The inclusion of high-dimensional omics data in prediction models has become a well-studied topic in the last decades. Although most of these methods do not account for possibly different types of ...variables in the set of covariates available in the same dataset, there are many such scenarios where the variables can be structured in blocks of different types, e.g., clinical, transcriptomic, and methylation data. To date, there exist a few computationally intensive approaches that make use of block structures of this kind.
In this paper we present priority-Lasso, an intuitive and practical analysis strategy for building prediction models based on Lasso that takes such block structures into account. It requires the definition of a priority order of blocks of data. Lasso models are calculated successively for every block and the fitted values of every step are included as an offset in the fit of the next step. We apply priority-Lasso in different settings on an acute myeloid leukemia (AML) dataset consisting of clinical variables, cytogenetics, gene mutations and expression variables, and compare its performance on an independent validation dataset to the performance of standard Lasso models.
The results show that priority-Lasso is able to keep pace with Lasso in terms of prediction accuracy. Variables of blocks with higher priorities are favored over variables of blocks with lower priority, which results in easily usable and transportable models for clinical practice.
The clinical and prognostic relevance of many recently identified driver gene mutations in adult acute myeloid leukemia (AML) is poorly defined. We sequenced the coding regions or hotspot areas of 68 ...recurrently mutated genes in a cohort of 664 patients aged 18 to 86 years treated on 2 phase 3 trials of the German AML Cooperative Group (AMLCG). The median number of 4 mutations per patient varied according to cytogenetic subgroup, age, and history of previous hematologic disorder or antineoplastic therapy. We found patterns of significantly comutated driver genes suggesting functional synergism. Conversely, we identified 8 virtually nonoverlapping patient subgroups, jointly comprising 78% of AML patients, that are defined by mutually exclusive genetic alterations. These subgroups, likely representing distinct underlying pathways of leukemogenesis, show widely divergent outcomes. Furthermore, we provide detailed information on associations between gene mutations, clinical patient characteristics, and therapeutic outcomes in this large cohort of uniformly treated AML patients. In multivariate analyses including a comprehensive set of molecular and clinical variables, we identified DNMT3A and RUNX1 mutations as important predictors of shorter overall survival (OS) in AML patients <60 years, and particularly in those with intermediate-risk cytogenetics. NPM1 mutations in the absence of FLT3-ITD, mutated TP53, and biallelic CEBPA mutations were identified as important molecular prognosticators of OS irrespective of patient age. In summary, our study provides a comprehensive overview of the spectrum, clinical associations, and prognostic relevance of recurrent driver gene mutations in a large cohort representing a broad spectrum and age range of intensively treated AML patients.
•We present comprehensive information on genetic driver events in a uniformly treated cohort of 664 adult AML patients aged 18 to 86 years.•Mutations in NPM1, FLT3, CEBPA, TP53, and, in patients <60 years, DNMT3A and RUNX1, are the most important molecular risk factors in AML.
The revised 2017 European LeukemiaNet (ELN) recommendations for genetic risk stratification of acute myeloid leukemia have been widely adopted, but have not yet been validated in large cohorts of AML ...patients. We studied 1116 newly diagnosed AML patients (age range, 18-86 years) who had received induction chemotherapy. Among 771 patients not selected by genetics, the ELN-2017 classification re-assigned 26.5% of patients into a more favorable or, more commonly, a more adverse-risk group compared with the ELN-2010 recommendations. Forty percent of the cohort, and 51% of patients ≥60 years, were classified as adverse-risk by ELN-2017. In 599 patients <60 years, estimated 5-year overall survival (OS) was 64% for ELN-2017 favorable, 42% for intermediate-risk and 20% for adverse-risk patients. Among 517 patients aged ≥60 years, corresponding 5-year OS rates were 37, 16, and 6%. Patients with biallelic CEBPA mutations or inv(16) had particularly favorable outcomes, while patients with mutated TP53 and a complex karyotype had especially poor prognosis. DNMT3A mutations associated with inferior OS within each ELN-2017 risk group. Our results validate the prognostic significance of the revised ELN-2017 risk classification in AML patients receiving induction chemotherapy across a broad age range. Further refinement of the ELN-2017 risk classification is possible.
Some patients with acute myeloid leukemia (AML) who are in complete remission after induction chemotherapy harbor persisting pre-leukemic clones, carrying a subset of leukemia-associated somatic ...mutations. There is conflicting evidence on the prognostic relevance of these clones for AML relapse. Here, we characterized paired pre-treatment and remission samples from 126 AML patients for mutations in 68 leukemia-associated genes. Fifty patients (40%) retained ≥1 mutation during remission at a VAF of ≥2%. Mutation persistence was most frequent in DNMT3A (65% of patients with mutations at diagnosis), SRSF2 (64%), TET2 (55%), and ASXL1 (46%), and significantly associated with older age (p < 0.0001) and, in multivariate analyses adjusting for age, genetic risk, and allogeneic transplantation, with inferior relapse-free survival (hazard ratio (HR), 2.34; p = 0.0039) and overall survival (HR, 2.14; p = 0.036). Patients with persisting mutations had a higher cumulative incidence of relapse before, but not after allogeneic stem cell transplantation. Our work underlines the relevance of mutation persistence during first remission as a novel risk factor in AML. Persistence of pre-leukemic clones may contribute to the inferior outcome of elderly AML patients. Allogeneic transplantation abrogated the increased relapse risk associated with persisting pre-leukemic clones, suggesting that mutation persistence may guide post-remission treatment.
A cute myeloid leukemia is a disease of the elderly (median age at diagnosis, 65-70 years). The prognosis of older acute myeloid leukemia patients is generally poor. While genetic markers have become ...important tools for risk stratification and treatment selection in young and middle-aged patients, their applicability in very old patients is less clear. We sought to validate existing genetic risk classification systems and identify additional factors associated with outcomes in intensively treated patients aged ≥75 years. In 151 patients who received induction chemotherapy in the AMLCG-1999 trial, we investigated recurrently mutated genes using a targeted sequencing assay covering 64 genes. The median number of mutated genes per patient was four. The most commonly mutated genes were
(42%),
(35%),
(32%),
(25%) and
(21%). The complete remission rate was 44% and the 3-year survival was 21% for the entire cohort. While adverse-risk cytogenetics (MRC classification) were associated with shorter overall survival (
=0.001),
and
-ITD mutations (present in 18%) did not have a significant impact on overall survival. Notably, none of the 13
-mutated patients (9%) reached complete remission. Consequently, the overall survival of this subgroup was significantly shorter than that of
-wildtype patients (
<0.001). In summary, even among very old, intensively treated, acute myeloid leukemia patients, adverse-risk cytogenetics predict inferior survival. The spectrum and relevance of driver gene mutations in elderly patients differs from that in younger patients. Our data implicate
mutations as a novel marker for chemorefractory disease and inferior prognosis. (AMLCG-1999 trial: clinicaltrials.gov identifier, NCT00266136).
MicroRNAs are subject to precise regulation and have key roles in tumorigenesis. In contrast to the oncogenic role of miR-22 reported in myelodysplastic syndrome (MDS) and breast cancer, here we show ...that miR-22 is an essential anti-tumour gatekeeper in de novo acute myeloid leukaemia (AML) where it is significantly downregulated. Forced expression of miR-22 significantly suppresses leukaemic cell viability and growth in vitro, and substantially inhibits leukaemia development and maintenance in vivo. Mechanistically, miR-22 targets multiple oncogenes, including CRTC1, FLT3 and MYCBP, and thus represses the CREB and MYC pathways. The downregulation of miR-22 in AML is caused by TET1/GFI1/EZH2/SIN3A-mediated epigenetic repression and/or DNA copy-number loss. Furthermore, nanoparticles carrying miR-22 oligos significantly inhibit leukaemia progression in vivo. Together, our study uncovers a TET1/GFI1/EZH2/SIN3A/miR-22/CREB-MYC signalling circuit and thereby provides insights into epigenetic/genetic mechanisms underlying the pathogenesis of AML, and also highlights the clinical potential of miR-22-based AML therapy.
Philadelphia-like B-cell precursor acute lymphoblastic leukemia (Ph-like ALL) is characterized by distinct genetic alterations and inferior prognosis in children and younger adults. The purpose of ...this study was a genetic and clinical characterization of Ph-like ALL in adults. Twenty-six (13%) of 207 adult patients (median age: 42 years) with B-cell precursor ALL (BCP-ALL) were classified as having Ph-like ALL using gene expression profiling. The frequency of Ph-like ALL was 27% among 95 BCP-ALL patients negative for BCR-ABL1 and KMT2A-rearrangements. IGH-CRLF2 rearrangements (6/16; P=0.002) and mutations in JAK2 (7/16; P<0.001) were found exclusively in the Ph-like ALL subgroup. Clinical and outcome analyses were restricted to patients treated in German Multicenter Study Group for Adult ALL (GMALL) trials 06/99 and 07/03 (n=107). The complete remission rate was 100% among both Ph-like ALL patients (n=19) and the "remaining BCP-ALL" cases (n=40), i.e. patients negative for BCR-ABL1 and KMT2A-rearrangements and the Ph-like subtype. Significantly fewer Ph-like ALL patients reached molecular complete remission (33% versus 79%; P=0.02) and had a lower probability of continuous complete remission (26% versus 60%; P=0.03) and overall survival (22% versus 64%; P=0.006) at 5 years compared to the remaining BCP-ALL patients. The profile of genetic lesions in adults with Ph-like ALL, including older adults, resembles that of pediatric Ph-like ALL and differs from the profile in the remaining BCP-ALL. Our study is the first to demonstrate that Ph-like ALL is associated with inferior outcomes in intensively treated older adult patients. Ph-like adult ALL should be recognized as a distinct, high-risk entity and further research on improved diagnostic and therapeutic approaches is needed. (NCT00199056, NCT00198991).
Primary therapy resistance is a major problem in acute myeloid leukemia treatment. We set out to develop a powerful and robust predictor for therapy resistance for intensively treated adult patients. ...We used two large gene expression data sets (n=856) to develop a predictor of therapy resistance, which was validated in an independent cohort analyzed by RNA sequencing (n=250). In addition to gene expression markers, standard clinical and laboratory variables as well as the mutation status of 68 genes were considered during construction of the model. The final predictor (PS29MRC) consisted of 29 gene expression markers and a cytogenetic risk classification. A continuous predictor is calculated as a weighted linear sum of the individual variables. In addition, a cut off was defined to divide patients into a high-risk and a low-risk group for resistant disease. PS29MRC was highly significant in the validation set, both as a continuous score (OR=2.39,
=8.63·10
, AUC=0.76) and as a dichotomous classifier (OR=8.03,
=4.29·10
); accuracy was 77%. In multivariable models, only
mutation, age and PS29MRC (continuous: OR=1.75,
=0.0011; dichotomous: OR=4.44,
=0.00021) were left as significant variables. PS29MRC dominated all models when compared with currently used predictors, and also predicted overall survival independently of established markers. When integrated into the European LeukemiaNet (ELN) 2017 genetic risk stratification, four groups (median survival of 8, 18, 41 months, and not reached) could be defined (
=4.01·10
). PS29MRC will make it possible to design trials which stratify induction treatment according to the probability of response, and refines the ELN 2017 classification.
Antibody-dependent cellular phagocytosis (ADCP) by macrophages, an important effector function of tumor targeting antibodies, is hampered by ‘Don´t Eat Me!’ signals such as CD47 expressed by cancer ...cells. Yet, human leukocyte antigen (HLA) class I expression may also impair ADCP by engaging leukocyte immunoglobulin-like receptor subfamily B (LILRB) member 1 (LILRB1) or LILRB2. Analysis of different lymphoma cell lines revealed that the ratio of CD20 to HLA class I cell surface molecules determined the sensitivity to ADCP by the combination of rituximab and an Fc-silent variant of the CD47 antibody magrolimab (CD47-IgGσ). To boost ADCP, Fc-silent antibodies against LILRB1 and LILRB2 were generated (LILRB1-IgGσ and LILRB2-IgGσ, respectively). While LILRB2-IgGσ was not effective, LILRB1-IgGσ significantly enhanced ADCP of lymphoma cell lines when combined with both rituximab and CD47-IgGσ. LILRB1-IgGσ promoted serial engulfment of lymphoma cells and potentiated ADCP by non-polarized M0 as well as polarized M1 and M2 macrophages, but required CD47 co-blockade and the presence of the CD20 antibody. Importantly, complementing rituximab and CD47-IgGσ, LILRB1-IgGσ increased ADCP of chronic lymphocytic leukemia (CLL) or lymphoma cells isolated from patients. Thus, dual checkpoint blockade of CD47 and LILRB1 may be promising to improve antibody therapy of CLL and lymphomas through enhancing ADCP by macrophages.