Supplementary code repository for: A stem cell differentiation model reveals two alternative fates in CBFA2T3::GLIS2-driven acute megakaryoblastic leukemia initiation

Authors: Mohamed R. Shoeb,¹ Dagmar Schinnerl,¹ Lisa E. Shaw,² Matthias Farlik,² Sabine Strehl,¹ Florian Halbritter,^1,# Klaus Fortschegger^1,#

Affiliations:

¹St. Anna Children’s Cancer Research Institute (CCRI), Vienna, Austria

²Medical University of Vienna, Department of Dermatology, Vienna, Austria

Abstract:

The CBFA2T3::GLIS2 (CG) fusion protein causes aggressive pediatric acute megakaryoblastic leukemia (AMKL). Although dysregulated molecular pathways in AMKL have been identified, their role in early pre-leukemic transformation remains poorly understood. We developed a disease model utilizing genetically modified human induced pluripotent stem cells (hiPSC) physiologically and conditionally expressing CG. Using in vitro differentiation and single-cell multi-omics, we captured the impact of oncogene activity on gene-regulatory networks during hematopoiesis. We discovered that CG interferes with myelopoiesis through two alternative routes: by locking aberrant megakaryocyte progenitors (aMKP) in a proliferative state, or by impeding differentiation of aberrant megakaryocytes (aMK). Transcriptionally and functionally, aMKPs mimic CG-AMKL cells and establish a self-renewal network with co-factors GATA2, ERG, and DLX3. In contrast, aMKs partially sustain regulators of MK maturation but fail to complete differentiation due to repression of factors like NFE2, SPI1, GATA1 and LYL1. These insights may inform new strategies for targeting AMKL cell states.

Repository structure:

• project.Dockerfile defines the environment used to carry out all analyses.
• config.yaml is used to set paths
• _targets scripts define the analysis pipeline
• src/ holds the scripts for functions utilized in the _targets file, manuscript_figures generating the figures, and raw data preprocessing.
• docker/ holds shell scripts to build and run the docker image, and to parse the config file

Analysis workflow:

We use targets workflow to coordinate and automate the different steps of the analysis pipeline. The _targets is the main file that sets the analysis pipeline and define the generated R objects. The functions used to process objects in pipeline are defined in src/ and separated by data type: atac-seq/, chip-seq/, multiome/, rna-seq/, scrna-seq/.

Instructions

1. Edit project_config.yaml

2. Compile docker image project.Dockerfile

3. Download and preprocess the data using the scripts in src/

4. Start container:

src/docker/run_docker_rstudio.sh PORT PWD

5. Use the function tar_visnetwork() to visualize the pipeline graph and targets::tar_make() to run the targets pipeline. The output of the pipeline will be saved to _targets/ folder.

Links:

• Paper: 10.1038/s42003-025-08730-4.
• Gene Expression Omnibus (GEO) entry: GSE277004
• The scRNA-seq data embeddings generated in this study can also be explored interactively at the following URLs: UMAP of control reference (Fig3) and Harmony of megakaryocyte subset (fig4)