Romagnoli, Simone. "Identification of Structural Variants in Acute Myeloid Leukemia with normal karyotype patients by using long-reads sequencing technology." Doctoral thesis, Università di Siena, 2021. http://hdl.handle.net/11365/1157520.
Анотація:
Acute Myeloid Leukemia (AML) accounts for approximately 25% of all leukemias in adults in the Western world, and therefore is the most frequent form of blood neoplasia. Leukemic stem cells show abnormal proliferation, activation of antiapoptotic pathways and the impairment normal cell differentiation resulting in the dysregulated production of not functional blood cells, known as blast. AML is an aggressive disease, with a relative survival rate for all ages 5 years after diagnosis of 29.5%, the clinical manifestations of AML reflect the accumulation of malignant, poorly differentiated myeloid cells within the bone marrow, peripheral blood and in other organs. Diagnostic tests are mainly constituted by blood cells count and morphology, AML diagnosis is established by the presence of >=20% myeloid blasts in the bone marrow or peripheral blood. The prognostic assessment of AML patients is of capital importance for the management of the disease and to set up risk adapted therapies. Although clinical factors play an important role in disease development, karyotype is the most independent prognostic factor to forecast patients’ survival and it is adopted to provide the framework for risk-adapted treatment approach (Deschler and Lübbert, 2006; De Kouchkovsky and Abdul-Hay, 2016). The European Leukemia Net (ELN) guidelines aims to standardize risk stratification in adult AML patients by incorporating cytogenetic and known molecular abnormalities in hot spot genes. Accordingly, AML patients could be stratified into distinct prognostic risk groups (favorable, intermediate or adverse) based on their cytogenetic and molecular profile. Although this classification is the gold standard for the stratification of patients, it is fulfilled for only the 75% of AML whereas it is poorly satisfying for those patients resulted with normal karyotype (nk) at the conventional cytogenetic analysis. Normal karyotype AML (nkAML) patients mostly belong to the intermediate risk category but they experience an extremely heterogeneous outcome that represents an unmet needs in the clinical context of AML (De Kouchkovsky and Abdul-Hay, 2016; Döhner et al., 2017). In the last few years, large-scale tumour-sequencing studies have demonstrated that the majority of cancers, including hematologic neoplasia, are driven by Structural Variants (SVs) that are, for instance, genomic rearrange- ments larger than 50 bp. SVs include insertions, translocations, inversions and Copy Number Alterations (CNAs) (deletions and duplications). The recent development of high-throughput sequencing platforms provided impressive insights into leukemia pathogenesis and contributed to consider SVs as the hallmark of the genome instability leading to the establishment of the neoplasia. Beside karyotype, SVs detection is currently addressed by Next Generation Sequencing (NGS) technologies that allow the simultaneous and accurate detection of recurrent SVs breakpoints (Schütte et al., 2019), nothwithstanding, NGS faces inaccuracy and limitations when applied to resolve wide and structurally complex SVs due to the short length (100-500 bp) of the sequencing read employed (Norris et al., 2016).
In this study, we exploited the long-reads Oxford Nanopore Sequencing technology to explore the genome of a cohort of 152 AML patient with normal cytogenetics, aiming to address the genomic analysis challenges and to identify new potential genomic biomarkers able to refine the prognostic forecasting for nkAML patients. Of 152 bone marrow samples collected at diagnosis, 85 referred to the hematology unit of the A.O.U.Careggi and 67 were prospectively collected for the AML #1310 study by the Italian Hematologic Network GIMEMA (Venditti et al., 2019).
The DNA purified from nkAML samples was used to sequence the whole genome by the nanopore long-reads approach and further analysed by the bioinformatic pipeline specifically developed for SVs calling. Two SVs caller, Sniffles (Sedlazeck et al., 2018) and cuteSV (Jiang et al., 2020), were employed for the identification of an high-confidency callset of SVs that were further clustered and filtered before correlating them with patients’ outcome data. We employed an univariate Cox proportional-hazards analysis to weight the correlation between patients’ survival and each predictor variables. Further, to better estimate the cumulative impact of multiple genome and clinical variables, we developed a multi- variate Cox regression model including those SVs selected by Cox univariate model (pvalue <.05) and other predictors such as age, white blood cells count and the known molecular abnormalities in specific hotspot genes included in the ELN guidelines (Fms related Receptor Tyrosine Kinase 3 (FLT3)-ITD, Nucleophosmin 1 (NPM1), CCAAT Enhancer Binding Protein alpha (CEBPa)). Multivariate analysis allowed to select 12 SVs, represented by genomic deletions or insertions, with high impact on patients’ leukemia free and Overall Survival (OS). Of those, 8 resulted with an HR >1 (also referred as High Risk SVs (hrSVs)), thus associated with an increased risk of death, the other with an Hazard Ratio (HR) <1 (also referred as Low-risk SVs (lrSVs)) were associated to a reduced risk of death. The following stratification of the study cohort based on the presence of hrSVs enabled the identification of a high risk group of patients (accounting for the 17% of the cohort) with an extremely poor survival (median OS time 8.27 months for the group harbouring the hrSVs compared to 62.7 month fo the other, LogRank pvalue <.0001) and a low rate of response to therapy (46% for the patients with hrSVs compared to the 80%, pvalue <.0001). Taking together, these data suggest that the employ of an emerging long-reads sequencing technology capable to detect wide SVs together with a dedicated analysis pipeline could represent a powerful tool to accurately screen the whole genome of AML patients and identify new genomic biomark- ers for the prognostic assessment of nkAML patients capable to refine the actual ELN prognostic assessment in our cohort.
inversions and Copy Number Alterations (CNAs) (deletions and duplications). The recent development of high-throughput se- quencing platforms provided impressive insights into leukemia pathogenesis and contributed to consider SVs as the hallmark of the genome instability leading to the establishment of the neo- plasia. Beside karyotype, SVs detection is currently addressed by Next Generation Sequencing (NGS) technologies that allow the simultaneous and accurate detection of recurrent SVs breakpoints (Schütte et al., 2019), nothwithstanding, NGS faces inaccuracy and limitations when applied to resolve wide and structurally com- plex SVs due to the short length (100-500 bp) of the sequencing read employed (Norris et al., 2016).
In this study, we exploited the long-reads Oxford Nanopore Se- quencing technology to explore the genome of a cohort of 152 AML patient with normal cytogenetics, aiming to address the genomic analysis challenges and to identify new potential genomic biomarkers able to refine the prognostic forecasting for nkAML patients. Of 152 bone marrow samples collected at diagnosis, 85 referred to the hematology unit of the A.O.U.Careggi and 67 were prospectively collected for the AML #1310 study by the Italian Hematologic Network GIMEMA (Venditti et al., 2019).
The DNA purified from nkAML samples was used to sequence the whole genome by the nanopore long-reads approach and further analysed by the bioinformatic pipeline specifically developed for SVs calling. Two SVs caller, Sniffles (Sedlazeck et al., 2018) and cuteSV (Jiang et al., 2020), were employed for the identification of an high-confidency call-set of SVs that were further clustered and filtered before correlating them with patients’ outcome data. We employed an univariate Cox proportional-hazards analysis to weight the correlation between patients’ survival and each predic- tor variables. Further, to better estimate the cumulative impact of multiple genome and clinical variables, we developed a multi- variate Cox regression model including those SVs selected by Cox univariate model (pvalue <.05) and other predictors such as age, white blood cells count and the known molecular abnormalities in specific hotspot genes included in the ELN guidelines (Fms related Receptor Tyrosine Kinase 3 (FLT3)-ITD, Nucleophosmin 1 (NPM1), CCAAT Enhancer Binding Protein alpha (CEBPa)). Multivariate analysis allowed to select 12 SVs, represented by genomic deletions or insertions, with high impact on patients’
leukemia free and Overall Survival (OS). Of those, 8 resulted with an HR >1 (also referred as High Risk SVs (hrSVs)), thus associ- ated with an increased risk of death, the other with an Hazard Ratio (HR) <1 (also referred as Low-risk SVs (lrSVs)) were as- sociated to a reduced risk of death. The following stratification of the study cohort based on the presence of hrSVs enabled the identification of a high risk group of patients (accounting for the 17% of the cohort) with an extremely poor survival (median OS time 8.27 months for the group harbouring the hrSVs compared to
62.7 month fo the other, LogRank pvalue <.0001) and a low rate of response to therapy (46% for the patients with hrSVs compared to the 80%, pvalue <.0001). Taking together, these data suggest that the employ of an emerging long-reads sequencing technology capable to detect wide SVs together with a dedicated analysis pipeline could represent a powerful tool to accurately screen the whole genome of AML patients and identify new genomic biomark- ers for the prognostic assessment of nkAML patients capable to refine the actual ELN prognostic assessment in our cohort.