Therapy-related leukaemias with balanced translocations can arise from pre-existing clonal haematopoiesis

Therapy-related leukaemia is a life-threatening complication of cancer treatment. The prognosis is poor and incidence is increasing, mainly due to the increasing survival from primary cancers, leading to an emerging major healthcare problem [1]. Currently two groups of tMN are recognised: the first, accounting for ~80% of cases, is characterised by exposure to alkylating agents, a latency period of 5–6 years, a complex or monosomal karyotype and/or mutations in TP53 [2]. These mutations have been detected in samples pre-dating therapy exposure [3] and presence of clonal haematopoiesis (CH) prior to cancer treatment is associated with a markedly increased risk of tMN with estimated hazard ratios of 5.8–13.7 [4, 5], indicating that cancer therapy may drive the expansion of pre-existing mutated clones, eventually leading to malignant transformation. The second group of tMN accounting for ~20% of cases is characterised by exposure to topoisomerase II (topo2) targeting agents, a latency period of 1–2 years and presence of recurrent fusion genes identical to those seen in de novo AML [6, 7]. In these cases, there is evidence that drugs targeting topo2 directly initiate chromosome rearrangements: in the presence of mitoxantrone or etoposide, topo2 cleaves the PML and RARA loci at the precise base pair positions observed to be fusion breakpoint junctions in patients who develop therapy-related acute promyelocytic leukaemia (tAPL) after exposure to these agents [8, 9]. Fusion-gene associated chromosome translocations are widely regarded as primary leukaemia-initiating events in AML. Genome-wide studies show striking exclusivity between fusion genes and other mutations shown to be leukaemiainitiating events. For example, mutations in DNMT3A are frequent in AML and are associated with a preleukaemic state [10–12]. These mutations either do not occur or are rare in leukaemias with fusion genes [11–13]. Thus, two distinct processes that lead to the generation of tMN appear to mirror two separate mechanisms of de novo leukaemogenesis, with two distinct types of primary leukaemia-initiating event (fusion genes and preleukaemic mutations, respectively). The relationship between CH and therapy-related leukaemias with recurrent chromosome translocations has not previously been investigated. Here, we focussed on a molecularly defined group of patients with PML-RARA tAPL and present evidence that in these patients, in contrast to de novo APL, chromosome translocations may be secondary events, arising on a background of pre-existing CH. We first performed whole-exome sequencing of 55 cases of de novo (dnAPL) and 13 cases of therapy-related APL (tAPL, exposure details shown in Supplementary Table 1). Information regarding library preparation, sequencing and analysis is provided in the Supplementary data. We identified * Richard Dillon richard.dillon@kcl.ac.uk


Patients and samples
All samples were taken from patients for the purpose of molecular disease monitoring. Patients with de-novo APL were enrolled in the UK NCRI AML17 study (ISRCTN55675535) [16] and all received treatment with the AIDA protocol and achieved molecular complete remission. Patients with therapy-related APL were treated with either the AIDA protocol or all-trans retinoic acid (ATRA) and arsenic trioxide (ATO) according to the AML17 schedule. Patients provided written informed consent for the use of excess diagnostic material for this study, which was approved by the London Westminster Research Ethics Committee (reference number 06/Q0702/140).

Whole-exome sequencing
Genomic DNA libraries prepared using the SureSelectXT Human All Exon V6 (Agilent Technologies, Santa Clara, CA) target enrichment system were sequenced on the Illumina HiSeq 2500 platform (Illumina, San Diego, CA) with 100bp paired-end reads. Reads were aligned to the reference human genome (hg19, b37) using novoalign (Novocraft Technologies, Selangor, Malaysia). Alignments underwent quality control and filtering as previously described [17]. In each sample, >87% of the GENCODE-defined coding bases of the exome was represented by at least 20 reads. Pairwise variant calling was performed using SAMtools and filtered using VarScan2. Only variant sites with coverage of ≥20 reads in both leukemic and remission samples were considered. Alleles present in ≥20% of reads in leukemic samples and <20% in remission samples were identified and annotated.

Targeted deep sequencing
A custom capture panel consisting all somatic variants identified by whole exome sequencing together with the coding regions of genes previously associated with clonal haematopoiesis (see supplementary table 3) was constructed using the HaloPlexHS system (Agilent). Sequencing libraries were constructed using the Agilent Bravo liquid handling system and sequenced using on a HiSeq 2500 instrument. Alignment and generation of consensus read families was performed using SureCall software (Agilent).

Amplicon sequencing
Sequencing libraries were generated using a two step PCR approach. In the first step, target specific primers incorporating adapter sequences were used to amplify the locus of interest. In the second step, a second PCR was used to incorporate 5' and 3' index and adapter sequences. Libraries were sequenced on a HiSeq 2500 instrument and reads were aligned to the reference genome using Novoalign as described above.

Reverse Transcription-quantitative Polymerase Chain Reaction (RT-qPCR)
RNA was isolated from cell samples of interest using Trizol reagent (Life Technologies, Inchinnan, UK) and cDNA was synthesised using the SuperScript3 reverse transcription kit (Life Technologies). qPCR was performed in triplicate using primer sets and conditions according to the Europe Against Cancer Programme using an ABI 7900 instrument (Life Technologies) and ABL was amplified in parallel as a control for RNA quality and quantity. Criteria for reporting positivity proposed by Gabert et al [18] were adopted.

Fluorescence in-situ hybridisation (FISH)
Cytospin preparations were fixed in methanol and hybridised with a PML/RARA dual fusion FISH probe (Cytocell, Cambridge, UK) for subsequent analysis using fluorescent microscopy.

Xenografts
NOD/SCIDIL2Rg -(NSG) mice between 10-14 weeks of age were irradiated (2Gy) and treated with human immunoglobulin 10mg per gram of body weight by intraperitoneal injection (Privigen, CSL Behring, Haywards Heath, UK). The following day, 10 6 human cells were injected into the tail vein. Mice received ciprofloxacin in their drinking water for four weeks post-transplant. Mice were assessed for engraftment at 12 weeks and then at monthly intervals. Assessment was by bone marrow aspiration from the tibia under general anaesthetic. Aspirates were stained with the following antibodies: human CD45 APC-Cy7, human CD19 APC, human CD33 PE, human CD 3 FITC (all from BioLegend) and analysed using a BD Fortessa flow cytometer. Where engraftment was detected, mice were sacrificed and the bones dissected and crushed. Cell extracts were stained with the same antibody cocktail and human CD45+ cells were sorted using the FACSAriaII. Double sorting was used to ensure >99% purity. Sorted cells were separated DNA and RNA extraction and cytospins were prepared for FISH.