Abstract
Purpose: RUNX1-mutated (RUNX1mut) acute myeloid leukemia (AML) is associated with adverse outcome, highlighting the urgent need for a better genetic characterization of this AML subgroup and for the design of efficient therapeutic strategies for this disease. Toward this goal, we further dissected the mutational spectrum and gene expression profile of RUNX1mut AML and correlated these results to drug sensitivity to identify novel compounds targeting this AML subgroup.
Experimental Design: RNA-sequencing of 47 RUNX1mut primary AML specimens was performed and sequencing results were compared to those of RUNX1 wild-type samples. Chemical screens were also conducted using RUNX1mut specimens to identify compounds selectively affecting the viability of RUNX1mut AML.
Results: We show that samples with no remaining RUNX1 wild-type allele are clinically and genetically distinct and display a more homogeneous gene expression profile. Chemical screening revealed that most RUNX1mut specimens are sensitive to glucocorticoids (GCs) and we confirmed that GCs inhibit AML cell proliferation through their interaction with the glucocorticoid receptor (GR). We observed that specimens harboring RUNX1 mutations expected to result in low residual RUNX1 activity are most sensitive to GCs, and that coassociating mutations as well as GR levels contribute to GC sensitivity. Accordingly, acquired glucocorticoid sensitivity was achieved by negatively regulating RUNX1 expression in human AML cells.
Conclusions: Our findings show the profound impact of RUNX1 allele dosage on gene expression profile and glucocorticoid sensitivity in AML, thereby opening opportunities for preclinical testing which may lead to drug repurposing and improved disease characterization. Clin Cancer Res; 23(22); 6969–81. ©2017 AACR.
This study characterized the effect of RUNX1 allele dosage on the gene expression profile and glucocorticoid sensitivity of primary RUNX1mut AML specimens. Our findings suggest a new role for RUNX1 in the glucocorticoid response and support the rationale to evaluate the addition of glucocorticoids in preclinical models of RUNX1mut AML.
Introduction
RUNX1 is a master regulator of definitive hematopoiesis where it regulates the differentiation of myeloid, megakaryocytic, and lymphocytic lineage progenitors (1, 2). RUNX1 is part of the core binding factor (CBF) transcriptional complex, and its transcriptional activity is dependent on the recruitment of its heterodimeric partner, CBFB. RUNX1 contains a RUNT domain at its N-terminus that is responsible for both DNA binding and protein heterodimerization (3). The C-terminal region of the protein encompasses domains for nuclear localization and regulation of DNA binding (4), as well as for the interaction with lineage-specific transcription factors, transcriptional coactivators, and corepressors.
Anomalies involving the RUNX1 gene or its partner CBFB have been implicated in the pathogenesis of subsets of human myeloid and lymphoblastic leukemias (5). The RUNX1 and CBFB genes are involved in the t(8;21)(q22;q22) and inv(16)(p13.1q22) chromosomal rearrangements, respectively, and these entities constitute the CBF acute myeloid leukemia (AML) subgroup (6, 7). Prognosis is favorable for patients carrying these cytogenetic anomalies when compared with other AML subtypes (8). In addition to chromosomal rearrangements, mutations in the RUNX1 gene are also found in myelodysplastic syndrome (MDS) and in 10%–21% of AMLs where they are associated with French-American-British (FAB) M0 morphology (9–11). In contrast to CBF AMLs, RUNX1mut AMLs are associated with adverse outcome (11–17). In a large proportion of cases, RUNX1mut AMLs harbor normal karyotype or noncomplex chromosomal imbalances, with a frequent association with trisomy 13 (12–15). RUNX1 mutations are also generally mutually exclusive of recurrent translocations in AML, and mutational analyses using targeted approaches revealed that RUNX1 mutations cooccur with mutations in epigenetic modifiers, such as ASXL1, splicing factors, STAG2, BCOR, and PHF6 (14, 17). Microarray analysis has been used to derive a RUNX1 mutation–associated gene expression signature (17), however, a complete assessment of the mutational and gene expression landscape of RUNX1mut AML is lacking.
Two types of mutations in RUNX1 have been described in AML: missense mutations found in the RUNT domain and nonsense or frameshift mutations distributed throughout the entire gene. Some frameshift mutations located in the C-terminal region produce elongated versions of the protein with intact DNA-binding activity that retain the ability to heterodimerize with CBFB, and that are believed to act as dominant negatives (10, 12, 15, 18). Approximately 30% of RUNX1 mutations occur in combination (i.e., double heterozygosity) or are associated with a loss of heterozygosity, both of which lead to a complete loss of wild-type RUNX1 in these leukemias (15). In line with this, it has been proposed that the greater the extent of RUNX1 inactivation in hematopoietic cells, the higher the propensity to develop leukemia, suggesting a dependence on RUNX1 protein dosage for disease onset (19, 20).
The unique genetic, biological, and clinical features of de novo AMLs with mutated RUNX1 prompted its suggestion as a distinct entity in the 2016 revision of the World Health Organization (WHO) classification of myeloid neoplasms (21). The poor outcome of patients suffering from RUNX1mut leukemias highlights the need to better understand the genetics of this disease and to develop more specific and efficient therapeutic strategies. In this study, we further dissected the mutational spectrum and gene expression profile of RUNX1mut AML and correlated these results to drug sensitivity. This was accomplished by RNA sequencing of the 47 RUNX1mut specimens included in the Leucegene cohort and by testing the sensitivity of these specimens to a collection of small molecules. This effort represents the first chemogenomic assessment of RUNX1mut AML.
Materials and Methods
Primary AML specimens
The Leucegene project is an initiative approved by the Research Ethics Boards of Université de Montréal and Maisonneuve-Rosemont Hospital. As part of this project, RNA sequencing of 415 primary AML specimens from various cytogenetic groups was performed as described previously (22). All leukemia samples were collected and characterized by the Quebec Leukemia Cell Bank (BCLQ). RNA sequencing data are available at GEO or through the Leucegene web page leucegene.ca/research/resources/.
Next-generation sequencing and mutation validations
Sequencing was performed as described previously (22). Sequenced data were mapped to the reference genome hg19 according to RefSeq annotations (UCSC, April 16, 2014). Variants were all identified using CASAVA 1.8.2 or km (https://bitbucket.org/iric-soft/km) approaches according to the previously reported pipeline (23). All variants present in 97 genes mutated in myeloid cancers or in acute leukemias were investigated (Supplementary Table S1). Genes and positions from Supplementary Table S5 were also investigated by km approach described previously, using a 5% variant allele frequency (VAF) cutoff for missense and nonsense mutations as well as for indels confirmed by another approach, of 10% for other indels. RUNX1 longest isoform (NM_001754/NP_001745.2) was used for representations.
Primary AML cell culture and chemical screens
Freshly thawed primary AML specimens were used for chemical screens. Cryopreserved cells were thawed at 37°C in Iscove's modified Dulbecco's medium (IMDM) containing 20% FBS and DNase I (100 μg/mL). Cells were resuspended in IMDM supplemented with 15% BIT (BSA, insulin, transferrin; StemCell Technologies), 100 ng/mL stem cell factor (Shenandoah 100-04), 50 ng/mL FLT3L (Shenandoah), 20 ng/mL IL3 (Shenandoah), 20 ng/mL G-CSF (Shenandoah), 10−4 mol/L β-mercaptoethanol, gentamicin (50 μg/mL), ciprofloxacin (10 μg/mL), SR1 (500 nmol/L, Alichem), and UM729 (500 nmol/L, IRIC) and 5,000 cells were plated per well of 384-well white plates in 50 μL. Compounds were dissolved in DMSO and diluted in media immediately before use. Compounds were added to plated cells by Biomek automatic pipettor at a final DMSO concentration of 0.1%. Glucocorticoids were tested in the exploratory screen at single doses of 2.5 μmol/L as described previously (24). In the confirmatory screen, compounds were added in serial dilutions (ranging from 10,000 nmol/L to 4.5 nmol/L). Cells were grown in culture in the presence of compounds for 6 days before determining cell viability using luminescent CellTiter Glo assay (Promega). Absolute IC50 values were calculated using ActivityBase SARview Suite. For cases where compounds failed to inhibit AML cell survival/proliferation, IC50 values were reported as the highest concentration tested (10,000 nmol/L). Compounds that showed more than 50% inhibition at the lowest concentration tested had IC50 assigned as the lowest dose tested (4.5 nmol/L). Dose–response curves were generated using GraphPad Prism 5.0. Heatmap representations of IC50 values were created using GENE-E software (http://www.broadinstitute.org/cancer/software/GENE-E).
AML cell lines and chemical screen
AML cell lines were purchased from the DSMZ German collection of Microorganisms and cell culture (Leibniz Institute), the ATCC, the University Health Network, or otherwise donated from collaborators. Cell lines were obtained from January to October 2015, and no authentication test was done by the authors. Cells were cultured according to manufacturer's or collaborator's instructions. The chemical screen performed with AML cell lines was carried out as described for primary AML cells with a few modifications. Compounds were added at concentrations ranging from 20,000 nmol/L to 1 nmol/L. Cell viability was evaluated after 7 days of culture in the presence of compounds using the CellTiter Glo assay (Promega).
Knockdown experiments
Lentiviral vectors carrying shRNAs targeting the RUNX1 and NR3C1 genes were generated by cloning appropriate shRNA sequences as described in ref. 25 into MNDU vectors comprising miR-E sequences as well as GFP or YFP. Control vector (shNT) contained shRNA-targeting Renilla luciferase. Lentiviruses were produced in HEK-293 cells and AML cell lines were infected with lentiviruses in media supplemented with 10 ng/mL polybrene for 48 hours. Infection efficiency, as determined by the percentage of GFP- or YFP-positive cells, was monitored by flow cytometry using a BD FACSCantoII flow cytometer. Infected cells were sorted using a BD Aria II cell sorter and knockdown efficiency was determined by quantitative RT-PCR and Western blotting using standard methods.
Immunofluorescence
AML cell lines treated or not with dexamethasone were applied to 0.01% poly-l-lysine–coated (Sigma) iBIDI chambers and were fixed with 4% paraformaldehyde (PFA) in PBS for 10 minutes at room temperature. Cells were incubated with permeabilization/blocking solution (0.25% Triton-X, 1% BSA in PBS) for 40 minutes and incubated with primary antibodies against GR (1:50; Cell Signaling Technology, 12041) and CD44-FITC (1:400; eBioscience, 11-0441-85) for one hour at room temperature. GR signal was revealed using Cy3-conjugated anti-rabbit secondary antibody (1:2,000) which was incubated with cells for 40 minutes at room temperature. Images were acquired on a Zeiss LSM 700 confocal laser scanning microscope.
Statistical analysis
Mutations and transcriptome.
Statistical tests for mutation and gene expression analyses were performed using R version 3.2.3. Fisher exact test was used in the analysis of contingency tables. Analysis of continuous variables and differential gene expression was performed using the Wilcoxon rank-sum test. False discovery rate (FDR) method was applied for global gene analysis.
Chemical screens and functional studies.
Figure 3B shows the Spearman correlation between the inhibitory responses of compounds of the glucocorticoid (GC) cluster. Growth inhibition is presented as rank transformed percentage of inhibition [100 − (100 × (number of cells (compound)/mean number of cells (DMSO controls)]. In Fig. 5B, Fisher exact test was used to test the association of the type of RUNX1 mutation and GC sensitivity in GraphPad Prism 5.0. In Fig. 5C, the difference in response to compounds between mutation groups was calculated on the basis of IC50 values using Wilcoxon rank-sum test in R. The highest dose tested (10,000 nmol/L) was arbitrarily assigned to samples when the compounds failed to inhibit 50% of cell proliferation. Similarly, the lowest dose tested (4.5 nmol/L) was reported in cases where the inhibitory response was higher than 50% at the lowest dose. Determination of differentially active compounds in mutation subgroups was performed using Wilcoxon rank sum test on IC50 values in R. In Fig. 6C, the differences in IC50 values between OCI-AML5 cells expressing shNT and OCI-AML5 cells expressing shRUNX1 were calculated using Wilcoxon rank-sum test in R. In Figs. 5C and 6B and D and Supplementary Fig. S8D and S8E, P values were calculated using two-tailed Student t test in GraphPad Prism 5.0.
Results
RNA sequencing data of 47 RUNX1mut primary AML specimens were compared with that of 368 control RUNX1 wild-type (RUNX1wt) samples, which were sequenced as part of the Leucegene project (ref. 22; Supplementary Table S1). RUNX1mut specimens were associated with older age, FAB M0 morphology, intermediate-risk cytogenetics with abnormal karyotype (Fig. 1A), and poor patient survival (Fig. 1B) compared with RUNX1wt specimens, in accordance with published characteristics (26, 14). Twenty-seven specimens (57%) carried a nonsense or frameshift mutation (RUNX1ns/fs), whereas 20 specimens (43%) were characterized by missense mutations only (RUNX1mis; Supplementary Table S2). As previously reported, most missense mutations were located in the RUNT domain, whereas nonsense/frameshift mutations were more widely distributed (Fig. 1C; ref. 18). No significant differences in clinical and laboratory characteristics were observed between RUNX1ns/fs and RUNX1mis AML samples (data not shown). Fourteen samples (30%) were characterized by either homozygous RUNX1 mutations, defined by a VAF greater than 75%, or by double heterozygous mutations (RUNX1−/− in Fig. 1D, on the left), suggesting that an important percentage of RUNX1mut cases have very little remaining RUNX1 activity.
Mutational landscape of RUNX1mut primary AML specimens. A, Characteristics of RUNX1mut and RUNX1wt cohorts. P values are based on two-tailed Fisher exact test or Wilcoxon rank-sum test. B, Patient survival according to RUNX1 mutation status. P value was calculated using the log-rank test. C, Primary structure and position of mutations on RUNX1 protein (NP_001745.2). RUNT: 85–206, TAD: 318–398, RUNXI: 389–480. D, Mutational profile of RUNX1mut primary AML specimens. Samples are grouped according to their RUNX1 mutation type, with samples carrying mutations at VAF ≥ 75% and double heterozygous mutations on the left (RUNX1−/−) and RUNX1 heterozygous mutations on the right (RUNX1−/+). Each column represents a patient sample. Cytogenetic and other clinical information is provided in the last rows, whereas mutation frequency within RUNX1mut cohort and enrichment between indicated comparison groups are shown in left and right panels, respectively. Enrichments were calculated using Fisher exact test. WBC, white blood cell; T-related, therapy-related; myelodysplasia, myelodysplasia-related changes; NK, normal karyotype; Inter, intermediate; FAB, French-American-British; TAD, transcriptional activation domain; RunxI, Runx1 inhibition domain; ITD, internal tandem duplications; CEBPAbi, biallelic CEBPA mutations.
Mutational landscape of RUNX1mut primary AML specimens. A, Characteristics of RUNX1mut and RUNX1wt cohorts. P values are based on two-tailed Fisher exact test or Wilcoxon rank-sum test. B, Patient survival according to RUNX1 mutation status. P value was calculated using the log-rank test. C, Primary structure and position of mutations on RUNX1 protein (NP_001745.2). RUNT: 85–206, TAD: 318–398, RUNXI: 389–480. D, Mutational profile of RUNX1mut primary AML specimens. Samples are grouped according to their RUNX1 mutation type, with samples carrying mutations at VAF ≥ 75% and double heterozygous mutations on the left (RUNX1−/−) and RUNX1 heterozygous mutations on the right (RUNX1−/+). Each column represents a patient sample. Cytogenetic and other clinical information is provided in the last rows, whereas mutation frequency within RUNX1mut cohort and enrichment between indicated comparison groups are shown in left and right panels, respectively. Enrichments were calculated using Fisher exact test. WBC, white blood cell; T-related, therapy-related; myelodysplasia, myelodysplasia-related changes; NK, normal karyotype; Inter, intermediate; FAB, French-American-British; TAD, transcriptional activation domain; RunxI, Runx1 inhibition domain; ITD, internal tandem duplications; CEBPAbi, biallelic CEBPA mutations.
RUNX1mut AMLs are genetically distinct
Specimens of our RUNX1mut cohort harbored mutations in 39 different genes (Fig. 1D; Supplementary Table S3). These included: ASXL1 (18/47, 38%), SRSF2 (13/47, 28%), TET2 (10/47, 21%), FLT3 (9/47, 19%), BCOR and NRAS (8/47 each, 17%), DNMT3A (7/47, 15%), KMT2A/MLL and STAG2 (6/47 each, 13%), CEBPA, EZH2, IDH1 and IDH2 (5/47 each, 11%), JAK2, TP53 and U2AF1 (4/47 each, 9%) as well as KRAS and NF1 (3/47 each, 6%; Fig. 1D; Supplementary Table S3). Statistical analysis revealed that ASXL1, SRSF2, BCOR, EZH2, JAK2, STAG2, and PHF6 mutations significantly associated with RUNX1 mutations, whereas an antiassociation was found between RUNX1 and NPM1 or FLT3 mutations (Fig. 1D). Mutations in components of the spliceosome such as SRSF2 and SF3B1 have been previously reported in RUNX1mut AML (27). When comparing RUNX1ns/fs and RUNX1mis specimens, we observed an association between SRSF2 and RUNX1ns/fs mutations (12/27 for RUNX1ns/fs vs. 1/20 for RUNX1mis, P = 0.003, Fig. 1D). This association between splicing genes and RUNX1 mutations remained highly significant (P = 0.009) when all splicing genes (SRSF2, U2AF1, SF3B1) were considered, suggesting a possible link between loss of RUNX1 function and altered splicing activity in AML. Moreover, FAB M0 morphology and trisomy 13, which are two characteristic features of RUNX1mut AMLs, were significantly associated with samples with no remaining RUNX1 wild-type allele (RUNX1−/−), compared with samples carrying RUNX1 heterozygous mutations (RUNX1−/+; 36% vs. 9% for trisomy 13 and 57% vs. 9% for M0, Fig. 1D, right). Furthermore, RUNX1−/− samples harboring either nonsense or frameshift mutations were also significantly associated with ASXL1 mutations (86% vs. 30%, Fig. 1D, right). On the other end of the spectrum, RUNX1−/+ AML samples, especially those with heterozygous missense mutations, are enriched for mutations in EZH2 and in CEBPA. Altogether, these data suggest that RUNX1mut samples lacking a wild-type RUNX1 allele display distinct genetic features (e.g., trisomy 13 and M0 in RUNX1−/− vs. EZH2 and CEBPA mutations in RUNX1−/+).
RUNX1 allele dosage determines gene expression signature
Comparative transcriptomic analysis of RUNX1mut and RUNX1wt specimens revealed a list of 100 differentially expressed candidate genes (Fig. 2A; Supplementary Table S4). Projecting RUNX1mut specimens on a principal component analysis (PCA) representation constructed with these 100 genes predictably identified most RUNX1mut specimens, but lacked the specificity previously reported for other AML subgroups (28, 29). Indeed, several RUNX1mut specimens were in the vicinity of control RUNX1wt AML specimens (Fig. 2B). Interestingly, we observed a correlation between RUNX1 allele dosage and the expression levels of the most specific transcripts of the RUNX1mut signature such as BAALC and DNTT (Fig. 2A, C; Supplementary Fig. S1). These two genes are expressed at much higher levels in specimens homozygous for nonsense/frameshift RUNX1 mutations than in those with heterozygous missense mutations (Fig. 2C). This highlights the importance of considering the degree of wild-type RUNX1 loss to accurately reveal the gene expression profile of this AML subgroup.
RUNX1 allele dosage determines RUNX1 mutation–associated gene expression signature. A, Differentially expressed genes in RUNX1mut specimens as revealed by RNA sequencing. The 100 most differentially expressed genes are indicated by diamonds. Scale: average [log10((RPKM + 0.0001) × 10,000)]. Genes with a value < 3 in both groups, corresponding to approximately 0.1 RPKM, were not included in this analysis. RPKM of 1 is equivalent to approximately 4 on the scale. B, PCA performed using sequencing data from 415 primary AML specimens using the RUNX1 100-gene signature. C, Expression levels of the BAALC and DNTT genes according to mutation type (ns/fs = nonsense/frameshift; mis = missense) and load (VAF) using scale defined in A. Double heterozygous cases with missense and nonsense/frameshift mutations were classified as nonsense/frameshift. VAF ≥ 75% for homozygous mutation, or sum of VAF ≥ 75% for double heterozygous mutations was used to label a sample RUNX1−/−. All RUNX1mut subgroup comparisons to RUNX1wt specimens were significant, except RUNX1mis versus RUNX1wt for DNTT. P values were calculated using Wilcoxon test. D, Correlation of the 100 most differentially expressed genes identified in A according to model based on mutational pattern identified in C. Correlations were performed in the RUNX1mut cohort only (n = 47), using the following template: 1 (missense−/+), 2 (nonsense/frameshift−/+), 3 (missense−/−), 4 (nonsense/frameshift−/−). Genes are ordered according to their correlation to this template. Genes with maximum expression levels in RUNX1mut specimens < 1 RPKM were not included in this analysis. A selection of the most correlated genes known to be related to leukemia is labeled, as well as the Glucocorticoid receptor gene, NR3C1.
RUNX1 allele dosage determines RUNX1 mutation–associated gene expression signature. A, Differentially expressed genes in RUNX1mut specimens as revealed by RNA sequencing. The 100 most differentially expressed genes are indicated by diamonds. Scale: average [log10((RPKM + 0.0001) × 10,000)]. Genes with a value < 3 in both groups, corresponding to approximately 0.1 RPKM, were not included in this analysis. RPKM of 1 is equivalent to approximately 4 on the scale. B, PCA performed using sequencing data from 415 primary AML specimens using the RUNX1 100-gene signature. C, Expression levels of the BAALC and DNTT genes according to mutation type (ns/fs = nonsense/frameshift; mis = missense) and load (VAF) using scale defined in A. Double heterozygous cases with missense and nonsense/frameshift mutations were classified as nonsense/frameshift. VAF ≥ 75% for homozygous mutation, or sum of VAF ≥ 75% for double heterozygous mutations was used to label a sample RUNX1−/−. All RUNX1mut subgroup comparisons to RUNX1wt specimens were significant, except RUNX1mis versus RUNX1wt for DNTT. P values were calculated using Wilcoxon test. D, Correlation of the 100 most differentially expressed genes identified in A according to model based on mutational pattern identified in C. Correlations were performed in the RUNX1mut cohort only (n = 47), using the following template: 1 (missense−/+), 2 (nonsense/frameshift−/+), 3 (missense−/−), 4 (nonsense/frameshift−/−). Genes are ordered according to their correlation to this template. Genes with maximum expression levels in RUNX1mut specimens < 1 RPKM were not included in this analysis. A selection of the most correlated genes known to be related to leukemia is labeled, as well as the Glucocorticoid receptor gene, NR3C1.
Following the hypothesis that mutation type and allelic burden are determinant in RUNX1mut AML, we derived a model which identified transcripts whose expression is determined by RUNX1 allele dosage (Fig. 2D; Supplementary Table S6). Consistent with the ability of RUNX1 to regulate its own expression (30), the RUNX1 transcript was among the most positively correlated ones (Fig. 2D). Interestingly, TCF4 was the second top candidate (Fig. 2D). TCF4 recognizes the CANNTG-binding site, an E-box found to be enriched in the promoter of genes identified by our model (q = 1.17 × 10−4),suggesting that it may play a role in establishing the RUNX1mut signature as recently suggested for blastic plasmacytoid dendritic cell neoplasms (BPDCN; ref. 31).
RUNX1 mutations are associated with glucocorticoid sensitivity
To identify small molecules selectively affecting the survival of RUNX1mut AML cells, we screened a panel of primary AML specimens, which included RUNX1mut and RUNX1wt samples, selected to represent the genetic heterogeneity of the disease, with a library of compounds enriched for clinically approved drugs. This strategy was rendered possible by our advanced cell culture conditions which transiently support the ex vivo activity of leukemia progenitor/stem cells (32). This initial screen revealed groups of compounds exhibiting similar patterns of inhibition across specimens, which we named "compound correlation clusters" (CCC; ref. 24). One such cluster identified in the screen was the GC cluster, which comprises 34 compounds sharing structural similarities (Fig. 3A; Supplementary Fig. S2). Interestingly, primary AML specimens found to be sensitive to GCs in the screen were enriched with samples harboring RUNX1 mutations (4/6 GC-sensitive specimens were mutated for RUNX1, P = 0.003, and Fig. 3B). To further explore the link between GC sensitivity and RUNX1 status, we interrogated a cohort of RUNX1mut specimens and of specimens carrying a t(8;21) translocation resulting in the RUNX1–RUNX1T1 fusion with compounds of the glucocorticoid cluster (Fig. 3C). We observed that RUNX1mut specimens and those presenting RUNX1–RUNX1T1 fusions were more frequently sensitive to GCs than RUNX1wt samples (Fig. 3C). Primary AML specimens responded similarly to all compounds of the GC cluster, indicating that these compounds most likely share a common target and operate via the same mechanism (Fig. 3B and C). Notably, an impressive difference in GC IC50 was observed between sensitive and resistant specimens, ranging from single digit nmol/L for sensitive samples to >10,000 nmol/L for resistant ones, with few cases of intermediate sensitivity (Fig. 3D; Supplementary Fig. S3). These results suggest that RUNX1 loss of function confers sensitivity to GCs in AML.
RUNX1mut primary AML specimens are sensitive to glucocorticoid treatment. A, Response of 20 primary AML specimens (1 AML per vertical line on x-axis) to different compounds of the glucocorticoid cluster (each compound represented by 1 line in graph) as indicated by % inhibition of cell viability (y-axis). B, Correlation of inhibitory response of 20 primary AML specimens to three representatives of the glucocorticoid cluster with decreasing potency (flumethasone > dexamethasone > hydrocortisone). Black dots, RUNX1wt specimens; blue dots, RUNX1mut specimens. ρ, Spearman rank correlation coefficient. C, Heatmap showing IC50 values for validation screen carried out on 25 additional primary AML specimens and 30 GCs. Drugs were tested in 8 serial dilutions ranging from 4.5 to 10,000 nmol/L. D, Dose–response curves for dexamethasone and associated IC50 values for two representative RUNX1mut and two RUNX1wt specimens. Dex, dexamethasone.
RUNX1mut primary AML specimens are sensitive to glucocorticoid treatment. A, Response of 20 primary AML specimens (1 AML per vertical line on x-axis) to different compounds of the glucocorticoid cluster (each compound represented by 1 line in graph) as indicated by % inhibition of cell viability (y-axis). B, Correlation of inhibitory response of 20 primary AML specimens to three representatives of the glucocorticoid cluster with decreasing potency (flumethasone > dexamethasone > hydrocortisone). Black dots, RUNX1wt specimens; blue dots, RUNX1mut specimens. ρ, Spearman rank correlation coefficient. C, Heatmap showing IC50 values for validation screen carried out on 25 additional primary AML specimens and 30 GCs. Drugs were tested in 8 serial dilutions ranging from 4.5 to 10,000 nmol/L. D, Dose–response curves for dexamethasone and associated IC50 values for two representative RUNX1mut and two RUNX1wt specimens. Dex, dexamethasone.
Glucocorticoid receptor mediates the GC response in AML
GCs are glucocorticoid receptor (GR) agonists that cause its translocation into the nucleus to modulate transcription of specific genes (33). The possibility that GR is the target through which GCs affect AML cell behavior was first evaluated using a panel of AML-derived cell lines (Supplementary Fig. S4). The ability of GCs to induce translocation of the GR from the cytoplasm to the nucleus was intact in the different AML cell lines tested (Fig. 4A and data not shown for other cell lines due to space limitation). We observed that treatment of GC-sensitive Kasumi-1 and OCI-AML3 cell lines with increasing concentrations of the GR antagonist RU486 progressively decreases their GC sensitivity (Fig. 4B for dexamethasone and Supplementary Fig. S5 for mometasone furoate), demonstrating that GR inhibition prevents GCs from exerting their effect on cell viability. To validate this hypothesis, we designed shRNAs targeting the GR gene, NR3C1, producing 75%–90% gene knockdown and a corresponding decrease in GR protein levels in OCI-AML3 cells (Fig. 4C), and observed that GR knockdown abrogates the antiproliferative effect of GCs, and enables proliferation of shNR3C1-expressing cells in GC-supplemented media (Fig. 4D for dexamethasone and Supplementary Fig. S6 for other GCs). Altogether, these results suggest that GCs inhibit AML cell proliferation through their interaction with the GR.
Inhibitory response to GCs in AML cells is dependent on glucocorticoid receptor activity. A, Evaluation of subcellular localization of the GR following dexamethasone treatment (Dex; 1 μmol/L for 1 hour) by immunofluorescence in HL-60 cells. Representative of three independent experiments. B, Dose–response curves and associated IC50 values for dexamethasone in Kasumi-1 and OCI-AML3 cells with increasing concentrations of GR full antagonist, RU486. Data is shown as mean ± SEM for three independent experiments. C, Assessment of GR knockdown efficiency by shRNAs targeting NR3C1 (sh1–sh4) in shRNA expressing OCI-AML3 cells (GFP+) by qRT-PCR (middle, results are shown for two independent infections and expressed as mean ± SEM) and Western blotting (right, representative of two independent infections). D, Determination of the impact of GR knockdown on inhibition of OCI-AML3 cell proliferation by dexamethasone. GFP+ cells were sorted, mixed with uninfected cells, and treated with a single dose of dexamethasone (IC75 = 200 nmol/L). Cell proliferation was evaluated at the indicated times after treatment by cell counting of the live GFP+ and GFP− populations, and expressed relative to DMSO controls. Line graph shows mean ± SEM of three independent experiments. Histograms are representatives of three replicates after 7 days of exposure to dexamethasone.
Inhibitory response to GCs in AML cells is dependent on glucocorticoid receptor activity. A, Evaluation of subcellular localization of the GR following dexamethasone treatment (Dex; 1 μmol/L for 1 hour) by immunofluorescence in HL-60 cells. Representative of three independent experiments. B, Dose–response curves and associated IC50 values for dexamethasone in Kasumi-1 and OCI-AML3 cells with increasing concentrations of GR full antagonist, RU486. Data is shown as mean ± SEM for three independent experiments. C, Assessment of GR knockdown efficiency by shRNAs targeting NR3C1 (sh1–sh4) in shRNA expressing OCI-AML3 cells (GFP+) by qRT-PCR (middle, results are shown for two independent infections and expressed as mean ± SEM) and Western blotting (right, representative of two independent infections). D, Determination of the impact of GR knockdown on inhibition of OCI-AML3 cell proliferation by dexamethasone. GFP+ cells were sorted, mixed with uninfected cells, and treated with a single dose of dexamethasone (IC75 = 200 nmol/L). Cell proliferation was evaluated at the indicated times after treatment by cell counting of the live GFP+ and GFP− populations, and expressed relative to DMSO controls. Line graph shows mean ± SEM of three independent experiments. Histograms are representatives of three replicates after 7 days of exposure to dexamethasone.
RUNX1 allele dosage and coassociated mutations contribute to GC sensitivity
About one third of RUNX1mut specimens did not respond to GC treatment (Fig. 3C). To gain further insight into the impact of RUNX1 mutations on GC response, we determined the IC50 values of at least one GC (dexamethasone, hydrocortisone, and/or flumethasone) in an enlarged cohort of 33 RUNX1mut primary AML samples. We observed that RUNX1fs/ns specimens showed increased sensitivity to GCs when compared with RUNX1mis specimens (Fig. 5A), suggesting a strong impact of RUNX1 allele dosage on GC sensitivity. In support of this, we observed that missense mutations reported to have no impact on RUNX1 function were enriched among GC-resistant specimens, whereas frameshift mutations predicted to produce elongated versions of RUNX1 with dominant-negative activity (10, 12, 18) were more frequent in the GC-sensitive group [P = 0.03, compare distribution of RUNX1mis (blue triangles) to that of dominant negatives (red diamonds) in Fig. 5B]. Nonetheless, 10%–18% of RUNX1mis specimens were sensitive to GCs, and similarly, 14%–18% of RUNX1ns/fs were resistant (Fig. 5A). To identify genetic lesions that may contribute to modulation of GC responsiveness, we analyzed GC sensitivity of various defined genetic groups of AML. We found that CEBPAbi and SRSF2-mutated specimens were significantly more sensitive to GCs than other leukemias (Fig. 5C), and GC-sensitive RUNX1mut specimens were enriched for these two mutations (Fig. 5B).
RUNX1 allele dosage dictates GC response in RUNX1mut AML. A, IC50 values of 28 RUNX1mut specimens in response to dexamethasone (left) and of 30 RUNX1mut specimens in response to flumethasone (right). Double-heterozygous samples were labeled as missense when both alleles had missense mutations or as frame-shift/nonsense when at least one mutated allele had a frameshift or nonsense mutation. B, Heatmap showing IC50 values for dexamethasone (Dex), flumethasone (Flu), and hydrocortisone (HC) for 33 RUNX1mut specimens. The effect of mutations on RUNX1 function was determined on the basis of previously published functional studies. C, Volcano plot showing integrative analysis of chemical screens using primary AML specimens from various genetic groups and GCs (flumethasone, dexamethasone, and hydrocortisone). Specimens from genetic groups significantly more sensitive to at least two GCs than wild-type specimens are represented as filled symbols.
RUNX1 allele dosage dictates GC response in RUNX1mut AML. A, IC50 values of 28 RUNX1mut specimens in response to dexamethasone (left) and of 30 RUNX1mut specimens in response to flumethasone (right). Double-heterozygous samples were labeled as missense when both alleles had missense mutations or as frame-shift/nonsense when at least one mutated allele had a frameshift or nonsense mutation. B, Heatmap showing IC50 values for dexamethasone (Dex), flumethasone (Flu), and hydrocortisone (HC) for 33 RUNX1mut specimens. The effect of mutations on RUNX1 function was determined on the basis of previously published functional studies. C, Volcano plot showing integrative analysis of chemical screens using primary AML specimens from various genetic groups and GCs (flumethasone, dexamethasone, and hydrocortisone). Specimens from genetic groups significantly more sensitive to at least two GCs than wild-type specimens are represented as filled symbols.
Given that acute lymphoblastic leukemias (ALL) are known to respond to GCs, we examined the expression of lymphoid markers in specimens of the Leucegene collection. t(8;21) specimens exhibited elevated expression of these markers (Supplementary Fig. S7); however, RUNX1mut specimens did not, thereby suggesting that lymphoid lineage–associated genes do not contribute to the GC response of RUNX1mut specimens. Interestingly, we observed that samples expressing the highest levels of NR3C1 were enriched with specimens harboring SRSF2 and RUNX1 mutations (Supplementary Fig. S7), suggesting that GR levels influence GC sensitivity for these mutation groups. However, the resistance of RUNX1mut specimens to GCs could not be explained by altered expression levels of wild-type RUNX1, NR3C1, CEBPA, or SRSF2 (Supplementary Fig. S7). These observations suggest that inactivation of RUNX1 is associated with NR3C1 upregulation and sensitivity to GCs, and that interference with the function of the splicing machinery (SRSF2) or of other transcription factors (CEBPA), and possibly other yet to be identified processes, also appears to be involved in the GC response in AML cells.
RUNX1 silencing sensitizes AML cells to GCs
To validate our model predicting that RUNX1 dosage and the resulting RUNX1 protein availability modulate GC responsiveness, we evaluated how RUNX1 silencing affects survival of GC-resistant AML cell lines in GC-supplemented media. We designed shRNAs targeting RUNX1 for which gene knockdown and reduction of RUNX1 protein levels were validated in OCI-AML5 cells (Fig. 6A). We observed that RUNX1 silencing was able to sensitize 5 of the 8 AML cell lines tested to dexamethasone (Fig. 6B). A RUNX1 level-dependent shift in dexamethasone sensitivity as revealed by a decrease in IC50 values for shRUNX1-expressing cells compared with controls, reaching 2 digit nanomolar range in these engineered cells, was observed in OCI-AML5 (Fig. 6C) and OCI-AML1 cells (Supplementary Fig. S8). Interestingly, a concomitant increase in NR3C1 expression was noted in these cells, further strengthening the idea that GR levels impact on the GC response (Fig. 6D; Supplementary Fig. S8). In accordance with the increased GC sensitivity observed for shRUNX1-expressing cells, dexamethasone induced pronounced apoptosis in OCI-AML5 cells upon RUNX1 knockdown (Fig. 6E). The GC-sensitizing effect of RUNX1 dosage was also observed for other GCs, such as flumethasone and budesonide (Supplementary Fig. S8), and appeared to be specific to the GC response, as it was not observed with other cytotoxic agents such as cytarabine and 6-thioguanine (Fig. 6F; Supplementary Fig. S8). In summary, our data provide functional evidence that RUNX1 dosage influences GC sensitivity, suggesting a novel role for RUNX1 in the response to GCs in AML cells.
RUNX1 silencing in AML cells increases sensitivity to GCs. A, Assessment of RUNX1 knockdown efficiency by shRNAs (sh1–sh3) in YFP+ sorted OCI-AML5 cells by qRT-PCR (middle, results are shown for two independent infections and expressed as mean ± SEM) and Western blotting (right, representative of two independent infections). B, IC50 values for dexamethasone (Dex) in eight AML cell lines expressing shRUNX1 or shNT (control). Cells were treated with increasing concentrations of dexamethasone (1 to 20,000 nmol/L) and the effect on cell viability was monitored. Results from three independent experiments are shown with SEM. P values were determined by two-tailed unpaired t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001. C, Dose–response curves and associated IC50s for dexamethasone in OCI-AML5 cells expressing shRUNX1 or shNT. Results from three independent experiments are shown with SEM. D, NR3C1 transcript levels in OCI-AML5 cells expressing shRUNX1 or shNT. Results from three independent experiments are shown with SEM. P values were determined by two-tailed unpaired t test. **, P < 0.01; ***, P < 0.001. E, Proportion of apoptotic cells determined by Annexin-V staining 24 hours after treatment of OCI-AML5 with 100 nmol/L dexamethasone or DMSO as control. Results from three independent experiments are shown with SEM. P values were determined by two-tailed unpaired t test. Comparisons between DMSO and dexamethasone for each shRNA are represented by connecting line with ***. **, P < 0.01; ***, P < 0.001. F, IC50s for dexamethasone, cytarabine, and 6-thioguanine in OCI-AML5 cells expressing shRUNX1 or shNT. Cells obtained from two independent infections were used for drug treatment, values are expressed with SEM. P values were calculated using two-tailed unpaired t test. ***, P < 0.001, NS, not significant.
RUNX1 silencing in AML cells increases sensitivity to GCs. A, Assessment of RUNX1 knockdown efficiency by shRNAs (sh1–sh3) in YFP+ sorted OCI-AML5 cells by qRT-PCR (middle, results are shown for two independent infections and expressed as mean ± SEM) and Western blotting (right, representative of two independent infections). B, IC50 values for dexamethasone (Dex) in eight AML cell lines expressing shRUNX1 or shNT (control). Cells were treated with increasing concentrations of dexamethasone (1 to 20,000 nmol/L) and the effect on cell viability was monitored. Results from three independent experiments are shown with SEM. P values were determined by two-tailed unpaired t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001. C, Dose–response curves and associated IC50s for dexamethasone in OCI-AML5 cells expressing shRUNX1 or shNT. Results from three independent experiments are shown with SEM. D, NR3C1 transcript levels in OCI-AML5 cells expressing shRUNX1 or shNT. Results from three independent experiments are shown with SEM. P values were determined by two-tailed unpaired t test. **, P < 0.01; ***, P < 0.001. E, Proportion of apoptotic cells determined by Annexin-V staining 24 hours after treatment of OCI-AML5 with 100 nmol/L dexamethasone or DMSO as control. Results from three independent experiments are shown with SEM. P values were determined by two-tailed unpaired t test. Comparisons between DMSO and dexamethasone for each shRNA are represented by connecting line with ***. **, P < 0.01; ***, P < 0.001. F, IC50s for dexamethasone, cytarabine, and 6-thioguanine in OCI-AML5 cells expressing shRUNX1 or shNT. Cells obtained from two independent infections were used for drug treatment, values are expressed with SEM. P values were calculated using two-tailed unpaired t test. ***, P < 0.001, NS, not significant.
Discussion
In this study, we used a chemogenomic approach to characterize RUNX1mut AML. RNA sequencing confirmed most mutations in other genes previously reported for this subgroup. Interestingly, RUNX1 allele dosage appears to identify a clinically and genetically distinct subgroup of AML patients lacking a wild-type RUNX1 allele (RUNX1−/−), as demonstrated by their frequent association with M0 morphology and trisomy 13, as well as with ASXL1 mutations in nonsense or frameshift cases. Our observations complement a recent report associating RUNX1−/− samples with adverse clinical outcome (34). Altogether, these data suggest that RUNX1mut AML, which has been recently added as a provisional entity in the WHO classification, may be more heterogeneous than previously believed. This RUNX1 allele dosage effect also revealed the complexity of the RUNX1 AML gene signature and appears to predispose AML cells to GC sensitivity or resistance. To the best of our knowledge, this is the first study linking RUNX1 mutations to GC sensitivity in AML.
In our study, primary AML specimens carrying RUNX1 C-terminal mutations showed increased sensitivity to GCs when compared with samples harboring N-terminal mutations (C-terminal mutations: 7/8 in GC sensitive group vs. 10/29 for N-terminal mutations, P = 0.014). All insertions located in the C-terminal region are predicted to lead to elongated RUNX1 proteins, which are known to act as dominant-negative inhibitors of RUNX1 (10, 12, 18), and these mutations were enriched in the GC-sensitive group. Interestingly, RUNX1 C-terminal mutations have been shown to affect protein function and leukemogenesis differently than N-terminal mutations involving the RUNT domain (10, 18, 35). Among the GC-sensitive specimens in our cohort, one specimen presenting a MLL-PTD fusion was later found to carry a novel translocation involving RUNX1 and SON, which results in a chimeric transcript comprising a truncated version of RUNX1 RUNT domain expected to result in RUNX1 loss of function, in accordance with our model (Supplementary Fig. S2). We also show that primary AML specimens carrying a t(8;21) translocation involving the RUNX1 gene are more frequently sensitive to GCs than RUNX1wt specimens (Fig. 3C), further supporting our model. In line with these results, Corsello and colleagues identified GCs as modulators of the gene expression signature associated with the AML1-ETO fusion in Kasumi-1 cells (36). They demonstrated that ectopic expression of the fusion in U937 cells sensitizes them to GCs, and that GC treatment decreases AML1-ETO fusion protein levels in a proteasome-dependent manner, suggesting that a similar mechanism might be at play in t(8;21) primary AML specimens. We also show that AML cell lines are susceptible to the modulation of RUNX1 dosage as RUNX1 silencing dramatically increased the sensitivity of various AML cell lines to GCs. Similarly, residual RUNX1 activity in primary AML specimens carrying heterozygous mutations or hypomorphic alleles could contribute to the resistance of these specimens to GCs. Overall, these results further support our theory that RUNX1 dosage dictates GC response in AML.
The mechanism by which RUNX1 loss of function mediates GC sensitivity is not immediately clear. We observed that elevated NR3C1 expression identifies primary AML specimens carrying RUNX1 mutations (Supplementary Fig. S7), suggesting an involvement of GR levels in GC sensitivity of RUNX1mut specimens. This finding is in line with a recent study by Malani and colleagues (37), which showed that acquired cytarabine resistance and GC sensitivity in AML cells is associated with increased expression of NR3C1. NR3C1 levels could not account for differences in GC response within the RUNX1-mutated group, however (Supplementary Fig. S7), implying that other factors are involved. This is supported by the observation that additional mutations (SRSF2 and CEBPAbi) are associated with GC sensitivity. One hypothesis to explain the interplay between RUNX1 and the GR in the GC response could be that RUNX1 negatively modulates the transcription of the NR3C1 gene. In accordance with this, we identified NR3C1 as one of the transcripts whose expression is determined by RUNX1 allele dosage in primary AML specimens (Fig. 2D; Supplementary Fig. S7) and we showed that RUNX1 silencing results in the upregulation of NR3C1 in AML cell lines (Fig. 6C; Supplementary Fig. S8). Moreover, RUNX1 has been shown to modulate NR3C1 expression (38) and to exert a protective effect against GC-induced apoptosis in lymphoma cells (39). Evidence for RUNX1 occupancy in the promoter region of the NR3C1 gene in normal hematopoietic stem/progenitor cells and AML cells also suggests that RUNX1 can directly regulate the expression of the NR3C1 gene (40). Interestingly, in ALL, activation of the GR by synthetic GCs, such as dexamethasone, leads to an apoptotic response by negative regulation of BCL2 expression and upregulation of proapoptotic BIM genes (41). We show that OCI-AML5 cells undergo apoptosis following GC treatment; therefore, one can envision that a similar mechanism exists in AML. It therefore appears that elucidation of the precise mechanism by which RUNX1 loss of function confers GC sensitivity in AML will require further investigation.
Considering that RUNX1 sequencing is now included in the initial prognostic assessment of AML patients, our study supports the rationale to evaluate the addition of GC therapy for a subset of patients, for example, those with dominant negative or RUNX1−/− mutations. GCs are largely used in the treatment of ALL, and correlation of in vitro prednisolone sensitivity of primary pediatric ALL specimens to clinical characteristics revealed that low IC50 values for this compound in vitro are associated with good short-term response and long-term clinical outcome (42). Similarly to what is observed for prednisolone, clinical trials for ALL have shown that dexamethasone treatment improves relapse-free survival for these patients (43, 44). Interestingly, patients with ETV6-RUNX1–positive B-cell precursor ALL were among the best responders to GC treatment in this study (45). Results from these clinical trials and case reports support the idea that as for ALL, in vitro sensitivity to GCs might correlate with good GC treatment outcome for AML patients. In support of this, effective clinical doses of dexamethasone and prednisolone have been reported for ALL (44, 46) and circulating plasma levels as well as disposition for these GCs have been determined after intravenous or oral administration in several clinical trials (47–49). The IC50 values determined for inhibition of primary AML cells viability by GCs in our study systematically fall well below the dexamethasone and prednisolone plasma levels that can be inferred from the above, suggesting that clinically relevant GC concentrations could be achieved in patients for the treatment of RUNX1mut AML.
In conclusion, our data suggests that RUNX1mut AMLs have distinct genetic and transcriptomic features, possibly impacting on their sensitivity to drugs such as glucocorticoids. Our data also indicate that SRSF2 loss of function contributes to the GC response, suggesting that GC treatment could be beneficial for AML patients carrying both SRSF2 and RUNX1 mutations. Of interest, such patients were recently reported to have particularly poor outcome (14). Collectively, our results reveal a potentially easy clinical intervention that may rapidly impact the outcome of patients suffering from RUNX1mut AML. Adequately designed preclinical studies will help determine the nature of these interventions.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: L. Simon, V.-P. Lavallée, M.-E. Bordeleau, J. Hébert, G. Sauvageau
Development of methodology: L. Simon, B. Lehnertz, T. MacRae, J. Hébert, G. Sauvageau
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Simon, J. Krosl, T. MacRae, R. Ruel, J. Hébert, G. Sauvageau
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Simon, V.-P. Lavallée, M.-E. Bordeleau, I. Baccelli, G. Boucher, Y.A. Chantigny, S. Lemieux, A. Marinier, G. Sauvageau
Writing, review, and/or revision of the manuscript: L. Simon, V.-P. Lavallée, M.-E. Bordeleau, I. Baccelli, R. Ruel, Y.A. Chantigny, S. Lemieux, A. Marinier, J. Hébert, G. Sauvageau
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Chagraoui, J. Hébert
Study supervision: M.-E. Bordeleau, S. Lemieux, A. Marinier, J. Hébert
Other (performed mutation and transcriptomic characterization and contributed to chemical screens analyses): V.-P. Lavallée
Other (provided all human AML samples with cytogenetic and clinical data for this study): J. Hébert
Acknowledgments
The authors wish to thank Muriel Draoui for project coordination, Sophie Corneau and Nadine Mayotte for sample coordination, Isabel Boivin for mutation validation and chemical screens assistance, and Simon Girard for assistance with protein analysis. The authors acknowledge Marianne Arteau and Raphaëlle Lambert at the IRIC genomics platform for RNA sequencing; Jean Duchaine, Dominic Salois and Sébastien Guiral at the IRIC high-throughput screening platform for assay optimization and chemical screens supervision; Danièle Gagné and Gaël Dulude at the IRIC flow cytometry platform for assistance with flow cytometry acquisition, analysis, and cell sorting; and Christian Charbonneau at the IRIC bioimaging platform for guidance with confocal analysis. The authors thank Dr. Trang Hoang, Dr. Brian Wilhelm, Dr. Toshio Kitamura, Dr. Kathy Borden, and Dr. Oliver Herault for the generous donation of cell lines used in this study. The authors also acknowledge the Banque de cellules leucémiques du Québec (BCLQ) team who characterized and provided all AML samples of the Leucegene cohort with special thanks to G. D'Angelo, C. Rondeau and S. Lavallée.
Grant Support
L. Simon is supported by the IRIC Members Ph.D. Awards and the Faculté de Médecine recruitment scholarship (Université de Montréal), and the Cole Foundation. V.-P. Lavallée is supported by a fellowship from the Cole Foundation. This work was supported in part by the Government of Canada through Genome Canada and the Ministère de l'économie, de l'innovation et des exportations du Québec through Génome Québec and the Fonds de partenariat pour un Québec innovant et en santé, with supplementary funds from AmorChem and, through grants awarded to G. Sauvageau, J. Hébert, A. Marinier, and S. Lemieux. The work was also supported by a CCSRI impact grant to G. Sauvageau. G. Sauvageau and J. Hébert are recipients of research chairs from the Canada Research Chair program and Industrielle-Alliance (Université de Montréal), respectively. The BCLQ is supported by grants from the Cancer Research Network of the Fonds de recherche du Québec-Santé (FRQS).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
References
Supplementary data
List of 97 gene mutations and fusions systematically included in mutational analysis.
Table S6