Deciphering Potential Molecular Signatures to Differentiate Acute Myeloid Leukemia (AML) with BCR::ABL1 from Chronic Myeloid Leukemia (CML) in Blast Crisis

Acute myeloid leukemia (AML) with BCR::ABL1 has recently been recognized as a distinct subtype in international classifications. Distinguishing it from myeloid blast crisis chronic myeloid leukemia (BC-CML) without evidence of a chronic phase (CP), remains challenging. We aimed to better characterize this entity by integrating clonal architecture analysis, mutational landscape assessment, and gene expression profiling. We analyzed a large retrospective cohort study including CML and AML patients. Two AML patients harboring a BCR::ABL1 fusion were included in the study. We identified BCR::ABL1 fusion as a primary event in one patient and a secondary one in the other. AML-specific variants were identified in both. Real-time RT-PCR experiments demonstrated that CD25 mRNA is overexpressed in advanced-phase CML compared to AML. Unsupervised principal component analysis showed that AML harboring a BCR::ABL1 fusion was clustered within AML. An AML vs. myeloid BC-CML differential expression signature was highlighted, and while ID4 (inhibitor of DNA binding 4) mRNA appears undetectable in most myeloid BC-CML samples, low levels are detected in AML samples. Therefore, CD25 and ID4 mRNA expression might differentiate AML with BCR::ABL1 from BC-CML and assign it to the AML group. A method for identifying this new WHO entity is then proposed. Finally, the hypothesis of AML with BCR::ABL1 arising from driver mutations on a BCR::ABL1 background behaving as a clonal hematopoiesis mutation is discussed. Validation of our data in larger cohorts and basic research are needed to better understand the molecular and cellular aspects of AML with a BCR::ABL1 entity.


Introduction
The BCR::ABL1 rearrangement, encoded by the Philadelphia (Ph1) chromosome and resulting from t(9;22)(q34;q11) translocation, is the molecular hallmark of chronic myeloid leukemia (CML), a well-known myeloproliferative neoplasm [1].Historically, in the absence of adequate treatment, CML progresses inexorably from the chronic phase (CP) to the blast crisis (BC), also known as the acute phase, with or without an intermediate accelerated phase (AP).Recently, the reality of the AP has been questioned, and a more global concept of the advanced phases of CML has been proposed [2].Today, with the use of ATPcompetitive tyrosine kinase inhibitors (TKIs), (imatinib, dasatinib, nilotinib, bosutinib, and ponatinib) or an allosteric inhibitor (asciminib), the risk of CML transformation has significantly decreased, becoming a rare event, with an estimated annual progression rate to the advanced phases (AP or BC) of approximately 1% [3].In most cases, blast crisis presents as acute myeloid leukemia (AML), less frequently as acute lymphoblastic leukemia (ALL), and rarely as mixed phenotype acute leukemia (MPAL) [1,4].
The BCR::ABL1 rearrangement is also found in de novo AMLs, called AML with BCR::ABL1 fusion or AML with t(9;22)(q34.1;q11.2) translocation in recent international classifications [2,5].More than 20% of blasts are required to diagnose such acute leukemias.Furthermore, they need to be differentiated from the myeloid transformation of CML without an apparent CP.Regarding acute leukemias, the BCR::ABL1 rearrangement can, therefore, be detected in AML arising from BC-CML, in de novo AML, and throughout the course of AML, as a secondary event.In this context, we believe that improved molecular analysis of AMLs harboring a BCR::ABL1 rearrangement holds excellent promise for refining their classification and selecting the best therapeutic strategies.
High-throughput genomic DNA sequencing (NGS) has paved the way for accurate characterization of the mutational landscape of AMLs [6][7][8].More recently, RNA sequencing (RNA-Seq) has enabled a more detailed diagnosis of AML, either by identifying fusion genes or by analyzing the transcriptome [9,10].Nowadays, such approaches, based on targeted gene panels and applied in clinical routine, allow comprehensive and accurate investigation of AMLs.
In this paper, we tried to determine whether AML with BCR::ABL1 could be molecularly distinguished from acute transformations of CML and definitively assigned as typical AML.To this end, we used targeted NGS to determine the molecular landscape of AML harboring a BCR::ABL1 fusion.We performed colony-forming cell (CFC) assays to understand the clonal architecture of these leukemias.We also examined CD25, CD26, and IL1-RAP mRNA expression to highlight a potential marker for BC-CML or AML.Finally, we analyzed the mRNA expression data obtained using RNA-Seq in an attempt to stratify patients according to their disease.

Clinical and Biological Characteristics of BCR::ABL1-Positive AML Patients
AMLs carrying the BCR::ABL1 rearrangement have often been described as de novo leukemias and, more rarely, as a secondary event during disease relapse.Two AML patients harboring a BCR::ABL1 fusion were identified in our institution and extensively analyzed.The first patient (AML/BA-1) is representative of a de novo AML with BCR::ABL1, without CML history.As such, it can be distinguished from BC-CML.He was a 47 year old man with hyperleukocytosis (292 × 10 9 /L) and 75% peripheral blasts at diagnosis (Table 1).Bone marrow aspiration and flow cytometry diagnosed AML-FAB-M2.The ISTH (International Society on Thrombosis And Haemostasis) score based on four parameters (prothrombin time, platelets, fibrinogen, and fibrin-related markers) was consistent with overt DIC (disseminated intravascular coagulation).Cytogenetic analysis identified a t(9;22)(q34;q11) translocation resulting in a classical e14a2 BCR::ABL1 rearrangement.No additional cytogenetic abnormality was observed.However, the IDH2 and NPM1 type A pathogenic variants were characterized using NGS.The patient started cytoreductive therapy (cytarabine + hydroxyurea) but died rapidly of multi-visceral failure associated with cerebral hematoma.The second patient (AML/BA-2) represents an example of BCR::ABL1 acquisition upon first AML relapse.She was a 61 year old woman with pancytopenia and 63.5% blasts in the peripheral blood at diagnosis.Bone marrow aspiration revealed massive infiltration by granular myeloblasts (92%), and flow cytometry confirmed the myeloid phenotype (Table 2).
The patient was subsequently diagnosed with AML-FAB-M1.A normal karyotype was found, and NGS identified three pathogenic variants in the TET2 (two variants) and NPM1 genes.The patient achieved complete cytologic remission after receiving the standard induction regimen (daunorubicin + cytarabine) and consolidation therapy (cytarabine).A hematologic relapse was observed 9 months after the disease onset, and a Philadelphia chromosome was detected.Molecular analyses characterized an atypical BCR::ABL1 e6a2 rearrangement in addition to previously identified variants.Before allogeneic stem cell transplantation, the patient was treated with salvage chemotherapy (cytarabine, amsacrine, dasatinib).Five years later, the patient remains in molecular remission (NPM1 and BCR::ABL1 mRNA transcript follow-up).

Clonal Architecture and Tumor Evolution of BCR::ABL1-Positive AMLs
To determine the sequence of mutation acquisition at the hematopoietic stem cell (HSC)/progenitor level, CFC assays were performed using a bone marrow sample at diagnosis (patient AML/BA-1) or at relapse (patient AML/BA-2).At day +14, clonogenic cells (CFU-GM and CFU-G) were harvested from the methylcellulose.BCR::ABL1 fusion and specific mutations were researched in each CFC using qRT-PCR.The clonal evolution of the disease was then described using CFC assays and NGS/RNA-Seq data.For the AML/BA-1 patient, the presence of BCR::ABL1 fusion and the IDH2/NPM1 variants was evaluated from 50 individual CFCs.These experiments revealed a major BCR::ABL1-NPM1-IDH2 sequence and a minor BCR::ABL1-IDH2 one (Figure 1).Most HSCs or progenitors carry at least the BCR::ABL1 fusion and the NPM1 mutation.Using these data, it was possible to apprehend the tumoral evolution of the de novo AML at the time of diagnosis.In this case, BCR::ABL1 fusion acts as a primary event, and the other mutations (NPM1 and IDH2) as secondary cooperative events.
Clonogenic assays were also used to understand the clonal architecture from the bone marrow NGS or RNA-Seq results.In contrast to the CFCs, the bone marrow bulk cells presented all three genetic abnormalities with equal frequency (Table 3).variants was evaluated from 50 individual CFCs.These experiments revealed a major BCR::ABL1-NPM1-IDH2 sequence and a minor BCR::ABL1-IDH2 one (Figure 1).Most HSCs or progenitors carry at least the BCR::ABL1 fusion and the NPM1 mutation.Using these data, it was possible to apprehend the tumoral evolution of the de novo AML at the time of diagnosis.In this case, BCR::ABL1 fusion acts as a primary event, and the other mutations (NPM1 and IDH2) as secondary cooperative events.For patient AML/BA-2, the presence of the TET2/NPM1 variants and BCR::ABL1 fusion was estimated from 44 individual CFCs representative of the relapse.Clonogenic assays highlighted a TET2 (K148fs)-TET2 (G1275R)-NPM1-BCR::ABL1 sequence (Figure 2).It should be noted that only a few hematopoietic stem/progenitor cells presented the complete sequence.In this case, BCR::ABL1 fusion operated as a secondary event (proliferating mutation), as expected.NGS and RNA-Seq analyses should also be interpreted based on the clonal architecture determined by the CFC assay.
As shown in Table 4, the NGS and RNA-Seq performed on the bone marrow bulk identified all four abnormalities with comparable frequency, suggesting a massive expansion of BCR::ABL1-expressing cells.For patient AML/BA-2, the presence of the TET2/NPM1 variants and BCR::ABL sion was estimated from 44 individual CFCs representative of the relapse.Clonogen says highlighted a TET2 (K148fs)-TET2 (G1275R)-NPM1-BCR::ABL1 sequence (Figu It should be noted that only a few hematopoietic stem/progenitor cells presented the plete sequence.In this case, BCR::ABL1 fusion operated as a secondary event (prolifer mutation), as expected.NGS and RNA-Seq analyses should also be interpreted base the clonal architecture determined by the CFC assay.As shown in Table 4, the NGS and RNA-Seq performed on the bone marrow identified all four abnormalities with comparable frequency, suggesting a massive ex sion of BCR::ABL1-expressing cells.The major clonal characteristics of the AML/BA-1 and AML/BA-2 patients are summarized in Table 5.The clonal architecture established at the progenitor/stem cell level identifies AML/BA-1 as a true AML with BCR::ABL1.We previously showed that CD25 mRNA expression was higher in advanced-phase CML than in CP-CML and AML [14].Here, we tested CD25, CD26, and IL1-RAP mRNA expressions as potential biomarkers to distinguish advanced-phase CML from AML.Three sub-cohorts, including 26 AML patients, 33 CP-CML patients, and 15 AP/BC-CML patients, were analyzed.It should be noted that the advanced phases of CML include AP-CML and BC-CML, as detailed in Table S1.To avoid bias in the results, the two patients with BCR-ABL1-positive AML were not included in the analysis.
As previously reported, CD25 mRNA was significantly overexpressed in advancedphase CML compared to CP-CML and AML (Figure 3A).It seems important to note that those results were obtained using independent patient cohorts from our first study [14].In addition, an analysis of the ROC curve suggests that CD25 mRNA expression behaves as a good marker of advanced-phase CML with an AUC > 0.8 (Figure 3B).Comparable results were obtained for CD26 mRNA expression (Figure 3C,D).In contrast, IL1-RAP mRNA expression did not vary significantly between the three cohorts.As a result, CD25 or CD26 mRNA expression may be relevant indicators in differentiating AML from advancedphase CML, allowing for better characterization of BCR::ABL1-positive AML.Moreover, a positive correlation was observed between CD26 and CD25 mRNA expression (Figure 3E).Consequently, only CD25 mRNA expression was used for the rest of the work.

Unsupervised Analyses of Gene Expression from RNA-Seq Data
The RNA-Seq expression data from AML and CML patients already analyzed for fusions and variant detection were extracted and subsequently analyzed (Table S2).These data included 30 AML samples (18 de novo AML and 12 secondary AML or sAML) and 20 CML samples (six patients with BC-CML, five with resistant CP-CML, and nine with CP-CML at diagnosis).To be consistent with our objective, only the myeloid BC-CML (M-BC-CML) samples were retained.Six PB samples from healthy donors and the two patients to be tested, AML/BA-1 (PB and BM at diagnosis) and AML/BA-2 (BM at diagnosis and relapse), were also analyzed.Three genes were discarded due to insufficient reads (MLLT4 and U2AF1) or a lack of expression in all samples (ROS1).
The transcriptome-normalized matrix was processed using principal component analysis for unsupervised analyses.These studies revealed a clear stratification of the healthy donor controls, CML, and AML samples (Figure 4A).BCR::ABL1-positive AML samples were found to be clustered within the AML samples (a, AML/BA-1 and b, AML/BA-2 at relapse).Regarding the AML/BA-2 patient, the progression from de novo AML (b, without BCR::ABL1, red circle) to relapse (b, with BCR::ABL1, black circle) occurred without any significant changes in the gene expression profile.The same investigations based on the sample subtypes allowed for good stratification of the CML samples

Unsupervised Analyses of Gene Expression from RNA-Seq Data
The RNA-Seq expression data from AML and CML patients already analyzed for fusions and variant detection were extracted and subsequently analyzed (Table S2).These data included 30 AML samples (18 de novo AML and 12 secondary AML or sAML) and 20 CML samples (six patients with BC-CML, five with resistant CP-CML, and nine with CP-CML at diagnosis).To be consistent with our objective, only the myeloid BC-CML (M-BC-CML) samples were retained.Six PB samples from healthy donors and the two patients to be tested, AML/BA-1 (PB and BM at diagnosis) and AML/BA-2 (BM at diagnosis and relapse), were also analyzed.Three genes were discarded due to insufficient reads (MLLT4 and U2AF1) or a lack of expression in all samples (ROS1).
The transcriptome-normalized matrix was processed using principal component analysis for unsupervised analyses.These studies revealed a clear stratification of the healthy donor controls, CML, and AML samples (Figure 4A).BCR::ABL1-positive AML samples were found to be clustered within the AML samples (a, AML/BA-1 and b, AML/BA-2 at relapse).Regarding the AML/BA-2 patient, the progression from de novo AML (b, without BCR::ABL1, red circle) to relapse (b, with BCR::ABL1, black circle) occurred without any significant changes in the gene expression profile.The same investigations based on the sample subtypes allowed for good stratification of the CML samples (CP-CML at diagnosis, resistant CP-CML, and myeloid BC-CML) (Figure 4B).Conversely, these transcriptome analyses did not stratify secondary AML and de novo AML.

AML vs. Myeloid BC-CML Differential Expression Signature
The transcriptome-normalized matrix was used for supervised analysis, highlighting the genes differentially expressed between the myeloid blast crisis CML and AML samples (Figure 5A).After false discovery rate adjustment of the p-values, nine genes appear to be differentially expressed between AML and M-BC-CML.In particular, MYD88, PI-CALM, and RARA are overexpressed in M-BC-CML with log2 fold changes of 1.22, 1.16, and 1.37, respectively (Table S3).On the other hand, ID4 is overexpressed in AML samples with a log2 fold change of 26.5.With a cutoff of a raw p-value ≤ 0.05, it was possible to identify an expression profile that tended to reclassify BC-CML samples into the left clusters of the heatmap after clustering by the Euclidean distances (Figure 5B).These results suggest that the myeloid BC-CML samples had a different expression profile from the AML samples.Moreover, ID4 (inhibitor of DNA binding 4) is a very attractive marker for distinguishing BC-CML from AML.

AML vs. Myeloid BC-CML Differential Expression Signature
The transcriptome-normalized matrix was used for supervised analysis, highlighting the genes differentially expressed between the myeloid blast crisis CML and AML samples (Figure 5A).After false discovery rate adjustment of the p-values, nine genes appear to be differentially expressed between AML and M-BC-CML.In particular, MYD88, PICALM, and RARA are overexpressed in M-BC-CML with log2 fold changes of 1.22, 1.16, and 1.37, respectively (Table S3).On the other hand, ID4 is overexpressed in AML samples with a log2 fold change of 26.5.With a cutoff of a raw p-value ≤ 0.05, it was possible to identify an expression profile that tended to reclassify BC-CML samples into the left clusters of the heatmap after clustering by the Euclidean distances (Figure 5B).These results suggest that the myeloid BC-CML samples had a different expression profile from the AML samples.Moreover, ID4 (inhibitor of DNA binding 4) is a very attractive marker for distinguishing BC-CML from AML.

BCR::ABL1-Positive AML Characteristic Profile
This study identified two molecular markers: ID4 using RNA-Seq and CD25 (or CD26) using qRT-PCR.To verify their efficacy in distinguishing BCR::ABL1 AMLs, we performed CD25 mRNA quantification on the entire RNAseq sub-cohort.The results for the CD25 and ID4 expression in the AML and M-BC-CML samples are presented in Figure 6A as the mean with a 95% confidence interval.The expression data from the AML/BA-1 and AML/BA-2 patients (circles in the graph) fall within the AML confidence interval.

BCR::ABL1-Positive AML Characteristic Profile
This study identified two molecular markers: ID4 using RNA-Seq and CD25 (or CD26) using qRT-PCR.To verify their efficacy in distinguishing BCR::ABL1 AMLs, we performed CD25 mRNA quantification on the entire RNAseq sub-cohort.The results for the CD25 and ID4 expression in the AML and M-BC-CML samples are presented in Figure 6A as the mean with a 95% confidence interval.The expression data from the AML/BA-1 and AML/BA-2 patients (circles in the graph) fall within the AML confidence interval.Lastly, differential expression analyses were performed between the BCR::ABL1-positive AML and other AML samples (Figure 6B).After false discovery rate adjustment of the p-values, a significant over-expression of ABL1 in the BCR::ABL1-positive AML as compared to the other AML samples (log2 Fold Change = +1.80,adjusted p-value = 0.039) is observed (Table S4).This result is consistent with the expression of the fusion transcript (assessed by the ABL1 expression) in the BCR::ABL1-positive AML subtype.We also show a significant down-regulation of PDCD1LG2 in BCR::ABL1-expressing AML (log2 Fold Change = −9.97,adjusted p-value = 0.039).PDCD1LG2 (programmed cell death 1 ligand 2) encodes an immune checkpoint receptor ligand.It should be noted that this gene was not differentially expressed between the BC-CML and AML samples.

Discussion
New AML classifications have validated "AML with BCR::ABL1", also called "AML with t(9;22)(q34.1;q11.2)",as a distinct entity [2,5].The diagnosis of AML with BCR::ABL1 could be established without evidence of CML history.Compared to myeloid BC-CML, AML with BCR::ABL1 seemed to be characterized by the absence of basophilia and a Lastly, differential expression analyses were performed between the BCR::ABL1positive AML and other AML samples (Figure 6B).After false discovery rate adjustment of the p-values, a significant over-expression of ABL1 in the BCR::ABL1-positive AML as compared to the other AML samples (log2 Fold Change = +1.80,adjusted p-value = 0.039) is observed (Table S4).This result is consistent with the expression of the fusion transcript (assessed by the ABL1 expression) in the BCR::ABL1-positive AML subtype.We also show a significant down-regulation of PDCD1LG2 in BCR::ABL1-expressing AML (log2 Fold Change = −9.97,adjusted p-value = 0.039).PDCD1LG2 (programmed cell death 1 ligand 2) encodes an immune checkpoint receptor ligand.It should be noted that this gene was not differentially expressed between the BC-CML and AML samples.
Although other mutations may play a role in disease prognosis, the BCR::ABL1 oncogene has long been known to be the trigger for CML [16,17].AMLs, on the other hand, require a known and complex multi-step process that ultimately leads to a blockage of the hematopoietic differentiation and the uncontrolled proliferation of immature leukemic cells [7,18].Acquisition of a BCR::ABL1 rearrangement during AML evolution is rare [19].In this context, secondary BCR::ABL1 fusion could act as a late cooperating event within the multi-step pathogenesis.It may also arise from de novo AML, in which CBFB::MYH11, GATA2::MECOM, KMT2A::AFDN, PML::RARA, or RUNX1::RUNX1T1 are rearranged at diagnosis [20][21][22][23][24][25][26][27].Secondary BCR::ABL1 fusion has also been found to occur in AML after treatment with an FLT3 inhibitor [28,29].Not all these cases meet the definition of AML with BCR::ABL1 since BCR::ABL1 acts as a secondary event and is likely to confer a proliferative advantage, such as NRAS, FLT3, or KIT mutations.Therefore, it appears essential to establish whether BCR::ABL1 is a primary or secondary event through the multi-step sequence of AML.Consequently, to identify a putative AML with BCR::ABL, it seems necessary not only (1) to verify that BCR::ABL1 is the first event of the multi-step process (using clonogenic assay and single-cell analysis), (2) to highlight the presence of typical AML abnormalities in addition to BCR::ABL1 (using NGS and RNA-Seq), but also (3) to eliminate "specific" BC-CML mutations, such as ABL1 TKD mutations.
One of the major issues is whether or not AMLs with BCR::ABL1 share a characteristic mutational landscape.Various approaches are available today to answer these questions, from older methods such as fragment analysis or Sanger sequencing to more global methods such as NGS or RNA-Seq.Overall, the AML driver genes include (according to frequency) FLT3, NPM1, DNMT3A, NRAS, IDH1/2, RUNX1, TET2, WT1, ASXL1, PTPN11, SRSF2, TP53, CEBPA, BCOR, KMT2A, and KRAS [30].However, these data need to be qualified because AMLs are heterogenous malignancies.For example, the mutational landscape may be specific to cytogenic groups, resulting in molecular subgroups [7,31].It may also differ between de novo AML and sAML or t-AML (therapy-related AML) [32].
Several studies focusing on M-BC-CML, particularly those based on high-throughput sequencing approaches, have been published in the last decade [33][34][35][36][37].They showed that in most cases, at least one pathogenic mutation (in addition to BCR::ABL1) is detected during the progression to the blast phase.Pathogenic variants of RUNX1, ASXL1, and BCR::ABL1 TKD mutations (involved in TKI resistance) are frequently found.BCR::ABL1 mutations are a well-known characteristic feature of TKI-resistant CMLs.RUNX1 mutations may not be a specific marker for M-BC-CML, as they are also found in AMLs (albeit less frequently).Interestingly, NPM1 mutations (common in AML) appear to be absent or extremely rare in BC-CML [38][39][40][41].Considering all these data, can we characterize AMLs with BCR::ABL1 by their mutational landscape?Recent studies based on NGS experiments have shown that the genetic profiles of AMLs harboring a BCR::ABL1 fusion are not significantly different from those of M-BC-CML [35,42].Therefore, it appears that only a few genetic markers other than the potential presence of an NPM1 variant and the absence of a BCR::ABL1-TKD mutation can distinguish Ph1-positive AMLs from M-BC-CMLs and link them to Ph1-negative AMLs [43].
Some cell surface markers have been shown to distinguish CML LSCs from AML LSCs (CD34 + CD38 − ) using multicolor flow cytometry [13,44].In contrast to AML LSCs, CML LSCs can display a characteristic expression profile including CD25 (interleukin-2 receptor alpha chain), CD26 (dipeptidyl peptidase-4), or IL1-RAP (interleukin-1 receptor accessory protein) [44].It has been proposed that CD25 and CD26 may be accurate markers of CML LSCs and potential therapeutic targets [11,13].Recently, we established a BC-CML model using several CML-derived iPSC (induced pluripotent stem cell) lines mutagenized with N-ethyl-N-nitrosourea (an alkylating agent) [14].Mutagenized CML iPSCs (representing BC-CML) and non-mutagenized CML iPSCs (representing CP-CML) were compared at the transcriptomic level after hematopoietic differentiation.Several genes were found to be upregulated in BC-CML, including CD25.These findings were confirmed in patients with CP-CML, BC-CML, AML, B-ALL (B-acute lymphoblastic leukemia), and healthy controls.CD25 mRNA transcripts were found to be over-expressed in BC-CML in comparison to all other categories.Based on independent patient cohorts, these data were corroborated in the present study, highlighting their relevance.Notably, CD25 mRNA was shown to be expressed in all samples and significantly overexpressed in BC-CML.All in all, CD25 mRNA expression may be an attractive marker to differentiate BC-CML from AML.
In our RNA-Seq analyses, ID4 mRNA appears undetectable in most M-BC-CML samples, while low levels are detected in AML samples.ID4 is critical for embryogenesis and fetal development and may also be involved in tumorigenesis [45].ID4 may play seemingly contradictory roles in cancer cells.It could be over-or under-expressed at the RNA level, and the DNA promoter could be hypermethylated [46].It has been reported that ID4 expression silencing using promoter methylation is increased during CP to BC progression in CML [47].The mRNA expression of PDCD1LG2, also called PD-L2, CD273, or B7-DC, was shown to be increased in a small proportion of bone marrow CD34+ or blood mononuclear cells from AML patients [48].In this study, we have shown that the PDCD1LG2 mRNA was under-expressed in BCR::ABL1-positive AML compared to de novo and secondary AML.Its impact on the disease and its potential interest in identifying AML harboring the BCR::ABL1 fusion is unknown.
Our work shows that a comprehensive molecular analysis might be necessary to characterize leukemias more likely to correspond to AMLs with the BCR::ABL1 WHO (World Health Organization) entity (Table 6).Identification of BCR::ABL1 as the primary genetic event is critical, as are the presence of AML-specific secondary abnormalities (especially NPM1 mutations) and the absence of ABL1 TKD point mutations.CD25 and ID4 mRNA expression may also be helpful, but this needs to be confirmed in larger patient cohorts.Overall, our results support the inclusion of AML with BCR::ABL1 within the AML category.However, the identification of the BCR::ABL1 fusion (considered the hallmark of CML) as the primary oncogenic event through the clonal ontogeny of these cases of acute leukemia raises several concerns.BCR::ABL1 rearrangement has been sporadically detected at low levels in the blood of healthy individuals [49][50][51][52].Several hypotheses could explain why they do not develop CML despite the presence of the BCR::ABL1 oncogene.Some studies have suggested that BCR::ABL1 expression alone (or low-level BCR::ABL1 expression) may not be sufficient to trigger CML leukemogenesis [53][54][55][56].BCR::ABL1 fusion could occur within a committed progenitor that has lost its stem cell properties [52].It is now well established that leukemogenesis is characterized by its progressive nature, molecular diversification, and the clonal evolution of founder cells [57][58][59].Six stages of cancer development were recently proposed, ranging from early phases to overt malignancy [60].Concerning pre-malignant conditions, several molecular abnormalities detectable in healthy individuals can be characterized.Passenger or low-level driver mutations would correspond to CHIP (clonal hematopoiesis of indeterminate potential) or CHOP (clonal hematopoiesis with oncogenic potential) conditions.In these circumstances, BCR::ABL1 could be considered a CHOP mutation likely to progress to overt leukemia subsequent to a variable latency period.
Based on these data, we can formulate a hypothesis worthy of discussion.In AMLs with BCR::ABL1, we can speculate that the BCR::ABL1 fusion (primitive event) would act as a clonal hematopoiesis mutation.The subsequent acquisition of leukemogenesis-driving abnormalities such as NPM1 or IDH1/2 mutations could lead to the sudden development of de novo AML without passing through CP-CML.While this hypothesis appears attractive, it nevertheless requires further validation.
In conclusion, our study provides some new molecular clues helping to characterize AML with BCR::ABL1.Distinguishing these AMLs from myeloid BC-CMLs remains challenging, and fundamental research is required to better understand the molecular and cellular specificity of AMLs with BCR::ABL1.Given the rarity of these AMLs, a multicenter study is needed to analyze a significant number of patients.It would also be helpful to consider a single-cell analysis approach to assess the clonal architecture.In addition, this work highlights the importance of gene expression signatures.In this context, a more global transcriptomic approach could be an attractive perspective for better characterization of AMLs with BCR::ABL1.

Patients
In this study, we analyzed bone marrow or peripheral blood samples from 63 CML patients (19 advanced-phase CML, 5 TKI-resistant CP-CML, 39 and CP-CML at diagnosis) and 54 AML patients (37 de novo AML and 17 secondary AML or sAML).Molecular and cellular experiments (NGS, RNA-Seq, qRT-PCR, and clonogenic assays) were performed on residual sample material (leftovers after routine analysis).Among all patients, sub-cohort studies provided expression data using RNA-Seq or RT-qPCR (CD25, CD26 or IL1-RAP).Two patients with AML and BCR::ABL1 rearrangements (AML/BA-1 and AML/BA-2) had their bone marrow and peripheral blood extensively analyzed.The details about the patients (disease, karyotype, molecular rearrangements, pathogenic variants, and the availability of the RNA-Seq and qRT-PCR experiments) are provided in Table S5.A cohort of 6 healthy donors was used as the control.The study was carried out in accordance with the Declaration of Helsinki.

Targeted High-Throughput DNA and RNA Sequencing
Myeloid-targeted NGS was used to characterize pathogenic variants with their VAFs (variant allele frequencies).The NGS analyses were performed using the Agilent SureSelect Target Enrichment System and a personalized myeloid panel of 85 genes (Agilent Technologies, Santa Clara, CA, USA).An in-house bioinformatics pipeline was used to analyze the raw data after sequencing.RNA-Seq was used to characterize the fusion transcripts, to identify some pathogenic variants with their TAFs (transcript allele frequencies), and to estimate the mRNA expression of the targeted genes (expressed as log2 expressions using CHMP2A, GPI, RAB7A, and VCP as housekeeping genes).RNA-Seq was performed using the Archer ® FusionPlex ® Myeloid kit with a target set of 84 genes (Archer DX, Boulder, CA, USA).After sequencing, the raw data were analyzed using the Archer Analysis bioinformatics system.All libraries (DNA/RNA) were sequenced on the Illumina MiSeq platform (Illumina, San Diego, CA, USA).

Figure 1 .
Figure 1.Clonal architecture and mutational evolution of patient AML/BA-1.The presence of e14a2 BCR::ABL1 fusion and IDH2 and NPM1 variants was assessed using qRT-PCR in 50 individual CFCs

Figure 1 .
Figure 1.Clonal architecture and mutational evolution of patient AML/BA-1.The presence of e14a2 BCR::ABL1 fusion and IDH2 and NPM1 variants was assessed using qRT-PCR in 50 individual CFCs from the bone marrow sample at diagnosis.The number of CFCs harboring a specific genetic pattern was indicated.This protocol highlighted a BCR::ABL1-NPM1A-IDH2 major sequence.

Figure 2 .
Figure 2. Clonal architecture and mutational evolution of patient AML/BA-2.Real-time RT-PC termined the presence of TET2 and NPM1 variants along with e6a2 BCR::ABL1 rearrangemen individual CFCs from the bone marrow sample at relapse.The number of CFCs harboring a sp genetic pattern was indicated.This procedure highlighted a TET2 (K148fs)-TET2 (G127 NPM1A-BCR::ABL1 sequence.

Figure 2 .
Figure 2. Clonal architecture and mutational evolution of patient AML/BA-2.Real-time RT-PCR determined the presence of TET2 and NPM1 variants along with e6a2 BCR::ABL1 rearrangement in 44 individual CFCs from the bone marrow sample at relapse.The number of CFCs harboring a specific genetic pattern was indicated.This procedure highlighted a TET2 (K148fs)-TET2 (G1275R)-NPM1A-BCR::ABL1 sequence.

17 Figure 3 .
Figure 3. CD25 and CD26 mRNA expression in CML and AML patients.(A,C).The graphs show the CD25 and CD26 mRNA expressions in CML patients in chronic phase at diagnosis (CP-CML), in patients in accelerated phase (AP) or blast crisis (BC), and in patients with AML.Data were shown as box-and-whisker plots (5-95 percentile).The median is shown in red for each group, and a twotailed Mann-Whitney test calculated the p-value.(B,D).ROC curves suggest that CD25 and CD26 mRNA expression can distinguish AP/BC patients (n = 15) from AML controls (n = 26).(E).Correlation between CD26 and CD25 mRNA expression.

Figure 3 .
Figure 3. CD25 and CD26 mRNA expression in CML and AML patients.(A,C).The graphs show the CD25 and CD26 mRNA expressions in CML patients in chronic phase at diagnosis (CP-CML), in patients in accelerated phase (AP) or blast crisis (BC), and in patients with AML.Data were shown as box-and-whisker plots (5-95 percentile).The median is shown in red for each group, and a two-tailed Mann-Whitney test calculated the p-value.(B,D).ROC curves suggest that CD25 and CD26 mRNA expression can distinguish AP/BC patients (n = 15) from AML controls (n = 26).(E).Correlation between CD26 and CD25 mRNA expression.

Figure 5 .
Figure 5. Genes differentially expressed between myeloid BC-CML and AML samples.(A).The volcano plot highlights genes overexpressed in the AML or BC-CML context.Scattered points represent genes, the x-axis is the log2 fold change, and the y-axis represents the −log10 (p-values).(B).Heatmap of differentially expressed genes.The list of significantly differentially expressed genes (pvalue < 0.05) is shown.BC-CML samples are reclassified into the left clusters of the heatmap.

Figure 5 .
Figure 5. Genes differentially expressed between myeloid BC-CML and AML samples.(A).The volcano plot highlights genes overexpressed in the AML or BC-CML context.Scattered points represent genes, the x-axis is the log2 fold change, and the y-axis represents the −log10 (p-values).(B).Heatmap of differentially expressed genes.The list of significantly differentially expressed genes (p-value < 0.05) is shown.BC-CML samples are reclassified into the left clusters of the heatmap.

Figure 6 .
Figure 6.Molecular features of BCR::ABL1-expressing AML samples.(A).The graph shows ID4 and CD25 mRNA expressions in AML and M-BC-CML represented by the mean with a 95% confidence interval, highlighted in red and blue for AML and M-BC-CML, respectively.This representation indicates AML/BA-1 and AML/BA-2 samples as plain or empty black circles.AMLs harboring the BCR::ABL1 fusion are separated from BC-CML and segregated with AML.(B).The volcano plot highlights genes overexpressed in BCR::ABL1-positive or -negative AML context.Scattered points represent genes, the x-axis is the log2 fold change, and the y-axis represents the −log10 (p-values).

Figure 6 .
Figure 6.Molecular features of BCR::ABL1-expressing AML samples.(A).The graph shows ID4 and CD25 mRNA expressions in AML and M-BC-CML represented by the mean with a 95% confidence interval, highlighted in red and blue for AML and M-BC-CML, respectively.This representation indicates AML/BA-1 and AML/BA-2 samples as plain or empty black circles.AMLs harboring the BCR::ABL1 fusion are separated from BC-CML and segregated with AML.(B).The volcano plot highlights genes overexpressed in BCR::ABL1-positive or -negative AML context.Scattered points represent genes, the x-axis is the log2 fold change, and the y-axis represents the −log10 (p-values).

Table 3 .
Results of NGS and RNA-Seq experiments for patient AML/BA-1.

Table 4 .
Results of NGS and RNA-Seq experiments for patient AML/BA-2.

Table 6 .
Genetic characteristics of AML with BCR::ABL1 compared with AML and BC-CML.