Deconvoluting Clonal and Cellular Architecture in IDH-Mutant Acute Myeloid Leukemia

Introduction

Mutations in Isocitrate dehydrogenase 1/2 ( IDH) occur in 20% of acute myeloid leukemias (AML) and frequently co-occur with DNMT3A, NPM1, SRSF2, and NRAS. In clonal hematopoiesis (CH), IDH mutations confer high risk of AML transformation. Recently developed methods linking single cell genotype to single cell transcriptomic profiles and cell lineage offer the opportunity to define the transcriptomic consequences of each mutation across leukemic cell lineages and clonal phylogenies. Here, we develop an integrative single cell laboratory and analytical framework to study the genetic, clonal and cellular heterogeneity of IDH mutated CH and AML, across disease stages (diagnosis and relapse) and deliver a high-resolution map of IDH-mutant AML.

Methods

Bone marrow or peripheral blood mononuclear samples from 15 AML patients, 2 CH individuals, and 12 unaffected age-matched donors were profiled by single cell RNA-sequencing. For the AML cases, a median of 3 samples per patient were profiled under treatment with IDH inhibitor (IDHi) monotherapy (n=5 patients), IDHi in combination with intensive chemotherapy (NCT02632708, n=6) or in combination with hypomethylating agent (n=2). This dataset comprised 259,610 cells across AML (n=190,133), CH (n=16,705) and normal controls (n=52,772). Genotyping of Transcriptomes (Nam et al., 2019) (for IDH1/2, NPM1, SRSF2, and NRAS mutations) across 6 AML patients enabled genotyping 23% of cells at diagnosis for at least one clone-defining mutation.

Results

Through meta-analysis of 3,653 diagnostic AML samples profiled by a targeted DNA panel (Papaemmanuil et al., 2016; Tazi et al., 2022), we identified DNMT3A, SRSF2 and NPM1 as critical co-operative mutations in IDH-mutant AML. Amongst the 15 AML patients, 6 had IDH1 mutation, 8 had IDH2 mutations, and 1 had both IDH1 and IDH2. The co-occurring mutations in DNMT3A (n=11 patients), NPM1 (n=5), SRSF2 (n=5) reflected the genomic landscape of IDH-mutant AML.

At diagnosis, IDH-mutant AML was enriched in immature hematopoietic stem and progenitor-like blasts. Compared to normal hematopoiesis, IDH-mutant AML blasts upregulated inflammatory response markers including TNFɑ/NFκB-signaling and downregulated proliferation-associated pathways including G2M Checkpoint, and MYC and E2F Targets, and DNA damage response.

To investigate how co-mutations modify the IDH-mutant phenotype, we established an integrative single-cell profiling approach to assign single cells to distinct phylogenetic subclones and extract clone-specific transcriptional programs. SRSF2 and NPM1 co-mutations modulated the AML blast phenotype, towards erythroid ( SRSF2) or granulomonocytic lineage ( NPM1). NRAS- or SRSF2-mutant subclones defined distinct gene expression-derived clusters and upregulated proliferation and metabolic pathways. Furthermore, NPM1 co-mutated blasts upregulated KMT2A-target genes.

Next, we tracked clonal responses to combination IDHi and chemotherapy using GoT and targeted DNA sequencing. In 4 cases, the ancestral clone, defined by IDH and DNMT3A mutations, retained clonal dominance in both mature and immature myeloid compartments at complete response (median 4 weeks). Conversely, subclones defined by mutations in NPM1, NRAS, or SRSF2, were either reduced or not detectable (limit of detection 2%). In two cases, mutations were not detected after therapy.

In patients who relapsed after IDHi monotherapy treatment, blasts upregulated stemness, fatty acid metabolism, OxPHOS, and inflammation at relapse compared to diagnosis. Furthermore, MHC Class II was downregulated at relapse, pointing to immune evasion of AML as a potential relapse mechanism.

Finally, we investigated the role of IDH2 mutations at CH. IDH-mutant CH-derived monocytes upregulated inflammation-associated pathways, mirroring the phenotypes observed in IDH-mutant AML at diagnosis.

Conclusion

We defined how IDH mutations and their frequent co-mutations dysregulated hematopoiesis. We leveraged state-of-the-art single cell techniques to assign leukemic cells to phylogenetic subclones and delineated the role of serially acquired co-mutations in shaping lineage bias and transcriptional programs. Furthermore, we defined genetic, clonal and transcriptional changes at three stages of the AML disease process: diagnosis, remission, and relapse under IDH inhibitor therapy.