To test the oncogenic activity of the MSI2-HOXA9 fusion, we generated an MSI2-HOXA9 gene reflecting the breakpoint reported in CML patients (t(7:17)(p15;q23)). Specifically, we fused the amino terminus of MSI2 (containing both RNA binding domains [RRMs]) with the carboxyl terminus of the HOXA9 protein (containing the homeobox domain) (Figure 1A). We infected murine hematopoietic stem/progenitor (ckit+, Lineage-, Sca1+ [KLS]) cells with viruses encoding BCR-ABL together with either MSI2-HOXA9 or an empty vector and tested their colony-forming ability (Figure 1B). Compared to BCR-ABL and empty vector (hereafter referred to as just BCR-ABL), the combination of BCR-ABL- and MSI2-HOXA9-infected cells exhibited significantly increased colony formation in primary (Figure 1C) and secondary plating (Figure 1D).

To test whether the MSI2-HOXA9 fusion confers a growth advantage in vivo, we transplanted hematopoietic stem and progenitor cells infected with BCR-ABL and MSI2-HOXA9 or an empty vector into lethally irradiated recipient mice and monitored disease development over time (Figure 1B). Animals transplanted with cells harboring both BCR-ABL and MSI2-HOXA9 exhibited far more aggressive disease with an ~3-fold increase in chimerism relative to the cohort receiving cells carrying BCR-ABL alone (Figure 1E). To determine whether the disease observed with BCR-ABL and MSI2-HOXA9 was more aggressive because it was more undifferentiated than the disease driven by BCR-ABL alone, we compared the BCR-ABL/MSI2-HOXA9 leukemic cells to the previously established model of bcCML driven by BCR-ABL/NUP98-HOXA9 (Dash et al., 2002). To this end, we transplanted mice with KLS cells infected with either BCR-ABL, BCR-ABL/NUP98-HOXA9, or BCR-ABL/MSI2-HOXA9 and sorted Lin- leukemia cells from these mice. Even though Lin- cells were compared, which are generally more undifferentiated, based on Giemsa staining, BCR-ABL/MSI2-HOXA9 cells displayed an increase in less differentiated blasts (82% vs. 61%) and a concomitant decrease in more differentiated cells compared to BCR-ABL-driven disease and, in this regard, was most consistent with the previously described bcCML model (Figure 1F–I, Figure 1—figure supplement 1A–C). Additionally, by FACS, total Lin- content was also increased in BCR-ABL/MSI2-HOXA9 relative to BCR-ABL alone (57% vs. 16%, Figure 1—figure supplement 1D). The change in differentiation state driven by MSI2-HOXA9 cells suggested that when coupled with BCR-ABL, MSI2-HOXA9 can drive chronic phase CML to blast crisis, and thus serve as an oncogene.

Reflecting the development of this more aggressive disease, mice transplanted with BCR-ABL/MSI2-HOXA9-infected cells displayed significantly reduced survival. Thus, while the median survival of mice carrying BCR-ABL alone was 37 days, that of the BCR-ABL/MSI2-HOXA9 group was just 18 days, an 8-fold increase in the risk of death (HR = 0.11) (Figure 1J). Collectively, these findings indicate that the MSI2/HOXA9 fusion protein has oncogenic activity and can cooperate with BCR-ABL to drive cancer progression.

Given that MSI2 and HOXA9 both play key roles in regulation of gene expression – MSI2 as an RNA binding protein and HOXA9 as a homeodomain transcription factor – we tested whether the oncogenic ability of MSI2-HOXA9 was dependent on the RNA binding domains of MSI2 (RRM1 and RRM2) or on the DNA binding domain (DBD) of HOXA9. We generated MSI2-HOXA9 domain mutants lacking either RRM1 or RRM2, as well as a truncated mutant lacking the HOXA9 DBD (Figure 2A). To test whether the domain mutants would blunt the growth advantage provided by MSI2-HOXA9 in the presence of BCR-ABL, we infected KLS cells with viruses encoding BCR-ABL and either full-length MSI2-HOXA9 or the domain mutants and performed colony assays. The loss of either RRM1 or RRM2 led to a significant loss in colony-forming ability compared to cells infected with BCR-ABL/MSI2-HOXA9 (>2-fold for both ΔRRM1 and ΔRRM2), indicating that each RNA binding domain was required for MSI2-HOXA9 activity (Figure 2B). While loss of the HOXA9 domain did not significantly impact primary colony formation, some impact (1.7-fold reduction) was seen at secondary plating (Figure 1—figure supplement 1E). These data suggest that the domains could be differentially important, with the RRMs having a dominant impact on bcCML growth. Consistent with the impact on growth in vitro, transplantation of BCR-ABL/MSI2-HOXA9ΔRRM1 (also listed as BCR-ABL/ΔRRM1) blunted leukemic growth in vivo, while loss of the HOXA9 portion of the MSI2-HOXA9 fusion did not impact chimerism compared to BCR-ABL/MSI2-HOXA9 (~30% for MSI2-HOXA9, ~25% for ΔHOXA9, ~7% for ΔRRM1; Figure 2C; compare with Figure 1E). Although each RRM was necessary for bcCML growth, expression of full-length wild-type MSI2 did not impart any growth advantage when combined with BCR-ABL in either primary (Figure 2D) or secondary colony assays (Figure 2E) relative to BCR-ABL alone. The fact that the 5’ portion of MSI2 but not the full-length protein is oncogenic suggests that the 3’ end may regulate or mask MSI2 oncogenic activity, consistent with the highly regulated activity of wild-type MSI2 proteins during development and in stem cell maintenance. Whether this regulatory effect is due to altered binding activity resulting from a change in MSI2 protein conformation, a shift in cellular localization relative to wild-type MSI2, or through some other mechanism is not clear and represents an important question for future study.

We also tested which domains may be important for driving the more undifferentiated state created by MSI2-HOXA9. Analysis of established primary cancer cells stained with Giemsa and May-Grunwald stains indicated that while both RRM1 and RRM2 similarly impacted growth in colony assays (78 colonies for MSI2-HOXA9, ~35 colonies for both ΔRRM1 and ΔRRM2; Figure 2B, Figure 1—figure supplement 1E), only deletion of RRM1 impacted differentiation (Figure 2F–I, Figure 1—figure supplement 1F–H), leading to a reduction in blast cell counts (~80% for MSI2-HOXA9, 60% for BCR-ABL only, 58% for ΔRRM1; Figure 2F) and an increase in differentiated myeloid cells (Figure 2G–I), and resulting in a more chronic phase disease similar in composition to that driven by BCR-ABL alone.

In comparing the two MSI2 RRMs, the data above suggested that RRM1 is more critical than RRM2. This is consistent with RRM1 having been shown to provide most of the specificity and higher binding energy relative to RRM2 (Ohyama et al., 2012), and thus we focused on RRM1 deletion for subsequent in vivo investigation. Further, supporting the observation that the MSI2 RRM1 is critical for growth and differentiation, the deletion of RRM1 from Msi2-HoxA9 led to significantly longer survival that was no different from the survival of mice transplanted with BCR-ABL alone (p=0.97) (Figure 2J). Thus, while the median survival of mice with leukemia driven by BCR-ABL/MSI2-HOXA9ΔHOXA9 (BCR-ABL/ΔHOXA9) was 26 days, the loss of RRM1 led to a median survival of 46 days, similar to BCR-ABL alone (43.5 days). This resulted in a 12-fold (HR = 0.08) increase in risk of death for BCR-ABL/ΔHOXA9 compared to BCR-ABL only and a 15-fold (HE = 0.06) increase in risk of death for BCR-ABL/ΔHOXA9 compared to BCR-ABL/ΔRRM1. Together, these results demonstrate that the MSI2 RRM1 domain is preferentially critical for the lethality seen in BCR-ABL/MSI2-HOXA9 bcCML .

To understand how the MSI2-HOXA9 translocation promotes bcCML, we compared gene expression patterns in BCR-ABL and BCR-ABL/MSI2-HOXA9-driven disease. RNA-seq analysis was carried out on lineage-negative cells from leukemia established with BCR-ABL or BCR-ABL/MSI2-HOXA9 (Figure 3A, Figure 3—figure supplement 1A). Network mapping of all differentially expressed genes (q-value<0.05) using nonredundant functional grouping revealed an enrichment of metabolic processes with oncogenic pathways and developmental programs (Figure 3B). Programs that were dominantly upregulated by MSI2/HOXA9 were those involved in development, including Aldh1a1, Erbb3, and Kit, consistent with BCR-ABL/MSI2-HOXA9 driving a more undifferentiated disease (Figure 3C), known oncogenes, including Frat1, Map7, and Fzd3 (Figure 3D), and components of the mitochondria, including mt-Co2, mt-Atp8, and mt-Nd5 (Figure 3E). Gene set enrichment analysis (GSEA) suggested an enrichment in genes associated with the mitochondrial transcription pathway (p-value<0.05) (Figure 3F); this finding was subsequently confirmed by qRT-PCR (Figure 3G). The impact on mitochondrial gene expression was unexpected, and we thus pursued this to gain insight into whether BCR-ABL/MSI2-HOXA9-driven mitochondrial changes may be pivotal in driving leukemia growth.

Among the genes most enriched in BCR-ABL/MSI2-HOXA9-driven bcCML were those encoding key components of the mitochondrial respiratory chain, including Complex I, Complex III, Complex IV, and the F0 subunit of ATP synthase (mt-Atp6 and mt-Atp8). Analysis of metabolic capacity and mitochondrial function on Lin- bcCML using the Seahorse Cell Mito Stress Test (Figure 4A) revealed that MSI2-HOXA9 significantly increased basal respiration (Figure 4B and C), ATP production (Figure 4D), and maximal respiration (Figure 4E). These three findings suggest that the introduction of MSI2-HOXA9 results in a cell operating at a higher energetic level. In contrast, most of the other mitochondrial parameters we evaluated were not significantly impacted (Figure 3—figure supplement 1B–E). Further, BCR-ABL/MSI2-HOXA9 increased basal glycolysis (Figure 4F and G), glycolytic reserves (Figure 4H), and maximum glycolysis (Figure 4I). Again, these data suggest that in addition to requiring a higher basal energy level, they also exhibit greater maximal energy capacity. We also observed an increase in the lactate output to glucose uptake ratio in BCR-ABL/MSI2-HOXA9 cells compared to BCR-ABL cells in vitro (Figure 3—figure supplement 1F–H), suggesting greater energy demands in these cells. A plot of the basal oxygen consumption rate (OCR) against extracellular acidification rate revealed a cellular energetics landscape wherein BCR-ABL/MSI2-HOXA9 cells are characteristically more energetic than BCR-ABL cells (Figure 4J). Finally, OCR data showed that BCR-ABL/MSI2-HOXA9 cells operate in a more energetically optimal way; their mitochondria were in a state close to 3.5 (Figure 4K), with state 3 being the theoretical maximal respiration state (Chance and Williams, 1955). Collectively, these findings suggest that MSI2-HOXA9 exerts its effects at least in part by augmenting mitochondrial function and basal respiration in leukemia cells.

To understand the basis of the alteration in mitochondrial gene expression in BCR-ABL/MSI2-HOXA9 leukemia and the concomitant increase in energy production, we tested whether MSI2-HOXA9 impacted mitochondrial gene expression directly in the mitochondria or indirectly through control of the polymerase and/or transcription factors. We thus assessed expression of the mitochondrial polymerase and selected mitochondrial transcription factors and found a 5-fold increase in expression of the lone mitochondrial polymerase (Polrmt) in BCR-ABL/MSI2-HOXA9 cells, as well as significant increases in the transcription initiation factor Tfb2m and the mitochondrial transcription factor regulator Nrf1, 1.5-fold and 1.4-fold, respectively (Figure 5A). While Tfb2m and Nrf showed a significant increase in expression, we chose to focus on the much greater impact seen on Polrmt expression. Since these components are transcribed in the nucleus, the data suggest that MSI2-HOXA9 controls mitochondrial gene expression by promoting expression of POLRMT and key transcription factors. Consistent with this, MSI2-HOXA9 was preferentially found in the nucleus relative to MSI2 (Figure 5B–D). Further, we identified 10 occurrences of the minimal MSI2 RRM1 consensus binding sequence UAGU within the Polrmt transcript. These data suggest a model in which MSI2-HOXA9 localizes to the nucleus and elevates Polrmt expression. POLRMT and the mitochondrial transcription factors, in turn, control expression of mitochondrial genes that encode the components of the respiratory chain, leading to increased mitochondrial function and cellular energetics that drive the aggressiveness of bcCML (Figure 5E).

All animal experiments were performed according to protocols approved by the University of California San Diego Institutional Animal Care and Use Committee (#S10274). Mice were bred and maintained in the animal care facilities at the University of California San Diego. The following mice were used: B6-CD45.2 and B6-CD45.1 (Strain: B6.SJL-Ptprc^aPepc^b/BoyJ); NSG mice (Strain: NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ). All mice were 8–16 weeks of age.

For BCR-ABL1/NUP98-HOXA9-driven bcCML, BCR-ABL-driven CML, or BCR-ABL/MSI2-HOXA9, bone marrow-derived KLS cells were isolated and sorted from CD45.2 B6 mice. All sorted cells were cultured overnight in X-Vivo15 media (Lonza) supplemented with 50 μM 2-mercaptoethanol, 10% (vol/vol) fetal bovine serum (FBS), 1% penicillin-streptomycin, SCF (100 ng/ml, R&D Systems) and TPO (20 ng/ml, R&D Systems). Cells were retrovirally infected with MSCV-BCR-ABL-IRES-NGFR and MSCV-NUP98-HOXA9-IRES-huCD2 to generate bcCML; MSCV-BCR-ABL-IRES-NGFR and MSCV-EMPTY-IRES-GFP to generate CML; MSCV-BCR-ABL-IRES-NGFR and MSCV-MSI2-HOXA9-IRES-GFP to generate BCR-ABL/MSI2-HOXA9 leukemia. Subsequently, cells were harvested 48 hr after infection. For primary transplants, infected and sorted cells were transplanted retro-orbitally into cohorts of sublethally (6 Gy) irradiated CD45.1 mice. Disease mice were analyzed as previously described (Zimdahl et al., 2014).

For primary plating, KLS cells isolated from B6-CD45.2 mice were infected as described above, incubated in the presence of virus for 48 hr, and then sorted for oncogene tags (NGFR and GFP). 250 cells were plated in triplicate in cytokine-free methylcellulose media (Methocult GM M3234, StemCell Technologies) supplemented with 20 ng/ml IL-3 and 20 ng/ml SCF (R&D Systems), 2% FBS, and 1% penicillin-streptomycin. For secondary plating, primary plates were dissociated, replicates were combined, and cells were collected and washed in PBS. 2500 cells were re-plated in fresh methylcellulose media and cytokines.

Cells were suspended in Hanks’ balanced salt solution (Gibco, Life Technologies) containing 5% (vol/vol) FBS and 2 mM EDTA and prepared for FACS analysis and sorting as previously described (Domen et al., 2000). Red blood cells were lysed using RBC Lysis Buffer (eBioscience) before antibody incubation. For KLS isolation, cells were incubated with CD117 (ckit) magnetic beads (Miltenyi Biotec) followed by positive selection using an AutomacsPro (Miltenyi Biotec). Following ckit enrichment, cells were incubated with 2B8 (ckit), D7 (Sca1) and the following antibodies to define lineage-positive cells: 145–2C11 (CD3ε), GK1.5 (CD4), 53–6.7 (CD8), RB6-8C5 (Ly-6G/Gr1), M1/70 (CD11b/Mac-1), TER119 (Ly-76/TER119), 6B2 (CD45R/B220), and MB19-1 (CD19). The same general procedure of RBC lysis and antibody incubation was used for all other flow cytometry experiments. All antibodies were purchased from BD Pharmingen, eBioscience, or BioLegend. All cell sorting and flow cytometry analysis was carried out on a FACSAria III machine (Becton Dickinson), and data were analyzed with FlowJo software (version 10.7.1).

The day prior to plating cells, four-well removable chamber slides (Lab Tek II) were coated with Cell-Tak and kept at 4°C overnight. The next day, FLAG-tag MSCV-MSI2-IRES-GFP or FLAG-tag MSCV-MSI2HOXA9-IRES-GFP-infected K562 cells were labeled with 100 nM MitoTracker Deep Red FM (Thermo Fisher) at 37°C. After washing twice with PBS, cells were fixed with 4% PFA at 37°C and washed with warm PBS. Cells were spun onto the Cell-Tak-coated chamber slides with the brake off and then allowed to settle for an additional 30 min. Cells were then permeabilized with 0.1% Triton X-100 for 3 min and then blocked in PBS with 10% normal goat serum, 5% bovine serum albumin, 0.3 M glycine, and 0.05% Tween-20 for 1 hr. After blocking, cells were incubated with primary antibody in 1:10 diluted blocking buffer overnight at 4°C. The following primary antibodies were used: chicken anti-GFP 1.5:1000 (Abcam) and rabbit anti-FLAG 1.5:1000 (Cell Signaling Technology). Alexa Fluor-conjugated secondary antibody (1:500) incubation was performed for 1 hr at room temperature. DAPI (4′,6-diamidino-2-phenylindole; Molecular Probes) was used to detect DNA. Images were obtained with a Zeiss LSM-700 confocal microscope.

Lin- cancer was sorted from transplanted mice and spun onto cytospin slides. Slides were allowed to air dry overnight before staining with May-Grunwald (Sigma) and Giemsa (diluted 1:20 with DI H₂O, Sigma) stains and again allowed to air dry overnight. Imaging was done using the Keyence BZX-700 fluorescent microscope.

MIG-BCR-ABL was provided by Warren Pear and Ann Marie Pendergast and was cloned into the MSCV-IRES-NGFR retroviral vector. MSCV-NUP98-HOXA9-IRES-YFP was provided by Gary Gilliland and was subcloned into the MSCV-IRES- huCD2 vector (or MSCV-IRES-GFP) retroviral vector. Virus was produced in 293T cells transfected using X-tremeGENE HP (Roche) with viral constructs along with VSV-G and gag-pol. Viral supernatants were collected for 3 days followed by ultracentrifugal concentration at 20,000 × g for 2 hr.

RNA was isolated using RNeasy Micro Kit (QIAGEN), and RNA was converted to cDNA using Superscript III reverse transcriptase (Life Technologies). Quantitative real-time PCRs were performed using an iCycler (Bio-Rad) by mixing cDNAs, iQ SYBRGreen Supermix (Bio-Rad), and gene-specific primers. Gene expression was normalized to the levels of Beta-2 microglobulin (B2m). All primers were designed using NCBI’s Primer-BLAST (https://www.ncbi.nlm.nih.gov/tools/primer-blast/). Primer sequences are listed in Supplementary file 2.

Lin- leukemic cells were sorted from mice transplanted with BCR-ABL/Control or BCR-ABL/MSI2-HOXA9-transduced KLS cells. Sorted cells were plated in round-bottom plates in X-Vivo15 media (Lonza) supplemented with 50 μM 2-mercaptoethanol, 10% (vol/vol) FBS, 1% penicillin-streptomycin, SCF (100 ng/ml, R&D Systems) and TPO (20 ng/ml, R&D Systems). The media supernatant was collected on days 1, 2, and 3, and cell counts were obtained by Trypan Blue counting on days 1 and 3. Glucose, glutamine, glutamate, and lactate concentration in the collected supernatant were measured using a YSI 2950 metabolite analyzer.

The day prior to running the assay, the wells of a XFp plate were coated with Cell-Tak and the plate was kept at 4°C overnight. Lin- leukemic cells were sorted from mice transplanted with BCR-ABL/Control or BCR-ABL/MSI2-HOXA9-transduced KLS cells. Sorted cells were washed twice with complete assay media: XF-DMEM pH 7.4, 1 mM pyruvate, 2 mM L-glutamine, and 10 mM D-glucose. The cells were then aliquoted, spun down, resuspended in 50 µl complete assay media, and added to the previously coated XFp plate. The XFp plate was then spun at 300×g for 1 min with the brake off, and cell sedimentation was observed by microscope before loading onto the Seahorse XFp analyzer.

Lin- leukemic cells were sorted from mice transplanted with BCR-ABL/Control or BCR-ABL/MSI2-HOXA9-transduced KLS cells. Total RNA was isolated using the RNeasy Micro Plus Kit (QIAGEN). RNA libraries were generated from 150 ng of RNA using Illumina’s TruSeq Stranded mRNA Sample Prep Kit (Illumina). Libraries were pooled and single-end sequenced (1×75) on the Illumina NextSeq 500 using the High Output V2 Kit (Illumina).

BCR-ABL/Control and BCR-ABL/MSI2-HOXA9 fastq files were processed into transcript-level summaries using kallisto (Bray et al., 2016). Transcript-level summaries were processed into gene-level summaries, and differential gene expression was performed using sleuth with the Wald test (Pimentel et al., 2017). GSEA was performed as previously described (Lytle et al., 2019). ClueGO was used for gene enrichment analysis of all differentially expressed genes (FDR<0.05) identified between BCR-ABL/Control and BCR-ABL/MSI2-HOXA9 mouse cells. GO and Reactome gene sets were used with medium network specificity, a p-value cutoff of <0.05 and a kappa score of 0.4. All other statistical parameters remained on default settings. CluePedia was used to identify genes found within enriched gene sets. All network analyses ran and were visualized in Cytoscape 3.7 (Cytoscape, ClueGO/CluePedia).

K562 and 293T cells were obtained directly from ATCC. Quality control through ATCC included testing for bacterial, fungal, and mycoplasma contamination. ATCC also performed STR profiling to authenticate the identity of the lines.

Statistical analyses were carried out using GraphPad Prism software version 9.2 (GraphPad Software Inc). All data are shown as mean ± SEM. Two-tailed unpaired Student’s t-tests were used to determine statistical significance. No statistical method was used to predetermine sample size and no data were excluded from the analysis. All experiments were reproducible.