Single-cell sequencing reveals shared clonal signatures in nonmalignant B and tumor cells in T-prolymphocytic leukemia

T-cell prolymphocytic leukemia (T-PLL) is an aggressive T-cell neoplasm that originates from mature T cells of post-thymic origin and is associated with an aggressive clinical course.¹ The annual incidence is 1 in 1 000 000 and, although rare, it remains 1 of the most common mature T-cell leukemias.¹ Patients with T-PLL usually present with widespread disease characterized by T-lymphocytosis, thrombocytopenia, anemia, and involvement of liver, spleen, lymph nodes and occasionally pleural effusions and skin. Treatment options remain limited for patients with T-PLL, however, most patients will initially achieve a complete response to single agent alemtuzumab, an anti-CD52 antibody (MabCampath). Unfortunately, responses can be short and many of these patients will eventually relapse and succumb to the disease.^2,3 That said, some patients exhibit prolonged survival without an immediate requirement for upfront treatment, however, active disease eventually emerges.^3,4 Of note, allogeneic hematopoietic stem cell transplantation (allo-HSCT) may yield durable remissions in suitable candidates.^5,6 

Aggressive high-grade T-cell neoplasias, such as T-PLL, have an inferior and dismal prognosis compared to high-grade B-cell malignancies. Recent advances in the development of targeted treatment options directed toward specific molecular pathways in the B-cell receptor signaling and NF-κB-signaling pathways have yielded successful outcomes.⁷ Mutations within the ataxia telangiectasia mutated (ATM) gene, located on chromosome 11, and the proto-oncogene T-cell leukemia/lymphoma 1A (TCL1A), located on chromosome 14, are putative main drivers of T-PLL development. In addition, genes involved in the Janus kinase (JAK) signal transducer and activator of transcription (STAT) pathways are recurrently mutated in T-PLL, becoming a secondary hallmark of the disease.⁸ Development of JAK/STAT inhibitors may result in new treatment options for T-PLL and lead to better outcomes in this aggressive disease.⁹ 

This study aimed to investigate the mutational profile of genes commonly involved in the leukemogenesis of T-PLL using targeted next generation sequencing (NGS) of paired samples obtained at diagnosis and relapse from patients diagnosed with T-PLL to elucidate the clonal origin and evolution of the disease.

Samples obtained from 16 patients (9 males and 7 females) diagnosed with T-PLL at the Department of Clinical Pathology, Uppsala University Hospital, Sweden over a period of 15 years (2002-2017) were included in the study. All diagnostic and relapse samples were re-evaluated by an experienced hematopathologist (R.-M.A.). Clinical information was retrieved from patient records and the study was approved by the Regional Ethical Committee (Dnr 2014/233) in accordance with the Declaration of Helsinki.

Bone marrow aspirates (n = 6) and peripheral blood (n = 10) samples were collected in heparinized tubes and multicolor flow cytometry analyses were performed as previously described.¹⁰ Flow cytometry was performed before magnetic bead separation to (1) identify the cell populations and (2) to determine the purity of malignant and nonmalignant cell fractions (supplemental Material).

Positive selection of CD19⁺ B cells from 14 of the 16 patients was performed using CD19 MicroBeads and Auto MACS (Miltenyi Biotech, Cologne, Germany), according to the manufacturer’s instructions, to obtain a germ line cell population. For the remaining 2 patients, germ line DNA was extracted from follow-up remission samples that demonstrated no evidence of disease. DNA extraction (tumor and germ line) was performed using the AllPrep DNA/RNA/microRNA KIT (Qiagen, Hilden, Germany). NGS was performed on genomic DNA (gDNA) extracted from blood or bone marrow. Sequencing libraries were prepared from 50 ng of gDNA using a customized HaloPlex high sensitivity (Agilent Technologies, San Diego, CA) gene panel and sequenced on the NextSeq instrument (Illumina, San Diego, CA). The panel contained 3729 unique amplicons covering either hotspots or the full coding sequence of 19 genes frequently mutated in T-PLL (ATM, BCOR, CHEK2, CREBP, EZH2, IL2RG, JAK1, JAK2, JAK3, STAT1, STAT2, STAT3, STAT4, STAT5A, STAT5B, STAT6, TET2, TP53, TYK2) (supplemental Table 1).

FASTQ demultiplexing and adapter removal were performed using bcl2fastq (version v2.20.0.422). Sequencing reads were aligned to the GRCh37 human reference genome using bwa-mem (version 0.7.16) and subsequently processed using Samtools (version v1.8). Pisces (version 5.2.10.49) was used for the detection of single nucleotide variants and small indels with the detection threshold set at 500× coverage, variant allele frequency (VAF) 5% and a minimum requirement of 100 reads supporting the variant. Variants were annotated using population variation databases and Cosmic (release v85) using VEP (version v91) and SnpEFF (version 4.3) were classified according to an in-house protocol as previously reported.^11-13 

Droplet digital polymerase chain reaction (ddPCR) was performed using the QX200 AutoDG Droplet Digital PCR system (Bio-Rad Laboratories, Hercules, CA) as previously described.^14,15 In short, 60 ng of gDNA was mixed with 1× ddPCR Supermix for probes (no deoxyuridine triphosphate (dUTP), 900 nM of each primer and 250 nM of each probe (FAM-labeled or HEXA-labeled). All consumables and reagents for ddPCR were purchased from Bio-Rad. Droplets were created using the AutoDG Droplet Generator, followed by PCR, adhering to the manufacturer-recommended cycling protocol. Droplets were subsequently read using a QX200 droplet reader. Analysis of the ddPCR data was performed using the QuantaSoft Analysis Pro software (Bio-Rad).

Single-cell analysis of gDNA combined with cell surface protein markers was performed on samples obtained from 2 patients, using the Tapestri Platform from Mission Bio (San Fransisco, CA) with version 3 software and reagents. A custom DNA panel consisting of 136 amplicons and covering regions of interest within 25 genes frequently mutated in hematological neoplasms (Mission Bio) and the commercially available TotalSeq-D Heme Oncology Cocktail (BioLegend, San Diego, CA) were used. Cryopreserved cells were thawed, diluted in RPMI-1640 (Sigma-Aldrich, Saint Louis, CA), washed once with phosphate-buffered saline/0, 5% bovine serum albumin, dead cells were removed using the dead cell removal kit with MS columns (Miltenyi Biotech, Cologne, Germany) and finally resuspended in cell staining buffer (BioLegend). The cell count and viability were assessed using a Countess 3 Automated Cell Counter (Thermo Fisher). One million cells at a concentration of 25 000 cells per μL were blocked, stained (Heme Oncology Antibody Cocktail) and washed according to manufacturer’s instructions. After staining, cells were resuspended in Mission Bio’s cell buffer and diluted to a concentration of 3000 cells per μL. Encapsulation, lysis, barcoding and targeted PCR were performed according to the manufacturer’s instructions. PCR products were digested and purified using HighPrep PCR beads (MagBio, Genomics Inc, Gaithersburg, MD) followed by library PCR with index primers. After PCR cleanup, each library was quantified using the Qubit dsDNA HS Assay kit (Thermo Fisher, Waltham, MA) and fragment size distribution was assessed by the Tapestation 4200 (Agilent Technologies, San Diego, CA) using D1000 Screen Tape. Libraries were pooled (equimolar) followed by sequencing on the NextSeq 2000 using a P3 reagent kit (Illumina, San Diego, CA).

FASTQ files were analyzed through the Tapestri Pipeline (DNA &Protein v3.4, Mission Bio), which trims adapter sequences, aligns reads to the human genome (hg19) using Burrows Wheeler Aligner (BWA), performs barcode correction, assigns sequence reads to unique cell barcodes and performs genotype calling using GATK (version 3.7). Loom and h5 files generated were analyzed using the Mosaic v3.1.1 with the aid of the curated Jupiter notebook downloaded from Mission Bio (https://missionbio.github.io/mosaic/notebooks/curated_notebooks/dna-protein.html). Figures produced were also generated using the Jupiter Notebook. Mutations identified in NGS analysis were tracked.

The median age at diagnosis was 69 years (range, 53-83); 14 of 16 patients were deceased (12 patients succumbed to T-PLL whereas 2 patients died due to other causes following allo-HSCT). Eight patients received alemtuzumab and 3 of these patients went on to have an allo-HSCT. Three patients (patient numbers 4, 5, and 12) did not require treatment at diagnosis and were treated when their disease progressed but 1 patient (patient number 4) progressed rapidly and died when treatment was about to start. One patient (patient number 16) was diagnosed incidentally when treated for other reasons and no treatment for T-PLL was initiated. Mean time-to-first treatment was 14 (range, 0.2-57) months. Clinical characteristics are presented in Table 1. Mean overall survival was 24 months (median [range], 24 [0.2-72] months) (supplemental Figure 1).

For 9 patients, samples from different time points were available (Table 1).

In these cases, 5 of 9 patients relapsed after treatment with alemtuzumab (Campath), and treatment data for 4 patients are lacking (Table 1). These samples were cell sorted and exhibited high purity for the selected cell populations, with the exception of the sample from patient number 6, where due to the low amount of CD19 positive cells (0.3%) (Table 2), the “normal” fraction used as a germ line sample, was obtained from a bone marrow sample taken during complete remission. A low level of T cells (CD3⁺ and/or CD7⁺) was observed in the enriched CD19-positive cells (median [range], 1.5% [0%-17%]). The CD19-negative cell population (tumor fraction) sorted from diagnostic and relapse samples had a mean cell fraction comprising CD3⁺ and/or CD7⁺ cells of 88% (median [range], 91% [67%-100%]) (Table 2). At diagnosis, the most frequently mutated gene was ATM (n = 10) followed by STAT5B (n = 7), JAK3 (n = 3), EZH2 (n = 3), BCOR (n = 1), and STAT6 (n = 1) (Table 2). Twenty-one pathogenic (P) or likely pathogenic (LP) variants were detected in 11 patients at diagnosis with a mean number of P/LP variants of 1.9 (range, 1-5) and a mean VAF of 43%, median 30% (range, 7%-91%) (Table 2). In the paired samples at the time of relapse (n = 9) 14 P/LP variants were detected at a mean VAF of 41%, median 40% (range, 1%-91%). Five patients had no variants in any of the genes analyzed at diagnosis.

In 7 patients, due to the lack of available sample material, only the diagnostic sample was investigated (patient numbers 3, 7, 8, 11, 12, 13, and 16) and in 3 of these patients (patient numbers 7, 12, and 13) no mutations were detected with our targeted NGS gene panel (Table 2).

Several patterns of clonal evolution were observed in the 9 patients that had both a diagnostic and relapse sample available for analysis. In 2 patients (patient numbers 1 and 9) no mutations were detected in any of the genes analyzed at either time point. Clonal dynamics for the remaining 7 patients could be described as follows: (1) 3 patients exhibited an increase in the VAFs of the gene mutations detected at diagnosis (cases 5 [JAK3: VAF 7%-59%], 10 [ATM: VAF 59%-91% and STAT5B: VAF 27%-48%], and 14 [STAT5B: VAF 8%-36% and EZH2: VAF 14%-45%]); (2) for patient 4, the VAF of the single mutation observed at diagnosis decreased at relapse (ATM: VAF 55%-33%); and (3) in patient 2, a significant decrease in the VAF for 1 mutation was observed (STAT5B: VAF 25%-1%), whereas the other variant remained stable (ATM: VAF 91%-89%) (Table 2; Figure 1; supplemental Figure 2). Of the 5 mutations detected at diagnosis in case 15, the VAFs remained stable for 4 of the 5 mutations with only a slight decrease observed for a mutation detected within the JAK3 gene which was detected at 11% at diagnosis vs 6% at relapse (Table 2; supplemental Figure 2). Finally, patient 6 demonstrated a stable mutation (ATM: VAF 44%-40%) detected at diagnosis but the emergence of additional mutations at relapse (Table 2; Figure 1; supplemental Figure 2). The changes in VAF over time for the most commonly involved genes; ATM, STAT5B, JAK3 and EZH2 are summarized in Figure 1.

In 3 patients (patient numbers 2, 6, and 15) the VAFs were higher than expected in the “normal samples that is, the B-cell sorted fraction” indicating that mutations or variants detected were also present in the non-tumor cells that is, B cells and myeloid cells. The variants in question were subsequently analyzed using ddPCR and were detectable at levels comparable to those obtained by NGS (Table 3). For example, in patient 2 an ATM mutation was detected by NGS at a VAF of 91% in the diagnostic tumor sample while NGS analysis of the B-cell fraction detected the same variant at a VAF of 31%. Patient 2 also harbored a STAT5B variant at a VAF of 25% at diagnosis (tumor fraction); however, this variant was detected at a VAF of 10% in the sorted B-cell fraction (Table 3). VAFs obtained from NGS analysis vs ddPCR for both the tumor (diagnosis) and germ line (B cell) fractions analyzed demonstrated concordance for all samples analyzed (Table 3).

For patient 2, a total of 7421 cells were analyzed and both the ATM and STAT5B mutations were detected in almost all T cells investigated by protein and DNA single-cell multiomics using the Mission Bio tapestri technology (Figures 2A, 3A, 4, and 6A). Because the number of B cells detected in this sample was negligible, the presence of mutations within the B cells could not be confirmed.

In patient 6 an ATM mutation was detected at a VAF of 44% in the diagnostic tumor sample by targeted NGS. This mutation was also detected in the “normal” cell fraction at a VAF of 63%. This “normal” fraction, acting as a germ line sample in this case, was obtained from a bone marrow sample taken during remission. VAF measured by ddPCR were concordant for the ATM gene both in the tumor sample and the normal cell fraction (VAF 49.2% vs 49.9%, respectively) (Table 3). Additional mutations, classed as pathogenic, were detected in the EZH2 (VAF 28%) and STAT5B genes (VAF 14% and 35%) at second relapse following allo-HSCT (Table 3; supplemental Figure 1). The EZH2 mutation was also identified at a very low level, as measured by ddPCR (0.02%), in the sample of “normal cells” with a low fraction of 0.3% CD19-positive B cells at primary diagnosis (Table 3). The EZH2 mutation was detected in both the T-cell fractions and the B-cell fraction (n = 6880 cells) by DNA + protein single-cell multiomics (Figures 2B, 3B, 5, and 6B). The STAT5B mutations in this case were not further investigated.

Genomic alterations of the ATM gene are found in >85% of T-PLL cases¹⁶ and this finding was reflected in our results with ATM mutations detected in 9 of the 11 patients harboring mutations in our patient cohort. To the best of our knowledge, no previous study has systematically investigated clonal evolution in T-PLL in longitudinal samples or performed single-cell analysis. A major advantage and strength of our study is thus, the multiomics approach taken to analyze longitudinal samples. Indeed, the varying clonal shifts observed between diagnostic and relapse samples indicate different patterns of evolution. Varying patterns of evolution, or evolutionary dynamics, have been described in several other hematological malignancies.^17-19 

The detection of mutations shared by both the malignant T-PLL cells and nonmalignant B cells suggests a possible shared common clonal origin. A plausible explanation is that mutations may arise in a hematopoietic progenitor cell as an early, initial clonal event. We detected mutations using 3 independent methods that is, targeted NGS analysis of tumor cell fractions and enriched fractions of normal cells (both enriched CD19 positive cells and whole bone marrow remission samples), ddPCR and DNA + protein single-cell multiomics.

Our finding of identical mutations in 3 of our patients (patient numbers 2, 6, and 15) in the nonmalignant, CD19 enriched fractions at high VAFs supports the notion that these cells share a common clonal ancestry. We hypothesize that while the B cells remain “normal,” the T cells acquire additional genetic aberrations and/or are more vulnerable to microenvironmental interactions leading to the development of T-PLL with mutations associated with clonal hematopoiesis being the first hit.

Activating mutations within the JAK1, JAK3, and STAT5B genes have been identified as recurrent genomic aberrations in T-PLL using different sequencing approaches.^8,20,21 Interestingly, in 4 of our patients that carried ATM and/or JAK/STAT mutations (patient numbers 3, 6, 14, and 15) mutations in genes such as BCOR and/or EZH2 co-occurred. Mutations within the BCOR and EZH2 genes have previously been reported in T-PLL²¹ at frequencies similar to what we observed in our patient series. Both genes play a role in clonal hematopoiesis lending further support to the idea of a common clonal progenitor cell. This hypothesis is strengthened by our finding of mutations within these genes in both normal B cells and T-PLL cells. However, it cannot be ruled out that genes involved in the JAK/STAT pathway may also be involved in clonal hematopoiesis.

Detection of an ATM mutation in a composite lymphoma of chronic lymphocytic leukemia and T-PLL has been described in a previous case report and suggested as a possible common pathogenetic mechanism.²² The presence of driver ATM mutations within the B-cell compartment indicates that B-cell lymphomas may arise as a consequence of mutated ATM which could then be screened for at the time of relapse. Pathogenic heterozygous germ line ATM variants occur in ∼1% of the population²³ with the carriers believed to have defective or decreased production of ATM kinase. Heterozygous pathogenic germ line variants within the ATM gene have been associated with an increased risk of numerous solid tumor cancers for example, breast, prostate, and pancreatic, but have recently been reported in patients with chronic lymphocytic leukemia.²⁴ In 1 of our patients (patient 6) the VAF of the ATM mutation was high in the normal cell sample, which in this case consisted of a bone marrow remission sample with no evidence of residual disease. This implies that this ATM variant may be of germ line origin. To the best of our knowledge this is the first study detailing a potential causative germ line ATM variant in a patient diagnosed with T-PLL. Although the question remains as to whether this variant may have conferred a predisposing risk for the development of T-PLL in this patient, this observation warrants further investigation regarding the potential role of germ line pathogenic ATM variants in the pathogenesis of T-PLL.

One limitation of our study was related to the targeted approach undertaken, which focused on key genes known to be recurrently mutated in T-PLL, an approach which most likely explains the lack of identified variants in 5 of the included patients. A more comprehensive approach utilizing whole exome sequencing or whole genome sequencing would allow us to better understand the impact of other genetic aberrations including structural or numerical variations that may be crucial for the development of T-PLL. That said, our multi-methodological approach in both B cells and clonal T cells provides robust proof of shared mutational signatures. Another limitation concerns the absence of cytogenetic data available for our patient series. This data would have provided an additional piece of the evolutionary puzzle because rearrangements involving chromosome 14 or X [t14(q11;q32) or, less commonly, the translocation t(X;14)(q28;q11) resulting in juxtaposition of TCL1A or mature T-cell leukemia 1 (MTCP1) proto-oncogenes to TCRAD gene enhancer elements] are frequently reported in T-PLL. Owing to the overall rarity of T-PLL, collecting large patient series with ample sample material available across several time points is difficult and underscores the need for collaboration to drive research forward.

In conclusion, T-PLL exhibits variable patterns of clonal evolution between diagnosis and relapse. Single-cell multiomics analysis reveals shared mutational signatures in both nonmalignant B cells and clonal T cells. Finally, the role of germ line ATM mutations in the pathogenesis of T-PLL should be further investigated.

The authors would like to acknowledge Clinical Genomics Uppsala, the Science for Life Laboratory, Department of Immunology, Genetics and Pathology, Uppsala University, Sweden, for performing the experiments.

Contribution: I.T. and M.M. performed experiments (flow cytometry, cell isolation, and DNA extractions); C.L. and J.A. performed bioinformatical analysis and analyzed and interpreted data; L.C., S.L., H.N., and S.W. performed experiments (ddPCR, sequencing, single-cell multiomics analysis), analysis, and interpretation of data; L.-A.S. developed the targeted deep sequencing panel; L.-A.S. and P.B. revised the manuscript and analyzed data; C.H. and R.-M.A. wrote the manuscript and analyzed and interpreted data; C.H. performed statistical analysis; and all authors agreed to the submission and approved the final manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Rose-Marie Amini, Department of Immunology, Genetics and Pathology, Rudbeck Laboratory, Entrance C5, Uppsala University and Uppsala University Hospital, SE-75185, Uppsala, Sweden; email: [email protected].