Rapid and high-sensitivity screening of pregnancy complications by profiling circulating placental extracellular vesicles

Abstract

Herein, we developed a specific, rapid sensor to quantify placental extracellular vesicle (EV) protein biomarkers of early pregnancy complications. A distinct tetraspanin CD9 and placental alkaline phosphatase (PLAP) expression pattern was observed via targeted multiple reaction monitoring of EVs from maternal plasma collected before 18 weeks of gestation. A classification model was developed using training and validation patient sets, distinguishing between individuals at high risk of developing complications from those with normal pregnancies, achieving 80% sensitivity, 90% specificity, 89% positive predictive value (PPV), and 82% negative predictive value (NPV). Superparamagnetic nanoflowers that captured target EVs (CD9⁺/PLAP⁺) were used to construct a 4-flex glass strip nanozymatic readout system. The sensor analyzes plasma for EVs, identifying gestational diabetes mellitus risk with a 95% combined sensitivity, 100% specificity, 100% PPV, and 96% NPV. This nanoplatform identifies individuals at risk of developing pregnancy complications with a >90% classification accuracy, exhibiting potential for clinical applications.

Fig. 1.. Study population characteristics.

A total of 201 pregnant women at early gestation were enrolled. For biomarker discovery and training, 141 women were included, with 71 controls (normal pregnancies) and 70 cases with pregnancy complications, specifically gestational diabetes mellitus (GDM; n = 20), preeclampsia (PE; n = 16), preterm birth (PTB; n = 7), and others (n = 27). The test set included 60 women, with 30 controls and 30 cases comprising GDM (n = 8), PE (n = 7), PTB (n = 3), and others (n = 12).

Fig. 2.. Individual performance of proteins associated with EV and placenta in predicting the risk of pregnancy complications.

(A to E) Peptides are colored for each protein structure based on the Protein Data Bank (PDB) with the best resolution and coverture available. The structures were prepared with PyMOL in surface mode, and biological assembly was available from PDB (monomer, dimer, and tetramer). (A) CD9, (B) PLAP, (C) CD63, (D) CD81, and (E) TSG101. a, area under the receiver operating characteristic curve (AUROC); b, area under the precision-recall curve (AUPRC); c, calibration curve. Data are presented as mean values generated via bootstrapping (discovery/training set, n = 140; and test set, n = 70) with a 95% confidence interval. Blue line, training set; red line, test set.

Fig. 3.. Predictive model performance under combined CD9, PLAP, CD63, CD81, and TSG101EV.

(A) Binominal model for cases of all pregnancy complications and controls. (B) Binominal model for cases of GDM and controls. (C) Binominal model for cases of PE and controls. (D) Binominal model for cases of PTB and controls. (E) Binominal model for cases of other complications (excluding GDM, PE, and PTB) and controls. a, AUROC; b, AUPRC; c, calibration curve. Data are presented as mean values generated via bootstrapping (discovery/training, n = 140; and test, n = 70) with a 95% confidence interval. Blue line, training set; red line, test set.

Fig. 4.. SMNF-based 4-flex glass nanosensor for PLAP⁺ EVs isolation and detection.

(A) Direct isolation of PLAP⁺ EVs using anti-PLAP modified SMNFs; (B) schematic representation of the assay on the 4-flex glass sensor; (C) relative activity of the assay with different types of antibodies obtained from BeWo and HTR-8 cell lines; (D) relative activity of assay with no antibody or with a PLAP antibody; (E) NTA responses after isolating EVs using unmodified SMNFs and anti-PLAP modified SMNFs; (F) responses of the assay for target samples against three controls [no-target (NoT): EV sample is replaced by PBS; No SMNFs: EVs are not extracted by PLAP modified SMNFs; No Abs: SMNFs are not modified with PLAP antibody (Ab)]; (G) relative activity (absorbance at 652 nm) for the designated concentration of exosomes extracted from a GDM (BeWo) cell line. The concentration varies between 10³ and 10⁹ particles/ml, equivalent to 8 × 10⁶ versus 8 × 10¹¹ vesicles per million cells/24 hours.

Fig. 5.. Patient sample analysis.

Relative activity of plasma and extract EV samples from (A) patients with GDM and (B) healthy patients; side-by-side comparison of responses obtained from (C) extracted EVs and (D) plasma samples.

Fig. 6.. Predictive performance of the sensor to identify individuals at risk of developing GDM.

(A) Plasma samples; (B) extracted EVs; a, AUROC; b, AUPRC; c, calibration curve. Data are presented as mean values generated via bootstrapping with a 95% confidence interval. Blue line, training set; red line, test set.