BCR::ABL1-Induced Mitochondrial Morphological Changes As a Novel Biomarker for the Artificial Intelligence-Driven Flow Cytometric Detection of Chronic Myeloid Leukemia Cells

Objectives:

The development of tyrosine kinase inhibitors (TKIs) has significantly improved the prognosis of chronic myeloid leukemia (CML) patients, making their life expectancy nearly equal to that of healthy individuals. However, lower rates of deep molecular response (DMR) and treatment-free remission (TFR) in Sokal or ELTS high-risk patients remain concerns. These scores are calculated based on age, peripheral blood blast count, platelet count, and splenomegaly, which is thought to worsen over time. Therefore, we hypothesize that early diagnosis and treatment intervention could improve the DMR achievement rate and potentially contribute to an increased rate of TFR. Recently, mitochondrial dynamics have drawn attention due to their reported involvement in many diseases; however, there are few reports on that in CML. In this study, we explored the mitochondrial alterations in CML cells and their potential as a novel biomarker for the early diagnosis.

Methods:

The BCR::ABL1 oncogene was transduced into UT-7/EPO cells. Mitochondrial observations were performed using transmission electron microscopy and confocal laser microscopy. Protein expressions were analyzed by Western blotting. For the detection of CML cells, we introduced a novel technology called ghost cytometry (GC), which is a combination of flow cytometry and artificial intelligence (AI). GC enables the acquisition of signals containing subtle morphological information of the cells in a label-free manner, which are then identified using AI (Ota S et al., Science 2018).

Results:

Through observing leukocytes from patients with CML and BCR::ABL1-transduced UT-7/EPO, we found that mitochondria were excessively fragmented in these cells compared to leukocytes from healthy individuals and gene-untransduced UT-7/EPO. We discovered that the phosphorylation of dynamin related protein (DRP) 1 and extracellular signal-regulated kinase (ERK) 1/2 in BCR::ABL1-transduced UT-7/EPO were enhanced, while inhibiting DRP1 suppressed excessive mitochondrial fragmentation. Moreover, inhibition of MEK1/2 reduced ERK1/2 and DRP1 phosphorylation, indicating that mitogen-activated protein kinase (MAPK) pathway was involved in BCR::ABL1-induced mitochondrial fragmentation. GC distinguished between BCR::ABL1-positive and negative cells with area under the receiver operating characteristic curve of >0.8, both in cell line and real-world clinical specimens. Subsequently, we created an AI learning model using the peripheral blood leukocytes of CML patients at diagnosis and healthy individuals. We then tested peripheral blood samples from CML patients after TKI initiation, which had white blood cell counts and BCR::ABL1 (IS) mRNA values of 5,960 (1,680 - 11,690) /uL and 48.8 (8.8 - 71.4) %, respectively. The percentages of AI-detected CML cells in these samples showed a high correlation (r = 0.87) with the actual IS values. In addition, we found that AI-based discrimination scores at diagnosis were significantly higher in patients whose IS values were higher than 1% at three months of treatment compared to those lower than 1% (p < 0.05).

Discussion:

In this study, we revealed that BCR::ABL1 induces excessive mitochondrial fragmentation by activating DRP1 through the MAPK pathway. These morphological alterations could be identified by AI-driven flow cytometry, enabling label-free detection of CML cells before numerical abnormalities in white blood cell counts appear. Moreover, AI-detectable morphological changes might occur between patient groups with different treatment responses. Since clinical symptoms are rarely observed in chronic phase CML, early diagnosis is challenging, especially in patients before the abnormal increase in white blood cell count. By employing this automated AI-driven label-free method as a screening test, GC holds potential for early diagnosis of CML, which may contribute to improving the rate of DMR. Previous studies have shown that patients are more likely to have a prolonged duration of DMR when the IS values are lower than 1.5% at three months of treatment (Sasaki K et al., Cancer 2018). Based on our findings, GC-derived morphological information may predict early treatment response to TKI therapy and subsequent DMR achievement. In conclusion, BCR::ABL1-induced mitochondrial fragmentation could serve as a novel biomarker for the early diagnosis and prediction of treatment efficacy in CML.

Suzuki:Sysmex: Current Employment. Inano:PharmaEssentia Japan: Research Funding. Yamada:Sysmex: Current Employment. Komatsu:PharmaEssentia Japan KK: Other: Board member, medical writing funding; Meiji Seika Pharma Co., Ltd: Research Funding; Torii Pharmaceutical Co., Ltd: Consultancy. Ando:AstraZeneca: Honoraria; Chugai Pharmaceutical: Research Funding; Century Therapeutics: Research Funding; Sumitomo Pharma: Research Funding; Kyowa Kirin: Research Funding; Novartis Pharma: Honoraria; AbbVie: Honoraria; Astellas Pharma: Honoraria. Takaku:Novartis: Honoraria; Otsuka: Honoraria, Research Funding; Phizer: Honoraria; Sysmex: Research Funding; ThinkCite: Research Funding.