Page 27 - 2021_12-Haematologica-web

Prognostic relevance of subclonal mutations in ALL
Genomic characterization of relapsed pediatric ALL has revealed multiple alterations that are enriched compared to diagnosis, including activating mutations in RAS path- way genes, HAT domain mutations in CREBBP and dele- tions or mutations in the B-cell transcription factor IKZF1.6-13 The presence of these aberrations at the time of diagnosis can be of potential prognostic relevance, as has been demonstrated extensively for IKZF1 in many differ- ent treatment protocols12,14-19 and can even lead to adjust- ments in stratification and treatment.14,20 However, it remains unclear whether mutations in relapse-associated genes when present in a minor subclone at initial diagno- sis are also clinically relevant.
Subclonal mutations can be identified using deep target- ed, next-generation sequencing techniques.21,22 Despite the sensitivity of these techniques, both amplification and sequencing can easily lead to errors that hamper the reli- able detection of low-level mosaic mutations. We previ- ously demonstrated that single molecule molecular inver- sion probes (smMIP), which use unique molecular identi- fiers to barcode each DNA copy, can correct for sequenc- ing and amplification artefacts, resulting in a reliable detection of low-level mosaic mutations, down to a vari- ant allele frequency of 0.4%.23
In this study we used the smMIP-based sequencing approach to perform deep targeted sequencing of seven relapse-associated genes in a cohort of 503 pediatric ALL samples taken at initial diagnosis, resulting in the detec- tion of 141 clonal and 469 subclonal mutations. In addi- tion, we performed real time quantitative polymerase chain reaction (PCR) to sensitively detect subclonal IKZF1 exon 4-7 deletions (del 4-7), which were found at a similar frequency as full-clonal deletions. Subsequently, we esti- mated their potential as drivers of clonal expansion and prognostic markers for relapse development.
Methods
In this study we analyzed two cohorts of diagnostic samples from B-cell precursor ALL patients treated according to the Dutch Childhood Oncology Group (DCOG) protocols DCOG-ALL9 (n=131)12,24 and DCOG-ALL10 (n=245) (Online Supplementary Table S1). Both cohorts were representative selections of the total stud- ies12,24 (Online Supplementary Table S2). The median age at diagno- sis of the patients in these cohorts was 4 and 5 years, and the median follow-up time, estimated with a reverse Kaplan-Meier method, was 138 and 104 months, respectively.25 Relapse occurred in 18% (24/131) and 11% (27/245) of the patients, while 0.7% (1/131) and 2.8% (7/245) died during the follow-up. DNA was iso- lated from mononuclear cells obtained from bone marrow or peripheral blood. The median blast percentage of the samples was 92% (Online Supplementary Table S3). To increase the number of patients for the comparisons between relapsed and non-relapsed cases, we used an extended cohort of diagnostic samples from 127 additional ALL patients treated according to the DCOG-ALL9 (n=76) or DCOG-ALL10 (n=51) protocols; this cohort was enriched for patients who had a relapse and also contained 55 patients with T-cell ALL. This latter cohort was not included in the survival analyses. In order to detect mutations preserved in major clones at relapse, we performed Sanger sequencing (73/171) or used previously published Ampliseq-based deep-sequencing data (98/171) to verify alterations observed at diagnosis.26 In accor- dance with the Declaration of Helsinki, written informed consent was obtained from all patients and/or their legal guardians before
enrollment in the study, and the DCOG institutional review board approved the use of excess diagnostic material for this study (OC2017-024).
In order to accurately detect subclonal alterations in diagnostic samples, 166 smMIP were designed in CREBBP, PTPN11, NT5C2, WHSC1, TP53, KRAS and NRAS, seven genes that are frequently mutated in relapsed ALL (Online Supplementary Table S4, Online Supplementary Materials and Methods). IKZF1 and ERG deletion sta- tus was assessed using the multiplex ligation-dependent probe amplification assay (MLPA) SALSA P335 ALL-IKZF1 and P327 iAMP-ERG kits, respectively (MRC-Holland, the Netherlands), according to the manufacturer’s instructions and as described before.12,24 Additionally, IKZF1 4-7 deletions were assessed with Sanger sequencing and real-time quantitative PCR, using an IQ SYBR Green supermix (Biorad, USA). For detailed descriptions of the smMIP-based sequencing, IKZF1 deletion detection and data analysis, see the Online Supplementary Materials and Methods (Online Supplementary Figures S1 and S2, Online Supplementary Tables S4-S6).
To test continuous and categorical variables, the nonparametric Wilcoxon signed rank and Fisher exact tests were used, respective- ly (R packages ggpubr version 0.2 and stats version 3.5.1). Cumulative incidence of relapse (CIR) was estimated by employ- ing a competing-risk model with death as a competing event.27 To assess the statistical difference between CIR, the Gray test28 was applied. To investigate the effect of prognostic factors on relapse, univariate and multivariate Cox proportional hazard regression models were estimated. Competing risk analysis was performed with the R packages cmprsk (version 2.2-7) and survminer (version 0.4.3). Univariate and multivariate Cox models were estimated using R package survival (version 3.1-12). Data were visualized using the R package ggplot2 (version 3.2.1) and cBioPortal MutationMapper.29,30
Results
A total of 503 diagnostic samples from children with ALL (Online Supplementary Table S1) was subjected to tar- geted deep sequencing of the relapse-associated genes TP53, CREBBP (HAT domain), KRAS, NRAS, PTPN11, NT5C2 and WHSC1 using smMIP, which contain random molecular tags to accurately detect low-level mosaic vari- ants.23 Each targeted region was covered with an average of 308 unique capture-based consensus reads (Figure 1, Online Supplementary Figure S1A, B), enabling the reliable detection of alterations with allele frequencies even below 1%. A total of 7,836 quality-filtered variants was detected, of which 610 were absent in public and private variant databases and were predicted as pathogenic. The allele frequency of these mutations ranged from 0.03-100% (Figure 2A, Online Supplementary Table S3). The majority of the mutations (473/610; 78%) was found in one of the three RAS pathway genes (KRAS, NRAS, PTPN11), of which 418 (88%) were known hotspot mutations.
In addition to sequencing the seven relapse-associated genes, we performed sensitive screening for IKZF1 dele- tions, which are strongly associated with the occurrence of relapse. We chose to focus on exon 4-7 deletions, which represent 25% of all IKZF1 deletions, have a similar unfa- vorable outcome as other IKZF1 deletions,31 and show the strongest clustering of deletion breakpoints, thus enabling their sensitive upfront detection by breakpoint-spanning semi-quantitative PCR.32 Applying this strategy to the 503 diagnostic samples revealed all 22 IKZF1 exon 4-7 dele-
haematologica | 2021; 106(12)
3047