Two cycles of chemotherapy modulate itaconic acid levels via ATF3/TET2/NF-κB pathways in ovarian cancer among Han Chinese women

Abstract

Background

The metabolic rationale chemotherapy cycles after debulking surgery in ovarian cancer (OC) remain unclear. This prospective cohort was aimed to define metabolic changes induced by chemotherapy and assess longitudinal effects.

Methods

Metabolites between treatment-naïve OC patients (n = 26) and age-matched healthy controls (n = 30) were identified by gas chromatography-mass spectrometry (GC-MS). Chemotherapy-related signaling molecules were quantified via enzyme-linked immunosorbent assay.

Results

Our analysis identified 38 differentially expressed metabolites, including critical intermediates in the tricarboxylic acid (TCA) cycle and revealed significant reductions in five TCA cycle intermediates, namely itaconic acid, citric acid, fumaric acid, cis-aconitic acid, and malic acid, in OC patients compared to controls. Following the first chemotherapy cycle, absolute quantification of 128 metabolites was performed, with targeted metabolomic analysis of TCA cycle intermediates revealing an increase in levels of itaconic acid prior to the second chemotherapy cycle. Remarkably, after two chemotherapy cycles, itaconic acid concentrations normalized to levels comparable to those in healthy controls, suggesting metabolic recovery. Furthermore, significant inverse correlations were observed between human epididymis protein 4 (HE4) levels and the concentrations of the five TCA cycle intermediates, highlighting the potential of these metabolites as biomarkers. Mechanistic studies revealed that two cycles of chemotherapy restored metabolic homeostasis, normalizing itaconic acid levels and reactivating the Activating transcription factor 3 (ATF3)/ten-eleven translocation 2 (TET2)/NF-kappaB (NF-κB) signaling pathway to levels seen in healthy participants.

Conclusions

Our pilot data suggest that two chemotherapy cycles may help restore metabolic homeostasis in patients who have undergone complete cytoreduction. These findings provide preliminary insights into chemotherapy-induced metabolic reprogramming and warrant further investigation in larger, outcome-driven studies.

Trial registration

ClinicalTrials.gov ID: ChiCTR2300069160.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13048-026-02074-1.

Background

Ovarian cancer (OC) is one of the most challenging gynecological malignancies owing to its frequent late-stage diagnosis, high propensity for metastasis, and significant risk of recurrence [1–4]. Epidemiological data underscore its global burden: the American Cancer Society estimated 12,740 deaths from OC in the United States in 2024 [5], while in China, the disease caused approximately 40,000 deaths in 2020, with mortality rates continuing to rise as of 2022 [6]. According to the National Comprehensive Cancer Network (NCCN) guidelines, the standard treatment for advanced-stage or high-risk OC patients typically involves surgery followed by post-surgery chemotherapy or neoadjuvant chemotherapy (NACT) [2, 7]. Nevertheless, the current guideline-based standard is six cycles [2, 6], and that potential number of chemotherapy cycles required after optimal debulking surgery (ODS) is inconsistent based on medical biological mechanisms. Although human epididymis protein 4 (HE4) is widely applied as a biomarker for OC diagnosis and prognosis monitoring [8], its low accuracy during chemotherapy limits its utility in routine clinical practice. Therefore, there is a critical need for supplementary biomarkers, such as metabolites, to monitor metabolic changes during chemotherapy, optimize post-chemotherapy regimens, and explore potential therapeutic mechanisms.

Longitudinal metabolomic studies tracking OC patients through chemotherapy remain scarce. Cui et al.. reported that the plasma arachidonic acid-to-prostaglandin E2 ratio and tumoral focal adhesion kinase (FAK) expression were significantly reduced in OC patients after chemotherapy relative to the pre-treatment levels [9]. Similarly, Corvigno et al.. identified the associations between specific metabolic alterations, including dysregulation of pyrimidine, aspartate and asparagine metabolism, and clinical response to chemotherapy in OC patients [10]. A complete metabolic response (CMR) on 18 F-fluorodeoxyglucose positron emission tomography/computed tomography (18 F-FDG-PET/CT) after three cycles of NACT has been put forward as a prognostic marker for the recurrence and mortality in advanced high-grade serous carcinoma [11]. However, most existing studies are constrained by cross-sectional designs or a narrow focus on single chemotherapy cycles, failing to systematically explore the longitudinal metabolic mechanisms underlying OC progression following ODS across multiple post-chemotherapy regimens [10].

This study was aimed to 1) characterize the chemotherapy-induced metabolic alterations of OC, 2) explore the metabolic rationale chemotherapy cycles and potential underlying mechanisms; 3) elucidate the longitudinal metabolic effects following ODS. Ultimately, this research seeks to provide actionable insights for clinical decision-making and uncover the molecular mechanisms underlying chemotherapy efficacy in OC.

Methods

Study populations

This prospective cohort study, conducted between March 2023 and March 2025, included 30 patients and 26 healthy controls matched for age and body mass index (BMI). All participants were recruited from the Affiliated Nanchong Central Hospital of North Sichuan Medical College (ClinicalTrials.gov ID: ChiCTR2300069160), with healthy individuals identified through the hospital’s physical examination center.

The inclusion criteria for the OC group were as follows: radiological suspicion (magnetic resonance imaging (MRI)/ultrasound) of OC with histopathological confirmation; planned treatment with one of two standard pathways—either primary debulking surgery (PDS) followed by platinum/paclitaxel-based chemotherapy, or (NACT) followed by interval debulking surgery (IDS) and continued chemotherapy; commitment to regular oncology follow-up; and commitment to study blood collection protocol.

Exclusion criteria were shown below: patients not undergoing chemotherapy after surgery, those not confirmed by histopathology, those not receiving platinum-based chemotherapy, those unwilling to provide blood according to the study schedule, and those who failed to collect blood samples according to the study schedule (Flow chart of the patient selection is displayed in Fig. 1). This study was approved by the Institutional Review Board of the Affiliated Nanchong Central Hospital of North Sichuan Medical College (NO.2023-010), and written informed consent was acquired from all participants.

Fig. 1.

Flowchart of patient selection. A total of 30 ovarian cancer patients and 30 healthy participants (control group) were initially enrolled. Four patients were excluded due to loss of follow-up

Blood collection

To characterize chemotherapy-associated metabolic trajectories while minimizing acute infusion-related perturbations, blood was collected at standardized pre-cycle time points, immediately before each infusion. Specifically, blood samples were collected at three chemotherapy timepoints: pre-treatment baseline (Chemotherapy-1 pre, Chemo-1 pre), prior to cycle 2 (Chemotherapy-2 pre, Chemo-2 pre), and prior to cycle 3 (Chemotherapy-3 pre, Chemo-3 pre). This design prioritizes comparability across patients by capturing the metabolic state after recovery from the previous cycle and reflecting the cumulative, persistent effects of treatment rather than transient post-infusion responses. Prior to blood collection, all participants required an overnight fasting period. Subsequently, 5 ml samples of venous blood were collected using a peripheral venous catheter in the following morning. Blood samples were centrifuged (3,000 × g, 10 min, 4 °C) using standardized protocols. Plasma aliquots were cryopreserved in microtubules (Lelystad, Netherlands) at -80 °C until analysis [12].

Gas Chromatography-Mass Spectrometry (GC-MS)

Metabolite extraction and derivatization from plasma

Plasma metabolite extraction was performed by mixing 100 µL aliquots with 200 µL ice-cold methanol and 4 µL isotope-labeled internal standards (d4-alanine, d5-phenylalanine, d2-tyrosine). After vortexing (30 s), samples were subject to incubation at − 20 °C for 30 min under intermittent shaking. The mixture was then centrifuged at 12,000 rpm for 15 min. Afterwards, the extracted samples were chemically derivatized through the methyl chloroformate (MCF) derivatization method, following the protocol published by Smart et al. [13]. Briefly, 200 µl of the extracted metabolites were transferred into a glass tube and mixed with 200 µl NaOH (1 M), 34 µl pyridine, and 20 µl MCF. After vortexing for 30 s, an additional 20 µl MCF was supplemented and vortexed for 30 s. Then, 400 µl chloroform was added and vortexed for 10 s, followed by the addition of 400 µl NaHCO₃ and vortexing for another 10 s. The upper aqueous and intermediate protein layers were carefully discarded after centrifugation of the derivatized sample at 2000 rpm for 10 min. Sodium sulfate was supplemented to absorb the remaining water. Finally, 150–200 µl of the chloroform phase was transferred to a vial for mass spectrometry analysis.

Gas chromatography-mass spectrometry analysis

The devitalized plasma metabolites were analyzed by Agilent Intuvo9000 Gas chromatograph coupled with an MSD5977B mass spectrometer. The DB-1701 GC capillary column (20 m × 180 μm ×0.18 μm, Agilent) was used for metabolite separation. The GC-MS inlet parameters were set as follows [12, 14]: injection volume (1 µl), pulsed splitless mode (290 °C, 180 kPa for 1 min), guard chip (300 °C), and helium carrier flow rate (1 mL/min). The GC oven temperature was initiated at 45 °C for 2 min, raised to 180 °C at 9 °C/min and held for 5 min, subsequently raised to 220 °C at 40 °C/min and held for 5 min, continued to raise to 240 °C at 40 °C/min and held for 5 min, and finally raised to 280 °C at 80 °C/min and held for 3 min. The GC-MS system maintained the following temperatures: ion source at 150 °C, quadrupole at 230 °C, and transfer line at 300 °C.The mass spectrometry was operated using electron-impact ionization at 70 eV. The solvent delay was set at 4.5 min, with the mass range being 30 to 550 m/z at a scan speed of 2.9 scan/s [12, 14].

Data extraction and normalization

The Automated Mass Spectral Deconvolution and Identification System (AMDIS) software was used for metabolite identification and deconvolution. The metabolites were identified by comparing their MS fragmentation patterns (mass-to-charge ratio and relative intensity of mass spectra to most abundant ion) and GC retention time to an in-house MS library established with chemical standards. The Mass Omics R-based script was used for extracting the relative concentration of the metabolites through the peak height of the most abundant fragmented ion mass. To improve the quantitative robustness and minimize the human and instrumental variability, the relative abundances of the identified compounds were normalized against multiple internal standards. Subsequently, blank samples were applied to subtract background contamination and any carryover from identified metabolites [14]. Calibration curves were acquired from the respective chemical standard, spanning a concentration range of 0 ~ 195.8 µM [15].

Metabolites in signaling pathways with chemotherapy

Chemotherapy-related signaling molecules were quantified by enzyme-linked immunosorbent assay (ELISA) at Lilai Biomedical Research Center using commercial kits, with analyses performed by blinded technicians (≥ 5 years’ experience) [16]. Aconitate decarboxylase 1 (ACOD1) concentrations were determined using Human ACOD1 Platinum ELISA Kit (eBioscience; Zhuocai Biotech, China) following the manufacturer’s protocol. Activating transcription factor 3 (ATF3)/ten-eleven translocation 2 (TET2)/NF-kappaB (NF-κB)/inhibitor of nuclear factor kappa B zeta (IκBζ)/Interleukin 6 (IL-6) levels were measured with human ATF3/TET2/NF-κB/IκBζ/IL-6 Platinum ELlSA Kit (eBioscience, an Zhuocai Biotechnology Company, Shanghai, China). The absorbance (OD value) was measured at 450 nm and the sample concentrations were calculated. All procedures were performed and analyzed in strict accordance with the manufacturer’s instructions.

Clinic data collection

Serum HE4 concentrations were retrospectively obtained from the patients’ clinical records. Following the standardized collection protocols, venous blood samples were drawn into serum separator tubes, allowed to clot for 30 min at room temperature, and later centrifuged at 3000 × g for 10 min to obtain serum aliquots [17, 18]. All samples were processed and analyzed within 24 h of collection to ensure sample integrity. Quantitative measurement of HE4 was performed using automated chemiluminescent immunoassays (CLIA) on the ARCHITECT i2000SR platform (Abbott Diagnostics, Chicago, IL, USA) according to the manufacturer’s specifications [19]. The assay demonstrated a measurable range of 15-1500 pmol/L, with intra- and inter-assay coefficients of variation being < 5% and < 8%, respectively. All the testing procedures were conducted in a CAP-accredited central laboratory by experienced technicians blinded to clinical outcomes to eliminate potential bias [19].

Statistical analysis

Continuous data were expressed as mean ± standard deviation, while categorical data were presented as proportions. Metabolomic analysis was conducted using MetaboAnalyst 6.0 (https://www.metaboanalyst.ca), with plasma metabolite levels being log-transformed and Pareto-scaled to achieve normal distribution prior to analysis. Univariate analyses included Student’s t-tests (two-group comparisons) and one-way ANOVA (multi-group comparisons). Multivariate analyses comprised partial least squares-discriminant analysis (PLS-DA), evaluation using R²Y (goodness-of-fit) and Q²Y (predictive ability via leave-one-out cross-validation), and linear support vector machine (SVM) classification to rank metabolite importance [12]. To mitigate overfitting in this small cohort, multivariate models were assessed using leave one out cross validation, and performance was summarized using cross-validated metrics rather than apparent (training) performance. Pathway enrichment analysis was conducted via KEGG (2023Q1 release) with the significance thresholds of p < 0.05, targeted metabolite and pathway enrichment were performed with false discovery rate (FDR) < 0.3 (Benjamini-Hochberg correction). Pearson correlation assessed associations between metabolites and key elements during chemotherapy, with multiple comparisons adjusted using the q-value R package (FDR < 0.05). Data were visualized using GraphPad Prism 10 (box/line plots), ggplot2 (heatmaps, forest plots, correlation matrices), and the CNSKnowAll platform (clustering, chord diagrams). All analyses were prespecified and hypothesis-driven, with appropriate multiple-testing correction; no data-driven feature selection or iterative model tuning was performed. All statistical analyses were conducted in SPSS 23.0 and ggplot 2 R package, with supplementary figures generated in Microsoft Excel 2019.

Results

Patient characteristics

Between March 2023 and May 2023, 26 patients and 30 healthy participants (controls) were enrolled (4 patients were excluded, Fig. 1). The majority of OC patients presented with advanced-stage disease (International Federation of Gynecology and Obstetrics [FIGO] stage III/IV) and epithelial histology (serous subtype: 96.2%).

Among the 26 patients, 18 (69.2%) underwent PDS followed by post-chemotherapy, and 8 (30.8%) received neoadjuvant chemotherapy followed by IDS and post-chemotherapy. Surgical outcome was documented as ‘complete cytoreduction’ (no macroscopic residual disease, R0) in 22 patients (84.6%) and ‘optimal’ (< 1 cm residual) in 4 patients (15.4%, shown in Table 1). Table 1 summarizes demographic and clinical characteristics. No significant differences were observed in age (median: 55.3 vs. 65.1 years, P = 0.093) or BMI 23.9 vs. 23.7 kg/m², P = 0.231) between the OC group and the control group.

Table 1.

Baseline clinical characteristics of the study participants (n = 56)

Variable	Ovarian cancer (n = 26)	Healthy participants (n = 30)	P‑value*
Age, years	53.5 (48.0–62.3)	65.0 (52.0–67.0)	0.659
BMI, kg/m⁻² †	23.9 (23.1–25.6)	22.3 (19.1–27.0)	0.231
FIGO stage, n (%)
Early (I–II)	7 (26.9)
Advanced (III–IV)	19 (73.1)
Histopathological type, n (%)
Epithelial	25 (96.2)
Other	1 (3.8)
Malignant ascites, n (%)
Yes	19 (73.1)
No	7 (26.9)
Chemotherapy regimen, n (%)
Platinum-based	20 (100)
Other	0 (0)
Treatment way before chemotherapy, n (%)
PDS followed by post-chemotherapy	18 (69.2)
NACT followed by IDS and post-chemotherapy	8 (30.8)
Residual disease (cm), n (%)
< 1	4 (15.4)
< 0	22 (84.6)

Values for continuous variables were presented as median (inter‑quartile range); categorical data were shown as number (percentage)

Abbreviations: PDS primary debulking surgery, NACT neoadjuvant chemotherapy, IDS interval debulking surgery

* P‑values derived from the Mann‑Whitney U test for continuous variables

† BMI recorded for patients with ovarian cancer before the first cycle of chemotherapy

Metabolic profiling in OC vs. healthy participants

Metabolomic analysis identified 128 metabolites. After performing a t-test with FDR correction for multiple comparisons, 38 of these metabolites were found to be significantly altered in OC patients compared to healthy controls (P < 0.05, FDR < 0.05, Supplementary Table 1). Univariate p-values were adjusted using the Benjamini–Hochberg FDR procedure, and key metabolites—including itaconic acid and multiple TCA cycle intermediates remained significant after correction (FDR < 0.05). Supervised PLS-DA modeling revealed a clear metabolic separation between OC patients and controls, indicating distinct metabolic signatures in OC (R²Y = 0.748, Q² = 0.481, Supplementary Fig. 1A-B). This analysis identified 33 down-regulated and 5 up-regulated metabolites in OC (P < 0.05, FDR < 0.05, FC > 1.1; Supplementary Fig. 1C). Notably, these TCA cycle intermediates were consistently depleted in OC, and the combined area under the curve (AUC) for the top five metabolites reached 0.852 (95% CI: 0.696–0.988; Supplementary Fig. 1D). An exploratory multivariate receiver operator characteristic curve (ROC) analysis, performed using models incorporating the top 5, 10, 15, 25, 50, and 100 ranked metabolites, indicated that a panel comprising the top 10 metabolites showed promising diagnostic potential within this cohort (Supplementary Fig. 1E), however, it is important to note that this biomarker panel was exploratory and required validation in an independent external dataset. The metabolic pathways associated with these candidate metabolites were detailed in Supplementary Fig. 1F.

PLS-DA analysis revealed that chemotherapy significantly altered the metabolic profile of OC patients. Pathway enrichment analysis further highlighted significant dysregulation in both the tricarboxylic acid (TCA) cycle and pyruvate metabolism (P < 0.05 for each, Supplementary Fig. 1F and Fig. 3). Across successive pre-treatment time points (Chemo-1 pre, Chemo-2 pre, and Chemo-3 pre), the metabolic patterns of OC patients progressively shifted toward those of the healthy control group. The PLS-DA model demonstrated satisfactory explanatory and predictive performance (R² = 0.55, Q² = 0.39; Fig. 2A). Cluster analysis identified significant alterations in amino acids, amino acid derivatives, TCA cycle-related metabolites (including itaconic acid), and their derivatives. Volcano plot analysis indicated significantly decreased itaconic acid levels in patients relative to healthy participants (FDR < 0.05, fold change (FC) > 1.2, Fig. 2B), with gradual normalization of levels following successive chemotherapy cycles. The metabolic disparities diminished with treatment progression. One-way ANOVA showed statistically significant differences in TCA cycle metabolites (including itaconic acid) between OC patients and healthy participants before the first chemotherapy cycle (Fig. 2C). However, these differences became statistically insignificant after the third chemotherapy cycle (Fig. 2D).

Fig. 3.

Pathway enrichment analysis. The Sankey plots show the distribution of affected metabolic athways across the three chemotherapy cycles, with the color intensity corresponding to pathway significance (top panels: Chemo-1 pre vs. healthy participants; middle: Chemo-2 pre vs. healthy participants; bottom: Chemo-3 pre vs. healthy participants)

Fig. 2.

Metabolic profiling of ovarian cancer (OC) versus healthy participants. Longitudinal metabolic changes during chemotherapy. A PLS-DA analysis revealed that chemotherapy significantly altered the metabolic profile of OC patients (R2Y = 0.55, Q2 = 0.39). B A heatmap illustrates metabolite levels with elevated metabolites shown in red and reduced metabolites in blue, comparing the OC group to the control group. Only metabolites with P < 0.05 were displayed. C The volcano plot illustrates the increase (shown in red)and decrease (shown in green) of metabolites across different chemotherapy stages compared to healthy participants. D Differential metabolites involved in the TCA cycle and pyruvate metabolism

Longitudinal metabolic changes during chemotherapy

Chemotherapy cycle 1 (chemo-2 pre. vs. healthy participants)

PLS-DA, heatmap analysis and clustering analysis revealed 9 phyla correlated with 13 differentially abundant metabolites, including itaconic acid, citric acid, fumaric acid, cis-aconitic acid, and malic acid, which were obviously higher in Chemo-1 pre group than those in control group (Fig. 2B). Compared with healthy participants, those TCA cycle intermediates, such as itaconic acid, and citric acid, partially recovered. Following the first cycle, TCA intermediates showed further elevation compared with Chemo-1 pre (P < 0.05), though still below healthy levels (P < 0.05). Pathway analysis indicated reactivation of pyruvate metabolism and TCA cycle after one-cycle chemotherapy (P < 0.05; Fig. 3).

Chemotherapy cycle 2 (chemo-3 pre group vs. healthy participants)

After two cycles, five TCA cycle intermediates: itaconic acid, citric acid, fumaric acid, cis-aconitic acid, and malic acid, resembled to levels of healthy participants (P > 0.05; Fig. 2B). Metabolic pathway disparities between Chemo-2 pre group and controls were not significant for both TCA cycle (P = 0.21) and pyruvate metabolism (P = 0.18) (Fig. 3, shown in Table 2).

Table 2.

Trajectory of key serum metabolites and HE4 across consecutive chemotherapy cycles in patients with ovarian cancer (n = 26)

Metabolite	Chemo-1 pre	Chemo-2 pre	Chemo-3 pre	P-value†
Pyruvic acid	2.34 (0.97–3.90)	2.74 (1.42–4.74)	5.54 (3.34–9.45)	0.015*
Citric acid	0.06 (0.04–0.08)	0.07 (0.06–0.08)	0.05 (0.03–0.07)	0.235
cis-Aconitic acid	0.64 (0.37–0.73)	0.65 (0.57–0.77)	0.51 (0.33–0.67)	0.072
Itaconic acid	0.04 (0.03–0.06)	0.05 (0.04–0.06)	0.07 (0.06–0.07)	0.018*
Fumaric acid	0.04 (0.04–0.05)	0.05 (0.04–0.06)	0.06 (0.05–0.09)	0.040*
Malic acid	0.10 (0.07–0.11)	0.11 (0.09–0.13)	0.11 (0.08–0.14)	0.106
HE4, pg mL	895 (316–1 090)	69.6 (46.0–129.9)	68.0 (43.2–109.4)	< 0.001***

Values were median (inter-quartile range)

P-values indicate statistical significance at *P < 0.05 and ***P < 0.001

† P-values derived from a non-parametric Friedman test for repeated measures across the three pre-chemotherapy time points

Dynamic changes in key signaling molecules in ELISA profiling

ELISA profiling demonstrated dynamic changes in key signaling molecules (Fig. 4). ACOD1restored to near-normal levels Chemo-2 pre group (P = 0.08 vs. controls). ATF3/NF-κB/IL-6 elevated in Chemo-1 (P < 0.01) but were normalized Chemo-3 pre group (P > 0.05). TET2/IκBζ were suppressed in Chemo-1 (P < 0.05) and recovered Chemo-3 pre group (P = 0.12). Univariate ANOVA indicated that ACOD1 increased with chemotherapy, resembling the levels of health participants after one-cycle-chemotherapy. While levels of ATF3, TET2, NF-κB, IκBζ and IL-6 resembled those of health participants after two-cycles of chemotherapy.

Fig. 4.

Protein levels of key signaling molecules (ATF3, TET2, ATP3, NF-KB, IKBζ, and IL-6) measured by ELISA. Changes in protein levels were evaluated across three chemotherapy cycles (Chemo-1pre, Chemo-2 pre, Chemo-3 pre) and compared to healthy participants. Data are presented as mean ± standard deviation. Statistical significance is indicated by asterisks ( *** P <0.01, ** P <0.05, ns: not significant). Abbreviations were as follows: Abbreviations were as follows: ACOD1, aconitate decarboxylase 1; ATF3, activating transcription factor 3; TET2, ten-eleven translocation 2; NF-κB, nuclear factor-kappaB; IkBζ, inhibitor of nuclear factor kappa B zeta; IL-6, interleukin 6

Chemotherapy-induced modulation of ATF3/TET2/NF-κB pathways

Prior to chemotherapy, OC patients exhibited significantly reduced TCA cycle intermediates, including itaconic acid, citric acid, fumaric acid, cis-aconitic acid, and malic acid, compared with healthy participants (P < 0.05). This metabolic suppression was associated with the diminished activity of ACOD1, resulting in the impaired conversion of cis-aconitate to itaconic acid. Following chemotherapy, a progressive restoration of TCA cycle activity was observed. Concomitantly, key pro-inflammatory mediators, which included ATF3, NF-κB, and IL-6, were significantly suppressed (P < 0.05), while epigenetic regulators TET2 and IκBζ showed marked up-regulation (P < 0.05). After two cycles of chemotherapy, the levels of TCA cycle intermediates and associated signaling molecules (ACOD1, TET2, IκBζ, NF-κB) approached those of healthy controls (P > 0.05, Fig. 4).

Associations among serum HE4, metabolites and dynamic changes in key signaling molecules

Following one cycle of chemotherapy, HE4 levels declined significantly from a median of 895 pg/mL (IQR: 316–1090) in the pre-cycle-1 samples to 69.6 pg/mL (IQR: 46.0–129.9) in the pre-cycle-2 samples, falling below 149 pg/mL (Table 2). Pearson correlation analysis was performed to study the associations among serum HE4, metabolites and dynamic changes in key signaling molecules. Our results demonstrated that among the blood biomarkers analyzed (including CA125 cancer antigen, white blood cells, neutrophils, monocytes, lymphocytes, hemoglobin, and platelets), HE4 showed significant associations with several metabolites: itaconic acid, citric acid, fumaric acid, cis-aconitic acid, and malic acid (all P < 0.05; Figs. 5 and 6). Notably, Fig. 6 revealed inverse correlations between serum HE4 levels and these metabolites. Furthermore, itaconic acid exhibited significant associations with ATF3 in both the Chemo-1 pre and Chemo-3 pre groups (P < 0.05, Figs. 5 and 6). Importantly, the normalization of HE4 and metabolite levels should not be interpreted as the eradication of microscopic disease or as conferring a guaranteed survival benefit.

Fig. 5.

Associations among hematological markers, serum metabolites, and signaling molecules across three chemotherapy cycles (Chemo-1 pre, Chemo-2 pre, Chemo-3 pre). A Correlation matrix showing the relationships between hematological markers and various metabolites. B Correlation matrix between hematological markers and signaling molecules. C Correlation matrix between metabolites and hematological markers Red indicates a positive association, while blue indicates a negative association. The intensity of the color reflects the strength of the association, with darker shades representing stronger correlations. The numbers within the ellipses represent the correlation coefficients. Abbreviations are as follows: HE4 (cancer antigen 125), CA125 (cancer antigen 125), WBC (white blood cell), N (neutrophil), M (monocyte), L (lymphocyte), Hb (hemoglobin), and PLT (blood platelet). Statistical significance is indicated for relevant associations (P<0.05)

Fig. 6.

Scatter plots and linear correlation analyses between serum HE4 and serum metabolites (citric acid, itaconic acid, aconitic acid, fumaric acid, and malic acid) as well as ATF3 across pre-chemotherapy cycles (Chemo-1 pre, Chemo-2 pre, and Chemo-3 pre). Significant associations (P <0.05) are indicated. The colors represent the different chemotherapy cycles: Chemo-1 pre (purple), Chemo-2 pre (teal), and Chemo-3 pre (yellow). Abbreviations: HE4, human epididymis protein 4; ATF3, activating transcription factor 3

Discussion

Although international guidelines (e.g., NCCN, European Society for Medical Oncology) clearly recommend 6 cycles of platinum-based chemotherapy as standard adjuvant treatment for advanced OC [2], the metabolic rationale for this duration, and whether it could be personalized, remains an area of investigation. This study investigated metabolomic changes in OC to identify effective post-chemotherapy regimens. We found that the concentrations of itaconic acid, citric acid, fumaric acid, cis-aconitic acid, and malic acid were significantly lower in OC patients than those in healthy participants. These metabolites increased with chemotherapy and reached levels comparable to those in healthy participants after two cycles. Additionally, itaconic acid levels were associated with the ATF3/TET2/NF-κB signaling pathways and HE4. Based on these findings, we proposed that two chemotherapy cycles may help restore metabolic homeostasis in patients who have undergone complete cytoreduction, offering insights into chemotherapy-induced metabolic reprogramming and warrant further investigation in larger, outcome-driven studies.

Metabolome profiles disparity between chemo-1 pre-group and healthy participant group

The Warburg effect, a hallmark of cancer metabolism, is characterized by the preference for anaerobic glycolysis in cancerous tissues, even in the presence of oxygen. This metabolic shift causes the increased glucose consumption and the secretion of large amounts of lactic acid, supporting the energy demands and biosynthetic needs of rapidly proliferating cancer cells [20, 21]. In OC, this phenomenon is accompanied by a corresponding reduction in TCA cycle activity during aerobic glycolysis [22]. Consistent with our findings, the levels of TCA cycle derivatives and intermediates, such as itaconic acid, citric acid, fumaric acid, cis-aconitic acid, and malic acid, were obviously lower in the Chemo-1 pre-group than the healthy participant group. This suggests a metabolic reprogramming that prioritizes alternative fuel sources and energy production to support tumor growth. These observations align with previous studies demonstrating a similar reduction in TCA cycle metabolites in cancerous tissues, further underscoring the metabolic reprogramming characteristic of OC [23]. For instance, Hanahan et al. [24]. reported a reduction in TCA intermediates in OC cells, highlighting the altered metabolic pathways that facilitated cancer progression. Similarly, DeBerardinis et al. [25]. identified a strong correlation between the diminished TCA cycle activity and the enhanced tumorigenesis in OC, emphasizing the critical role of metabolic adaptations in promoting cancer cell survival and proliferation. Collectively, these findings provide compelling evidence that OC cells rely on the reprogrammed metabolic pathways to sustain their growth and survival. This metabolic rewiring not only supports the bioenergetic and biosynthetic demands of tumors but also represents a potential target for therapeutic intervention.

At least two chemotherapy cycles post-ODS was recommended for OC

The NCCN guidelines recommend 3–6 cycles of chemotherapy to maintain favorable outcomes in OC [2, 7]. Compared to the tri-weekly regimen, the weekly schedule of paclitaxel combined with carboplatin has been shown to be safe and well-tolerated [26]. While this regimen has demonstrated a modest improvement in progression-free survival (PFS) based on moderate-certainty evidence, it has not shown a significant benefit in overall survival, for which high-certainty evidence is available [27]. Therefore, the tri-weekly paclitaxel plus carboplatin regimen remains the standard choice for ovarian cancer patients in our cohort. Recent metabolomic studies have revealed that metabolites such as glutamine are uniquely enriched in populations of cancer stem cells and polyploid giant cancer cells, where they are associated with stemness and therapy-induced senescence [28]. However, current clinical guidelines do not incorporate longitudinal metabolomic profiling to assess whether ODS achieves sufficient tumor cytoreduction, nor do they use metabolome data to define the optimal number of chemotherapy cycles after ODS.

Based on our PLS-DA and heatmap analyses of metabolite profiles in this study, a clear overlapping metabolite profile was observed between healthy and chemo-2 pre groups (Fig. 2). Particularly, the levels of itaconic acid, citric acid, fumaric acid, cis-aconitic acid, and malic acid increased following two cycles of chemotherapy, indicating the alternations of TCA cycle. Furthermore, after two cycles of chemotherapy, the levels of these metabolites returned to levels comparable to those in healthy participants. These results suggested that at least two cycles of chemotherapy were recommended for OC following ODS. The underlying mechanism remains unclear, though one possible explanation is the up-regulation of TCA cycle activity in aerobic glycolysis following chemotherapy [29]. We explored the association between TCA cycle derivatives, intermediates, and HE4, a widely used biomarker for the diagnosis and prognosis of OC. We observed that increased itaconic acid levels were associated with a reduction in serum HE4, which fell below 149 pg/ mL after one cycle of chemotherapy. This finding is consistent with previous studies demonstrating HE4 as a reliable biomarker for assessing post-chemotherapy response and predicting survival outcomes [30–33].

Chemotherapy is intended to improve disease control and survival. In this pilot cohort, after two post-ODS cycles, several metabolites shifted toward levels observed in healthy participants, and HE4 showed an inverse association with TCA-cycle intermediates (including itaconic acid), findings consistent with an early systemic response to initial chemotherapy exposure. However, these biomarker–metabolite relationships remain surrogate signals: although biologically plausible, they cannot be translated into survival benefit or cure in the absence of outcome data. Indeed, the broader ovarian cancer literature shows that normalization of CA-125 during or after neoadjuvant chemotherapy does not necessarily indicate eradication of disease at interval debulking surgery, and patients may still harbor macroscopic or microscopic residual tumor despite normalized markers [34]. By analogy, marked declines in HE4 and apparent metabolic normalization could likewise occur in the presence of residual disease [34]. Therefore, without linkage to clinical endpoints (PFS, OS, recurrence) and without parallel CA-125 measurements in our cohort, these changes should not be interpreted as evidence of residual disease clearance or as justification for defining chemotherapy duration or treatment completeness. Rather, they are hypothesis-generating signals that warrant prospective, adequately powered trials integrating longitudinal sampling, standardized stratification by disease status, and survival endpoints to test whether early metabolic normalization after two cycles can safely inform any treatment de-escalation. Until such validation is available, clinical practice should continue to follow established guideline-based recommendations.

Chemotherapy-associated signaling pathways

Our study further investigated the impact of chemotherapy on key signaling pathways, including ATF3/TET2/NF-κB, in OC patients. We observed significant alterations in the levels of ACOD1, ATF3, TET2, NF-κB, IκBζ, and IL-6 in OC patients (Fig. 7), which were gradually restored to levels compared with those in healthy participants following chemotherapy. ACOD1, an enzyme responsible for converting cis-aconitate to itaconic acid, returned to near-healthy levels after one cycle of chemotherapy, leading to the increased production of itaconic acid. Itaconic acid, a metabolite with emerging significance in cancer research, has been shown to exert a critical role in immune regulation and inflammation suppression [35, 36], and recent work highlights its immunomodulatory and potential anti-tumor roles [37, 38]. The up-regulation of ACOD1 after one cycle of chemotherapy, coupled with the normalization of ATF3, TET2, NF-κB, IκBζ, and IL-6 levels after two cycles, suggests that chemotherapy may enhance antitumor immunity and suppress inflammation through these pathways [39–41]. These findings align with previous studies that emphasize the roles of NF-κB and IL-6 in tumor progression and chemoresistance [42]. Additionally, studies have demonstrated that paclitaxel, a key component of chemotherapy, is metabolized into detectable plasma metabolites in treated patients [43]. Platinum- and paclitaxel-based chemotherapy enhances the antitumor activity by reducing apoptosis and inhibiting tumor metastasis and invasion [44]. Furthermore, chemotherapy may down-regulate biosynthesis and mitochondrial metabolism in highly aggressive OC cells, thereby increasing TCA cycle activity and suppressing inflammation [45, 46]. Based on these observations, we speculate that chemotherapy may enhance antitumor immunity, reduce inflammation, and restore metabolic homeostasis in OC patients after two treatment cycles. Consequently, targeting metabolic pathways in combination with platinum-based therapy may represent a novel therapeutic strategy for OC. This metabolic rewiring not only supports the bioenergetic and biosynthetic demands of tumors but also represents a potential target for therapeutic intervention.

Fig. 7.

Chemotherapy-induced modulation of the ATF3/TET2/NF-κB pathways. A Prior to chemotherapy, ovarian cancer patients displayed a suppression of key metabolites in the TCA cycle, including itaconic acid, citric acid, fumaric acid, cis-aconitic acid, and malic acid, accompanied by reduced activity of ACOD1. This metabolic disruption hindered the conversion of cis-aconitate to itaconic acid. B Following chemotherapy, a progressive recovery of TCA cycle intermediates was observed. Alongside these metabolic changes, key pro-inflammatory signaling molecules, such as ATF3, NF-κB, and IL-6, were significantly downregulated, while epigenetic regulators, including TET2 and IκBζ, were upregulated. C After two cycles of chemotherapy, both metabolic and signaling pathways normalized, reaching levels similar to those observed in healthy participants. Components directly measured in this study (including Pyruvic acid, itaconic acid, citric acid, fumaric acid, cis-aconitic acid, malic acid, ACOD1, ATF3, TET2, NF-κB, IκBζ, and IL-6) were highlighted in solid boxes; molecules inferred from prior literature (such as acetyl-CoA, isocitric acid, α-ketoglutarate, succinic acid, and oxaloacetic acid) and their associated pathways were indicated with dashed boxes. Abbreviations: IRG1, Immune response gene 1; ACOD1, aconitate decarboxylase 1; ATF3, activating transcription factor 3; TET2, ten-eleven translocation 2; NF-κb, nuclear factor-kappaB; Iκbζ, inhibitor of nuclear factor kappa B zeta; IL-6, interleukin 6

Strengths and limitations

A key strength of this study was the longitudinal collection of plasma samples at standardized intervals, which permitted the differentiation of metabolic profiles across successive chemotherapy cycles. While the dynamic metabolic changes during chemotherapy warrant further investigation, the longitudinal framework established here provided a valuable approach for monitoring treatment-induced metabolic alterations. Moreover, this methodology holds promise not only for elucidating the temporal patterns of metabolic reprogramming but also for informing the future development of personalized chemotherapy regimens.

Nevertheless, this study still had the following several limitations. First, as a single-centre pilot study with a modest sample size, we were underpowered to perform stratified analyses by clinically relevant variables, including postoperative residual disease, preoperative tumor burden (e.g., Peritoneal Cancer Index, where available), FIGO stage (I–II vs. III–IV), histological subtype, and lymph node involvement/lymphadenectomy, and long-term survival outcomes (PFS, OS, and recurrence). Consequently, the generalizability of our metabolic findings remains to be established. Second, sampling was restricted to three pre-cycle time points (baseline, pre-cycle 2, and pre-cycle 3). More frequent on-treatment or post-treatment sampling was not feasible in this cohort due to patient burden, visit scheduling, and biobanking resources. This design may have missed acute post-infusion shifts and longer-term recovery; thus, our data mainly capture between-cycle cumulative changes rather than full time-resolved metabolic dynamics across chemotherapy. Third, the proposed 10-metabolite panel (including itaconic acid) was derived as an exploratory signature from this pilot cohort and lacks external validation; independent cohorts will be required to confirm its robustness and translational relevance. Such a cohort should ideally include on-treatment and post-treatment samples to assess metabolic trajectory and its clinical relevance, a logistical complexity beyond the scope of this initial investigation. Furthermore, while most patients achieved complete cytoreduction, our metabolic observations may not generalize to patients with substantial postoperative residual disease, which could independently influence metabolic recovery. Besides, owing to the relatively short duration of the study and the limited number of patients reporting related symptoms, chemotherapy-related toxicities and drug resistance were not included as outcomes. We plan to collect and analyze this information in future research. Finally, while our manuscript focused on a selected number of metabolites and evaluated surrogate metabolic endpoints, the correlation between early metabolic normalization and long-term survival outcomes remains unknown and should be investigated in future trials.

Conclusion

In this exploratory study, our pilot data suggest distinct metabolic profiles in OC patients, with itaconic acid levels returning to levels observed in healthy participants after 2 cycles of chemotherapy. These observations suggest that early postoperative chemotherapy may be accompanied by partial metabolic recovery following ODS, but they are not yet clinically actionable. Without survival endpoints, the clinical relevance of these metabolic changes remains uncertain. Outcome-powered prospective studies should test whether early metabolic trajectories are associated with disease control and survival and whether they can inform treatment decisions.

Abbreviations

OC

Ovarian Cancer

Chemo-1

Chemotherapy-1

Chemo-2

Chemotherapy-2

Chemo-3

Chemotherapy-3

GC-MS

Gas Chromatography-mass Spectrometry

TCA

Tricarboxylic Acid

NCCN

National Comprehensive Cancer Network

NACT

Neoadjuvant Chemotherapy

ODS

Optimal Debulking Surgery

PFS

Progression-free Survival

OS

Overall Survival

HE4

Human Epididymis Protein 4

FAK

Focal Adhesion Kinase

CMR

Complete Metabolic Response

BMI

Body Mass Index

MRI

Magnetic Resonance Imaging

AMDIS

Automated Mass Spectral Deconvolution and Identification System

ELISA

Enzyme-linked Immunosorbent Assay

ACOD1

Aconitate decarboxylase 1

CLIA

Chemiluminescent Immunoassays

ATF3

Activating Transcription Factor 3

TET2

Ten-eleven Translocation 2

NF-κB

NF-kappaB

CA125

Ca125 Cancer Antigen

Authors’ contributions

YT, MH, and JL designed the study. JW, QW, BS, LL, SN and LS collected the data. YS, JH and HH performed the analyses. All authors drafted the first manuscript with the help of JL and MH. All authors have given approval of the final version prior to submission.All authors reviewed the manuscript.

Funding

This work was supported by National Key R&D Plan for Intergovernmental Cooperation, the Ministry of Science and Technology of China (Grant No.2022YFE0133100), Foundation of State Key Laboratory of Ultrasound in Medicine and Engineering (Grant No. 2024KFKT016), Sichuan Provincial Natural Science Foundation Project (Grant No. 26NSFSC0004) and Nanchong Municipal Bureau of Science and Technology Project (Grant No. 25YYJCYJ007).

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author Ting-li Han and Jun Li on reasonable request. With regard to the publication and utilization of datasets, all contributing authors to this article have executed consent declarations.

Declarations

Ethics approval and consent to participate

The written informed consent was acquired from each participant to gather their data anonymously for research purpose. The approval of this study was obtained from Ethics Committee of the Affiliated Nanchong Central Hospital of North Sichuan Medical College (NO.2023-010), and written informed consent was acquired from all participants. All procedures performed in studies involving human participants were in accordance with the 1963 Helsinki Declaration.

Consent for publication

This work is original and has not been published elsewhere. All co-authors have reviewed and approved the final version of the manuscript. We agree to the journal’s copyright/license terms (e.g., CC BY, transfer of copyright).

Competing interests

The authors declare no competing interests.