In this study, we collected placenta samples and clinical data from 10 normal pregnancy patients and 10 ICP patients who underwent cesarean or vaginal delivery. The diagnostic criteria of ICP were gestational pruritus, and the concentration of serum bile acids was greater than 10 umol/L, and the details of all patients are shown in the supplementary file (Table 1). All samples were obtained with informed patients’ consent before collection and approved by the Guangzhou Women and Children’ Medical Center Ethics Committee. Moreover, the collected samples were fixed in formalin or maintained at −196°C (liquid nitrogen) for further study.

Quantitative proteomics of data-independent acquisition (DIA) integrated “shotgun” and a technique of selected reaction monitoring (SRM), also known as multiple reaction monitoring (MRM), and it performed quantitative and qualitative assessments based on mass-to-charge (m/z). The collected placental specimens underwent a series of protein extraction, protease dissociation, database establishment, mass spectrum, DIA, quality control, and protein identification steps, and we obtained proteins of samples, according to the detected spectrum counting and peak. The mass spectrometry data were collected using a Q Exactive Plus mass spectrometer in tandem EASY-nLC 1,200 liquid phase. The details of the parameters are as follows: analysis column (50 um*15 cm, C18, 2 um, 100 Å), mobile phase (A: 0.1% formic acid, B: 0.1% formic acid and 80% ACN), velocity (300 nL/min), MS1 (R = 70K, AGC = 3e6, Max IT = 30 ms, scan range = 350–1,250 m/z), MS2 (R = 17.5 K, AGC = 1e6, Max IT = 50 ms), and collisional energy (value = 28). The differentially expressed proteins were identified according to the fold change (FC) of the protein expression (FC > 1.2 or FC < 0.83, p < 0.05).

OPLS-DA were powerful statistical modeling tools (Worley and Powers, 2016). In this study, it was used to provide insights into separations between the ICP group and NC group based on high-dimensional spectral measurements.

In order to study the function of proteins, the identified proteins were annotated in the functional database. BLAST comparison (blastp, evalue ≤ 1e-5) was used to annotate GO, KEGG, and COG databases, and the comparison result with the highest score was selected as the annotation result. BLAST software combined with UniProt and InterProScan databases was used for domain annotation. The variation trend of differential proteins was evaluated by a volcano map and heatmap display analysis. The GO items, KEGG pathway, and COG functions of the differential proteins were performed to determine functional enrichment analysis. Metascape was used to further verify the summary of hub protein enrichment analysis in PaGenBase, DisGeNET, and Cell Type Signature.

STRING is a database used to predict protein–protein interactions (PPI) including direct (physical) and indirect (functional) associations (Szklarczyk et al., 2021); If there were no corresponding species in the database, the sequences of proximal species were extracted. Then, blast comparison was performed between the differential protein sequences to obtain the corresponding interaction information and construct the network. In this study, we matched PPI data in the human database, and STRING in this study was used for exploring the interactions within differentially expressed proteins of ICP placenta.

Cytoscape Web was an interactive tool, and it visualized protein interactions (Shannon et al., 2003). In this study, Cytoscape was used to visualize the PPI network from STRING and classify proteins according to their expression levels. Moreover, Cytoscape was also used to construct a competing endogenous RNA (ceRNA) network. CytoHubba, a plugin of Cytoscape, was used to rank proteins according to degree algorithms. Plugin-MCODE was used to explore chief sub-PPI networks.

BioGPS was a free extensible and customizable gene annotation portal (Wu et al., 2016). In this study, BioGPS was used to study the function and position of core proteins.

NetworkAnalyst was a visual analytics platform for gene regulation, gene co-expression, and generic and drug–gene interaction networks (Zhou et al., 2019). NetworkAnalyst in our study was used for exploring transcription factors that regulated core proteins.

In this study, PITA, miRanda, and TargetScan databases were used to identify the upstream miRNAs of core proteins. StarBase and LncBase Predicted v.2 were used to find the upstream lncRNAs of miRNAs. Furthermore, the results were taken, the intersection and network were constructed by Cytoscape.

The 18 female and 18 male Sprague–Dawley rats (body weight: 250 ± 20 g) were purchased from Guangzhou Ruige Biotechnology Co., LTD. The animals were given access to food and water. One female rat was mated with one male overnight, and the day when a vaginal plug was found was defined as day 0. The ICP rat model was constructed as follows: from day 13, the rats were injected intraperitoneally with estradiol benzoate (5 mg/kg, MCE, China) until day 20, which we named it the ICP group. The NC group was treated with equal DMSO consecutively. The pregnant rats were sacrificed, and the placenta was harvested in day 21. All experimental operations were approved by the animal ethics committee of Guangzhou Women and Children’ Medical Center Ethics Committee. As for placental explant models, the placenta was obtained from normal pregnancy patients, and it was shredded and stimulated for 48 h with 100 uM cholid acid (MCE, China) to construct the ICP model.

R packages including pheatmap, clusterProfiler, and igraph were used in this work. The antibodies used in Western blotting are as follows: actin (GB12001; 1:3000; Servicebio, China), LC3 (14600-1-AP; 1:1,000; Proteintech, United States), and Beclin 1(GB112053; 1:1,000; Servicebio, China). The AlphaView tool was used to analyze the optical density values of the target band.

Results expressed as mean ± standard error (m ± SE) were analyzed by two-tailed Student’s t-test. Fisher’s test was used for clinical data table. The statistical significance was defined as p < 0.05 by GraphPad Prism software (version 6.0, GraphPad Software, United States).

We collected placentas from 10 normal pregnancies and 10 ICP patients. In this study, OPLS-DA was an alternative method of principal component analysis. The results of OPLS-DA analysis areshown (Figure 1A). Based on the quantitative proteomics of DIA and the standard of significance (FC > 1.2 or FC < 0.83, p < 0.05), we found 77 upregulated proteins and 99 downregulated proteins which were rendered as visual heatmap (Figure 1B), volcano map (Figure 1C), and histogram (Figure 1D). GO enrichment analysis showed that the differentially expressed proteins were implicated in a wide range of biological processes, such as the process utilizing the autophagic mechanism, autophagy, protein acetylation, porphyrin-containing compound metabolic process, and pigment metabolic process. The proteins that participated in the cellular component contained hemoglobin complex, haptoglobin–hemoglobin complex, autophagosome, vacuole, vacuolar membrane, and Schaffer collateral-CA1 synapse. The molecular functions were tetrapyrrole binding, oxygen carrier activity, molecular carrier activity, organic acid binding, and protein serine/threonine kinase inhibitor activity, etc. (Figure 1E). KEGG pathway analysis demonstrated that these proteins were involved in biosynthesis of second metabolites, microbial metabolism in the diverse environment, biosynthesis of cofactors, NOD-like receptor signaling pathway, autophagy, and carbon metabolism. (Figure 1F). Domain enrichment analysis revealed that Atg8-like, Ubiquitin-like_domsf, Globin/Proto, Keratin_II, Serpin_B9/Maspin, and Oxidoreductase_N were included (Figure 1G). KOG enrichment analysis showed that signal transduction mechanisms (T), lipid transport and metabolism (I), transcription (K), energy production and conversion (C), and posttranslational modification, protein turnover, and chaperones (O) were covered (Figure 1H).

The PPI network of differentially expressed proteins was made by using the STRING database and Cytoscape (Figure 2A). To further capture the relationships between proteins, Metascape was used in this study for enrichment analysis and enriched term network construction, where terms with a similarity >0.3 are connected by edges. The results displayed that the core terms included autophagy, regulation of histone H4 acetylation, regulation of TP53 activity, nucleotide biosynthetic process, and oxygen transport (Figure 2B). The proteins in the PPI network were ranked by the degree algorithm, and the results showed top 20 proteins were NEDD8, ATG5, FECH, ACAT2, HBE1, ACOX3, CTPS2, GABARAPL2, GABARAPL1, MLST8, PLEK, RANBP6, HMBS, PRKAG2, SDC1, PBDC1, ECI2, ACE, PEX14, and GNG10 (Table 1). The MCODE algorithm was applied to identify densely connected network components and showed that FBXO7, LTN1, UBE2O, NEDD8, HPRT1, GANARAPL1, GABARAPL2, and ATG5 were included in MCODE1, and CD2BP2, PPIL3, PAPOLA, and BCS2 were components of MCODE2. Furthermore, the proteins in MCODE 1 were associated with autophagy of the mitochondrion, mitochondrion disassembly, and cellular response to nitrogen starvation. The proteins in MCODE 2 were connected with the mRNA splicing-major pathway, mRNA splicing, and processing of capped intron-containing pre-mRNA (Figure 2C).

The summary of enrichment analysis in the cell type signature showed that Descartes fetal pancreas, intestine, stomach, and kidney erythroblasts were included (Figure 2D). Summary of enrichment analysis in DisGeNET showed that red blood cell disorder, slow progression, upper limb amyotrophy, myopathy, and feeding difficulties in infancy, etc., were included (Figure 2E). To seek core proteins, we ranked the proteins according to the value of expression differences (Figure 2F), and our results showed that the top 12 upregulated proteins were PSAT1, HBG1, HBM, SPI1, PIP4K2B, HBG2, HBE1, FOXK1, YOD1, KRT72, PKLR, and SIGLEC6 (Figure 3A). The top 12 downregulated proteins were HSPA13, EVA1A, EGFL7, SLC13A3, MBD2, TSFM, SP9, GPLD1, GH2, BLOC1S1, F7, and MYH7 (Figure 3B).

We further performed the enrichment analysis in PaGenBase, and our results showed that most core proteins had tissue-specific expression patterns, especially in the liver and placenta. SIGLEC6, HBG2, and GH2 were highly expressed in the placenta predominantly among core proteins (Figure 4A). The tissue-specific results were also verified by the BioGPS database (Figure 4B). Furthermore, we collected clinical data on patients, including age, delivery gestational age (week), fetal weight (g), and expression of core proteins. Our results showed that ICP development was closely associated with delivery gestational age, and HBG1, SPI1, HBG2, HBE1, FOXK1, KRT72, SLC13A3, MBD2, SP9, GPLD1, MYH7, and BLOC1S1 were linked to the occurrence of ICP. Moreover, the KRT72 expression remained very low in normal human pregnancy and extremely high in ICP patients (Table 2).

To explore the regulatory mechanism, we constructed the transcription factor-protein network. Our results showed HBG1, SPI1, HBE1, SP9, GPLD1, BLOC1S1, KRT72, and SLC13A3 were regulated by FOXC1, and HBG1, HBG2, SPI1, SP9, BLOC1S1, KRT72 and SLC13A3 were regulated by GATA2. Moreover, YY1 regulated six protein expressions, and transcription factor NFIC regulated five protein expressions (Figures 5A,B). In addition, combined with PITA, miRanda, and TargetScan databases, we found hsa-miR-155-5p regulated the SPI1 expression, hsa-miR-371a-5p was on the upstream of FOXK1, hsa-miR-383-5p and hsa-miR-543 regulated the SLC13A3 expression, and hsa-miR-520e. was on the upstream of MBD2 (Figure 5C). No miRNAs were found to target HBG1, HBG2, HBE1, KRT72, SP9, GPLD1, MYH7, and BLOC1S1. Moreover, we further studied the lncRNA on the upstream of miRNAs based on Starbase and LncBase Predicted v.2 database. MIR17HG, HCG18, MALAT1, INE1, CASC2, and SNHG5 were involved in the regulation of hsa-miR-155-5p. NEAT1, XIST, SNHG1, CKMT2-AS1, COLCA1, and TUG1 regulated hsa-371a-5p expression. NEAT1 and KCNQ1OT1 were coregulatory factor of hsa-miR-543 and hsa-miR-383-5p. As for MBD2, the targeted ceRNA network was very complex, and a variety of miRNAs and lncRNAs participated in the regulation of MBD2. KCNQ1OT1 and XIST were common regulatory molecules of all four proteins (Figure 5D).

The placental explant from normal placenta was stimulated for 48 h with 100 uM cholid acid. HE staining results showed that compared with the NC group, the placenta in the ICP group had narrower intervillous space and increased fibrinous necrosis of placental villi (Figure 6A). The Western blotting results of the explant demonstrated that compared with the NC group, the expressions of autophagy-related proteins—Beclin 1 and LC3 II/1 were reduced significantly, which was consistent with our quantitative proteomics (Figure 6B). ICP animal models demonstrated immature fetal rats and placentas were more common, and most fetal rats were not viable in vitro. However, in the NC group, all fetal rats developed well (Figures 6C–6D).