Preeclampsia (PE) is a common obstetric multisystem disorder that occurs in 2–5% of pregnancies worldwide [1]. Numerous efforts have been made to predict PE development in certain targeted pregnant populations of different gestational ages. However, due to the relatively low prevalence, screening the general population in early pregnancy did not offer an effective method for the prediction of PE [1, 2]. Recently, the serum levels of soluble fms-like tyrosine kinase-1 (sFlt-1), placental growth factor (PlGF), and especially the ratio of sFlt-1 to PlGF were suggested to exclude PE in suspected women with elevated blood pressure or proteinuria. More specifically, even though the negative predictive value of the sFlt-1/PlGF ratio was adequate and as high as 99.3%, its positive predictive value was only 36.7%, suggesting its limited value in the prediction of the presence of PE in high-risk populations [3].

As part of the continual efforts in identifying the novel serum predictors, we employed the isobaric tags for relative and absolute quantitation (iTRAQ) technique to reveal the proteomic changes in the sera from a prospective cohort suspected to develop PE, followed by validation experiments by enzyme-linked immunosorbent assays (ELISAs) and western blotting. For the prospective cohort, the pregnant women with PE-related clinical and/or laboratory presentations had their blood drawn at their first visits and were followed up for PE development until delivery (see details in the Supplementary Methods).

As summarized in Table 1, the patients of the iTRAQ PE-positive group (n = 33) and the iTRAQ PE-negative group (n = 33) whose sera were used in the iTRAQ and ELISA experiments had no difference in their ages or sampling gestational weeks. Similar findings were seen between the subjects of the healthy group (n = 11) (“ELISA PE negative” in Tables 1 and 2) and those of the PE-diagnosed group (n = 11) (“ELISA PE positive” in Tables 1 and 2), in which both age and sampling GWs were indistinguishable (see the definitions of the above groups in the Supplementary Methods).

Table 1 Characteristics of the study cohort in the iTRAQ proteomics and the ELISA experiments
Full size table
Table 2 The significantly different serum biomarkers by ELISA validation
Full size table

In the iTRAQ proteomics experiments, a total of 262,852 mass spectra were generated; 4685 peptides and 760 proteins were identified with a 1% false discovery rate. There were 37 upregulated and 90 downregulated proteins in the iTRAQ PE-positive versus PE-negative (PE-VS-CTR) comparison analysis. The ten selected protein candidates were downregulated proteins with fold changes ranging from 0.31 to 0.82 (Supplementary Table 1), of which the serum levels of fibrinogen beta chain, nidogen-2, and reelin were significantly lower in the PE-positive group than in the PE-negative group according to the ELISA validation results (Table 2). In the receiver operator characteristic (ROC) analysis for the evaluation of the prospective differentiation power of the candidate markers, the AUCs were 0.74 for fibrinogen beta chain, 0.70 for nidogen-2, and 0.72 for reelin (Supplementary Fig. 1A). When combined, the diagnostic power of fibrinogen beta chain, nidogen-2, and reelin reached an AUC of 0.79 (Supplementary Fig. 1B).

In the western blotting analysis with the protein extracts of the placental tissues, reelin expression was found to be significantly decreased in the PE patients compared with the healthy controls. However, there was no significant difference in the expression of fibrinogen beta chain or nidogen-2 between the PE group and the healthy group (Supplementary Fig. 2).

There were three protein marker candidates that showed significant differences in concentrations between the PE-positive group and the PE-negative group in our ELISA validation experiment, including fibrinogen beta chain, nidogen-2, and reelin. A few articles have shown that decreased fibrinogen beta chain levels are associated with PE. For example, in their study, Anand et al. observed a decrease in fibrinogen beta chain in the serum of pregnant women who later developed PE compared with those with uncomplicated pregnancies [4]. In another study of pregnant women with autonomic dysfunction, it was reported that hypertensive pregnant women presented lower fibrinogen levels in serum than normotensive pregnant women [5]. Although the underlying mechanism was not fully understood, it was suggested that the decreased levels of fibrinogen in PE pregnancies might result from extensive platelet activation and consumption compared with normal pregnancies [6].

Nidogen-2 is an extracellular matrix protein that is ubiquitous in renal basement membranes, linking the collagen IV, and laminin networks. An article has shown that blood pressure and albuminuria were significantly higher in nidogen-2-deficient mice than in wild-type littermates, suggesting that nidogen-2 may be important in the regulation of systemic blood pressure and essential for the formation and maintenance of the glomerular basement membrane. In addition, the lack of nidogen-2 may directly promote the disruption of tubular basement membranes, causing subsequent increased fibrosis and inflammation [7].

Reelin is reported as a secreted glycoprotein expressed in granule cells and Cajal–Retzius cells in the cerebellum and olfactory bulb [8]. In addition, reelin plays a vital role in the development and maturation of the central nervous system, such as neuronal excitability, neuronal migration, and dendritic morphology [9]. However, there has been no direct study published to elucidate the connection between reelin and PE in pregnancy. Interestingly, two known receptors of reelin are the very low-density lipoprotein receptor and the apolipoprotein E receptor type 2 [10], through which reelin may be speculated to contribute to PE by altered lipid metabolism.

In the ROC analysis (Supplementary Fig. 1), the AUCs of the individual markers or the combination of fibrinogen beta chain, nidogen-2, and reelin reached 0.7–0.8, suggesting that the three markers have potentially predictive value in the cohort suspected to develop PE. As mentioned above, for a pilot discovery study, it was unlikely to evaluate those markers in a fixed time window, as in Zeisler’s paper [3]. Nevertheless, with more laboratory evidence accumulated in the future, the clinical utility of the three markers reported in the present study could be further refined for better prediction performance.