https://orcid.org/0000-0001-8414-253X
TO THE EDITOR:
The bone marrow (BM) and peripheral blood (PB) create different microenvironments for acute myeloid leukemia (AML) cells.1 Moreover, AML remodels the BM microenvironment to create a proinflammatory environment favoring malignant over normal hematopoietic cells.2
The BM microenvironment supports AML cell resistance to therapy through direct and indirect interactions among the stromal and leukemic cells and metabolic and signaling modulations in hypoxic conditions.3 However, clonal diversity plays a key role in poor therapeutic outcomes in patients with AML.4 This observation is supported by several studies revealing AML-inducing somatic mutations (eg, DNMT3A, TP53, KDM6A) are associated with drug resistance.5-7 Therefore, we evaluated whether a clonal composition of AML at diagnosis might contribute to different therapeutic outcomes in PB and BM.
Previous reports delivered conflicting results. A bulk RNA sequencing of the BEAT-AML cohort revealed some differences in cell type composition between PB and BM, which might lead to difference in response to treatment.8 However, whole-genome sequencing (WGS) followed by software analysis revealed similar clonal architecture in PB and BM in most patients with AML.9 Although small relative quantitative differences were detected in a few AML cases, all variants detected in BM were present in PB. WGS and targeted sequencing of clonal hematopoiesis revealed that variant allele frequencies of specific mutations (eg, DNMT3A, ASXL1) were highly concordant or displayed significant differences in PB and BM samples of the same patients.10,11
To pinpoint and compare the baseline of the clonal landscape in PB and BM of the karyotypically normal AML samples, we applied a single-cell targeted DNA sequencing (sctDNAseq). Lin-negative CD34+ cells from paired PB and BM samples obtained at diagnosis from 5 patients with AML were subjected to sctDNAseq using our expanded panel detecting myeloid somatic mutations.12 Results in Figure 1 clearly show that AML clonal composition is very similar in PB and BM. Modest differences were detected only in the small founder clones (eg, TTN R17838H in 16888, KMT2A E502K in 18037, and SMC1A G1139W in 17550), which may represent clonal hematopoiesis. Because the founder clones are usually not equivalent to the AML-initiating clones,9 these differences may not have a significant impact on the treatment outcomes. Moreover, no major variant allele frequency variations were detected between the clones from the corresponding PB and BM samples. Therefore, we postulate that BM represents a microenvironmental, but not a clonal, challenge for the treatment of karyotypically normal AML at diagnosis.
Clonal composition of paired AML samples of PB and BM. AML samples 16888, 17385, 17550, 17570, and 18037 were from the ECOG-ACRIN E1900 clinical trial.13 Lin-negative CD34+ cells were obtained from mononuclear fractions by magnetic sorting using the EasySep Lin-negative selection cocktail followed by CD34+ selection (StemCell Technologies). sctDNAseq was performed as described before using our custom-made expanded myeloid panel.12 Briefly, sequencing data generated from the Tapestri platform were processed using Mission Bio’s Tapestri Pipeline (Tapestri 2.0.2) for adapter trimming (Cutadapt), sequence alignment (reference genome hg19), barcode correction, cell finding, and variant calling (GATK). Annotations for the filtered variants were curated using the Integrative Genomics Viewer. Initial clonal architectures were determined using genotype clustering analysis, including zygosity information with the Tapestri Insight software package. We used the SCITE software to infer phylogenetic trees of the driver mutations from the sctDNAseq data. SCITE phylogenetic inference is based on Bayesian approach which will allow us to quantify uncertainty in the inferred clonal architectures by sampling trees based on the model’s posterior distribution. Steps were performed using the methodology described previously.12 SCITE software was run with a chain length of 900 000 for each repetition. We used an estimated allele dropout rate of 4.5% and a false-positive rate of 1.0%. Further data analysis was performed using a customized R script. Left panels: The phylogenetic trees visualize the predicted evolutionary descent of the AML clones based on sctDNAseq data. The connecting lines represent the link between the consecutive clones. Right panels: The circles illustrate the clone sizes. The percentage of cells carrying indicated mutations (clones) is illustrated.
Clonal composition of paired AML samples of PB and BM. AML samples 16888, 17385, 17550, 17570, and 18037 were from the ECOG-ACRIN E1900 clinical trial.13 Lin-negative CD34+ cells were obtained from mononuclear fractions by magnetic sorting using the EasySep Lin-negative selection cocktail followed by CD34+ selection (StemCell Technologies). sctDNAseq was performed as described before using our custom-made expanded myeloid panel.12 Briefly, sequencing data generated from the Tapestri platform were processed using Mission Bio’s Tapestri Pipeline (Tapestri 2.0.2) for adapter trimming (Cutadapt), sequence alignment (reference genome hg19), barcode correction, cell finding, and variant calling (GATK). Annotations for the filtered variants were curated using the Integrative Genomics Viewer. Initial clonal architectures were determined using genotype clustering analysis, including zygosity information with the Tapestri Insight software package. We used the SCITE software to infer phylogenetic trees of the driver mutations from the sctDNAseq data. SCITE phylogenetic inference is based on Bayesian approach which will allow us to quantify uncertainty in the inferred clonal architectures by sampling trees based on the model’s posterior distribution. Steps were performed using the methodology described previously.12 SCITE software was run with a chain length of 900 000 for each repetition. We used an estimated allele dropout rate of 4.5% and a false-positive rate of 1.0%. Further data analysis was performed using a customized R script. Left panels: The phylogenetic trees visualize the predicted evolutionary descent of the AML clones based on sctDNAseq data. The connecting lines represent the link between the consecutive clones. Right panels: The circles illustrate the clone sizes. The percentage of cells carrying indicated mutations (clones) is illustrated.
However, PB and BM clonal compositions of more aggressive AML might display substantial differences. For example, unique and overlapping somatic mutations were detected by WGS in PB and BM from AML cases harboring mutations in Fanconi anemia genes and/or carrying chromosomal aberrations.14 In conclusion, similar clonal composition of karyotypically normal AML was detected in PB and BM at diagnosis, but the microenvironment-dependent unique clones may be detected when the disease displays high level of genomic instability. In addition, because we analyzed Lin-negative CD34+ cells, we cannot exclude that PB and BM differentially influence clonal outgrowth at later stages of AML differentiation.
Studies were approved by the ethics committee of Temple University Lewis Katz School of Medicine and met all requirements of the Declaration of Helsinki.
Acknowledgments: Single-cell targeted DNA sequencing was performed at the Single-Cell Multiomics Facility, Fels Cancer Institute for Personalized Medicine, Temple University Lewis Katz School of Medicine.
This work was supported by R01CA237286, R01CA244044, R01244179, and R01CA1247707 (T.S.).
Contribution: M.M.T. and M.N.-S. performed the experiments; A.K. provided bioinformatic analysis; and T.S. conceived the idea, supervised the project, and wrote the manuscript.
Conflict-of-interest disclosure: The authors declare no competing financial interests.
Correspondence: Tomasz Skorski, Fels Cancer Institute for Personalized Medicine, Lewis Katz School of Medicine, Temple University, MRB, Room 548A, 3400 N Broad St, Philadelphia, PA 19140; email: tskorski@temple.edu.
