Metabolites and CG site methylation as markers for the diagnosis of liver cancer

Abstract

The present invention discloses metabolites and CG site methylation markers for liver cancer diagnosis. The metabolite is phosphatidylcholine, and the CG site methylation includes cg05166871, cg14171514, and cg18772205. Using the markers of the present invention, the cumulative incidence of HCC in the low-risk group and the high-risk group was 1.19% and 21.40%, respectively. The cumulative incidence of HCC in the high-risk group was 10.2% at 12 months and 21.40% at 16.33 months, which was significantly higher than that in the low-risk group, where the cumulative incidence of HCC was 1.19% at both 12 and 16.33 months. Therefore, this marker helps to identify HBVLD patients at risk of HCC and conduct early diagnosis and intervention screening.

Description

Technical Field

[0001] The present invention belongs to the technical field of cancer diagnosis, and specifically relates to a metabolite and a CG site methylation marker for liver cancer diagnosis. Background Art

[0002] Liver cancer refers to malignant tumors that develop in the liver. It includes primary liver cancer and metastatic liver cancer. When people refer to liver cancer, they often refer to primary liver cancer. Primary liver cancer can be categorized by cell type: hepatocellular carcinoma, cholangiocarcinoma, and mixed liver cancer. Tumor morphology can be divided into nodular, massive, and diffuse types.

[0003] Chronic hepatitis B and cirrhosis are common hepatitis B-related liver diseases with a large population base. Some patients are at risk of developing liver cancer. How to evaluate the risk of hepatitis B-related liver diseases developing into liver cancer is an urgent problem to be solved by technicians in this field. Summary of the Invention

[0004] The purpose of the present invention is to provide a metabolite and CG site methylation marker for liver cancer diagnosis.

[0005] A marker for diagnosing liver cancer, wherein the marker is phosphatidylcholine (PC) and CpG.

[0006] The phosphatidylcholine is 1-hexadecyl-2-dihydroglycyl-enoyl-sn-glycero-3-phosphocholine (O-16:0 / 20:3(8Z,11Z,14Z)) and 1-palmitoleoyl-2-(1-enyl-stearoyl)-sn-glycero-3-phosphocholine (16:1(9Z) / P-18:0).

[0007] The CpGs include cg05166871, cg14171514 and cg18772205.

[0008] Compared with patients with hepatitis B-related liver disease, the plasma phosphatidylcholine content in patients with liver cancer is decreased.

[0009] Compared with patients with hepatitis B-related liver disease, the differential expression of phosphatidylcholine in patients with liver cancer was positively correlated with total bile acid.

[0010] Compared with patients with hepatitis B-related liver disease, the differential expression of phosphatidylcholine in patients with liver cancer was negatively correlated with high-density lipoprotein, low-density lipoprotein, triglycerides, and cholinesterase.

[0011] Use of phosphatidylcholine and / or CpG in preparing a reagent for detecting liver cancer.

[0012] Beneficial effects of the present invention: Using the marker of the present invention, the cumulative incidence of HCC was 1.19% in the low-risk group and 21.40% in the high-risk group, respectively. The cumulative incidence of HCC in the high-risk group was 10.2% at 12 months and 21.40% at 16.33 months, significantly higher than that in the low-risk group, where the cumulative incidence was 1.19% at both 12 and 16.33 months. Therefore, this marker can help identify HBVLD patients at risk for HCC, enabling early diagnosis and intervention screening. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is a flow chart of the research of the present invention.

[0014] Figure 2 Methylation levels of the four CpG sites in the liver disease group and HCC group (p value of the two-sided Wilcoxon rank sum test).

[0015] Figure 3 Pearson correlation of QC samples in plasma for quality control examination, ESI- (A) and ESI+ (B) models; ESI-, negative electrospray ionization; ESI+, positive electrospray ionization; QC, quality control.

[0016] Figure 4 3D scores for all samples using principal component analysis (PCA), including QC samples, for ESI- (A) and ESI+ (B) models; ESI-, negative electrospray ionization; ESI+, positive electrospray ionization; QC, quality control.

[0017] Figure 5 Metabolite profiles of HCC and HBVLD plasma samples. OPLS-DA score plots depicting the differentiation of metabolomic features in the ESI-model (A) and the ESI+ model (D). OPLS-DA cross-validation with 200 repetitions of permutation tests in the ESI-model (B) and the ESI+ model (E). Volcano plots showing metabolites in HCC and HBVLD plasma samples in the ESI-model (C) and the ESI+ model (F). The horizontal dashed line indicates the p = 0.05 threshold, and the vertical dashed line indicates the 0-fold threshold. The two groups were compared using the Student t-test. HBVLD, HBV-related liver disease; HCC, hepatocellular carcinoma; VIP, variable importance in projection; ESI-, negative electrospray ionization; ESI+, positive electrospray ionization; orthogonal partial least squares discriminant analysis.

[0018] Figure 6Matching analysis of key differential metabolites between the HCC group and the HBVLD group in terms of ESI- (A) and ESI+ (B) * represents significance, *0.01 < p < 0.05, **0.001 < p < 0.01, ***p < 0.001; VIP, variable importance in projection; HBVLD, HBV-related liver disease; ESI-, negative electrospray ionization; ESI+, positive electrospray ionization; HCC, hepatocellular carcinoma.

[0019] Figure 7 Pathway enrichment analysis of pathways significantly altered in HCC compared with HBVLD patients. (A) Significantly downregulated (A) and upregulated (B) pathways. Red fonts indicate mitochondrial β-oxidation of fatty acids of different chain lengths. (C) Non-esterified fatty acid levels were significantly reduced in HCC plasma. Volcano plots show lipids esterified with fatty acids, including CE, ceramide, DG, PA, LysoPA, PE, LysoPE, PI, PS, SM (D) and PC, LysoPC (E). Labeled metabolites were processed with thresholds of VIP ≥ 1.0 and p < 0.05. The horizontal dashed line indicates the p = 0.05 threshold, and the vertical dashed line indicates the 0-fold threshold. Red fonts indicate the two PCs used to develop the HCC model. HBVLD, HBV-related liver disease; HCC, hepatocellular carcinoma; VIP, variable importance in prediction; CE, cholesterol ester; DG, diacylglycerol; PA, phosphatidic acid; lysophosphatidic acid; PC, phosphatidylcholine; lysophosphatidylcholine; PE, phosphatidylethanolamine; lysophosphatidylethanolamine; PI, phosphatidylinositol; PS, phosphatidylserine; SM, sphingomyelin.

[0020] Figure 8 Box plots of the chemical structures and normalized peak heights of 11 PCs between the HBVLD group and the HCC group.

[0021] Figure 9 Pearson correlation analysis between 11 PCs and (A) non-esterified fatty acids, (B) total bile acid, (C) high-density lipoprotein, (D) low-density lipoprotein, (E) triglycerides, and (F) cholinesterase. Correlation analysis was performed using linear regression analysis. Each colored dot represents an individual patient.

[0022] Figure 10Evaluation of the diagnostic and predictive performance of the HCC model. (A) Area under the receiver operating characteristic curve (AUROC) of the HCC model in the discovery dataset. (B) Calibration curve of the HCC model in 20 stacked imputed complete datasets. The x-axis represents the predicted probability; the y-axis represents the actual HCC incidence. The black diagonal dashed line represents perfect performance of the ideal nomogram. The red dashed line represents the performance of the model; the green solid line represents the bootstrap-corrected performance of the model. (C) Time-dependent receiver operating characteristic curves and AUROC of the model at 6, 12, and 18 months in the HBVLD group. (D) Cumulative incidence of HCC based on risk group scores calculated by the model. Kaplan-Meier curves are used to show the cumulative incidence of HCC in the low-risk group (score <0.45) and the high-risk group (score ≥0.45). P values were determined using the log-rank test. HBVLD, HBV-related liver disease; HCC, hepatocellular carcinoma. DETAILED DESCRIPTION

[0023] To facilitate understanding of the present invention, the present invention will be described more fully below. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive.

[0024] Example 1

[0025] 1. Materials and Methods

[0026] 1. Patients and data collection

[0027] Between August 2010 and July 2019, a total of 400 plasma samples were collected at Beijing You'an Hospital, Capital Medical University, including 200 hepatitis B-related liver disease (HBVLD) samples: 100 chronic hepatitis B (CHB) and 100 liver cirrhosis (LC); and 200 hepatocellular carcinoma (HCC) samples. Fresh blood was collected in citrate and EDTA, and plasma and PBMC were separated by Ficoll-Hypaque density gradient centrifugation within 4 h. Plasma samples were stored at −80°C. The diagnostic criteria for CHB were based on the Asia-Pacific Clinical Practice Guidelines for the Management of Hepatitis B: 2015 Update (14), the diagnosis of HBV-related LC was based on the Guidelines for the Prevention and Treatment of Chronic Hepatitis B (2010 edition), and the diagnosis of HCC was based on the 2012 EASL Clinical Practice Guidelines: Management of Chronic Hepatitis B Virus Infection and the Barcelona Clinic Cancer Staging System (BCLC). The methylation levels of 34 CpGs were analyzed using multiple bisulfite sequencing (MBS) and the results of the inventors' previous studies. The raw data of CpGs methylation were analyzed by CapitalBio Technology. The HBLVD group adopted a retrospective design. Follow-up data of HBLVD patients were obtained by reviewing electronic medical records or telephone communications. The last follow-up was on August 5, 2022, and the primary endpoint was the incidence of HCC. The workflow of the present invention is as follows Figure 1 All patients and their families signed informed consent, and ethical approval was obtained from the Ethics Committee of Beijing You'an Hospital, Capital Medical University (Jing You Lun Zi

[2020] #123). This study was conducted in accordance with the principles of the Declaration of Helsinki.

[0028] 2. Non-targeted metabolomics by HPLC-MS

[0029] Metabolic extracts were obtained from plasma samples and analyzed by liquid chromatography-mass spectrometry (LC-MS) using an ultra-high-performance liquid chromatography (UHPLC) system (Vanquish, Thermo Scientific, San Jose, CA, USA) equipped with a Waters ACQUITY UPLC BEH Amide column (2.1 mm × 100 mm, 1.7 μm). The mobile phase consisted of 25 mmol / L ammonium acetate and 25 mmol / L ammonia in water (phase A) and acetonitrile (phase B). Detailed parameters were as follows: shield gas flow rate: 50 Arb, auxiliary gas flow rate: 15 Arb, capillary temperature: 320°C, full MS resolution: 60,000, MS / MS resolution: 15,000, collision energy: 10 / 30 / 60 in NCE mode, and spray voltage: 3.8 kV (positive) or -3.4 kV (negative).

[0030] Raw data were processed by Shanghai Sangon Biotech. ProteoWizard software converted the raw data into mzXML format. The R package "XCMS" was used for peak identification, extraction, alignment, and integration. Substances were then annotated using the in-house developed secondary mass spectrometry database, BiotreeDB (V2.1). The algorithm score cutoff was set to 0.3.

[0031] 3. Statistical analysis

[0032] Continuous variables are expressed as mean ± standard deviation or median (interquartile range, IQR). Categorical parameters were analyzed for differences between groups using the chi-square test, and continuous parameters were analyzed for quantitative differences using the Student t-test or Wilcoxon rank-sum test. All statistical tests were performed using R software (version 4.2.2). Quality control (QC) samples were analyzed using principal component analysis (PCA) to test the repeatability and stability of the method. Orthogonal partial least squares discriminant analysis (OPLS-DA) was used to calculate the differences in metabolites between the HCC and HBVLD groups. The variable importance (VIP) value of each metabolite was calculated using OPLS-DA. The Student t-test was used to compare the metabolite levels between the two groups, where a VIP ≥ 1.0 and a p-value < 0.05 were considered to have changed significantly. Potential metabolite biomarkers were used for further pathway enrichment analysis in MetaboAnalyst 5.0 (https: / / genap.metaboanalyst.ca / MetaboAnalyst / ).

[0033] Missing data were imputed 20 times using multiple imputation with chained equations (MICE). The DIABLO method in the R package “mixOmics” was applied to integrate the 20 imputed complete data sets in a for loop, including clinical data, CpGs data, and metabolome analysis. The variables that appeared most frequently in the 20 groups of results were selected as potential biomarkers. The 20 imputed complete data sets were stacked together to create a data set for model construction. The HCC model was weighted merged by using the R package “StackImpute”. Internal validation was performed using a 10-fold cross-validation method. Kaplan-Meier curves and log-rank tests were used to detect the cumulative incidence of HCC in the HBVLD group, and p < 0.05 was considered statistically significant. The R packages “ggplot2” and “ggexcelle” were used for Volcano plots, and “riskRegression” was used for time-dependent AUC curves.

[0034] 2. Results

[0035] 1. Clinical data analysis

[0036] The inventors used strict inclusion and exclusion criteria to conduct a cross-sectional analysis of two groups of HCC and HBVLD patients, including 200 HCC patients and 200 HBVLD patients. Their characteristics and clinical characteristics are shown in Table 1. Compared with HBVLD patients, HCC patients were older, had higher levels of γ-glutamyl transpeptidase (γ-GT), alkaline phosphatase (ALP), AFP, and neutrophil count, but lower levels of lymphocyte count. The methylation data of the four sites are shown in Table 1. Figure 2 .

[0037] Table 1. Clinical characteristics of enrolled patients

[0038]

[0039]

[0040] 2. Overall metabolomic profiles and annotated metabolites of 400 plasma samples

[0041] Non-targeted metabolomics analysis was performed on 400 plasma samples using high-performance LC-MS to compare the correlation between HBVLD and HCC metabolic changes. 41 quality control (QC) samples were performed to confirm whether the systematic error of the experiment was within the controllable range. Pearson correlation was used to analyze the QC data to obtain stable and accurate metabolomics results. The Pearson correlation of plasma negative electrospray ionization (ESI-) and positive ESI (ESI+) QC samples was high ( Figure 3 ). Meanwhile, in the principal component analysis (PCA) score 3D plots of ESI- and ESI+ models ( Figure 4 ), and the 41QC samples were closely clustered. These results demonstrate reliable data quality and good method reproducibility. A total of 1,140 annotated metabolites were identified in 400 plasma samples, including 320 (ESI-) and 820 (ESI+) small molecule metabolites.

[0042] 3. Changes in plasma metabolism in HCC patients compared with HBVLD patients

[0043] OPLS-DA was used to analyze different metabolites in ESI- Figure 4 A) and ESI+ mode ( Figure 5 D). In order to test the robustness and predictive ability of the OPLS-DA model, 200 permutation tests were further performed. The goodness of fit (R2) and the intercept of the goodness of prediction (Q2) verified the reliability of the model and the lack of overfitting ( Figure 5 B and 5E). The volcano plot shows the ESI-( Figure 5 C) and ESI+ model ( Figure 5In F), fold changes in the levels of identified plasma metabolites in HCC patients relative to HBVLD, taking into account statistically significant differences (p-values) and variable importance in prediction.

[0044] Based on the thresholds of VIP ≥ 1.0 and p < 0.05 (Student’s t-test), 125 differentially expressed metabolites (DEMs) were selected, including 36 (ESI-) and 89 (ESI+) small molecule metabolites, and their retention time m / z and group differences are shown in Table S1. The key DEMs in HCC patients were significantly increased and decreased in the ESI- group compared with the HBVLD group ( Figure 6 A) and ESI+( Figure 6 B).

[0045] 4. Reprogramming of lipid metabolism in HCC leads to changes in fatty acid esterified lipids

[0046] Pathway enrichment analysis showed that significantly downregulated ( Figure 7 A) and upregulated ( Figure 7 B) Pathway. Alterations in metabolic pathways suggest that HCC undergoes a reprogramming of lipid metabolism. Mitochondrial β of long-chain (≥14 carbon atoms) saturated fatty acids (LC SFAs) is increased, whereas mitochondrial β of short-chain (<6 carbon atoms) saturated fatty acids is decreased in HCC compared with HBVLD patients. In fact, clinically, the levels of non-esterified fatty acids (NEFA, or free fatty acids) with even-numbered chains of 16-24 carbon atoms and 0-6 double bonds are significantly decreased in HCC ( Figure 7 C). These long-chain and very long-chain fats are cut into a composition of lipids esterified with fatty acids, such as fatty acid esters of cholesterol, sphingolipids, glycerol, and glycerophospholipids. Finally, the inventors identified 152 lipids esterified with fatty acids, including two fatty acid esterified cholesterol esters (CE), six ceramides, six diacylglycerols (DG), seven phosphatidic acids (PA) and lysophosphatidic acids (LysoPA), 83 phosphatidylcholines (PC) and LysoPC, 7 phosphatidylserines (PS) and 13 sphingomyelins (SM). Overall, 97 of the 152 lipids esterified with fatty acids (63.8%) were downregulated, while 15 of the 17 DEMs (88.23%) were downregulated ( Figure 7 D and 7E).

[0047] 5. Significant changes in PC in HCC and their correlation with circulating lipids

[0048] Among them, PC and LysoPC are the most abundant and most significantly changed dominant subclasses. Among the 1140 annotated metabolites, there are 83 PC and LysoPC, of which 57 decreased and 26 increased. At the same time, among the total 125 DEMs, 10 PCs decreased in expression and 1 increased ( Figure 8 ).

[0049] The inventors speculate that excessive β-oxidation of LC SFAs and VLC SFAs, which produces more energy and leads to a significant reduction in fatty acids, leads to a decrease in PC production, especially fatty acyl chains of 16 to 24 carbon atoms in length. This hypothesis is further supported by our data showing a significant correlation between NEFA and the nine differentially expressed downregulated PCs ( Figure 9 A), and a recent study showed that all PCs were significantly downregulated in HCC plasma, regardless of their unsaturation degree.

[0050] Meanwhile, most of the 11 differentially expressed PCs were positively correlated with total bile acid ( Figure 9 B) and high-density lipoprotein ( Figure 9 C), and low density ( Figure 9 D), lipoprotein, triglyceride ( Figure 9 E), cholinesterase ( Figure 9 F) showed a negative correlation, although these circulating lipids in the clinical test data did not show significant differences in HCC compared with HBVLD. This result suggests that most of the 11 differentially expressed PCs can reflect the specific effects of remodeled lipid metabolism on circulating lipids in HCC and serve as potential plasma predictive biomarkers in HCC.

[0051] 6. Comprehensive analysis of clinical data, CpGs data and metabolomics for selecting candidate biomarkers. Progress of comprehensive analysis: Figure 1As shown. Missing values in the original dataset were imputed 20 times. The mixOmics:DIABLO method was nested in a for loop to analyze the 20 imputed complete data. The program identified 20 groups of outcomes, including clinical, methylation, and metabolite characteristics that distinguished HCC from HBVLD. The most common variables in the 20 groups of outcomes were used as potential biomarkers. Seven clinical variables were selected 20 times; a total of four CpGs were screened 20 times; and 19 metabolites, including ESI- and ESI+ models, were selected 20 times. Referring to our previous selection of 17 clinical factors and 34 CpGs as outcomes of HCC with HBVLD, this study selected three clinical factors, including age, sex, and AFP levels, and three CpGs, including cg05166871, cg14171514, and cg18772205, as candidate biomarkers. Among the 19 selected metabolites, 5 metabolites, including laudanin, melibiose, zabotine, PC(O-16:0 / 20:3(8Z,11Z,14Z)), and PC(16:1(9Z) / P-18:0), met the threshold of VIP ≥ 1.0, with P < 0.05. After eliminating the possibility of being drug or traditional Chinese medicine metabolites, PC(O-16:0 / 20:3(8Z,11Z,14Z))(HMDB0039527) and PC(16:1(9Z) / P-18:0)(HMDB 0008028) were finally identified as candidate metabolic biomarkers.

[0052] 7. Development and internal validation of HCC models based on comprehensive analysis

[0053] The model was constructed using the R package "StackImpute" with weights. The parameters of the HCC model are shown in Table 2. The model showed an area under the receiver operating characteristic (AUROC) of 0.82 (95% confidence interval (CI) = 0.81-0.83), a cutoff value of 0.63, a sensitivity of 0.67, a specificity of 0.82, a negative predictive value of 0.71, a positive predictive value of 0.79, and an accuracy of 0.74 ( Figure 10 A). The calibration curve showed good agreement between predictions and observations, with a brier score of 0.24 ( Figure 10 B). There is also some sign of overestimation (calibration intercept = -2.87) but no overfitting (calibration slope = 5.79). This model has good diagnostic performance.

[0054] Table 2 Coefficients and variances of logistic regression of HCC model

[0055]

[0056] Internal validation was performed using 10-fold cross validation to evaluate the specificity, sensitivity, accuracy, C-statistic, and brier score in Table 3 , showing the stability of the HCC model.

[0057] Table 3 Ten-fold cross validation of HCC model classification performance

[0058]

[0059]

[0060] 8. Predictive Performance of HCC Model

[0061] A total of 133 patients (including 62 chronic hepatitis B and 71 LC) in the HBVLD group (n = 200) had complete clinical data recorded and were followed up for a mean of 13.43 months (IQR, 11.23-18.53 months), of whom 7 patients developed HCC. A score was calculated for each HBVLD patient according to the HCC model and the low-risk group (score < 0.45, including 40 chronic hepatitis B and 45 LC) and high-risk group (score ≥ 0.45, including 22 chronic hepatitis B and 26 LC). By the last follow-up, one chronic hepatitis B patient in the low-risk group and six LC patients in the high-risk group had developed HCC. The Harrell C index for predicting HCC using the HCC model was 0.70 (95% CI 0.50-0.89). The time-dependent ROC curves were shown, and the AUROC values of the model for predicting HCC at 6, 12, and 18 months were 0.65, 0.61, and 0.89, respectively ( Figure 10 C).

[0062] The Kaplan-Meier method was used to calculate the cumulative incidence of HCC in the low-risk group and the high-risk group. The cumulative incidence of HCC was 1.19% (95% confidence interval 0-3.48%) in the low-risk group and 21.40% (1.30%-37.40%) in the high-risk group, respectively. The cumulative incidence of HCC in the high-risk group was 10.2% (95% CI = 1.00-19.20%) at 12 months and 21.40% (1.30%-37.40%) at 16.33 months, which was significantly higher than that in the low-risk group, where the cumulative incidence of HCC was 1.19% (95% CI = 0-3.48%) at both 12 and 16.33 months (log-rank test, p = 0.0039) ( Figure 10 D). Therefore, this model may help identify HBVLD patients at risk for HCC who should undergo early diagnosis and intervention screening.

[0063] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the patent for this invention shall be determined by the appended claims.

Claims

1. A marker for diagnosing liver cancer, characterized in that: The markers are phosphatidylcholine and CpG; the phosphatidylcholine is 1-hexadecyl-2-dihydroglycyl-enoyl-sn-glycero-3-phosphocholine and 1-palmitoleoyl-2-(1-enyl-stearoyl)-sn-glycero-3-phosphocholine; the CpG includes cg05166871, cg14171514 and cg18772205; Compared with patients with hepatitis B-related liver disease, the phosphatidylcholine content in the plasma of liver cancer patients is decreased; the differential expression of phosphatidylcholine in liver cancer patients is positively correlated with total bile acid; the differential expression of phosphatidylcholine in liver cancer patients is negatively correlated with high-density lipoprotein, low-density lipoprotein, triglycerides and cholinesterase; the hepatitis B-related liver disease patients are patients with chronic hepatitis B and patients with cirrhosis.

2. Use of the marker for liver cancer diagnosis according to claim 1 in the preparation of a reagent for detecting liver cancer.

Images