An Inv(16)(p13.3q24.3)-encoded CBFA2T3-GLIS2 fusion protein defines an aggressive subtype of pediatric acute megakaryoblastic leukemia

Abstract

To define the mutation spectrum in non-Down syndrome acute megakaryoblastic leukemia (non-DS-AMKL), we performed transcriptome sequencing on diagnostic blasts from 14 pediatric patients and validated our findings in a recurrency/validation cohort consisting of 34 pediatric and 28 adult AMKL samples. Our analysis identified a cryptic chromosome 16 inversion (inv(16)(p13.3q24.3)) in 27% of pediatric cases, which encodes a CBFA2T3-GLIS2 fusion protein. Expression of CBFA2T3-GLIS2 in Drosophila and murine hematopoietic cells induced bone morphogenic protein (BMP) signaling and resulted in a marked increase in the self-renewal capacity of hematopoietic progenitors. These data suggest that expression of CBFA2T3-GLIS2 directly contributes to leukemogenesis.

Figure 1. inv(16)(p13.3;q24.3) encodes a CBFA2T3-GLIS2 chimeric transcript

(A) Schematic of chromosome 16 with locations of GLIS2 and CBFA2T3 shown. Arrows indicate orientation of the gene and the green and red lines the probes used for FISH. The protein structure of the genes is shown below chromosome 16 and is not drawn to scale. Breakpoints are indicated by arrows. TAD, transactivation domain; TRD, transcriptional regulatory domain; ZF, zinc finger; NHR, nervy homology region. (B) Schematic of CBFA2T3-GLIS2 chimeric protein. (C) Interphase FISH analysis of two representative patient samples carrying CBFA2T3-GLIS2. The GLIS2 probe is green, the CBFA2T3 probe is red. White arrows indicate the fusion event. Scale bars, 10 m. (D) RT-PCR for CBFA2T3-GLIS2 and GAPDH on the discovery cohort. See also Figure S1 and Tables S1-S6.

Figure 2. Somatic mutations in whole genome sequenced AMKL cases

Plots depict structural genetic variants, including DNA copy number alterations, intraand inter-chromosomal translocations, and sequence alterations (Krzywinski et al., 2009). DNA copy number alterations: Loss of heterozygosity (LOH), orange; amplification, red; deletion, blue. Sequence mutations in Refseq genes: silent single nucleotide variants (SNVs), black; UTR, brown; non-silent SNVs, blue. Genes at structural variant breakpoints: genes involved in in-frame fusions, red; others, green.

Figure 3. Low frequency chimeric transcripts in pediatric AMKL

Four chimeric transcripts were identified in one case each of the discovery cohort: GATA2-HOXA9, MN1-FLI1, NIPBL-HOXB9, and NUP98-KDM5A. (A) RT-PCR validation of the discovery cohort. Primers and conditions are described in supplementary experimental procedures. (B) Interphase FISH analysis of M703 carrying the MN1-FLI1 chimeric protein. The MN1 probe is red, the FLI1 probe is green. White arrows indicate the fusion event. Scale bar, 10 m. (C) Schematic of chimeric proteins. Exons and domains are not drawn to scale. TAD, transactivation domain; NRD, negative regulatory domain; ZNF, zinc finger; MIM, Meis interaction motif; HD, Hox domain; Ets, E-twenty six domain; FLS, Fli1 specific region; CTA, C-terminal transactivation domain; GLN, glutamine rich domain; NLS, nuclear localizing signal; HEAT, Huntingtin/EF3/PP2A/TOR1 domain; FG, phenylalanine-glycine repeats; JMJ, jumonji domain; ARID, AT-rich interaction domain; PHD, plant homeodomain. See also Figure S2.

Figure 4. CBFA2T3-GLIS2 defines a unique subtype of AML with a distinct gene expression signature and poor outcomes

(A) Principal component analysis of the gene expression profiles of the AMKL discovery cohort and 32 other non-AMKL AML samples representing all other known genetic subtypes of pediatric AML. Clusters were generated using 1000 genes selected by k-means algorithm. A detailed description of the samples included in this analysis can be found at NCBI gene expression omnibus, accession GSE35203. (B) Heat map of differentially expressed genes in the top scoring network module of CBFA2T3-GLIS2 positive and negative AMKL patient samples. For gene relationships please see Figure S3. For a detailed list of the top 500 differentially expressed the genes (not limited to this network), please see Table S7. (C) Overall survival of 40 pediatric non-DS AMKL cases treated at multiple institutions (CBFA2T3-GLIS2 negative cases n=28, and CBFA2T3-GLIS2 expressing cases, n=12). The curves for the two groups were tested by logrank method and exact test using permutation which yielded a p value of p=0.05. (D) Overall survival of 19 pediatric non-DS AMKL cases treated at St. Jude Children's Research Hospital (CBFA2T3-GLIS2 negative cases, n=9, and CBFA2T3-GLIS2 expressing cases, n=10). The curves for the two groups were tested by logrank method and exact test using permutation which yield a p value of p=0.03. See also Figure S3 and Table S7.

Figure 5. CBFA2T3-GLIS2 leads to enhanced replating of hematopoietic cells

(A) Experimental design. Murine bone marrow cells were transduced with retroviral vectors expressing mCherry alone (MSCV-IRES-mCherry, “MIC”), or mCherry along with GLIS2, or CBFA2T3-GLIS2. Transduced cells were purified by sorting mCherry positive cells and plated onto methylcellulose containing IL3, IL6, SCF, and EPO. Colonies were counted after 7 days of growth and replated serially. (B) Semi-quantitative RT-PCR of GLIS2 utilizing cells harvested from first round of plating. GLIS2 primers are specific for the 3’ half of the transcript and thus pick up both full length GLIS2 as well as CBFA2T3-GLIS2. Expression in MIC cells was defined as 1, and data is pooled from two separate experiments with similar results. p=<0.0001 as determined by one-way ANOVA. Error bars represent mean ± SEM of two independent experiments. (C) Number of colonies detected at 7 days following each plating. Error bars represent mean ± SEM of two independent experiments. (D) Colony morphology detected in GLIS2 and CBFA2T3-GLIS2 modified cells from the 2^nd plating and beyond. a, CFU-Meg; b, CFU-GM. Scale bars, 500 m. Representative cytospins and morphology of each colony type are shown. c, CFU-Meg; d, CFU-GM. Scale bars, 50 m. (E) Cells harvested from colony forming assays after 3 or more replatings were subjected to flow cytometry. Cells were negative for acetylcholinesterase (data not shown).

Figure 6. CBFA2T3-GLIS2 activates the BMP pathway

(A) The Hedgehog (HH) signaling pathway. In addition to classic hedgehog targets such as PTCH and HHIP, WNT and BMP gene expression have been demonstrated to be affected by the GLI transcription factor in various models (Dahn and Fallon, 2000; Ingham and McMahon, 2001; Vokes et al., 2007). (B) Gene expression profiles from CBFA2T3-GLIS2 containing AMKL cases and other AML subtypes were evaluated for expression levels of BMP, WNT, and HH target genes. CBFA2T3-GLIS2 negative AMKL cases are not shown in this analysis. Significantly upregulated probe sets (FDR less than 0.05) are designated with red font: BMP2 FDR 1.06×10-17, BMP4 FDR 0.015976, PTCH1 FDR 2.05×10-6, and HHIP FDR 0.0038. (C) Murine bone marrow cells were transduced with retroviral vectors carrying mCherry alone (MSCV-IRES-mCherry, “MIC”), mCherry plus GLIS2, or CBFA2T3-GLIS2. mCherry positive cells were sorted and plated in methylcellulose containing IL3, IL6, SCF, and EPO. Following one week of growth, RNA was isolated, reverse transcribed, and amplified with Bmp4 or Hprt specific primers. Error bars represent mean ± SEM of four independent experiments. A representative gel is shown (-, neg; M, MIC; G, GLIS2; and C-G, CBFA2T3-GLIS2). p=0.047 as determined by one-way ANOVA. (D) GLIS2 and CBFA2T3-GLIS2 sensitize murine hematopoietic cells to BMP receptor type I inhibition. Colony formation assays were conducted in the presence or absence of dorsomorphin at the indicated concentrations (Yu et al., 2008). IC50 values were calculated as the amount of drug required to inhibit 50% of the colony formation as determined by colony counts. Error bars represent mean ± SEM of two independent experiments. p=0.036 as determined by one-way ANOVA. See also Figure S4.

Figure 7. CBFA2T3-GLIS2 is present in AMKL cell lines

(A) GLIS2 expression as determined by gene expression arrays in 991 human cancer cell lines. Log₂ transformed expression levels are shown. Data obtained from the Broad-Novartis Cancer Cell Line Encyclopedia (http://www.broadinstitute.org/ccle/home). 34 AML cell lines are included, the extreme outlier of this subtype, M07e, is indicated. The GLIS2 probe set recognizes the end of the transcript and thus does not distinguish between wild type GLIS2 and CBFA2T3-GLIS2. (B) RT-PCR on 5 AMKL cell lines: MEG-01, CHRF-288-11, RS-1, WSU-AML, and M07e. The three cell lines carrying CBFA2T3-GLIS2 were validated by FISH. Scale bars, 10 μm. (C) Real time semi-quantitative RT-PCR of GLIS2, BMP2, BMP4, and ID1 on the 5 AMKL cell lines. Expression levels relative to Beta Actin are shown. CHRF-288-11 expression levels were set to 1 for comparison across cell lines. Error bars represent mean ± SEM of two independent experiments. (D) Dorsomorphin sensitivity in the cell lines as determined by MTT assay. Error bars represent mean ± SEM of two independent experiments. For cell line information and MTT assay, please see supplemental experimental procedures.

Figure 8. Transgenic CBFA2T3-GLIS2 Drosophila ectopically express Dpp

(A) CBFA2T3-GLIS2 was expressed under control of Apterous-Gal4 (strong epithelial dorsal driver). dpp-lacZ serves as a reporter for dpp induction. Wing imaginal discs were isolated at the late 3^rd instar, stained for β-gal as a read out for dpp (green), CBFA2T3 (red), and DAPI (blue) followed by immunofluorescence analysis. Nuclear localization of CBFA2T3-GLIS2 can be seen by the pink signal (inset). Scale bars, 100 μm. (B) CBFA2T3-GLIS2 was expressed under control of C765, a weak epithelial driver. Pharate adults were dissected from pupal casings and imaged. Arrows indicate ectopic notum, broadened and shortened legs. No C765>GLIS2 Drosophila matured to adulthood. Arrows indicate ectopic veins in wings of rare C765>CBFA2T3-GLIS2 escapers. See also Figure S5.