early disease progression and highlight potential metabolite signatures that may be leveraged for improved detection and mechanistic insight, as illustrated by the following results. The non-targeted lipidomic datasets from both positive and negative modes were aligned into a cohesive dataset before being median-normalized and Pareto-scaled for the principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) models via MetaboAnalyst v6.0 [24] . For the volcano plot analyses, the positive and negative mode datasets were kept separate and individually normalized and scaled using the aforementioned methods. Statistical significance was evaluated using an independent two-sample t -test (Welch’s t -test, unequal variance). The resulting p -values were log10-transformed, and the corresponding fold change (FC) values were log2-transformed, both for visualization purposes. To further analyze the datasets based on individual lipids’ fatty acid chains and degrees of unsaturation, the normalized and scaled intensities of specific lipid species were calculated as a ratio between the individual lipid’s intensity versus the summed intensity of the entire lipid subclass. The targeted steroidomic dataset included normalization of metabolite abundances using a probabilistic quotient normalization (PQN) reference sample (standards mix at 50 ng·L-1) prior to log10-transformation and Pareto scaling. A differences-in-differences (DID) linear logistic regression model was applied to the steroid concentrations to estimate the causal impact of both P4- and mifepristone treatments by comparing individual steroid average concentration changes over time between the hormone-treatment groups and their respective placebo groups. Consistent across the statistical analyses of both the non-targeted lipidomic and targeted steroidomic datasets, the thresholds for statistical significance include p-values 〈 0.05 and FC values〉 2 or <- 2, corresponding to increases in opposite treatment groups within a binary comparison.

Our study reveals distinct and early-arising metabolic changes that precede the development of HGSC in a BRCA1/2-resembling mouse model by a P4 treatment sharply elevating key lipid subclasses, PC, SM, and TG, as well as corticosterone and estrogenic levels across both premalignant and early-stage disease. Increased P4 signaling via the progesterone receptor (PR) has been recognized as a key regulator of epithelial metabolic homeostasis. Nevertheless, PR activation has also been shown to reprogram several lipid metabolic pathways, including de novo lipogenesis and fatty acid desaturation through transcriptional regulation of enzymes such as fatty acid synthase (FASN) and stearoyl-CoA desaturase-1 (SCD1), both of them frequently dysregulated in HGSC [ 32 , 33 ]. FASN supports membrane biosynthesis and metabolic adaptation, while SCD1 controls the saturation index of membrane lipids and promotes cell survival under metabolic stress [34] . Collectively, these pathways position lipid metabolism as a downstream effector of steroid hormone signaling in epithelial tissues. These lipid metabolic alterations, such as sphingolipid signaling, directly translate into changes in membrane phospholipid composition, which can regulate the membrane fluidity, curvature, and lipid raft organization. These biophysical properties critically influence receptor clustering and signal transduction efficiency. Thus, an increased membrane fluidity has been associated with an enhanced activation of oncogenic mechanisms, including PI3K/AKT and MAPK signaling, which are fundamental to proliferation and survival in HGSC [35] . In addition, lipid raft reorganization has been implicated in sustained growth factor signaling and malignant progression [36] . In this context, P4-driven maintenance of lipid homeostasis may help stabilize membrane composition and prevent activation of oncogenic signaling in HGSC premalignant stages. In parallel, an increased glucocorticoid signaling can suppress immune surveillance and modulate inflammatory responses, thereby enhancing tumor cell initiation under stress conditions. In the early stages of HGSC, disruption of steroidome profiling further contributes to tumor progression. Reduced P4 signaling favors a hormonal shift toward relatively increased corticosterone and estrogenic activity via E1-to-E2 conversion. Estrogen signaling has been strongly implicated in promoting proliferative and genotoxic stress in fallopian tube epithelium and in the early pathogenesis of HGSC [ 37 , 38 ]. However, estrogens have also been previously shown to attenuate the development of HGSC metastases in the same DKO mouse model [20] and their role still remains unclear. Together, these endocrine alterations may converge to create a permissive stromal–epithelial environment that supports early neoplastic evolution in HGSC. Overall, the results obtained in this study support a model in which P4–PR signaling keeps fallopian tube epithelial homeostasis through coordinated regulation of lipid metabolism, membrane biophysics, and systemic steroid balance. Loss or attenuation of this axis promotes metabolic reprogramming, altered signal transduction, and endocrine imbalance, thereby facilitating early events in the HGSC initiation and progression. Building on these first observations, the analysis of the sixty mice across four treatment groups, P4 (ovariectomized), P4 placebo (ovariectomized), mifepristone, and mifepristone placebo, yielded the annotation of 336 individual lipids and highlighted multiple steroid hormone pathways whose abundances shifted significantly in association with tumor progression. This model recapitulates early HGSC microenvironmental biology with high fidelity and expands previous evidence establishing the fallopian tube as the site of HGSC origin [22] and P4 as a modifier of early tumor initiation [20] . Despite the genetic similarity of the mice [22] , P4 and mifepristone produced opposing biological outcomes: P4 accelerated HGSC development ( Fig. 4 A), whereas mifepristone suppressed it ( Fig. 4 B) by blocking P4 signaling, inhibiting glucocorticoid activity, and diverting androstenedione away from estrogen synthesis. These findings not only align with established literature but also clarify the specific lipid and steroid pathways most closely associated with divergent tumor trajectories in this BRCA-resembling model [20] . Beyond the broader lipidomic and steroidogenic alterations observed with P4 treatment, several individual metabolites also showed early, consistent perturbations that further support the metabolic trajectory toward HGSC development. Among the notable differences, two metabolites, tetrahydrocortisone and LPE(18:1), previously studied in the context of OC development, were found to be statistically or close to statistically significantly elevated at the three-week, pre-cancerous stage across all three treatment group comparisons: P4 vs. P4 placebo ( Figure. S2B ), P4 vs. mifepristone ( Fig. 4 D), and mifepristone vs. mifepristone placebo ( Figure. S2H ). Earlier research by Wang et al . identified altered levels of both tetrahydrocortisone and tetrahydrodeoxycorticosterone when comparing uterine fluid from patients with benign ovarian tumors, early-stage OC, and late-stage OC [39] . The concordance between these prior findings and the elevated tetrahydrocortisone observed in our study suggests that dysregulation of adrenal-derived metabolites may represent early stress-related responses during tumor initiation. Such dysregulation of the hypothalamus-pituitary-adrenal axis has been previously linked to the development of various types of cancer, due to both the psychological stress of a cancer diagnosis and the physical stress associated with these diseases [40] . In addition, LPE(18:1) was markedly elevated in the three-week serum samples from P4-treated mice, detectable in both positive- and negative-ion modes ( Fig. 4 D). Within a broader panel, this specific lipid emerged as a potential biomarker for OC development in pelvic fluid samples, as reported by Tang et al. [41] . Collectively, these metabolites, together with the following shifts in broader lipid and steroid pathways, reinforce the distinct metabolic signature associated with early tumorigenic processes in this BRCA1/2-like model. Building on these observations, the potential relevance of tetrahydrocortisone and LPE(18:1) as biomarkers warrants further consideration. Additional prior studies support the biological relevance of LPE signaling in cancer, including evidence that LPE species can promote migration and invasion in OC cell models [42] as well as broader roles for lysophospholipid metabolism in tumor progression [43] . Although direct reports of tetrahydrocortisone in OC are limited, dysregulation of glucocorticoid metabolism has been implicated in tumor-associated inflammation and stress signaling, and steroid metabolomic profiling, including glucocorticoid pathway intermediates, has demonstrated diagnostic potential in endocrine-related malignancies [44] . Importantly, the specificity of these metabolites as standalone indicators of OC is likely limited, as lysophospholipid metabolism is altered in inflammatory and metabolic conditions [45] , and glucocorticoid metabolites are strongly influenced by systemic stress-axis activity [46] . Accordingly, a more realistic translational approach is to incorporate these features into multi-analyte metabolite panels, in which combinations of metabolites may provide improved specificity relative to any individual metabolite. In this context, such metabolite signatures may ultimately serve as complementary biomarkers to established OC biomarkers rather than replacements. Clinically, CA-125 remains the most widely used biomarker for OC detection [47] while human epididymis protein 4 (HE4) has demonstrated improved specificity in certain settings, and the risk of ovarian malignancy algorithm (ROMA) integrates both biomarkers to enhance risk stratification, although these approaches remain limited in sensitivity for early-stage disease [ 48 , 49 ]. The hormone-responsive metabolic features identified herein may therefore capture aspects of tumor biology not reflected by existing protein-based assays, particularly in genetically susceptible populations such as BRCA1/2 Mutation carriers. While these findings are derived from a controlled mouse model and require validation in human cohorts, they provide a rationale for future studies in evaluating whether metabolite-based panels, in combination with established clinical markers, can improve early detection performance while maintaining acceptable specificity. Importantly, the early and consistent perturbations observed in these individual metabolites also suggest that they may reflect broader shifts in lipid metabolism, motivating further examination of pathway-level changes within the lipidome. Among the changes in the lipidome, the phospholipid subclasses of PC, SM, and TG are particularly interesting due to their established roles in cell division and proliferation (PC and SM) and energy storage (TG), which can be substantially altered during cancer development [50] . The notable trends in HGSC development observed when comparing mice treated with P4, which subsequently exhibited accelerated tumor progression ( Fig. 4 A), and those treated with mifepristone, which developed ovarian cysts ( Fig. 4 B), generated further interest in analysis. While lipid dysregulation has previously been associated with benign ovarian cysts and other gynecological maladies like endometriosis [ [51] , [52] , [53] ], the specific trends among the PC and TG lipid subclasses observed in our study, including the changes in fatty acid length and degrees of unsaturation, have been more notably linked to the progression of various cancers, including OC [ 54 , 55 ]. Higher levels of long-chain SM species have been previously observed in ovarian tumors compared to healthy tissues [56] and among patients who were later diagnosed with OC [57] . Considering the biological importance of variations in fatty acid chain length and the number of double bonds, the potential differences within the fatty acid chains across these three lipid subclasses were analyzed. A broad spectrum of fatty acid chain lengths and levels of unsaturation was seen among the annotated PC, SM, and TG species, with the most significant differences appearing when comparing mice from the P4 and mifepristone treatment groups ( Figures S4, S6, & S8 ), suggesting a lipid remodeling pattern more consistent with malignant transformation than with benign cystic pathology. As shown in Figure. S4 (D-F) , individual PC species appear to increase in concentration across the one-week, three-week, and three-month intervals, particularly as HGSC development progressed. This observation aligns with previous research by Sah et al., which found that most PC species showed decreased abundance in the premalignant stages of the disease [58] . Guo et al . conducted a study examining variations in PC species abundance, focusing on fatty acid chain length and degrees of unsaturation across five cancer types [54] . Their findings, which show that certain PC lipids vary in abundance with chain length and degree of unsaturation across cancer stages and types, underscore the importance of examining these structural features in OC, given their direct influence on membrane fluidity and the remodeling required for tumor progression. Within the SM lipid subclass, the largest shifts in FC occurred at the three-week premalignant stage, with pronounced elevation of long-chain SM species ( Figure. S6E ). This observation is consistent with prior reports of rising long-chain SM levels in OC patients compared with healthy controls [56] and among patients who later received an OC diagnosis between 3 and 23 years after analysis [57] . As SM lipid species decrease membrane fluidity and promote lipid raft formation, the accumulation of long-chain SM lipids at this key time point may support signaling pathways that facilitate early oncogenic activity. Among TG lipid subclasses, the most significant alterations occurred during early HGSC development. As shown in Figure. S8F , specific TG lipids with longer acyl chains and higher degrees of unsaturation became markedly elevated at the three-month interval, mirroring findings from Zeleznik et al ., linking long-chain, highly unsaturated TG species to increased OC risk, particularly in serous subtypes [55] . Long-chain fatty acids, derived either from dietary sources or from hepatic synthesis, are typically transported as TG and mobilized via lipid metabolic pathways to supply energy. Their increased abundance during early tumor development, as seen in Figure. S8 , is therefore biologically consistent with the elevated energetic demands of proliferating cancer cells and the broader metabolic shifts observed in HGSC [31] . To assess the extent to which P4-driven metabolic alterations are mechanistically linked to hormone signaling, a trajectory-based reversal analysis was performed to compare P4- and mifepristone-treated mice over one week, three weeks, and three months ( Figure. S10 ). Metabolites were classified by the extent to which mifepristone shifted their abundance toward the P4 placebo baseline, providing insight into the temporal dependence of these features on P4 receptor signaling. At the earliest time point of one week ( Figure. S10A ), several metabolites exhibited strong reversal, including corticosterone and multiple PC species. In these cases, metabolite levels in the mifepristone-treated group closely approximated those of the P4 placebo group, indicating that these early alterations are tightly coupled to P4 signaling and remain readily reversible. By three weeks ( Figure. S10B ), a more heterogeneous pattern emerges. While some metabolites continue to show reversal, others, including LPE(18:1), demonstrate only partial shifts toward baseline. Notably, LPE(18:1) was previously highlighted for its potential biomarker relevance and established roles in cancer-associated signaling [ 42 , 43 ]. Its incomplete reversal in response to mifepristone, therefore, suggests that its regulation is not solely dependent on P4 receptor signaling but may also reflect contributions from broader physiological responses, such as inflammation or microenvironmental remodeling. At three months ( Figure. S10C ), several metabolites, including sphingosine, ceramide-related species, and select TG, show limited reversal, indicating more stable metabolic adaptations. The persistence of these alterations despite P4 receptor antagonism suggests that early hormone-driven changes may become progressively embedded within the metabolic landscape. Together, the temporal progression from strongly reversed early changes to partially or non-reversible later alterations supports a model in which P4 signaling plays a central role in initiating metabolic remodeling while subsequent disease progression involves stabilization and propagation of these changes through additional regulatory mechanisms. Among the steroidome changes shown in Fig. 5 , the sulfate/sulfite and E1 metabolic pathways were the most enriched among those associated with P4 treatment at the premalignant stage of HGSC development. The enrichment of these pathways emphasizes the crucial role of E1 and its dynamic balance with the inactive E1 sulfate reservoir. This regulation of the ongoing source of active hormonal stimulation via the reversible conversion between E1 and E2 may contribute to early OC risk. In fact, previous studies by Johansson et al . and Koziel et al. found a significant correlation between higher E2 levels and increased OC risk, underscoring that local E2 availability may promote tumor initiation and progression [ 59 , 60 ]. In our study, increases in E2 were also observed and confirmed by DID linear logistic regression analysis ( Table S9 ) at both the premalignant and early stages of HGSC development, suggesting a possible role in the risk of OC development. Interestingly, estrogen supplementation has previously been shown to attenuate the development of HGSC metastases in the same DKO mouse model [20] . This also agrees with our results, as shown in Fig. 6 A and Table S10 : mifepristone treatment increased E1 and E2 at premalignant and early stages of HGSC development. Precursors of estrogen metabolites, specifically androstenedione, are more readily converted into E1 and E2 rather than into other androgens [61] . Additionally, excessive androgen secretion in females can also worsen endocrine disorders like polycystic ovary syndrome or congenital adrenal hyperplasia [62] . Once HGSC has begun to develop, as confirmed by surgical verification of tumor presence (three weeks to three months comparison), steroidogenesis becomes the most significantly enriched pathway ( Fig. 5 B). On the one hand, the concentration of P4 continues to decrease, while corticosterone increases, suggesting that the conversion between both metabolites by 11β-hydroxyprogesterone promotes OC oncogenic activity ( Fig. 5 , Table S9). On the other hand, the levels of cortisol, cortisone, and tetrahydrocortisol glucocorticoids also increased at the premalignant time points ( Table S9 ), while their concentrations began to decrease concurrently with tumor development. The observed increase in cortisol at the premalignant stage might suggest an initial stress- and inflammatory-mechanism aimed at inhibiting early tumor development, with concurrent elevations in cortisone and tetrahydrocortisol levels. However, previous findings by Veneris et al., reported correlations between higher glucocorticoid receptor gene expression and decreased survival in patients with advanced-stage OC [63] . Similar results were reported in a breast cancer study, demonstrating that steroid receptor expression levels often decline or lose their predictive value for treatment response, thereby necessitating more aggressive interventions such as chemotherapy or targeted therapies [64] . Nevertheless, a potential contributing factor to the variability observed in the concentrations of glucocorticoids is the well-established circadian rhythm of these hormones, as both corticosterone and cortisol exhibit pronounced diurnal oscillations in mice [65] . Thus, differences in sample collection timing may partially influence the magnitude of these steroidogenic shifts, and further studies are needed to rule out this effect. Treatment with mifepristone revealed multiple concurrent hormonal effects ( Fig. 6 A). In addition to functioning as a P4 receptor antagonist blocking P4 conversion to corticosterone and, in consequence, OC oncogenic activity (see also Table S10 ); mifepristone also acts as a glucocorticoid, rising cortisol and cortisone levels at later stages. Furthermore, it exhibits androgenic properties by promoting the conversion of to E1 and E2 estrogens through the enzyme aromatase. Supporting this mechanism, enrichment analysis ( Fig. 6 B) identified androstenedione metabolism as one of the most enriched pathways in both the one week to three weeks and three weeks to three months time intervals. Collectively, these results support that mifepristone may be a potential early-stage OC risk-reducing agent by inhibiting P4 conversion to corticosterone, and promoting aromatase activity and thereby favoring the conversion of androgens to estrogens. Nonetheless, further studies are required to validate these findings. Lipids and steroids metabolic alterations previously mentioned in this study may have implications for risk stratification and early detection strategies in BRCA1/2-mutation carriers. The coordinated perturbations observed across steroidomic and lipidomic profiles suggest that hormone-driven metabolic signatures could serve as intermediate biomarkers of early disease susceptibility, potentially enabling a more refined risk assessment beyond genetic status alone. In this context, the observed effects of P4 on metabolic remodeling raise the possibility that hormone-modulating strategies, including P4-based chemoprevention, warrant further investigation although their clinical utility remains uncertain and must be carefully evaluated given the complex and context-dependent roles of steroid signaling in ovarian carcinogenesis. Additionally, the integration of steroidomic and lipidomic features into multi-analyte metabolite panels may offer a promising avenue for improving early OC detection, particularly when combined with emerging minimally invasive sampling approaches. While prior studies have highlighted the potential of circulating metabolites as cancer biomarkers, further validation in longitudinal human cohorts will be essential to determine the sensitivity, specificity, robustness, and clinical feasibility of such approaches in BRCA1/2-mutation carriers. Our findings provide evidence of the acceleratory effects of P4 on premalignant and early stages of HGSC development, particularly among BRCA1/2 mutation carriers. However, it is crucial to acknowledge the limitations of these results. Although the BRCA-resembling DKO mouse model recapitulates key genetic and pathological features of HGSC, it does not fully capture the molecular heterogeneity, tumor microenvironment complexity, or inter-patient variability observed among human patients with both BRCA1/2 mutations and OC diagnoses. In fact, DKO mice do not harbor a BRCA mutation. However, despite of it, DKO model has been demonstrated to be a valid surrogate for BRCA1/2 carriers due to its capacity to replicate the human disease’s origin in the fallopian tube epithelium before spreading to the ovaries and metastasize throughout the pelvic and peritoneal cavities, and because HGSC’s exhibit BRCAness, defects in homologous recombination repair, the hallmark molecular features of BRCA-mutant tumors, suggesting the relevance of the DKO model to BRCA1/2 [20] . Furthermore, studies using human fallopian tube epithelial explants have demonstrated that P4 directly regulates epithelial function, including modulation of secretory activity and ciliary dynamics, pointing out its role in maintaining epithelial homeostasis [ 66 , 67 ]. These observations are also consistent with the DKO mouse model employed here, in which the loss of P4 signaling (conversion from P4 to corticosterone in our P4 treatment) promotes tumor progression, indicating conserved mechanisms between humans and the DKO mouse model. Additionally, the sample size ( n = 5) was limited by the complexity and resource-intensive nature of the animal model and thus, precludes definitive conclusions regarding biomarker discovery or broad generalizability. Nevertheless, the consistency of observed steroidomic and lipidomic trends across biological replicates supports the dataset's internal robustness. Importantly, the lack of validation in human biospecimens further limits assessment of clinical translatability, and future studies using human patient-derived samples will be essential to determine whether the identified metabolic alterations are conserved in human disease. The cross-sectional design, with individual analyses and surgical interventions performed at discrete time points (one week, three weeks, and three months), provides only a partial view of temporal metabolic dynamics and does not fully resolve the transitional states during tumor initiation and progression. Moreover, systemic endocrine perturbations induced by P4 and mifepristone administration may introduce confounding effects that extend beyond tumor-intrinsic metabolism, complicating the attribution of observed metabolic changes to oncogenic processes. Finally, while larger-scale studies, particularly in BRCA1/2 mutation carriers, would be critical for capturing population heterogeneity and enabling statistically powered comparisons between healthy and diseased states, such efforts remain challenging due to the rarity of early-stage HGSC sampling and the logistical and financial constraints associated with human cohort studies. Despite these limitations, the integrated -omic framework presented herein provides a valuable foundation for future longitudinal and translational investigations of hormone-driven metabolic reprogramming of HGSC.

This study provides a detailed analysis of how lipid profiles and steroid hormone metabolism change in BRCA1/2-resembling DKO mice after treatment with P4 and anti-progestins. By analyzing 336 unique lipids across 20 subclasses, we identified specific trends that accelerated and slowed HGSC development at different stages of growth and spread. Detecting HGSC early in BRCA1/2 mutation carriers remains difficult, but the elevated levels and known functions of PC, SM, and TG during the premalignant phase strongly suggest that these lipid subclasses could serve as useful markers of metabolic reprogramming, offering important insights for early detection of high-grade serous OC in BRCA mutation carriers. Complementing these findings, an extensive LC-MS/MS-based method enabled, to our knowledge, the first identification and quantification of twelve steroid hormones in early HGSC, revealing key shifts in hormone metabolism across the premalignant and early stages of HGSC development. In P4-treated mice, where tumor growth was promoted, an altered balance between P4 and corticosterone through 11-hydroxyprogesterone suggests an increased risk of early-stage OC. The equilibrium between E1 and E1 sulfate facilitated reversible conversion between E1 and E2, thereby increasing estrogen activity and reducing early-stage OC risk. Additionally, androstenedione contributed to this effect by serving as a substrate for aromatase-mediated synthesis of E1 and E2. Conversely, treatment with the anti-progestin mifepristone substantially inhibited P4 conversion to corticosterone and promoted aromatase activity, thereby potentially preventing further HGSC development. Together, these results improve the current understanding of how specific lipid subclasses and steroid hormones reflect metabolic processes associated with early-stage OC risk and malignancy in BRCA1/2 mutation carriers. Future research should focus on identifying additional reliable biomarkers and developing personalized treatment strategies based on hormone receptor status and transcription factor profiles.

Ovarian cancer (OC) remains the deadliest gynecological malignancy affecting women. Often referred to as a silent killer, OC typically manifests with vague symptoms, such as abdominal bloating and abdominopelvic pain, which are frequently mistaken for less severe conditions, resulting in delayed diagnoses [1] . Although OC is often regarded as a single disease in global cancer statistics, it comprises several distinct conditions, including high-grade serous carcinoma (HGSC), also known as high-grade serous ovarian cancer. HGSC is the most prevalent and deadly subtype of OC, accounting for the majority of OC-related deaths, with a five-year survival rate of merely 32 percent [ [1] , [2] , [3] , [4] ]. This is because >75 percent of HGSC diagnoses occur at advanced stages (stages III and IV), when widespread metastases have already taken place [ 1 , 2 , 4 ]. When considering all subtypes, OC ranks as the third leading cause of cancer-related deaths among women worldwide, with cancer incidence rates varying by ethnicity [4] . The incidence rates rise in women as they age, especially among those with mutations in the breast cancer genes (BRCA1/2) and other familial cancer risks [5] . Women with a BRCA1 mutation have a 40 to 60 percent higher lifetime risk of developing OC, while those with a BRCA2 mutation face a 10 to 30 percent increased risk [6] . Individuals carrying BRCA1/2 mutations are often advised to undergo prophylactic risk-reducing salpingo-oophorectomies (RRSO), which have been shown to reduce the likelihood of developing OC or other gynecological conditions in BRCA1/2-positive patients by over 75% [ 7 , 8 ]. Current OC detection methods, the coupling of transvaginal ultrasounds (TVU) and cancer antigen-125 (CA-125) blood tests, frequently lack sensitivity and specificity for early-stage OC, as higher levels of CA-125 are also present during menstruation and in conditions like endometriosis and coronary heart disease [9] . Nevertheless, these ultrasounds and prophylactic surgeries are invasive and could be avoided in lower-risk individuals, with or without BRCA1/2 mutations, if a more sensitive and specific screening and risk-assessment approach were developed. To identify an effective screening tool or early-stage biomarkers for OC, it is essential to first gain a deeper understanding of the early-stage disease processes in BRCA1/2-mutation carriers. Understanding this is essential to grasping why the disease often manifests more aggressively in these patients. For individuals carrying BRCA1/2 mutations, the disruption of deoxyribonucleic acid (DNA) repair mechanisms encoded by these genes does not fully account for the specific organ localization of reproductive cancers. Previous studies have shown that mastectomies and RRSO procedures are linked to a reduced risk of developing reproductive cancers, indicating a possible role of hormonal dysregulation involving the two primary ovarian sex hormones, estradiol (E2) and progesterone (P4), in those with these mutations [ 10 , 11 ]. The involvement of these reproductive hormones in the development of both breast cancer and OC has been a contentious issue in previous literature. Most OC cases are diagnosed in post-menopausal patients aged between 50 and 69 [12] . This has led to questions about the role of ovarian sex steroids in the development of OC, as the levels of circulating steroids decrease with age [ 13 , 14 ]. Moreover, among the two primary ovarian sex hormones, P4 has long been considered "protective" against OC. Previous studies have found that individuals who use hormonal contraceptives long-term or have multiple full-term pregnancies experience a reduced risk of OC [ 15 , 16 ]. However, both perspectives overlook the realities faced by BRCA1/2-mutation carriers, who are often diagnosed with OC at pre-menopausal ages, before ovarian regression, and experience dysregulation in their sex steroid metabolic systems [ [17] , [18] , [19] ]. Additionally, individuals with BRCA1/2 mutations exhibit elevated levels of both E2 and P4 during the menstrual cycle [10] . Our recent study revealed that P4 treatment induces the development of metastatic HGSC in BRCA1/2-resembling double-knockout (DKO) mice [20] . In contrast, blocking P4 signaling with mifepristone, a known P4 antagonist, significantly reduced HGSC development and metastases and markedly improved mouse survival [20] . These findings highlight P4 as a critical endogenous hormonal factor in the development and metastasis of HGSC. Building on these findings, we hypothesize that P4 induces distinct metabolomic alterations during HGSC development and progression. The present study focuses on applying both targeted and non-targeted ultra-high performance liquid chromatography mass spectrometry (UHPLC-MS) methods, including triple quadrupole mass spectrometry and simultaneous quantitation and discovery (SQUAD) analysis [21] , to profile serum samples from DKO mice at a premalignant stage as well as at an early stage of HGSC. The targeted analysis aimed to explore potential direct or inverse correlations between steroid metabolites and P4 treatment in mice during tumor development. Non-targeted analysis, on the other hand, involved detecting broader metabolomic and lipidomic changes within the mice sera. Together, these combined approaches offer unique opportunities to map metabolic alterations specific to the initiation and development of early-stage HGSC, paving the way for the potential identification of effective biomarkers for early detection and elucidating the early carcinogenesis of HGSC.

The authors declare no conflicts of interest.

The data obtained in this study are accessible at the National Institute of Health’s Common Fund’s NMDR (supported by the NIH grant, U01-DK097430) website, the Metabolomics Workbench, https://www.metabolomicsworkbench.org/ , with the study ID ST004688.