The complexity of genomic mutation dictates the prognosis of acute leukemia with ambiguous lineage

Acute leukemia with ambiguous lineage (ALAL) is a rare and highly aggressive malignancy with limited molecular characterization and therapeutic recommendations. In this study, we retrospectively analyzed 1635 acute leukemia cases in our center from January 2012 to June 2018.The diagnose of ALAL was based on either EGIL or 2016 WHO criteria, a total of 39 patients were included. Four patients diagnosed as acute undifferentiated leukemia (AUL) by both classication systems. The mutations detected in bi-phenotypic acute leukemia enriched in genes related to genomic stability and transcriptional regulation; while AUL cases frequently mutated in genes involved in signaling pathway. Survival analysis of all patients suggested that the prognosis of ALAL was independent of immunophenotype, chromosome karyotype, treatment, but signicantly associated with the mutation complexity, also termed numbers of the mutations carried by each patient (Log rank p = 0.009 for progression-free survival [PFS] and Log rank p = 0.047 for overall survival [OS], respectively). Similar results were obtained when the WHO diagnostic system were applied (Log rank p < 0.001 for both PFS and OS). Among these patients, those excluded by WHO criteria had even worse clinical outcome than the patients included (Log rank p = 0.023 for PFS and Log rank p = 0.031 for OS). Collectively, the complexity of genomic mutation of ALAL patients is signicantly associated with the clinical outcomes. The rationality and clinical applicability of the diagnostic criteria of WHO system need to be evaluated by more large-scale clinical studies. undifferentiated leukemia (AUL), which means no lineage related antigens are expressed, and bi-phenotypic acute leukemia (BAL) when either two separate blast populations are encountered or a single blast population demonstrating evidence of both myeloid and lymphoid differentiation concurrently 2 . WHO (2008/2016) classication system collectively groups these rare disorders as acute leukemia with ambiguous lineage (ALAL), which may possibly arise from hematopoietic pluripotent stem cells 3 . The new classication system proposes the diagnosis of mixed-phenotype acute leukemia (MPAL) and emphasizes the impact of CD19 for B-cell lineage, CD3/cCD3 for T-cell lineage, and cytoplasmatic myeloperoxidase (MPO) for myeloid lineage, respectively 4,5 . The diagnosis of ALAL largely depends on the immunophenotype, and cases with AML specic recurrent cytogenetic abnormalities are classied to AML as dened in WHO classication system. Based on the markers expressed on the blast cells, BAL/MPAL can be further divided in to B/M, T/M, B/T, and B/T/M subtypes. completed genetic mutations in PTEN, KRAS, JAK3, KMT2C, ETV6, TP53, JAK3, PHF6,


Introduction
Majority of acute leukemia (AL) can be accurately diagnosed and classi ed into acute myeloid leukemia (AML) or acute lymphoid leukemia (ALL) according to morphologic, cytochemical, and immunologic features 1 . However, less than 5% of patients can't be clearly assigned to the well-established subsets of AL as lacking expression of lineage-speci c markers, or presenting mixed immune phenotype more than one lineage. In the past two decades, the diagnostic criteria for such kinds of leukemia were controversial, and the nomenclature was constantly improving. The European Group for the immunological characterization of leukemia (EGIL) separately de nes them as acute undifferentiated leukemia (AUL), which means no lineage related antigens are expressed, and bi-phenotypic acute leukemia (BAL) when either two separate blast populations are encountered or a single blast population demonstrating evidence of both myeloid and lymphoid differentiation concurrently 2 . WHO (2008/2016) classi cation system collectively groups these rare disorders as acute leukemia with ambiguous lineage (ALAL), which may possibly arise from hematopoietic pluripotent stem cells 3 . The new classi cation system proposes the diagnosis of mixed-phenotype acute leukemia (MPAL) and emphasizes the impact of CD19 for B-cell lineage, CD3/cCD3 for Tcell lineage, and cytoplasmatic myeloperoxidase (MPO) for myeloid lineage, respectively 4,5 . The diagnosis of ALAL largely depends on the immunophenotype, and cases with AML speci c recurrent cytogenetic abnormalities are classi ed to AML as de ned in WHO classi cation system. Based on the markers expressed on the blast cells, BAL/MPAL can be further divided in to B/M, T/M, B/T, and B/T/M subtypes. molecular level 6,13,14 . Several studies demonstrated the genetic landscape of MPAL and showed patients clustered to either AML-like or ALL-like MPAL based on methylation pro ling could be bene ted from corresponding therapies 13,14 . However, the correlation between the genetic mutation and the clinical outcomes of ALAL was not fully elucidated.
In this study, we retrospectively analyzed the patients with de novo acute leukemia from January 2012 to June 2018 in Tongji Hospital, Tongji Medical Collage of Huazhong University of Science and Technology and aimed to examine a wide spectrum of gene mutations in patients with ALAL with de ned under WHO or EGIL criteria, to determine their clinical relevance with clinical outcomes of such dire leukemia and the clinical applicability of these two diagnostic systems.

Basic characteristics of patients enrolled in this study
In all 1635 patients with acute leukemia admitted to our center over a period of more than ve years, 39 patients (2.4%) satis ed the criteria of EGIL, and 30 (1.8%) patients were diagnosed as AUL or MPAL according to the WHO standards. For all 39 patients, mean age was 39 years and 23 (59%) of them were male. The median value of WBC count on presentation was 11×10 9 , hemoglobin 80 g/L, platelet 71×10 9 , and bone marrow blasts 80% (Table 1). There was no signi cant difference regarding to the baseline and clinical characteristics between the patients included (WHO 2016, n = 30) and excluded (EGIL-WHO, n = 9) by WHO criteria, except that patients excluded by WHO system had a lower level of platelet count (p = 0.039) and did not express MPO (p = 0.001). According to the FAB criteria, 14 (35.9%) patients were classi ed as ALL, and 5 (12.8%) patients displayed an AML morphology, and others were classi ed as AUL (7.7%), mix phenotype leukemia (2.6%), or inclusive (41%) (Supplementary Table 1). The clinical diagnosis of ALAL can hardly be realized by FAB diagnosis system Among all, 32 patients underwent karyotyping, and 17 (53.1%) of them were abnormal. Complex karyotypes were observed in 9 (28.1%) patients with a high degree of heterogeneity as shown in Table 2. The most frequent abnormal karyotype is t (9; 22) (q34; q11.2), resulting in the BCR-ABL1 fusion gene, appeared in 7 patients. Interestingly, 6 patients were B-M MPAL while one of them was T-M MPAL. MLL rearrangement was observed in a patients diagnosed as T-M MPAL ( Table 2). In addition to BCR-ABL and SET/CAN, which was previously reported, we also rstly detected E2A-PBX1 fusion genes in a B-M MPAL patient ( Table 2). Based on the data of the 19 patients who had been sequenced in the next-generation platform, 53 high-con dence somatic mutations were detected in 33 genes in 17 patients. The two patients with no detectable mutations were both B-M phenotype. The median number of gene mutation was 3 (0-8) per sample. The most frequently mutated genes were NRAS (4, 21%), CEBPA (4, 21%), JAK3 (3, 16%), RUNX1 (3, 16%). Mutations in genes such as DNMT3A, ETV6, IDH2, KMT2D, KRAS, NOTCH1, PHF6, TP53, and WT1 were detected in two different patients. CEBPA mutations were detected in 3 patients with M-T phenotype and one with B-T phenotype. Notably, the mutations in NRAS gene (reoccurring in 4 cases) were concentrated at its 12th/13th amino acid site.
We divided genes into the following categories according to gene attribute: transcription factor, chromatin regulation, epigenetics, cell apoptosis, signaling pathway (RTK-RAS, NOTCH, MAPK-ERK, PI3K/AKT, JAK-STAT), and others. The results showed that mutations detected in BAL cases enriched in transcription factor, chromatin regulation and epigenetic, all related to genomic stability and transcriptional regulation; while AUL cases frequently mutated in genes in signal pathway regulation such as RTK-RAS, NOTCH, and PI3K/AKT (Fig. 1A).There was no signi cant difference in the distribution of B/M and T/M gene mutations, except for the particularly high frequency of transcription factor mutations in T/M cases (71.4%) (Fig. 1B). The detailed information pro le was shown in Fig. 1C.

Survival Analyses
Survival analyses were used to compare the prognosis of clinical subsets and types of treatment. Up to January 16, 2019 (mean follow up time 428 days), 34 patients received chemotherapy, while others, according to patients' wishes or limitation of general condition, only received symptomatic support treatment. Among the patients who received chemotherapy, 7 cases received ALL-like therapy, 6 cases received AML-like therapy, and 21 cases received AML+ALL combined therapy. Decitabine was included in the therapy of 9 patients, 5 patients received tyrosine kinase inhibitors (TKI), and 4 patients received allogeneic hematopoietic stem cell transplantation (HSCT).
Among all the patients, 21 patients (53.8%) achieved complete remission (CR) at least once. The CR rate was not affected by immunophenotype (p=0.167), complex karyotype (p=0.761), whether mutated (p=0.156), or chemotherapy regimen (p=0.326); while showed a slight association with no. of mutation/person (p=0.096) (Supplementary Table 3). During the follow up, 27 cases (69.2%) progressed or died. Of all factors considered, mutation complexity, de ned as no. of mutation/person was signi cantly associated with PFS and OS (Log rank p=0.009 and Log rank p=0.047 respectively) ( Table 3 and Fig. 2A-B). When the WHO diagnostic criteria were applied, we obtained the similar results in the 30 patients included. In speci c, mutation complexity was the only clinical risk factor that remarkably associated with PFS (Log rank p<0.001), and OS (Log rank p<0.001), manifested as patients with more complex mutation pro les had much worse clinical outcomes (Supplementary Table 3 and Fig. 2C-D). Of note, 9 patients were excluded according to WHO 2016 classi cation system. Eight of them were diagnosed as acute lymphoblastic leukemia, and one was determined as unclassi ed. We found the patients excluded by WHO criteria had an even worse prognosis than those patients included, characterized as shorter PFS (Log rank p=0.023) and OS (Log rank p=0.031) ( Table 3 and Fig. 2E-F). Whether the clinical diagnosis and treatment of these subsets of patients is the optimal strategy may need to be further explored.

Discussion
ALAL is a rare and aggressive entity with heterogeneous immunophenotypic, cytogenetic and molecular features. Our study provided the genetic landscape of ALAL by screening the hot-spots in 173 genes strongly associated myeloid and lymphoid malignancies and evaluated the clinical characteristics with the dismal clinical outcomes. Apart from immunophenotype, complex karyotype, treatment, somatic mutation complexity is exclusively associated PFS and OS in ALAL. The increased number of mutations in blast cells predicted higher risk of relapses and deaths, no matter using the WHO or the EGIL diagnostic system to evaluate. Recent studies also emphasized the important roles of somatic mutations in the diagnoses and treatment strategy of ALAL,but most of published studies focused on MPAL 6,13,14 . In this study, we rstly analyzed 4 patients with AUL by high throughput sequencing and demonstrated these two sub-types of ALAL possessed different mutation les. We anticipate that the genetic information revealed by high throughput sequencing will innovate the medical care of ALAL in the future.
Another important point of view of our study is the deep exploration of the diagnostic system of ALAL. EGIL classi cation guideline was rst proposed in 1995 and adopted in WHO criteria in 2001 2 . Updated WHO 2008/2016 classi cation implicated signi cant modi cation to the EGIL score system and quite a number of patients were eliminated outside ALAL diagnosis, just as we found in our study 4,5 . This part of patients was diagnosed as AML or ALL, and received corresponding treatment. However, whether the remarkable change is suitable and bene cial for the patients is still controversial, and the con ict focuses on the part of patients excluded by the new WHO system 3 . In this study, we found the patients excluded by WHO system had an even worse prognosis (Log rank p = 0.023 for PFS and Log rank p = 0.031 for OS), which was consistent with a previous reported clinical study 15 . In spite of this, we also found that the patients, satis ed the EGIL criteria but excluded by WHO diagnose system, shared the similar basic characteristics and immunophenotype with the patients included as showed in Table 1. Of the 9 patients excluded by WHO from ALAL, 2 cases underwent high throughput sequencing, and the results showed both of them carried completed genetic mutations distributed in genes like PTEN, KRAS, JAK3, KMT2C, WT1, ETV6, TP53, JAK3, RUX1, NOTCH, PHF6, and CREBBP. One of them showed no response to conventional chemotherapy, and unfortunately died 83 days after the diagnose. This is the rst study to investigate the genetic background of this part of patients, however, the number of patients is limited to describe the whole picture of these patients. As the diagnosis signi cantly dictates the therapeutic decision making, the clinical appliance of the WHO 2008/2016 criteria needs to be in more prospective clinical studies.
The role of genetic (cytogenetic and molecular) markers in guiding chemotherapy strategies as well as targeted therapy is inexperienced. According to the WHO2016 criteria, t (9; 22)/Ph + was considered as a separate entity. And the Ph chromosome was revealed as the most common cytogenetic abnormality, with the incidence ranging from 17-41% in MPAL especially with B-M phenotype 10 . All Ph + patients should be con rmed as soon as possible due to the bene t probability from TKI 16,17 . In our study, we not only screened out 7 abnormal karyotyping patients with Ph + resulting in BCR/ABL fusion gene, but also further detected 1 case with T315I mutation. Therefore, it is evident for us to make a better choice on the utility of TKIs. New molecular information in combination with cytogenetic characteristics will be valuable for the identi cation of genetic changes in ALAL, and undoubtedly play an important role in precision medicine of ALAL.
There is a proof that a patient with IDH1-mutated AUL achieved molecular complete remission following ivosidenib 19 . This might be a promising step to open the door for targeted therapy in IDH-mutated ALAL 20 .
Interestingly, we rst found E2A/PBX1 in MPAL, B/M, NOS, which has not been reported in ALAL before. To our knowledge, it is one of the most common translocations in pediatric B-ALL. In adult, although not exceptional, it appears approximately 3% 20 . E2A/PBX1 fusion gene is associated with unfavorable prognosis, eventually leading to the intensive therapeutic approaches. The prognostic signi cance of E2A/PBX1 fusion gene in ALAL is acquired further research.
Given that the drugable genetic target is still limited, analyzing the aberrant or speci c antigen expression on leukemia cells seems to be another potential strategy. We tried to detect some canonical surface markers which can be targeted by off-the-shelf chimeric antigen receptor T-cell (CAR-T) immunotherapy. Strikingly, we de nitely found 11 patients (47.8% of tested, data not shown) expressed CD123 on the surface of the blast cells. It has been reported that CD123, as a more speci c marker for AML cells, is expressed at low levels on normal hematopoietic stem/progenitor cells 21 . And CD123 CAR-T therapy exhibited a good effect in patients who are no longer responding to standard therapies 22,23 . It is potentially considered as a novel magic bullet for the therapeutics of ALAL with CD123 expression.
Collectively, as the ALAL appears heterogeneous, it seems reasonable to establish strati cation system of diagnosis and prognosis, especially based on the genetic features. Elucidating the genetic heterogeneity implicated in the process of leukemogenesis will provide insights into the pathogenesis and improve the management of the unusual subtype of acute leukemia.

Study Design and Participants
A retrospective analysis was performed based on data from 1635 de novo acute leukemia patients in Tongji Hospital (Wuhan, China) from January 2012 to June 2018. Standard procedures for diagnosis were performed, including morphology examination, multi-parameter ow cytometry analysis, molecular detection and karyotype analysis. According to EGIL or 2016 WHO criteria, 39 (2.4%) patients were enrolled in this study 1,4,5 . A total of 35 patients were designated as BAL according to EGIL scoring system and/or as MPAL if 2016 WHO criteria were ful lled and 4 patients were diagnosed as AUL. The clinical characteristics, treatments and prognostic information were objectively collected from the medical records and validated by the physicians in charge or the telephone follow-up to the patients. ALL-like therapy was de ned as regimens including vincristine, doxorubicin, cyclophosphamide, L-asparaginase, methotrexate and steroids, and AML-like therapy was de ned as initial chemotherapy including anthracyline, cytarabine, with or without hypomethylating (demethylating) agents.
The study design was approved by the Ethics Committee of Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology according to the Declaration of Helsinki and all patients gave written informed consents to participate in this study.
Targeted gene next-generation sequencing and mutation calling Among the 19 patients underwent second-generation sequencing, 3 patients carried next-generation sequencing reports when they were admitted to our hospital, the data of next-generation sequencing was obtained from their medical record. For the others, samples from 15 patients were sequenced in Ion Personal Genome Machine and 1 was sequences in Illumina NextSeq 500/550 platform in our center. The panel designed for variants detection covered the main mutation hotspots of a total of 173 genes which were the most frequently detected in myeloid and lymphoid malignancies (Supplementary Table 2). Genomic DNA was isolated from tumor blasts from bone marrow samples with the QIAamp DNA Blood Kit (Qiagen GmbH, Hilden, Germany). Quality veri ed DNA extracted from tumor cells of patients was diluted to 5 ng/µL and prepared for library construction with the Ion AmpliSeq™ Library Kit 2.0 (Applied Biosystems, Foster City, CA, USA). The concentration of library was measured by Qubit 3.0. All consumable items and reagents used in Library pooling, Emulsion PCR and Enrichment of template were provided in Ion PGM™ Hi-Q™ View OT2 Kit and used according to the manufacturer's instructions. The next-generation sequencing was conducted in the Ion torrent PGM (Applied Biosystems, Foster City, CA, USA) with Ion 318 chip. The procedure of transferring raw signal to base and reads alignment were conducted in the local server by default settings. The VCF (Variant Call Format) les containing the raw mutation information was uploaded to Ion Reporter server for further analysis. After getting MAF (Mutation Annotation Format) les, we ltered mutations by conditions: 1) MAF (Minor Allele Frequency) value < 0.01; 2) Mutation reads number > 10; 3) Coding region mutation; 4) Nonsynonymous mutation; 5) Not recorded in our false positive mutation database and human genome database of 1000 Genomes. IGV 24 software was used to screening all mutations in BAM to guarantee no false positive mutations to be left 25 .

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementaryMaterial.doc