-Author name in bold denotes the presenting author
-Asterisk * with author name denotes a Non-ASH member
Clinically Relevant Abstract denotes an abstract that is clinically relevant.

PhD Trainee denotes that this is a recommended PHD Trainee Session.

Ticketed Session denotes that this is a ticketed session.

63 Abundance of Relapse-Predictive Cells Can be Estimated at Diagnosis and Is Strongly Associated with Outcome in Pediatric AML

Program: Oral and Poster Abstracts
Type: Oral
Session: 618. Acute Myeloid Leukemias: Biomarkers and Molecular Markers in Diagnosis and Prognosis: Multi-omic Applications for Disease Evolution and Response to Therapy
Hematology Disease Topics & Pathways:
Research, Acute Myeloid Malignancies, AML, Translational Research, Genomics, Pediatric, Diseases, Myeloid Malignancies, Biological Processes, Molecular biology, Study Population, Human, Animal model
Saturday, December 7, 2024: 10:00 AM

Mohammad Javad Najaf Panah, MS1*, Alexandra McLean Stevens, MD2, Michael Krueger1*, Max P Rochette1*, Sohani K Sandhu1*, Lana E Kim1*, Joanna S. Yi, MD3, Tsz-Kwong Man, PhD1*, Michele S. Redell, MD1 and Pavel Sumazin, PhD1*

1Texas Children's Cancer Center, Baylor College of Medicine, Houston, TX
2Baylor College of Medicine/ Texas Children's Hospital, Houston, TX
3Texas Children's Cancer and Hematology Centers, Baylor College of Medicine/Texas Children's Hospital, Houston, TX

Background

Risk assessment for pediatric AML (pAML) determines which patients receive chemotherapy only and which receive a stem cell transplant in first remission. Current risk algorithms use cytomolecular features (cyto) and measurable residual disease at the end of Induction 1 (EOI1 MRD). Nevertheless, 30% of low risk patients relapse. Distinct AML subclones can be identified at diagnosis. At relapse, chemosensitive subclones are diminished and chemoresistant subclones are expanded. We applied single-cell RNA-sequencing (scRNA-seq) to 13 pAML diagnosis-relapse pairs to identify and characterize chemoresistant cells.

Methods

Diagnosis-relapse pairs were obtained from 6 local cases and 7 Children’s Oncology Group (COG) cases. All patients were treated on or following the COG Phase 3 trial AAML1031. 10K-15K cells were prepared for scRNA-seq using the 10X Genomics Chromium platform. We inferred the abundance of cell types of interest in public bulk RNA-seq profiles using the SQUID framework (PMC10394903). Patient-derived xenograft (PDX) models were treated with cytarabine (50 mg/kg/dose) or saline on days 1-4. Residual human AML (hAML) was quantified in the bone marrow on day 8.

Results

Normalized scRNA-seq data were integrated and clustered. We merged clusters with similar pseudobulk profiles to yield 90 subclones. We applied SQUID-based deconvolution to pAML RNA-seq data for 870 FLT3-ITD-negative cases in TARGET to estimate the abundance of each subclone in these cases. We identified 5 relapse-predictive subclones (R1-R5) whose abundance was significantly associated with event-free and overall survival (EFS, OS). Considering the total (T) of R1-R5 cell types in aggregate, TARGET cases with T above the median of 7% had a 3 yr OS of 54.4% v. 79.6% for those with T below the median (log rank p<10-15).

We combined T abundance (above or below the median) with cyto and/or EOI1 MRD risk. We considered t(8;21), inv(16) and normal karyotype as low risk, and KMT2A fusions and other cytogenetics as high risk. Combining T abundance with cyto and MRD risk significantly improved outcome prediction. T abundance was most valuable when cyto and MRD risk conflicted. For example, among cases with high risk cyto and negative (low risk) MRD, those with low T (n=97) had a 3 yr EFS of 78%, whereas those with high T (n=226) had a 3 yr EFS of 54%. Multivariate Cox proportional hazard model analysis indicated that after controlling for cyto and MRD, T abundance remained significantly predictive with a hazard ratio of 1.3 (p=3.69e-05).

To further evaluate the association of the R subclones with chemoresistance, we treated 8 PDX models with cytarabine or saline and assessed residual hAML by flow cytometry, RNA-seq and scRNA-seq. Two models (AML005 and AML006) showed a significant decrease in %hAML with cytarabine, 2 had intermediate responses, and 4 models had no difference in %hAML between the cytarabine and saline groups. T abundance was <3% in AML005 and AML006, and up to 15% in the other models. By scRNA-seq we discriminated 3 clusters in AML005 and AML006. R2, R3 and R4 cell types were identified in the clusters most enriched for cytarabine-treated cells (i.e. chemoresistant), but not in the clusters depleted of cytarabine-treated cells. These results support the contention that the R cell types are chemoresistant.

R populations were significantly enriched for genesets including OxPhos, MYC targets, and genes co-expressed with FLT3 and CDK6. Further, we identified transcription factors whose inferred activity was dysregulated in R1-R5. Retinoic X Receptor Alpha (RXRA) targets were significantly downregulated in the R5 subclone relative to a chemosensitive subpopulation from the same patient. The RARA/RXRA complex represses transcription in the absence of retinoic acid; transcription can be activated by the agonist tamibarotene (tami). We previously reported that tami reduced cell viability in vitro and prolonged PDX survival for AML001 but not AML006 (PMC9153032). Re-analysis of this RNA-seq dataset revealed that R5 abundance was high in AML001 at baseline and depleted after tami. The R5 cell type was not detected in AML006.

Conclusion

We identified 5 cell types whose abundance at diagnosis strongly predicted EFS and OS in pAML patients. We validated the association of these cell types with chemoresistance in PDX models. We identified retinoic acid receptor agonism as a potential strategy to deplete one of these cell types.

Disclosures: Stevens: Gilead Pharmaceuticals: Research Funding; AbbVie Pharmaceuticals: Research Funding. Yi: Syros: Consultancy.

*signifies non-member of ASH