An Enhanced Sensitivity DNA Sequencing Protocol for the Detection in AML of Measurable Residual Disease (MRD) Applicable for All Mutations

Background: One of the more important prognostic factors used to predict the outcome in acute myeloid leukemia (AML) is the persistence of leukemic cells after treatment. The reliable measurement of residual disease (MRD) offers many other clinical uses besides. An assay that was facile, affordable, and applicable to the broadest group of patients would find immediate favor.

Next-generation sequencing (NGS) combined with various enrichment methods allow high sequencing depth on restricted targets. For instance, by enriching for <100 targets raw sequencing depths of 10⁵ to 10⁶ are routinely achieved. However, after excluding PCR duplicates identified using either end-point diversity of paired-end fragments or molecular tags, we found that NGS libraries effectively capture only a few percent of the input genomes (typically 1000 genomes or fewer for a 100ng input DNA corresponding to 30,000 genomes) effectively restricting the limit of detection (LOD) to approximately 1%. Hybrid capture and primer extension methods including protocols available from Nimblegen, Illumina for whole exome sequencing, exon capture using IDT probes and Nugen yielded similar results: all showed extensive loss of input genome diversity restricting the LOD (see Table 1 Section A). Most surprising was whole exome sequencing (WES). Sequencing 1000 ng input representing 300,000 genomes performed worst: for a given locus, loss of genomic diversity of approximately can be as high as 99.99% and is always more than 95% (Table 1A). More input is clearly not better.

Methods and Results: We used the ThermoFisher multiplex PCR-based method to capture the 50 genomic fragments most frequently mutated in AML and labeled each with unique DNA tags. Sequence generated on an Ion Torrent showed that for 100ng of input DNA, >50% of the genomic diversity of that input DNA was preserved, i.e., >15,000 unique genomes were captured and sequenced from an initial 30,000 genomes resulting in a LOD of 0.1% with high precision as a mutation present at 0.1% in 15,000 genomes could be observed ~15 times. As shown in Table 1B, this method scales (non-linearly) as input DNA is increased; samples containing 500ng of DNA (~150,000 genomes) reproducibly permit an LOD of 0.01%, a sensitivity far exceeding any other non-allele specific method for measuring residual disease.

In a real world application, we followed 3 AML patients from diagnosis to clinical CR. We tracked all mutations present at diagnosis using both IDT hybrid capture (HC) and the ThermoFisher (TF) method. One patient had molecular evidence of residual disease (VAF3%) that was detected and similarly quantitated by both HC and TF, in the second patient mutation were present at 1% by HC and 3% by TF . The TF value is the reliable one as it is derived from multiple independent templates. Finally the third patient was in molecular remission by HC but had a mutation allele detected by the TF with a frequency of 0.07%. We have extended these studies to include more than a dozen patients followed through relapse with similar results.

These pilot studies provide clear evidence that standard sequencing methods cannot reliably promise LODs below 1%; further, the ThermoFisher method enhances the LOD at least 20-fold in an assay that can be applied to the great majority of AML patients.

Table 1.

View largeDownload slide

Table 1.

View largeDownload slide

Close modal