Optimized digital polymerase chain reaction enables detection of telomerase reverse transcriptase C228T mutation for prognostic assessment in hepatocellular carcinoma

Abstract

BACKGROUND

Recurrence remains the leading cause of poor prognosis in hepatocellular carcinoma (HCC), particularly among patients infected with hepatitis B virus (HBV). The telomerase reverse transcriptase (TERT) promoter is the most frequently mutated site in HBV-related HCC; however, its prognostic significance is not fully established.

AIM

To evaluate the prognostic impact of TERT promoter mutations and efficiency of digital polymerase chain reaction (dPCR).

METHODS

A total of 66 HBV-related HCC patients who underwent hepatectomy were enrolled in this study. DNA extracted from fresh tumor tissues was analyzed for TERT promoter mutations using Sanger sequencing and dPCR. The dPCR assay was optimized by adding 7-deaza-dGTP, CviQ1, and ethylenediaminetetraacetic acid to improve detection sensitivity. Concordance between methods was assessed, and nomogram survival prediction models were developed to evaluate prognostic value based on mutation status.

RESULTS

TERT promoter mutations were detected in 26/66 (39.39%) cases by Sanger sequencing and 30/66 (45.45%) by dPCR. The two methods showed high concordance (93.939%, κ = 0.876), with dPCR demonstrating 100% sensitivity and 90% specificity. Patients harboring TERT promoter mutations exhibited reduced overall survival and higher recurrence risk. Nomogram models successfully distinguished mutant from non-mutant cases for both overall survival (C-index: 0.7651) and disease-free survival (C-index: 0.6899).

CONCLUSION

TERT promoter mutation predicts poor prognosis in HBV-related HCC and serves as a biomarker for risk stratification. Optimized dPCR outperforms Sanger sequencing, and nomograms with TERT status guide precision therapy.

Key Words: Hepatocellular carcinoma; Hepatitis B virus; Mutation; Polymerase chain reaction; Hepatectomy; Nomograms; Prognosis

Core Tip: Telomerase reverse transcriptase (TERT) promoter mutation is the most frequent genetic alteration in hepatitis B virus-related hepatocellular carcinoma, yet its prognostic role remains uncertain. In this study, optimized digital polymerase chain reaction outperformed Sanger sequencing in detecting TERT promoter mutations. Patients with TERT mutations had significantly worse survival and higher recurrence risk after hepatectomy. Incorporating TERT mutation status into a nomogram model provides a practical tool for individualized prognosis prediction and may guide precision treatment in hepatitis B virus-related hepatocellular carcinoma.

INTRODUCTION

Primary liver cancer is the third leading cause of cancer-related death worldwide, with hepatocellular carcinoma (HCC) accounting for 75%-85% of cases, most often due to chronic hepatitis B virus (HBV) infection[1]. Despite hepatectomy being the main curative treatment, HBV-related HCC shows high recurrence rates, significantly worsening prognosis[2]. Thus, identifying reliable biomarkers for recurrence is crucial to improve risk stratification and guide precision therapies. Somatic mutation detection is a central application of precision medicine[3]. In HCC, recurrent mutations frequently involve tumor protein p53, and catenin beta 1, and telomerase reverse transcriptase (TERT) promoter, with the latter considered an early driver of tumorigenesis[4,5].

Telomerase, composed of TERT and telomerase RNA components, counteracts telomere shortening and is a well-recognized hallmark of carcinogenesis[5]. In HCC, telomerase activity is frequently reactivated through HBV X protein, HBV integration into the TERT promoter, or oncogene regulation[6]. Among these alterations, TERT promoter mutations occur in approximately 60% of cases, creating novel E-twenty-six transcription factor binding sites that enhance telomerase activity and drive hepatocarcinogenesis[7,8]. HBV integration at the TERT locus further promotes telomerase activation and cellular proliferation. TERT promoter mutations (26.0% vs 0%) and HBV-TERT integration (37.0% vs 3.0%) are significantly enriched in HCC compared with non-tumorous tissues[9]. Analysis of 4133 HCC samples has shown that 49.2% of HCC cases harbor TERT promoter mutations, with C228T as the predominant hotspot (93.3%)[5]. In addition, TERT modulates oncogenic signaling pathways, including nuclear factor-κB and Wnt/β-catenin, thereby contributing to tumor progression[10,11].

In recent years, nomograms have emerged as practical prognostic tools by integrating clinical and pathological variables into quantitative models, and several studies have shown they outperform conventional staging systems in predicting outcomes after liver resection[12]. Key factors such as albumin, bilirubin, microvascular invasion (MVI), tumor burden, and surgical variables have been incorporated, improving predictive accuracy and clinical utility[13]. More recently, radiomics has further advanced this field by extracting high-throughput features from computed tomography or magnetic resonance imaging to capture tumor heterogeneity[14]. Radiomics-based nomograms, especially those using multiphase computed tomography, susceptibility-weighted imaging, or T2-weighted imaging, have demonstrated high accuracy in predicting MVI, early recurrence, and in stratifying patients for adjuvant therapies[15]. However, nomogram models incorporating molecular biomarkers such as TERT mutations are lacking, limiting the integration of genomic predictors into current prognostic tools.

TERT promoter mutations are among the most common genetic alterations in HCC and hold promise as prognostic biomarkers, yet their detection has been hindered due to the guanine-cytosine (GC)-rich promoter region[16-18]. While conventional gold standard Sanger sequencing has limited sensitivity, digital polymerase chain reaction (dPCR) has recently emerged as a more reliable method for detecting TERT promoter mutations across cancers, including HCC[19]. In this study, we optimized a dPCR assay, established detection thresholds, and evaluated the prognostic value of the C228T mutation by developing nomogram-based survival prediction models in HBV-related HCC, aiming to integrate molecular and clinical predictors for precision treatment strategies.

MATERIALS AND METHODS

Patients and tissue samples

This study included 66 HBV-related HCC patients who underwent hepatectomy at the Cancer Hospital of Peking Union Medical College and Chinese Academy of Medical Sciences between January 2013 and May 2016, including 54 males and 12 females. The median diagnostic age of enrolled HBV-related HCC patients was 52 years (ranging from 46.25 to 59). We also collected fresh tumor tissue samples from the surgical specimens of these 66 patients and all the tissues were frozen immediately after surgery and stored at -80 °C. All patients had a clear postoperative pathological diagnosis by two expert pathologists and were confirmed to have HCC.

We collected patient information and biomarker test results, including cirrhosis and alcohol status, alpha fetoprotein (AFP), alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyl transferase (GGT), lactate dehydrogenase (LDH), and activated partial thromboplastin time (APTT) retrospectively. Clinical data on the number of tumors, tumor diameter, liver capsule invasion (LCI), blood vessel invasion (BVI), Barcelona Clinic Liver Cancer (BCLC) stage, and MVI were also obtained from medical records. Patients were divided into the well-differentiated group (high, n = 10), moderately differentiated group (medium, n = 42), and poorly differentiated group (low, n = 14) according to histological grade using the criteria of Edmondson and Steiner[20]. Follow-up evaluations occurred in 2018 (first follow-up), 2019 (second follow-up), June 2021 (third follow-up), and August 2022 (fourth follow-up). The maximum follow-up time was 9.3 years.

DNA extraction

An incubation solution containing 500 μL of DNA lysis solution was used to dissolve the frozen tissue samples [proteinase K 1 mg/mL, Tris-HCl (pH 8.0) 10 mmol/L, ethylenediaminetetraacetic acid (EDTA) (pH 8.0) 0.1 mol/L, sodium-dodecyl sulfate 0.5% (w/v)] at 55 °C, overnight. After digestion, DNA was extracted by phenol and chloroform, and stored at -20 °C until use. A Nanodrop spectrophotometer ND-1000 (Thermo Scientific, United States) was used to verify DNA concentration and quality. All DNA samples had an absorbance ratio equal to or above 1.80 at 260 nm and 280 nm.

Sanger sequencing for detecting TERT C228T mutations

We performed polymerase chain reaction (PCR) reactions with genomic DNA in a 50 μL reaction mixture, with 2.5 μL of each primer, 0.25 μL HotStartTaq Plus DNA Polymerase (Qiagen, Germany), 10 μL Q-solution 5 × (Qiagen, Germany), 5 μL Coralload PCR buffer 10 × (Qiagen, Germany), and 1 μL deoxyribonucleoside triphosphate 10 mmol/L (TaKaRa, Japan). In a PTC-200 Peltier Thermal Cycler (Applied Biosystems, Thermo Fisher Scientific, United States), primers for the TERT promoter region were amplified as follows: (forward) 5′-ACGAACGTGGCCAGCGGCAG-3′, and (reverse) 5′-CTGGCGTCCCTGCACCCTGG-3′. Cycling conditions were as follows: Denaturing 30 seconds at 98 °C, after 35 cycles of denaturing for 15 seconds at 98 °C, annealing for 15 seconds at 62 °C, and elongation for 30 seconds at 72 °C, the final elongation at 72 °C was performed for 2 minutes. As a result, the 474 bp amplicon was obtained[21]. Products were visualized by 1% agarose gel electrophoresis, gel-purified, and sequenced. The product was then sequenced with (forward) 5′-CAGCGCTGCCTGAAACTC-3′ and (reverse) 5′-GTCCTGCCCCTTCACCTT-3′ primers. Somatic mutations have been observed in the 124 bp hotspot of TERT promoter region.

dPCR for detecting TERT C228T mutations

Our experimental conditions were verified on positive samples with known TERT promoter mutation. We performed the droplet generation process according to the manufacturer’s protocol using the Naica Crystal PCR system (Stilla Technologies, France). The reaction mixtures used in this assay were as follows: 12.5 μL ddPCR Supermix for Probes (no deoxyuridine triphosphate) (Bio-Rad, United States), 1.25 μL 20 × primer/probe mixture [TERT C228T_113 (dHsaEXD72405942), Bio-Rad, United States], restriction enzyme 0.2 μL (New England BioLabs; CviQI; R0639S, United States), 0.2 μL 7-Deaza-dGTP (New England BioLabs; N0445S, United States), 1 μL of 25 mmol/L EDTA (Thermo Fisher Scientific, United States), 2.5 μL of 10 μmol/L fluorescein sodium salt (Apexbio, United States), DNA template, and finally ultrapure DNAse/RNAse free H₂O (Thermo Fisher Scientific/Gibco, United States) was added to a final volume of 25 μL. The cycling protocol was as follows: 1 cycle of 10 minutes at 37 °C, 1 cycle of 10 minutes at 95 °C, followed by 50 cycles of 95 °C for 30 seconds and 62 °C for 1 minutes, and a final hold at 4 °C (2 °C/seconds). The dPCR data were analyzed using Crystal Miner for the Nacia^TM System v2.3.0 software (Stilla Technologies, France), and positive and negative controls were used to determine the thresholds for each mutation assay.

To ensure the detection quantity of TERT C228T mutations, we used mutation probes in the channel to determine a false positive measure and therefore detected a threshold for the smallest mutation droplet, twenty wild-type DNA samples for TERT mutations were tested by their respective dPCR. To determine the limit of blank (LoB), negative controls were analyzed, and background signals were obtained. At least five low-level (LL) samples were prepared independently (LL1, LL2, LL3, LL4, and LL5), and at least four replicates were performed per sample. We calculated the limit of detection (LoD) by determining the concentration of mutant alleles at which LL samples should comprise representative positive samples or representative matrix samples with a maximum target concentration of one to five times the calculated LoB. The LoB of TERT C228T was detected in pure wild-type samples in 0.35 cp/μL per well. The LoD of TERT C228T in LL samples was 0.55 cp/μL/well.

Independence of the TERT mutation from clinical features

Our 66 HBV-related HCC cohort completed clinical and survival information, including, age, gender, ALT, AST, LDH, GGT, APTT, cirrhosis, alcohol, tumor number, tumor diameter, MVI, tumor differentiation, LCI, BVI, and BCLC grade. Cox regression analysis with univariate and multivariate variables was conducted to confirm that the predictions of the prognostic model were independent of clinical characteristics in HBV-related HCC.

Construction and evaluation of the nomogram

Multivariable Cox models were used to calculate survival probability estimates for overall survival (OS) and disease-free survival (DFS) nomograms. Up to 5 years after surgery, the accuracy of the nomogram models was evaluated based on Harrell’s C-index discrimination, calibration, and overall accuracy. To evaluate model performance, calibration and discrimination are commonly used methods, and they were calculated using a bootstrapping approach with 1000 resamples. Furthermore, C-index and time-dependent receiver operating characteristic analyses were also performed to compare predictive accuracies between the nomogram and prognostic variates separately. Additionally, to verify the clinical validity of the established nomogram, calibration curve analysis and decision curve analysis (DCA) were conducted.

Statistical analysis

The demographic and clinical characteristics of the patients were summarized using descriptive statistics. We performed χ² or Fisher’s exact testing for categorical variables and Wilcoxon rank sum tests for continuous variables to examine the causal relationship between clinicopathological factors and TERT mutation status (no vs yes). Based on Cohen’s kappa index, we evaluated the degree of agreement between dPCR and Sanger sequencing. Cohen proposed a categorical interpretation of the kappa index. Weak agreement ≤ 0, slight agreement is 0-0.2, moderate agreement is 0.21-0.40, average agreement is 0.41-0.60, and basic agreement is 0.61-0.80, while 0.81-1.00 was almost perfect. This index is considered too lenient for health-related research. Calculations for sensitivity, specificity, and accuracy of the TERT promoter mutations were performed using the Clopper-Pearson method. We performed Kaplan-Meier plots and Cox proportional hazards models using the R packages (v 3.6.3) “survminer” and “survival”. The OS was defined as the period from surgery until death due to any cause, and the DFS was the period after surgery until any recurrence or death. The nomograms for predicting the OS and DFS were generated using the R “rms” package. Other two-sided statistical tests were conducted, and P < 0.05 was considered statistically significant. Statistical analyses were performed using the packages “survminer”, “rms”, “rmda”, and “timeROC” in R version 3.6.3.

RESULTS

Assay optimization

It was challenging to amplify the TERT promoter due to its high GC content (more than 80%) and the way it forms secondary structures, such as hairpin structures[22] (Figure 1). Thus, developing specific PCR assays for this region is challenging, posing a major obstacle to dPCR analysis with distinct droplet clustering. Therefore, to overcome the challenges presented by this issue, we planned to find suitable additives to accurately detect variants of the TERT promoter region using dPCR. In previous studies, 7-deaza-dGTP has been shown to improve cluster resolution in dPCR in agreement with Colebatch et al[23]. CviQ1 was added to a dye-binding detection assay to increase cluster tightness[24]. An EDTA titration was performed to determine the applicable Mg²⁺ concentration in the PCR mixture. We conducted the assays with or without these additives in the PCR reactions and found a suitable concentration of these chemical additives. As a result, droplet clusters were separated in patient HC2873 after we optimized the assays. It was also necessary to vary amplification parameters and additives so that cluster separation and tightness could be maximized. The detailed amount of each reagent is shown in Figure 2.

Figure 1 Schematic representation of the commonly mutated nucleotides C228T (124 bp) and C250T (146 bp) in relation to the telomerase reverse transcriptase promoter positions and the genomic positions within chromosomal region 5 p15.33. TSS: Transcriptional start site; ATG: Adenine-thymine-guanine; TERT: Telomerase reverse transcriptase.

Figure 2 Two-dimensional digital polymerase chain reaction plots showing the telomerase reverse transcriptase C228T promoter assay with and without 7-deaza-dGTP, CviQI and EDTA reagents in patient HC2873. A: Preliminary results of amplification and detection of the C228T fragment using standard supermixtures and assays (left); the detailed amount of each reagent (right); B: The optimized detection conditions, including 7-deaza-DGTP, CviQI and EDTA reagents, and the experimental results of generating distinct droplet clusters for a C228T fragment (left). The optimized detailed dosage of each reagent (right). Color of line: Green, VIC probe telomerase reverse transcriptase (TERT) C228T wild-type; dark blue, 6-carboxyfluorescein probe for TERT C228T mutant type; light blue, droplets with TERT C228T mutant and wild-type allele; dark grey, droplets negative for template DNA. The green and blue lines are manually set thresholds for a 6-carboxyfluorescein mutant probe and VIC wild-type probe for digital polymerase chain reaction mutation detection, respectively.

dPCR specificity, sensitivity, and limits of detection

An orthogonal method, such as Sanger sequencing, was used to validate the specificity of the optimized assays on mutant templates. We measured the sensitivity of the assay by serially diluting wild-type genomic DNA into TERT promoter mutant genomic DNA. We established the LoB from 20 replicate wells of tumor samples from the patient who did not have TERT mutation in his biopsy tissue. The LoB for the C228T was 0.35 cp/μL. LoD, the lowest frequency of TERT C228T that can be accurately distinguished from five independent LL samples (each level performing four replicates) was found to be 0.55 cp/μL/well, as shown in Supplementary Table 1. Thirty out of 66 cases (45.45%) had 6-carboxyfluorescein concentrations higher than 0.55 cp/μL (LoD), which were treated as positive for TERT C228T promoter mutations in HCC cohort samples detected by dPCR (Figure 3).

Figure 3 Altered landscape of hepatocellular carcinoma cohort samples, combined with clinical variables (the lower half) and a bar graph of 6-carboxyfluorescein concentration (the upper half) of telomerase reverse transcriptase C228T mutation. A minimum concentration of 0.55 cp/μL (limit of detection) was evaluated to be positive in digital polymerase chain reaction. Comparison of telomerase reverse transcriptase status detected by Sanger sequencing and digital polymerase chain reaction in our hepatocellular carcinoma cohort (the table at the top of this picture). +: Telomerase reverse transcriptase C228T mutated; -: Telomerase reverse transcriptase C228T non-mutated; dPCR: Digital polymerase chain reaction; TERT: Telomerase reverse transcriptase; ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; LDH: Lactate dehydrogenase; GGT: Gamma-glutamyl transferase; APTT: Activated partial thromboplastin time; AFP: Alpha fetoprotein; MVI: Microvascular invasion; LCI: Liver capsular invasion; BVI: Blood vessel invasion; BCLC: Barcelona Clinic Liver Cancer; CI: Confidence interval.

Clinical characteristics of the HBV-related HCC patients

In our study, 66 HBV-related HCC patients who underwent hepatectomy were enrolled. All patients had HBV infection. Cirrhosis was present in 55 patients (83.3%). Ten patients (15.2%) had multiple malignant tumor foci. The patients were staged according to the BCLC classification. Seventeen patients (25.8%) had early HCC (BCLC A), 45 (68.2%) had intermediate HCC (BCLC B), and 4 (6.1%) had advanced (BCLC C). Most patients had HCC related to chronic alcohol abuse (n = 34, 51.5%). This included 38 cases with LCI and 17 with the BVI. Regarding tumor differentiation, the majority of tumors were moderately differentiated (n = 42, 63.6%), 21.2% were poorly differentiated (n = 14), and the remaining 15.2% (n = 10) were well differentiated. Elevated AFP (> 200 ng/mL) was seen in 19 patients (28.8%). These patient characteristics are summarized in Table 1.

Table 1 Clinicopathological characteristics of hepatectomy patients stratified by survival and recurrence status, n (%).

Characteristics	Survival			Recurrence
	Live, n = 44	Death, n = 22	P value	No, n = 24	Yes, n = 42	P value
Age			0.432			0.080
≤ 55	26 (59.1)	10 (45.5)		17 (70.8)	19 (45.2)
> 55	18 (40.9)	12 (54.5)		7 (29.2)	23 (54.8)
Gender			0.193			0.512
Male	38 (86.4)	16 (72.7)		21 (87.5)	33 (78.6)
Female	6 (13.6)	6 (27.3)		3 (12.5)	9 (21.4)
ALT			0.501			0.817
≤ 40 U/L	33 (75.0)	14 (63.6)		18 (75.0)	29 (69.0)
> 40 U/L	11 (25.0)	8 (36.4)		6 (25.0)	13 (31.0)
AST			0.117			0.560
≤ 40 U/L	37 (84.1)	14 (63.6)		20 (83.3)	31 (73.8)
> 40 U/L	7 (15.9)	8 (36.4)		4 (16.7)	11 (26.2)
LDH			0.281			0.736
≤ 214 U/L	39 (88.6)	17 (77.3)		21 (87.5)	35 (83.3)
> 214 U/L	5 (11.4)	5 (22.7)		3 (12.5)	7 (16.7)
GGT			0.025a			0.058
≤ 55 U/L	32 (72.7)	9 (40.9)		19 (79.2)	22 (52.4)
> 55 U/L	12 (27.3)	13 (59.1)		5 (20.8)	20 (47.6)
APTT			1.000			0.511
≤ 34.6 seconds	37 (84.1)	18 (81.8)		19 (79.2)	36 (85.7)
> 34.6 seconds	7 (15.9)	4 (18.2)		5 (20.8)	6 (14.3)
AFP			0.003a			0.817
≤ 200 ng/mL	37 (84.1)	10 (45.5)		18 (75.0)	29 (69.0)
> 200 ng/mL	7 (15.9)	12 (54.5)		6 (25.0)	13 (31.0)
Cirrhosis			0.739			0.733
No	8 (18.2)	3 (13.6)		3 (12.5)	8 (19.0)
Yes	36 (81.8)	19 (86.4)		21 (87.5)	34 (81.0)
Alcohol			0.258			0.561
No	24 (54.5)	8 (36.4)		10 (41.7)	22 (52.4)
Yes	20 (45.5)	14 (63.6)		14 (58.3)	20 (47.6)
Tumor number			0.720			0.306
Single	38 (86.4)	18 (81.8)		22 (91.7)	34 (81.0)
Multiple	6 (13.6)	4 (18.2)		2 (8.3)	8 (19.0)
Tumor diameter			1.000			1.000
≤ 2 cm	3 (6.8)	1 (4.5)		1 (4.2)	3 (7.1)
> 2 cm	41 (93.2)	21 (95.5)		23 (95.8)	39 (92.9)
MVI1			0.052			0.722
Negative	30 (75.0)	10 (45.5)		15 (68.2)	25 (62.5)
1	7 (17.5)	8 (36.4)		4 (18.2)	11 (27.5)
2	3 (7.5)	4 (18.2)		3 (13.6)	4 (10.0)
Differe			0.132			0.322
High	9 (20.5)	1 (4.5)		5 (20.8)	5 (11.9)
Medium	28 (63.6)	14 (63.6)		16 (66.7)	26 (61.9)
Low	7 (15.9)	7 (31.8)		3 (12.5)	11 (26.2)
LCI			1.000			1.000
No	19 (43.2)	9 (40.9)		10 (41.7)	18 (42.9)
Yes	25 (56.8)	13 (59.1)		14 (58.3)	24 (57.1)
BVI			0.022a			0.325
No	37 (84.1)	12 (54.5)		20 (83.3)	29 (69.0)
Yes	7 (15.9)	10 (45.5)		4 (16.7)	13 (31.0)
BCLC			0.551			0.388
A	13 (29.5)	4 (18.2)		7 (29.2)	10 (23.8)
B	29 (66.0)	16 (72.7)		17 (70.8)	28 (66.7)
C	2 (4.5)	2 (9.1)		0 (0.0)	4 (9.5)
Digital polymerase chain reaction TERT mutation			0.066			0.005a
No	28 (63.6)	8 (36.4)		19 (79.2)	17 (40.5)
Yes	16 (36.4)	14 (63.6)		5 (20.8)	25 (59.5)
Sanger TERT mutation			0.786			0.038a
No	28 (63.6)	13 (54.5)		19 (79.2)	21 (50.0)
Yes	16 (36.4)	9 (45.5)		5 (20.8)	21 (50.0)

¹Four patients had missing data.

^aP < 0.05 was significant.

ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; LDH: Lactate dehydrogenase; GGT: Gamma-glutamyl transferase; APTT: Activated partial thromboplastin time; AFP: Alpha fetoprotein; MVI: Microvascular invasion; Differe: Tumor differentiation; LCI: Liver capsular invasion; BVI: Blood vessel invasion; BCLC: Barcelona Clinic Liver Cancer; TERT: Telomerase reverse transcriptase.

TERT promoter mutations and tumor characteristics

Correlations between TERT promoter mutation status and clinicopathological variables in Sanger sequencing and dPCR are shown in Figure 3. 26 cases were examined for TERT C228T mutation by both Sanger sequencing and dPCR. However, 4 patients were positive for C228T by dPCR assay but negative by Sanger sequencing. The in-depth sequencing results of these 4 samples (all dPCR positive, but Sanger negative) associated with 6-carboxyfluorescein concentration and prognostic status were examined (Figure 4).

Figure 4 Illustration of sequencing results and follow-up data in four samples, which were positive for telomerase reverse transcriptase C228T by digital polymerase chain reaction assay, but were all negative by Sanger sequencing. Each column represents a sample of one patient with hepatocellular carcinoma after surgical resection. At the top of the figure, Sanger sequencing results (all were non mutated); In the middle of the figure, the two-dimensional digital polymerase chain reaction plot (the color of line: Green, wild-type; dark blue, mutant type; light blue, droplets with mutant and wild-type allele; dark grey, negative droplets) and the concentration (cp/μL) of the telomerase reverse transcriptase C228T mutant allele targeted by the 6-carboxyfluorescein dye labeled probes. 6-carboxyfluorescein concentration higher than LoD was positive. Clinical follow-up data collected from each patient are shown at the bottom of the figure. The orange and grey bars represent the time from surgery to tumor recurrence and death, respectively. PCR: Polymerase chain reaction.

We further analyzed the relationship between the TERT C228T mutation and clinicopathological features (Supplementary Table 2). Among 66 HCC patients, TERT C228T mutations were detected in 30 cases (45.5%). The mutation was significantly enriched in patients > 55 years (66.7% vs 27.8%, χ² = 9.981, P = 0.002). No significant associations were found with other clinicopathological factors (gender, ALT, LDH, APTT, AFP, cirrhosis, tumor characteristics, MVI, differentiation, LCI, BVI, BCLC stage). Trends toward higher mutation frequency were observed with elevated AST (33.3% vs 13.9%, P = 0.061), elevated GGT (50.0% vs 27.8%, P = 0.064), and heavy alcohol consumption (63.3% vs 41.7%, P = 0.079).

Sanger sequencing found that 39.39% (26/66) were positive, whereas dPCR found that 45.45% (30/66) were positive. The consistency of the two methods for detecting TERT C228T was good (concordance = 93.939). The Cohen’s kappa coefficient was 0.876 (95% confidence interval: 0.760-0.991) suggesting perfect agreement. While Sanger sequencing is regarded as the gold standard for detecting sequence variation, dPCR offers improved sensitivity over Sanger sequencing. By dPCR, TERT C228T mutation was detected with 100% sensitivity and 90% specificity, respectively (Figure 3). Therefore, in the next analysis we decided to choose the TERT mutation status detected by dPCR.

TERT mutation detected by dPCR assay is independent of conventional clinical characteristics associated with tumor recurrence

Tumor recurrence rates were 59.5% vs 50% in TERT C228T positive patients detected by dPCR vs Sanger sequencing (Table 1). Among those with available data, there was a significant correlation between tumor recurrence in the presence of the TERT C228T mutation detected by dPCR (P < 0.01). There were no significant correlations between tumor recurrence status and age, cirrhosis, AFP level, alcohol, or other clinical variables (Table 1). Kaplan-Meier analysis of the differences in OS and DFS in the study with and without TERT promoter mutations are shown in Figure 5. Kaplan-Meier analysis revealed significantly shorter OS (Figure 5A) and DFS (Figure 5B) in HBV-related HCC patients with TERT C228T mutation than in those without mutation. DFS was not significantly different between patients with AFP > 200 ng/mL and AFP ≤ 200 ng/mL, but was significantly different in terms of OS (Figure 5C and D). In all TERT C228T mutated patients, significant differences were found between those with AFP > 200 ng/mL and those with AFP ≤ 200 ng/mL in both OS (Figure 5E) and DFS (Figure 5F) (OS: P = 0.005, DFS: P = 0.01).

Figure 5 Kaplan-Meier curves based on telomerase reverse transcriptase C228T mutation status and alpha fetoprotein level. A and B: Kaplan-Meier curve of overall survival (OS; P = 0.028) and disease-free survival (DFS; P = 0.002) in telomerase reverse transcriptase C228T mutated patients; C and D: Kaplan-Meier curve of OS (P = 0.004) and DFS (P = 0.278) stratified by alpha fetoprotein; E and F: The telomerase reverse transcriptase C228T mutated patient’s OS (P = 0.005) and DFS (P = 0.01) stratified by alpha fetoprotein. TERT: Telomerase reverse transcriptase; HR: Hazard ratio; AFP: Alpha fetoprotein.

We performed univariate and multivariate Cox regression analyses to examine whether TERT mutations in our HCC cohort were independent of other clinical variables. Following univariate Cox regression analysis significant variables (P < 0.05) were selected for multivariate Cox regression analysis. The univariate analysis determined five parameters that were slightly or significantly associated with OS: TERT mutation status, AST, GGT, AFP, BVI and MVI were slightly associated with OS (Table 2). Following multivariate Cox regression analysis TERT mutation status, GGT and BVI, significantly and independently determined DFS (Table 2). In addition, the C-index was also compared in the nomograms for OS and DFS, TERT mutation and 18 clinical characteristics. The nomograms successfully differentiated TERT C228T mutant and non-mutant patients with HBV-related HCC associated with OS and DFS (Figure 6A and B). Overall, these results showed that TERT mutations consistently outperformed conventional clinical characteristics in survival prediction.

Figure 6 Construction and validation of overall survival and disease-free survival nomograms. A and B: The prognostic performance between the telomerase reverse transcriptase mutation status and different conventional clinical characteristics was compared by calculating the overall survival (OS) and disease-free survival (DFS) C-index. The maximum C-index of the OS and DFS nomogram model was 0.7651 and 0.6899, respectively. The maximum C-index of telomerase reverse transcriptase (TERT) mutation in OS and DFS was 0.6224 and 0.6228, respectively; C: OS-nomogram was used to predict aspartate aminotransferase, gamma-glutamyl transferase, microvascular invasion, blood vessel invasion and TERT mutation status; D: DFS-nomogram was used to predict gamma-glutamyl transferase, blood vessel invasion and TERT mutation status; E and F: Receiver operating characteristic curves showed the 1-, 2-, and 5-year predictive efficiency of the OS and DFS nomogram model. OS: Overall survival; DFS: Disease-free survival; AST: Aspartate aminotransferase; GGT: Gamma-glutamyl transferase; AFP: Alpha fetoprotein; MVI: Microvascular invasion; BVI: Blood vessel invasion; TERT: Telomerase reverse transcriptase; AUC: Areas under the curve; TPR: True positive rate; FPR: False positive rate.

Table 2 Univariate and multivariate Cox regression analyses of overall survival and disease-free survival in hepatocellular carcinoma patients after hepatectomy.

Characteristics	OS, univariate analysis		OS, multivariate analysis		DFS, univariate analysis		DFS, multivariate analysis
	HR (95%CI)	P value	HR (95%CI)	P value	HR (95%CI)	P value	HR (95%CI)	P value
Age (> 55/≤ 55)	0.65 (0.28-1.50)	0.311			0.64 (0.35-1.18)	0.149
Gender (female/male)	1.81 (0.71-4.64)	0.214			1.43 (0.68-3.00)	0.341
ALT (≤ 40/> 40 U/L)	1.70 (0.71-4.05)	0.235			1.45 (0.76-2.80)	0.264
AST (≤ 40/> 40 U/L)	2.84 (1.17-6.90)	0.021a	1.17 (0.42-3.24)	0.766	1.70 (0.85-3.39)	0.132
LDH (≤ 214/> 214 U/L)	1.71 (0.63-4.63)	0.295			1.22 (0.54-2.74)	0.639
GGT (≤ 55/> 55 U/L)	3.40 (1.43-8.10)	0.006a	2.40 (0.88-6.54)	0.087	2.53 (1.37-4.68)	0.003a	2.14 (1.13-4.05)	0.019a
APTT (> 34.6 seconds/≤ 34.6 seconds)	0.99 (0.33-2.93)	0.984			1.45 (0.61-3.46)	0.398
AFP (> 200/≤ 200 ng/mL)	0.28 (0.12-0.66)	0.004a	0.36 (0.13-0.98)	0.044a	0.91 (0.49-1.68)	0.756
Cirrhosis (yes/no)	0.85 (0.25-2.86)	0.788			1.41 (0.65-3.06)	0.380
Alcohol (no/yes)	2.13 (0.89-5.11)	0.090			0.93 (0.51-1.70)	0.806
Tumor number (single/multiple)	1.46 (0.49-4.33)	0.495			1.61 (0.75-3.49)	0.224
Tumor diameter (> 2 cm/≤ 2 cm)	0.62 (0.08-4.64)	0.644			1.44 (0.44-4.68)	0.545
MVI		0.038a				0.634
Negative	Reference				Reference
2	3.58 (1.11-11.53)	0.032a	1.24 (0.22-6.84)	0.806	1.25 (0.43-3.59)	0.683
1	2.68 (1.05-6.82)	0.038a	0.89 (0.20-4.04)	0.877	1.40 (0.69-2.86)	0.353
Differe		0.086				0.086
Medium	Reference				Reference
High	0.27 (0.04-2.08)	0.210			0.74(0.29-1.94)	0.544
Low	2.14 (0.85-5.37)	0.105			2.02 (1.00-4.11)	0.052
LCI (yes/no)	0.97 (0.41-2.29)	0.948			1.08 (0.59-2.01)	0.797
BVI (no/yes)	4.36 (1.82-10.43)	< 0.001a	2.23 (0.50-9.84)	0.292	2.19 (1.13-4.23)	0.020a	2.27 (1.15-4.50)	0.018a
BCLC		0.230				0.096
A	Reference				Reference
B	1.66 (0.55-4.97)	0.365			2.50 (1.01-6.14)	0.047a
C	2.64 (0.48-14.53)	0.264			1.67 (0.75-3.72)	0.206
TERT mutation (no/yes)	2.70 (1.12-6.54)	0.028a	2.27 (0.84-6.14)	0.107	2.69 (1.45-5.02)	0.002a	2.43 (1.28-4.63)	0.007a

^aP < 0.05 was significant.

OS: Overall survival; DFS: Disease-free survival; ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; LDH: Lactate dehydrogenase; GGT: Gamma-glutamyl transferase; APTT: Activated partial thromboplastin time; AFP: Alpha fetoprotein; MVI: Microvascular invasion; Differe: Tumor differentiation; LCI: Liver capsular invasion; BVI: Blood vessel invasion; BCLC: Barcelona Clinic Liver Cancer; TERT: Telomerase reverse transcriptase; HR: Hazard ratio; CI: Confidence interval.

A nomogram for TERT mutations was constructed and validated

Based on the final multivariable model, the nomogram was developed to predict survival at 1-year, 2-years, 5-years and recurrence based on risk factors in HBV-related HCC patients who underwent hepatectomy. The total score was 0-100 and 0-300, respectively, and each variable was calculated and merged. A high score indicates a high risk of death and recurrence in relation to OS and DFS, respectively (Figure 6C and D). For example, an HBV-related HCC patient with TERT mutation (100 points) GGT > 55 U/L (86 points), and without BVI (0 points) had a total score of 186 points. The 1-year recurrence probability was 71%. Compared with other clinical information, TERT mutations contributed most to the risk score (0 to 100), based on Cox multiple regression results. The inclusion of TERT mutation status improved the predictive accuracy of both OS and DFS nomograms. The OS and DFS models incorporating TERT mutation had maximum C-index values of 0.7651 and 0.6899, respectively, compared to 0.7498 and 0.6492 for models excluding TERT mutation. These findings underscore the significant contribution of TERT mutation in predicting prognosis in HBV-related HCC patients after hepatectomy (Supplementary Figure 1).

Areas under the curve for OS at 1-year, 2-years, and 5-years follow-up were 0.614, 0.800, and 0.834, respectively (Figure 6E). Areas under the curve for DFS at 1-year, 2-years, and 5-years follow-up were 0.849, 0.688 and 0.772, respectively. The models’ performance was evaluated by calibration and discrimination (Figure 6F). Nomogram predictions are represented on the X-axis, and actual survival and recurrence probabilities are displayed on the Y-axis. An ideal curve was a 45° line, indicating good agreement between prediction and observation. Calibration curves for OS and DFS were plotted at 1-year, 2-year, and 5-year intervals, revealing consistent results between predicted and observed data. (OS: Figure 7A-C; DFS: Figure 7D-F). Based on the results of our study, we found that particularly in the 5-year prediction, the performance of nomograms plotted as solid lines were satisfactory (Figure 7C and F). A DCA was used to determine whether the models could guide treatment in our HBV-related HCC cohort. Additionally, high predictive efficiency of the nomogram was observed for OS and DFS in HBV-related HCC patients after hepatectomy, based on DCA curves at 1-year, 2-years, and 5-years (OS: Figure 8A-C; DFS: Figure 8D-F). Overall, the nomogram predicted whether HBV-related HCC patients would relapse after hepatectomy more accurately than individual prognostic factors.

Figure 7 Validation of the overall survival and disease-free survival nomogram. A-C: Overall survival; D-F: Disease-free survival. Calibration plots were used to predict the 1-, 2-, and 5-year overall survival and disease-free survival in our hepatitis B virus-related hepatocellular carcinoma cohort, respectively.

Figure 8 Decision curve analysis. A-C: Overall survival; D-F: Disease-free survival. Decision curve analysis showed the 1-, 2-, and 5-year predictive efficiency of the overall survival and disease-free survival nomogram model. OS: Overall survival; DFS: Disease-free survival; TERT: Telomerase reverse transcriptase; AFP: Alpha fetoprotein; AST: Aspartate aminotransferase; GGT: Gamma-glutamyl transferase; MVI: Microvascular invasion; BVI: Blood vessel invasion.

DISCUSSION

In hepatocarcinogenesis, TERT promoter mutation is recognized as an early event. Somatic mutation status, particularly TERT mutations, has been associated with poor prognosis and is considered a potential predictor of liver cancer outcomes[9]. However, the GC-rich region of the TERT promoter hampers amplification efficiency, reducing assay performance[17]. To address this, we employed dPCR to achieve more accurate detection of TERT promoter variants.

With its high sensitivity, precision, and ability to provide absolute quantification without a standard curve, dPCR has been widely applied to detect pathogens and somatic mutations[25]. In our study, after optimizing the assay, dPCR demonstrated high concordance, sensitivity, and specificity compared with Sanger sequencing when detecting DNA from fresh HBV-related HCC tumor tissues. Notably, all patients who tested positive for the TERT C228T mutation by dPCR but negative by Sanger sequencing experienced postoperative recurrence and subsequently died during follow-up. Some of these cases were only slightly above the detection threshold (LoD 0.55 cp/μL), underscoring the importance of carefully defining this cutoff. These findings highlight the superiority of dPCR in detecting GC-rich TERT promoter mutations, and its potential to provide stronger prognostic information. Previous studies comparing the two methods yielded unsatisfactory results[26], possibly due to limited postoperative follow-up data, which may have underestimated the prognostic relevance of TERT mutations. In this study, TERT C228T mutation was enriched in patients older than 55 years, suggesting an age-related accumulation of somatic mutational burden or selective retention under chronic liver injury and replicative stress. Mutation-positive cases showed a higher proportion of poor differentiation, implying a potential role in tumor dedifferentiation. Trends toward association with elevated AST, GGT, and heavy alcohol consumption suggest links with chronic hepatic inflammation and metabolic stress. Given the limited sample size and statistical power, future studies with larger cohorts, multivariate analyses, and survival evaluation are warranted to clarify the biological and prognostic significance of TERT C228T mutation in HCC.

Furthermore, Jang et al[27] reported that TERT promoter mutations detected in resected HCC tissues were significantly correlated with survival but not with recurrence. Another study found that TERT promoter mutations alone were not independently associated with overall or recurrence-free survival; however, when combined with single nucleotide polymorphism rs2853669, they were linked to poorer outcomes[28]. In contrast, our HBV-related HCC cohort demonstrated that the TERT C228T mutation was strongly associated with both OS and DFS. Notably, elevated preoperative AFP levels (> 200 ng/mL) in TERT C228T-mutant patients correlated with worse survival (P = 0.005) and higher recurrence risk (P = 0.01), a finding likely attributable to the exclusive HBV etiology in our cohort. The TERT promoter has been identified as the most frequent HBV integration site in HCC[29]. Although HBV integration and TERT promoter mutations may occur independently, both events have been linked to increased TERT transcriptional activation[30,31]. Jang et al[32] further reported that 65.4% (17/26) of HBV-related HCC cases harbored either TERT promoter mutations or HBV insertions into TERT or MLL4, underscoring the role of these alterations in HBV-related hepatocarcinogenesis[33]. In addition, Sant’Anna et al[7] reported that the C228T mutation in the TERT promoter was detected in 27.1% of samples, with higher prevalence in HCC tissues (45.2%) compared to cirrhotic (19%) and hepatitis tissues (0%). Significant differences were observed between HCC and hepatitis tissues, as well as between tumor and non-tumor tissues overall, while the difference between HCC and cirrhotic tissues showed a borderline trend (P = 0.0542). Regardless of mutation status, TERT expression was markedly elevated in HCC compared with cirrhotic and hepatitis tissues, indicating that promoter mutations are not the sole drivers of its upregulation[34]. This suggests that TERT activation is mainly associated with malignant transformation[35]. Given its central role in hepatocarcinogenesis, TERT upregulation may result from inflammatory signaling, secondary events such as p53 Loss, or HBV integration mechanisms involving E-twenty six transcription factor E74-like factor 4-driven transcriptional activation[10,36]. Therapeutic strategies targeting TERT, including vaccines (such as GV1001, GRNVAC1) and combinatorial approaches, are under development[5,37]. Collectively, these findings support the prognostic relevance of TERT promoter mutations, although further large-scale validation is required.

To further validate the prognostic value of TERT promoter mutations, we constructed a nomogram to predict outcomes in HBV-related HCC patients after hepatectomy. Our findings suggest that detection of the TERT C228T mutation by dPCR improves recurrence prediction and should be considered alongside clinical variables such as AST, GGT, AFP, MVI, and BVI for OS, and GGT, BVI, and TERT C228T for DFS. Calibration curves demonstrated good agreement between predicted and observed outcomes at 1-, 2-, and 5-year intervals. This model provides complementary insights into tumor biology and enables patient-specific risk stratification. Notably, Zhu et al[38] reported that patients with TERT mutations may benefit from atezolizumab plus bevacizumab in unresectable HCC, further highlighting its clinical relevance. With continued validation in larger cohorts, this nomogram may serve as a valuable tool to guide precision management in HBV-related HCC.

Several limitations in this study should be acknowledged. First, this study was constrained by a relatively small sample size and the absence of validation in a large-scale cohort, which may have introduced selection bias. Second, the lack of a randomized control group represents another major limitation, partly due to the low incidence of non-HBV-related HCC in our center. Third, the findings may be influenced by epidemiological differences in HBV-related HCC across race, ethnicity, and region, as well as variations in surgical techniques and indications among institutions.

In summary, we optimized a dPCR assay and established detection thresholds, demonstrating that this highly sensitive and specific method identifies HBV-related HCC patients at higher risk of recurrence more effectively than Sanger sequencing. This study is the first to develop nomogram models incorporating TERT mutation status to validate its prognostic value in HBV-related HCC, while also highlighting the clinical utility of dPCR for somatic mutation detection. Nonetheless, larger multicenter and prospective studies are needed to confirm the reliability of this model. Our findings may provide a useful reference for precision treatment strategies in HBV-related HCC and contribute to precision oncology. Moreover, extending this approach to diverse sample types, including circulating tumor DNA from liquid biopsy, may enhance postoperative monitoring and broaden its applicability to other cancers such as urothelial carcinoma.

CONCLUSION

In this study, we optimized a highly sensitive and specific dPCR assay for detecting TERT promoter mutations in HBV-related HCC patients and demonstrated its superiority over conventional Sanger sequencing. We established and validated the first nomogram models incorporating TERT mutation status to predict recurrence risk, highlighting its potential clinical utility for precision treatment strategies.

ACKNOWLEDGEMENTS

The authors are grateful not only to the consent of patients and their family members, but also the investigators and research staff involved.