Biomarkers for complete hydatidiform moles

Abstract

The application belongs to the technical field of medical diagnosis, and provides a biomarker of complete hydatidiform mole. The biomarker comprises RASA1, the biomarker shows significant differential expression in normal, complete hydatidiform mole trophoblast stem cells and normal, complete hydatidiform mole tissues; can be used for distinguishing complete hydatidiform mole, early complete hydatidiform mole, non-hydatidiform mole villous edema and normal placental tissue, and shows good robustness. Compared with the serum β-hCG level, the expression level of RASA1 shows higher diagnostic efficiency for complete hydatidiform mole detection. The biomarker of the application provides a basis for future development of targeted diagnosis and treatment.

Description

Technical Field

[0001] This invention belongs to the field of medical diagnostic technology, and specifically relates to a biomarker for complete hydatidiform mole. Background Technology

[0002] The information disclosed in this background section is intended to enhance understanding of the overall background of the invention and is not necessarily to be construed as an admission or in any way implying that such information constitutes prior art known to those skilled in the art.

[0003] Hydatidiform mole poses a significant threat to human reproductive health, characterized by abnormal trophoblastic proliferation, impaired embryonic development, and increased invasive potential. Early treatment of hydatidiform mole can achieve a cure and preserve fertility. However, the molecular and cellular heterogeneity of the pathogenesis of hydatidiform mole remains poorly understood, hindering the development of reliable early diagnostic biomarkers and accurate risk stratification for invasive progression. Delayed diagnosis increases the risk of serious complications, including hyperthyroidism, hyperemesis gravidarum, early-onset preeclampsia, and life-threatening bleeding. Patients with a history of complete hydatidiform mole (CHM) have a significantly increased risk of developing gestational trophoblastic neoplasms (GTNs, including invasive hydatidiform mole and choriocarcinoma), estimated to be approximately 1000 times higher than in normal pregnancies.

[0004] The pathological diagnosis of hydatidiform mole relies on assessing villous edema to differentiate it from non-hydatidiform mole, and on immunohistochemical detection of p57 expression in cytotrophoblastic and villous stromal cells to differentiate between complete and partial hydatidiform mole. A significant proportion of very early complete hydatidiform mole (VCEM, defined as complete hydatidiform mole diagnosed at ≤12 weeks of gestation) and approximately 50% of true partial hydatidiform mole present significant diagnostic challenges due to the lack of distinctive histomorphological features, making accurate diagnosis by routine histopathology difficult. Given the varying risks of progression and monitoring needs among different types of hydatidiform mole, accurate differentiation between hydatidiform and non-hydatidiform mole, as well as precise subtyping of hydatidiform mole, are crucial for appropriate clinical management. Inaccurate pathological diagnosis not only hinders the establishment of reliable clinical and research cohorts but also impairs the credibility and reproducibility of clinical and translational studies. Summary of the Invention

[0005] To address the problems in the prior art, this invention provides a biomarker for detecting complete hydatidiform mole, which has high diagnostic efficacy and good robustness.

[0006] To achieve the above objectives, the present invention adopts the following technical solution.

[0007] A composition for classifying placental tissue types, comprising: a reagent that specifically binds to RASA1 (Ras p21 protein activator 1) mRNA or protein;

[0008] The placental tissue type is hydatidiform mole and non-hydatidiform mole; the non-hydatidiform mole is non-hydatidiform mole with villous edema and normal placenta; the hydatidiform mole is complete hydatidiform mole.

[0009] In some embodiments, the placental tissue type is very early complete hydatidiform mole, complete hydatidiform mole, non-hydatidiform mole with villous edema, and normal placenta.

[0010] The reagents are selected from primers for amplifying RASA1 mRNA, aptamers or antibodies targeting RASA1 mRNA or protein.

[0011] The mRNA is either a precursor mRNA or a mature mRNA.

[0012] The antibody is a polyclonal antibody or a monoclonal antibody. Monoclonal antibodies are preferred.

[0013] The antibody is also conjugated with labeled molecules, such as fluorescent groups and horseradish peroxidase.

[0014] The above composition may also include reagents that specifically bind other biomarker mRNA or proteins; said biomarkers include at least one of MYCN (human N-myc proto-oncogene protein), CSH1 (human chorionic gonadotropin 1), and PAPPA (pregnancy-associated protein A).

[0015] The present invention also provides a kit comprising the above-described composition.

[0016] The present invention has the following advantages:

[0017] This invention utilizes snRNA-seq to detect different cell types in placental tissues from normal and complete hydatidiform moles, obtaining information such as the types of upregulated and downregulated genes in different cell types. By constructing different machine learning models, the biomarker RASA1 for complete hydatidiform moles was screened and obtained. This biomarker showed significant differential expression in trophoblastic stem cells of normal and complete hydatidiform moles, as well as in normal and complete hydatidiform mole tissues. It can be used to differentiate between complete hydatidiform moles, very early complete hydatidiform moles, non-hydatidiform moles with chorionic villus edema, and normal placental tissue, and exhibits good robustness. Compared with serum β-hCG levels, RASA1 expression levels showed higher diagnostic efficacy for complete hydatidiform moles. The biomarkers of this invention provide a foundation for the future development of targeted diagnostics and treatments. Attached Figure Description

[0018] Figure 1 These are abnormal developmental features of the main trophoblast cell types in HM;

[0019] Figure 2 The distribution of gene counts among samples in snRNA-seq data;

[0020] Figure 3 The expression of marker genes used for annotation in different cell types is shown, where the color intensity represents the relative expression level and the size of the dot represents the percentage of cells expressing the corresponding gene;

[0021] Figure 4 The classification accuracy of the three models for different cell types (A), the number of positive and false positive cells in SCT and VCT cells in the MLP model (B), and the gene ranking in SCT cells predicted by the MLP model based on SHAP value (C).

[0022] Figure 5 The expression density curves of RASA1 (A), CSH1 (B), and CDKN1C (C) in SCT cells from different sources are shown.

[0023] Figure 6 Immunofluorescence images of sections of hydatidiform mole and normal trophoblast tissue (A) and relative fluorescence intensities of RASA1 and hCGB (B);

[0024] Figure 7 The protein blots (A) and relative mRNA expression levels (B) of RASA1 in hydatidiform mole and normal trophoblast stem cells.

[0025] Figure 8 Immunohistochemical analysis of RASA1 in different placental tissues (A) and relative content (B).

[0026] Figure 9 ROC curves for RASA1 and β-hCG. Detailed Implementation

[0027] This invention, through snRNA-seq analysis of complete hydatidiform mole and normal placenta, shows that ( Figure 1The presence of abnormal developmental features in the main trophoblastic cell types in hydatidiform mole (HM) is a key indicator. Specifically, during trophoblastic cell development, the three main trophoblastic cell lineages exhibit different differentiation relationships and functional associations: VCT (Vascular Tissue Trophoblasts) are stem trophoblasts that can differentiate into hormone-secreting SCT (Self-Tissue Trophoblasts) and invasive migratory EVT (Electrotrophic Vessels). In VCT, HM lacks the VCT1 progenitor cell subset, and its core stem cell transcription factor TP63 is inactivated. EVT exhibits enhanced invasiveness and migration in HM, with overactivation of the transcription factor MYCN and increased crosstalk with the immune microenvironment. SCT-Mature1 shows impaired maturation and downregulated expression of placental-specific proteins (including PSG, CSH, and PAPPA). Machine learning analysis of all differentially expressed genes revealed RASA1 as a novel key gene, and its involvement in SCT differentiation was further validated using the hTSC cell line. These abnormal trophoblastic cell lineages collectively contribute to the pathophysiological phenotype of hydatidiform mole.

[0028] Based on the above results, the present invention provides a composition for distinguishing different placental tissue types, comprising: a reagent that specifically binds to RASA1 mRNA or protein; such as primers for amplifying RASA1 mRNA, aptamers or antibodies targeting RASA1 mRNA or protein. The reagent is used to determine the content of RASA1 mRNA and the expression level of RASA1 protein.

[0029] The mRNA can be a precursor or mature mRNA. Primers for amplifying RASA1 mRNA and aptamers targeting RASA1 mRNA can target different RASA1 transcripts.

[0030] The antibody can be a polyclonal or monoclonal antibody. To complete the signal output, the antibody may also be conjugated with a labeling molecule, such as a fluorescent group or horseradish peroxidase.

[0031] The above composition may also include reagents that specifically bind to other biomarker mRNA or proteins; the biomarkers include at least one of MYCN (human N-myc proto-oncogene protein), CSH1 (chorionic chorionic gonadotropin 1), and PAPPA (pregnancy-associated protein A); the reagents are used to determine the content of other biomarkers mRNA or protein expression levels; such as primers for amplifying mRNA, aptamers or antibodies targeting the mRNA or protein of the biomarker.

[0032] The placental tissue type is either complete hydatidiform mole or incomplete hydatidiform mole. Specifically, the placental tissue type is either complete hydatidiform mole, non-hydatidiform mole with villous edema, or normal placenta.

[0033] The proteins, mRNA precursors, and different transcripts (mature mRNAs) of human RASA1, MYCN, CSH1, and PAPPA can all be obtained from publicly available databases. Therefore, mRNA primers, aptamers, or antibodies targeting the above biomarkers can be obtained through existing design principles and screening methods.

[0034] The present invention will be further described below with reference to the embodiments and accompanying drawings, but the present invention is not limited to the following embodiments.

[0035] The human placental tissue analyzed in this embodiment was obtained through standardized clinical collection procedures, including: pathological specimens from confirmed cases of hydatidiform mole and control specimens from cases of selective termination of pregnancy. All donors of placental tissue voluntarily signed informed consent forms regarding the use of placental tissue for research.

[0036] Example 1: Biomarkers for complete hydatidiform mole

[0037] 1. Identification of complete hydatidiform mole

[0038] Placental tissue was collected from hydatidiform moles and identified as complete hydatidiform moles by P57 and Ki67 staining. STR genotyping confirmed the absence of maternal alleles in the villi of complete hydatidiform moles. Copy number variation (CNV) analysis showed that complete hydatidiform moles were uniparental diploid, while control placentas were biparental diploid. These results confirmed the uniparental origin of complete hydatidiform moles. All analyzed androgenic complete hydatidiform moles had a 46,XX karyotype; therefore, female control placentas were selected for sex-matched comparison.

[0039] 2. Molecular pathological changes in complete hydatidiform mole

[0040] Six cases of complete hydatidiform mole placentas (HM2, 4, 6, 15, 16, 17) and seven healthy controls (PLA1, 5, 8, 9, 10, 12, 14) underwent 10×Genomics snRNA-seq. After quality control, the snRNA-seq data preserved 114,662 cell nuclei, demonstrating high-quality data. Figure 2 The snRNA-seq data were integrated and batch-corrected, followed by dimensionality reduction, clustering, and cell type annotation.

[0041] Eleven major cell types were annotated based on placental markers in early pregnancy. Figure 3Based on proliferation and fusion capabilities, VCTs were classified into proliferating VCTs (VCT-p) and fusing VCTs (VCT-fusing). Furthermore, stromal cells were classified into decidual stromal cells (stromal-decidua) and stromal stromal cells (stromal-villi) based on specific biomarkers. The results showed sample-level consistency and high cell subtype correlation with published early placental snRNA-seq datasets (Vento-Tormo et al., 2018), thus validating the accuracy of cell annotation.

[0042] Significantly differentially expressed genes were identified in VCT, proliferating VCT (VCT-p), fusing VCT, SCT, EVT, stromal-villi, stromal-decidua, endothelial, epithelial, Hofbauer, and dNK cells (Table 1).

[0043] Table 1. Significantly Differential Genes in Each Cell

[0044]

[0045] 3. Machine Learning for Identification of Key Genes in Hydatidiform Mole

[0046] Three binary classifier models were constructed for each major cell type with sufficient cell count: XGBoost, LightGBM, and multilayer perceptron (MLP) classifier to distinguish between hydatidiform moles and control cells.

[0047] First, feature engineering was performed to retain genes showing significant differential expression between the two groups, and cells with consistently high data quality were selected. The resulting dataset was then stratified by class label and divided into an 80% training + validation set and a 20% independent test set using StratifiedShuffleSplit. The former was further split into 80% / 20% to obtain the final training set and internal validation set, while maintaining class balance.

[0048] The hyperparameters are optimized using Optuna (TPE sampler, ≤50 trials) with the goal of minimizing 1-accuracy. The search is terminated via a custom callback function when the best validation accuracy does not improve by ≥5% for three consecutive trials.

[0049] For XGBoost, eval_metric is fixed at "mlogloss", allowing a maximum of 200 rounds of improvement, and early stopping is applied (patience value is 10).

[0050] For LightGBM, set metric="multi_logloss" and optimize for ≤1000 iterations (with the same patience value).

[0051] The MLP consists of two fully connected layers (Hidden Layer 1: 64-256 units; Hidden Layer 2: half the size of Hidden Layer 1), followed by batch normalization, ReLU activation, and dropout with a rate of 0.2-0.5. The Adam optimizer is used (learning rate 1×10⁻⁶). -5 -1×10 -2 Training was conducted with batch sizes of 16-128 and 50-2000 epochs, and the process was stopped when the validation accuracy stabilized for 10 consecutive epochs.

[0052] After finding the optimal hyperparameters, each model was retrained on the merged training and validation data. Accuracy was reported on an untouched test set, and a confusion matrix was plotted. Model interpretability was assessed using SHAP (v0.47). TreeExplainer was applied to tree-based models, and DeepExplainer was applied to MLPs. 10% of cells were randomly selected as background, and 5% as the evaluation subset. The average absolute SHAP values ​​among cells were calculated, sorted, and visualized as a hierarchical violin plot.

[0053] Feature importance was further evaluated by: (i) retraining a simplified model containing only the top SHAP genes and comparing its test accuracy with the corresponding full model; (ii) performing gene-level ROC analysis, for each top SHAP gene, displaying the log-normalized expression distribution in hydatidiform mole and control cells using a nuclear density estimation plot, calculating the ROC curve and AUC, determining the optimal threshold using |TPR-FPR|, and labeling the threshold and AUC on the KDE plot to illustrate the discriminative power of single genes.

[0054] Classification performance evaluations of different cell types showed that MLP achieved the highest overall accuracy. Figure 4 (Among them, A), the trophoblast lineage provides the most information. Specifically, in the MLP model, SCT and VCT have the highest accuracy ( Figure 4 (B) To analyze the contribution of gene levels to SCT-based identification of hydatidiform mole, SHapley Additive Explanation (SHAP) analysis was used to interpret the SCT-MLP model. RASA1 expression became the main driving factor distinguishing hydatidiform mole from the control group ( Figure 4 (C)

[0055] To validate the specificity of RASA1 in semi-contrast cytosis (SCT) of hydatidiform mole, the SCT expression matrix was extracted, receiver operating characteristic (ROC) curves were plotted based on RASA1 expression, the area under the curve (AUC) was calculated, and the optimal cutoff value was determined using the Youden index. These metrics were benchmarked against SHAP-ranked candidate gene CSH1 and the recognized negative marker for complete hydatidiform mole, CDKN1C (P57). Notably, RASA1 exhibited the highest diagnostic performance, with robust differentiation between hydatidiform mole and control achieved with single-gene expression. Figure 5 ).

[0056] Example 2: Identification of RASA1 in hydatidiform mole and normal trophoblast.

[0057] Immunofluorescence analysis of tissue sections from hydatidiform mole and control groups (n=3 per group) confirmed that RASA1 was lowly expressed in the villi of hydatidiform mole, but significantly expressed in normal placenta. Figure 6 (A). Furthermore, RASA1 exhibits subcellular localization differences in the normal placenta, primarily cytoplasmic in the VCT and nuclear in the SCT, suggesting that RASA1 may play a biological role in the differentiation of the syncytiotrophoblast (SCT). Figure 6 (B)

[0058] Human trophoblastic stem cell (TSC) lines were established using hydatidiform mole and normal placental villi from early pregnancy. Western blot and qPCR analyses confirmed that RASA1 expression was significantly reduced in HM-TSCs, consistent with the RASA1 downregulation observed in HM placental tissue. Figure 7 (A, B)

[0059] Application Example 1: Detection of Complete Hydatidiform Mole

[0060] An independent cohort was used to validate RASA1 expression in different placental tissues. The cohort included 38 cases of complete hydatidiform mole, 7 cases of very early complete hydatidiform mole (VECM, ≤8 weeks of gestation), 4 cases of non-hydatidiform mole with chorionic villi, and 21 gestationally matched normal placenta controls.

[0061] Immunohistochemical analysis showed that RASA1 expression was low in all cases of complete hydatidiform mole (including VECM); villous edema in non-hydatidiform mole showed normal RASA1 expression levels. Figure 8 (A, B). This suggests that RASA1 can serve as a robust biomarker even in challenging situations, such as identifying VECM and differentiating complete hydatidiform mole from non-hydatidiform mole edematous villi.

[0062] The diagnostic performance of RASA1 and β-hCG was further compared. Compared with serum β-hCG levels, RASA1 expression levels showed higher power in the detection of complete hydatidiform mole (AUC = 0.98 vs. 0.85). Figure 9 ).

[0063] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. Use of a composition for the preparation of a reagent for classifying placental tissue types, characterized in that, The composition includes a reagent that specifically binds to the RASA1 protein; The placental tissue type is hydatidiform mole and non-hydatidiform mole; the non-hydatidiform mole is non-hydatidiform mole with villous edema and normal placenta; the hydatidiform mole is complete hydatidiform mole.

2. Use according to claim 1, characterized in that, The reagent is an aptamer or antibody for a protein.

3. Use according to claim 2, characterized in that, The antibody is a polyclonal antibody or a monoclonal antibody.

4. Use according to claim 2, characterized in that, The antibody is a monoclonal antibody.

5. Use according to claim 4, characterized in that, The antibody is also conjugated with a labeling molecule.

6. Use according to claim 5, characterized in that, The labeled molecule is selected from fluorescent groups or horseradish peroxidase.

7. Use according to claim 1, characterized in that, The composition also includes reagents that specifically bind to other biomarker proteins; said biomarkers include at least one of MYCN, CSH1, and PAPPA.

Images