AUTHORS
Mariateresa Giaimo, MD, PhD, Salvatore Benfatto, PhD2, Alexia C. Ferreira3, Maria Reyes Reyes Capilla-Guerra, MSc4, Til Steinicke5, Evan C. Chen, MD6, Kimberly Stegmaier, MD7, Giovanni Roti, MD, PhD8, Marian H. Harris, MD, PhD9, Harrison Tsai, MD, PhD10, Volker Hovestadt, PhD11, Gabriel Griffin, MD12, Yana Pikman, MD13
AFFILIATIONS
1 Dana-Farber Cancer Institute, Boston, Massachusetts, United States, University of Parma, Parma, Italy
2 Broad Institute of MIT & Harvard, Cambridge, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
3 Broad Institute of MIT & Harvard, Cambridge, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
4 Broad Institute of MIT & Harvard, Cambridge, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
5 Broad Institute of MIT & Harvard, Cambridge, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
6 Dana-Farber Cancer Institute, Boston, Massachusetts, United States
7 Boston Children’s Hospital, Boston, Massachusetts, United States, Broad Institute of MIT & Harvard, Cambridge, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
8 University of Parma, Parma, Italy
9 Boston Children’s Hospital, Boston, Massachusetts, United States
10 Boston Children’s Hospital, Boston, Massachusetts, United States
11 Boston Children’s Hospital, Boston, Massachusetts, United States, Broad Institute of MIT & Harvard, Cambridge, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
12 Brigham and Women’s Hospital, Boston, Massachusetts, United States, Broad Institute of MIT & Harvard, Cambridge, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
13 Boston Children’s Hospital, Boston, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
ABSTRACT
Acute Leukemia (AL) is the most common childhood cancer and a leading cause of cancer-related death. Timely genomic characterization is required to guide treatment selection and risk stratification, yet standard diagnostics are time consuming and sometimes fail to reveal genetic drivers. To address these problems, we recently established a clinically-applicable framework to classify AL using genome-wide DNA methylation profiling and a deep neural network classifier (MARLIN; Steinicke, Benfatto et al. Nature Genetics, 2025). Trained on >2,500 adult and pediatric AL samples, MARLIN distinguishes 38 methylation classes covering the majority of described molecular subtypes in Acute Myeloid Leukemia (AML, n=21), B-cell Acute Lymphoblastic Leukemia (B-ALL, n=11), T-cell ALL (T-ALL, n=5), Acute Leukemia of Ambiguous Lineage (ALAL, n=1), in addition to 4 normal control classes. When coupled to nanopore sequencing, MARLIN generates “real-time” classifications typically within 2 hours from sample receipt. However, MARLIN and nanopore sequencing have not yet undergone dedicated validation on a large pediatric acute leukemia cohort.
Here, we assessed 105 pediatric acute leukemia samples (n=52 AML, n=42 B-ALL, n=4 T-ALL, n=5 MPAL-T/myeloid, n=2 ALAL) from patients enrolled in the LEAP leukemia genomics study (NCT02670525), at diagnosis and at relapse (R/R). We systematically correlated MARLIN classifications with available clinical, pathology, and molecular information, including fluorescent in situ hybridization (FISH), cytogenetics, and DNA/RNA-based panel sequencing. Archival DNA from all samples was barcoded and subjected to multiplexed nanopore (PromethION platform) sequencing. Mean coverage over reference CpGs was 113,513 per sample (range 21,014 to 263,701), sufficient for MARLIN classification.
Overall, 76/105 (72.4%) of samples received a confident MARLIN classification score (≥0.8), including 31/37 (83.8%) diagnostic and 45/68 (66.2%) R/R cases. Among these cases, 72/76 (94.7%) were concordant with standard-of-care lineage assignments, including 31/31 (100%) diagnostic and 41/45 (91.1%) R/R samples. T-Myeloid MPALs were classified as T-ALL (3/4) or AML (1/4), while ALALs were classified as AML into groups consistent with known drivers (CBFA2T3::GLIS2 and KMT2A-r). The 4 remaining cases that were lineage discordant were assigned to control or control-adjacent methylation classes, suggesting low malignant cell content in sequenced samples.
For methylation-based molecular classifications, 68/72 (94.4%) lineage-correct cases were concordant with known genetics. For example, these included 7 AML cases with RUNX1::RUNX1T1 fusions, 4 AML cases with CEBPA mutations, and 3 B-ALL cases with KMT2A-r. Furthermore, MARLIN indicated a refined classification in 15 cases not identified by standard cytogenetics/FISH but ultimately validated by RNA-based fusion testing, including 2 Ph-like B-ALL cases, one with ZMYND8::JAK2 and another with RCSD1::ABL2, and 2 additional B-ALL cases, one with PAX5::FKBP15 and another with ZNF384::EP300. Likewise, MARLIN refined classifications in 10 AML samples, including 1 case classified as acute promyelocytic leukemia found to harbor a non-canonical TBL1XR1::RARB fusion, and 9 cases classified as distinct subtypes of HOX-activated AML that were consistent with known drivers (e.g., NPM1-mut or KMT2A-r) and co-mutations (e.g., IDH2-mut). In one of these samples, MARLIN indicated the HOX-activated Group 4 subtype often associated with NUP98::NSD1 fusions, which was subsequently confirmed by follow-up sequencing. Finally, in one T-ALL found to carry a MYB::TRB fusion, methylation-based classification into the TAL1 group refined risk stratification.
In summary, MARLIN and nanopore sequencing showed strong concordance with standard-of-care lineage and molecular classifications in a large cohort of pediatric acute leukemia. Refinement in lineage and/or molecular classifications was also achieved in some cases, particularly for categories defined by gene-fusions that are difficult to detect by routine cytogenetics/FISH. Thus, MARLIN and nanopore-based methylation profiling represent promising tools for rapid molecular classification and risk-stratification of pediatric acute leukemia. Evaluation in prospective clinical settings and in larger retrospective cohorts with annotated outcomes is needed for further validation.
DISCLOSURES
Mariateresa Giaimo, MD, PhD: No relevant conflicts of interest to declare.