Genomic profiling for clinical decision making in myeloid neoplasms and acute leukemia

Abstract

Myeloid neoplasms and acute leukemias derive from the clonal expansion of hematopoietic cells driven by somatic gene mutations. Although assessment of morphology plays a crucial role in the diagnostic evaluation of patients with these malignancies, genomic characterization has become increasingly important for accurate diagnosis, risk assessment, and therapeutic decision making. Conventional cytogenetics, a comprehensive and unbiased method for assessing chromosomal abnormalities, has been the mainstay of genomic testing over the past several decades and remains relevant today. However, more recent advances in sequencing technology have increased our ability to detect somatic mutations through the use of targeted gene panels, whole-exome sequencing, whole-genome sequencing, and whole-transcriptome sequencing or RNA sequencing. In patients with myeloid neoplasms, whole-genome sequencing represents a potential replacement for both conventional cytogenetic and sequencing approaches, providing rapid and accurate comprehensive genomic profiling. DNA sequencing methods are used not only for detecting somatically acquired gene mutations but also for identifying germline gene mutations associated with inherited predisposition to hematologic neoplasms. The 2022 International Consensus Classification of myeloid neoplasms and acute leukemias makes extensive use of genomic data. The aim of this report is to help physicians and laboratorians implement genomic testing for diagnosis, risk stratification, and clinical decision making and illustrates the potential of genomic profiling for enabling personalized medicine in patients with hematologic neoplasms.

Figure 1.

How molecular profiling can inform clinical decision making in MDS. IPSS-M, Molecular International Prognostic Scoring System; MDS, myelodysplastic syndrome; MRD, minimal/measurable residual disease. Professional illustration by Patrick Lane, ScEYEnce Studios.

Figure 2.

Key concepts in sequencing-based diagnostics. (A) VAF represents the ratio of sequencing reads that contain a variant divided by the total number of reads at that position. Because most somatic mutations are heterozygous, doubling the VAF generally indicates the fraction of cells with the mutation (except when the mutation occurs in a region of copy number alteration). (B) Coverage represents the number of sequencing reads (red and blue indicating forward and reverse reads, respectively) that span a particular region. Approximate coverage levels for different sequencing approaches are compared. Higher coverage (or more independent observations) generally yields more sensitive sequencing. Shown on the right is the coverage depth required to detect mutations at various VAFs. Binomial sampling probability for detection of variants with VAFs of 50% (typical inherited variants; black), 2% (general sensitivity for targeted panels; red), and 0.1% (MRD assays; blue) assuming each variant must be seen at least twice. (C) DNA-sequencing methods. In WGS, libraries are created by ligating sequencing adapters (gray and orange) to the 3′ and 5′ ends of short genomic DNA fragments called “inserts.” Gene panels or exome sequencing enriches DNA of interest form a library using antisense capture probes (green) labeled with biotin, which are then hybridized to DNA inserts from the sequencing and then physically enriched using streptavidin-coated magnetic beads. (D) High-sensitivity sequencing for MRD detection requires error correction to reliably identify mutations below the intrinsic error rate of the sequencer and to account for PCR errors. Error-corrected deep sequencing reduces false-positive calls for low VAF variants by tagging individual DNA molecules with unique molecular identifiers (UMIs). In this example a “true” mutation “T” is present in a single DNA molecule that labeled with a UMI (green). Library amplification and sequencing will result in duplicate DNA molecules each labeled with the same UMI. Randomly accumulated sequencing and PCR errors (orange) will be present in only a subset of reads with a particular UMI (green, purple, red). During sequencing analysis, variants present on only a subset of reads from a particular “read family” with the same UMI will be discarded as errors; true mutations present in the original DNA molecule will be detected in all reads within a read family with the same UMI. UMI methods can be further improved by tracking both DNA strands using “duplex sequencing,” which can yield sensitivities of 10⁻⁶. Professional illustration by Patrick Lane, ScEYEnce Studios.

Figure 3.

Sequencing-based tumor burden monitoring in myeloid neoplasms. (A) In this AML example, sequencing identifies NPM1 and DNMT3A mutations at diagnosis with the NPM1 mutation representing the founding clone (green: based on higher mutation VAF) and DNMT3A representing a subclone (gray: based on lower mutation VAF). Estimated sensitivity to detect mutations for different sequencing approaches is shown below. As the patient enters remission, the NPM1-mutated clone is partially cleared, becoming undetectable by panel-based sequencing and WGS but remains detectable by high-sensitivity MRD sequencing. In this example, the DNMT3A-mutated clone remains without significant change in VAF, consistent with persistent clonal hematopoiesis. Finally, the patient relapses with the same NPM1-mutated founding clone plus a newly acquired NRAS mutation. (B) A comparison of sequencing methods for MRD monitoring. Although WGS offers the greatest sequencing breadth and is capable of detecting structural variants such as CNAs and chromosomal translocations, standard coverage is generally only ∼60×, limiting detection to mutations with VAFs > 10%. Targeted sequencing is generally limited to 50 to 500 genes, providing minimum sequencing breadth but can achieve high coverages (1000×) at a lower cost than WGS and provides adequate sensitivity (2% VAF) for initial diagnostic evaluation. MRD panels are similar to targeted panels but use much higher sequencing depths and use UMIs to achieve sensitivities of ∼0.1% VAF allowing for MRD monitoring. MRD panels are generally easy to implement but may be of limited clinical utility for patients with few mutations covered by the panel. In patient-specific MRD sequencing, mutations are identified at diagnosis using broad methods such as exome sequencing or WGS, ensuring an adequate number of mutations to track. These mutations are then individually targeted using custom probes at subsequent time points. By focusing sequencing on known mutations, patient-specific MRD can achieve extremely high detection sensitivities for nearly all patients. Patient-specific MRD, however, can be logistically challenging and expensive to implement because it requires custom probe designs and validation for every patient. Patient-specific methods also cannot detect newly acquired mutations that were not targeted by probes at diagnosis. Professional illustration by Patrick Lane, ScEYEnce Studios.

Figure 4.

Identification of distinct subtypes of ALL through gene expression profiling. (A) Representative break-apart FISH for KMT2A::AFF1 fusion. The upper panel shows a cell with DNA FISH for KMT2A 5′ and 3′ showing 1 intact allele and 1 disrupted allele. The lower panel shows a second hybridization added on top of the first with AFF1 3′, which confirms disruption of KMT2A and fusion to AFF1 3′. (B) Illustration showing overexpression of CRLF2 and detection by flow cytometry. The image was created in Biorender (https://biorender.com/). (C) Schematic representation of NUTM1 rearrangements with multiple fusion partners and multiple breakpoints detected by WTS and visualized in ProteinPaint (https://proteinpaint.stjude.org/). Ex, exon. The approach in parenthesis (WGS) is alternative to WTS. (D) Integrative Genomics Viewer visualization of BCL11B Enhancer Tandem Amplification (BETA), observed in 20% of BCL11B-activated lineage ambiguous leukemia. (E) t-distributed stochastic neighbor embedding (t-SNE) representation from WTS data of B-ALL subtypes highlighted in different colors. Each dot represents a sample (N = 2004). Image is from Kimura et al.