To the editor:

Mixed phenotype acute leukemias (MPALs) are a rare subset of acute leukemias that cannot be unambiguously assigned to a single hematopoietic lineage because of the significant expression of antigens from additional lineages. This subset comprises several entities, including biphenotypic leukemias, in which leukemic blasts co-express markers of two lineages, and bilineal leukemias (BLLs), which demonstrate 2 immunophenotypically distinct populations, each belonging to a different lineage. These 2 entities may overlap significantly. How to properly treat MPALs is still an undecided question that can be addressed by combining international clinical experience with biological knowledge.1 

Whether the mixed phenotype of MPAL is driven by the bi- or multilineage potential of the cell of origin or whether it is triggered by 1 or more specific oncogenes is not yet fully understood; both mechanisms may potentially play a role. At the genomic level, acute leukemias often have a complex clonal architecture, and subclonal genetic lesions are a common phenomenon.2  In our study, we aimed to determine whether leukemia bilineality is triggered by a subclonally derived genetic aberration driving the transdifferentiation or reprogramming of individual subclones and transforming leukemia that was originally uniform into BLL.

In addition to standard routine molecular and immunologic diagnostics, we used high-throughput genomic methods to analyze the genomic landscape of immunophenotypically distinct subpopulations in 2 BLL patients (patient characteristics are provided in supplemental Table 1, available on the Blood Web site). Bulk leukemic samples and sorted subpopulations (preT and myeloid cells from patient 2073 and preT and biphenotypic preT myeloid cells from patient 2279; Figure 1A-B; supplemental Data) were analyzed via whole transcriptome sequencing (RNA sequencing [RNAseq]) and whole exome sequencing (WES); in addition, bulk leukemic samples were analyzed with whole genome single nucleotide polymorphism (SNP) arrays.

Figure 1

Immunophenotypic and gene expression patterns of distinct BLL subpopulations. Immunophenotypic analysis performed by flow cytometry revealed 2 distinct BLL subpopulations in (A) patient 2073 and (B) patient 2279. Cells in the representative dot plots are colored according to the expression of 2 other CD markers measured within the same tube. Although the 2 distinct subpopulations are clearly visible in patient 2073, the immunophenotypic architecture of the leukemic clone in patient 2279 is more complex (see supplemental Figure 1). However, by using a combination of several markers, which are displayed in the sorting algorithms below the dot plots, it was possible to sort 2 nonoverlapping subpopulations in both patients: preT and myeloid (My) in patient 2073 and biphenotypic preT-myeloid (preT-My) and preT in patient 2279. (*) The presence or absence of intracellular myeloperoxidase (iMPO) was determined in a different tube. (C) A heat map representing the expression levels of the 200 top-scoring genes that are differentially expressed in preT (2279-T, 2073-T) compared with myeloid/preT-myeloid (2279-M, 2073-M) subpopulations. Expression values were transformed into a scale ranging from −1 to +1; for each gene, the minimal value was set to −1, and the maximal value was set to +1. (D) The normalized expression levels (ncounts) of selected lineage-specific genes involved in the 200-gene set represented in panel C document the distinct lineage characteristics of sorted subpopulations.

Figure 1

Immunophenotypic and gene expression patterns of distinct BLL subpopulations. Immunophenotypic analysis performed by flow cytometry revealed 2 distinct BLL subpopulations in (A) patient 2073 and (B) patient 2279. Cells in the representative dot plots are colored according to the expression of 2 other CD markers measured within the same tube. Although the 2 distinct subpopulations are clearly visible in patient 2073, the immunophenotypic architecture of the leukemic clone in patient 2279 is more complex (see supplemental Figure 1). However, by using a combination of several markers, which are displayed in the sorting algorithms below the dot plots, it was possible to sort 2 nonoverlapping subpopulations in both patients: preT and myeloid (My) in patient 2073 and biphenotypic preT-myeloid (preT-My) and preT in patient 2279. (*) The presence or absence of intracellular myeloperoxidase (iMPO) was determined in a different tube. (C) A heat map representing the expression levels of the 200 top-scoring genes that are differentially expressed in preT (2279-T, 2073-T) compared with myeloid/preT-myeloid (2279-M, 2073-M) subpopulations. Expression values were transformed into a scale ranging from −1 to +1; for each gene, the minimal value was set to −1, and the maximal value was set to +1. (D) The normalized expression levels (ncounts) of selected lineage-specific genes involved in the 200-gene set represented in panel C document the distinct lineage characteristics of sorted subpopulations.

RNAseq data were used to analyze the gene expression profiles of sorted subpopulations. Comparing the preT subpopulations from both patients to the myeloid and preT myeloid populations, we defined a set of 200 top-scoring differentially expressed genes (Figure 1C; supplemental Table 4). Importantly, this gene set included several lineage-specific genes whose expression correlated with subpopulation immunophenotypes, thus further endorsing their differential lineage engagement as assigned by flow cytometry (Figure 1D).

Although patient 2073 had a complex karyotype with several chromosomal rearrangements (supplemental Table 2), in-frame fusion transcripts were not detected by RNAseq in either patient 2073 or patient 2279.

At the genomic level, we identified a total of 26 nonsynonymous mutations affecting gene coding regions in the bulk populations of patient 2073 and 11 in patient 2279 (Figure 2A-B; supplemental Table 5). These included mutations in recurrently affected genes in T-cell acute lymphoblastic leukemia3-6 : PHF6, STAT5B, WT1, NRAS, and FBXW7 in patient 2279 and MED12, NRAS, NOTCH1, CCND3, and EZH2 in patient 2073. Thus, both patients harbored mutations activating Notch1 signaling. Similarly, the NRAS gene was also affected in both patients, although a rather rare mutation, G60E, detected in patient 2279 was shown to have weaker activation potential compared with G13 mutations7  and was present only at the subclonal level (variant allele frequency <15%). The variant allele frequencies in the vast majority of mutations indicated their presence in the major proportion of bulk samples. Accordingly, all of these mutations were present in both sorted subpopulations. Importantly, there were no significant differences in the patterns of subclonal mutations in either of the patients.

Figure 2

Nonsynonymous mutations detected by WES and amplicon sequencing. Nonsynonymous mutations detected by WES in (A) patient 2073 and (B) patient 2279. The first column shows gene symbols, the second column shows detected mutations. Mutations with homozygous variant allele frequencies are displayed in dark red. These include hemizygous mutations of PHF6, MED12, and CLDN34 on chromosome X and the mutation of a single RREB1 allele on chromosome 6 (the second RREB1 allele is completely lost because of a monoallelic deletion on 6p identified by SNP array; supplemental Table 6). Mutations with the heterozygous variant allele frequency are displayed in light red, and subclonal mutations with variant allele frequencies <15% are displayed in pink. Genes that are recurrently mutated in T-cell acute lymphoblastic leukemia are in bold. Mutated alleles that were detectable in the RNAseq data are marked in black. The positions of the remaining mutations that were not detected in the RNAseq data were either not covered at all or were covered by only a low number of reads (supplemental Table 5). (C) Levels of 6 different mutated WT1 alleles and of germline allele were determined by amplicon sequencing in patient 2279. The y-axis shows the proportion of mutated and/or germline reads within the total number of reads, which was set to 100%. M, myeloid subpopulation in patient 2073 and preT-myeloid subpopulation in patient 2279; T, preT subpopulation.

Figure 2

Nonsynonymous mutations detected by WES and amplicon sequencing. Nonsynonymous mutations detected by WES in (A) patient 2073 and (B) patient 2279. The first column shows gene symbols, the second column shows detected mutations. Mutations with homozygous variant allele frequencies are displayed in dark red. These include hemizygous mutations of PHF6, MED12, and CLDN34 on chromosome X and the mutation of a single RREB1 allele on chromosome 6 (the second RREB1 allele is completely lost because of a monoallelic deletion on 6p identified by SNP array; supplemental Table 6). Mutations with the heterozygous variant allele frequency are displayed in light red, and subclonal mutations with variant allele frequencies <15% are displayed in pink. Genes that are recurrently mutated in T-cell acute lymphoblastic leukemia are in bold. Mutated alleles that were detectable in the RNAseq data are marked in black. The positions of the remaining mutations that were not detected in the RNAseq data were either not covered at all or were covered by only a low number of reads (supplemental Table 5). (C) Levels of 6 different mutated WT1 alleles and of germline allele were determined by amplicon sequencing in patient 2279. The y-axis shows the proportion of mutated and/or germline reads within the total number of reads, which was set to 100%. M, myeloid subpopulation in patient 2073 and preT-myeloid subpopulation in patient 2279; T, preT subpopulation.

Two different frameshift mutations in exon 7 of the WT1 gene were detected by WES in patient 2279. In addition, another frameshift mutation affecting this exon was also called, but it did not pass subsequent filtering criteria because it was supported by a single read only. Therefore, the mutational spectrum of WT1 exon 7 was re-analyzed via custom-designed deep amplicon sequencing. In addition to the 2 most abundant mutations detected by WES, we identified 4 additional frameshift mutations with variant allele frequencies of 0.8% to 4.6% in the bulk sample (Figure 2C). We suspect that the most abundant mutation inactivating 1 WT1 allele appeared early during leukemogenesis and was inherited by all blasts, whereas all of the other WT1 mutations affecting the second allele were subsequently gained in (nonoverlapping) subclones. Interestingly, all WT1 mutations, including those affecting the minor subclones, were detected in both sorted subpopulations. These data demonstrated the genetic heterogeneity of the leukemia in patient 2279 and showed that all distinct genetic subclones, even the rare ones, contributed to both immunophenotypically distinct subpopulations.

By using an SNP array, we detected a total of 3 regions involving copy number aberrations (CNAs) or uniparental disomy (UPD) in patient 2279 and 9 regions in patient 2073 (supplemental Table 6; supplemental Figure 1). There were 3 adjacent regions of loss on 9p in patient 2279, who presented with a normal karyotype; the middle region involved the loss of the MTAP, CDKN2A, and CDKN2B genes, which are recurrently deleted in T-cell acute lymphoblastic leukemia.8  In accordance with the complex karyotype reported by conventional cytogenetics, SNP array identified large aberrations on 5qter, 6pter, 12p, and 16qter in patient 2073. In addition, this patient carried 3 focal aberrations on chromosomes Y, 9, and 3, the latter resulting in a loss of the TBL1XR1 gene, which is associated with resistance9  to glucocorticoids in acute lymphoblastic leukemia cell line–based models. All 3 CNAs in patient 2279 and 7 of 9 CNAs and/or UPDs in patient 2073 were present at levels corresponding to the involvement of whole leukemic populations. However, the deletion on 3q and the aberration on 9p (proximal to the CDKN2A/CDKN2B gene region but not affecting any gene) in patient 2073 exhibited a mosaic character, indicating their subclonal presence. We did not have sufficient material from the sorted subpopulations to carry out SNP array analysis; however, we used WES data to analyze the presence of the CNAs and/or UPDs (identified by SNP array in bulk samples). We found that all CNAs and/or UPDs, including the 2 subclonal aberrations in patient 2073, were present in both sorted subpopulations (supplemental Figure 1).

Routine analysis of the immunoglobulin/T-cell receptor (TCR) gene repertoire revealed 5 different clonal leukemia-specific rearrangements in patient 2279, whereas there were no clonal rearrangements detected in patient 2073. All 5 rearrangements (immunoglobulin H, TCR-δ, TCR-β, and 2 rearrangements in TCR-γ) in patient 2279 can theoretically be present simultaneously in a single originating cell. By using conventional quantitative polymerase chain reaction (qPCR) –based analysis, we demonstrated the presence of all of these rearrangements in both sorted subpopulations at similar levels (supplemental Figure 2), further complementing our data demonstrating the genetic concordance of both immunophenotypic subpopulations in patient 2279.

In summary, although we used complex genomic analyses, we did not identify any genetic aberrations that were specifically present in only 1 of the immunophenotypically distinct subpopulations in either of the 2 patients investigated upon the diagnosis of BLL. Although we cannot exclude that such aberrations triggering the immunophenotypic lineage split (or contributing to it) do exist in certain BLLs, this does not appear to be a general mechanism, at least in T/myeloid BLL. Conversely, mutational analysis of WT1 in patient 2279 suggests that lineage plasticity is inherent to the vast majority of leukemic blasts, which alter their transcriptional programs and differentiate into different lineages in a stochastic manner. Our results shed new light on BLL: not only can all immunophenotypically heterogeneous leukemic populations be derived from an identical founder cell, but multiple leukemic cells possess the potential to differentiate into very distinct cell types. Further studies with large cohorts are needed to elucidate whether this lineage plasticity in BLL associates with certain specific genetic and/or epigenetic backgrounds.

The online version of this article contains a data supplement.

Authorship

Acknowledgments: This work was supported by grants from the Czech Ministries of Health (AZV 15-30626A and AZV 15-28525A), and Education (NPU I LO1604), and by the Project for Conceptual Development of Research Organization 00064203 (University Hospital Motol, Prague, Czech Republic).

Contribution: M.K. and M.Z. performed the genetic analyses; E.M. and O.H. performed the immunophenotyping and cell sorting; A.M., J. Stuchly, and K.F. performed the bioinformatic analyses; M.K. and M.Z. performed the postbioinformatic analyses; M.K., M.Z., J. Starkova, J. Stary, J.Z., O.H., and J.T. integrated and interpreted the results; O.H., J.T., and M.Z. designed the study; M.Z. led the study and wrote the manuscript; and all authors revised and approved the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Jan Trka, Childhood Leukaemia Investigation Prague, Department of Paediatric Haematology and Oncology, Second Faculty of Medicine, Charles University, V Uvalu 84, 150 06 Prague, Czech Republic; e-mail: jan.trka@lfmotol.cuni.cz.

References

References
1
Hrusak
 
O
Luks
 
A
Janotova
 
I
, et al. 
Acute Leukemias of Ambiguous Lineage; Study on 247 Pediatric Patients [abstract].
Blood
2015
, vol. 
126
 
23
 
Abstract 252
2
Jan
 
M
Majeti
 
R
Clonal evolution of acute leukemia genomes.
Oncogene
2013
, vol. 
32
 
2
(pg. 
135
-
140
)
3
Tosello
 
V
Mansour
 
MR
Barnes
 
K
, et al. 
WT1 mutations in T-ALL.
Blood
2009
, vol. 
114
 
5
(pg. 
1038
-
1045
)
4
Zhang
 
J
Ding
 
L
Holmfeldt
 
L
, et al. 
The genetic basis of early T-cell precursor acute lymphoblastic leukaemia.
Nature
2012
, vol. 
481
 
7380
(pg. 
157
-
163
)
5
Kontro
 
M
Kuusanmäki
 
H
Eldfors
 
S
, et al. 
Novel activating STAT5B mutations as putative drivers of T-cell acute lymphoblastic leukemia.
Leukemia
2014
, vol. 
28
 
8
(pg. 
1738
-
1742
)
6
Liu
 
Y
Easton
 
J
Shao
 
Y
, et al. 
The Genomic Landscape of Childhood T-Lineage Acute Lymphoblastic Leukemia [abstract].
Blood
2015
, vol. 
126
 
23
 
Abstract 691
7
Gremer
 
L
Merbitz-Zahradnik
 
T
Dvorsky
 
R
, et al. 
Germline KRAS mutations cause aberrant biochemical and physical properties leading to developmental disorders.
Hum Mutat
2011
, vol. 
32
 
1
(pg. 
33
-
43
)
8
Mullighan
 
CG
Goorha
 
S
Radtke
 
I
, et al. 
Genome-wide analysis of genetic alterations in acute lymphoblastic leukaemia.
Nature
2007
, vol. 
446
 
7137
(pg. 
758
-
764
)
9
Jones
 
CL
Bhatla
 
T
Blum
 
R
, et al. 
Loss of TBL1XR1 disrupts glucocorticoid receptor recruitment to chromatin and results in glucocorticoid resistance in a B-lymphoblastic leukemia model.
J Biol Chem
2014
, vol. 
289
 
30
(pg. 
20502
-
20515
)