Molecular Alterations of Circulating Cell-Free DNA in the Pathological Progression of Hepatocellular Carcinoma

Background Hepatocellular carcinoma (HCC) is one of the most malignant cancers. Early diagnosis of HCC is important to reduce the mortality rate. The aim of this study is to explore the plasma cell-free DNA (cfDNA) mutation profile in the pathological progression of HCC and to investigate the significance of plasma cfDNA mutations in the early diagnosis of HCC. Methods Thirty-seven patients with chronic hepatitis B (CHB), eight with liver cirrhosis (LC), and eleven with HCC were enrolled in this cohort. Plasma cfDNA and white blood cell DNA were isolated, and plasma cfDNA mutation profiles were detected using a targeted gene panel. Results The sequencing results of plasma cfDNA showed that HCC-related gene mutations were present in patients with CHB and LC. The mutation burden of HCC-related genes increased from CHB and LC to HCC. In patients with HCC, the average mutation burden of NRAS (10.1%), TP53 (7.4%), PTEN (4.2%), and APOB (2.6%) was the highest. The average mutation burden of PTEN, APOB, FRAS1, KDM6A, DDR2, TTK, NRAS, TP53, PTPRB, MPL, FCRL1, HN1, and SFN gradually increased from CHB and LC to HCC. The mutation burden of 18 HCC-related genes had an area under the receiver operating characteristics of 0.92 for the diagnosis of HCC. Conclusions The mutation burden of HCC-related genes increased from CHB and LC to HCC. An optimal combination of cfDNA mutations in the gene panel for diagnosing HCC in patients with CHB and LC was selected. Our study indicates that somatic mutations in plasma cfDNA may serve as potential biomarkers for early HCC diagnosis.


Introduction
Hepatocellular carcinoma (HCC) is the sixth most common cancer worldwide and the third most deadly cancer [1]. Viral infection, especially hepatitis B virus, increases the risk of liver cirrhosis (LC) and HCC [2][3][4], with approximately 60-65% of HCC cases caused by hepatitis B virus infection [5]. Reports show that 1.3% of patients with LC develop HCC each year [6], underscoring the necessity of surveillance for cirrhotic patients. Generally, the high mortality rate of HCC is due to the low rate of early diagnosis, lack of clinically effective systemic treatment, high intratumoral heterogeneity, and high relapse rates.
Liver biopsy is the gold standard for the diagnosis of liver diseases despite its invasiveness. However, it is unlikely that the biopsy can reflect the entire liver condition owing to the tissue heterogeneity. Current methods such as alpha-ferroprotein blood tests and imaging modalities lack sensitivity and specificity, which limits their use in early diagnosis. e use of liquid biopsies has been developed and validated over the past few years, making the diagnosis of HCC and liver fibrosis possible [7][8][9].
Circulating cell-free DNA (cfDNA) detected in the blood of cancer patients carries tumor genomic alterations and has been widely investigated in various cancers as a biomarker for diagnosing, predicting, and monitoring response to therapy and monitoring tumor burden and relapse due to its comprehensive capture of somatic mutations [10][11][12][13]. Several studies have shown that the level of cfDNA is markedly elevated in patients with HCC and can be detected early in tumorigenesis [14][15][16]. erefore, somatic mutations in cfDNA may provide necessary information to distinguish between benign and malignant tumors. To date, there have been only a few studies that utilized cfDNA to track the development of chronic hepatitis B (CHB), LC, and HCC [17][18][19]; hence, further research is required to better understand the genetic mutational progression in plasma cfDNA of patients with CHB, LC, and HCC. e present study was conducted to profile the plasma cfDNA mutations in the pathological progression from CHB and LC to HCC and to investigate the significance of plasma cfDNA mutations in the early screening for HCC.

Library Construction and cfDNA Sequencing.
Genomic DNA extracted from white blood cells was fragmented into 200-bp DNA pieces with a M220 Focused-Ultrasonicator (Covaris, Woburn, MA) according to the manufacturer's protocol. Both the cfDNA library and the genomic DNA library were constructed using a KAPA Hyper Preparation Kit (Kapa Biosystems, Wilmington, MA). A 2100 bioanalyzer with a DNA 1000 kit (Agilent, Santa Clara, CA) was used to determine the fragment length. DNA was hybridized to a designed 543-gene panel (Genecast, Wuxi, China) that included major tumor-related genes, covering 1.7 Mb of the genome. e sequencing libraries were quantified using a Qubit dsDNA HS Assay Kit (Cat#Q32854, ermo Fisher Scientific, Waltham, MA, USA). After library construction, the collected samples were sequenced on an Illumina HiSeq X platform (paired end, 150 bp).

Bioinformatics Pipeline.
Bioinformatics pipeline analysis was performed as previously described [20]. Burrows-Wheeler Aligner (version 0.7.12) [21] was used to align the hg19 reference genome with the sequences. SAMtools was used to process the resulting alignments. e bam files were processed to distinguish somatic SNP and indel mutations using VarScan v2.4.2 [22]. ANNOVAR was used to annotate variants. Blood cell samples were used as negative controls. e copy number variation was analyzed using the CNVkit [23].

Statistical
Analysis. Statistical analyses were conducted using SPSS software (version 19.0; IBM, Armonk, NY, USA). Heatmap and clustering analyses were conducted using Python (v3.6). Experimental data were reported as the mean ± standard error of the mean (SEM). Student's t-test and nonparametric tests were used for normally distributed data and nonnormally distributed data, respectively. Statistical significance was set at P < 0.05.

Mutation Profile of cfDNA in Patients with CHB, LC, and HCC.
To investigate molecular alterations in patients with CHB, LC, and HCC, we collected blood samples from 37 patients with CHB, 8 patients with LC, and 11 patients with HCC (clinical characteristics are shown in Table 1). Plasma cfDNA and white blood cell DNA were sequenced using a targeted next-generation sequencing gene panel. e quality control is shown in Supplementary Figure 1

Mutation Burden of HCC-Related Genes Increased from CHB and LC to HCC.
To explore the gene mutation variations in the pathological progression of HCC, we analyzed the numbers and mutation burden of mutated genes in the plasma cfDNA of patients with CHB, LC, and HCC. e number of mutated genes increased from CHB and LC to HCC. e number of mutated genes in patients with HCC was significantly higher than that in patients with CHB ( Figure 2(a), P � 0.0071). ere were no significant differences in the number of mutated genes between HCC and LC cases (Figure 2(a), P � 0.1437). e mutation burden of mutated genes increased from CHB and LC to HCC. Compared with that in patients with CHB, the mutation burden was significantly increased in patients with HCC ( Figure 2(b), P � 0.0037). ere were no significant differences in the mutation burden of mutated genes between patients with HCC and those with LC ( Figure 2 Journal of Oncology

HCC Diagnosis Model Based on Gene Mutation
Burden. e finding that the mutation burden of HCC-related genes increased from CHB and LC to HCC indicated that HCCrelated gene mutations may be potential biomarkers for HCC diagnosis. Based on the gene mutation profiles among patients with CHB, LC, and HCC, we used the decision tree method to select optimal combinations of gene panels for HCC diagnosis. e optimal HCC gene panel included DST, SYNE2, APOB, KDM6A, USP9X, SPAG17, TERT, TP53, AMPH, EP300, TTK, CCND1, FCRL1, IL6ST, PTEN, SFN, STK11, and NRAS. A panel of gene mutation profiles among patients with CHB, LC, and HCC is shown in Figure 5(a). e mutation burden of panel genes in patients with HCC was significantly higher than in those with CHB and LC ( Figure 5(b), P < 0.01 and P � 0.0018, respectively). e gene mutation burden of 18 HCC-related genes had an area under the receiver operating characteristic (AUROC) of 0.92, with 100% sensitivity, 82.2% specificity, 57.9% positive predictive value, and 100% negative predictive value (Figure 6). e diagnostic power of the 18 HCC-related gene panels was superior to that of cfDNA (AUROC � 0.87) and alpha fetoprotein (AUROC � 0.69). e cutoff value for the HCC gene panel was 5%.

Discussion
HCC is one of the most malignant cancers worldwide and mainly occurs in East Asia and the sub-Saharan region [24]. CHB and LC are the two major risk factors for HCC. Patients with either CHB or LC could be symptomatic or asymptomatic, depending on the severity of their disease. As CHB or LC progresses, some of them will develop into HCC. Patients with HCC could obtain more clinical benefits from early therapy based on early diagnosis, reducing mortality by approximately 37% [25]. erefore, we enrolled patients with CHB, LC, and HCC to evaluate potential biomarkers that could be employed to screen high-risk populations early.
Clinically efficient and reliable biomarkers for the early detection of HCC are critical. Alpha fetoprotein and imaging modalities are still the most frequently used means in clinics for HCC diagnosis. Nevertheless, they are not ideal for early diagnosis of HCC due to their limited specificity and sensitivity. Additionally, they lack efficiency to detect HCC at the early onset. Meanwhile, cfDNA has shown potential as a minimally invasive and efficient biomarker for risk prediction of different cancers. is study indicated that the mutation burden of HCC-related genes increased from CHB and LC to HCC.
Previous studies have indicated that liquid biopsy can be used to detect cancer by assessing circulating plasma cfDNA mutations [13,19,26,27]. While most studies enrolled healthy individuals as controls, screening and continued surveillance of high-risk populations may be more meaningful in clinical practice. Early detection and confirmation of HCC in high-risk individuals usually involves different treatment strategies. erefore, we enrolled patients with CHB, LC, and HCC. Based on the gene mutation profiles among patients with CHB, LC, and HCC, we used the decision tree method to select the optimal combination forms of the gene panel for diagnosing HCC in patients with CHB and LC. e optimal HCC gene panel included a combination of 18 HCCrelated genes, which had an AUROC of 0.92, with 100.0% sensitivity, 82.2% specificity, and 100% negative predictive value, and was generally higher than the parameters in the 40-gene panel.
e diagnostic power of the 18  Journal of Oncology HCC-related gene panel was superior to that of cfDNA concentrations [14] and alpha fetoprotein [28]. More importantly, the positive predictive value of this HCC gene panel for diagnosis was 57.9%, which is more powerful than in previous studies [17,19].
Most patients with HCC have a history of chronic liver diseases, such as CHB and LC. Our study explored the plasma DNA mutation profile in the pathological progression of CHB and LC to HCC. e results indicated that HCC-related gene mutations existed as early as in CHB and LC. Furthermore, we found that the mutation burden of HCC-related genes increased from CHB and LC to HCC. In the progression of HCC, gene mutations continue to accumulate gradually.
ese results reveal that the plasma cfDNA mutation burden may predict disease progression in liver diseases. is finding implies that patients with CHB and LC with high molecular mutations might have a high risk of carcinogenesis.
Our study has limitations. Firstly, the sample size is relatively small. One reason for this is the difficulty to obtain a large sample size in clinics, including, but not limited to, the consent to participate in the study. Another reason for this is the purpose of the study. Our aim in this study is to explore the cfDNA mutation profile in the pathological progression of HCC and to investigate the possibility of using cfDNA in the adjunctive diagnosis of HCC. Secondly, we enrolled patients mostly at the late stage instead of early HCC such as carcinoma in situ, who are more suitable for our study. e main reason for this is the difficulty to enroll the very early patients with HCC due to the limitations of currently available diagnostic methods and the patients' late visit to clinicians. In fact, few patients with HCC are diagnosed at the early stage. We hope to solve the aforementioned limitations by enrolling more patients with CHB and LC and following their progression closely in a long time span.

Conclusions
Our study shows that the plasma cfDNA mutation burden accumulates from CHB and LC to HCC. Plasma cfDNA may contain potential biomarkers for predicting disease progression during the development of HCC. More importantly, we used the decision tree method to select optimal combinations of cfDNA mutations in the gene panel for diagnosing HCC in patients with CHB and LC. Our study broadens the knowledge of the progression of CHB and LC to HCC and deserves a prospective investigation in clinics to screen for HCC early.

Data Availability
Data are available on reasonable request.

Conflicts of Interest
Debin Sun and Longlong Gong work for Genecast Biotechnology Co., Ltd. Journal of Oncology 7