Page 95 - 2019_02-Haematologica-web

Targeted RNA-sequencing for MRD in AML
A
B
Figure 1. The targeted RNA-sequencing assay for measurable residual disease detection can be applied to over two-thirds of patients with acute myeloid leukemia.
(A) The acute myeloid leukemia (AML) measurable residual disease (MRD) RNA-sequencing assay comprises eight targets, including insertion mutations, fusion tran- scripts, and wild-type transcript expression. The frequency of each target was determined using AML cases in The Cancer Genome Atlas (TCGA) for whom clinical infor- mation, mutation analysis, and RNA-sequencing data were available (n=173). The expression of insertions and fusion transcripts was evaluated first. For the remaining patients, WT1 and PRAME wild-type transcript overexpression was evaluated and defined as overexpressed if greater than the mean of the entire cohort. (B) The AML MRD RNA-sequencing assay begins with targeted reverse transcription utilizing a pool of primers consisting of a gene-specific region (GSP1), random 12-nucleotide unique molecular identifier (UMI), and conserved sequence region (RS2) per target. The resulting barcoded complementary DNA (cDNA) is subjected to limited ampli- fication using a reverse primer complementary to the RS2 sequence and a pool of forward primers consisting of a gene-specific region (GSP2) and conserved sequence region (FS2) per target. The targeted amplicons are then subjected to amplification, library construction, and sample indexing for Illumina sequencing.
The final assay design addresses these factors by utiliz- ing a pool of target-specific primers containing 12 nucleotide UMIs (Online Supplementary Table S1) which capture and individually tag RNA molecules of interest during reverse transcription, followed by targeted PCR of the barcoded cDNA, and library construction (Figure 1B). Unlike most targeted RNA-sequencing approaches, this simplified design reduces protocol steps while maximiz- ing utilization of the RNA input by performing target enrichment during the reverse transcription step, as opposed to after cDNA generation. The concurrent addi- tion of molecular barcodes during this first step also allows for a digital output, increasing the accuracy of tran- script quantification. Additionally, the amplicon-based enrichment design for fusion detection limits the sequenc- ing requirements, since only fusion transcripts and not wild-type transcripts, will be amplified. Finally, with a total of only three steps, the hands-on time is minimized, allowing for the entire protocol to be completed in less than a day.
Validation of assay performance and limit of detection
To assess the sensitivity and dynamic range of the AML MRD RNA-sequencing panel, cell lines expressing fusion transcripts or patient cells positive for the NPM1 mutA insertion mutation (94% blasts) were serially diluted (1:10 to 1:100,000) into healthy adult donor peripheral blood mononuclear cells and RNA was isolated. A total of 250 ng of RNA from each dilution was subjected to targeted RNA-sequencing library preparation and sequencing.
Sequencing files were processed by extracting the UMI from each read, alignment to the human genome, and clustering of sequences which correspond to the intended panel targets (Online Supplementary Figure S1). Targets were quantified by the number of unique UMIs, with a library-specific UMI cutoff value established to eliminate background due to sequencing errors (Online Supplementary Figure S2).
The expression of all assay targets exhibited significant correlation (linear regression r2 ≥0.97) and leukemic cells could be detected at a level as low as one leukemic cell in 100,000 healthy donor cells (Figure 2). For the fusion and mutated NPM1 transcripts, detection sensitivity was between 1:10,000 to 1:100,000. Many of the cell lines expressing fusion transcripts also displayed aberrant WT1 and/or PRAME transcript expression, which was also highly correlated and for which the assay showed varying degrees of sensitivity, ranging from 1:1,000 to 1:100,000. Detection of all targets was highly reproducible across replicates.
Determination of assay sequencing requirements
Variations in sequencing depth depending on the leukemic burden present in a sample and baseline error rates between sequencing platforms are important ele- ments which can affect assay performance and reliability. Additionally, sequencing read requirements play an impor- tant role in the feasibility of assay adaptation into practice.
To address these factors, we first examined the impact of sequencing depth and platform on assay detection met-
haematologica | 2019; 104(2)
299