Integrated analysis of telomerase enzymatic activity unravels an association with cancer stemness and proliferation

Abstract

Active telomerase is essential for stem cells and most cancers to maintain telomeres. The enzymatic activity of telomerase is related but not equivalent to the expression of TERT, the catalytic subunit of the complex. Here we show that telomerase enzymatic activity can be robustly estimated from the expression of a 13-gene signature. We demonstrate the validity of the expression-based approach, named EXTEND, using cell lines, cancer samples, and non-neoplastic samples. When applied to over 9,000 tumors and single cells, we find a strong correlation between telomerase activity and cancer stemness. This correlation is largely driven by a small population of proliferating cancer cells that exhibits both high telomerase activity and cancer stemness. This study establishes a computational framework for quantifying telomerase enzymatic activity and provides new insights into the relationships among telomerase, cancer proliferation, and stemness.

Fig. 1. EXTEND Validation.

a Correlation between EXTEND score and TRAP assay readouts in 28 glioma sphere-forming cell lines. Two ALT cell lines are labeled in blue. For (a–c), Spearman’s correlation was used to calculate P value and Rho, and shade indicates 95% CI of the regression calculated by ggplot2 using default parameters. b Correlation between EXTEND score and direct enzymatic assay results in 11 bladder cancer cell lines. c Correlation between EXTEND score and digital droplet TRAP assay in 15 lung cancer cell lines. d EXTEND scores across ALT and telomerase-positive tumors in liposarcomas (“GSE14533”). Both telomerase-positive tumors (n = 8) and cell lines (n = 8) show significantly higher EXTEND scores than ALT samples (cell lines, n = 8 and tumors, n = 10). e EXTEND scores across the five TMM groups of neuroblastoma (“GSE120572”). Telomerase-positive groups (MYCN amplification (n = 52), TERT high (n = 9), and TERT rearrangement (n = 21)) show significantly higher scores than ALT (n = 31) and no TMM (n = 99) groups. Boxplots, interquartile ranges (25–75th percentile); middle bar defines median and the minima and maxima are within 1.5 times the interquartile range of the lower and higher quartile. Statistical differences were assessed using two-sided Student’s t test in (d) and (e). Data used are available in Source Data.

Fig. 2. EXTEND scores across normal tissues and during embryonic development.

a EXTEND scores and TERT expression across 52 sub-tissues across Genotype Tissue Expression (GTEx) data (n = 11,688). The left y-axis indicates TERT expression (sea green), while the right y-axis represents EXTEND scores (pink). Boxplot interquartile ranges (25–75th percentile); middle bar defines median and the minima and maxima are within 1.5 times the interquartile range of the lower and higher quartile. b, c TERT expression (blue) and EXTEND scores (red) across tissue development phases in (b) heart tissue and (c) liver tissue. Red and blue lines represent the mean of EXTEND scores and TERT expression of samples from each age group. d EXTEND predicts higher telomerase activity in PARN mutant (n = 2) and PARD5 knockout (n = 2) sample (left two bars), whereas no such effect is observed in PARN wild-type iPS cells (n = 1 for each case) (right two bars). Data are downloaded from “GSE81507.” Data used for the figure is available in Source Data.

Fig. 3. EXTEND scores in cancer.

a EXTEND scores for TERT-expressing tumors and normal samples across 16 TCGA cancer cohorts. Violin width indicates data point density. Middle bar, median. b EXTEND scores across tumor stages (CRC = 338, HNSC = 420, KIRC = 375, KIRP = 254, LUAD = 497, STAD = 374, and THCA = 499 cases in total). Only tumor types with a minimum of ten cases for each stage were used in this analysis. P values were calculated using two-sided t test. *P < 0.05; **P < 0.01. Box indicates 25–75th quantile range. Middle bar indicates median. The minima and maxima are within 1.5 times the interquartile range of the lower and higher quartile. c EXTEND scores of cutaneous melanomas (TCGA SKCM). Violin width indicates data point density. Middle bar, median; two-sided t test. *P < 0.05; **P < 0.01. d Hazard ratio plot based on univariate Cox regression model for 31 cancer types. Vertical bars indicate upper and lower limits of the 95% confidence interval of the hazard ratio estimates. P values are highlighted for significant cases only (*P < 0.05; **P < 0.01). The y-axis is in the natural log scale. e Correlations between EXTEND scores and gene expression of ten oncogenic signaling pathways across 31 cancer types. Significant correlations (FDR < 0.05) are shown in either red (positive correlation) or blue circles (negative correlation). The frequency of each gene’s positive and negative correlation patterns across different cancer types is summarized on the right. Data used for the figure is available in Source Data.

Fig. 4. Association between EXTEND score, cancer stemness, and proliferation.

a Correlation between telomerase activity and cancer stemness across TCGA cohorts. Node size is proportional to correlation coefficients. In all cancer types, tumor stemness and telomerase activity is significantly correlated (FDR < 0.05). Shade indicates 95% CI of the regression. Spearman’s correlation was used to calculate Rho and P values for all plots unless otherwise stated. b, c Correlation between EXTEND score and cancer stemness at single-cell level in (b) glioblastoma and (c) head and neck cancer. Each dot represents one cell. d Top ten pathways enriched in the high stemness, high telomerase cells from head and neck cancer. P values are calculated using two-sided t test. Box indicates 25–75th quantile range. Middle bar indicates median. The minima and maxima are within 1.5 times the interquartile range of the lower and higher quartile. e, f EXTEND scores of cycling cells (G1–S phase and G2–M phase) and non-cycling cells in (e) glioblastoma and (f) head and neck cancer. g Correlation between EXTEND scores and proliferation marker MKI67 expression across 31 cancer types. This correlation is significant in all but two cancer types (FDR < 0.05 in blue). Shade indicates 95% CI of the regression similar to (a).