Long-Read Amplicon Sequencing for the Detection of TERT Promoter Variant Clonal Hematopoiesis in Patients with Telomere Biology Disorders

Abstract

Acquired TERT promoter (TERTp) variants are found in the blood of patients with telomere biology disorders (TBDs) and carry diagnostic and prognostic significance. Detection of these variants is challenging due to low variant allele frequencies (VAFs) and high GC content. We tested the sensitivity of long-read amplicon sequencing with deepSNV analysis, referred to as LR-deep AmpSeq, for TERTp variant detection. Among 47 patients with TBD features, we achieved an average depth of coverage of 7,943x and detected seven TERTp variants in six individuals (13%) with VAFs ranging from 0.006-0.33. Our results demonstrate that LR-deep AmpSeq is a sensitive, cost-effective method to detect low VAF TERTp variants.

Introduction

Acquired variants in the promoter region of TERT (TERTp) are among the most frequent mutations found in solid cancers,^1-5 recurring at “hotspots” located at the −57, −124, and −146 base pair (bp) positions from the transcriptional start site.^{6, 7} These variants activate telomerase expression in cells where it is normally silenced, enhancing proliferative capacity.^8-10

Several groups have identified acquired TERTp variants in blood cells of patients with telomere biology disorders (TBDs),^11-16 which arise through a process called clonal hematopoiesis (CH). Patients with a TBD have germline defects in telomere maintenance genes which lead to critically shortened telomere lengths and/or telomere dysfunction, resulting in premature cellular senescence. It is therefore not surprising that highly proliferative tissues, such as bone marrow, would benefit from acquiring a TERTp variant. Indeed, studies have identified TERTp variants more frequently in patients with TBD who do not develop myelodysplastic syndrome (MDS) or acute myeloid leukemia (AML),^{12, 14, 15} suggesting a protective effect on blood cancer risk. However, the presence of a TERTp variant has also been associated with adverse outcomes in patients with TBD such as an increased risk of multiorgan involvement and decreased overall survival.¹⁵

Identification of TERTp variants can also aid in TBD diagnosis. These variants are rare among patients with other hematologic disorders, including other bone marrow failure syndromes, MDS, myeloproliferative neoplasms, and AML.^{13, 17} These variants are also rare in the general population with a large-scale UK biobank-based CH study identifying a TERTp hotspot variant in only 148 of 489,545 (0.03%) individuals.¹⁸ Interestingly, these individuals had significantly shorter genetically predicted leukocyte telomere length compared to those without a TERTp variant.

The estimated prevalence of TERTp variant CH in patients with TBD ranges from 5-35%^11-16 with the variation partially attributed to the sensitivity and specificity of the detection methods used. TERTp variants are challenging to detect using first- and second- generation sequencing methods because: (1) they often occur at low variant allele frequencies (VAFs), even below 1% and (2) the TERTp region has very high GC content (>80% G+C), posing a challenge for PCR amplification and short-read sequencing technologies, which suffer from biases in GC-rich regions leading to uneven coverage and reduced sensitivity.^{19, 20} To address these challenges, we developed an Oxford nanopore technologies (ONT) long-read amplicon sequencing pipeline combined with deepSNV analysis²¹ for detecting low VAF acquired variants in the TERTp region. We demonstrate that this method is a cost-effective and sensitive alternative targeted approach for variant detection in this critical cancer- and TBD-associated region.

Materials and Methods

Patient Population

The University of Wisconsin-Madison Hereditary Hematology Registry was searched to identify all consented subjects who met the following criteria: (1) Clinically evaluated for a TBD including a comprehensive medical evaluation and family history, peripheral blood lymphocyte telomere length measurement by Flow-FISH, and germline genetic testing of known TBD genes; (2) Meeting clinical TBD diagnosis criteria based on clinical phenotype plus the presence of one or more pathogenic or likely pathogenic (P/LP) germline genetic variant(s) on a clinical genetic test in a TBD gene and/or critically short lymphocyte telomere lengths (defined as <1^st percentile for age) as measured by Flow-FISH; and (3) Having an available DNA sample from peripheral blood (PB) or bone marrow (BM). Individuals with TBD phenotypic features and a rare telomere gene variant of uncertain significance (VUS) without diagnostic telomere lengths were also included. The UW-Madison Institutional Review Board approved this study.

Sample Preparation and PCR

DNA from cell lines or patient PB or BM samples was extracted using the Maxwell RSC Whole Blood DNA Extraction Kit and PCR-amplified using primers flanking the TERTp to generate a 474 bp amplicon (Forward Primer: 5’-ACGAACGTGGCCAGCGGCAG-3’; Reverse Primer: 5’-CTGGCGTCCCTGCACCCTGG-3’).²² PCR using Takara LA Taq and a GC Rich Buffer in a 50 μL reaction was performed with the following conditions: initial denaturation at 94°C for 1 minute, 30 cycles of denaturation at 94°C for 30 seconds, annealing at 60°C for 30 seconds, elongation at 72°C for 2 minutes, final elongation of 72°C for 5 minutes, and a 4°C hold. PCR products were run on a 2% agarose gel for visualization. Samples were run in duplicates, and the resulting PCR products were pooled for PCR clean-up which included an isopropanol wash step to increase purity and DNA concentration (ThermoScientific GeneJET PCR Purification Kit). DNA concentration was determined via QuBit (1x dsDNA High Sensitivity). Samples were diluted to comparable amounts targeting 30 ng/μL to achieve optimal read depth. A known negative control (healthy subject DNA with no hotspot variant) PCR product was included with each sequencing batch.

Dilution Series

TERTp PCR product DNAs generated from the above protocol from a known negative control (human PB) and positive control (TPC1 thyroid cancer cell line, which is heterozygous for the −124C>T bp hotspot variant) were utilized to generate a dilution series. Dilutions were prepared in quadruplicate to generate samples with a −124C>T bp variant at VAFs of 0.50, 0.25, 0.10, 0.05, 0.025, 0.01, and 0.005 in a final volume of 20 μL. Samples were submitted to the UW-Madison Biotechnology Center (UWBC) for long-read ONT sequencing (LRS) along with a negative control (VAF of 0).

Long-Read Sequencing

Approximately 20 ng of PCR product was tagmented using the ONT Rapid Barcoding Kit (SQK-RBK114-96), followed by LRS on a PromethION 24 instrument using a PRO114M Flow Cell with established UWBC protocols. Reads were base called on-instrument using MinKNOW (v23.11.7) and bundled with Dorado (v7.2.13) using the SUP model (dna_r10.4.1_e8.2_400bps_5khz). FASTQ files were demultiplexed using Dorado (v7.2.13) with subsequent removal of barcode, adapter, and chimeric sequences and filtered for a minimum read quality of 10. Reads were aligned to a 474 bp portion of Homo sapiens TERT 5' regulatory region (LOC110806263) located on chromosome 5 (hg38 coordinates: 1294907-1295380) using Minimap2 (v2.17-r941). Variants were called using deepSNV²¹ (v1.48), which is designed to detect SNVs with VAFs as low as 1×10⁻⁴. Aligned reads were manually visualized using IGV (v.2.16.2) (reference genome = hg38 including per-sample alignment (BAM) and variant (VCF) tracks to inspect variant calls and allele depths at the three recurrent TERT promoter variant sites (hg38: chr5:1295113G>A for “C228T,” chr5:1295135G>A for “C250T”, and chr5:1,295,046T>G for “A161C”)).

Statistics

Statistical analyses were performed using GraphPad Prism 10. Fisher’s exact tests were used to compare the frequency of TERTp variants in different subgroups with p < .05 set as the threshold for statistical significance.

Results

LR-deep AmpSeq Pipeline Development and Validation

To develop a pipeline that is sensitive enough to detect variants occurring at extremely low VAF in the GC-rich TERTp region, we established an ONT long-read amplicon sequencing workflow that incorporates deepSNV.²¹ deepSNV is an acquired variant calling computational tool originally developed and used for short-read sequencing pipelines that uses a matched negative control for each test sample as an error correction method, helping to differentiate a low VAF acquired variant from background noise. Our LRS pipeline has been tailored to incorporate this computational tool for variant detection (hereafter called LR-deep AmpSeq) (Figure 1A). To determine the accuracy and VAF limit of detection of our method, a dilution series was conducted (Figure 1B-E). DNAs from the TPC1 human thyroid cancer cell line, which harbors a −124C>T bp hotspot TERTp variant in a heterozygous state, and from a healthy control subject with no TERTp variant (i.e., the required matched negative control) were PCR amplified to yield a 474 bp TERTp region amplicon. PCR products were quantified and mixed at different ratios to yield expected VAFs ranging from 0.005-0.50. These samples then underwent LR-deep AmpSeq in quadruplicates, yielding an average read depth of 3,933x with uniform coverage across the amplicon (Figure 1B) and no evidence of strand bias (Figure 1C). After deepSNV analysis, the −124C>T variant was reliably detected in all four replicates at VAF of 0.01 and higher and in two of four samples at 0.005 (Figure 1D, E, Supplemental Table 1).

Figure 1. Validation of LR-deep AmpSeq pipeline for TERTp variant detection.

A. Schematic of LR-deep AmpSeq pipeline using Oxford Nanopore Technology (ONT). Made with Biorender. B. Representative plot demonstrating filtered alignment coverage based on strand. Arrows indicate positions of TERTp hotspot variants. C. Representative plot demonstrating coverage across GC-rich regions of the amplicon. Arrows indicate positions of TERTp hotspot variants. D. Representative output plot from deepSNV analysis. A test sample (y-axis) is plotted against a known negative control sample (x-axis) with colored dots corresponding to statistically significant single nucleotide variants. E. Dilution series results. Samples with VAFs ranging from 0.5 to 50% were generated and submitted for LRS in quadruplicates. The VAF outputs from the LRS pipeline (y-axis) are plotted against the expected VAFs (x-axis). Replicates are plotted with averages and SEM. The dilution series was performed three times with a representative run shown.

LR-deep AmpSeq Hotspot CH Variant Detection in Patients with a Telomere Biology Disorder

We next tested the ability of this pipeline to detect TERTp variants in patients with TBD features, a group highly enriched for TERTp CH in their blood cells.^11-16 In total, 47 patients met inclusion criteria (Supplemental Table 2), including five sets of biological relatives (n=11 individuals). Median patient age was 58 years (range 24-77 years). Eighteen (38%) were female. Forty-one patients met TBD clinical diagnostic criteria based on lymphocyte telomere lengths <1st percentile by Flow-FISH (n=31) and/or from identification of a P/LP germline genetic variant (n=23). Six patients had TBD phenotypes and a rare TBD-associated variant of uncertain significance (VUS) with non-diagnostic telomere lengths. TBD phenotypes included: interstitial lung disease (ILD) (n=35), cytopenias/bone marrow failure/macrocytosis (n=21), myelodysplastic syndrome/hematologic malignancies (n=6), and liver disease (n=6). Germline variants (P/LP/VUS) included ACD (n=2), PARN (n=4), POT1 (n=2), RTEL1 (n=19), TERC (n=2), TERT (n=15), and TINF2 (n=1). Seven patients had two germline variants identified with either a P/LP variant and a VUS (n=2) or two VUSs (n=5). Six patients had prior clinical genetic testing for CH using a hybrid capture short-read sequencing platform which included the TERTp region (Tempus, Inc) (Supplemental Table 2). This clinical testing identified a TERTp variant at the −124 bp site for two patients (24.1 and 34.1) (Table 1) and did not identify any TERTp variants in four others (38.1, 41.1, 41.31, 43.1), allowing these cases to serve as a comparison between LR-deep AmpSeq and an established, commercial short-read sequencing-based method.

Table 1.

Long-read amplicon sequencing identified seven TERTp variants in six individuals with TBD features.

	Germline Mutation		Patient Demographics & Diagnosis Status				Known Positive Sample (Clinical Testing through Tempus)*		Initial LR-Deep AmpSeq		Confirmatory LR-Deep AmpSeq
ID	Germline Variant (p., c.)	Variant Class	Sex	Age	Phenotype	TBD Criteria Met	Variant Detected	Reported VAF	Variant Detected	VAF (# variant reads/total reads)	Variant Detected	VAF (# variant reads/total reads)
24.1	TERC noncoding r.114_115delTT	LP	M	44	ILD, BMF	YES: P/LP Variant	−124C>T	0.037	−124C>T	0.071 (625/8445)	-	-
34.1	RTEL1 p.Arg1264His c.3791G>A	P	M	66	ILD, macrocytosis	YES: P/LP Variant AND TLs <1st %ile	−124C>T	0.262	−124C>T	0.334 (1859/5514)	-	-
29.1	TERT p.Val777Met c.2329G>A	LP	M	53	ILD, thrombocytopenia, macrocytosis, early gray	YES: P/LP Variant AND TLs <1st %ile	None detected*		−124C>T	0.0142 (118/6794)	−124C>T	0.0190 (80/3569)
27.1	PARN p.Ile415Thr c.1244T>C	VUS	M	59	ILD, thrombocytopenia	YES: TL <1st %ile	-	-	−124C>T	0.0061 (39/4122)	−124C>T	0.00652 (119/11965)
									−146C>T	0.237 (986/4093)	−146C>T	0.268 (3332/12349)
37.1	RTEL1 p.Arg754Gln c.2261G>A	VUS	F	61	ILD	YES: TL <1st %ile	-	-	−124C>T	0.0164 (146/7448)	−124C>T	0.0198 (383/16,511)
	TERT p.Arg622His c.1865G>A	VUS
2.1	RTEL1 p.? c.2214-7C>G	VUS	F	58	ILD, BMF, MDS	NO	-	-	−124C>T	0.0118 (212/15589)	−124C>T	0.00876 (42/3448)

Abbreviations: VAF-variant allele frequency; ILD-interstitial lung disease; MDS-myelodysplastic syndrome; BMF-bone marrow failure’ P-pathogenic; LP-likely pathogenic; NA-not available; VUS-variant of uncertain significance; TBD-telomere biology disorder; TL-telomere lengths

^*The limit of detection for this clinical assay is 5%, although molecular pathologists can report variants at VAFs <5% at their discretion based on data quality and metrics.

LR-deep AmpSeq was performed using PB (n=45) and BM (n=2) DNAs and achieved an average read depth at the TERTp region (hg38 chr5:1,294,981-1,295,189) of 7,943x (Figure 2A). Overall, a TERTp variant was found in 12.8% (6/47 patients), including our two positive control patients with known hotspot variants (Table 1). Variants were validated by conducting a second round of LR-deep AmpSeq, which detected the hotspot variants at comparable VAFs (Figure 2B and Supplemental Table 3). Six of seven variants occurred at the −124 bp site. One individual had variants at both the −124 and −146 bp positions. These two variants were mutually exclusive, occurring on separate reads, suggesting they occur in distinct clonal populations. No variants were detected at the −57 bp site. VAFs ranged from 0.0061 to 0.33 with four of seven variants occurring at a VAF < 0.02. By age, TERTp variants were found in 7% (1/14) of those under age 50 versus 15% (5/33) of patients age 50+ years (p = .65). By diagnosis subcategories, TERTp variants were found in 12.2% (5/41) of those meeting and 16.7% (1/6) of those not meeting a clinical TBD diagnosis (Figure 2C, Supplemental Table 2; p = .99). In those meeting a clinical diagnosis, TERTp variant proportions did not differ by germline genetic variant status (2/23, 8.7%; versus 3/9, 33%; versus 0/9, 0%; in those with a P/LP, VUS, or no variant identified, respectively; p > .05 for all pairwise comparisons) or by telomere length (4/31, 13% in those <1^st percentile versus 2/16, 10% in those >1^st percentile; p = .99). By phenotype, 5/35 (14%) patients with ILD and 5/21 (24%) with cytopenias, macrocytosis or BMF had TERTp CH (p = .47) (Figure 2C). TERTp variants were observed in individuals with germline variants in PARN (n=1/4; 25%), RTEL1 (n=2/14; 14%), TERC (n=1/2; 50%), and TERT (n=2/14; 14%). Of these, 4/6 (66%) were in individuals whose TBD gene variant was considered a VUS.

Figure 2. LR-deep AmpSeq Performance and Yield for TERTp detection in patients with Telomere Biology Disorders.

A. Plot of average read depth at TERTp loci (hg38 chr5:1,294,981-1,295,189) from patient samples and controls. Red open circles indicate patient samples that were repeated due to low DNA concentrations and low read depths. Red solid circles indicate the improved read depths after repeating the PCR and generating a DNA concentration of 30 ng/μL for long-read sequencing. B. Solid circles represent the initial LR-deep AmpSeq assay per sample. Open triangles correspond to VAFs detected through clinical CH testing (positive controls). Open circles represent the second round of LR-deep AmpSeq assay per sample. Black represents the −124C>T hotspot variant; blue represents the −146C>T hotspot variant. C. TERTp variants detected in TBD patients by diagnostic, genetic, and phenotypic subtypes. Percentages indicate the percent in each category with a TERTp mutation detected. For the phenotypic categories, ILD=interstitial lung disease and Heme=cytopenias, bone marrow failure, and/or macrocytosis.

LR-deep AmpSeq Common Variant Detection in Patients with a Telomere Biology Disorder

LR-deep AmpSeq enables detection of other variants within the amplified region, including a common SNP which occurs at the −245 bp position (hg38: chr5:1295234A>G). The minor allele of this common SNP was present in 22 individuals, of which 17 were heterozygous and five were homozygous. Of patients with TERTp variants detected, three individuals were heterozygous and one was homozygous for the minor allele (Supplemental Table 4).

Discussion

Acquired variant detection in PB and BM is becoming increasingly critical for diagnosis, prognosis, and monitoring for disease progression in numerous hereditary and acquired blood disorders. GC-rich regions, such as the TERT promoter, present challenges to standard short-read sequencing pipelines and are often not included on standard myeloid malignancy acquired mutation panels yet are critical to detect. To overcome these challenges, we developed a long-read amplicon-based sequencing approach that implements deepSNV, a low VAF variant calling tool previously only used in short-read sequencing pipelines. We show that this method achieves a high depth of coverage and reliably detects low VAF CH variants in patients with TBD features.

Although other targeted approaches exist for detecting variants in this region, our ONT LRS-based method offers key advantages over prior approaches. First, generating contiguous reads across full-length targets reduces mapping ambiguity and enhances the sensitivity needed for detecting variants present at very low VAFs. Second, LRS is less susceptible to GC bias compared to short-read sequencing platforms.¹⁹ Third, LRS of amplicons generated by GC-rich optimized PCR with primers that target the TERTp region as opposed to native strand LRS enriches for fragments of interest to improve efficiency and reduce costs. Specifically in comparison to digital PCR, LR-deep AmpSeq generates sequence-specific information across the entire amplicon, allowing detection of all three hotspot variants as well as other variants in this region without requiring a multiplexed assay. Further, this method involves a straightforward sample preparation, requiring only DNA extraction, a single PCR amplification reaction, and tagmentation for LRS. This method also offers a low per-sample sequencing cost (approximately $20 USD exclusive of local staffing and overhead costs) with rapid turnaround time as short as one to several days depending on local infrastructure, making LR-deep AmpSeq a highly accessible approach well suited to clinical translation.

Using this method, we identified TERTp CH at a frequency comparable to previous studies. We detected TERTp variants in 12.2% of individuals meeting TBD diagnostic criteria, while previous studies have reported rates ranging from 5-35%.^11-16 The variation between studies can be partially attributed to differences in sensitivity of sequencing methods used as well as differences in the specific TBD populations studied. Sanger sequencing identified the lowest prevalence of 5% in a TBD ILD cohort,¹¹ HaloPlex targeted short-read sequencing achieved a prevalence of 19% in a cohort of TBD patients with and without hematologic malignancies,¹² and droplet digital PCR identified the highest prevalence of 35% in a TBD lung transplant population (Supplemental Table 5). Most prior studies primarily used short-read sequencing pipelines that achieved an average read depth at the TERTp region of 100-2963x but often required multiple PCR and amplicon enrichment steps prior to sequencing (Supplemental Table 5). Our method yielded an average read depth of 7,943x, enabling us to detect VAFs as low as 0.006, while previous studies had a lowest detected VAF ranging from 0.008-0.05.^11-16

TERTp hotspot variant CH detection is increasingly important in patients with TBD. Several studies in TBD patient populations have suggested these variants may have a protective effect on risk of MDS and AML, finding that fewer patients with MDS/AML than those without MDS/AML have a TERTp variant.^{12, 14, 15} However, these variants have also been associated with a higher risk of multiorgan involvement and decreased overall survival.¹⁵ Reliable detection of these variants in larger cohorts of patients with TBD across the phenotypic and age spectrum will be essential for clarifying their independent prognostic significance and whether their presence in blood, bone marrow, or other tissues warrants action.

Diagnostic criteria that reliably differentiate adults with and without a TBD are currently lacking. Because TERTp CH has recently been shown to be rare in the general population¹⁸ and at the same time is significantly enriched in patients with various TBD phenotypes in our study and others,^11-16 these data suggest that TERTp CH can serve as supportive molecular evidence for a TBD diagnosis. This additional evidence is especially useful when telomere lengths are non-diagnostic, as they often are in adults with TBD phenotypes,²³ and there is no clear, causative P/LP genetic variant identified. Our study detecting TERTp CH in 1 of 6 (17%) individuals with strong TBD features who lack the molecular evidence needed for diagnosis fits well with similar findings in prior studies,^{11, 15} strengthening the rationale for incorporating TERTp CH into TBD diagnostic criteria. Given that TBD-specific management guidelines exist and improve outcomes in urgent scenarios such as lung and BM transplantation for this patient population,²⁴ the ability to quickly and reliably detect TERTp variants to aid in diagnosis and for long-term monitoring is essential.

Additionally, LR-deep AmpSeq can detect any variant in the amplified and sequenced region. We identified the minor allele of the common SNP at −245T>C in 22 of our patients with TBD features. Several studies have reported variable associations between cancer risk or cancer outcomes and this SNP.^25-28 In the context of hematologic malignancies, the −245T>C SNP has been shown to have a negative effect on overall survival in AML.²⁹ In several solid cancers, the combination of the −245T>C SNP and a TERTp hotspot variant has been shown to increase mortality and cancer recurrence.^{30, 31} However, the effect of a −245T>C SNP alone or in combination with a TERTp hotspot variant in patients with TBD has yet to be elucidated. LR-deep AmpSeq is therefore a powerful tool to address these questions, as it enables detection and phasing of the −245T>C SNP and of all three hotspot variants found in the TERTp region.

In conclusion, this study demonstrates that integrating deepSNV with long-read amplicon sequencing enables reliable detection of common and acquired TERTp variants at VAF as low as 0.006 in PB or BM. LR-deep AmpSeq effectively addresses the challenges faced by short-read sequencing-based approaches and offers a fast, sensitive, and cost-effective alternative method for this challenging GC-rich region. It holds promise for broader applications in detecting low VAF variants in blood or other tissues beyond the TBD population and the TERTp region.

Key Points.

• LR-deep AmpSeq is a long-read amplicon sequencing pipeline that incorporates deepSNV analysis for ultra-low frequency acquired variant detection.

• In the TERT promoter region, which is >80% GC, LR-deep AmpSeq achieved an average read depth of 7,943x and detected hotspot acquired variants with variant allele frequencies as low as 0.006.

• LR-deep AmpSeq is a rapid, sensitive, and cost-effective approach to detect low allelic fraction variants in the TERT promoter that can be adapted to other GC-rich genomic regions.