Preeclampsia (PE) poses a critical global health challenge, contributing substantially to maternal, fetal, and neonatal morbidity and mortality (Magee et al., 2022; Abalos et al., 2014). Affecting approximately 2%–8% of pregnancies worldwide, this complex multisystem disorder manifests through a diverse spectrum of clinical symptoms, ranging from mild hypertension to severe complications including eclampsia and HELLP syndrome (Hemolysis, Elevated Liver enzymes, Low Platelet count) (Gestational Hypertension and Preeclampsia: ACOG Practice Bulletin, 2020). PE can develop at various time points in pregnancy after 20 weeks of gestation and vary in severity. The condition is often classified into subtypes based on onset timing: early-onset (<34 gestational weeks) versus late-onset (≥34 weeks), or preterm (delivery <37 weeks) versus term (delivery ≥37 weeks). Additionally, PE can be further categorized as mild and severe depending on maternal symptom severity, or whether complicated with fetal growth restriction (FGR) (Dimitriadis et al., 2023).

Although PE subtypes present similar clinical symptoms, a common pathophysiological mechanism currently fails to explain the aetiology of all PE cases. Substantial evidence from clinical, epidemiologic, histologic and biological studies supports placental dysfunction as a central factor in PE pathophysiology (Dimitriadis et al., 2023). It has been proposed that dysfunctional placenta releases pathogenic factors into maternal circulation, triggering endothelial dysfunction and systemic inflammation responses, leading to clinical manifestation of PE (Pankiewicz et al., 2021; Michalczyk et al., 2020; Ngene and Moodley, 2018).

The cause and degree of placental dysfunction varies among preeclampsia subtypes, likely reflecting various pathophysiological processes. Early-onset PE is often associated with inadequate trophoblast invasion and poor remodelling of the uterine spiral arteries, leading to placental hypoxia and oxidative stress. This defective placentation is believed to be influenced by aberrant maternal immune responses to the feto-placental unit (Burton et al., 2019). Additionally, early-onset PE is characterized by more pronounced systemic inflammation and disruption of the angiogenic balance (Chuah et al., 2018; Pinheiro et al., 2014). Another potential aetiology is suboptimal maternal cardiovascular function secondary to uteroplacental malperfusion, which may contribute to placental dysfunction in certain PE cases (Melchiorre et al., 2022). Epidemiological evidence has revealed shared risk factors between PE and cardiovascular disease (Wu et al., 2017; Leon et al., 2019), and echocardiographic studies have found cardiac parameter abnormalities in women several weeks prior to the manifestation of clinical signs of both preterm and term PE (Thilaga and nathan, 2020; Castleman et al., 2016; Garcia-Gonzalez et al., 2020; Melchiorre et al., 2013).

Previous transcriptomic studies have provided valuable insights into the molecular differences between early-onset and late-onset PE, supporting the hypothesis that these subtypes may be driven by different pathogenic mechanisms. As early as 2007, Nishizawa et al. conducted a microarray analysis of placental samples from severe PE cases and identified 11 differentially expressed genes between early-onset and late-onset subtypes (Nishizawa et al., 2007). Later, Sitras et al. reported 168 differentially expressed gene between these two PE subtypes, with pathways related to oxidative stress, inflammation, and endothelin signalling involved in early-onset PE (Sitras et al., 2009). Similarly, Junus et al. found significant downregulation of angiogenesis-related genes specifically in early-onset type, suggesting its association with more severe placental vascular dysfunction (Junus et al., 2012). Subsequent transcriptomic investigations have consistently shown that late-onset PE exhibits fewer placental gene alterations compared to early type (Ren et al., 2021; Liang et al., 2016). Furthermore, dysregulation of the placental innate immune system has been identified as a feature specific to early-onset PE but not observed in the late-onset subtype (Broekhuizen et al., 2021). Most recently, a single-cell transcriptomics study of PE placentae reinforced this evidence, showing widespread cell-type–specific gene dysregulation in early-onset PE but fewer changes in late-onset (Admati et al., 2023).

The classical analytic method for those transcriptomic studies focuses on differential gene expression, examining individual genes based on fold changes and statistical significance. However, this approach cannot fully capture the complex gene-gene interactions and regulatory networks underlying multifactorial diseases like PE. Advanced bioinformatics methods like weighted gene co-expression network analysis (WGCNA) can identify co-expression modules of functionally related genes that can be correlated with clinical phenotypes and disease pathophysiology (Langfelder and Horvath, 2008). Our study employs an integrated approach, combining WGCNA with differential expression analysis to systematically characterise molecular signatures across three PE subtypes. This investigation elucidates distinct molecular mechanisms underlying subtype-specific placental pathologies in PE.

The exact aetiology of PE remains elusive, and its clinical management continues to be challenging due to its multifactorial and heterogenous nature. Through integrated analysis of placental transcriptomics, we have identified both subtype-specific molecular signatures and overlapped biological processes with a core placental dysregulation signature underlying the three PE subtypes (severe, early-onset, and late-onset). Co-expression gene modules showed stronger association with severe and early-onset PE and these subtypes also have a greater number of differentially expressed genes. In contrast, late-onset PE presents modest correlation with WGCNA gene modules and fewer dysregulated genes. This may indicate that placental dysfunction is closely related to disease severity and early manifestation. We also identified 20 commonly dysregulated placental genes across PE-related modules in all PE subtypes, with four upregulated genes (LEP, FSTL3, HTRA4, and HK2) validated in the external dataset, suggesting a potential shared pathogenic feature despite the clinical and molecular heterogeneity among subtypes.

In this study, we found a robust association between severe PE and the WGCNA green module. GO functional annotation of dysregulated genes in this module revealed enrichment of several biological processes, including lipid storage, epidermis development, and response to decreased oxygen levels. These dysregulated pathways, particularly abnormal lipid metabolism and hypoxia response, appear to be key features of severe PE placental pathology. Previous research found increased levels of phospholipids, total cholesterol and lipid peroxides in preeclamptic decidua basalis tissue (Staff et al., 1999). Subsequent lipidomic studies also confirmed significantly higher lipid content in preeclamptic placental tissue (Zhang et al., 2022; Brown et al., 2016). In addition, maternal blood lipidomic profiling study has identified a significant correlation between oxidized phospholipids (OxPLs) and PE. They also found specific lipid species are uniquely associated with severe PE (He et al., 2021). Additionally, a study found that hypoxia promotes accumulation of lipid droplets in primary human trophoblast, and that perilipin (PLIN) proteins play key roles in the process (Bildirici et al., 2018). This evidence suggests there may be a potential link between placental hypoxia and altered lipid metabolism. Despite established research for both placental hypoxia and dysregulated lipid metabolism in PE, the relationship between these processes and how hypoxia-induced alterations in placental lipid metabolism may drive PE development and progression remains unclear.

The enrichment of lipid storage pathways was also observed in early-onset PE within the yellow module. Four genes involved in lipid metabolism (SCARB1, LEP, PLIN2, LPL) are upregulated in both severe and early-onset subtypes: SCARB1 mediating cholesterol uptake (West et al., 2009), LEP encoding leptin, a hormone regulating energy consumption and adiposity (LeDuc et al., 2021), PLIN2 facilitating lipid storage droplets formation (Itabe et al., 2017), and LPL hydrolysing triglycerides (Mead et al., 2002). These molecular alterations in placental lipid processing may contribute to both PE severity and early-onset manifestation. Moreover, DEGs genes (LEP, INHBA, INHA and CRH) in the yellow module are enriched endocrine hormone secretion pathways. This molecular signature suggests that disruption of endocrine and gonadotropin secretion processes may be a pathogenic mechanism in early-onset PE. INHA and INHBA encode inhibin A and activin A, modulating placental hormone synthesis. Elevated levels of activin A and inhibin A have been previously reported in placenta and maternal circulation as potential endocrine markers for PE (Florio et al., 2002; Muttukrishna et al., 2000; Spencer et al., 2008). Furthermore, dysregulated genes in the black module are mostly downregulated. We found IDO1, VNN1, S100A8 are enriched in chronic inflammatory and immune response processes, indicating possible altered inflammatory or immune regulation in early-onset PE. IDO1 is an interesting gene encoding indoleamine 2,3-dioxygenase (IDO), with functions involved in chronic inflammatory response, T cell tolerance induction, and L-tryptophan catabolism (Seo and Kwon, 2023). Reduced expression and activity of IDO1 have been reported in preeclamptic placentae (Kudo et al., 2003; Iwahashi et al., 2017), with one study suggesting this downregulation only occurs in early-onset PE but not in late-onset PE (Broekhuizen et al., 2021). Overall, these findings provide molecular evidence of complex interactions among metabolic, endocrine, and immune-inflammatory pathways in the pathogenesis of early-onset PE.

Previous research suggests that late-onset PE is less associated with placental dysfunction than severe and/or early-onset forms (Ren et al., 2021). These differences likely reflect distinct underlying pathophysiological mechanism. Early-onset PE is primarily characterised by defective placentation in early gestation, resulting in widespread transcriptomic and histopathological disruption. In contrast, late-onset PE is believed to being predominantly driven by maternal factors, such as preexisting cardiovascular and metabolic conditions, with placental stress and aging emerging as secondary contributors in later gestation (Melchiorre et al., 2022; Redman et al., 2022; Robillard et al., 2022; Staff, 2019; Khodzhaeva et al., 2016). This is further supported by clinical evidence demonstrating higher frequencies of fetal growth restriction in early-onset PE compared to late-onset PE, as well as placental pathology analyses reporting a higher rate of maternal vascular malperfusion lesions in early-onset cases (Freedman et al., 2023; Ogge et al., 2011; Gilgannon et al., 2023; Hung et al., 2018). Consistent with these established findings, our analysis found that late-onset PE exhibited fewer differentially expressed genes and only modest correlations with WGCNA gene modules, which may indicate more subtle placental transcriptomic alterations in the late subtype. The ivory module has a moderate correlation with disease status, with DEGs primarily enriched in the p38MAPK signalling pathway (SASH1/FLT1/NPNT/LEP/OPRK1). This pathway plays a critical role in stress response and inflammatory signalling (Cuenda and Rousseau, 2007). The enrichment of p38MAPK signalling in late-onset PE placenta may reflect activation of stress-response mechanisms proximal to term.

We externally validated twenty placental genes that are consistently dysregulated across PE subtypes and confirmed that four genes (LEP, FSTL3, HTRA4, HK2) were significantly upregulated. However, clinical validation by RT-PCR only presented moderate upregulation, which may be attributed to the predominance of term PE cases (19/22) in our validation cohort, all of which developed and delivered at or beyond 37 weeks of gestation. Previous studies support the clinical utility of three of these candidates as maternal biomarkers. LEP plays a multifunctional role in the placenta such as regulating endocrine processes, angiogenesis, and inflammatory responses (Zeng et al., 2023). Maternal serum and plasma leptin levels have been found to differ between preeclamptic women and normotensive pregnant women, with higher concentrations in severe and early-onset cases (Taylor et al., 2015; Hao et al., 2020; El et al., 2013; Salimi et al., 2014). Similarly, increased follistatin-like 3 (FSTL-3) levels is reported with increased likelihood of developing PE (Found et al., 2015; Han et al., 2014), although another study found that FSTL-3 did not alter in early-onset PE (Nevalainen et al., 2017). Elevated serum HtrA4 levels were also higher in the PE group compared to the control group, and this biomarker showed predictive value when combined with first-trimester uterine artery Doppler measurements (Siricharoenthai and Phupong, 2023). HK2 encodes hexokinase 2, a key glycolytic enzyme that is upregulated in preeclamptic and FGR placentas (Wong et al., 2024). Currently, no studies have investigated whether hexokinase 2 levels are elevated in the maternal circulation in PE cases.

Our study identified subtype-specific mechanisms and key dysregulated genes associated with PE. Future research should validate key dysregulated placental genes through functional experiments such as placenta organoid models to define their roles in placental dysfunction. Moreover, determining whether candidate genes such as LEP, FSTL3, HTRA4, and HK2, or their protein products, can be reliably detected and quantified in maternal circulation is essential for translating these findings into clinical applications as potential biomarkers. Several limitations should be considered when interpreting these results. First, heterogeneity in sample sources and transcriptomic platforms may impact reproducibility. Datasets GSE75010 and GSE25906 were generated using microarray technology, whereas dataset GSE234729 utilized RNA-sequencing. Such technique and platform differences introduce technical variations that may affect gene expression comparison across datasets. For the current analysis, we also selected only studies with greater than 60 samples; this was done to provide a good level of statistical power, but may have introduced selection bias by excluding smaller studied. Additionally, potential confounding factors like maternal clinical characteristics may also influence placental gene expression patterns. Second, although PE cases and controls were matched for key maternal variables in RT-PCR validation, several factors are likely to have limited our capacity to detect gene expression with significant differences, including the modest sample size, the predominance of term PE cases (19/22 delivering ≥37 weeks gestation), and potential RNA degradation during sample processing. Third, the computational methodologies employed generate preliminary findings that require experimental validation. While WGCNA is a powerful tool for identifying gene co-expression modules, this approach is susceptible to various sources of bias, including technical artifacts, suboptimal experimental design, and analytical decisions (e.g., sample clustering, module selection). Similarly, predicted PPI networks need experimental confirmation at the protein level to establish biological relevance and functional significance. These methodological limitations collectively affect the reproducibility and clinical interpretation of our results, indicating that further experimental validation is required.

In conclusion, this study presents a detailed analysis of placental transcriptomic data across different PE subtypes, revealing both distinct molecular signatures and shared potential pathogenic mechanisms. Severe and early-onset PE are characterized by significant molecular dysregulation in placenta, while late-onset PE shows more modest alterations. There is evidence that disrupted lipid storage pathways are a common molecular feature in both early-onset and severe PE, suggesting that altered placental lipid homeostasis may be a critical determinant of disease severity and early manifestation. Whilst these findings provide evidence of placental transcriptomic changes associated with PE, they are preliminary and require further experimental confirmation in additional cohorts to determine the potential translation of evidence into clinical care.