We obtained approval for this study from the Oxfordshire Research Ethics Committee C (Reference Number: 07/H0606/148). We defined normal pregnancy as a healthy singleton pregnancy conceived naturally without known fetal abnormality or abnormal maternal blood pressures. Placentas were collected from normal pregnant women who were in labor and undergoing elective Caesarean sections. Serum samples were obtained from gestationally matched PE and NP patients. We defined PE as hypertension (systolic blood pressure ≥140 mmHg/diastolic pressure ≥90 mmHg) and proteinuria (protein/creatinine ratio of 0.3 mg/mmol or more). In addition, for the serum samples, we selected EOPE patients with a soluble Flt: PLGF (sFLT: PLGF) ratio >110 and the NP samples with a ratio <38. The soluble Flt: PLGF (sFLT: PLGF) ratio helps characterize the level of placenta ischemia in Preeclampsia and is now routinely used in risk-stratifying Preeclampsia (17) We obtained informed consent from all participants in this study.

We enriched for m/lSTB-EVs as described in our previous study (18). Briefly, after examining the placenta, a satisfactory cotyledon of the placenta devoid of rupture, ischemia, clots, or calcification was identified. This cotyledon was perfused at a flow rate of 4–5 ml/min for three hours to obtain a maternal placenta perfusate using dual lobe perfusion. STB-EVs were enriched by differentially centrifuging the maternal placental perfusate; an initial 1,500 g spin at 4°C (Beckman Coulter Avanti J-20XP centrifuge using a Beckman Coulter JS-5.3 swing-out rotor) to remove red blood cells and cellular debris followed by enrichment for m/lSTB-EVs by centrifugation at 10,000 × g in a swing bucket centrifuge (Beckman L80 ultracentrifuge and Sorvall TST28.39 swing-out rotor) for 30 min at 4°C. Pellets were resuspended in filtered phosphate-buffered saline (fPBS, 0.22 µm Millipore Stericup Filtration Device (Merck, SCVPU11RE) and aliquots with a protein concentration around ≈ 5 µg/µl (measured using a Pierce BCA protein assay) were stored or set aside for downstream characterization. Downstream characterization experiments included nanoparticle tracking analysis, transmission electron microscopy, western blot, and flow cytometry. The STB-EV enrichment and categorization process has been deposited and assessed on EV Track [(http://www.EVTRACK.org), EV-TRACK ID: EV210382] with a score of 78% (the average score on EV track for 2021 is 52%). For the serum validation, a cohort of serum samples from pregnant women with early onset preeclampsia (EOPE, n = 8) and gestationally matched normal pregnancy women (NP) were obtained (n = 8). The serum samples were prepared using standard protocol. Briefly, blood samples were collected from the antecubital fossa and centrifuged at 2,000 g for 10 min. The supernatant was collected, and the serum samples were frozen at −80°C until analysis. All blood samples were processed within 1 h upon collection.

To prepare the placenta tissue for RNA extraction, placental tissue samples were promptly collected post-delivery and precisely sliced before being immersed in pre-cooled homogenization vessels containing 500 µl of RNAprotect Tissue Reagent. Ensuring RNA stability, these prepared samples were swiftly stored at −80°C. For RNA extraction, the RNAprotect-treated tissue was transferred to appropriately sized vessels, and 600 µl of Buffer RLT was added. Under ice-cooled conditions, the samples underwent thorough homogenization for 40 s with a tissue homogenizer, resulting in a consistent lysate. The lysate was centrifuged at full speed for 3 min, yielding an RNA-enriched supernatant. To prevent contamination, the supernatant was meticulously transferred to fresh RNase-free tubes. The RNA lysates were then securely stored at −80°C until further RNA extraction procedures were performed following the standard manufacturer protocol with the RNeasy Mini kit. Stringent measures, including the use of RNase-free equipment, were consistently applied throughout the process to mitigate the risk of RNA degradation.

We phenotyped the isolated m/lSTB-EV with flow cytometry (BD Biosciences, LSRII), transmission electron microscopy (TEM, for morphology), and western blot (for immunophenotyping). We used antibodies to placental alkaline phosphatase—PLAP- (to confirm syncytiotrophoblast origin), CD41 (to identify co-isolated platelet EVs, CD235a/b (to identify co-isolated red blood cell EVs), HLA class I and II (to identify co-isolated white blood cell EVs) for flow cytometric analysis as previously described in our previous paper (16). For immunophenotyping, we characterized the m/lSTB-EVs by probing for known EV markers CD63 [(200 mg/ml) at 1:1,000 dilution, Sc-59286, Santa Cruz Biotechnology], placental alkaline phosphatase [PLAP (1.667 mg/ml) at 1:1,000 dilution, in house antibody], and the known negative EV marker Cytochrome C [(200 mg/ml) at 1:1,000 dilution, Sc-13156, Santa Cruz Biotechnology] as recommended by the international society for extracellular vesicles (ISEV) (19). The immunoblot was subsequently incubated with either anti-mouse or anti-rabbit polyclonal goat immunoglobulins/HRP (at 1 2,000 dilution, Dako UK Ltd, Cambridgeshire UK).

Morphology was assessed by using TEM. Briefly, STB-EVs were diluted with filtered PBS to obtain a protein concentration between 0.1–0.3 µg/µl. 10 µl of the solution was added to freshly glowing discharged carbon formvar 300 mesh copper grids for two minutes, dried with filter paper, stained with 2% uranyl acetate for ten seconds, and air-dried. The grid was negatively stained to intensify the contrast between the pellets and the background. Imaging was done with an FEI Tecnai 12 TEM at 120 kV with a Gatan OneView CMOS camera.

Finally, we characterized the STB-EVs with nanoparticle tracking analysis ((NTA) NanoSight NS500 instrument equipped with a 405 nm laser (Malvern UK), sCMOS camera, and NTA software version 3.5, Build 0033 (Malvern UK)). We checked the instrument performance with silica 100 nm microspheres (Polysciences, Inc.). Each STB-EV sample was diluted in fPBS to obtain the concentration within 2 × 10² particles/ml to 1 × 10⁹ particles/ml. The samples were automatically injected into the sample chamber with a 1 ml syringe with the following script for EV measurements: PRIME, DELAY 5, CAPTURE 60, REPEAT 4. Images of the analyzed samples were captured on camera at 12 (Camera shutter speed; 15 ms and Camera gain; 350), and NTA post-acquisition settings were optimized and kept constant between samples. Each video recording was analyzed to infer STB-EV size and concentration profile.

For sequencing, RNA extraction and sample preparations were performed for both placenta lysates and medium/large STB-EVs (6 NP, 6 PE) with the RNeasy Mini kit and miRCURY™ RNA isolation kit for biofluids (Exiqon Services, Denmark) respectively based on the manufacturer's protocol. RNA purity was assessed with Nanodrop, and samples with a 260/280 ratio greater than 1.8 were sent to the Welcome Centre for Human Genetics (WCHG) for sequencing. RNA libraries were prepared using the standard procedure using the TruSeq small RNA library preparation kits (Illumina). Details can be found in the Supplementary Data.

We performed bioinformatics analysis for small RNA analysis on Oasis 2.0 (https://oasis.ims.bio/small_rna_classification.php) (20) using the default small RNA pipeline. More information about the default parameter setting for OASIS-db can be found in the documentation and the Supplementary Methods. Differential expression was performed with DESeq2 package (21) (v.1.32.0 in R v.4.0.5), and the Benjamin Hochberg method was used to correct for multiple testing and calculate the adjusted P value. Significantly differentially expressed microRNAs were defined by fold change (FC) ≥ ± 1 with an adjusted P value ≤ 0.05. The unprocessed fastQ files and the processed data have been archived in the National Center for Biotechnology Information (NCBI) database. These files are publicly accessible under the accession number GSE190972.

Complementary DNA (cDNA) from the extracted RNA was generated with the TaqMan^TM Advanced microRNA cDNA Synthesis Kit (A28007). Each reaction was performed in a G-storm^TM (G-storm, UK) thermocycler with settings provided by the manufacturer. We selected the following m/lSTB-EV microRNAs to validate based on fold change in PE compared to NP; hsa-miR-30b-5p, hsa-miR-221-3p, hsa-miR-222-3p, hsa-miR-324-5p, hsa-miR-194-5p, hsa-miR-519c-3p, hsa-miR-652-3p, hsa-miR-584-5p, hsa-miR-421, hsa-let-7e-5p, hsa-miR-3196, hsa-miR-4488, hsa-miR-4516, hsa-miR-9-5p, hsa-miR-210-3p, and hsa-miR-193b-5p (see Supplementary Material). We opted for hsa-miR-9-5p for serum validation due to its described role in angiogenesis, a process widely reported dysregulated in Preeclampsia. All microRNAs tested were detected before the 35th cycle. The following settings were used for the QuantStudio 3 and QuantStudio Design and Analysis desktop software (Thermofisher) qPCR: hold 50°C for 2 min, hold at 95°C for 20 s, followed by forty 95°C for 1 s and 60°C for 20 s.

For placenta and m/lSTB-EVs, data was normalized to hsa-miR-30d-5p. We chose hsa-miR-30d-5p from the sequencing data based on an absent differential expression (FC approximately <0.001) between PE and NP and a low standard error (LFSE) and verified by qPCR for expression stability. For serum samples, data was normalized to spiked in Cel_miR-39. Analysis was performed with the 2^−ΔΔCt method (22) and statistical testing on the Δ Ct values using a one-tailed Student t-test (significance = P < 0.05). Data was visualized using GraphPad Prism software (version 9) and expressed as fold change, and the standard error of the mean was denoted as error bars. All gene expression assays, their corresponding assay IDs, and all other details used in this study are listed in Supplementary Methods.

Frozen serum samples from gestationally age-matched women with EOPE (n = 8) and those with normal pregnancies (n = 8) were thawed at 37°C in a water bath. Subsequently, 800 µl of serum was loaded onto ExoRneasy Midi kits (Qiagen) columns to extract high-quality RNA. We then added 1 μl aliquot of a 5 nanomolar working solution of cel-miR-39 (Integrated DNA Technologies, 100 nmoles RNA Oligo) spike-in to each serum sample. Complementary DNA (cDNA) synthesis was performed using the TaqMan^TM Advanced microRNA cDNA synthesis kit (A28007). All reactions were meticulously executed on ice, adhering to the manufacturer's protocol. Post reverse transcription (RT), 10 µl of the reverse-transcribed samples underwent pre-amplification. The pre-amplified samples were then diluted at a 1:5 ratio with 0.1× Tris-EDTA buffer for subsequent qPCR reactions, as earlier described.

Given our specialized interest in microRNAs and their regulatory role in transcription, a comprehensive functional enrichment analysis was conducted on microRNAs extracted from both placenta and m/lSTB-EVs. Utilizing DIANA-miRPath v3.0 (http://www.microrna.gr/miRPathv3), the microRNAs were stratified into upregulated and downregulated categories for analysis. We performed distinct functional enrichment analyses, focusing on Gene Ontology categories encompassing biological processes, molecular functions, cellular components, and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. This analysis was conducted separately for the top 100 upregulated and downregulated microRNAs identified in the placenta and m/lSTB-EVs. Restricting the examination to up to 100 microRNAs (microRNAs) with a fold change (LFC ≥ 0.75) and an adjusted p-value (<0.05), the imposed limit was imposed by the default settings of the tool and the computational demands associated with processing extensive lists of microRNAs. The prediction of mRNA targets for the loaded microRNAs was accomplished using the microT-CDS algorithm with a microT-threshold of 0.8. To refine the results further, the outcomes of this analysis were fused at the level of categories or pathways. Additionally, removing, summarizing, and visualizing redundant Gene Ontology (GO) terms were undertaken using REVIGO (http://revigo.irb.hr/) to address the computational intensity and streamline the data.

We subsequently focused on predicting and visualizing mRNA targets modulated by the differentially expressed microRNAs detected in our study to elucidate the regulatory dynamics associated with Preeclampsia. A vital part of this study was identifying some microRNAs responsible for the differentially expressed mRNAs previously reported in our transcriptomics paper (16). The raw fastQ files and the processed file of the transcriptomics data can be accessed on NCBI with the following ID: GSE190971. This analysis used the normalized expression matrices for microRNAs and mRNAs with miRTarVis + (http://sehilyi.com/mirtarvisplus/)-an online interactive visual analytics tool. MicroRNA expression matrices were filtered based on a fold change (FC) threshold of 1.5 and an adjusted p-value criterion of 0.05. Concurrently, mRNA expression data underwent a similar filtration process, with a fold change (FC) threshold of 1.5 and a stringent adjusted p-value of 0.005. The filtered datasets were harnessed to predicticroRNA target interactions with the TargetScan prediction algorithm (23). After making the prediction, we used Pearson correlation, mutual information, and maximal information-based nonparametric exploration (MINE) to find microRNA-mRNA pairs that were strongly correlated. We selected the top 300 interactions for the placenta and the top 1,000 interactions for m/lSTB-EVs. A correlation coefficient cutoff of 0.5 and a neighbor cutoff score of 3 were used for correlation and mutual information analysis. We only considered microRNA-mRNA interactions identified by all three algorithms for enhanced reliability. The results of the microRNA-mRNA interactions were visualized through a treemap, providing a comprehensive overview of the intricate regulatory network governing the interplay between microRNAs and mRNAs in Preeclampsia.

PE women were (as expected) significantly more likely to deliver prematurely (PE = 31.88 weeks' gestation NP = 39.33 weeks' gestation) and have proteinuria (PE = 2.19 pluses on urine dipstick, NP = 0 pluses on urine dipstick) compared to normal pregnancy. PE mothers had a significantly higher average systolic (163.12 mmHg) and diastolic (104.62 mmHg) blood pressure compared to normal pregnant mothers average systolic (121.47 mmHg) and diastolic (71.2 mmHg). PE babies weighed less than normal babies (PE = 1,515.31 g, NP = 3,671.33 g). However, we found no significant difference in body mass index, the gender of the child, and maternal age (Table 1).

The 10 K STB-EVs isolated from placenta perfusion showed the characteristic cup shape under the negative staining on transmission electron microscopy (TEM). Western blot confirmed that 10 K STB-EVs expressed the classic syncytiotrophoblast membrane marker (CD63), placenta alkaline phosphatase (PLAP), confirming placental origin (Figure 1C). In addition, 10 K STB-EVs lacked the negative EV marker cytochrome C, confirming that these vesicles were not derived from apoptotic bodies (Figure 1C). Nanoparticle tracking analysis (NTA) confirmed the heterogeneity of the 10 K STB-EVs (Figure 1D) with a size range of (479.4 ± 145.6) nm, with the majority over 200 nm. According to ISEV, 10 K STB-EVs were defined as medium/large STB-EVs (m/lSTB-EVs). Finally, flow cytometric analysis (Figures 1E–I) of m/l STB-EVs showed PLAP positive 89 ± 1.3% predominantly. m/l STB-EVs were also shown to be vesicular due to minimal signal after adding detergent to break the lipid membrane (Figure 1I). These data suggested that we had successfully enriched m/lSTB-EVs from placental perfusions. Full details of m/lSTB-EV characterization can be seen in the Supplementary Methods.

In the placenta, a comparative analysis between preeclampsia (PE) and normal pregnancy revealed significant dysregulation of microRNAs. Specifically, 14 microRNAs were upregulated, while 120 were downregulated (p-value < 0.05, as shown in the Supplementary Figure S1). The full list of the placenta data can be found in the Supplementary Material. In comparison between PE and normal pregnancy m/lSTB-EVs, 47 microRNAs were upregulated, and 53 were downregulated (p-value of <0.05) (see Supplementary Files for full list). Table 2 shows the top forty DEMs in m/lSTB-EVs in PE. As discussed earlier in the methodology section, due to our interest in m/lSTB-EVs as potential biomarkers, we selected the most differentially expressed microRNAs from the RNA sequencing data. For m/lSTB-EVs (Figure 2A), of the microRNAs tested, hsa-miR-193b-5p (RNA Seq FC = 1.46 qPCR FC = 2.02), hsa-miR-324-5p (RNA Seq FC = −1.65 qPCR FC = −1.66), hsa-miR-652-3p (RNA Seq FC = −1.24 qPCR FC = −1.53), hsa-miR-3196 (RNA Seq FC = 1.59 qPCR FC = 0.89), hsa-miR-9-5p (RNA Seq FC = 0.84 qPCR FC = 4.73), hsa-miR-421 (RNA Seq FC = −0.84 qPCR FC = −1.03), and hsa-miR-210-3p (RNA Seq FC = 1.42 qPCR FC = 4.58) were all significantly differentially expressed. We meticulously scrutinized our validated differentially expressed microRNA repertoire to uncover novel insights. We sought a microRNA with a positive fold change while concurrently exhibiting involvement in angiogenesis—a process intricately linked to the pathology of Preeclampsia. Our investigation led us to hsa-miR-9-5p, a microRNA that remains relatively underexplored within the field despite its pivotal role in angiogenesis. Figure 2B shows that in the placenta (FC = 31.66 SE = 4.98, P < 0.001) and maternal serum EVs (FC = 2.04 SE = 0.34, P < 0.001), hsa-miR-9-5p is significantly more abundant in PE compared to NP.

Table 3 presents the functional enrichment analysis results on differentially expressed microRNAs found in both placenta and m/lSTB-EVs. The analysis revealed ten biological processes influenced by upregulated microRNAs, including the neurotrophin TRK receptor signaling pathway, cellular protein modification process, and epidermal growth factor receptor signaling pathway (as detailed in Table 3). Moreover, the KEGG pathway analysis highlighted the significance of ECM-receptor interaction among the upregulated microRNAs (Table 3). In contrast, the downregulated microRNAs exhibited a broader range of shared KEGG pathways, encompassing ECM-receptor interaction, glycosphingolipid biosynthesis (lacto and neolacto series), morphine addiction, TGF-beta signaling pathway, and ErbB signaling pathway. Refer to the Supplementary Material for the complete list of functional enrichment analysis results.

To comprehensively assess the potential influence of microRNA modifications on mRNA expression, our investigation systematically cross-referenced the results of microRNA analysis with messenger RNA sequencing data, as outlined in the methods section. Within extracellular vesicles derived from placental and medium/large syncytiotrophoblasts, our analysis revealed 155 and 57 distinct microRNA-mRNA interactions, respectively. Notable instances include the interaction between downregulated hsa-miR-222-3p and upregulated TPBG and PLEKHA2. The visual representation of these interactions is depicted in Figures 3A,B for the placenta and m/lSTB-EVs, respectively. It is essential to highlight that among the microRNAs validated by quantitative polymerase chain reaction (qPCR) in m/lSTB-EVs, only two, specifically hsa-miR-324-5p and hsa-miR-421, exhibited previously confirmed mRNA targets based on other experimental studies, as elaborated in Table 4.