Increased ferroptosis in leukocytes from preeclamptic women involving the long non‐coding taurine upregulated gene 1 (TUG1)

Ferroptosis plays a key role in placental development and physiology, and abnormal ferroptosis has been implicated in trophoblast injury leading to preeclampsia (PE). We hypothesize that leukocytes isolated from PE exhibit increased ferroptosis and that extracellular vesicles contain long non‐coding (lnc) RNA/mRNAs that modulate oxidative stress and iron toxicity in vascular endothelial cells.


Introduction
Ferroptosis refers to non-apoptotic programmed cell death mediated by the iron-dependent lipid peroxidation of cell membranes [1].Dysregulated ferroptosis has been implicated in disease states, including cancer [2] and cardiovascular disease [3,4].Ferroptosis also plays a key role in placental development and physiology, and abnormal ferroptosis has been implicated in trophoblast injury [5], leading to pregnancy complications such as preeclampsia (PE) [5][6][7].Ferroptosis is regulated by multiple metabolic signaling pathways, including cellular oxidative redox status and iron metabolism, and is triggered by exposure to excess iron or agents that increase extracellular reactive oxygen species (ROS).Iron absorption increases during pregnancy, and little is known about the mechanisms regulating maternal iron levels and transfers through the placenta.In the trophoblasts decreased levels of erythroferrone, a hormone that increases plasma iron availability by suppressing hepcidin to facilitate hemoglobin and erythropoiesis was found in early-onset PE [8].However, the maternal plasma erythroferrone levels were increased in PE.Fetal hemoglobin has previously been reported increased in PE and alpha globin were discovered in syncytiotrophoblasts in PE, and induced by hypoxia [9].The two-stage model of PE is well known, the first stage with poor placentation and the second stage having the syncytiotrophoblast stress as the main feature that causes the maternal syndrome, including endothelial dysfunction [10].Studies to date on ferroptosis in PE have focused primarily on placental dysfunction; less is known about its impact on the vasculature.PE is characterized by the activation of maternal immune cells [11].Ferroptosis may act directly by affecting the number and function of immune cells or indirectly by triggering both innate and adaptive immune responses through the recognition of ferroptotic cells [12].
Women with overt PE dysregulated iron metabolism characterized by high circulating levels of iron, ferritin, and transferrin-saturation suggestive of iron overload [6], which correlates with increased systemic oxidative stress, are presented [13,14].Glutathione peroxidase 4 (GPX4) is an antioxidative enzyme that prevents the generation of lipid hydroperoxide caused by ROS and is considered the key regulator of ferroptotic cell death [1].T cells can undergo ferroptosis, thereby impairing the immune response to infection.For example, CD4+ and CD8+ T cells lacking GPX4 undergo rapid ferroptosis with an accumulation of lipid peroxides [15].Enhanced monocyte production of ROS has been demonstrated in PE [16], and exposures of endothelial cells to monocyte supernatants from PE women further enhance oxidative stress [17].Thus, an imbalance in favor of pro-ferroptotic mechanisms in immune cells could promote endothelial dysfunction.Excessive oxidative damage in the placenta of PE women may further enhance the leakage of cellular debris, including extracellular vesicles, containing a range of molecules such as mRNAs and non-coding (nc) RNAs (including microRNA) that are known regulators of ferroptosis [7].The extracellular vesicles in maternal blood may be derived from different sources, including the placenta-derived vesicles (syncytiotrophoblasts).Markers that are specific for trophoblast-derived extracellular vesicles origin are suggested to be placental-type alkaline phosphatase, and the microRNA cluster C19 [18][19][20].Transcription factor nuclear factor erythroid 2-like 2 (NRF2) and the long nc (lnc) RNA taurine upregulated gene 1 (TUG1), both of which regulate anti-ferroptosis factors [21,22], are dysregulated in the placenta of PE women and released in exosomes during cellular stress [23].PE-specific exosomes derived from damaged trophoblasts could modulate ferroptotic cell death in circulating maternal cells and organs contributing to endothelial dysfunction [24,25].Thus, immune cells in PE women could undergo ferroptosis through both maternal and placenta-derived factors.
We hypothesize that leukocytes from PE women exhibit increased regulators of ferroptosis and that extracellular vesicles from these women contain lncRNA/mRNAs modulating oxidative stress and iron toxicity leading to ferroptosis in vascular endothelial cells.To explore our hypothesis, we measured the expression of key regulators of ferroptosis in maternal leukocytes and extracellular vesicles as well as circulating markers of iron homeostasis, oxidative stress, and leukocyte activation in plasma collected from women with/without PE at different timepoints during pregnancy.For markers that were dysregulated, we assessed their temporal correlation with established markers of disease activity in PE (soluble fms-like tyrosine kinase-1 [sFlt-1]/placental growth factor [PlGF] ratio) as well as other markers that were dysregulated in PE women.For markers dysregulated in early pregnancy, we assessed their ability to predict the subsequent development of PE.

Materials and methods
The STORK study, a prospective longitudinal cohort study in which 1031 women of Scandinavian heritage with low-risk singleton pregnancies gave birth at Oslo University Hospital, Rikshospitalet, Oslo, Norway between 2002 and 2008, was followed throughout pregnancy [26].Exclusion criteria included the presence of one or more severe chronic diseases (such as pregestational diabetes, lung, cardiac, gastrointestinal, and/or renal disease).None of the included individuals had any symptoms of acute infections, and none were using antibiotics or antiviral agents when included in the study.Each woman had four studyrelated antenatal visits at 14-16, 22-24, 30-32, and 36-38 weeks' gestation.In the current study, additional exclusion criteria included gestational diabetes mellitus.A substudy [27] including 215 normotensive controls and 38 PE women were included in this study, and women developing gestational diabetes were excluded to reduce the heterogeneity in the control group.In the ACUTE-PE study, there were 34 PE and 61 control subjects.Blood samples were collected at 14-16, 22-24, and 36-38 in the controls and 24-34 in the earlyonset group and 34-38 in the late-onset group [28].Measurements of brachial arterial systolic and diastolic blood pressure (BP) were made with an automated oscillometric technique (Dinamap ProCare 300-Monitor, Criticon, GE Medical Systems).Systolic and diastolic BP were assessed as the mean of three recordings.All clinical investigations were conducted in accordance with the principles enshrined in the Declaration of Helsinki.The study was approved by the Regional Committee for Medical Research Ethics of Southern Norway in Oslo, Norway (REK# S-01191a, S-07392a, and s-07425a).Written informed consent was obtained from all study participants.Preeclampsia PE was diagnosed by new-onset hypertension (sustained elevation in BP ≥140/90 mmHg) and significant proteinuria (urinary total protein/creatinine ratio >30 mg/mmol or +1 on urine dipstick), using old criteria as the cohort was collected between 2002 and 2008.In the STORK study, almost all cases (n = 35) were diagnosed after 34 weeks' gestation (late-onset PE), and three cases before 34 weeks (early-onset PE).The ACUTE-PE study included cases diagnosed after 34 weeks (lateonset PE [n = 23]) and (early-onset PE [n = 11]).No women were treated with low-dose aspirin as this recommendation was not yet implemented in Norway during the period of data collection.

Uterine artery pulsatility index
The ultrasound examinations in the STORK study were done using the same equipment for all participants (Acuson Aspen).The mean uterine artery pulsatility index (PI) was obtained by bilateral Doppler flow velocity measurements using an abdominal approach.The measurements were performed close to the crossing of the external iliac arteries.The insonation angle was as low as possible.PI is defined as the difference between the peak systolic flow and minimum diastolic flow velocity, divided by the mean velocity recorded throughout the cardiac cycle.The formula for PI = (peak systolic velocity-end diastolic velocity)/mean flow velocity.The PI was calculated as the mean of three heart cycles on each side.For each woman, we used the mean values of the right and left side.Some of the uterine artery PI data was missing (approximately 10 PE and 40 normal pregnancies in the largest validation (leukocyte) cohort).

Collection, storage, and RNA extraction of extracellular vesicles
Whole blood (8 mL) was drawn (after an overnight fast) directly into ml BD Vacutainer CPT Tubes (Becton Dickinson Vacutainer Systems) with sodium citrate additives at weeks 22-24 and 36-38.The tubes were centrifuged at room temperature in a horizontal rotor (swing-out head) for 20 min at 1800 relative centrifugal force with the use of a FICOLL Hypaque gradient centrifugation, and maternal plasma was collected from the top and above the mononucleated cell layer and stored at −80°C.The unthawed plasma stored at − 80°C was thawed at room temperature centrifuged at 3000g for 5 min (to eliminate residual cellular material, including platelet fragments, but still retain the vast majority of extracellular vesicles), and total RNA from the total pool of extracellular vesicles was isolated from 350 μL plasma, using the exoRneasy serum/plasma kit (Qiagen), as previously published [27].Briefly, the total procedure for isolating RNA from extracellular vesicles comprises two phases: extracellular vesicles purification and RNA isolation.In the EV purification stage, precentrifuged sample was mixed with Buffer XBP and bound to an exoEasy membrane affinity spin column (500 g, 1 min).The bound extracellular vesicles were washed with Buffer XWP (5000 g, 5 min) and then lysed with QIAzol (5000 g, 5 min).In the RNA extraction step, chloroform was added to the QIAzol eluate, and the aqueous phase was recovered and mixed with ethanol.Total RNA binds to the spin column, where it was washed three times and eluted.Purity and concentration of isolated total RNA were measured using a Nanodrop ND-1000 spectrophotometer (Thermo Fisher Scientific Inc.), and RNA integrity was investigated using an Agilent 2100 Bioanalyzer (Agilent Technologies).

Collection, storage, and RNA extraction of maternal leukocytes
Whole blood was drawn in CPT tubes as described before, and leukocytes (peripheral blood mononuclear cells, PBMC) were collected by pipetting the mononucleated cell layer and stored at −80°C.RNA was extracted using MagNa pure Isolation Kit (Roche Life Science) at weeks 22-24 and Magmax Isolation Kit (Applied Biosystems) at weeks 36-38, due to change in instruments at the laboratory, as previously published [29].

Quantitative real-time polymerase chain reaction
Reverse transcription was performed using a High-Capacity cDNA Archive Kit (Applied Biosystems) for PBMC RNA and miScript II RT Kit (Qiagen) for extracellular vesicle RNA.RNA quantification was performed using SYBR Green PCR Fast Mix (Quantabio) for PBMC and miScript SYBR Green PCR Kit (Qiagen) for extracellular vesicles using the standard curve method on an ABI Prism 7900 (Applied Biosystems).Primers for TUG1 RT 2 lncRNA qPCR Assay (LPH18394A), miRNA PCR Assay has-miR-515-5p (YP00204431), has-miR-518b (YP00204405), and Ce_miR-39_1 miScript Primer Assay were from Qiagen.Sequence-specific intron spanning oligonucleotide primers for IL10, GPX4, NRF2, hemoglobin subunit beta (HBB), and hemoglobin subunit alpha 1 (HBA1) were designed by NCBI Primer Blast (Supplemental file).Transcript expression levels were normalized to ACTB and GAPDH and expressed as relative RNA levels in PBMC and RPLP0 in extracellular vesicles.

Cibersort analysis
Deconvolution of cell-type proportions from the previous microarray data [27] was performed with CIBERSORTx using the validated LM22 signa-ture as Ref. [30].Non-PBMCs were excluded after deconvolution, and related cell populations were merged.

Statistical analysis
Data are expressed as mean ± SD when normally distributed and median (25th, 75th percentile) when skewed.Comparison of demographics between women with/without PE was performed using t-test or Mann-Whitney U test, depending on distribution, and chi-square test for categorical variables.Temporal changes in PBMC mRNA expression, plasma markers, and extracellular vesicle mRNA expression during pregnancy between control and PE women were assessed using linear mixed model analysis with subject as random effect and time and PE diagnosis as fixed effects (also as interaction) in addition to age.We further used linear mixed models to model the association between PE disease

Study population
The final study population included 38 women with PE and 215 normotensive controls from the STORK cohort and 34 women with PE and 61 normotensive controls from the ACUTE-PE study.The clinical and demographic characteristics of the cohort are presented in Table 1.We did not find any significant differences (p = 0.423) between the BMI of the father in groups of PE (n = 14) and controls (n = 91) (availably data on only half of the STORK cohort).

Decreased expression of ferroptosis markers in leukocytes in PE
Expression of NRF2 at weeks 22-24 and 36-38 (p < 0.001 and p < 0.01, respectively) and GPX4 at 22-24 weeks (p < 0.01) was significantly decreased in PE compared to controls (Fig. 1A), indicating increased ferroptosis in leukocytes.The hemoglobin genes, HBA1 and HBB, were significantly decreased at weeks 22-24 in PE patients (p < 0.001 and p < 0.01, respectively).The long ncRNA TUG1 was decreased at weeks 22-24 and 36-38 in PE patients (p < 0.001 for both).Similar results were observed when evaluating differences within groups of fetal sex (Table S2) We investigated the same markers in extracellular vesicles and found no significant difference in RNA expression between PE and controls (Fig. 1B).
As PE-specific extracellular vesicles derived from damaged trophoblasts could modulate ferroptotic cell death in circulating maternal cells and control for the inclusion of placental-derived extracellular vesicles in our experiment, we next assessed the expression of mir-515-5p and mir-518b in our isolated extracellular vesicles.As shown in Fig. S1, we detected these placenta-specific extracellular vesicles markers in our sample, with an increase during pregnancy, and very low levels in follow-up samples (i.e., 5 years after pregnancy).However, no differences between PE and control samples or correlation between these microRNAs and PBMC expression of the ferroptosis-related genes was observed (all p-values >0.4).

Increased circulating ferritin and AOPP in PE
Figure 2A shows the level of the sFlt-1/PlGF ratio, an established predictive risk marker for PE [31], in our study populations during pregnancy.Individual graphs for sFlt-1 and PlGF are shown in Fig. S2.In the STORK study (left panel), the sFlt1/PlGF ratio was increased in PE women at weeks 14-16 but with larger differences compared to a healthy pregnancy from weeks 30-32.Similar and larger differences were observed in the ACUTE-PE study, in particular in early-onset PE, taking into account the gestational age when samples were obtained.
Figure 2B shows markers of iron metabolism in our study populations at similar timepoints as shown for the sFlt-1/PlGF ratio.Circulating ferritin levels decreased during pregnancy in the STORK cohort but were significantly elevated in PE compared to controls from weeks 22-24 through the end of pregnancy (p < 0.05 at all timepoints).Although a similar temporal decline in ferritin was observed in healthy women in the ACUTE-PE study, we found no differences compared to earlyand late-onset PE women.Hepcidin levels had a similar temporal profile as ferritin, whereas transferrin receptor levels increased during pregnancy in both study populations.However, levels of both of these biomarkers were largely similar in PE and controls.Lower transferrin receptor levels were observed in PE women in the ACUTE-PE study, but taking into account the gestational age when samples were obtained, these differences seem minor.Finally, erythroferrone levels, a major regulator of hepcidin, displayed stable levels in both groups of women in both studies, during pregnancy.
The activities of the oxidative stress markers in the circulation were unchanged for SOD, whereas TBARS increased during pregnancy; however, there were no significant differences between PE and controls.However, AOPP, a marker  reflecting the oxidation of proteins, increased during pregnancy and PE women in the STORK cohort displayed higher levels throughout the study.A similar pattern was observed in the ACUTE-PE study, and although the levels in earlyonset PE were not statistically different compared to healthy women at 26-28 weeks, the levels appeared higher when evaluated against earlier and more comparable gestational age.Finally, plasma levels of GPX4, one of the ferroptosis markers showing decreased mRNA expression in PBMC from PE women, showed no temporal regulation during pregnancy, and levels in PE women were similar to healthy women.For both oxidative stress and iron metabolism markers, similar results were observed when evaluating differences within groups of fetal sex (Table S2).

Decreased expression of IL-10 in leukocytes in PE
To assess the inflammatory phenotype of the PBMC, we assessed mRNA expression levels of pro-inflammatory IL-6 and anti-inflammatory IL-10.Although expression levels of IL-6 were too low to be adequately quantitated in our cDNA, IL-10 mRNA levels were lower in PE women at 22-24, but not 36-38 weeks (Fig. 3A).
Adjustment for reference RNA should account for cellularity; however, information on the proportion of cell types could be informative.Although we did not have differential counts of our PBMC samples, we have performed a deconvolution based on a previous microarray study [27] to estimate cell type proportions with CibersortX, using PBMCs from 22-24 weeks in pregnancy.As shown in Fig. S3A, most of the cell type proportions were similar between controls and PE, however, lymphocytes, and in particular, the B cell population was lower in the PE group.
In addition, we analyzed activation markers for major leukocyte subsets in PBMC in both study populations.As shown in Fig. S3, leukocyte activation markers reflecting B cell (BAFF), monocyte/macrophage (sCD163), and T cells (sCD25) were similar in healthy and PE women in the STORK cohort apart from a minor increase in sCD25 at 14-16 weeks in PE women.In addition, we assessed MPO, which may primarily reflect activation of neutrophils but is also secreted by lymphocytes and monocytes and considered a marker of oxidative stress [32].Although MPO levels were similar in PE women and healthy controls in the STORK study, we found an increase in MPO in PE compared to controls in the ACUTE-PE study (Fig. S3E).Furthermore, we found an increase in BAFF in LPE and decrease in EPE compared to controls (Fig. S3C).

Associations between markers of ferroptosis and disease activity
Using mixed models, we next performed a temporal correlation between ferroptosis markers (Ferritin, GPX4, NRF2, TUG1, HBA1, and HBB), and markers of disease severity and other markers that were dysregulated in PE (i.e., ferritin, IL10 mRNA expression, AOPP, and uterine artery PI).A t-score above (positive association) 1.96 or below (negative association) −1.96 signifies a significant correlation.As shown in the heatmap in Fig. 3B, there were few associations detected between ferroptosis and the Flt-1/PlGF ratio or uterine artery PI.Notably, all ferroptosis markers correlated with IL10 mRNA expression in PBMC with varying degree in PE and healthy women.The strongest association was seen with TUG1 as shown in Fig. 3C.In addition, HBA1 and HBB correlated negatively with ferritin levels in healthy women.

Predictors of PE development
Finally, we evaluated the discriminatory ability of the RNA species evaluated in leukocytes at weeks 22-24, compared to the established plasma sFlt-1/PlGF ratio.As shown in Fig. 4A, ROC analysis revealed that all RNA markers had moderate discriminatory properties, whereas TUG1 had good properties with an AUC of 0.82 (0.76-0.89).This gave a sensitivity for maternal leukocyte TUG1 at 22-24 weeks of 90.3%, specificity of 97.4%, NPV of 63.3%, and PPV of 30.1% for the prediction of PE.Dichotomizing TUG1 RNA levels according to Youden's index indicated a 12.9-fold higher risk of developing PE with levels below this threshold in univariate analysis (Fig. 4B).Furthermore, this association was not influenced when including more established markers in multivariable analysis (Model 1).Evaluation of a more linear association with PE development using log transformed and standardized levels revealed a 4.2-and 5.1fold increase in risk per SD decrease in TUG1 in univariate and multivariable analysis (Model 2), respectively.

Discussion
In this study, we prospectively measured ferroptotic markers in maternal leukocytes and systemic indices of iron metabolism, oxidative stress, and leukocyte activation at timed intervals throughout pregnancy in relation to PE diagnosis and progression.Our major findings were as follows: (i) Maternal leukocytes from PE women displayed decreased lncRNA/mRNAs that modulate oxidative stress and iron toxicity and decreased expression of antiinflammatory IL-10; (ii) leukocyte expression of anti-ferroptotic markers correlated with IL-10, irrespective of diagnosis; (iii) PE women were characterized by increased levels of ferritin and the oxidative stress marker AOPP throughout pregnancy; and (iv) low lncRNA TUG1 RNA levels in leukocytes at 22-24 weeks were strongly associated with the subsequent development of PE.Taken together, our findings suggest that decreased antiferroptotic mechanisms are present in PE leukocytes.This does not appear to be related to maternal disease activity or plasma oxidative stress status but rather to attenuated anti-inflammatory expression in these cells.
The temporal decrease in ferritin and hepcidin levels and increase in transferrin receptor levels during pregnancy in our study is similar to previous reports [33][34][35] and explained by an increased fetal demand for iron.In our study, ferritin levels were modestly elevated in PE women, possibly due to hemoconcentration, although we had no data on hemoglobin or hematocrit to verify this.We did not detect differences in other measures of iron metabolism compared to healthy women, or between early-and late-onset PE as suggested by others [36,37].However, although the impact of systemic iron metabolism in ferroptosis is obscure [38], placental ischemia may also be a source of iron toxicity in PE [14].Indeed, gene expression profiling and functional enrichment at the placental level have identified distinct molecular pathways in early-and late-onset PE, with hypoxia-and iron-related pathways being selectively enriched in early-onset PE [39].The WHO recommends routine iron supplementation of 30-60 mg per day throughout pregnancy.In experimental models, iron supplementation reduced iron deposi-tion in the decidua in a spontaneous abortion model and improved pregnancy outcome.However, local iron overload caused ferroptosis in the decidual stromal cells by downregulating GSH and GPX4 levels [40].An exogenous supplement of iron increased the sensitivity of cells to ferroptosis inducers [41] and enhanced RSL3-induced ferroptosis [42] in vitro.Future studies should evaluate if iron-supplementation in pregnancy may impact ferroptosis in vivo.
A major finding in our study was the consistently higher AOPP levels, mostly reflecting the oxidation of albumin and fibrinogen [43], in PE women as reported by most [44][45][46], but not all previous studies [47].However, these studies obtained samples late in gestation or with differences in gestational age between comparison groups, which is not optimal based on the increase in AOPP during pregnancy observed in both our study populations.Thus, our study suggests that enhanced oxidative stress is present already at 14-16 weeks in women with late-onset PE.Interestingly, an increased expression of AOPP has also been observed in PE placentas [48] and shown to promote trophoblast damage and dysfunction by inhibiting the Nrf-2/ARE/HO-1 antioxidative pathway in these cells [49].Furthermore, PE trophoblasts exposed to AOPP, secrete extracellular vesicles that trigger inflammatory responses and pyroptosis in endothelial cells [49].The discrepancy between AOPP and the activity markers (i.e., TBARS and SOD) in our study could be due to the latter being more affected by long-term storage [50].Finally, as GPX4 seems to be relatively steadily expressed in most tissues, cell-specific regulation may not be reflected in circulating levels, which explain the lack of regulation of plasma GPX4 in our study.
A major discovery in our study was that leukocytes from PE women displayed markedly decreased mRNA expression of hemoglobin genes and lncRNA/mRNAs known to attenuate oxidative stress and ferroptosis.These findings are not explained by differences in cell numbers as levels were adjusted for cellularity using reference RNA.Furthermore, the lower levels do not seem to be driven by differences in activation status in leukocytes, as immune activation markers for relevant leukocyte subsets were similar in PE and healthy women.Experimental knockdown of GPX4, NRF2, and TUG1 may enhance oxidative stress and ferroptosis [51][52][53] and attenuate T cell immune response [15].Furthermore, singlenucleotide polymorphisms in the GPX4 gene have been associated with severe and early-onset PE [54,55], and lymphocytes with this genotype have higher levels of 5-lipoxygenase [56], which may contribute to the cellular pool of lipid peroxides that promote ferroptosis [57].Similarly, mice with a deletion in Nrf2 in leukocytes have enhanced oxidative stress and pro-inflammatory responses [52].Moreover, decreased lncRNA TUG1 may increase ferroptosis, partly via the NRF2 signaling pathway [53].Although we hypothesized that extracellular vesicles containing lncRNA/mRNAs that could modulate oxidative stress and iron toxicity would be regulated in PE women, we found no difference in ferroptosis-related transcripts in extracellular vesicles from PE compared to healthy women.Furthermore, we found no difference in the specific markers, mir-515-5p and mir-518b, from the placenta-derived extracellular vesicles in PE women compared to controls.However, this does not imply that these vesicles have no role in PE progression as their uptake in maternal cells could be different as well as their cargo.We speculate that decreased levels of these protective factors in PE leukocytes could promote placental oxidative stress and ferroptotic mechanisms potentially impacting spiral artery remodeling and trophoblast invasion.
Hemoglobin gene expression has been shown to increase in response to oxidative stress and hypoxia in non-erythroid cells and is believed to reduce it [58].At the placental level, HBA1 has been found increased in syncytiotrophoblasts in PE and suggested to be increased by hypoxia [9].The observed decrease in HBB and HBA1 gene expression levels in PE leukocytes in early pregnancy suggests that these cells may have a lower protection against oxidative stress and hypoxia early in pregnancy.Furthermore, the marked increase in HBB and HBA1 in women with and without PE in later pregnancy could reflect a compensatory mechanism in response to the increasing hypoxia and oxidative stress during pregnancy, irrespective of a PE diagnosis [58].Evaluations of expression profiles of HBB and HBA1 from the human protein atlas (https://www.proteinatlas.org/)reveal that in PBMC, these transcripts are almost exclusively expressed in B cells, potentially linking our findings to abnormal B cell regulation in PE [59].We did observe lower estimated B cell numbers in PE using Cibersort, as reported by others [60], and a differential regulation of the B cell factor BAFF, in early-and late-onset PE in the ACUTE-PE study.Furthermore, regulatory B cells are found to be enhanced in healthy pregnancies, where they contribute to establishing a tolerant environment through the production of the anti-inflammatory IL-10.Lower plasma IL-10 has been reported lower in both early-and late-onset PE at the time of diagnosis [61], and our finding of decreased IL-10 expression at 22-24 weeks in late-onset PE suggests the attenuation of IL-10 occurs quite early in pregnancy.As maternal immune cells are present at the fetal-maternal interface early in gestation [62], one hypothesis of PE development is that the overactivation of immune cells around the time of implantation stimulates trophoblast apoptosis leading to suboptimal placentation and PE in later pregnancy [63].Interestingly, the level of anti-ferroptosis markers correlated with IL-10 expression rather than maternal oxidative stress as reflected by AOPP levels, and we speculate that the lower anti-inflammatory potential of maternal immune cells from PE women could have early adverse effects on placental development.
Finally, we evaluated the discriminatory properties of HBA1, HBB, GPX4, NRF2, and TUG1 at 22-24 weeks in predicting the subsequent development of PE.All displayed modestly good AUC in ROC analysis.However, TUG1 was strongly associated with PE development (AUC of 0.82), and the association was augmented, rather than attenuated, in adjusted analysis including age, BMI, parity, MAP, and the sFlt-1/PlGF ratio.Thus, decreased TUG1 expression in maternal leukocytes seems to reflect a different pathophysiological impact on PE progression than conventional markers with comparable discriminatory properties as recently published work investigating cell-free RNA as an early predictor of PE [64].TUG1 has also been shown to be markedly reduced in placental samples from patients with PE, and its absence decreased cell proliferation, invasion, and angiogenesis in trophoblasts and endothelial cells in vitro [65][66][67][68] suggesting that TUG1 may have a protective function in PE.TUG1 has, as mentioned above, been shown to directly bind to NRF2 protein and may decrease ferroptosis, partly via the NRF2 signaling pathway [22].

Strengths and limitations
A limitation of our study is the lack of PBMC from early/severe PE women.As the STORK study focused on low-risk singleton pregnancies, the number of early PE cases was very limited and it was not feasible to obtain a sizeable number prior to a PE diagnosis.Moreover, other markers of iron imbalance and lipid peroxidation like transferrinsaturation, non-glycosylated ferritin, and hydroxyalkenals, respectively, have not been measured and are a limitation of the study.In addition, we did not have PBMC that could be used for FACS analysis that would have been given us details of the proportion of the different immune cells in the PE and control groups.Another limitation of the study is the relatively low number of PE women in the STORK cohort, although a representative percentage in a healthy population.Even so, our clinical cohort and diagnostic criteria are well described, with a non-selection bias during inclusion and consistent temporal sampling during pregnancy.Another limitation is the lack of data on asymptomatic infections in our cohort, which could influence mRNA expression in PBMC as well as markers of iron metabolism and oxidative stress.Emerging evidence shows that pathogenic infections are closely related to ferroptosis, especially viral infections [69].Ferroptosis occurs when the rate of intracellular lipid ROS production exceeds the cell's detoxification capacity.The importance of ferroptosis manipulation in pathogens' propagation, pathogenesis, or immune evasion is evident from their numerous tactics that control lipid ROS levels [70].
We have no data on iron supplementation, antimicrobials, or antihypertensive medication during pregnancy.Commonly prescribed heart medications might have previously unidentified antiferroptotic activity.For example, carvedilol, which is widely used to treat hypertension and heart failure, has been shown to block ferroptosis independent of its effect on β-adrenergic receptors, and the underlying mechanism might contribute to its capacity to scavenge lipid peroxides and chelate iron [71].

Conclusions
Ferroptosis has been implicated in trophoblast injury leading to PE.Our study shows that markers of ferroptosis are also significantly regulated in the leukocytes in early pregnancy in women who go on to develop PE which could provide novel insights into the pathogenesis of this dangerous and poorly understood disease.The lncRNA TUG1 is shown dysregulated in the placenta of PE women, and in current study, we show that it is also dysregulated in the leukocytes and could be used as an early predictive marker of PE.The predictive value of TUG1 should be further explored in independent forthcoming studies, also at earlier gestations.

Table 1 .
Clinical and demographic characteristics of the study population.
Note: Data given as mean ± SD when normally distributed and median (25th, 75th percentile) when not normally distributed.Abbreviations: BMI, body mass index; MAP, mean arterial pressure; PE, preeclampsia, PI, pulsatility index.a At blood sampling.index was identified and utilized in logistic regression with adjustment for age, BMI, mean arterial pressure (MAP), parity, and sFlt-1/PlGF.A similar model using log10 transformed and normalized TUG1 (i.e., log10TUG1/SD) was also evaluated.