Telomerase reverse transcriptase mutations in plasma DNA in patients with hepatocellular carcinoma or cirrhosis: Prevalence and risk factors

PATIENTS, DATA SOURCES, AND BIOSPECIMENS

This study was approved by the institutional review boards of all collaborating institutions. The Cancer Genome Atlas (TCGA) HCC data were extracted from a recent study,6 the cBioPortal online platform, and Firebrowse.10 Clinical and demographic parameters from 196 HCCs with available TERT promoter mutations data in TCGA are summarized in Supporting Table S1. Plasma samples from 218 histologically confirmed patients with HCC were collected between 2002 and 2010, prior to treatment at MD Anderson Cancer Center. The demographic and clinical parameters of these 218 patients with HCC are described in Supporting Table S2. Formalin-fixed paraffin-embedded HCCs of 15 patients with HCC were collected at the University of Texas Medical Branch. The demographic and clinical parameters of these patients with HCC are described in Supporting Table S3. Plasma samples from 81 patients with cirrhosis but no HCC were collected at Baylor-St Luke's Hospital. The demographic and clinical parameters of these 81 patients with cirrhosis are described in Supporting Table S4. cfDNA was extracted from plasma samples (250-500 μL) using the QiAamp circulating nucleic acid kit (Qiagen). cfDNA peaks at a size range of 75-225 base pairs were visualized on a fragment analyzer (Advanced Analytical Technologies) and their heights used to estimate cfDNA amount. DNA was extracted from tumor, adjacent liver, and distant nontumoral liver areas, using the QIAamp DNA formalin-fixed paraffin-embedded tissue kit (Qiagen). DNA samples were then quantified using the Qubit fluorometer and the double-stranded DNA high-sensitivity assay kit (Thermo Fisher Scientific). DNA quality was assessed using a fragment analyzer and high sensitivity genomic DNA analysis kit (Advanced Analytical Technologies).

TCGA HCC MESSENGER RNA EXPRESSION ANALYSIS

Tumoral messenger RNA (mRNA) expression data and TERT promoter mutation data were available for 193 HCCs in TCGA. We compared gene expression in tumors with and without TERT promoter mutations, using feature-by-feature t tests and fit the resulting P values to the beta-uniform mixture model to allow for estimation of the false discovery rate (FDR) and to determine P value cutoffs for specified FDR values. Linear gene expression fold changes were then calculated for each gene between the two tumor groups. We considered genes to be differentially expressed if the linear fold change had an absolute value greater than 1.5 and the P value was less than the modeled cutoff at an FDR of 5% (corresponding P value of 0.0038). Gene expression analysis was conducted in R version 3.3 using the ClassComparison and ClassDiscovery packages.

DROPLET DIGITAL POLYMERASE CHAIN REACTION

TERT C228T and C250T promoter mutations were detected by polymerase chain reaction (PCR) using the QX200 droplet digital PCR (ddPCR) system (Bio-Rad Laboratories, Inc.) and the dHsaEXD72405942 assay for TERT C228T and dHsaEXD46675715 assay for TERT C250T. Mutant and wild-type (wt) TERT alleles were differentiated by the fluorophores attached to the probes, with HEX fluorescence for wt TERT alleles and FAM fluorescence for mutant TERT alleles. A total of 4 μL of DNA per 20 μL ddPCR reaction was used in each reaction. Thermocycling conditions were 95°C for 10 minutes followed by 50 cycles of 96°C for 30 seconds and 62°C for 1 minute followed by 98°C for 10 minutes. The sealed plates were then placed in the droplet reader for detection of complete ddPCR reactions in individual droplets. The data were analyzed using QuantaSoft software (Bio-Rad Laboratories, Inc.). Samples with a mutant TERT allele fraction ≥0.1% and with at least two visible mutant signals were considered positive for the mutation, as described.11, 12 Because the two TERT promoter mutations are mutually exclusive,6, 7 we measured TERT C250T mutation in samples that were negative for TERT C228T mutation.

SANGER SEQUENCING

TERT promoter PCR was performed in 20-μL reactions and using a hotstart DNA polymerase (KAPA Biosystems) activation (95°C, 15 minutes), 40 denaturation cycles (94°C, 30 seconds), primer annealing (62°C, 30 seconds), and extension (72°C, 30 seconds) followed by a final extension (72°C, 5 minutes). PCR products were purified using the QIAquick Gel Extraction kit (Qiagen). Purified PCR products were sent to the MD Anderson Sequencing and Microarray Facility for Sanger sequencing. The primers used for PCR amplification and sequencing were forward CAGCGCTGCCTGAAACTC, reverse GTCCTGCCCCTTCACCTT. The sequences were visualized and analyzed using FinchTV software (Geospiza).

STATISTICAL ANALYSIS

Two-tailed χ² tests and Fisher exact tests were used to test independence of patient characteristics and TERT promoter mutations in all data sets. Univariable logistic regression was used to generate odds ratios (ORs) and 95% confidence intervals (CIs) for the association of each patient characteristic with TERT promoter mutations. In addition, we calculated the adjusted OR (AOR) and 95% CI for each clinical parameter after adjusting for age and sex. To ensure that our results were not confounded by the amount of cfDNA extracted from plasma, we repeated the logistic regression analysis controlling for cfDNA concentration. We assessed the difference in overall survival in patients with and without TERT promoter mutations by visualizing Kaplan-Meier survival curves and testing the difference of the survival curves using the log-rank test. Analyses were conducted in R version 3.3.

DEMOGRAPHIC AND CLINICAL CHARACTERISTICS OF PATIENTS WITH HCC WITH TERT MUTATIONS IN TUMORS

Two TERT promoter mutations, chr5 1,295,228 G>A (C228T) and 1,295,250 G>A (C250T), were measured by Sanger sequencing in 196 patients with HCC in TCGA and recently reported.6 The clinical and demographic characterization of these 196 patients are summarized in Supporting Table S1. The majority of these patients were male (65.8%) and non-Hispanic Caucasian individuals (57.8%), and the prevalence of obesity was 22.9%. These HCCs were associated with HCV (17.9%), HBV (22.4%), or NAFLD (4.3%), and for 27.4% of them, no known risk factor was identified. Only 25.3% of these patients had underlying cirrhosis. The majority of these HCCs were early stage (41.3% stage I and 23.4% stage II). The overall prevalence of TERT promoter mutations was 44.4%, and the two mutations C228T and C250T were mutually exclusive. We further characterized the patients with HCC with TERT promoter mutations (Table 1). We found that the presence of TERT promoter mutations was associated with male sex (OR, 2.5; 95% CI, 1.3-4.7; P = 0.005), older age (>65 years) (OR, 2.3; 95% CI, 1.3-4.3; P = 0.003), and HCV (OR, 2.5; 95% CI, 1.2-5.4; P = 0.002) and negatively associated with HBV (OR, 0.4; 95% CI, 0.2-0.8; P = 0.011). We also found that TERT promoter mutations were associated with family history of cancer (OR, 2.1; 95% CI, 1.1-4.1; P = 0.018), the presence of cirrhosis (OR, 2.0; 95% CI, 1.0-3.9; P = 0.044), and alcoholic etiology (OR, 2.3; 95% CI, 1.2-4.3; P = 0.005). Indeed, TERT promoter mutations were found in most patients with alcoholic etiology (58.5%), alcoholic cirrhosis in particular (66.7%). After adjusting for age and sex, there were no changes in significance or direction of association with male sex, older age, and HBV; significance became stronger for family history of cancer (AOR, 2.7; 95% CI, 1.3-5.9; P = 0.007), and association with cirrhosis and alcoholic etiology became nonsignificant. Finally, we did not find any association with obesity, NAFLD, tumor stage, or any laboratory test (alpha-fetoprotein, platelet count, albumin, and bilirubin).

DEMOGRAPHIC AND CLINICAL CHARACTERISTICS OF PATIENTS WITH HCC WITH TERT PROMOTER MUTATIONS IN PLASMA cfDNA

We then measured TERT C228T and C250T promoter mutations in cfDNA extracted from plasma collected from 218 patients with HCC. The demographic and clinical parameters of these 218 patients are summarized in Supporting Table S2. The ethnicity/race distribution among these 218 patients with HCC was non-Hispanic Caucasian (54.6%), Hispanic (16.1%), Asian (4.6%), Black (7.8%), and unknown (17%). The majority of these patients were male individuals (72%), and 37.9% were obese, 40.8% had diabetes, 32.5% were positive for HCV, 20.9% were positive for HBV, and 10.9% had NAFLD. Underlying cirrhosis was present in 54.5% of the patients, with 19.1% alcoholic cirrhosis and 35.4% nonalcoholic cirrhosis. No known risk factor could be identified in 9.5% of the patients. The distribution of well-, moderately, and poorly differentiated tumors was 39.6%, 34.3%, and 26.1%, respectively. We measured TERT C228T and C250T mutations in plasma cfDNA using ddPCR. Droplets positive for mutant alleles, positive for wt alleles, and double positive for both mutant and wt alleles were clearly separated, as shown in Fig. 1. Black, blue, green, and orange dots represent empty droplets, mutant positive droplets, wt-positive droplets, and double-positive wt and mutant droplets, respectively. PLC/PRF-5 and HepG2 cell lines, not harboring and harboring TERT C228T mutation, respectively, were used as controls (Fig. 1A). HCC tumor samples HCC-T13 and HCC-T9, not harboring and harboring TERT C250T mutation, respectively, as determined by Sanger sequencing (Supporting Fig. S1), were also used as controls (Fig. 1B). TERT C228T or C250T mutation was detected in 104 out of the 218 HCCs (47.7%), with mutant allele fractions ranging from 0.1% to 50.2%. As observed in the tumors, the distribution of TERT promoter mutations differed according to demographic and clinical characteristics (Table 2). As for TCGA HCCs, we found that the presence of TERT promoter mutations was associated with male sex (OR, 3.1; 95% CI, 1.6-6.3; P = 0.0003) and the presence of cirrhosis (OR, 1.7; 95% CI, 1.0-3.1; P = 0.041) with alcoholic cirrhosis in particular (OR, 2.2; 95% CI, 1.1-4.8). TERT promoter mutations were again found in most patients with alcoholic cirrhosis (62.5%). Associations with HCV (OR, 1.5; 95% CI, 0.8-2.6; P = 0.192) and with family history of cancer (OR, 1.7; 95% CI, 0.9-3.0; P = 0.080) were also found but did not reach significance. As for TCGA HCCs, the prevalence of TERT promoter mutations was low in patients with NAFLD. We further found an inverse association between TERT promoter mutations and obesity (OR, 0.6; 95% CI, 0.3-0.8; P = 0.039) or history of diabetes (OR, 0.5; 95% CI, 0.3-0.8; P = 0.010). In contrast to TCGA HCCs, no associations with older age or inverse association with HBV were found. After adjusting for age and sex, there were no changes in significance or direction of association with male sex, the presence of cirrhosis, HCV, and family history of cancer; the inverse association became stronger with diabetes (AOR, 0.4; 95% CI, 0.2-0.7; P = 0.003), while association with alcoholic cirrhosis and inverse association with obesity became not significant. Additional associations were identified, such as place of birth in the United States or Canada (AOR, 2.7; 95% CI, 1.4-5.6; P = 0.047) and multicentric tumor (AOR, 2.2; 95% CI, 1.2-4.2; P = 0.012). In addition, the odds of having a TERT promoter mutation increased with tumor stage (overall P value 0.0003). However, tumor stage was also associated with male sex (P = 0.0171), with over 75% of stage III/IV tumors diagnosed in male patients (data not shown). Therefore, male sex partially explains the association between stage and TERT promoter mutations.

To rule out the possibility that detection of TERT promoter mutations correlated with amounts of extracted cfDNA, we reanalyzed the data after controlling for cfDNA amounts. Only minor changes in measures of association were noted. The most extreme changes were that the OR for birthplace increased from 2.7 to 3.3 and the OR for Child score C relative to A decreased from 1.2 to 1.0, with no changes in statistical significance (Supporting Table S5).

INDEPENDENT VALIDATION IN PATIENTS WITH HCC WITH HCV AND ALCOHOLIC CIRRHOSIS

Altogether, the results from both HCC tumor and HCC cfDNA data sets suggest that the TERT promoter is very frequently mutated in male individuals with HCV and alcoholic cirrhosis. Indeed, while the prevalence of TERT promoter mutations in the HCC tumors and HCC cfDNAs analyzed above was 44.4% and 47.7%, respectively, it reached 100% and 80% in tumors and cfDNAs from male patients with HCC with combined HCV and alcoholic cirrhosis, respectively. We further confirmed this result in an independent set of 15 HCC tumors from male patients with HCV and alcoholic cirrhosis. Demographic and clinical parameters of these 15 patients are summarized in Supporting Table S3. The distribution of well-, moderately, and poorly differentiated tumors was 20%, 46.7%, and 33.3%, respectively. We used ddPCR to measure TERT promoter mutations C228T and C250T in tumors, adjacent liver, and distant liver. Using a cutoff for a mutant allele's fractional abundance of 2.5%, TERT promoter mutations were detected in 13 out of the 15 tumors (86.6%). TERT promoter mutations were also detected in six of the corresponding adjacent liver and in one distant liver (Fig. 2). Among those samples with TERT promoter mutations, the mean mutation fractional abundance in tumors, adjacent liver, and distant liver was 35.2% (range, 5.6%-59.5%), 6.0% (range, 2.5%-12.5%), and 2.75%, respectively, confirming the association of TERT mutations with hepatocarcinogenesis.

DEMOGRAPHIC AND CLINICAL CHARACTERISTICS OF PATIENTS WITH CIRRHOSIS AND NO HCC WITH TERT PROMOTER MUTATIONS IN PLASMA cfDNA

Finally, we measured TERT promoter mutations C228T and C250T in cfDNA isolated from plasma collected from 81 patients with cirrhosis without HCC. The demographic and clinical parameters of these 81 patients are shown in Supporting Table S4. The ethnicity/race distribution among these 81 patients was 69.1% non-Hispanic Caucasian, 19.8% Hispanic, 3.7% Asian, and 7.4% Black. Among these 81 patients, 45.7% were male individuals, 42% were positive for HCV, 2.5% were positive for HBV, 13.6% had alcoholic cirrhosis, and 38.3% had NAFLD. The prevalence of TERT promoter mutations in cfDNA of these patients with cirrhosis was 8.6%. Among the 7 patients with TERT promoter mutations, 5 were male individuals (71.4%), 3 were positive for HCV (42.8%), 2 had alcoholic cirrhosis (28.6%), and 1 had NAFLD (14.3%), demonstrating that, similar to patients with HCC, TERT promoter mutations were more prevalent in males with HCV or alcoholic disease and less prevalent in patients with NAFLD.

OVERALL SURVIVAL OF PATIENTS WITH HCC WITH TERT PROMOTER MUTATIONS IN TUMOR OR IN PLASMA cfDNA

We further evaluated whether noninvasive detection of TERT promoter mutations in plasma cfDNA could have utility in prognosis. To that end, we calculated and plotted Kaplan-Meier survival curves. The overall survival of these patients, stratified by the presence or absence of TERT promoter mutations, is shown in Fig. 3A. In the HCC patient group with TERT promoter mutations measured in plasma cfDNA, overall survival decreased rapidly after diagnosis, with overall survival of patients with TERT promoter mutations significantly lower than for those without TERT promoter mutations (P = 0.0062). In multivariate analysis adjusting for tumor stage, the association was not significant (P = 0.19). When repeating a similar analysis only in patients with cirrhosis, overall survival of patients with TERT promoter mutations was again significantly lower than for those without TERT promoter mutations (P = 0.0042), and in multivariate analysis adjusting for tumor stage, the association remained significant (P = 0.051) (Fig. 3B).

TUMORAL TRANSCRIPTOMIC SIGNATURE ASSOCIATED WITH TERT PROMOTER MUTATIONS

To determine whether TERT promoter mutations are associated with a specific tumor transcriptomic signature, we downloaded mRNA expression data from 193 TCGA HCCs with TERT promoter mutation information available. The expression of 499 genes was significantly changed in tumors with mutated TERT compared to tumors with no TERT mutations (350 down-regulated and 149 up-regulated) (Fig. 4A; Supporting Table S6). The largest expression changes were observed for PR/SET domain 13 (PRDM13), potassium voltage-gated channel subfamily B member 2 (KCNB2), aldo-keto reductase family 1 member B10 (AKR1B10), SRY (sex determining region Y)-box 2 (SOX2), solute carrier family 22 member 12 (SLC22A12), and F-box and WD repeat domain-containing 10 (FBXW10), all overexpressed in samples with TERT promoter mutations compared to samples without mutation (Fig. 4B). Epithelial cell adhesion molecule (EPCAM), GATA binding protein 5 (GATA5), tripartite motif containing 50 (TRIM50) neuronal pentraxin 2 (NPTX2), and protease serine 8 (PRSS8) were the strongest underexpressed genes in tumors with TERT promoter mutations (Fig. 4B). Ingenuity Pathway Analysis (IPA) of all 499 genes identified xenobiotic metabolism signaling (P = 6.02 × 10^–5), nicotine degradation III (P = 1.58 × 10^–4), and serotonin degradation (P = 2.17×10^–4) as the top canonical pathways and vitamin K2 (P = 4.91 × 10^–5) and thioacetamide (P = 1.50 × 10^–4) as the top upstream regulators affected by TERT promoter mutations (Fig. 4C). Most genes included in the identified top canonical pathways were genes involved in alcohol metabolism.