One of the most important application spectrums of transcriptomic data is cancer phenotype classification. Many characteristics of transcriptomic data, such as redundant features and technical ...artifacts, make over-fitting commonplace. Promising classification results often fail to generalize across datasets with different sources, platforms, or preprocessing. Recently a novel differential network rank conservation (DIRAC) algorithm to characterize cancer phenotypes using transcriptomic data. DIRAC is a member of a family of algorithms that have shown useful for disease classification based on the relative expression of genes. Combining the robustness of this family's simple decision rules with known biological relationships, this systems approach identifies interpretable, yet highly discriminate networks. While DIRAC has been briefly employed for several classification problems in the original paper, the potentials of DIRAC in cancer phenotype classification, and especially robustness against artifacts in transcriptomic data have not been fully characterized yet.
In this study we thoroughly investigate the potentials of DIRAC by applying it to multiple datasets, and examine the variations in classification performances when datasets are (i) treated and untreated for batch effect; (ii) preprocessed with different techniques. We also propose the first DIRAC-based classifier to integrate multiple networks. We show that the DIRAC-based classifier is very robust in the examined scenarios. To our surprise, the trained DIRAC-based classifier even translated well to a dataset with different biological characteristics in the presence of substantial batch effects that, as shown here, plagued the standard expression value based classifier. In addition, the DIRAC-based classifier, because of the integrated biological information, also suggests pathways to target in specific subtypes, which may enhance the establishment of personalized therapy in diseases such as pediatric AML. In order to better comprehend the prediction power of the DIRAC-based classifier in general, we also performed classifications using publicly available datasets from breast and lung cancer. Furthermore, multiple well-known classification algorithms were utilized to create an ideal test bed for comparing the DIRAC-based classifier with the standard gene expression value based classifier. We observed that the DIRAC-based classifier greatly outperforms its rival.
Based on our experiments with multiple datasets, we propose that DIRAC is a promising solution to the lack of generalizability in classification efforts that uses transcriptomic data. We believe that superior performances presented in this study may motivate other to initiate a new aline of research to explore the untapped power of DIRAC in a broad range of cancer types.
Deregulated microRNA (miRNA) expression has been implicated in the pathogenesis of acute myeloid leukemia (AML). We previously showed that miR-193b is a STAT5-regulated miRNA that controls ...hematopoietic stem and progenitor cell (HSPC) expansion by modulating cytokine receptor signaling. Here we demonstrate that the miR-193 family members miR-193a and 193b are potent tumor suppressors in AML. Both miRNAs were downregulated in several cytogenetically-defined subgroups of pediatric and adult AML (n=202), whereas low miR-193b expression was an independent indicator for poor prognosis and survival. Accordingly, ectopic retroviral Hoxa9-Meis1 expression in HSPCs from miR-193b-/- mice resulted in a more aggressive disease with significantly shortened latency and survival as compared to miR-193bWT/WT HSPCs. Inversely, ectopic miR-193 expression in leukemic cells belonging to various AML subgroups decreased leukemic growth in vitro and prolonged survival of mice suffering from Hoxa9-Meis1-induced leukemia through a G1/S phase block. These effects were mediated by targeting c-KIT, KRAS and SOS2 - key factors of the KIT-RAS-RAF-MEK-ERK signaling cascade - as well as the downstream cell cycle regulator CCND1. Knockdown of each of these genes partially recapitulated the anti-proliferative effect of ectopic lentiviral miR-193 expression. As the tumor suppressive function is independent of patient age or AML cytogenetic background, these observations suggest an opportunistic role for miR-193 in future AML therapies. With the notion that a single miRNA can control aberrant MAPK signaling at multiple levels, restoring miR-193 expression in AML cells with constitutive activation of this cascade would assure high antileukemic efficacy, while avoiding the fast development of resistance mechanisms.
Heuser:Bayer Pharma AG: Research Funding; Novartis: Consultancy, Research Funding; BerGenBio: Research Funding; Tetralogic: Research Funding; Karyopharm Therapeutics Inc: Research Funding; Celgene: Honoraria; Pfizer: Research Funding. Mulaw:NuGEN: Honoraria. Baruchel:Jazz: Consultancy; Servier: Consultancy; Celgene: Consultancy; Novartis: Consultancy; Baxalta: Research Funding.
Pediatric acute myeloid leukemia (AML) is a heterogeneous disease with respect to biology as well as outcome. In this study, we investigated whether known biological subgroups of pediatric AML are ...reflected by a common microRNA (miRNA) expression pattern. We assayed 665 miRNAs on 165 pediatric AML samples. First, unsupervised clustering was performed to identify patient clusters with common miRNA expression profiles. Our analysis unraveled 14 clusters, seven of which had a known (cyto-)genetic denominator. Finally, a robust classifier was constructed to discriminate six molecular aberration groups: 11q23-rearrangements, t(8;21)(q22;q22), inv(16)(p13q22), t(15;17) (q21;q22), NPM1 and CEBPA mutations. The classifier achieved accuracies of 89%, 95%, 95%, 98%, 91% and 96%, respectively. Although lower sensitivities were obtained for the NPM1 and CEBPA (32% and 66%), relatively high sensitivities (84%-94%) were attained for the rest. Specificity was high in all groups (87%-100%). Due to a robust double-loop cross validation procedure employed, the classifier only employed 47 miRNAs to achieve the aforementioned accuracies. To validate the 47 miRNA signatures, we applied them to a publicly available adult AML dataset. Albeit partial overlap of the array platforms and molecular differences between pediatric and adult AML, the signatures performed reasonably well. This corroborates our claim that the identified miRNA signatures are not dominated by sample size bias in the pediatric AML dataset. In conclusion, cytogenetic subtypes of pediatric AML have distinct miRNA expression patterns. Reproducibility of the miRNA signatures in adult dataset suggests that the respective aberrations have a similar biology both in pediatric and adult AML.
The most important reason for therapy failure in pediatric acute myeloid leukemia (AML) is relapse. In order to identify miRNAs that contribute to the clonal evolution towards relapse in pediatric ...AML, miRNA expression profiling of 127 de novo pediatric AML cases were used. In the diagnostic phase, no miRNA signatures could be identified that were predictive for relapse occurrence, in a large pediatric cohort, nor in a nested mixed lineage leukemia (MLL)-rearranged pediatric cohort. AML with MLL- rearrangements are found in 15-20% of all pediatric AML samples, and reveal a relapse rate up to 50% for certain translocation partner subgroups. Therefore, microRNA expression profiling of six paired initial diagnosis-relapse MLL-rearranged pediatric AML samples (test cohort) and additional eight paired initial diagnosis-relapse samples with MLL-rearrangements (validation cohort) was performed. A list of 53 differentially expressed miRNAs was identified of which the miR-106b~25 cluster, located in intron 13 of MCM7, was the most prominent. These differentially expressed miRNAs however could not predict a relapse in de novo AML samples with MLL-rearrangements at diagnosis. Furthermore, higher mRNA expression of both MCM7 and its upstream regulator E2F1 was found in relapse samples with MLL-rearrangements. In conclusion, we identified the miR-106b~25 cluster to be upregulated in relapse pediatric AML with MLL-rearrangements.
Abstract
Mixed lineage leukemia (MLL) rearrangements represent about 20% of pediatric acute myeloid leukemia (AML) cases. Survival rates have increased over the past decades due to intensified ...chemotherapy protocols and improved supportive care. Still, approximately 30-40% will relapse during or after therapy.
The role of microRNAs (miRNAs) in leukemogenesis of MLL-rearranged AML, especially towards the development of relapse is unknown. To determine whether specific miRNAs are involved in relapse development, we performed miRNA profiling, using TaqMan Low Density Array (TLDA) in pediatric AML with a special focus on MLL-rearranged cases. First, we compared miRNA profiles of de novo pediatric AML patients that relapsed with those that did not relapse. Secondly, we did the same comparison, only focussing on the MLL-rearranged subset. Third, we investigated the role of miRNAs in clonal evolution by investigating the differential expression of 6 paired initial diagnosis-relapses MLL-rearranged AML cases. An independent set of 6 paired initial diagnosis-relapse cases with MLL-rearrangements was used to confirm the identified miRNAs using single stemloop RT-qPCR.
In a cohort of 127 de novo AML cases, 59/127 (46.5%) relapsed after complete remission. Comparing miRNA profiles of relapsing and non-relapsing cases, we could not identify differentially expressed miRNAs signatures between the two groups at diagnosis (p=0.643). Likewise, in a nested cohort of de novo pediatric AML MLL-rearranged cases (4 AF6, 7 AF9, 9 AF10, 1 FNBP1, 1 SEPT6, and 1 NRIP3), of which 14/23 (60.9%) relapsed we did not identify differentially expressed microRNAs (p=0.429).
In the 6 paired initial diagnosis-relapse MLL-rearranged cases (2 AF9 and 4 AF10), 53 miRNAs were significantly differentially expressed (FDR <0.1).
Among those, miR-106b, miR-93 and miR-25, were most highly overexpressed at relapse. These microRNAs cluster together in intron 13 of minichromosome maintenance complex component 7 (MCM7) and are actively cotranscribed in the context of MCM7 primary RNA transcript. E2F transcription factor 1 (E2F1) acts as an upstream regulator. Overexpression of MCM7has been associated with poor prognosis in solid cancers and this may be linked to overexpression of the hosted microRNAs, the miR-106b~25 cluster. A possible role of the miR-106b~25 cluster in leukemia has so far not been investigated.
Therefore, we validated the expression of MCM7 and E2F1 mRNA (RT-qPCR) and found them to be differentially overexpressed in the paired samples with MLL-rearrangements at relapse (n=12, p=0.006 and p=0.003, respectively). Overexpression of miR-106b and miR-25 at relapse was confirmed by stem loop RT-qPCR in the patient samples as well as in 6 additional paired samples with MLL-rearrangements at relapse (4 AF10, 1 ELL, and 1 unknown)(n=12, p<0.05). MiR-93 overexpression could however not be confirmed by stem loop RT-qPCR (p=0.08).
Expression levels of two other predicted targets of the miR-106b~25 cluster as taken from the literature (Petrocca et al., 2008), cyclin-dependent kinase inhibitor 1A (p21WAF1/CIP1) and BCL2-like 11 (BIM), were analysed in 12 paired initial diagnosis-relapse samples. We did not find differences in mRNA expression of p21WAF1/CIP1 (p=0.17) or BIM (p=0.27) between initial diagnosis and relapse. As microRNAs regulate genes expression through translational repression or target mRNA cleavage we investigated protein level (Western blot, n=3 paired samples) of p21WAF1/CIP1, BIM, MCM7, and E2F1. In case of mRNA cleavage, a reverse correlation between predicted target gene and miRNA is expected. In 2/3 paired initial diagnosis-relapse samples the expression of p21WAF1/CIP1, BIM, E2F1, and MCM7 was downregulated in relapse samples as compared to initial diagnosis. Only one patient showed increased protein expression of E2F1 and MCM7, a modest downregulation of BIM, and no expression of p21WAF1/CIP1.
Together, our data indicate that miR-106b-25 cluster may play an important role in relapse pediatric AML with MLL-rearrangements. Further research is warranted to identify the role of this cluster for clinical resistance and as treatment target.
No relevant conflicts of interest to declare.
Primary refractory and relapsed pediatric acute myeloid leukemia (AML) still lead to a significant number of childhood cancer deaths, despite the current chemotherapeutic regimens. AML leukemogenesis ...is driven by collaborative genetic abnormalities that induce hematopoietic maturation arrest and cell proliferation. Particular AML-associated maturation inhibiting aberrations are known to target chromatin regulators, thus directly influencing the transcriptional program of leukemic cells. Therapies targeting epigenetic processes, e.g. with hypomethylation-inducing agents, are therefore becoming an attractive therapeutic strategy in adult AML. AML biology in children is not equivalent to that of adults, thus methylation patterns seen in adult AML cannot be extrapolated to pediatric AML. Therefore there is a need to unravel the mechanism behind changes in epigenetic processes as the result of AML-causing genetic abnormalities in order to develop new drugs for pediatric AML.
We hypothesized that pediatric AML samples have distinct DNA-methylation patterns which may provide a rationale for treatment with demethylating agents in specific pediatric AML subtypes. Furthermore, these differences in methylation could be characteristic for AML subgroups and that particular methylation patterns drive the expression of specific genes which may play a key role in the tumorigenesis of these AML leukemias.
We performed genome-wide CpG-island methylation profiling on a representative and molecularly characterized cohort of pediatric patients with de novo AML. Empirical Bayes Wilcoxon rank-sum test showed that AML patients carrying inv(16)(p13;q22) (n=9) have distinct DNA methylation patterns when compared to non-inv(16) AML patients (n=143) (consisting mainly of MLL-rearranged, t(8;21), t(15;17), t(8;16) AML and AML cases with a normal karyotype). The MN1 gene ranked as most significantly differentially methylated in inv(16) AML compared to non-inv(16) AML, with inv(16) AML cases having significantly (p=2x10-6) lower methylation levels compared to non-inv(16) AML cases. Hypomethylation of specific regions of the MN1-associated CpG-island was confirmed by methylation specific PCR and bisulfite sequencing. Subsequent gene expression (GEP) data on 294 pediatric AML patients showed that MN1 was 8 fold higher expressed in patients carrying inv(16) compared to all other patients (9.9, n=35 vs 6.9, n=259, p<0.001). Furthermore, integrating GEP and methylation array data showed that MN1 expression negatively correlated (ρs= 0.82, p=0.011) with methylation levels, which is in agreement with the biological assumption of methylation and gene expression.
Since genes known to regulate DNA methylation have frequently been shown to be mutated in adult AML we determined whether a decreased expression of DNA methyltransferases, DNMT1, DNMT2, DNMT3A, DNMT3B, could be the cause of a hypometylated MN1 locus in inv(16) AML. Our findings show that only DNMT3B expression was significantly (p=8x10-15) lower in inv(16) cases compared to non-inv(16) cases. To test whether hypomethylation of the MN1 CpG-island and the overexpression of MN1 is the result of decreased DNMT1 expression, HL60 cells which express negligible levels of MN1 were treated with the DNMT1 inhibitor Decitabine. This showed that treatment of HL60 cells with Decitabine led to increase of MN1 transcript levels, however, not as high as those observed in patient samples. This suggests that DNMT1 activity may not be the only DNA methyltransferase influencing expression of MN1 in inv(16) patients. Interestingly, we observed a high (ρs= 0.42) correlation between MN1 methylation and DNMT3B expression, which suggests DNMT3B could be an important DNA methyltransferase involved in regulating MN1expression.
Overall we show that pediatric AML patients carrying and inv(16) have a characteristic DNA methylation pattern compared to other AML patients carrying specific cytogenetic aberrations. Furthermore, our data suggest that hypomethylation of the MN1 gene is an underlying mechanism for high MN1 expression in inv(16)(p13;q22) patients possibly regulated by multiple DNA methyltransferases.
No relevant conflicts of interest to declare.
Pediatric acute myeloid leukemia (AML) is a heterogeneous disease and 30-40% of the patients still die. Prognosis is dependent on relevant genetic aberrations. Although many driving genetic ...alterations causing AML have been defined, in ~20% of the pediatric AML patients the oncogenic events remain unidentified. The ETS-Variant gene 6 (ETV6) encodes a transcription factor that functions as a tumor suppressor gene and is required for proper hematopoiesis in the bone marrow niche. Point mutations, deletions and translocations can lead to silencing of the gene, resulting in loss of transcriptional repression activity. ETV6 aberrations strongly associate with leukemia. In pediatric B-cell precursor acute lymphoblastic leukemia, translocation ETV6/RUNX1 occurs in ~25% of cases. Mutations in ETV6 are identified in ~25% of early immature T-cell ALL and also reported as event in adult AML (Van Vlierberghe et al, J Exp Med 2011; Barjesteh van Waalwijk van Doorn-Khosrovani et al, Oncogene 2005). We previously reported that pediatric AML patients can be divided in three clusters based on HOX-expression; (1) low HOXA/B expression, (2) high HOXA and low HOXB expression, and (3) high HOXA/B expression, and identified new repetitive genetic abnormalities in the third cluster, especially in NUP98. Cluster 1 is mainly represented by core-binding factor (CBF) AML, but in ~20% of these cases we did not find specific genetic abnormalities. Helton et al presented ETV6 aberrations in pediatric CBF-AML at ASH 2011, identified with whole genome sequencing, and with poor clinical outcome. We hypothesized that ETV6 aberrations might reduce the number of patients without known driving abnormality, especially in the low HOXA/B cluster.
We screened a large representative de novo pediatric AML cohort for ETV6 mutations in exons 2-8 with direct sequencing, for ETV6 deletions by multiplex ligation-dependent probe amplification and for ETV6 translocations using split signal FISH, and analyzed outcome.
In a well-characterized de novo pediatric AML cases with available gene-expression data, 6/275 (2.2%) patients had mutations affecting the predicted amino acid sequence of ETV6 and one had a silent mutation, 4/259 (1.5%) had an ETV6 deletion and 6/65 (9.2%) patients an MNX1/ETV6 translocation. Additionally, we identified 3 cases with a positive split signal FISH suggestive of a break in which ETV6 is involved, and a similar gene expression profile was found in these three cases. The aberrations of ETV6 were seen in patients of all three HOX-groups; n=9, n=6 and n=4 for cluster 1, 2 and 3 respectively.
In patients with an ETV6 mutation (n=6) or deletion (n=4) 13 and 38 genes, respectively, were significantly up-regulated, including CLDN5,DPEP1 and BIRC7. This is consistent with the up-regulated genes in functional studies silencing ETV6 in LOUCY cells (Van Vlierberghe et al, J Exp Med 2011). High expression of BIRC7 has been associated with poor prognosis in adult acute leukemia (El-Mesallamy et al, Leuk Res 2011).
The median age of patients with an ETV6-mutation or deletion (n=10) was 11.3 years (range 4.0-15.3) and 40% were female. Median WBC was significantly lower (15.1x109/L vs 47.0x109/L, p<0.01) in comparison to other pediatric AML cases. Other cytogenetic aberrations found in the ETV6-mutated or deleted cases were RUNX1/RUNX1T1 (n=3), PML/RARA (n=1), MLL/AF6 (n=1) and one case with an NPM1-mutation. Six out of ten patients encountered a relapse and one patient died of treatment-related mortality. The median age for patients with an ETV6-split signal FISH (n=9) was significantly lower compared to other pediatric AML cases (median 1.1 years vs 9.8, p<0.01), median WBC and sex did not differ. Five out of 9 encountered relapsed/refractory disease whereas 2 cases died of treatment-related mortality. The 3-yr pOS for all ETV6-aberrated patients taken together (n=19) was 37±11% vs 65±3% for the other pediatric AML patients (n=242, p<0.01); the 3-yr pEFS was 26±10% vs 46±3% (p=0.07), and 3-yr pCIR 47±13% vs 37±3% (p=0.24).
We conclude that ETV6 aberrations are rare but recurrent in pediatric AML. ETV6 aberrations predict a poor survival, although there was no evidence for an increased relapse incidence in this small cohort.
No relevant conflicts of interest to declare.
IKAROS family zinc finger 1 (IKZF1) is a transcription factor involved in lymphoid differentiation that acts as a tumor suppressor. In B-cell precursor acute lymphoblastic leukemia (BCP-ALL) loss of ...IKZF1 is found in ~15% of the patients, is associated with the presence of BCR/ABL1 (t(9;22)(q34;q11)) and confers a poor clinical outcome. Recent studies suggest that IKZF1 is also involved in myeloid differentiation. The best indication that loss of IKZF1 may contribute to myeloid leukemogenesis are deletions of the short arm of chromosome 7 associated with myeloproliferative-preceded secondary acute myeloid leukemia (AML) in adults, where the commonly deleted region is mapped to the IKZF1 locus (Jager et al,Leukemia 2010).
To investigate whether IKZF1 deletions play a role in pediatric AML we screened a representative well-characterized panel of 258 de novo pediatric AML samples with available gene expression data, obtained from the DCOG (The Hague, the Netherlands), the AML–Berliner-Frankfurt-Münster Study Group (Germany and Czech Republic), the Saint-Louis Hospital (Paris, France) and the Royal Hospital for Sick Children (Glasgow, UK) for deletions of the IKZF1 locus on chromosome 7p12.2 using multiplex ligation-dependent probe amplification (MLPA).
Median age of the patients was 9.5 years (range 0.1-18.5 years), median white blood cell count was 46.7x109/L (range 1.2-483x109/L). All major cytogenetic subgroups were included and all patients had received intensive cytarabine-anthracycline based pediatric AML therapy. Of 11 patients with an IKZF1 deletion, 8 cases presented with monosomy 7, and 3 cases showed a focal deletion of IKZF1. These focal deletions included the complete IKZF1 gene (n=2) or exons 1-4 (n=1), leading to a loss of IKZF1 function. Focal deletions were confirmed by high-resolution array comparative genome hybridization (array-CGH). Patients with a focal deletion included a 1.5 year old boy diagnosed with AML with a fusion of MNX1/ETV6 who died from relapse; an 11.3 year old girl diagnosed with AML M5 who remains in continuous complete remission (CCR) after salvage therapy for relapse; and a 2.3 year old boy diagnosed with AML M5 in CCR. IKZF1 deleted cases (n=11) did not differ significantly from the other pediatric AML cases (n=247) with regards to age at diagnosis (median age 9.1 years compared to 9.5 years respectively, p=0.41); gender (females 55% versus 42%, p=0.41); or white blood cell count at diagnosis (median 30.2 x109/L versus 47.5 x109/L, p=0.24). No specific FAB morphology subtypes were related to IKZF1 deletions. IKZF1 deleted samples showed either none or various different additional somatic mutations, most frequently activating the RAS pathway with mutations in NRAS or PTPN11 (n=4,). In BCP-ALL IKZF1 deletions are associated with BCR/ABL1 fusions and to test if this was also true for IKZF1 deletions in AML we screened samples for the BCR/ABL1 fusion and all were negative. The 3-year pOS in IKZF1 deleted patients (n=11) was 70±14% versus 63±3% (p=0.82) in IKZF1 wild-type patients (n=231). The 3-year pEFS was 36±15% versus 46±3%, and the 3-year pCIR was 64±16% versus 36±3% (p=0.87 and p=0.09) respectively. Genes differentially expressed in monosomy 7 cases correlated significantly with gene expression changes in focal IKZF1 deleted cases when comparing significant differences to non-IKZF1-deleted cases (n=247). Genes showing increased expression in IKZF1 deleted samples included those involved in myeloid cell cycling and self-renewal, such as HEMGN, FHL2, FZD6, and SETBP1.
In summary, we found focal IKZF1 deletions to be rare but recurrent events in pediatric AML. Gene expression patterns suggest that the loss of IKZF1 may be an important determinant in the biology of pediatric AML with monosomy 7.
No relevant conflicts of interest to declare.
This paper presents the R/Bioconductor package stepwiseCM, which classifies cancer samples using two heterogeneous data sets in an efficient way. The algorithm is able to capture the distinct ...classification power of two given data types without actually combining them. This package suits for classification problems where two different types of data sets on the same samples are available. One of these data types has measurements on all samples and the other one has measurements on some samples. One is easy to collect and/or relatively cheap (eg, clinical covariates) compared to the latter (high-dimensional data, eg, gene expression). One additional application for which stepwiseCM is proven to be useful as well is the combination of two high-dimensional data types, eg, DNA copy number and mRNA expression. The package includes functions to project the neighborhood information in one data space to the other to determine a potential group of samples that are likely to benefit most by measuring the second type of covariates. The two heterogeneous data spaces are connected by indirect mapping. The crucial difference between the stepwise classification strategy implemented in this package and the existing packages is that our approach aims to be cost-efficient by avoiding measuring additional covariates, which might be expensive or patient-unfriendly, for a potentially large subgroup of individuals. Moreover, in diagnosis for these individuals test, results would be quickly available, which may lead to reduced waiting times and hence lower the patients’ distress. The improvement described remedies the key limitations of existing packages, and facilitates the use of the stepwiseCM package in diverse applications.