Table 2 Summary of key findings related to heterogeneity indicators or classifying frameworks in leukemia by single-cell sequencing

AML

scRNA-seq; Targeted DNA sequencing; Single-cell short/long read sequencing

Machine learning was performed on high-throughput single-cell data and identified six malignant AML cell types, HSC-like, progenitor-like, GMP-like, promonocyte-like, monocyte-like, or cDC-like malignant cells, along the HSC to myeloid axis.

Related the AML developmental hierarchies to genotypes, providing information on how primitive AML cell types are prognosis informative.

[58]

AML

Bulk transcriptome deconvolution using single-cell references

AML hierarchy was subtyped into four overall classes, spanning Primitive, Mature, GMP, and Intermediate. LSPC cells were divided into Quiescent, Primed, and Cycling LSPC.

Noted that Primitive vs. GMP axes are chemotherapy responsive whereas Primitive vs. Mature axes is associated with drug sensitivity.

[59]

AML

scRNA-seq; SMRT-seq

AML progenitor cells cluster with novel AML markers associated with dysregulated RP expression were identified.

Highlighted that the high ribosomal protein involved in the p53 pathway in the progenitor cells subtype was associated with poor outcome.

[60]

cALL

scRNA-seq

Ribosomal protein expression profile is distinctive and inversely correlated with the presumptive ALL developmental state.

Highlighted that ribosomal protein may be considered as a marker for intra-individual heterogeneity in cALL patients.

[61]

AML

scRNA-seq

C1Q + macrophage-like leukemia subset was identified and verified in multiple patients with AML. C1Q + leukemia cells represent a highly tissue-infiltrative leukemia population and could reconstitute extramedullary infiltration phenotype of AML. C1Q interacts with C1Q–globular C1Q receptor on fibroblasts, regulating the cancer infiltration pathways and promoting the chemoresistance of C1Q + leukemia cells.

Put forwarded that C1Q can serve as a marker for AML with adverse prognosis and the cancer infiltration pathways. Also, C1Q is a great therapeutic target.

[62]

AML

scRNA-seq

NFIC protein is significantly overexpressed in 69% of acute myeloid leukemia patients, and increased expression of growth and survival genes in monocytes. NFIC knockdown in an ex vivo mouse a pre-leukemic stem cell model decreased their growth and colony formation and increased expression of myeloid differentiation markers Gr1 and Mac1.

Noted that NFIC is an important transcription factor in myeloid differentiation as well as AML cell survival, and is a potential marker for therapeutic targeting in AML.

[63]

CLL

scRNA-seq; Whole-genome bisulfite sequencing

High level of methylation heterogeneity in CLL arose from stochastic methylation dysregulation.

Identified that dysregulation of methylation is associated with poor prognostic outcome in CLL patients.

[67]

CLL

Multiplexed single-cell reduced representation bisulfite sequencing (MscRRBS); scRNA-seq; ChIP-seq

Coordination between different layers of CLL epigenome layers and epigenomic expression was disrupted, attributing to cell-cell heterogeneity.

Noted that corrupted epigenetic layers residing in CLL may stochastically activate heterogenous expression programs, associating poor prognosis.

[68]

Computational framwork

Developed sc-compReg for comparative analysis between disease and healthy samples based on scRNA-seq data and scATAC-seq data

Sc-compReg in CLL samples identifies TOX2 as a key regulator of tumor-specific subtypes.

Enabled the integrative comparison between healthy and disease states based on transcriptomic and chromatin accessibility. Further application in other leukemia subtypes could review more distinct subtypes.

[69]

MPAL, AML

CITE-seq; scRNA-seq; scATAC-seq

Single-cell epigenetics baseline for healthy blood samples was established, which was used to deconvolve aberrant molecular features of MPAL. 91,601 putative peak-to-gene linkages and transcription factors regulating leukemia-specific genes were identified.

Demonstrated that single-cell multiomics study may provide novel shared molecular mechanisms among different leukemia types for clinical targeting.

[35]

Experimental and Computational framework

Developed MutaSeq and mitoClone for single-cell targeted mutation analysis of nuclear and mitochondrial genes on scRNA-seq data.

Application of Mutaseq and MitoClone in AML implied that LSC, HSC and pre-LSC can be more confidently distinguished based on the combination of transcriptome, genetic and mitochondrial variants. Genetic mutations can distinguish between healthy and diseased states, and expression profiles can identify stem or progenitor cell states.

Demonstrated that mitochondrial mutations may also indicate leukemia heterogeneity and underlie therapeutic targets.

[75]

Experimental and Computational framework

Developed Optimized 10X and CloneTracer for clone tracing specifying nuclear and mitochondrial mutation on scRNA-seq data.

Application of CloneTracer to 19 AML patient samples revealed healthy or preleukemic state in a dormant HSC subset. Discovered that LSCs resemble HSCs expression but formed differential-blocked aberrant myeloid progenitors in downstream.

Demonstrated that mitochondrial mutations may also indicate leukemia heterogeneity and underly therapeutic targets. LSCs may be distinguished from HSCs by forming aberrant myeloid progenitors in downstream.

[76]