The 'Next-in-Cml' Study: A Prospective Multicenter Study of Deep Sequencing of the BCR-ABL1 Kinase Domain in Philadelphia Chromosome-Positive Patients with Non-Optimal Responses to Tyrosine Kinase Inhibitor Therapy

Introduction

Some retrospective studies in tyrosine kinase inhibitor (TKI)-resistant Philadephia-positive (Ph+) leukemia patients (pts) have suggested that deep sequencing (DS) may provide a more accurate picture of BCR-ABL1 kinase domain (KD) mutation status as compared to conventional sequencing (CS). However, the frequency and clinical relevance of low burden mutations remains to be explored prospectively in large series of unselected pts. In addition, the implementation of routine BCR-ABL1 DS in multiple molecular diagnostic laboratories has never been attempted. These open issues led us to design a multi-center, multi-laboratory prospective study ('NEXT-IN-CML') aimed to assess the feasibility, performance and informativity of DS for BCR-ABL1 KD mutation screening.

Aims

The first phase of the study was aimed to establish a network of 5 reference labs sharing a standardized DS workflow, a joint database for clinical and mutational data storage and a common pipeline of data analysis, interpretation and clinical reporting. The second phase of the study, involving 54 Italian Hematology Units, is aimed to assess the frequency and clinical significance of low burden mutations detectable by DS by prospective collection and analysis of samples from chronic myeloid leukemia (CML) pts who exhibit failure (F) or warning (W) responses and relapsed Ph+ acute lymphoblastic leukemia (ALL) pts.

Methods

A PCR and an amplicon DS protocol already set up and optimized for the Roche GS Junior in the framework of the IRON II international consortium was adopted. In the first phase, 5 batches of blinded cDNA samples were prepared and shipped to evaluate individual lab performances. The batches included archival samples with known BCR-ABL1 mutation status as assessed by CS and serial dilutions of BaF3 T315I+ cells in BaF3 unmutated cells, simulating mutation loads of 20% down to 1%. In the ongoing second phase prospectively, consecutively collected CML and Ph+ ALL samples are being analyzed in parallel by CS and DS. Clinical history and follow-up data are used for correlations.

Results

In the first phase of the study, 312/320 amplicons were successfully generated and sequenced. A median of 124,686 (range, 48,181-170,687) high quality reads were obtained across the 5 labs. Median number of forward and reverse reads was 1,757 (range 884-7,838), with no coverage dropouts for any amplicon or index. Comparison of observed vs expected mutations showed that 76/78 evaluable samples were accurately scored. In the remaining two, the analysis software failed to detect the 35bp insertion ('35INS') commonly detectable between exons 8 and 9. Quantitation of point mutation burden was highly reproducible across the entire range of frequencies, from 100% to 1%.

The second phase of the study has started in Jan 2016. As of Jul 31^st, a total of 106 consecutive pts (CML, n=96; Ph+ ALL, n=10) have been enrolled. The present analysis focuses on the first 75 CML pts (60 F and 15 W), for whom sequencing results are currently available (analysis of the entire population of patients enrolled up to Nov 2016 will be presented at the meeting). Clinically actionable mutations have been detected in 10/75 (14%) pts by CS and in 20/75 pts (27%) by DS. Notably, among the 10 pts positive for clinically actionable mutations by DS but not by CS, 3 had a low burden T315I (2 F [dasatinib, imatinib] and 1 W [dasatinib]). In 5 additional pts negative for mutations by CS (3 F and 2 W), DS identified multiple low burden mutations with unknown IC₅₀, suggesting that the cooperation of individually 'weak' mutants may be a new mechanism underlying reduced TKI efficacy. Longitudinal analysis and follow-up of pts are shaping the clinical significance of different types of low burden mutations and will be presented.

Conclusions

The 'NEXT-in-CML' study is demonstrating that DS of BCR-ABL1 can successfully be implemented in national lab networks and is an important step forward towards routine use of this technology. We have now adapted the protocol for both the Ion Torrent PGM and the Illumina Miseq platforms. For a minimum of 15 samples per sequencing run, DS costs are estimated to equal those of CS (cost per sample, reagents only: ≈100€ for PGM (314 chip) and Miseq (nano kit v2) vs ≈95€ for CS) with comparable turnaround times for delivery of results.

Our study is also contributing useful data for the clinical interpretation of DS findings.