Patients (n=96) were recruited at the CHUSJ either at the time of delivery with uncomplicated pregnancies (N=54, CTL) or when admitted for PPPE (N=42) and gave written informed consent. PPPE was defined as de novo hypertension (140/90mmHg) with end-organ damage, most commonly proteinuria (≥300mg/day), in the postpartum period (48h to 25 days after birth following an uncomplicated pregnancy) (13, 15). Demographic and obstetrical data were obtained through clinical chart review. Maternal blood samples were collected 24 hours after delivery (CTL) or at the time of diagnosis (PPPE). Approximately 8 ml blood was collected in 10mL K2 EDTA tubes (ThermoFisher) and processed immediately. From whole blood, 150µl was transferred to a round bottom tube for immunolabelling (see below) and the remaining blood was centrifuged at 845g (10 minutes, 4°C) and plasma was collected, aliquoted, and stored at −80°C. The residual blood was used for peripheral blood mononuclear cells (PBMC) isolation (see below).

Residual blood was diluted with PBS-10%FBS in a 1:1 ratio and added to Sepmate-50 tubes (Stemcell) containing 15mL Lymphoprep density gradient medium (StemCell). Tubes were centrifuged at 1200g for 10 minutes. PBMCs were collected and diluted with PBS for a final volume of 45mL and centrifuged at 300g for 10 minutes. Supernantants were removed and the pellet was resuspended with 10mL of PBS. A 10µl aliquot was taken and combined with 10µl trypan blue solution (Bio-Rad) for cell count on a Bright-Line hemacytometer (Hausser Scientific). The resuspended pellet was centrifuged at 400g for 10 minutes and the supernatant was discarded. Pellets were resuspended with FBS-10%DMSO at a final concentration of 2M cells/mL and aliquoted to be frozen down slowly at -80°C for 24-48 hours and then transferred to liquid nitrogen for long term storage.

Whole blood (150μl) was immunolabeled for 15 minutes at room temperature using the following antibodies [with catalog numbers; concentration]: 5μl for each of the following APC-Cy7-CD4 [557871; 0.2mg/ml], PerCP-Cy5.5-CD8 [560662; 0.05mg/ml], Alexa Fluor 700-CD14 [557923; 0.2mg/ml], BV605-CD25 [562660; 0.025mg/ml], PE-Cy7-CD196 [560620; 0.2mg/ml], BV421-CD194 [562579; 0.2mg/ml], PE-CF594-CD183 [562451; 0.2mg/ml], PE-CD127 [557938; 0.2mg/ml], V500-CD19 [561121; 0.2mg/ml] and 20μl of the following: FITC-CD3 [555339; 0.05mg/ml], and APC-CD56 [555518; 0.025mg/ml] (all from BD Biosciences) and 35μl buffer solution (BD Canada). After the immunolabelling was completed, red blood cells were lysed (0.06M NH4Cl, 4mM KHCO3, 0.052mM EDTA), and peripheral blood mononuclear cells were washed with PBS, PFA-fixed, and analyzed using a flow cytometer (LSR Fortessa, BD Biosciences). A minimum of 300,000 events were acquired per sample. Data analysis was performed using FlowJo (Tree Star Inc).

Intracellular cytokine staining was performed on a subset of our patients to assess immune cell mediator production. PBMCs were thawed and resuspended in RPMI/10% FBS (Wisent). Cells were plated in 96-well plates for a final count of 1 million/well, and treated for 4h (37°C, 5% CO₂) with one of the following: untreated (PBS/1%FBS), lipopolysaccharide (LPS from E. coli, 100ng/ml), monosodium urate crystals (MSU, 100mg/ml) or phorbol myristate acetate (PMA, 50ng/ml). After 1h of incubation, GolgiPlug (BD Biosciences) was added to wells. Cells were washed and incubated with the extracellular antibody cocktail [with catalog numbers; concentration] including 5μl of each of the following: (anti-CD3 [564001; 0.1mg/ml], anti-CD4 [557871; 0.2mg/ml], anti-CD8 [560662; 0.05mg/ml], anti-CD14 [557923; 0.2mg/ml], and anti-CD16 [563690; 0.2mg/ml] (all from BD Biosciences), viability dye (ThermoFisher) and brilliant stain buffer (BD Biosciences) for 20 minutes at room temperature. Cells were washed and were fixed and permeabilized in 200μl Cytofix/Cytoperm solution (BD Biosciences) for 20 minutes at room temperature, followed by a washing step with Perm/Wash (1:10 dilution; BD Biosciences). Cells are then incubated with an intracellular antibody cocktail [with catalog numbers] including 5μl of each of the following (anti-IL-10-PeCy7 [501420; 0.02mg/ml], anti-IFNγ-Pe [506507; 0.05mg/ml], anti-IL-1β-FITC [508206; 0.05mg/ml] (Bio legend), and anti-TNFα-BV421 [562783; 0.2mg/ml], anti-IL-12-APC [554576; 0.2mg/ml] (BD Biosciences), with brilliant stain buffer (BD Biosciences) for 20 minutes at room temperature. Cells were washed, resuspended in PBS/1%FBS, and passed by Flow Cytometry (LSR Fortessa, BD Biosciences) 300,000 events were acquired per sample. Gating was performed on live cells and data analysis was performed using FlowJo (Tree Star Inc).

Plasma was analyzed for a panel of immune mediators via multiplex analysis (Human Cytokine 48-Plex Discovery Assay, Eve Technologies). This panel included the following mediators: sCD40L, EGF, Eotaxin, FGF2, Flt3 ligand, Fractalkine, GCSF, GMCSF, GROA, IFNA2, IFNγ, IL1α, IL1β, IL1RA, IL2, IL3, IL4, IL5, IL6, IL7, IL8, IL9, IL10, IL12p40, IL12p70, IL13, IL15, IL17A, IL17E, IL17F, IL18, IL22, IL27, IP10, MCP1, MCP3, MCSF, MDC, MIG, MIP1α, MIP1β, PDGFAA, PDGFAB/BB, RANTES, TGFα, TNFα, TNFβ, VEGFA.

PBMCs were sequenced using 10X Chromium Next GEM Single Cell 3’ Reagent Kits v3.1 (dual index) with feature barcode technology for cell multiplexing at the McGill Genome Center using the Illumina NovaSeq6000 S4 v1.5, PE100 (16). The Chromium protocol was followed, including GEM generation and barcoding, followed by GEM-RT cleanup and cDNA amplification, gene expression library construction, and lastly, cell multiplexing library construction and sequencing. Libraries were prepared to obtain 5000 cells/patient and to sequence approximately 30,000 reads/cell. Files were demultiplexed using CellRanger (17) pipeline version 7.0.1 default parameters, whereby only cell-associated barcodes assigned exactly one cellular multiplexing oligo (CMO) were assigned to a sample and with reference to the Homo sapiens genome, build GRCh38 2020-A.

Data were analyzed with the ROSALIND® platform (https://rosalind.bio/), with HyperScale architecture. Quality scores were assessed using FastQC (18). Cell Ranger was again used to count UMIs, call cell barcodes, and perform default clustering. Individual sample reads were normalized via Relative Log Expression (RLE) using DESeq2 R library (19). Read Distribution percentages, violin plots, identity heatmaps, and sample MDS plots were generated as part of the QC step using RSeQC (20). DEseq2 was also used to calculate fold changes and p-values and perform optional covariate correction. The clustering of genes for the heatmap of differentially expressed genes was done using the PAM (Partitioning Around Medoids) method using the fpc R library (21). Final clustering was done using Seurat graph-based method. Hypergeometric distribution was used to analyze the enrichment of pathways, gene ontology, domain structure, and other ontologies. The topGO R library (22) was used to determine local similarities and dependencies between GO terms to perform Elim pruning correction. Enrichment was calculated relative to a set of background genes relevant to the experiment. BioPlanet and biological processes from the GO terms were the only pathway catalogs used for our analyses.

Placental biopsies, at least four, were obtained and included the entire thickness of the placenta. For immunohistochemistry, sections were processed as previously described (23). Briefly, sections were deparaffined and incubated with primary antibody overnight at 4°C. The following antibodies were used (dilution; concentration; catalog number, company): mouse anti-CD163 (1:100; 0.2mg/100μl; MCA1853, Bio-Rad), mouse anti-HO2 (1:50; 0.74mg/mL; NBP2-01407, Novus Biologicals), and rabbit anti-LC3b (1:100; 1mg/mL; NBP2-46892, Novus Biologicals). Sections were then washed with PBS, exposed to the matched horseradish peroxidase (HRP)-conjugated secondary antibody (goat anti-mouse IgG-HRP (1:100; 1mg/mL; 1706516, Bio-Rad), or goat anti-rabbit IgG-HRP (1:100; 1mg/mL; 1706515, BioRad), for 45 minutes at room temperature, and washed before the signal was revealed using DAB (Sigma) and counterstained with hematoxylin. Slides were imaged using a microscope (Revolve, Discover Echo), where 6–8 pictures were randomly taken within the placental villi or captured using a slide scanner (Axioscan, ZEISS). Image analysis was performed using ImageJ (NIH Image). For immune cells, counting was performed by a blinded individual, and data were reported as the number of positive cells by 1.2×10⁵μm² (24).

Receiver Operating Characteristic (ROC) curve analysis was employed to assess individual prenatal biomarkers for their predictive utility in PPPE outcome prediction. To explore the synergistic potential of 2 biomarkers together, logistic regression modeling was employed to create a combined predictive model. The ROC curve visually represents the trade-off between true positive rate (sensitivity) and false positive rate (1-specificity) across varying classification thresholds.

Data are presented as mean ± standard error of the mean (SEM). As appropriate, data were analyzed using the Mann-Whitney or Fisher’s exact test using GraphPad Prism, version 10.0.3 (GraphPad Software). A probability (p) value of <0.05 was considered statistically significant. A batch correction was performed for cluster comparisons, and the statistical cut-off for differentially expressed genes was p.adj<0.05, and a fold change≥ ± 1.5 and a p.adj<0.05 was used for pathway analyses.

This work was supported by grants from Canadian Institute of Health Research - CIHR to SG and CHG, Mayo Clinic to SG, and Scholarships from the Quebec Research Network in Perinatal Determinants of Children Health and the Fonds de Recherche du Quebec – Santé to CC. Funding sources were not involved in any part of this work’s study design, analysis, interpretation, and writing. Approval was obtained from the Centre Hospitalier Universitaire (CHU) Sainte-Justine (CHUSJ) Ethics Board for patient recruitment (IRB 2015-840) in Montreal, QC, Canada, and for experiments and data analysis at the Mayo Clinic (IRB 21-012205) in Rochester, MN, USA. Patients gave written informed consent either when they were recruited either at the time of delivery with uncomplicated pregnancies or when admitted for PPPE.

Ninety-six women were included in this study, all with uncomplicated pregnancies delivered at term (gestational age at delivery: 39.6 ± 0.13 CTL vs 39.1 ± 0.16 PPPE). Demographic and obstetrical characteristics are shown in Table 1 . PPPE patients were more likely to be of black race (57.1% vs. 24.1%, p<0.01), had higher pre-pregnancy body mass index (BMI, 28.4 ± 1.0 vs. 24.9 ± 0.6 kg/m², p<0.01), were more likely to have a personal and family history of hypertension and PE (33.3% vs. 1.8%, p<0.001 and 57.1% vs. 27.8%, p<0.01 respectively), and had lower gestational age at delivery (39.1 vs. 39.6 weeks, p<0.05) compared to controls. Both groups had similar maternal age, rate of primiparity, birth weight, and incidence of gestational diabetes mellitus during their pregnancy. Additionally, both groups had equal proportions of labor inductions, cesarean sections, and the proportion of male/female newborns. For women diagnosed with PPPE, the mean postpartum presentation day was 7.4 (range of 2-25 days), with the most common secondary symptoms presented in Supplementary Figure S1 .

To deepen our understanding of the transcriptional profiles of immune cells in PPPE, we performed single-cell RNA-seq on whole PBMCs from a subset of control (n=5) and PPPE (n=6) patients ( Figure 1A ). Using pseudo-bulk Seurat graph-based clustering, our analyses uncovered 16 cell clusters, with no clustering based on fetal-sex or race (data not shown) and for which proportions were similar between patient groups and which were identified using known immune markers ( Figures 1B–D ).

To address changes in the maternal circulating immune system, we also analyzed the proportions of circulating PBMCs via flow cytometry from a larger subset of our patients. Here we found changes in the proportions of CD3- cells, which were decreased in PPPE (45.0 ± 2.2% vs. 52.6 ± 1.8%, p<0.05), mainly driven by the large decrease in monocytes (35.8 ± 2.8% vs. 52.2 ± 1.3%; p<0.001); whereas NK cell subsets were increased (13.1 ± 1.1% vs. 9.6 ± 0.9%, p<0.05; Figures 5A–C ). Other CD3- subsets, such as B lymphocytes, remained unchanged (data not shown). Reciprocally, CD3+ cells were increased in PPPE (54.71 ± 2.2% vs. 47.25 ± 1.8%, p<0.05) and their CD8+ T cells subset tended to being decreased (31.21 ± 1.9% vs. 35.47 ± 1.3%; p=0.0598; Figures 5D, E ). Other CD3+ subsets, namely NKT and CD4+ T cells (including T-regulatory and T-helper cells), remained unchanged (data not shown). No correlations were observed between the percentage of any immune cells studied with the number of days postpartum ( Supplementary Figure S2 ), supporting that the observed changes are due to the pathology itself.

Immune cells function primarily via the production of cytokines – we thus investigated the cytokine production profile from monocytes following exposure to classical stimuli. At baseline (unstimulated) and under LPS (pathogen-derived) stimulation, the proportion of CD14+ cells producing IL1β, IL10, IL12, and IFNγ, and the level of their production (mean fluorescence intensity - MFI), did not differ between CTL and PPPE patients (data not shown). However, under the stimulation with phorbol myristate acetate (PMA), a diminished proportion of cells from PPPE patients were found positive for pro-inflammatory cytokines IL1β and IL12 (p<0.01 and p=0.068, respectively), while an increased proportion was positive for the anti-inflammatory cytokine IL10 (p<0.01; Figure 6A ). Similarly, exposing the cells to sterile-inflammatory stimuli, namely monosodium urate crystals (MSU), shown to be involved in PE (25–28), we found that fewer CD14+ cells produced IL12 (p<0.01, Figure 6B ), all supportive of a dampened inflammatory response.

T-cell subsets showed different cytokine production profiles in both in vitro LPS stimulated and unstimulated conditions in PPPE. There was a consistent trend to decreased numbers of IL12-producing CD4+ lymphocytes in PPPE patients regardless of stimulation (p=0.068, p=0.076, and p<0.05 respectively; Figures 7A–C ). CD4+ cells also had a diminished capacity to produce IFNγ (p<0.05, Figure 7D ) although the number of IFNγ-producing cells was unchanged. On the other hand, the number of IFNγ-producing cells was decreased in the CD8+ cell population at baseline (p<0.01; Figure 7E ). Overall, the response in CD8+ cells was not as consistent as in CD4+ cells. Importantly, the proportions of cells producing IL10 were altered in a different direction based on the type of treatment. LPS caused a tendency to have a higher proportion of CD8+ cells producing IL10 (p=0.071; Figure 7F ), whereas MSU led to a diminished proportion of positive cells (p=0.066; Figure 7G ).

As there is a net decrease in the proportion of monocytes and T-cells producing inflammatory cytokines, and their overall levels of production, we sought to investigate if PPPE immune cells displayed a phenotype, similar to what is observed in PE, of exhaustion (29–32). Using pseudo-bulk comparisons of our sequencing data presented above, we cross-referenced specific exhaustion markers, including inhibitory receptors and senescence markers. We found that immune cells from PPPE patients have decreased LMNB1 and increased KI67, PD1, TCF7L1, KLRG1, LAG3, TIGIT, CTLA4, and CD57, supporting an exhaustion phenotype in circulating immune cells ( Figure 8 ).

Considering the important shift in the profile of immune cells at the transcriptomic and functional (i.e., cytokine production) levels, we aimed to evaluate the profile of cytokines in the maternal circulation via a wide panel of inflammatory mediators, chemokines, and growth factors. Considering the range of time at which PPPE can be diagnosed, namely between 48 hours to 6 weeks postpartum, we first investigated if the mediators were correlated with the time of diagnosis. We found five, namely IL3, IL8, IL9, sCD40L, and MDC (presented in bold in Figure 9 ), that were correlated with the timing of diagnosis ( Supplementary Figure S3 ). Seeing as hypertension is a diagnostic criterion of PPPE, we wanted to differentiate mediators that were and were not correlated to blood pressure as these could be indicative of different pathological mechanisms. We identified 7 mediators in total (underlined in Figure 9 ), of which 5 that were negatively correlated (i.e., sCD40L, CXCL9, PDGFAA, PDGFAB/BB, and RANTES), and 2 that were positively correlated (i.e., IL12p40 and MDC; Supplementary Figure S4 ). Only sCD40L and MDC were both found to be correlated to days postpartum and blood pressure.

Aside from the pro-inflammatory mediators highlighted above, three additional mediators were increased (i.e., IL1α, IL7, IL17A) and three decreased (i.e., IL6, IL18, and IL27) all independently of the timing of diagnosis, suggesting a direct association with the pathology ( Figure 9A ). We also found decreased anti-inflammatory IL10 and increased IFNα2 ( Figure 9B ). The maternal circulation in PPPE also displayed alterations in chemokines whereby there was an increase in CXCL1, IP10, and MIP1β and decreased RANTES ( Figure 9C ). Vascular endothelial growth factor (VEGFA) was increased in PPPE, but GCSF levels decreased ( Figure 9D ). We lastly analyzed uric acid and HMGB1, both of which had been associated with PE, and found that uric acid was elevated; but HMGB1 was not ( Figures 9E, F ).

In addition to characterizing PPPE at the time of diagnosis, we investigated whether signs of prenatal initiation could be detected during pregnancy and used to identify women at risk for developing the pathology. We first investigated the placenta because its dysfunction is a cause of multiple pregnancy complications, and we have previously reported, in a smaller cohort, a placental increase in CD163+ cells. In our current cohort, we observed that the placentas of patients who went on to develop PPPE had skewed immune profiles with an elevated number of CD163+ cells ( Figure 10A ). We also investigated the presence of oxidative stress and autophagy markers, HO2 and LC3b respectively, and found an increase in HO2 ( Figure 10B ), but no changes in the levels of LC3b ( Suplementary Figure 6 ).

We also examined patients’ blood pressure collected in the 1^st trimester. It was observed that although women who went on to develop PPPE did not reach the levels to be classified as hypertensive (i.e. 140/90 mmHg), they presented with significantly higher mean arterial blood pressures (MAP) in their first trimester, with independent increases in both systolic and diastolic pressures, compared to the uncomplicated pregnancies who did not develop PPPE ( Figures 10C–E ).

The placenta is a valuable tool since samples collected before symptom onset are required for the prediction of diagnosis, and the placentas from both CTL and PPPE groups are gestational age-matched. Blood pressure measurements during the 1^st trimester are also valuable since they are non-invasive and consistently taken for all pregnant patients, providing insights early in gestation. Therefore, we investigated the potential use of placental CD163+ cells analysis and blood pressure in the 1^st trimester to identify women at risk of developing PPPE. To determine the predictive capacity of these biomarkers, we assessed them using Receiver Operating Characteristic (ROC) curves ( Figure 11 ). We began with MAP at the 1^st trimester and found an AUC of 0.67, suggesting a reasonable but limited discriminative power. We next explored the number of CD163+ cells as another potential biomarker. We found that CD163 alone yielded a substantially higher AUC of 0.86, indicating strong predictive accuracy. A logistic regression model combining both 1^st trimester MAP and placental CD163, showed an improved AUC of 0.9, demonstrating excellent ability to distinguish between the CTL and PPPE, which could be used for PPPE prediction. Thresholds for individual biomarkers indicate that 83.4mmHg for the 1st trimester MAP has the strongest combination of sensitivity and specificity, both at 67%. Likewise, the threshold for CD163+ cells is 50 per defined placental area (see Methods and Figure 10A ), which provides 84% sensitivity with 73% specificity in discriminating between patients that are CTL or going on to develop PPPE. We also investigated if MAP or numbers of CD163+ cells were driven by the differences in-between racial groups and found no differences amongst subgroups within the PPPE population (data not shown).