Analysis of transcriptomic and genomic sequencing in acute myeloid leukemia

During the late breaking abstracts (LBA) session on Tuesday 10^th December 2019 at the 61^st American Society of Hematology (ASH) meeting, Orlando, US, Ilaria Iacobucci, St Jude’s Children’s Research Hospital, Memphis, US, presented LBA-4. This abstract summarizes the results from a genome-wide analysis of adult patients with acute myeloid leukemia (AML) and myelodysplastic syndromes (MDS).

Background¹

Our understanding of the genetic landscape of AML and MDS has advanced significantly in recent years. For example, in AML:

• 2005: gene-expression profiling and discovery of NPM1 mutations
• 2008: first AML genome sequenced
• 2013: analysis of the genomic and epigenomic landscape in AML was conducted
• 2016: AML genetic risk groups are identified
• 2017: European LeukemiaNet (ELN) recommendations are released
• 2018: BEAT AML project—using genomic information to determine signatures of drug response in AML

However, since most analysis is conducted by characterizing only specific subtypes and using targeted DNA-sequencing, it is less likely that novel mutational patterns and gene expression clusters will be identified.

Study design¹

This present study, part of the 5K Project, in association with the Munich Leukemia Laboratory, set out to analyze genetic alterations in AML and MDS by integrating genomic and transcriptome data with clinico-pathologic features and clinical outcomes. Overall, the study aimed to define myeloid leukemia subtypes of diagnostic, prognostic, and therapeutic relevance.

Unpaired whole genome sequencing and transcriptome sequencing was conducted in 1,304 patient samples. A tissue bank of patient-derived xenografts was also created to enable researchers to perform preclinical therapeutic studies.

• AML: n= 598
• Median age: 68 years (range, 17.8–93.1)
• De novo vs secondary AML: 97% vs 3%
• Classification according to World Health Organization (WHO) 2016 criteria:
  • AML not otherwise specified: 31.7%
  • NPM1-mutated: 27.3%
  • RUNX1-mutated: 9.7%
  • AML with known cytogenetic alterations: 29.9%
• MDS: n= 706
• Median age: 73.2 years (range, 23.3–93.1)
• Classification according to WHO 2016 criteria:
  • MDS with excess blasts: 37.0%
  • MDS with multilineage dysplasia: 20.9%
  • MDS with ring sideroblasts: 26.3%
  • MDS, 5q- syndrome: 14.6%
  • MDS unclassifiable: 1.1%

Results¹

• There were over 7,000 variants in 839 genes (33% of which were potential driver gene mutations)
  • Median of five mutations per case
• The frequency of mutations in NPM1, epigenomic modifiers, transcriptional regulators, and signaling molecules was higher in AML
• Conversely, mutations in splicing pathways were more common in MDS
• In addition, AML had distinct gene expression subgroups with the following mutated genes found in three favorable prognosis AML subtypes:
• RUNX1-RUNX1T1 leukemias: KIT, ZBTB7A, ASXL2, RAD21, CSF3R, and DNM2
• PML-RARA leukemias: FLT3, DDX54, WT1, and CALR
• CBFB rearranged leukemias: KIT, NRAS, and NF1
• The two most frequently mutated genes were:
  • ET2 (MDS > AML, p= 0.0011)
  • DNMT3A (AML > MDS, p< 0.0001)
  • These mutations were also enriched in NPM1-mutated leukemias

Classification of genetically different AML and MDS subtypes

NPM1 mutations

• NPM1 mutations accounted for 13% of cases overall
• AML: 27.4%
• MDS: 0.8%
• Additional mutations were found in PTPN11, IDH1/2, RAD21, and SMC1A
• Three mutually exclusive expression signatures were associated with mutations in RUNX1, TP53, and CEBPA

Overexpression of MN1

• In total, 9% of patients in the cohort had overexpression of MN1 (meningioma 1), a transcriptional coactivator
• Commonly co-occurring and mutually exclusive mutations occurred in RUNX1 (29%) and TP53 (23%)

RUNX1 mutations

• Overall, RUNX1 alterations occurred in 12.5% of patients with AML and 9.5% with MDS
• Most mutations occurred in the Runt domain and the Atrophin-1 domain

TP53 mutation

• TP53 mutations, mainly found in the DNA-binding domains, accounted for 12% of AML cases and 10% of MDS cases
• These mutations were often associated with a complex karyotype (AML vs MDS: 78% vs 49%) and older age

Chromosome 5q loss CREBBP/EP300 mutations

• Mutations in CREBBP/EP300 occurred in 3% of patients with AML and MDS
• Most frequently with MDS with excess blasts

Prognosis of genetically different AML and MDS subtypes

• Prognosis of MDS strictly dependent on combinations of mutations (genetic subtype)
• Co-occurring mutations in NPM1 and FLT3 were associated with poorer outcome
• In patients with AML and MDS, survival at 50 months was < 25%
• Co-occurring mutations in NPM1 and cohesin conferred a better outcome
• In patients with AML and MDS, survival at 50 months was ~ 75%
• Overexpression of MN1 was associated with a poor outcome
  • Survival rate (MNI high vs low): 21.2 vs1 months
• RUNX1 mutations were generally associated with a poor outcome in both AML and MDS
• In AML, survival at 100 months was ~ 40% without RUNX1 mutation and < 20% with RUNX1. This was also true in MDS with survival at 140 months of ~ 40% without RUNX1 mutations and < 5% with RUNX1
• In MDS, mutations in TP53 and RUNX1 or epigenomic modifiers confer the worst prognosis: median survival < 70 months
  • Transcription regulator mutations confer a better prognosis, with a median survival not reached at 420 months
• In AML and MDS CREBBP/E300 mutations were associated with a poor prognosis: median survival < 50 months and < 70 months respectively

Conclusion¹

This study has demonstrated that combining mutational and expression data from a large cohort of patients with AML and MDS can identify subtypes and constellations of mutations with prognostic significance. More specifically, it highlights three points:

1. The power of integrated gene expression/sequencing mutation for classifying AML and MDS
2. The importance of integrating structural variations into classification schema
3. The importance of combinatorial gene interactions for prognostication

Whilst this study confirmed some subgroups currently identified by the WHO and ELN classification system, it also identified additional genetic clusters that allow further improvement based on constellations of mutations with prognostic significance that supersede previously described classification systems.

Due to the extent of various mutations in different pathways, classification remains a challenge. However, further enhancing our understanding and refining classification will assist with improving diagnosis and therapeutic decisions in the future.