All decidual samples from healthy donors undergoing elective abortion in the first trimester between 6 and 12 weeks of gestation were obtained at Anhui Provincial Hospital, Hefei. Peripheral blood samples were collected from age-matched non-pregnant healthy women. This study was approved by the Medical Ethics Committee (No. 2022KY063) of the First Affiliated Hospital of the University of Science and Technology of China. All donors supplied informed consent.

The cells were processed within 4 h of collection. Peripheral blood samples were diluted 1:2 in PBS. Mononuclear cells were isolated by Ficoll-Hypaque centrifugation using standard procedures (29). CD3⁻CD56⁺ NK cells were isolated from peripheral blood mononuclear cells with the MACS isolation system according to the manufacturer’s instructions (Miltenyi Biotec). Decidual NK cells were isolated from decidual samples as previously described (29, 31). The cell purity was determined to be >95% by post-purification FACS analysis.

Frozen samples were transferred into low protein binding tubes (1.5 ml Eppendorf) and lysed with 300 µL lysis buffer supplemented with 1 mM PMSF with sonication. After sonication, the samples were centrifuged at 15,000 g for 15 min to remove insoluble particles. Protein concentration was determined by BCA and aliquoted for storage at -80°C. The filter-aided sample preparation (FASP) approach was used to decrease and trypsinize the isolated proteins (32). The digested peptides were desalted by the C18-Reverse-Phase SPE Column.

All analyses were performed using a Q-Exactive mass spectrometer (Thermo, USA) equipped with a Nanospray Flex source (Thermo, USA). Samples were loaded and separated in a C18 column (15 cm × 75 µm) on an EASY-nLCTM 1200 system (Thermo, USA). The flow rate was 300 nL/min and the linear gradient was 90 min (0-55 min, 8% B; 55-79 min, 30% B; 79-80 min, 50% B; 80-90 min, 100% B; mobile phase A = 0.1% FA in water and B = 80% ACN/0.1% FA in water). Full MS scans were acquired in the mass range of 300 -1,600 m/z with a mass resolution of 70,000 and the AGC target value was set at 1e6. The ten most intense peaks in MS were fragmented with higher-energy collisional dissociation (HCD) and analyzed by MS/MS. MS/MS spectra were obtained with a resolution of 17,500 with an AGC target of 2e5 and a max injection time of 80 ms. Q-E dynamic exclusion was set for 15.0 s and run under positive mode.

Peak lists were generated from raw data files and searched against the Uniprot Human Protein Database using MaxQuant (Version 1.3.0.5). All peptides and proteins were filtered with false discovery rate (FDR) below 0.01. Label-free protein quantification was carried out using LFQ intensities by MaxQuant 1.5.2.8, and overlapped proteins between replicates were used for the following analysis.

A 250 μL mixture of methanol and water (7/3, vol/vol) was added to each sample. QC samples were prepared by mixing aliquots of all samples for a pooled sample. An Acquity UHPLC system (Waters Corporation, Milford, USA) coupled with an AB SCIEX Triple TOF 5600 System (AB SCIEX, Framingham, MA) was used to analyze the metabolic profiling in both ESI positive and ESI negative ion modes. An Acquity UPLC BEH C18 column (1.7 μm, 2.1 × 100 mm) was employed in both positive and negative modes. The binary gradient elution system consisted of (A) water (containing 0.1% formic acid, v/v) and (B) acetonitrile (containing 0.1% formic acid, v/v), and separation was achieved using the following gradient: 0 min, 5% B; 2 min, 20% B; 4 min, 60% B; 11 min, 100% B; 13 min, 100% B; 13.5 min, 5% B and 14.5 min, 5% B. The flow rate was 0.4 mL/min and the column temperature was 45°C. All samples were kept at 4°C during analysis. The injection volume was 5 μL. Data acquisition was performed in full scan mode (m/z ranges from 70 to 1,000) combined with IDA mode. Parameters of mass spectrometry were as follows: ion source temperature, 550°C (+) and 550°C (−); ion spray voltage, 5,500 V (+) and 4,500 V (−); curtain gas of 35 PSI; declustering potential, 100 V (+) and −100 V (−); collision energy, 10 eV (+) and −10 eV (−); and interface heater temperature, 550°C (+) and 600°C (−). For IDA analysis, the range of m/z was set as 50–1,000, and the collision energy was 30 eV. Peak lists were generated from raw data files and searched against the Uniprot Human Protein Database using MaxQuant (Version 1.5.2.8).

The Gene Ontology (GO) processes and kyoto encyclopedia of genes and genomes (KEGG) pathways of proteomics data were enriched using the Metascape web-based platform (33). KEGG pathways of metabolomics data were enriched by MetaboAnalyst 5.0 (34). The principal component (PC) contribution plots were generated using the factoextra (v.1.0.7) package and only the top 25 contributions were displayed in the contribution plot. R statistical software (v.4.1.2) was used for the correlation analysis. The P value of the correlation coefficient was computed by the corPvalueStudent function with the WGCNA (v.1.71) package. Networks were created by Cytoscape (v.3.9.1) software (35).

Cells were stained with the following human mAbs purchasing from Biolegend for FACS: anti-CD45 conjugated with fluorescein isothiocyanate (FITC), anti-CD3 conjugated with allophycocyanin-Cy7 (APC-Cy7), and anti-CD56 conjugated with phycoerythrin (PE). CellROX Deep Red Reagent was purchased from Invitrogen as reactive oxygen species (ROS) dye. Hoechst stain was purchased from Beyotime. FACS staining was performed according to the manufacturer’s instructions. The data were analyzed using FlowJo software (Version 10).

We conducted metabolomics analysis on pNK cells from blood (n=4) and dNK cells from humans in the first trimester of pregnancy (n=4) to assess metabolic differences between dNK and pNK cells ( Figure 1 ). The orthogonal partial least squares discrimination analysis (OPLS-DA) clearly distinguished dNK from pNK cells ( Figure 2A ). Metabolomics detected 77 differential metabolites (variable importance in projection (VIP) > 1, p value < 0.05) and showed that differential metabolites were mainly composed of lipids, organic acids, nucleosides, and benzenoids ( Figure 2B ). We found 17 metabolites that were upregulated (fold change > 1) and 60 metabolites that were downregulated (fold change < 1) ( Figure 2C ).

To illustrate alterations in different classes of metabolites, we displayed all metabolites in a bubble plot ( Figure 2D ). The bubble plot revealed that the majority of elevated metabolites were found in lipids and lipid-like compounds, particularly fatty acyls and glycerophospholipids, pointing to complex alterations in lipids between pNK and dNK cells ( Figure 2D ). Together, these findings shed light on the differences in metabolite levels between pNK and dNK cells.

To characterize the differential metabolic pathways between dNK and pNK cells, we performed pathway enrichment analysis of the differential metabolites and observed that glycerophospholipid metabolism, glutathione metabolism, purine metabolism, glycerolipid metabolism, and glycosylphosphatidylinositol (GPI)-anchor biosynthesis were all significantly changed in dNK cells ( Figure 3A ). Glycerolphospholipids, which included 24 differential metabolites, were the most abundantly altered class of metabolites so we analyzed glycerophospholipid subgroups to further determine the alterations. The radar map shows counts of upregulated/downregulated metabolites in these subgroups ( Figure 3B ). Most subgroups had more downregulated metabolites in dNK cells compared to pNK cells except phosphatidic acid (PA). For metabolites contained in each subgroup, phosphatidylcholine (PC), lysophosphatidylcholine (LPC), phosphatidyl-ethanolamine (PE), and lysophosphatidylethanolamine (LPE) were significantly reduced in dNK cells and phosphatidylinositol (PI), phosphatidylserine (PS), and phosphatidylglycerol (PG) showed a trend of reduction ( Figure 3C ), indicating downregulation of glycerophospholipid metabolism in dNK from pNK cells.

A previous study showed that PS exposure was proportional to the degree of NK cell activation during the NK cell activation process (36), suggesting that the reduced cytotoxicity may be linked to the decreased content of glycerophospholipids in dNK cells. In addition to the increase of PA, individual lipids in PG, PC, and PI, such as PG (16:0/0:0), PC (16:0/3:1(2E)), and PI (16:0/16:0) were significantly upregulated in dNK cells ( Supplementary Figure 1 ). Plotting chain carbon and double-bonds showed that upregulated glycerophospholipids tended to have fewer double bonds and downregulated glycerophospholipids might have more fatty acid chain carbons in dNK cells ( Figure 3D ). This finding demonstrated that the level of unsaturation of glycerophospholipids was dramatically reduced, implying that dNK cell membrane fluidity may be affected (37). In summary, these results indicate considerable alterations of complex lipids between dNK and pNK cells.

Strong correlations among metabolites suggest that the metabolites have similar roles and belong to related metabolic networks (38). To find potential co-regulatory relationships of differential metabolites, we performed Spearman correlation analysis and constructed a correlation network for all metabolites. In total, 85% of differential metabolites were significantly correlated with each other ( Figure 4A ). Notably, most differential metabolites had strong positive correlations with a threshold of absolute correlation coefficient greater than 0.95 ( Figure 4B ). There was a strong correlation between metabolites in the same metabolic pathway such as glycerophospholipid metabolism ( Figure 4B ). Furthermore, different pathways such as glutathione metabolism and purine metabolism seemed to have a significant positive correlation ( Figures 4A, B ). Overall, our survey demonstrated that the metabolic alterations between dNK and pNK cells are highly coordinated.

To further explore the comprehensive metabolic alterations between dNK and pNK cells, we plotted the schema of metabolic pathways ( Figure 4C ). Previous research has validated that glycolysis and oxidative phosphorylation (OXPHOS) are crucial to immune metabolism in NK cells (39–41). Our data demonstrated that glycolysis and the tricarboxylic acid cycle (TCA) cycle have a slight disturbance, indicating that glycolysis and OXPHOS are not vital in metabolic distinctions between pNK and dNK cells ( Figure 4C ). In glycerophospholipid metabolism, PA significantly accumulated in dNK cells ( Figure 4C ). Most metabolites of glycerophospholipid metabolism significantly decreased and affected pyruvate metabolism, leading to significant downregulation of the downstream glutathione pathway ( Figure 4C ). In addition, decreased NADPH could not donate enough electrons to reduce glutathione (GSH) from glutathione disulfide (GSSG), implying a redox balance shift in dNK cells ( Figure 4C ). Sphingomyelin in sphingolipid metabolism showed a significant increase in dNK cells ( Figure 4C ), consistent with a previous report showing that sphingolipid metabolism is a crucial signal for decidualization (42). Taken together, these results highlight interrelated metabolic network modifications in dNK cells compared to pNK cells.

The above results indicated that metabolism changed significantly between dNK and pNK cells, so we also collected normal dNK and pNK cells and performed proteomics to identify the protein alterations. In total, 2,724 proteins were found and 394 were identified as differentially expressed proteins (DEPs, p-value < 0.05, fold change > 3 or fold change < 1/3) ( Supplementary Figure 2A ). We found 232 upregulated DEPs (fold change > 3) and 162 downregulated DEPs (fold change < 1/3) in dNK cells compared to pNK cells ( Supplementary Figures 2A, B ). Principal component analysis (PCA) showed significant variance in proteomes between dNK and pNK cells, which had a separation in PC1 of 98.3% ( Figure 5A ). We focused on PC1 and displayed the top 25 DEPs that contributed to it ( Figure 5B ). Furthermore, Gene Ontology (GO) enrichment analysis of the DEPs contributing to PC1 (loading > 0.95 or loading < -0.95) showed that DEPs related to oxidoreductase activity were enriched ( Figure 5C ), suggesting that redox equilibria have shifted in dNK cells. In addition, for all GO-enriched DEPs the term oxidoreductase activity was upregulated ( Supplementary Figure 2C ). Some enriched GO terms (supramolecular fiber organization, GTPase regulator activity and actin binding) displayed alterations in the cytoskeleton ( Supplementary Figure 2C ), consistent with previous research showing that dNK cells failed to polarize their microtube to the synapse in contrast to pNK cells (43).

To investigate modified molecular pathways in dNK and pNK cells, we performed KEGG enrichment analysis of all DEPs. NK cell mediated cytotoxicity was the most significantly downregulated pathway in dNK cells ( Figure 5D ). Pathway protein processing in the endoplasmic reticulum had notable upregulation because of active cytokine production in dNK cells ( Figure 5D ). Additionally, leukocyte transendothelial migration appeared to be downregulated because dNK cells are tissue-resident ( Figure 5D ). Apart from the well-known proteins PRF1, CXCR4, HSPA2, etc., most DEPs involved in those three pathways have enormous potential to be investigated in NK cells in the future ( Supplementary Figure 2D ). The most surprising aspect of the KEGG enrichment analysis was DEPs enriched in a set of metabolic pathways, including arginine and proline metabolism, biosynthesis of amino acids, fatty acid metabolism, glutathione metabolism, and glycolysis/gluconeogenesis ( Figure 5D ). Additionally, 110 DEPs were involved in focal adhesion, cytoskeleton remodeling, oxidoreductase activity, and fatty acid metabolism based on pathway enrichment analysis ( Figure 5D , Supplementary Figure 2C ). Next, we plotted protein-protein interaction (PPI) networks to find interactions between the NK cell mediated cytotoxicity pathway and metabolic pathways ( Figure 5E ). The PPI network showed that ANPEP and TFRC had strong connections with DEPs in the NK cell mediated cytotoxicity pathway, indicating that metabolism of fatty acids and glutathione influenced NK immune function ( Figure 5E ). Overall, our results indicated various proteome profiles that revealed phenotypic alterations in pNK and dNK cells.

To understand the comprehensive inner network alterations between dNK and pNK cells, we combined metabolomic and proteomic datasets. The amounts of metabolites altered more than just proteins in glycerophospholipid metabolism, suggesting that upstream metabolites affect more than proteins ( Figure 6A ). Upregulation of GPD2 might result in production of PA, resulting in significant accumulation of PA ( Figure 6A ). Apart from this, glutathione metabolism was altered both in protein and metabolite levels ( Figure 6A ). Increased GSH might enhance cytotoxic functions in NK cells (44–46). We found that lactoylglutathione considerably decreased despite little change in glycolysis or pyruvate metabolism, suggesting that the glutathione source may be limited ( Figures 5C , 6A ). Glutathione peroxidase 1 (GPX1) was increased in dNK cells, suggesting a compensatory elevation for inner redox environment disruption ( Figure 6A ).

As mentioned above, redox equilibrium might shift in dNK compared to pNK cells. We further explored the correlation between selected DEPs and all differential metabolites. Proteins in the reactive oxygen species (ROS) metabolic process exhibited a negative correlation with common metabolites including glutathione metabolism, implying that proteins in oxidation-reduction (redox) reactions increased activity related to downregulated metabolites ( Figure 6B ). Metabolic reactions are inseparable from redox reactions (47). GSH and GSSG is the major redox pair in cells (48), and the ratio of GSH/GSSG is often used as an indicator of the cellular redox state (49). A notable aspect of our results was that glutathione metabolism was significantly downregulated, which included five downregulated metabolites ( Figure 3A ). Both GSH and GSSG significantly decreased and there was also a downregulated trend in the ratio of GSH/GSSG ( Figure 6C ), indicating alterations in amounts of glutathione metabolism and an imbalance in the conversion of GSH and GSSH. These findings pointed to significant modifications in the redox metabolic pathways between dNK and pNK cells. ROS is the kernel of redox systems and GSH is the major antioxidant for eliminating ROS (44). Therefore, it is reasonable to speculate that dNK has more ROS than pNK cells. To validate this speculation, we detected the ROS contents in dNK and pNK cells using FACS. The data showed that ROS in dNK cells were significantly higher than in pNK cells ( Figure 6D ).

To demonstrate that altered metabolites are interrelated with NK cell functions, we also plotted the correlation chord between proteins in NK cell mediated cytotoxicity and common differential metabolites ( Figure 6E ). The chord chart revealed that cytotoxicity was proportional to metabolite alteration ( Figure 6E ), implying that low cytotoxicity was due to the reduced level of whole cellular metabolites. LCK and PRF1, which were shown to affect cytotoxicity in NK cells, were both significantly downregulated, consistent with the reduced trend of whole metabolite alteration in dNK cells ( Supplementary Figure 2D ). Together, our metabolomics and proteomics data highlight that the inner redox state may affect dNK cell functions.