Page 99 - 2019_02-Haematologica-web

Targeted RNA-sequencing for MRD in AML
Figure 5. Evaluation of assay performance in serial samples from a patient. RNA isolated from peripheral blood mononuclear cells of a leukemia patient positive for the NPM1 mutA insertion mutation at serial time-points was subjected to acute myeloid leukemia (AML) measurable residual disease (MRD) RNA-sequencing library preparation and sequencing. The data are presented as the ratio of NPM1 mutA copies/ABL1 copies (log10) on the y-axis and the date of sample collection relative to allogeneic stem cell transplantation (in days) on the x-axis. Important clinical time-points are depicted above the graph. Clinical relapse occurred on day 175 while MRD was detectable by targeted RNA-sequencing on days 87 and 119. The background ratio of NPM1 mutA copies/ABL1 copies (log10) detected in the normal donor is represented as a dashed line. Pre-SCT: before stem cell transplantation; Path CR: pathological cytomorphological complete remission; Allo-SCT: allo- geneic stem cell transplantation.
tion sequencing allows for both target multiplexing and flexibility in identifying mutations that could vary between patients (variations in fusion breakpoints or insertion sequences, etc.); (ii) use of primers targeting recurrent AML abnormalities greatly increases the sensi- tivity of the assay over bulk RNA- or DNA-sequencing; (iii) RNA as the starting material allows for the simultane- ous examination of mutations/fusions and changes in transcript expression; (iv) RNA increases the limit of detection over that provided by DNA if the transcript is expressed at a level greater than the genomic equivalent per cell; (v) UMIs allow for the absolute quantification of target levels; and (vi) targeted primers during reverse tran- scription increase the limit of detection and allow for effi- cient use of the starting material. However, it is conceiv- able that this RNA-sequencing assay could be comple- mented in the future by the use of a DNA-based AML MRD next-generation sequencing approach (tracking, for example, somatic mutations in TP53, IDH1, IDH2, FLT3, etc.) which together would allow coverage of almost all cases of AML.
Utilizing cell lines and patient samples expressing the targets included in our assay, we demonstrated that the AML MRD RNA-sequencing assay has a sensitivity for residual disease detection down to as low as 1 in 100,000 cells (Figure 2). Importantly, this detection limit is well below the threshold of 1 in 1,000 cells currently suggested by the ELN MRD consortium6 and is comparable to that of the gold standard single-target molecular techniques (Figure 3). Furthermore, due to the MRD-focused design of our assay, it can achieve up to a 1,000-fold greater sen- sitivity than that of the most similar targeted RNA- sequencing assay for myeloid malignancies available on the commercial market (Figure 4). However, since RNA expression levels can vary from one patient to another, it is important to note that sensitivity levels can vary and a detection limit of 1 in 100,000 cells may not always be attainable. This is a general difficulty affecting all RNA- based methods of MRD detection, including qPCR, and is reflected in the recommendation that molecular relapse be
determined by the progression of trends across multiple time-points in an individual patient rather than by a single landmark assessment.6
Several important considerations for the clinical utility of an MRD assay include sample input requirements, cost, time, and ease of assay adaptation. Each of these factors were considered in the assay design. The use of targeted primers and the addition of UMIs during the reverse tran- scription step allow for maximal usage of the starting mate- rial while simplifying the workflow (Figure 1B), which can be completed in a single day and is easy to adopt and/or automate. Additionally, with a minimal sequencing require- ment of only 1-3 million reads (Online Supplementary Figure S4), a single patient sample can be assessed for immediate results or multiplexed on larger scale runs to minimize costs (Online Supplementary Table S4). Importantly, this multiplex assay and analytic workflow is both flexible and expand- able. The design allows for all targets to be screened at diag- nosis, with the ability to tailor analysis to a subset of patient-specific targets at later time-points to further mini- mize sequencing costs. Additional targets can easily be added to the assay design, potentially allowing for the detection of other AML MRD markers, chimerism,28,29 and HLA loss30 for those poor-risk AML patients who are not optimally covered by this assay but who will often undergo allogeneic stem cell transplantation.
While we confirmed the feasibility of use of this test in patient blood and bone marrow (Table 1, Figure 5), future work is needed to test the utility of this technique in large cohorts of patients and to determine the specific impact of MRD detection on AML patient outcomes in this setting. Overall, we believe that this UMI-based RNA-sequencing assay provides a high-throughput, reproducible, and broadly applicable tool for standardized detection of residual disease in patients with AML.
Acknowledgments
This work was supported by the Intramural Research Program of the National Heart, Lung, and Blood Institute (NHLBI) of the National Institutes of Health (NIH). GWR is
haematologica | 2019; 104(2)
303