Diagnosis of malignant body fluids via cancer-universal methylation in cell-free DNA

Abstract

BACKGROUNDDifferentiating malignant from nonmalignant body fluids remains a clinical challenge because of the unsatisfying performance of conventional cytology. We aimed to improve the sensitivity and ubiquity of cancer cell detection by assaying universal cancer-only methylation (UCOM) markers in supernatant cell-free DNA (cfDNA).METHODSAn observational prospective cohort including 1,321 nonmalignant and malignant body fluids of multiple cancers was used to develop and validate a cfDNA UCOM methylation diagnostic assay. All samples were divided into 2 portions for cytology and supernatant cfDNA methylation analysis.RESULTSThe significant hypermethylation of a potentially novel UCOM marker, TAGMe, together with the formerly reported PCDHGB7, was identified in the cfDNA of malignant body fluid samples. The combined model, cell-free cancer-universal methylation (CUE), was developed and validated in a prospective multicancer cohort with markedly elevated sensitivity and specificity, and was further verified in a set containing additional types of malignant body fluids and metastases. In addition, it remained hypersensitive in detecting cancer cells in cytologically negative malignant samples.CONCLUSIONcfDNA methylation markers are robust in detecting tumor cells and are applicable to diverse body fluids and tumor types, providing a feasible complement to current cytology-based diagnostic analyses.TRIAL REGISTRATIONThis study was registered at Chictr.org.cn (ChiCTR2200060532).FUNDINGNational Natural Science Foundation of China (32270645, 31872814, 32000505, 82170088), the National Key R&D Program of Ningxia Hui Autonomous region (2022BEG01003), Shanghai Municipal Key Clinical Specialty (shslczdzk02201), Science and Technology Commission of Shanghai Municipality (20DZ2261200, 20DZ2254400), and Major Special Projects of Basic Research of Shanghai Science and Technology Commission (18JC1411101).

Keywords: Cancer; Epigenetics; Genetics; Molecular diagnosis; Oncology.

Figure 1. Diagram of workflow.

PE, pleural effusion; MPE, malignant pleural effusion; BFE, benign pleural effusion; MA, malignant ascites; BA, benign ascites; BCF, benign cerebrospinal fluid; MCF, malignant cerebrospinal fluid; BPCE, benign pericardial effusion; MPCE, malignant pericardial effusion; BBAL, benign bronchoalveolar lavage fluid; MBAL, malignant bronchoalveolar lavage fluid.

Figure 2. UCOM markers demonstrated to identify tumor cells using cfDNA in malignant pleural effusion (MPE).

(A) Significant cfDNA hypermethylation of PCDHGB7 was detected in MPE samples from lung cancer patients by pyrosequencing. (B) MSRE-qPCR–based results highly correlated with the methylation percentages using cfDNA pyrosequencing. Spearman’s r correlation and P value were calculated using GraphPad Prism 9.3.0. (C) MSRE-qPCR simultaneously revealed PCDHGB7 cfDNA hypermethylation in MPE samples. (D) ROC analysis demonstrates complementary performance of the marker identified in this study, TAGMe, and PCDHGB7 across multiple cancer types in TCGA database. P values were calculated by pair-wise comparison of ROC curves test with SPSS 20.0. *P < 0.05; ***P < 0.001; ****P < 0.0001.NS, not significant. (E) cfDNA hypermethylation of TAGMe was also validated in MPE samples by MSRE-qPCR. (F) ROC analysis demonstrated the performance of the 3 assays. (G) Complementary results for 2 markers were found in the same sample. (H) The sensitivity can be maximized by a 2-marker detection in a single-positive-for-positive method. Black circles: nonmalignant samples; pink circles: malignant samples with positive cytology; green squares: malignant samples in which cytology failed to detect tumor cells. BLCA, bladder cancer; BRCA, breast cancer; CESC, cervical cancer; CHOL, bile duct cancer; COADREAD, colon and rectal cancer; ESCA, esophageal cancer; GBM, glioblastoma; HNSC, head and neck cancer; KIRC, kidney clear cell carcinoma; KIRP, kidney papillary cell carcinoma; LIHC, liver cancer; LUNG, lung cancer; PAAD, pancreatic cancer; PRAD, prostate cancer; UCEC, endometrioid cancer. P values in A, C, and E were calculated using a 2-tailed, nonparametric Mann-Whitney test as determined by GraphPad Prism 9.3.0.

Figure 3. Development and validation cfDNA UCOM diagnostic model in a prospective multicancer cohort.

(A) Methylation levels of cfDNA UCOM markers are significantly higher in MPE and ascites than in benign ones from patients with benign diseases or cancers in the training set. BBF, benign body fluid. (B) ROC analysis shows that the CUE model yielded the highest AUC compared with a single UCOM marker as well as the conventional tumor markers. (C) The clinical performance of the CUE model using an optimal threshold is illustrated. (D) Consistent performance was observed in the test set at the same threshold. (E and G) The AUC of the CUE model in different types of sample (E) and primary cancer (G) in training and test sets. (F and H) The clinical performance of the CUE model between sample types (F) and cancer types (H) was evaluated. P values in A, C, and D were calculated by the 2-tailed, nonparametric Mann-Whitney test, whereas those in B were determined by the Kruskal-Wallis test, and calculations were all performed via GraphPad Prism 9.3.0.

Figure 4. Additional evidence of the CUE model’s capacity for malignancy detection.

(A–C) cfDNA methylation measured by CUE value was significantly associated with meningeal, pericardial, and pulmonary metastases in CSF (A), PCE (B), and BALF (C) samples, respectively. (D) The AUC of the CUE model for detection of different metastases is shown, as well as the clinical performance (E). The CUE model also serves to identify malignant mesothelioma in effusion samples (F) with robust sensitivity (G). P values in A–C and F were calculated using a 2-tailed, nonparametric Mann-Whitney test as determined by GraphPad Prism 9.3.0.

Figure 5. Performance of the CUE model in malignant samples with cytologically undetectable cancer cells.

(A) The clinical information of 38 malignant body fluids that were diagnosed as negative by cytology analysis. (B–D) The CT images of 3 typical cases are demonstrated. (B) Primary malignant pleural mesothelioma. (C) Recurrent lung adenocarcinoma. (D) Pleural metastasis of laryngeal cancer. Yellow arrow: tumor site. (E) The positive rate of the CUE model was optimal compared with an individual UCOM marker and conventional cancer markers.