The current study was approved by the Ethics Committee of the Department of Hepatobiliary Surgery at Sir Run Run Shaw Hospital and Zhejiang Hospital. All patients with HCC provided written informed consent. A total of 160 HCC samples and paired ANT samples (3 cm from the tumor) were collected from patients who underwent surgical liver tumor resection between July 2008 and February 2014. The samples were immediately frozen in liquid nitrogen and then stored at −80°C for DNA/RNA extraction. Pathological diagnosis was based on the morphological and immunohistochemical criteria provided by the World Health Organization (23). The tumor stages were classified according to the AJCC tumor-node-metastasis staging system (24).

Of the 160 patients with HCC, 111 were treated at Zhejiang Cancer Hospital and 49 at Sir Run Run Shaw Hospital. Their ages ranged between 31 and 76 years (mean age ± SD, 52.6±9.9 years), and the male-to-female ratio was 131:29. A total of 136 patients were positive for hepatitis B (HBV) and 53 patients were ≥TII according to the tumor-node-metastasis staging system. Detailed clinical information is summarized in Table I.

Total DNA was extracted from 25 mg tissue (both ANTs and tumor tissues) using a QIAamp^® DNA mini kit (Qiagen GmbH), according to the manufacturer's protocol. The DNA concentration was determined using a NanoDrop 2000 (Thermo Fisher Scientific, Inc.). Subsequently, a total of 500 ng DNA from each sample was modified by sodium bisulfite using the EpiTect Fast DNA Bisulfite kit (Qiagen GmbH) according to the manufacturer's protocol.

CSF1R sequences were obtained from the human reference genome (GRch37/hg19; http://genome.ucsc.edu/) and were utilized to design the methylation genotyping primers using the online EpiDesigner software (www.epidesigner.com). For amplification of CSF1R from the DNA extracted from ANTs and tumor tissues, the following PCR primer pair was used: Forward, 5′AGGAAGAGAGTTTAGAGAGAGTAAGGGAGGGGTTA-3′ and reverse, 5′-CAGTAATACGACTCACTATAGGGAGAAGGCTTCATAATCAAACCCCAAATAAAAAA-3′. PCR was performed using the GeneAmp 9700 system (Applied Biosystems; Thermo Fisher Scientific, Inc.) in a 10-µl reaction containing 2 µl bisulfite-converted DNA (~10 ng/µl), 0.08 µl PCR enzyme (5 U/µl), 2 µl of each primer (1 µM), 1 µl PCR buffer (10×; Sequenom) and 0.08 µl deoxyribonucleotide triphosphates mix (25 mM each). The amplification process began with an initial 4-min denaturation at 94°C, followed by 45 cycles of 20 sec at 94°C, 30 sec at 56°C and 1 min at 72°C, and finally an extension step at 72°C for 3 min. Subsequently, the PCR products were treated with three standard procedures of the MassARRAY EpiTYPER (Sequenom), according to the manufacturer's protocol: Shrimp Alkaline Phosphatase cleanup, T cleavage and clean resin. Finally, the treated DNA was transferred to a MassARRAY Analyzer 4 (Sequenom) to analyze CSF1R promoter methylation, according to the manufacturer's protocol.

An in-house RNA sequencing (RNA-seq) dataset, which included data from tumor and ANTs of 11 patients with HCC, was established as previously described and was used to investigate the alteration of CSF1R expression between HCC tissues and ANTs (25). Additionally, the methylation and expression levels of CSF1R in 50 paired samples of patients with HCC were downloaded from The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov/) for validation. A total of 17 CpG probes in the CSF1R gene included in the Illumina HumanMethylation450 array from the TCGA database were analyzed, and 14 probes were successfully genotyped, but the methylation data of the remaining 3 probes were not available.

16 pairs (10% of patients with HCC) of ANTs and tumor tissues were randomly selected from the 160 patients with HCC to examine CSF1R protein expression via staining. Tissues used for immunohistochemistry were fixed in 10% neutral formalin at room temperature, embedded in paraffin and cut into 3-µm-thick sections. For staining, the sections were deparaffinized in xylene and rehydrated in a descending ethanol series (100, 95, 90, 80 and 70%, sequentially), and washed in water at room temperature. Antigen retrieval was performed using 0.01 M citrate buffer (pH 6.0) at high temperature in a pressure cooker for 5 min. Subsequently, the sections were treated with 3% hydrogen peroxide (diluted with methanol) for 10 min at 20°C and then incubated with 10% bovine serum albumin (cat. no. A8010; Beijing Solarbio Science & Technology Co., Ltd.) at room temperature for 30 min to block non-specific antibody binding. The slides were further incubated overnight with CSF1R antibody (1:100; cat no. ab52864; Abcam) at 4°C and then incubated with biotinylated secondary antibodies (ready to use; cat. no. GK600710/100; Gene Tech Biotechnology Co., Ltd.) for 30 min at room temperature. After light counterstaining with hematoxylin (cat. no. MB9897; Dailan Meilun Biology Technology Co., Ltd.) for 2 min at room temperature, the slides were dehydrated in an ascending ethanol series (70, 80, 90, 95 and 100%, sequentially), mounted with a coverslip and observed under a light microscope (magnification, ×100 and ×400; Nikon Eclipse 80i; Nikon Corporation).

Immunostaining was scored according to the German immunoreactive score (26). This 13-point method is used to determine the percentage of positive cells by assigning them 0–4 points: 0, no positive cells; 1, <10% positive cells; 2, 10–50% positive cells; 3, 50–80% positive cells; and 4, >80% positive cells. Staining intensity was graded as follows: 0, negative; 1, weak; 2, moderate; and 3, strong. The final score was calculated as the multiplication of these two indicators and ranged between 0 and 12. The bioinformatic analysis software TBtools (v0.67361) was used to generate the heat-maps (27).

SPSS software (version 25.0; IBM Corp.) was utilized in the present study for statistical analysis. The continuous type of clinical characteristics and methylation levels of each CpG sites were presented as the mean ± SD, and the categorical type of clinical characteristics were presented as number and percentage. Statistical differences in CSF1R methylation, RNA expression and HCC protein levels between tumor and paired ANT samples were analyzed using a paired t-test. Correlation between methylation and expression was analyzed by Pearson's correlation coefficient analysis. Step-wise linear regression was performed to investigate the association between clinicopathological characteristics and methylation levels in ANTs. Unpaired t-test and one-way ANOVA with the Scheffe post hoc test were used to evaluate methylation difference between binary variables (such as diabetes status, number of primary tumors and tumor stages) or multiple group variables (tumor size), respectively. Receiver operating characteristic (ROC) curves and the area under the curve (AUC) were utilized to evaluate if CSF1R methylation can be used as a predictor of biomarkers. P<0.05 was considered to indicate a statistically significant difference.

Clinicopathological characteristics of 160 patients with HCC are detailed in Table I. Of those, 131 were male, 55 had a drinking habit (>50 g/day), 20 had diabetes and 37 had hyperlipidemia. Additionally, there were 104 patients with α fetoprotein (AFP) scores of >20 and 136 had HBV. A total of 142 patients exhibited cirrhosis.

The CSF1R gene is located on chr5:149432854-149492935, and the target amplicon for methylation analysis was located on chr5:149492491-149492958. Therefore, the 468 bp amplicon ranged from −23 to 445 bp of the CSF1R gene, and only included the transcription start sites. A total of 9 CpG sites (sequentially named CpG1-9) were included in this amplicon region, but 2 sites (CpG1 and CpG7) failed to be detected due to limitations of the MassARRAY EpiTYPER technology (Fig. 1, Table SI). The methylation levels of 7 successfully genotyped CpG sites (CpG2, 3, 4, 5, 6, 8 and 9) demonstrated significant correlation with each other in both HCC and ANTs (Tables SII and SIII).

Subsequently, paired t-tests were used to analyze the methylation levels of each CpG site in HCC and normal tissues. The results demonstrated that CSF1R methylation levels in all 7 successfully analyzed CpG sites were significantly decreased in HCC compared with their paired ANT samples (Table II; Fig. 2A) and that the mean methylation difference between HCC and ANTs ranged between 4.9 and 11.0% in the 7 CpG sites (Table II). The mean methylation level in the CSF1R promoter was 57.3±14.4% in HCC tissues and 65.3±7.5% in ANTs, respectively (P<0.0001; Table II). Additionally, the present results were supported by data from TCGA database. The methylation data of CSF1R in 50 HCC tissues and their paired ANTs from TCGA (https://portal.gdc.cancer.gov/) database were assessed, and 12 CpG probes of the Illumina HumanMethylation450 array in the CSF1R gene were hypomethylated in HCC tissues (Fig. 2B).

ROC curve analysis was conducted to compare CSF1R methylation levels in ANTs and HCC tissues and to verify whether the methylation level of the CSF1R promoter region could be used as a biomarker for HCC diagnosis and treatment (Fig. 3). All AUC values were >0.5 and the mean value of all methylation sites was 0.713.

To analyze the alteration of CSF1R expression in HCC, an in-house RNA-seq dataset containing 11 HCC tissues and their paired ANTs was used. In this dataset, a significantly higher CSF1R expression was detected in ANTs compared with HCC tissues [normal tissues vs. cancer tissues; reads per kilobase of exon model per million mapped reads (FPKM), 23.94 vs. 11.43; P=0.0026; Fig. 4A]. In addition, CSF1R expression was then examined from TCGA dataset, from which 50 paired HCC RNA-seq data were downloaded and analyzed using paired t-tests. The results from TCGA database revealed that CSF1R expression was significantly decreased in HCC tissues compared with that in ANTs (ANTs vs. HCC tissues; FPKM, 10.21 vs. 9.19; P<0.0001; Fig. 4B).

Furthermore, immunohistochemistry was performed to detect CSF1R protein expression in 16 paired HCC and ANTs, and to verify expression differences at the protein level. The results indicated that CSF1R protein levels in ANTs were markedly higher compared with HCC tissues (Fig. 4C), which was consistent with the data that was sequenced and obtained from TCGA database. Using the 13-point method, it was also suggested that CSF1R protein expression in ANTs was significantly higher compared with that in tumor tissues (P=0.0025; Fig. 4D and E).

CSF1R expression and methylation data were downloaded from TCGA datasets to identify the correlation between the expression and methylation status. A significantly negative correlation was identified between methylation and expression of CSF1R in ANTs (Fig. 5A and B), particularly in sites cg12862231 and cg16492211 (P<0.001). Sites cg07260017 and cg01875467, which were adjacent to the selected CpG islands according to TCGA data, were also analyzed and revealed to follow the same trend (Fig. 5C and D). However, when combining the ANT and paired HCC data together, there was no correlation between CSF1R methylation and expression (Fig. SI).

To determine the potential effects of CSF1R methylation and expression in HCC, the correlation between the methylation status of CSF1R and comprehensive clinicopathological features was analyzed. According to the high correlation among the methylation levels of the CpG sites (Tables SII and SIII), the mean methylation levels of the 7 CpG sites were used to evaluate their association with clinicopathological characteristics. The results revealed that diabetes status was associated with methylation levels of CSF1R. The mean value of the methylation site in the ANTs of patients with diabetes was significantly different from that in tissues from patients without diabetes. Patients with HCC that concurrently presented with diabetes exhibited low ANT methylation (66.2% in 140 non-diabetic vs. 61.3% in 20 diabetic patients; P=0.0062; Fig. 6A and Table I).

Furthermore, the results demonstrated that patients with HCC who presented with multiple tumors exhibited an average methylation level that was significantly decreased compared with patients with single tumors (65.9 vs. 61.4%; P=0.0332; Fig. 6B and Table I). Additionally, there was a significant difference between tumor diameters of >10 cm and 5–10 cm (P=0.0494; Fig. 6C and Table I) as demonstrated by ANOVA analysis followed by the Scheffe post hoc test. However, a linear decrease with an increase in tumor diameter was not reported. This could be due to the current sample size being too small. Concurrently, the results revealed that the average ANT methylation levels of patients with high-stage HCC also exhibited a significant decrease (66.5 vs. 63.6%; P=0.0237; Fig. 6D and Table I).