Detecting​ myeloid malignancy minimal residual disease

By contrast, dual-strand or so-called duplex consensus sequencing, as pioneered by Lawrence Loeb, MD, PhD, and colleagues at the University of Washington, considers evaluation of both strands of a DNA molecule and requires the variant (and its complement) to be seen in both strands (Schmitt MW, et al. Proc Natl Acad Sci USA. 2012;​109​[36]:14508–14513). In this regard, Dr. Wu says, “duplex sequencing is more stringent and thus more accurate than single-strand consensus sequencing in that it seeks to exclude early cycle PCR error by specifically tagging the paired Watson/Crick strands of the DNA template to confirm that a mutation and its complement are indeed both seen on the two strands.” If a variant is observed on only one strand of the DNA template, but not its complement, then this variant is interpreted to be an error and is discarded from further consideration. Due to this additional stringency for defining a true variant, duplex consensus sequencing is considerably more accurate for identifying extremely low-level VAF variants well below 0.1 percent (one in 1,000) and can reach one in 1 million, Dr. Wu says. “The cost, however, is that one has to sequence more deeply to identify the paired sequences, which may be more challenging to achieve.”

Next-generation sequencing is just one of the several methods being applied now for MRD detection. Conventional approaches such as RT-PCR are being complemented with newer approaches, such as RNA-Seq for fusion detection (Dillon LW, et al. Haematologica. 2019;104[2]:297–304), Droplet Digital PCR, and error-corrected methods of sequencing, as well as ultra-sensitive chimerism assessment as a surrogate for relapse and evaluation of circulating tumor DNA. While these methods have largely been tested in research studies, Dr. Wu says, “there is obvious interest in applying these technologies in the clinical realm moving forward.”

Detection of circulating tumor DNA is emerging as a potential biomarker that has now also been explored for hematopoietic neoplasms, including AML. For example, Sousuke Nakamura, MD, and colleagues of the University of Tokyo recently demonstrated in AML/MDS patients who underwent stem cell transplantation that “in fact you can achieve the same sensitivity as bone marrow testing,” Dr. Wu says (Nakamura S, et al. Blood. 2019;​133[25]:2682–2695). In their work, the authors collected tumor and available matched serum samples from 53 patients at diagnosis and post-transplant. After identifying driver mutations in 51 patients using NGS, the authors then designed at least one patient-specific Droplet Digital PCR assay for each patient. Diagnostic ctDNA and matched tumor DNA exhibited excellent correlations with variant allele frequencies upon testing, and both mutation persistence in bone marrow post-allogeneic stem cell transplantation and corresponding ctDNA persistence in the matched serum were comparably associated with higher three-year cumulative incidence of relapse. This approach thus appears promising, Dr. Wu says, and could be advantageous for patients, particularly for those who may have poor marrow cellularity and low blood count recovery post-treatment.

Another emerging approach is the use of blocker displacement amplification probes to enhance detection of variant alleles, as developed by David Zhang, PhD, and his research group at Rice University. Dr. Zhang’s strategy is based on blocking amplification of the wild-type allele, resulting in potential variant enrichment by several hundredfold to enable rare variant detection below 0.1 percent VAF, using low read-depth sequencing on the order of about 300× coverage, versus the higher depth of coverage typically needed for consensus sequencing-based approaches. In this work, Dr. Zhang’s group showed the detection of single nucleotide polymorphisms with a VAF of approximately 0.02 percent in a multiplexed panel with limited sequencing coverage (Song P, et al. Nat Biomed Eng. 2021;​5[7]:690–701). Though this approach can readily enhance sequencing of multiplex hotspots using a low-depth sequencing method, Dr. Wu says, it may be more difficult to target the full coding regions of genes such as needed for some genes like TP53. “This approach nevertheless is likely one to have potential clinical relevance,” Dr. Wu says, “as many clinical labs do not typically have the scale of DNA sequencers that can achieve the depth of coverage needed for sequencing a comprehensive panel of AML gene targets for MRD testing.”

Evaluation using Droplet Digital PCR is another approach for MRD detection. ddPCR is a highly accurate molecular approach that uses microfluidics to partition a sample into tens of thousands of discrete reaction chambers for PCR analysis and subsequent discretized detection and quantitation. A notable advantage of ddPCR, Dr. Wu says, is its reliance on using Poisson counting statistics for quantitation of rare events, and thus it does require external standards for quantitation as does RT-qPCR.

Many in the field, including investigators at the University of Michigan, have shown the potential of using ddPCR to monitor patients post-therapy in AML with a limit of detection reported in their work as low as 0.002 percent VAF, he says (Parkin B, et al. J Clin. Invest. 2017;​127[9]:3484–3495). The authors used patient-specific assays targeting on average about two to three mutations per patient and showed the potential to detect clones at very low levels. ddPCR is an important platform for labs to consider, Dr. Wu says, adding it’s an approach that is somewhat constrained, however, “by the fact that only a few targets can be tested simultaneously, and as such it’s harder to conceive of developing a broad, multigene panel-based test for MRD testing using ddPCR technology alone.” Nevertheless, he says, clinical labs have developed assays using ddPCR to target frequent gene mutations, such as in NPM1, a gene mutated in nearly 20 to 25 percent of normal karyotype AML (Mencia-Trinchant N, et al. J Mol Diagn. 2017;19[4]:537–548), as well as in IDH1 and IDH2 (Ferret Y, et al. Haematologica. 2018;​103[5]:822–829).

As another approach for MRD detection, labs have increasingly turned toward sensitive molecular tests to assess chimerism for patients who have undergone stem cell transplantation as a way to monitor engraftment (Khan F, et al. Bone Marrow Transplant. 2004;​34[1]:1–12). These approaches are limited to patients who are post-allogeneic transplantation. “In this approach, molecular assays are designed to target and quantitate discordant alleles—either single nucleotide variants or other insertion/deletion mutations or copy number polymorphisms—that differ between the patient and donor. In this way, quantitation of the host cells or donor cells may inform the status of the stem cell engraftment and serve as a surrogate for leukemia relapse,” Dr. Wu says, noting various groups have achieved this using RT-qPCR approaches. More recent work by some, including by Dr. Wu and UW colleague Stephen Salipante, MD, PhD, used single-molecule molecular inversion probes and NGS to target deletion copy number polymorphisms (Wu D, et al. Clin Chem. 2018;64[6]:938–949). Dr. Wu and others envision such ultrasensitive chimerism tests as complements to other approaches for MRD assessment.

Lastly, for myeloid MRD monitoring, NGS detection of insertion/deletion mutations can be sensitive without a need for significant bioinformatic or technical effort. Unique to insertion/deletion mutations, such as in NPM1 and FLT3, is that the background error profile by NGS is quite clean, so that deep sequencing of these specific gene mutations can be performed without the need to use complex error correction approaches, such as is required for detecting SNVs. Currently, many labs use RT-PCR approaches for detecting NPM1 gene mutation (a 4-base pair insertion), as highlighted in the seminal study in which an RT-qPCR approach was used (Ivey A, et al. N Engl J Med. 2016;374[5]:422–433). However, NGS approaches can detect this same NPM1 insertion mutation, Dr. Wu says, “because the common NPM1 4-base pair insertion mutation does not commonly occur as sequencing or PCR artifact by chance.” An advantage of an NGS approach for detecting NPM1 mutation is that one can monitor MRD without a priori knowledge of NPM1 allele and therefore can capture all of the different NPM1 mutations, as well as assess clonal evolution (Thol F, et al. Genes Chromosomes Cancer. 2012;51[7]:689–695; Salipante SJ, et al. Mod Pathol. 2014;27[11]:1438–1446; Bacher U, et al. Haematologica. 2018;​103[10]:​e486–e488).

As several groups have also shown, the ability to deeply sequence NPM1 mutation allows for potential MRD monitoring in a substantial proportion of normal karyotype AML patients, similar to RT-PCR (Patkar N, et al. Oncotarget. 2018;9[93]:36613–36624; Ritterhouse LL, et al. Mol Diagn Ther. 2019;​23​[6]:791–802). A comparable approach for detecting FLT3-internal tandem duplication mutations using highly sensitive NGS has been described (Blatte TJ, et al. Leukemia. 2019;33[10]:2535–2539).

As per the 2018 European LeukemiaNet Working Party guide­lines, post-treatment testing for MRD is now standard of care in AML. Patients with mutant NPM1, RUNX1-RUNX1T1, CBFB-MYH11, or PML-RARA typically should have molecular testing for post-treatment monitoring. For other subtypes of AML, particularly normal karyotype AML, flow cytometry is commonly used. “It is hopeful that next-generation sequencing can play an increasing role,” Dr. Wu says. The field still has important work to do, with clinical colleagues, to ensure assay performance including accuracy of results, to define the clinical validity of reported variants, and to optimize time-points for testing. Many groups worldwide are advancing these efforts, Dr. Wu says, citing a recent review (Yoest JM, et al. Front Cell Dev Biol. 2020;8:249).

While one goal of clinical testing could be to detect only those mutations present at AML diagnosis, Dr. Wu’s view is that NGS is likely to be used “to detect any variant clone reliably in the post-treatment context using generic panel-based tests.” The challenge, he says, is developing an appropriate lab infrastructure to sequence broadly (as many AML genes as possible) and deeply enough (beyond a VAF of 0.1 percent), while minimizing false-positives and defining the clinical significance of variants that are most likely to correlate with imminent risk for disease relapse in a relevant clinical time frame.

“And all of this has to be done with a reasonable turnaround time,” he says, “and with the typical challenges of cost and oftentimes a lagging reimbursement landscape.”

Charna Albert is CAP TODAY associate contributing editor.