Early characterisation and prediction of liver diseases in pregnancy by plasma cell-free RNAs.
Jinghua Sun,Wen-Jing Wang,Xuanyou Zhou,Tingyu Yang,Mujin Ye,Weihui Shi,Zhongzhen Liu,Lanlan Zhang,Qing Zhou,Yulong Qin,Yiyao Chen,Fang Chen,Hefeng Huang,Songchang Chen,Chenming Xu
DOI: https://doi.org/10.1002/ctm2.1439
IF: 8.554
2023-01-01
Clinical and Translational Medicine
Abstract:Liver dysfunction during pregnancy can result in adverse pregnancy outcomes. In particular, intrahepatic cholestasis of pregnancy (ICP) increases the risk of stillbirth, while hepatitis B virus (HBV) infection raises the likelihood of developing ICP.1 Despite extensive research, the molecular pathways underlying ICP are not fully understood, and early predictive biomarkers for ICP are still unavailable. Here, we conducted a comparative analysis of plasma cell-free RNA (cfRNA) from ICP (N = 74) or HBV (N = 40) patients and healthy pregnant women (N = 171) to explore the biological processes involved in ICP and HBV infection (Supporting Information Figure S1). We identified cfRNA biomarkers to predict ICP before the onset of symptoms, demonstrating the potential clinical applications of cfRNA in liver diseases. The detailed methods are described in Supporting Information. Nineteen ICP patients with a total bile acids (TBA) concentration of more than 10 μmol/L (ICP_TBA > 10 group) and 55 patients with normal TBA levels (TBA < 10 μmol/L, pre-ICP group) at sampling time were included in the ICP group (Figure 1A). The average diagnosis time of patients in the ICP_TBA > 10 group was 10.89 gestational weeks, which was significantly earlier than that of the pre-ICP group (33.20 gestational weeks) (p = 1.05E-10). The detailed clinical information of ICP patients is summarized in Supporting Information Table S1. Differentially abundant genes (DAGs) analysis identified 34 genes with altered abundance in the plasma of ICP patients (Figure 1B, Supporting Information Figure S2A and Table S2). Some DAGs are involved in transmembrane transport, biosynthesis and metabolism (Figure 1C). Gene set enrichment analysis (GSEA) of mRNA and enrichment analysis of differentially abundant miRNAs (DA-miRNAs) showed that pathways related to immune response, cell junction, extracellular matrix organization, biosynthesis and metabolism, transmembrane transport and coagulation were enriched among genes found to be more abundant in patients with ICP, which reflects the known pathogenesis of ICP (Figure 1D and Supporting Information Figure S2B).2, 3 Single-sample GSEA (ssGSEA) analysis found that most of the pathways related to liver function were only significantly increased in the pre-ICP group but not in the early-diagnosed ICP group, consistent with the differences in clinical liver transferases, implying the heterogeneity between these two groups (Figure 1E and Supporting Information Table S1). Ten samples with positive HBeAg and 30 samples with negative HBeAg were included in the HBV group (Figure 2A). The detailed clinical information of HBV patients is summarized in Supporting Information Table S3. HBV RNA can be detected in 28 (70%) HBV samples and the HBeAg-positive samples presented a higher abundance than the HBeAg-negative samples (Figure 2B). Seven samples were assembled successfully and determined the HBV genotype (Supporting Information Figure S3A). Their sequencing coverage was uniform, avoiding technical errors in sequencing (Supporting Information Figure S3B & C). DAG analysis revealed 125 genes with altered abundance in HBV patients including well-established liver damage biomarkers (Figure 2C, Supporting Information Figure S4A and Table S2).4 The DAGs annotation and GSEA showed that molecular changes of HBV were associated with liver function and cell apoptosis, and these pathways were more significant in HBeAg-positive samples (Figure 2D-F). The enrichment pathway of DA-miRNA target genes is similar to that of the mRNA GSEA result (Supporting Information Figure S4B-C). Among the DAGs, DCTN3, PROESR3, SCD, SLC7A2 and hsa-miR-4443 also increased their abundance in the ICP patients, suggesting a shared molecular alteration in the pathogenesis of HBV and ICP (Supporting Information Table S2). Meanwhile, pathways related to biosynthesis and metabolism, transmembrane transport and complement and coagulation cascades were enriched in both HBV and pre-ICP groups, indicating a shared mechanism of HBV and ICP. We next investigated the feasibility of using cfRNA as a non-invasive tool for assessing liver health. Only one DAG in ICP was liver-specific, while 18 (21%) DA-mRNAs and 5 (18%) DA-miRNAs were liver-specific in HBV, and these genes were upregulated in both HBV and ICP (Figure 3A-C). Additionally, the liver and hepatocyte-specific signature increases in the HBV and pre-ICP groups and the hepatocyte score were positively correlated with alanine aminotransferase, triglyceride, γ-glutamyl transpeptidase and HBV RNA level (Figure 3D-F and Supporting Information Figure S5). Elevated hepatocyte scores from samples with liver cancer further confirmed that hepatocyte signatures are a robust indicator to assess liver health status (Figure 4G).5, 6 The placental and fetal-specific signatures were also investigated. However, no significant alteration was observed except for the fetal liver lymphoid cell (Supporting Information Figure S6), possibly due to the scarcity of fetal signals in early pregnancy and the low TBA concentration. To assess whether plasma cfRNA signatures could identify mothers at risk of ICP, we built a machine learning model using samples from the pre-ICP group and healthy pregnant women. Samples were split into training and validation sets according to the time of blood collection (Figure 4A). A multi-split method with the least absolute shrinkage and selection operator and random forest algorithms was used to select stable features. A gradient boosting machine with seven-fold cross-validation was applied to construct predictive models using mRNA, miRNA and lncRNA, respectively (Figure 4A and Supporting Information Figure S7A). Five mRNA genes (ABCA2, SLC7A2, PXK, RNF141 and FCHO2) were finally selected to construct the model (Figure 4B and Supporting Information Figure S7B). PXK and FCHO2 have been reported to be associated with alkaline phosphatase, aspartate aminotransferase and total cholesterol levels from the genome-wide association studies.7-9 The predictive model using mRNA obtained an AUROC of 0.80 and an accuracy of 0.81 in the validation set (Figure 4C&D), which is comparable with the model using clinical features.10 The model can be applied to ICP risk stratification after being validated by a large multi-center population, complementing the efforts based on clinical data. Eight miRNAs were selected to construct the model, but the AUROC is 0.54 in the validation set (Supporting Information Figure S7C and Table S4). LncRNA cannot produce stable features to train a model after multi-split iterations. In conclusion, we demonstrated that ICP pathophysiology-related changes were identified from plasma cfRNAs in early pregnancy. These changes were distinct in patients with early- and late-diagnosed ICP, with the latter showing similar changes to HBV carriers, implicating different biological processes that drive the underlying pathophysiology of these subtypes of ICP. Additionally, plasma cfRNAs provided a non-invasive means to early identify women at risk for ICP during pregnancy, which could help guide the precision management of pregnancy. We are grateful to the participants for supporting this study. We thank Dr. Jianming Zeng (University of Macau), and all the members of his bioinformatics team, and biotrainees, for generously sharing their experience and codes. The authors declare that they have no competing interest. This work was supported by the research grant from the sub-project of the National Key R&D Program (2022YFC2703702, 2021YFC2701002), National Natural Science Foundation of China (Nos. 81971344, 82171677, 82192864, 82088102 and 81901495), the Shanghai Municipal Commission of Science and Technology Program (22S31901500, 23ZR1408000, 21Y21901002), Shanghai Municipal Commission of Health and family planning (202140110), and CAMS Innovation Fund for Medical Sciences (2019-I2M-5-064), Collaborative Innovation Program of Shanghai Municipal Health Commission (2020CXJQ01), Clinical Research Plan of SHDC (SHDC2020CR1008A), Shanghai Clinical Research Center for Gynecological Diseases (22MC1940200), Shanghai Urogenital System Diseases Research Center (2022ZZ01012) and Shanghai Frontiers Science Research Center of Reproduction and Development. The plasma cfRNA data that support the findings of this study have been deposited into the Genome Sequence Archive for Human Database with accession numbers HRA003387. Plasma cfRNA data from samples with liver cancer and healthy subjects were downloaded from the Gene Expression Omnibus database with accession numbers GSE182824, GSE174302 and GSE142987. The code of cfRNA alignment and quantification is available at https://github.com/wonderful1/PALM-Seq-cfRNA. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.