Short NK- and Naïve T-Cell Telomere Length Is Associated with Thyroid Cancer in Childhood Cancer Survivors: A Report from the Childhood Cancer Survivor Study

Subsequent malignant neoplasms (SMN) of the thyroid are an established late effect of childhood cancer treatment (1), and are more common among females and children diagnosed at a younger age (2). Radiation-exposed survivors are at especially high risk for SMN (3). However, not all survivors with at-risk exposures develop thyroid SMN, suggesting a role for genetic contributors. Telomeres are the repetitive DNA-protein structures at chromosome ends that protect chromosome integrity. Telomere length (TL) is genetically determined and further moderated by endogenous and exogenous (e.g., environmental) factors. TL is maintained in cell populations that express sufficient telomerase, but shortens with chronologic age in hematopoietic cells. Hematopoietic cells are highly sensitive to reductions in telomerase, so that a pathogenic variant in telomerase reverse transcriptase (TERT) and/or associated proteins produces progressive pronounced leukocyte telomere shortening over successive generations (4).

In a meta-analysis of population studies, short leukocyte TL (LTL) was associated with gastrointestinal and head and neck cancers (5). Most population-based studies rely on high-throughput techniques that require a minimal amount of DNA, such as qPCR (6). This technique is performed on the diverse peripheral blood mononuclear cell fraction to determine the “average” quantity of telomere DNA in leukocytes (telomere content), but does not account for differential shortening in leukocyte subsets. Cancer chemotherapy and radiation may further accelerate leukocyte telomere shortening via various pathways that include a heightened proliferative demand for hematopoietic reconstitution and direct damage to telomere DNA (7, 8), although rare studies have found an opposite, lengthening effect after chemotherapy alone (9). In a recent study conducted in the St. Jude Lifetime Cohort of childhood cancer survivors, shorter LTL was associated with an increased prevalence of chronic health conditions (10). We previously conducted a matched case–control study in the Childhood Cancer Survivor Study (CCSS), which identified less telomere content measured by qPCR in whole blood samples among radiation-exposed survivors with thyroid SMN, compared with radiation-exposed survivors without SMN (11). Subsequent studies, also conducted in CCSS, indicated no association between genotypically-estimated LTL and thyroid SMN (12), and an association between thyroid SMN and a common, low frequency POT1 variant (13). This study builds on our prior work, with the objective of assessing absolute TL in leukocyte subsets relative to normative data among survivors with and without thyroid SMN.

This nested case–control study was conducted in the CCSS, a multicenter, retrospective cohort with longitudinal follow-up of 5-year survivors of childhood cancer diagnosed between 1970 and 1986 (14). All participants to this study provided informed consent, and the study was approved at each participating institution by the local institutional review board. Blood samples were collected from consenting participants at a minimum of 15 years off therapy and banked for future use. Radiation-exposed survivors who developed thyroid SMN (cases) were ascertained through self-report questionnaires. SMNs were validated from pathology reports, medical records, or the National Death Index and confirmed by a CCSS pathologist. Survivors with thyroid SMN were matched 1:1 with survivors that did not have SMN (controls) by primary cancer diagnosis, chemotherapy (yes/no), radiation field (yes/no for brain/neck, chest/spine, abdomen/pelvis), and follow-up duration (exceeding time to SMN for the case) by incidence-density sampling. Stem cell transplant (SCT) recipients were excluded. The sample size was determined by the number of CCSS cases with available viable leukocyte samples and without history of SCT, and represents a subset of thyroid SMN cases included in our prior matched case-control study (which only required DNA; ref. 11). The study was conducted in accordance with the Declarations of Helsinki. All subjects provided written consent to participate in CCSS, and each participating institution obtained approval to conduct this research through their respective institutional review board.

Median LTL was measured in duplicate for total lymphocytes, and in B cells, memory T cells, naïve T cells, and NK cells by telomere flow cytometry FISH (telomere flow-FISH), using established procedures and with quality control measures as described previously (Repeat Diagnostics; ref. 15). Briefly, LTL was assessed by denaturation in formamide at 87°C and quantitative hybridization with a Alexa Fluor 488-conjugated (CCCTAA)₃ peptide nucleic acid (PNA) probe specific for telomere repeats (Panagene) with a bovine thymocyte quantitative reproducibility control. Flow cytometry was used to define leukocyte subsets by a gating strategy that first distinguished lymphocytes based on LDS751 DNA staining (Thermo Fisher Scientific) and forward and light scatter properties, and then further distinguished naïve T cells, memory T cells, B cells, and NK cells based on staining with antihuman CD45RA, CD20, and CD57 (15). Median absolute LTL was determined for each subject for total lymphocytes and for each leukocyte subset. Age- and sex-adjusted LTL percentiles for both total lymphocytes and for subsets were derived from normative data obtained from 400 healthy volunteers between the ages of 0 to 100 years (16).

To further characterize the CD20⁻/CD57⁻/CD45RA⁺ cell population identified as naïve T cells by telomere flow-FISH, a subset of viably frozen samples from case–control pairs were analyzed locally on a BD LSR-II flow cytometer to identify CD3 positive T cells that coexpressed CD45RA and CD197 (CCR7) using the following antibodies: CD45RA-FITC, CD3 PerCP-Cy5.5, CD45 APC-H7, and CD197-V450, per manufacturer's recommendation (BD Biosciences).

Genomic DNA was extracted from saliva and/or leukocytes isolated by the Ficoll method, using the appropriate Qiagen Kit. WGS was performed by the Baylor College of Medicine Human Genome Sequencing Center. Five hundred nanograms of subject DNA was prepared and sequenced on an Illumina Novaseq 6000 S4 platform with 30x coverage. Variant call format (VCF) specifications were generated from the Genome Reference Consortium Human Build 38 (GRCh38), trimmed, and filtered using VCFtools (17) to remove variants that (i) mapped to genomic regions outside chromosomes 1 to 22, X, and/or Y, (ii) failed standard quality control, (iii) had an allele fraction <0.25 and/or Q fraction <10. The filtered data were analyzed using Ensembl Variant Effect Predictor (VEP) v97.3 (18) with the dbNSFP plugin v2.9.1 to extract exome population minor allele frequency (MAF). Variants analyzed in VEP were then further annotated for pathogenicity in ClinVar (version 201904; ref. 19). Variants were included if they were designated as “pathogenic” or “likely pathogenic,” or if they had a MAF <1% in the Genome Aggregation Database (gnomAD) r2.1 (exome; ref. 20) and/or the 1,000 Genomes Project (21).

The gnomAD MAF of non-Finnish European, African, and East Asian/South Asian were used to identify rare variants in CEU, AFR, and ASN samples, respectively. Custom shell and R scripts were generated to process the results and calculate the unique gene and variant frequencies in the output files. Unique variants were determined on the basis of reference sequence identifiers. We then further restricted our analyses to rare and/or pathogenic variants in 14 genes previously implicated in telomere biology disorders: ACD, CTC1, DKC1, NAF1, NHP2, NOP10, PARN, POT1, RTEL1, STN1, TERC, TERT, TINF2, and WRAP53 (22), including within a 5kbp flanking region. Sequence ontology was used to capture protein altering variants, including missense, in-frame insertion/deletion (indel), and frameshift variants.

Median LTL, transformed to sex- and age-adjusted percentiles based on the age at the time of sample collection, was the primary predictor considered in the analysis. The primary outcome was thyroid SMN. A paired t test was used to compare median LTL between cases and controls for each leukocyte subset. McNemar test with continuity correction was used to compare proportions of survivor pairs with very low (VL, <1^st percentile) or low (L, ≥1^st to <10^th percentile) TL (exposed) versus normal, high, or very high TL (unexposed). In consideration of the matched design, ORs for LTL <10^th percentile were calculated from the ratio of discordant pairs for each leukocyte subset. A Wilcoxon matched pairs signed rank test was used to compare proportions of each leukocyte subset relative to all leukocytes. All comparisons were conducted using matched analyses. A P value <0.05 was considered statistically significant.

Of 110 CCSS participants who developed thyroid SMN, 52 met inclusion criteria, had an available biospecimen, and were subsequently matched with 52 controls. However, for six of these pairs either the case failed quality control (insufficient cell count or viability for flow-FISH), or the control failed quality control and there were no suitable replacements available. The remaining final set of 46 matched controls were comparable in characteristics, including chemotherapy exposure (P = 0.88), cancer diagnosis (matched), sex (P = 0.06), race (P = 0.30), age at diagnosis (0.35), and radiation exposure/site (matched), to the pool of potential controls that matched to cases, regardless of sample availability. Sufficient subject-level data were available to conduct all analyses. Four samples were collected prior to thyroid SMN diagnosis, and 42 after SMN diagnosis. Survivor pair characteristics, including age at sample collection, are shown in Table 1. Because chemotherapy matching was considered as a yes/no variable, nine pairs were discordant for exposure to alkylators (five pairs with only the control exposed, and four pairs with only the case exposed).

Telomere flow-FISH was performed for all 92 survivors. We were able to measure naïve T-cell and memory T-cell TL for all subjects, and B-cell and NK-cell TL for 91 and 84 subjects, respectively, due to limitations in either cell recovery or integrity. Figure 1 shows survivor sex/age-adjusted median LTL relative to population normative data indicating the 1st, 10th, 50th, 90th, and 99th TL percentiles. Survivor LTL was downward distributed relative to normative data across all leukocyte subsets, suggestive of shorter TL compared with population norms. For example, seven of 92 survivors (8%) had total lymphocyte TL <1st percentile, and 28/92 (30%) had total lymphocyte TL <10th percentile. Figure 2 illustrates the deviation of survivor median LTL below the population median, by SMN status. Median LTL was not associated with primary diagnosis, age at primary diagnosis, time between diagnosis and blood sample, educational achievement, or income.

The distribution of differences in survivor median LTL relative to the population median supported analysis of these data using parametric tests for all subsets (Supplementary Fig. S1). When comparing differences in median LTL between cases and controls, cases had a greater negative deviation below the population median than controls. This difference reached statistical significance for total lymphocytes and for three out of four leukocyte subsets (Table 2; Fig. 2). We then compared proportions of survivor pairs that had one, both, or neither survivor with sex/age-adjusted LTL <10th percentile (Table 2). Eight survivors, all cases, had NK cell LTL <10th percentile (P = 0.01). Thirty-three out of 92 survivors, 22 cases and 11 controls, had naïve T-cell LTL <10th percentile. Cases were 2.83 times more likely to have naïve T-cell LTL <10th percentile than controls [confidence interval (CI), 1.06–8.78; P = 0.04). All cases with VL NK-cell TL had L or VL TL in all other subsets, as well as in total lymphocytes. All cases with L or VL NK-cell TL also had L or VL naïve T-cell TL.

Given that many blood samples were obtained from cases after the SMN diagnosis, a sensitivity analysis was conducted for total lymphocytes and subsets that were significant in our initial comparison, limited to case–control pairs in which the case's blood sample was obtained within 5 years of the SMN diagnosis (n = 23). Similar trends were observed, with the difference in NK-cell TL remaining significant (difference in median TL between cases and controls, P = 0.012; eight survivors, all cases, with NK-cell LTL <10th percentile, P = 0.013, Table 3).

Shorter LTL identified in naïve T cells by telomere flow-FISH (CD45RA⁺/CD20⁻) corresponded with a lower percentage of naïve T cells relative to all lymphocytes (cases: median 24%, interquartile range 19–32%, vs. controls: median 39%, interquartile range 30–50%, P = 0.002), suggestive of accelerated aging-related naïve T-cell LTL decline among cases. We conducted additional CCR7 immunophenotyping of the CD45RA⁺ cell population in a subset of eight matched survivor pairs with L and VL LTL (n = 16) to further distinguish naïve T cells from other potential CD45RA expressing cells, such as effector memory T cells that re-express CD45RA (TEMRA cells), which are characterized by very short telomeres (Supplementary Table S1; ref. 23). Out of all CD45RA⁺ T cells, approximately half were CCR7-defined naïve T cells across all subjects tested (mean 49%, median 47%). Cases had an equivalent proportion of CCR7-defined naïve T cells relative to controls (cases: mean 46%, median 47% vs. controls: mean 53%, median 50%), suggesting that a disproportionate admixture of CD45RA⁺ T cells was not a strong contributing factor to the differences in naïve T-cell TL observed between the groups (Fig. 3).

WGS data were available for 88 survivors, who were largely of CEU ancestry (85 CEU, 2 AFR, 1 ASN). WGS data identified coding sequence indel and missense variants with MAF < 1% in 23 survivors: 1 indel (in one case and one control) and 19 heterozygous missense variants (in 14 cases and 7 controls). None of the variants were designated as pathogenic. Seven out of 15 cases with indels/missense variants had L or VL TL in at least one of four subsets, compared with only one out of eight controls (Supplementary Table S2). Five out of 88 survivors sequenced had a rare missense or indel TERT variant (6%), comparable with the frequency of germline TERT variants we previously observed in a childhood AML cohort (24).

Adaptive and innate immunity decline with aging, due to a progressive reduction in naïve T cells, T-cell receptor diversity, and NK-cell function. In survivors of childhood cancer, the long-term impact of cancer treatment on the immune system is understudied (25). There is some evidence of impaired immune reconstitution following treatment, but studies are limited by small sample size and short follow-up duration (26–28). Exposure to chemotherapy in childhood depletes CD45RA⁺ naïve T cells during active therapy (29), an effect that extends to off-therapy survivors, for example, children ≥18 months following leukemia therapy evidenced ongoing naïve T-cell depletion and quantitative deficiencies in NK cells (30). In that study, in vitro gene expression response to radiation also showed a reduction in DNA damage response compared with sibling controls. Further, there is strong evidence that exposure to ionizing radiation accelerates thymic involution and aging, even after a low dose exposure (31). Together, these observations suggest that cancer treatment may induce defects in both adaptive and innate immunity, findings that at least persist into the early years after cancer treatment and coincide with an impaired DNA damage response. No studies to date have examined the long-term (15 years+) effects of childhood cancer therapy on immune function.

Our results suggest excess LTL shortening in 15+ year survivors of childhood cancer (mean age 39 years) that may be associated with the development of thyroid SMN. These observations are suggestive of premature immunosenescence, or, at minimum, a premature decline in LTL with age. Our findings are in line with prior reports of physical and molecular indicators of premature, accelerated aging in survivors of childhood cancer (32, 33), and expand on prior work evaluating TL associations in specific leukocyte subsets. Here, we show that survivors have shorter than expected LTL relative to sex/age-based norms, and nearly a third have TL at the <10th percentile. Cases had shorter mean LTL than matched controls in total lymphocytes, and in three out of four leukocyte subsets. Naïve T-cell and NK cell LTL <10th percentile were significantly associated with thyroid SMN.

NK cells play a key role in cancer immunosurveillance and cytotoxicity, particularly with respect to the control of hematologic malignancies and solid tumor metastases (34). Naïve T cells can also differentiate into various functional subsets with antitumor activity (35). Aging-related decline in NK-cell function, naïve T-cell number, and capacity for tumor neo-antigen expression contribute to the rising cancer incidence in the elderly (36, 37). In our study, naïve T-cell TL shortening corresponded with an overall decrease in the naïve T-cell population more pronounced among cases compared with controls. This difference did not appear to be related to differences in the CD45RA⁺/CD20⁻ cell admixture, as equivalent proportions in a subset of cases and controls expressed CCR7, a characteristic naïve T-cell surface marker. Our results suggest a pattern of naïve T-cell TL shortening and population decline in long-term cancer survivors, warranting further investigation of decline in adaptive and/or innate immunity in survivors relative to risk for SMN.

Prior studies detected rare variants in telomere biology disorder genes in individuals with thyroid cancer that impact LTL (38, 39). Our study was not designed to detect differences in the frequency of rare variants between survivor cases and controls, but the frequency of rare variants approximated the frequency observed in conditions within the telomere biology disorder spectrum. Of note, the TERT p.H412Y variant that occurred in two cases has been associated with telomere biology disorders (40), but is designated in ClinVar as benign. We and others have shown that p.H412Y telomerase activity and processivity is similar to wild type (24). However, the location of this variant in TERT may affect TERT-TERC binding, which would not be detected in activity and processivity assays and supports follow-up studies to assess the functional impact of this variant on telomere maintenance and extension.

Limitations to our study include lack of baseline LTL pre-primary cancer diagnosis and the inability to assess LTL over time, which would inform our understanding of LTL attrition postcancer therapy. Second, case/control matching was considered by chemotherapy yes/no rather than by individual agents, but in settings of lower radiation dose, exposure to alkylators may affect risk for thyroid SMN (41). However, only 4 of 46 pairs were discordant for alkylator exposure so that the case was exposed and the control was not. Third, only half of thyroid SMN cases had a biospecimen available, and the majority were obtained after the diagnosis of thyroid SMN. In our study, all survivors had shorter than expected LTL, likely related to the long-term impact of primary cancer treatment on telomere shortening over time. Regarding the shorter TL we observed among survivors with thyroid SMN, we cannot exclude the possibility of reverse causation, that is, that the SMN diagnosis led to the excess TL shortening observed in cases compared with controls. However, sensitivity analyses of these data limited to only those pairs with cases whose blood was drawn within 5 years of the SMN diagnosis showed similar trends as the larger sample set, retaining significance for the differences observed in NK-cell TL. Recognizing the limitations of biospecimens collected in CCSS (the majority of specimens obtained following the SMN diagnosis), for this study we deliberately limited our outcome comparison to thyroid SMN versus no SMN, given that the standard approach to thyroid cancer treatment is localized and often limited to surgical excision and/or radioiodine ablation, rather than systemic.

LTL is a convenient proxy for TL in other tissues, highly correlated between tissues obtained from the same individual (42, 43), and is associated with a variety of clinical outcomes. Our own data also support correlation of LTL between leukocyte subsets. In our study, LTL was not associated with primary diagnosis, age at diagnosis, time between diagnosis and blood sample, or demographic and socioeconomic factors. The degree of LTL shortening experienced by survivors may depend on LTL at baseline (prior to diagnosis) and vary by cellular response to intensive chemotherapy and radiation, among other factors. Prospective studies that include biobanking of leukocyte specimens are needed to further elucidate the relationship between LTL and risk for SMN and other late effects of primary cancer treatment. Our findings suggest that telomere flow-FISH may facilitate screening survivors at risk for thyroid SMN. The flow-FISH method has been adopted as the gold standard for clinical screening of telomere biology disorders. Specifically, lymphocyte TL <1st percentile has an 85% positive predictive value (PPV) for correctly diagnosing dyskeratosis congenita, a prototypic disorder of telomere biology (44). In our study, NK cell TL <10th percentile has a high specificity for detecting thyroid SMN. Considering a thyroid SMN prevalence of approximately 2% among survivors of childhood cancer (45), our data suggest a 100% PPV of NK cell TL <10th percentile detecting thyroid SMN, with a 98% negative predictive value (CI, 98.1–98.5). Replicating these findings in a larger cohort of survivors may support use of this test to inform clinical decision-making regarding the need for ultrasound-based screening and fine needle aspiration among radiation-exposed survivors with nodules detected on physical exam (46).

G. Aubert reports other support from Repeat Diagnostics Inc. (RDx) during the conduct of the study. W.M. Leisenring reports grants from NIH during the conduct of the study. L.L. Robison reports grants from NCI during the conduct of the study. G.T. Armstrong reports grants from NIH during the conduct of the study. No disclosures were reported by the other authors.

T.-K. Man: Resources, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. G. Aubert: Formal analysis, investigation, visualization, methodology, writing–review and editing. M.A. Richard: Formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. W. LeJeune: Investigation, methodology, writing–review and editing. E. Hariri: Investigation, methodology, writing–review and editing. T. Goltsova: Investigation, visualization, methodology, writing–review and editing. A. Gaikwad: Investigation, visualization, methodology, writing–review and editing. Y. Chen: Data curation, formal analysis, investigation, methodology, writing–review and editing. J. Whitton: Data curation, investigation, methodology, writing–review and editing. W.M. Leisenring: Resources, data curation, supervision, investigation, methodology, writing–review and editing. M.A. Arnold: Resources, data curation, investigation, methodology, writing–review and editing. J.P. Neglia: Resources, data curation, investigation, methodology, writing–review and editing. Y. Yasui: Resources, data curation, formal analysis, supervision, investigation, methodology, writing–review and editing. L.L. Robison: Resources, data curation, supervision, funding acquisition, investigation, methodology, writing–review and editing. G.T. Armstrong: Resources, data curation, supervision, funding acquisition, investigation, methodology, writing–review and editing. S. Bhatia: Conceptualization, resources, data curation, supervision, investigation, methodology, writing–review and editing. M.M. Gramatges: Conceptualization, supervision, funding acquisition, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

The authors would like to thank Dr. Alison Bertuch for her critical review of this manuscript, and the survivors and families who participate in the CCSS. This work was supported by the Texas Children's Cancer and Hematology Centers Flow Cytometry Core Laboratory, the Texas Children's Cancer and Hematology Centers Bioinformatics Core Laboratory, and the Baylor College of Medicine Human Genome Sequencing Center. This work was supported by the NCI at the NIH (R01 CA194473: M.M. Gramatges, Principal Investigator). CCSS was supported by the NCI at the NIH (U24 CA55727: G.T. Armstrong, Principal Investigator), the St. Jude Children's Research Hospital Cancer Center Support (CORE) grant (CA21765, C. Roberts, Principal Investigator), and the American Lebanese-Syrian Associated Charities (ALSAC).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.