A total of 177 frozen tumour specimens and paired paratumor noncancerous tissues were collected from patients with HCC who underwent surgery at Nagoya University Hospital (Nagoya City, Japan) between January 1998 and April 2014. All fresh tissues were immediately frozen in liquid nitrogen and stored at −80°C until use. Patient characteristics are summarized in Table I. After surgery, all patients were monitored via blood examinations, ultrasonography, and computed tomography once every six months. Angiography was performed for further information whenever recurrence was suspected. The median follow-up duration of all patients was 48.8 months (range, 0.3 to 191 months). This study and all procedures were approved by the Institutional Review Board at Nagoya University (Nagoya City, Japan), and all patients provided written informed consent. All clinical investigations were conducted in accordance with the principles of the Declaration of Helsinki.

Total RNA was extracted from tissue samples using a Qiagen miRNeasy mini-kit (Qiagen, Hilden, Germany). We used DNAse to avoid contamination, and RNA quality was analysed by a NanoDrop (Thermo Scientific Fisher, Waltham, MA, USA). Total RNA was converted to complementary DNA by reverse transcription with M-MLV Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA). This total cDNA was used as a template for the next step of quantitative PCR (qPCR). qPCR was performed using SYBR Premix Ex Taq II (Takara Clontech, Kyoto, Japan) under the following conditions: Denaturation at 95°C for 10 sec and 40 cycles of denaturation at 95°C for 5 sec and annealing/extension at 60°C for 30 sec. The SYBR Green signal was detected in real-time using a StepOne Plus Real-Time PCR System (Life Technologies, Carlsbad, CA, USA). The relative quantification method was used, and the expression level of each gene was normalized to the expression level of the control gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH) for each sample. The relative gene expression levels were determined using the comparative threshold cycle (2^−ΔCT) method.

The PCR primers used in the current study were specific for the 78-base-pair fragment of YTHDF1 (sense, 5′-TCCATCTTCGACGACTTTGCT-3′; antisense, 5′-TCGACTCTGCCGTTCCTTG-3′) and for the 50-base-pair fragment of YTHDF2 (sense, 5′-GAGGATCTGAGAGCCATGTCG-3′; antisense, 5′-ATTTTGTACTGCTCCAAGAGGC-3′). GAPDH primers (sense, 5′-GAGTCCACTGGCGTCTTCAC-3′; antisense, 5′-GTTCACACCCATGACGAACA-3′) were used to quantify the expression in each sample as an internal control. The primers were designed as intron spanning. All qPCR experiments were performed in duplicate, including the template-omitted negative controls.

Normalized TCGA RNA-sequencing data of HCC were downloaded from the Broad GDAC Firehose (http://gdac.broadinstitute.org/, accessed on January 1st, 2020). This dataset consists of 50 noncancerous cases and 360 HCC cases, including seven HCC cases mixed with hepatocholangiocarcinoma and two cases with fibrolamellar carcinoma. Of the 360 cases, there were 266 cases with recurrence-free survival (RFS) information and 336 cases with overall survival (OS) information.

Continuous variables are expressed as the median (range), and the expression of each target gene was compared by a Wilcoxon signed-rank test and paired t-test. Categorical variables were compared using χ² or Fisher's exact tests, as appropriate. The OS and RFS rates at each point of the follow-up time were estimated using the Kaplan-Meier method and compared using a log-rank test. A Cox proportional hazard regression model was used to perform univariate analysis and multivariate analysis for OS and RFS. In the multivariate analysis, variables that showed statistical significance in the univariate Cox proportional hazard regression were included. All statistical analyses were performed using R version 3.5.3 (http//www.r-project.org/), and P<0.05 obtained using two-tailed tests was considered to indicate statistical significance.

First, expression analyses of m⁶A readers were conducted with our surgically resected specimens. The expression levels of YTHDF1 and YTHDF2 were measured by qPCR. The expression of YTHDF1 and YTHDF2 was not significantly different between tumour tissues and noncancerous tissues (P=0.93 and P=0.7, respectively, Fig. 1A). Fig. 1B shows individual changes in YTHDF1 and YTHDF2 expression in paired analysis. Based on the results obtained by qPCR, 177 HCC cases were divided into two groups according to YTHDF1 and YTHDF2 expression in tumour tissues. We selected the cut-off values that showed the best statistical difference. Clinical features of the groups stratified by YTHDF1 and YTHDF2 expression are shown in Table II. In HCC tissues, low YTHDF1 expression was associated with a lack of capsule (65% vs. 81%, P=0.02), whereas low YTHDF2 expression was associated with septal formation (73% vs. 53%, P=0.02).

Next, the effects of the expression levels on RFS and OS were evaluated. In HCC tissues, both high YTHDF1 expression and high YTHDF2 expression were significantly correlated with shorter RFS (YTHDF1: MST=34.0 vs. 19.0 months, P=0.014; YTHDF2: MST=30.1 vs. 12.9 months, P=0.0032, Fig. 2A), whereas YTHDF1 expression and YTHDF2 expression were not correlated with OS (YTHDF1: MST=99.4 vs. 70.2 months, P=0.74; YTHDF2: MST=98.4 vs. 64.1 months, P=0.28, Fig. 2B).

YTHDF1 and YTHDF2 expression levels and their correlation with HCC prognosis in a publicly available dataset

We analysed the expression levels of YTHDF1 and YTHDF2 in HCC and noncancerous tissues using a TCGA RNA-sequence dataset. This analysis revealed that the expression of YTHDF1 was significantly higher in HCC tumour tissues and that YTHDF2 expression was significantly lower in HCC tumour tissues than in noncancerous tissues (YTHDF1, P=0.0011; YTHDF2, P=0.025, Fig. 3A and B). We then confirmed the prognostic impact of YTHDF1 and YTHDF2 expression in resected HCC patients using the same TCGA dataset. HCC cases were divided into two groups according to the YTHDF1 and YTHDF2 expression in HCC tissues in the normalized RNA-sequencing data. We also selected the cut-off values that showed the best statistical difference. This analysis revealed that the patients with high YTHDF1 expression had significantly worse RFS (MST=754 vs. 489 days, P=0.006, Fig. 3C), and the patients with high YTHDF2 expression tended to have worse RFS (MST=636 vs. 315 days, P=0.06, Fig. 3D). In addition, the patients with high YTHDF1 expression had significantly worse OS (MST=2456 vs. 1,372 days, P=0.015, Fig. 3E), and the patients with high YTHDF2 expression had significantly worse OS (MST=2,456 vs. 724 days, P=0.0002, Fig. 3F).

Since the survival curves showed that YTHDF1 and YTHDF2 expression levels in HCC tissues were correlated with RFS, we performed Cox proportional hazards analyses to further investigate the prognostic value of YTHDF1 and YTHDF2 expression. The multivariate analysis identified serosal invasion (hazard ratio (HR): 2.39, 95% confidence interval (95% CI): 1.30–4.42, P=0.005), portal vein or hepatic vein invasion (HR, 2.82, 95% CI: 1.26–6.28, P=0.01) and YTHDF2 expression in HCC tissues (HR, 1.85, 95% CI: 1.09–3.15, P=0.02) as significant independent factors for RFS (Table III) and AFP (HR, 1.79, 95% CI: 1.10–2.92, P=0.02), serosal invasion (HR, 1.99, 95% CI: 1.17–3.34, P=0.01), and portal vein or hepatic vein invasion (HR, 3.02, 95% CI: 1.38–6.61, P=0.006) as significant independent factors for OS (Table IV). Consequently, high expression of YTHDF2 in HCC tissues was significantly associated with recurrence after HCC surgery.