Page 94 - 2019_02-Haematologica-web

L.W. Dillon et al. et al.
however not all targets can reliably serve as a surrogate of disease burden. For example, mutations found in preleukemic founder clones can persist at significant lev- els even during complete remission8-11 and have also been observed in healthy individuals without hematologic malignancies with increasing age.12-15 To help harmonize AML MRD detection efforts, the European LeukemiaNet (ELN) recently released consensus guidelines for MRD detection in AML, including recommended markers.6
Despite advances in the field, it has remained difficult to develop a single MRD detection assay which is highly reproducible and sensitive, has limited operator depend- ence, and is capable of quantifying multiple targets simul- taneously. To overcome these limitations, we developed a multi-gene, targeted RNA-sequencing-based method for the sensitive detection and quantification of MRD in AML, which we present here. This novel assay has a min- imal sequencing budget requirement and outperforms a commercially available, conventional myeloid-targeted RNA-sequencing assay. It has a demonstrated limit of detection as low as one leukemic cell in 100,000 cells measured, so its performance equals that of current gold- standard, single-target qPCR MRD assays.
Methods
Cell lines and clinical samples
Leukemia cell lines positive for fusion genes targeted by the AML MRD panel were cultured according to the supplier’s guidelines. Peripheral blood samples were collected from a healthy adult donor and from a 46-year old female with mono- cytic AML who underwent myeloablative matched related donor allogeneic stem cell transplantation (NHLBI protocol # 07- H-0113). Additional samples from patients were collected in local institutional review board-approved biobanking protocols by collaborators. The samples were processed and RNA isolated as described in the Online Supplementary Appendix.
Preparation and analysis of targeted RNA-sequencing libraries for acute myeloid leukemia measurable residual disease detection
Targeted RNA-sequencing libraries were prepared and sequenced as described in the Online Supplementary Appendix. In short, unique molecular identifier (UMI) assignment and com- plementary DNA (cDNA) generation were performed on 250 ng of RNA using the SuperScript IV First Strand Synthesis System (Invitrogen, Carlsbad, CA, USA) and a pool of 100 nM of each barcoded (BC) primer. The barcoded cDNA was subjected to eight cycles of amplification with a pool of 100 nM of each lim- ited amplification (LA) primer and 600 nM RS2 primer, followed by final library amplification and sample indexing. Single-end 150 bp sequencing was performed on a Miseq or Hiseq 2500 platform (Illumina) using the QIAseq Read 1 Primer 1 custom primer (Qiagen, Germany). Raw sequencing FASTQ files were processed, aligned, and panel targets called as outlined in the Online Supplementary Appendix.
Real-time quantitative polymerase chain reaction
Expression of the “type A” mutation of the nucleophosmin gene (NPM1 mutA) and the RUNX1-RUNX1T1 fusion gene was determined for cell dilutions and patient samples by qPCR using the ipsogen NPM1 mutA MutaQuant kit (Qiagen, cat# 677513) or ipsogen RUNX1-RUNX1T1 kit (Qiagen, cat# 675013), respective- ly, according to the manufacturer’s instructions, and the Rotor-
Gene Q 5plex HRM (Qiagen). A standard curve for each gene was obtained from the plasmid serial dilutions and used to cal- culate the copy number for each sample. A patient’s sample was considered positive for NPM1 mutA or RUNX1-RUNX1T1 if the copy number detected was contained within the standard curve and the water control was negative.
Digital droplet polymerase chain reaction
CBFB-MYH11 type A expression was determined with digital droplet PCR (ddPCR) by converting the established European Against Cancer (EAC) assay for CBFB-MYH11 type A to the Raindance platform (RainDance Technologies, Billerica, MA, USA), as described in the Online Supplementary Appendix.
ArcherDx Myeloid FusionPlex library preparation and analysis
Anchored multiplex PCR-based enrichment RNA-sequencing libraries were generated from 250 ng of RNA using the ArcherDx Myeloid FusionPlex assay for Illumina (ArcherDx, Boulder, CO, USA) and analyzed using Archer Analysis software version 5.1.3, as described in the Online Supplementary Appendix.
Statistical analysis
Data were analyzed using Prism statistical software (v.7.0b, GraphPad software, La Jolla, CA, USA).
Results
Targeted RNA-sequencing panel target selection for acute myeloid leukemia measurable residual disease detection
Targets for this multi-gene RNA-sequencing panel were chosen from those recommended for AML MRD detection by the ELN6 and for which standardized qPCR assays for MRD detection have already been well established5,16-19 (Figure 1A). These targets include: (i) recurrent chimeric fusion transcripts PML-RARA, CBFB-MYH11, RUNX1- RUNX1T1, and BCR-ABL1, present in 20.8% of AML patients in The Cancer Genome Atlas (TCGA) dataset;20 (ii) the recurrent insertion site in exon 12 of NPM1, which is found in an additional 27.2% of patients; and lastly (iii) aberrant expression of WT1 and PRAME transcripts, which may be used for MRD detection for up to another 20.2% of patients not covered by the preferred fusion or mutated NPM1 targets. Based on the TCGA AML cohort, WT1 and PRAME could also serve as a secondary tracking target, for orthogonal validation, in over 80% of those patients who also express fusion or NPM1 mutant transcripts. Similar to the established qPCR assays, wild-type ABL1 transcript expression was included as a normalizing control.16
Development of the targeted RNA-sequencing method for acute myeloid leukemia measurable residual disease detection
In developing a targeted RNA-sequencing strategy that would be optimal for MRD detection, we took into account several important factors including: (i) the need to optimize target capture efficiency while minimizing the amount of RNA required, as this resource is often limited when dealing with patient specimens; (ii) the incorpora- tion of strategies for the digital quantification of tran- scripts; (iii) the need to minimize sequencing burden; and (iv) the generation of a workflow that is simple, efficient, and easy to adopt.
298
haematologica | 2019; 104(2)