Abstract
Acute leukemia is a rapidly progressing blood cancer with low survival rates. Unfavorable prognosis is attributed to insufficiently characterized subpopulations of leukemia stem cells (LSC) that drive chemoresistance and leukemia relapse. Here we utilized a genetic reporter that assesses stemness to enrich and functionally characterize LSCs. We observed heterogeneous activity of the ERG+85 enhancer–based fluorescent reporter in human leukemias. Cells with high reporter activity (tagBFPHigh) exhibited elevated expression of stemness and chemoresistance genes and demonstrated increased clonogenicity and resistance to chemo- and radiotherapy as compared with their tagBFPNeg counterparts. The tagBFPHigh fraction was capable of regenerating the original cellular heterogeneity and demonstrated increased invasive ability. Moreover, the tagBFPHigh fraction was enriched for leukemia-initiating cells in a xenograft assay. We identified the ubiquitin hydrolase USP9X as a novel ERG transcriptional target that sustains ERG+85–positive cells by controlling ERG ubiquitination. Therapeutic targeting of USP9X led to preferential inhibition of the ERG-dependent leukemias. Collectively, these results characterize human leukemia cell functional heterogeneity and suggest that targeting ERG via USP9X inhibition may be a potential treatment strategy in patients with leukemia.
This study couples a novel experimental tool with state-of-the-art approaches to delineate molecular mechanisms underlying stem cell-related characteristics in leukemia cells.
Introduction
Acute leukemia is a highly aggressive group of blood malignancies that originate from hematopoietic stem cells (HSC). Accumulation of blast cells in the bone marrow due to deregulation of molecular pathways controlling self-renewal and differentiation of immature blood cells is the main feature of leukemia (1, 2). Well-recognized and prognostic genetic heterogeneity, as well as functional variability, exists among the subsets of leukemia cells obtained from the same patient (3, 4). Functional heterogeneity model posits that a fraction of acute leukemia cells displays sufficient regenerative capacity to propagate the disease, withstand chemotherapy, and cause leukemia relapse. Functional resemblance of these leukemic cells to normal hematopoietic stem cells (HSC) contributed to their nomination as leukemia stem cells (LSC; refs. 5, 6).
Although recent studies have shown that phenotypic and genetic heterogeneity within tumors constitutes a major source of therapeutic resistance (7), efficient tools to identify and pull out functional stem cell from the heterogeneous cell population are still lacking. Experimentally, the presence of functional human LSCs can be proved by their capacity to engraft immunodeficient mice and induce leukemia growth in their hematopoietic organs (8, 9). Up to date, enrichment for LSCs has been reached by focusing on: cell surface markers (10), metabolism (11, 12), cell-cycle quiescence (13), and miRNA bioactivity (14). These studies and others demonstrated extraordinary inter- and even intra-sample (3) heterogeneity for LSC activity that was influenced by disease stage and type of therapy (10, 15, 16). Furthermore, prior studies emphasized the need to devise approaches that enable identification and isolation of viable human LSCs based on stemness state of the single cell to pinpoint molecular regulators that maintain LSC properties.
Here, we capitalized on our recent findings that a +85 enhancer of ERG transcription factor (TF) can be used as a probe of cellular stemness state in human normal and leukemic cells when integrated into a lentiviral reporter system (17). Indeed, we and others demonstrated that the endogenous ERG+85 enhancer is particularly active in human HSCs and a subset of leukemias as well as sensitive to the net activity of the multiple TFs, termed heptad, implicated in stemness program regulation (18–20).
Herein, by barcoding leukemia cell lines and patient sample with ERG+85 reporter, we attempted to enrich and characterize leukemia subpopulations endowed with LSC properties that included superior leukemia initiation, invasion, and drug resistance. Gene expression analysis of the subpopulations with different levels of ERG+85 reporter activity uncovered ERG/USP9X feed-forward–regulatory relationships that can be targeted therapeutically.
Materials and Methods
Full list of references can be found in the Supplementary Data.
Acute myeloid lymphoma patient samples
The study was approved by the Institutional Review Boards of Tel Aviv University (Tel Aviv, Israel) and University Health Network (Toronto, Ontario, Canada). Written informed consent (according to the Declaration of Helsinki) was obtained from all patients. Acute myeloid lymphoma (AML) samples were cultured in StemSpanTM SFEM II medium (StemCell Technologies) supplemented with growth factors [IL3 (10 ng/mL), IL6 (10 ng/mL), G-CSF (10 ng/mL), TPO (25 ng/mL), SCF (50 ng/mL), and FLT3L (50 ng/mL)] on preestablished confluent MS-5 stromal cells.
Cell lines and drug treatments
ELF-153, KASUMI-1, AML193, and ME1 were cultured as recommended by the manufacturer (DSMZ). Jurkat, THP1, and K562 were described elsewhere and grown in RPMI supplemented with FBS (10%), l-glutamine (1%), and penicillin/streptomycin (1%). TEX cell line was grown as described elsewhere (21). All cell lines were authenticated by short tandem repeat profiling using PowerPlex16 HS kit (Promega). Cell number and viability was estimated using hemocytometer and Trypan blue exclusion assay, respectively. For radiation treatment experiments, cells were exposed to ionizing radiation using Cs-137 source at the dose rate 2.8 Gy/minute using GMBH BioBeam 8000 gamma irradiation device (Gamma service). To quantitate clonogenic growth potential of the leukemic cells, we plated Jurkat cells (500 cells/plate for nontreated and 10,000 cells/plate for irradiated) in MethoCult H4100 (StemCell Technologies) supplemented with FBS (30%, MultiCell), 5637 cells’ conditioned medium (10%), l-glutamine (1%), and penicillin/streptomycin (1%), 2-mercaptoethanol (50 μmol/L). Serial replating experiments were performed by harvesting all the colonies from methylcellulose, followed by plating 1,000 cells into new methylcellulose. Colonies were counted under the microscope after 12–14 days of incubation. All cells were maintained in a humidified incubator at 37°C and 5% CO2. All cell lines were tested negative for Mycoplasma by PCR at least once during the period of this study. Ara-C and doxorubicin were purchased from Sigma, WP1130 from Cayman Chemicals, S63845 from Apex, and G9 was provided by Dr. N. Donato (University of Michigan, Ann Arbor, MI).
Leukemia xenotransplantation model
All animal experimental protocols were approved by the Institutional Animal Care and Use Committee of Tel-Aviv University (Tel Aviv, Israel). Jurkat cells were transplanted intrafemorally (IF) as described previously (22) into 8- to 10-week-old NOD-scid IL2Rgnull-3/GM/SF (NSG-SGM3) mice, which were injected intraperitoneally with busulfan (30 mg/kg) 24 hours before transplantation. Human engraftment in the injected and noninjected bones was analyzed 6–8 weeks posttransplantation by flow cytometry analysis using human-specific CD45-Alexa Fluor 750–conjugated antibody, EGFP, and tagBFP positivity. The frequency of repopulating cells was calculated using ELDA software (23).
DNA constructs and cloning
Dual promoter lentiviral reporter vectors (pMIN and pMIN-ERG+85) were described elsewhere (17). The ERG+85 fragment was amplified using Prime GXL high fidelity polymerase (Clontech) from PGL2-ERG+85 vector (24) and cloned into pMIN upstream of mCMV. Full-length human ERG cDNA was subcloned instead of EGFP into a bidirectional lentiviral vector MA1 (25).
USP9X shRNAs were cloned into pLKO1-puro TRC plasmid using EcoRI and AgeI restriction sites. Target sequences for shRNA experiments were CCTAAGGTTAAGTCGCCCTCG (shScramble), CCACCTCAAACCAAGGATCAA (shUSP9X#1), CGCCTGATTCTTCCAATGAAA (shUSP9X #2), and GAGAGTTTATTCACTGTCTTA (shUSP9X #3).
Virus preparation and transduction procedure
Viral particles were generated by transient transfection of 293T cells using CMVDeltaR8.91 and pMD2G constructs as described elsewhere (26). Leukemia cells were infected by addition of viral supernatant to obtain 10%–30% infection rate. Jurkat cells were selected with puromycin (3 μg/mL) for 3 days in the USP9X knockdown experiments. After selection completion, cells were replated for the downstream experiments.
Quantitative RT-PCR and microarrays
RNA was extracted with the TRIzol reagent (Invitrogen) and reverse transcribed with SuperScript III (Invitrogen). Real-time PCR reactions were prepared with SYBR Green PCR Master Mix (Applied Biosystems) in triplicates and analyzed on Applied Biosystems 7900HT instruments. Absolute gene expression was quantified with SDS software (Applied Biosystems) based on the standard curve method and presented after normalization for GAPDH. List of primers is presented in Supplementary Table S1.
RNA from 100,000 or more sorted cells was used for microarray analysis by Human Gene Clariom S Assay (Thermo Fisher Scientific). Data were processed, normalized, and log2 transformed. Microarray raw data are included as Supplementary Data. The data can be viewed in the following link (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE131079).
Flow cytometry sorting and analysis
EGFP and tagBFP expression were analyzed using Gallios Flow Cytometer and Cytoflex Flow Cytometer (Beckman Coulter, Inc.) For the intracellular ERG protein-level analysis, cells were fixed in formaldehyde (1.4%), permeabilized in ethanol (100%), followed by labeling with anti-ERG rabbit mAb conjugated with PE (clone A7L1G, Cell Signaling Technology, 1:50) or rabbit anti-IgG control (sc-2027, Santa Cruz Biotechnology). Jurkat cells were sorted using BD FACSAria Fusion sorter (BD Biosciences). In ERG overexpression experiments, anti-CD271-APC (Miltenyi Biotec, 1:50) was used for sorting of ERG-transduced Jurkat cells.
Apoptosis analysis
Apoptosis was measured using Annexin V (Invitrogen) and Zombie NIR Fixable Viability Dye (BioLegend) according to the manufacturer's protocols. Cells were analyzed with Gallios flow cytometer (Beckman Coulter). Flow cytometry data were analyzed using Kaluza flow cytometry analysis software (Beckman Coulter).
Western blotting analysis
The following primary antibodies were used: mouse anti-β-actin (clone 8H10D10, Cell Signaling Technology, 1:2,000), rabbit anti-USP9X (gene ID 8239, Bethyl Laboratories, 1:1,000), rabbit anti-ERG (sc-354, polyclonal, Santa Cruz Biotechnology, 1:2,000), mouse monoclonal anti-ubiquitin (Santa Cruz Biotechnology, sc-8017, 1:1,000), and mouse anti-GAPDH (clone 258, Invitrogen, 1:2,000).
Migration assay
Costar Transwells (insert diameter 6.5 mm, 5 μm/pore) were coated with 100 μL Matrigel (Corning, 356231), covered with medium, and polymerized for 30 minutes at 30°C. The bottom chamber was filled with 600 μL of MS-5 conditioned medium. Sorted leukemia cells were allowed to migrate at 37°C for 48 hours. The transwells were removed and the cell suspension was analyzed by flow cytometer to determine the cell number and phenotype.
Gene-set enrichment analysis
Gene-set enrichment analysis (GSEA) was performed using the GSEA Java Desktop tool (v3). Gene expression levels were obtained from microarray analysis of three independent replicative experiments comparing tagBFPhigh and tagBFPneg cells. Expression levels and genes were ranked using the SAM algorithm (27, 28). The GSEA preranked tool was used to interrogate the enrichment of various expression signatures including: HSC's super enhancer (29), HSC_R (30), Ara-C resistance (31), doxorubicin resistance (32), activated β-catenin pathway (33), early T-cell development (34), targets of HOXA9 and MEIS1 up/down (35), high BAALC AML (36), epithelial-to-mesenchymal transition (EMT) up (37), and TAL1 bound and negatively/positively correlated (38). Gene-set enrichment analysis for RNA-seq data was performed using GSAASeqSP 2.0 (39).
Analysis of ERG ubiquitination
ERG ubiquitination analysis was performed essentially as described elsewhere (40) with slight modifications. Briefly, ELF153 cells were pretreated with MG132 proteasome inhibitor (5 μmol/L) and then treated with WP1130 USP9X inhibitor (5 μmol/L) for 16 hours. Cells were lysed in lysis buffer [HEPES (25 mmol/L, pH 7.5), NaCl (400 mmol/L), IGEPAL CA-630 (0.5%), DTT (1 mmol/L), glycerol (5%) and protease inhibitors]. The soluble fraction of the lysate was diluted to adjust NaCl and IGEPAL CA-630 concentrations to 100 mmol/L and 0.125%, respectively, and then precleared with magnetic Dynabeads Protein G for immunoprecipitation (Invitrogen, 10003D). Dynabeads were incubated with ERG antibody for 10 minutes at room temperature and lysates were incubated with ERG antibody-Protein G bead complexes at 4°C for 3 hours, followed by bead wash (four times) with NaCl (100 mmol/L) and IGEPAL CA-630 (0.1%). Then, immunoprecipitated complexes were boiled in Laemmli buffer for Western blot analysis. ERG ubiquitination was detected in the above prepared extracts using anti-ubiquitin antibody (Santa Cruz Biotechnology, sc-8017).
Survival analysis
Patients from different AML cohorts (41, 42) were used to determine association between USP9X levels (measured by RNA-seq RPKM analysis or microarray) and survival [overall survival (OS) and event-free survival]. Survival curves were analyzed according to the Kaplan–Meier method and compared using the log-rank test.
Chromatin immunoprecipitation sequencing gene-set enrichment analysis
ChIP-seq diagrams
Protein–protein analysis
Protein–protein interaction analysis was performed using esyN platform (50).
Gene expression resources for USP9X expression analysis
USP9X expression in the different stages of hematopoietic hierarchy was extracted from GSE42414 (http://jdstemcellresearch.ca/node/32; ref. 51). USP9X expression for each AML cytogenetic group in The Cancer Genome Atlas (TCGA) was extracted from https://cancergenome.nih.gov/cancersselected/acutemyeloidleukemia. USP9X expression in chronic myeloid leukemia phases was extracted from GSE4170 (52). USP9X expression in the different AML cytogenetic-risk clusters was extracted from GSE1159 (53).
Results
Identification and characterization of leukemia cells with heterogeneous ERG+85 reporter activities
To characterize functional heterogeneity among cells in human leukemia, we utilized stem cell enhancer element from ERG TF regulatory region (ERG+85). This element contains binding sites for numerous stem cell regulators and possesses high activity in the most primitive normal hematopoietic cells and in the subset of human bulk AML and T-ALL (20). On the basis of our findings that ERG+85 enhancer is particularly active in human HSPCs and a subset of AMLs, we recently developed a lentiviral fluorescent reporter in which ERG+85 enhancer regulates tagBFP expression, while a constitutively active EF1/SV40 promoter drives EGFP cassette (Fig. 1A; ref. 17). Using this tool, we measured ERG+85 reporter activation in a wide range of human leukemia lines and found that its activity positively correlates with the endogenous ERG protein levels (Supplementary Fig. S1; ref. 17). As ERG+85 enhancer can bind numerous high variance TFs implicated in regulation of stem cells in both T- and myeloid leukemogenesis, we hypothesized that this reporter can be a tool to track, characterize, and link transcriptional and functional heterogeneity of leukemia cells. To validate this hypothesis, we selected three leukemia lines (Jurkat, ELF-153, and Kasumi-1) in which we discovered a distinct cell population characterized by the elevated tagBFP activity as compared with the same line infected with a vector lacking ERG+85 enhancer (pMIN; Fig. 1B). Jurkat cells exhibited the lowest activation of the ERG+85 reporter, while ELF-153 and Kasumi-1 lines had the intermediate and high activities, respectively, in correlation with their respective ERG levels (Supplementary Fig. S1B).
ERG+85 enhancer–based reporter demonstrates heterogeneous activity among human leukemia lines. A, Diagram of the lentiviral vector with “stemness-related TFs” (colored geometric shapes) interacting with cis elements identifiable in ERG+85 enhancer. B, Jurkat, ELF-153, and Kasumi-1 cell lines were infected with pMIN and pMIN-ERG+85 vectors. Representative EGFP and tagBFP intensity and distribution in each cell line are shown.
ERG+85 enhancer–based reporter demonstrates heterogeneous activity among human leukemia lines. A, Diagram of the lentiviral vector with “stemness-related TFs” (colored geometric shapes) interacting with cis elements identifiable in ERG+85 enhancer. B, Jurkat, ELF-153, and Kasumi-1 cell lines were infected with pMIN and pMIN-ERG+85 vectors. Representative EGFP and tagBFP intensity and distribution in each cell line are shown.
Relative chemo- and radioresistance are principal hallmarks of cancer stem cells in leukemia (54) and solid tumors (55, 56). However, the identity and the dynamics of therapy-resistant leukemia cells among the bulk malignant cells remain unclear. To test whether cells with heterogeneous ERG+85 activity differ in their response to genotoxic stress, we selected Jurkat leukemia cell line with a distinctive tagBFPHigh and tagBFPNeg subpopulations (Fig. 1B). We detected no difference in the expansion rate between tagBFPHigh and tagBFPNeg cells during in vitro culturing (Supplementary Fig. 2A). However, upon exposure of cells to ionizing radiation, we observed time-dependent enrichment in the proportion of the viable tagBFPHigh cells pointing to their relative resistance. (Fig. 2A; Supplementary Fig. S2B). FACS analysis revealed a diminished induction of the ionizing radiation (IR)-induced cell death in the tagBFPHigh cells, which is consistent with their relative resistance to this injury. (Fig. 2B; Supplementary Fig. S2C). To further elaborate these findings, we flow sorted EGFP+tagBFPHigh cells and EGFP+tagBFPNeg cells (Fig. 2C) and measured their colony-forming capacity. Analysis showed that individual tagBFPHigh cells had higher clonogenic capacity in methylcellulose as compared with the tagBFPNeg cells under both normal and irradiation conditions (Fig. 2D). Importantly, treatment of the ERG+85–expressing Jurkat cells with the additional genotoxic stressor doxorubicin led to a similar increase in the proportion of tagBFPHigh cells, suggesting an inherent resistance of this subpopulation to certain DNA-damaging agents (Fig. 2E). To validate our findings in primary samples, we infected patient-derived AML samples (n = 4) with the ERG+85 reporter followed by treatment with Ara-C or doxorubicin, both commonly used antileukemia drugs. Remarkably, we observed an elevated proportion of tagBFPHigh cells among Ara-C and doxorubicin survivors in three of four samples tested (Fig. 2F; Supplementary Fig. S2D).
Functional characterization of leukemic cells expressing various levels of the ERG+85 reporter. A, Effect of ionizing radiation exposure (3 Gy) on the relative fraction of tagBFPHigh cells in the bulk Jurkat cells infected with the reporter as assessed by flow cytometry at indicated time points. n = 5 independent irradiation and recovery experiments. B, Ionizing radiation (5 Gy) induced apoptosis in tagBFPNeg and tagBFPHigh Jurkat cells at 24 hours postirradiation, quantitated using Annexin V and Zombie-NIR assay. n = 3 independent irradiation experiments. C, Gating strategy to sort tagBFPNeg and tagBFPHigh subpopulations in Jurkat cell line model. D, Clonogenic potential of tagBFPNeg and tagBFPHigh Jurkat cells exposed or not to irradiation (1 Gy; n = 3). E, ERG+85 enhancer activity analysis [measured as mean fluorescence intensity (MFI) ratio] after treatment of Jurkat cells with doxorubicin (100 nmol/L) for the indicated time (n = 3). F, ERG+85 enhancer activity analysis (measured as tagBFP mean fluorescence intensity) after treatment of AML 8227 sample with Ara-C (1 μmol/L) for 48 hours. One representative plot of three independent experiments is shown. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.
Functional characterization of leukemic cells expressing various levels of the ERG+85 reporter. A, Effect of ionizing radiation exposure (3 Gy) on the relative fraction of tagBFPHigh cells in the bulk Jurkat cells infected with the reporter as assessed by flow cytometry at indicated time points. n = 5 independent irradiation and recovery experiments. B, Ionizing radiation (5 Gy) induced apoptosis in tagBFPNeg and tagBFPHigh Jurkat cells at 24 hours postirradiation, quantitated using Annexin V and Zombie-NIR assay. n = 3 independent irradiation experiments. C, Gating strategy to sort tagBFPNeg and tagBFPHigh subpopulations in Jurkat cell line model. D, Clonogenic potential of tagBFPNeg and tagBFPHigh Jurkat cells exposed or not to irradiation (1 Gy; n = 3). E, ERG+85 enhancer activity analysis [measured as mean fluorescence intensity (MFI) ratio] after treatment of Jurkat cells with doxorubicin (100 nmol/L) for the indicated time (n = 3). F, ERG+85 enhancer activity analysis (measured as tagBFP mean fluorescence intensity) after treatment of AML 8227 sample with Ara-C (1 μmol/L) for 48 hours. One representative plot of three independent experiments is shown. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.
In summary, we developed a tool to identify and characterize leukemia cells with distinct stress tolerance characteristics. In addition, we defined ERG+85High cells as “stress-resistant” subset in a population of leukemia cells.
ERG+85High cells exhibit regenerative and invasive properties
To characterize the stability of cellular states identified by ERG+85 reporter, we analyzed tagBFP expression dynamics in sorted tagBFPHigh and tagBFPNeg Jurkat and ELF153 cells. FACS-assisted tagBFP analysis revealed that tagBFPHigh cells capable to regenerate the original population consisting of tagBFPHigh, tagBFPIntermediate, and tagBFPNeg cells. On the other hand, tagBFPNeg cells did not give rise to tagBFPHigh cells even over the extended period of 30 days in culture (Fig. 3A; Supplementary Fig. S3). Similar pattern of tagBFP distribution was found when we measured ERG+85 reporter activity in the progeny of an individual Jurkat tagBFPHigh or tagBFPNeg cells plated in liquid or methylcellulose colony-forming assays (Supplementary Fig. S3A and S3B).
Regeneration and migration/invasion potential of tagBFPHigh and tagBFPNeg fractions. A, Flow cytometry–based analysis of tagBFP dynamics in sorted tagBFPHigh (blue histogram) and tagBFPNeg (red histogram) cells allowed to regenerate for a month. Dotted histograms represent tagBFP levels in tagBFPHigh and tagBFPNeg cells after the sort. A representative FACS plot is shown. Results are representative of eight independent experiments. B, Flow cytometric analysis of the endogenous ERG protein level in tagBFPNeg (red) and tagBFPHigh (blue) subpopulations. Pale histograms represent the isotype control staining for each fraction. A representative plot of three independent experiments is shown. C, Migration/invasion transwell assay scheme. D, Differential migration and invasion potential of tagBFPNeg and tagBFPHigh cells sorted from Jurkat and ELF153 cell lines. Mean ± SD of three (ELF153) and four (Jurkat) independent experiments is shown. **, P < 0.01; ***, P < 0.001; ns, nonsignificant.
Regeneration and migration/invasion potential of tagBFPHigh and tagBFPNeg fractions. A, Flow cytometry–based analysis of tagBFP dynamics in sorted tagBFPHigh (blue histogram) and tagBFPNeg (red histogram) cells allowed to regenerate for a month. Dotted histograms represent tagBFP levels in tagBFPHigh and tagBFPNeg cells after the sort. A representative FACS plot is shown. Results are representative of eight independent experiments. B, Flow cytometric analysis of the endogenous ERG protein level in tagBFPNeg (red) and tagBFPHigh (blue) subpopulations. Pale histograms represent the isotype control staining for each fraction. A representative plot of three independent experiments is shown. C, Migration/invasion transwell assay scheme. D, Differential migration and invasion potential of tagBFPNeg and tagBFPHigh cells sorted from Jurkat and ELF153 cell lines. Mean ± SD of three (ELF153) and four (Jurkat) independent experiments is shown. **, P < 0.01; ***, P < 0.001; ns, nonsignificant.
ERG protein–level analysis by intracellular flow cytometry and Western blotting revealed that tagBFPHigh Jurkat cells had a 2- to 3-fold higher levels of ERG protein as compared with tagBFPNeg cells (Fig. 3B; Supplementary Fig. S3C). Of note, sorted tagBFPHigh and tagBFPNeg Jurkat cells contained a similar number of viral integrations (Supplementary Fig. S3D). Collectively, these results argue against a lentivirus integration bias effect or stochastic variation in tagBFP expression. Furthermore, continuum of cells with various tagBFP levels in the progeny of the tagBFPHigh cells suggests gradual rather than on/off switch mechanism regulating ERG+85 reporter dynamics.
To determine a relative regenerative potential of tagBFPHigh and tagBFPNeg cells, we performed methycellulose colony serial replating assay. We found that tagBFPHigh ELF-153 cells gave rise to more colonies compared with tagBFPNeg cells upon the initial, secondary, and tertiary platings. Strikingly, no colonies were generated by the tagBFPNeg cells upon the third plating suggesting severely diminished self-renewal potential of this subpopulation (Supplementary Fig. S4A). To quantitate the regenerative capacity of patient-derived AML blasts expressing various levels of ERG+85 reporter, we utilized stroma-supported cultures and applied limiting dilution analysis (LDA) conditions. The LDA results revealed 1.7- to 23-fold higher frequency of culture-initiating cells in the tagBFPHigh population relative to the tagBFPNeg fraction. In agreement with this, tagBFPHigh cells were able to regenerate and demonstrated a substantial proliferative potential (up to 40-fold expansion), whereas the majority of the tagBFPNeg cell–initiated cultures failed to expand (Supplementary Fig. S4B–S4D). To better characterize the stability of these functionally distinct states, we analyzed tagBFP dynamics in sorted tagBFPHigh and tagBFPNeg primary AML cells at the day of sorting and after two weeks on stroma. Our results revealed that while tagBFPHigh cells could regenerate cells with high, intermediate, and low ERG+85 reporter activity, the tagBFPNeg blasts did not give rise to the tagBFPHigh cells (Supplementary Fig. S4B–S4D). These results obtained with primary AML samples agree with our findings in the reporter-expressing leukemia cell lines.
Elevated chemotactic migration and invasion capacities are characteristics of stem-like cells in solid tumors as well as in leukemia (57, 58). Thus, we hypothesized that leukemia cells, which are characterized by the higher levels of stem cell–specific ERG+85 enhancer, would exhibit distinct migration and invasion characteristics. To test this hypothesis, we utilized Transwell assay to examine migration of tagBFPHigh and tagBFPNeg cells toward MS5 conditioned medium and their invasion characteristics through reconstituted basement membrane (Matrigel; Fig. 3C). Sorted tagBFPHigh Jurkat cells had higher migration rate as compared with sorted tagBFPNeg cells. In the additional cell model, ELF153, we detected similar migration characteristics in sorted ELF-153 ERG+85–expressing cells. Strikingly, in Jurkat and ELF-153–based models, but not in Kasumi-1, only tagBFPHigh cells demonstrated the invasion ability through the Matrigel layer while virtually no such ability was observed in the tagBFPNeg population (Fig. 3D; Supplementary Fig. S4F). Thus, these functional assays reveal an increased invasive potential of ERG+85High cells. Furthermore, we provide evidence of ERG+85High cells’ capability to regenerate the original “phenotypic” heterogeneity. Combined with elevated stress tolerance, these functional characteristics indicate that ERG+85High cells persist preferentially in a “stem-like” state.
ERG+85-tagBFP reporter enriches for LSCs
The ability to initiate and sustain leukemia is the gold-standard assay for LSCs identification (59). To test whether ERG+85 activity and associated phenotypic heterogeneity correlate with leukemogenic potential, we performed an in vivo limiting dilution analysis (LDA). To that end, increased doses of tagBFPHigh and tagBFPNeg Jurkat cells were transplanted intrafemorally into immunodeficient recipients (Fig. 4A). Both tagBFPHigh and tagBFPNeg cells initiated local leukemic engraftment in the injected bone with the higher blast load formed by the tagBFPHigh cells (Fig. 4B).
Engraftment capability of tagBFPHigh cells in NSG-mice. A, Study design for leukemia establishment and engraftment analysis. B, Leukemic engraftment in the injected bone [right femur (RF)] and in the noninjected collateral bones was analyzed 8 weeks postinjection of Jurkat cells (1 × 105 cells/mouse). Each symbol represents engraftment level in a single recipient. Horizontal lines represent mean for each group. Welch t test was used. C, Leukemia-initiating cell frequency of tagBFPHigh (black line) and tagBFPNeg (red line) Jurkat cells was determined using LDA. The dotted lines indicate the 95% confidence interval. Supplementary Table S2 contains information regarding cell dose and the number of recipients used to calculate the leukemia-initiating cell frequency. D, Flow cytometry analysis of tagBFPHigh (blue) and tagBFPNeg (red) Jurkat cells regeneration capacity in vivo. A representative FACS plot of a single mouse recipient is shown. ns, nonsignificant.
Engraftment capability of tagBFPHigh cells in NSG-mice. A, Study design for leukemia establishment and engraftment analysis. B, Leukemic engraftment in the injected bone [right femur (RF)] and in the noninjected collateral bones was analyzed 8 weeks postinjection of Jurkat cells (1 × 105 cells/mouse). Each symbol represents engraftment level in a single recipient. Horizontal lines represent mean for each group. Welch t test was used. C, Leukemia-initiating cell frequency of tagBFPHigh (black line) and tagBFPNeg (red line) Jurkat cells was determined using LDA. The dotted lines indicate the 95% confidence interval. Supplementary Table S2 contains information regarding cell dose and the number of recipients used to calculate the leukemia-initiating cell frequency. D, Flow cytometry analysis of tagBFPHigh (blue) and tagBFPNeg (red) Jurkat cells regeneration capacity in vivo. A representative FACS plot of a single mouse recipient is shown. ns, nonsignificant.
Migration and leukemia regeneration in the noninjected bones characterize LSCs with the highest initiating potential (60). Strikingly, only tagBFPHigh cells were able to repopulate collateral bones in agreement with their elevated regenerative, migratory, and invasive potentials observed in vitro. Using repopulation of the noninjected bone marrow as LDA read-out, we found an 8-fold enrichment in LSCs in tagBFPHigh cells as compared with tagBFPNeg fraction (Fig. 4C; Supplementary Table S2). Interestingly, while ERG+85 reporter activity remained negative in the leukemic engraftment formed by the tagBFPNeg cells, strong regeneration of tagBFPNeg offspring was observed in mice that received tagBFPHigh cells (Fig. 4D). This in vivo heterogeneity regeneration behavior agrees with our aforementioned in vitro findings. To assess directly whether tagBFPHigh fraction contains functional LSCs with self-renewal potential we performed serial transplantation assay using patient-derived AML sample (#120791) infected with the ERG+85 reporter. We found a gradual elevation in the proportion of the tagBFP+ cells with every round of transplantation (Supplementary Fig. S4E) that paralleled an increase in the LSC frequency (primary 1:50,000, tertiary 1:100; ref. 17). Thus, our in vivo results reveal that ERG+85 reporter–positive cells demarcate LSC-enriched cell population and LSCs properties, such as migration and invasion, are tightly associated with high ERG+85 activity.
Transcriptional profiling of ERG+85High and ERG+85Neg states
To reveal potential regulators that are responsible for the distinct functional traits of ERG+85-tagBFPHigh and ERG+85-tagBFPNeg Jurkat cells, we performed gene expression analysis on these fractions. 391 genes (366 upregulated and 25 downregulated, FDR < 0.01) were differentially expressed between ERG+85High and ERG+85Neg states, albeit with the modest fold change (2.5–4.8 fold change range; Fig. 5A; Supplementary Fig. S5A). To better understand the regulation mode of the tagBFPHigh transcriptome, we investigated enrichment for the transcription factors known to interact with ERG+85 enhancer in the regulatory regions of the differentially expressed genes. By utilizing a TF-specific ChIP-seq datasets, we revealed a strong enrichment for ERG and TAL1 factors in the upregulated genes in the ERG+85High fraction (Fig. 5B), demonstrating preferential role of these transcription factors in governing ERG+85High gene expression profile. In addition, unsupervised gene ontology analysis of regions bound by ERG in Jurkat cells highlighted pathways regulating T-cell activation and HSC function and implicated additional transcriptional regulators such as MEIS1 and β-catenin (Supplementary Fig. S5B).
tagBFPHigh cells enriched with stemness and chemo-resistant gene sets. A, Statistical analysis of microarray (SAM) plot. The two parallel dashed lines are the cut-off threshold specified by the actual false discovery rate, and the total number of upregulated (red dots) and downregulated (green dots) genes are given. B, ChIP-seq enrichment analysis of the relevant TFs in tagBFPHigh signature genes. Scores are calculated using -log10(P) × (Enrichment Factor-EF)2 formula and the algorithm explained elsewhere. C, GSEA. Transcripts were ranked from top upregulated to top downregulated (in tagBFPHigh vs. tagBFPNeg fractions) and the ranked list was investigated using GSEA analysis. D, Clinical relevance of tagBFPHigh signature enrichments in different AML/T-ALL–related gene sets. AML cohort of TCGA with profiled transcriptome was used: the top 15% most expressing ERG AML patients and the 15% lowest were defined as ERGhigh and ERGlow AML patients, respectively. Database: AML, TCGA.
tagBFPHigh cells enriched with stemness and chemo-resistant gene sets. A, Statistical analysis of microarray (SAM) plot. The two parallel dashed lines are the cut-off threshold specified by the actual false discovery rate, and the total number of upregulated (red dots) and downregulated (green dots) genes are given. B, ChIP-seq enrichment analysis of the relevant TFs in tagBFPHigh signature genes. Scores are calculated using -log10(P) × (Enrichment Factor-EF)2 formula and the algorithm explained elsewhere. C, GSEA. Transcripts were ranked from top upregulated to top downregulated (in tagBFPHigh vs. tagBFPNeg fractions) and the ranked list was investigated using GSEA analysis. D, Clinical relevance of tagBFPHigh signature enrichments in different AML/T-ALL–related gene sets. AML cohort of TCGA with profiled transcriptome was used: the top 15% most expressing ERG AML patients and the 15% lowest were defined as ERGhigh and ERGlow AML patients, respectively. Database: AML, TCGA.
GSEA of the differentially expressed genes (Supplementary Table S3) revealed positive enrichment for gene signatures implicated in maintaining HSC identity (HSC super-enhancer–associated genes), HSC/LSC function (HSC_R, HOXA9 and MEIS1), chemotherapy resistance (Ara-C, doxorubicin), EMT, and early T-cell development in ERG+85High cells (Fig. 5C; Supplementary Fig. S5C). In agreement with ChIP-seq–based functional annotations, GSEA independently revealed enrichment of MEIS1, TAL1 targets, and activated β-catenin pathway in the ERG+85High cells. This analysis of ERG+85High and ERG+85Neg transcriptomes suggests that cell-to-cell variability in ERG+85 activation defines leukemia cells with distinct developmental, differentiation, and fitness potentials, which are consistent with the functional heterogeneity we described previously.
To determine the clinical relevance of the transcriptional signature identified in ERG+85High Jurkat cells, we tested its performance in the stratification of various primary AML and T-ALL datasets. We found significant enrichment of ERG+85High-specific transcripts in AML samples that expressed high levels of ERG as compared with samples expressing low levels of ERG. Strikingly, analysis of three independent studies (61–63) demonstrated that ERG+85High characteristic gene program was significantly enriched in AML and T-ALL samples obtained at relapse as compared with samples obtained at the diagnostic stage (Fig. 5D). In conclusion, our combined transcriptomics and bioinformatics analysis uncovered distinct transcriptional fingerprints associated with dynamic ERG+85High subpopulation with potential functional and clinical importance.
ERG/USP9X–positive feedback loop regulates heptad TFs network stability
Our results demonstrate that ERG+85 reporter activity varies among Jurkat cells, which is consistent with the highly dynamic nature of the TF network interacting with this genomic element (64–66). Positive and negative feedback loops that can reinforce or diminish TF expression (and thus stabilize ERG+85Neg or ERG+85High states) are general phenomenon in the developmental TF networks (67). Thus, we hypothesized that potential regulators of the heptad TFs can be differentially expressed in ERG+85High versus ERG+85Neg states. To validate this hypothesis, we performed bioinformatics analysis of potential gene-regulatory interactions between heptad TFs and differentially expressed genes and observed a number of regulatory interactions between the two groups (Fig. 6A). Furthermore, we confirmed by qRT-PCR assay differences suggested by microarrays for several genes implicated in these regulatory interactions (e.g., USP9X and SMAD3) as well as in leukemogenesis (e.g., ID2, EZH2, ALCAM, CD28) in independently sorted Jurkat subpopulations (Fig. 6B; Supplementary Fig. S5D).
USP9X is an ERG target gene with a prognostic impact. A, Literature-based functional interaction analysis of the heptad TFs and genes differentially expressed between tagBFPNeg and tagBFPHigh cell subsets. esyN online tool was used to generate this interaction network. B, qRT-PCR based validation of the selected genes (SMAD3 and USP9X) predicted by microarray profiling to be differentially expressed in tagBFPNeg and tagBFPHigh subpopulations. n = 3 independent experiments. CHIP-seq datasets are listed in Materials and Methods. C, USP9X gene expression difference between ERG-expressing AMLs (the highest and the lowest 30% cutoff was used to define ERGHigh and ERGLow groups). D, Kaplan–Meier estimates of OS according to USP9X expression levels that were calculated using RNA-seq datasets from AML-TCGA study. E, Normalized expression of ERG and USP9X genes in Jurkat tagBFPNeg cells infected with the control (1074) or ERG-overexpressing (1074-ERG) vectors. Bars represent means ± SD; n = 3 qRT-PCR experiments. F, ERG+85 reporter activity analysis in tagBFPNeg Jurkat cells 3 or 10 days postinfection with control (1074, black contour plot) or ERG-overexpressing (1074-ERG, green contour plot) vectors. A representative contour plot analysis of four independent experiments is shown. G, ERG protein enrichment in USP9X regulatory region located 3-kb upstream of the transcriptional start site (TSS) was detected by querying the published ERG ChIP-seq datasets for normal CD34+ cells, SKNO1, and KASUMI-1 leukemia cell lines. ****, P < 0.0001. Welch t test was used.
USP9X is an ERG target gene with a prognostic impact. A, Literature-based functional interaction analysis of the heptad TFs and genes differentially expressed between tagBFPNeg and tagBFPHigh cell subsets. esyN online tool was used to generate this interaction network. B, qRT-PCR based validation of the selected genes (SMAD3 and USP9X) predicted by microarray profiling to be differentially expressed in tagBFPNeg and tagBFPHigh subpopulations. n = 3 independent experiments. CHIP-seq datasets are listed in Materials and Methods. C, USP9X gene expression difference between ERG-expressing AMLs (the highest and the lowest 30% cutoff was used to define ERGHigh and ERGLow groups). D, Kaplan–Meier estimates of OS according to USP9X expression levels that were calculated using RNA-seq datasets from AML-TCGA study. E, Normalized expression of ERG and USP9X genes in Jurkat tagBFPNeg cells infected with the control (1074) or ERG-overexpressing (1074-ERG) vectors. Bars represent means ± SD; n = 3 qRT-PCR experiments. F, ERG+85 reporter activity analysis in tagBFPNeg Jurkat cells 3 or 10 days postinfection with control (1074, black contour plot) or ERG-overexpressing (1074-ERG, green contour plot) vectors. A representative contour plot analysis of four independent experiments is shown. G, ERG protein enrichment in USP9X regulatory region located 3-kb upstream of the transcriptional start site (TSS) was detected by querying the published ERG ChIP-seq datasets for normal CD34+ cells, SKNO1, and KASUMI-1 leukemia cell lines. ****, P < 0.0001. Welch t test was used.
To investigate whether predicted regulatory interactions affect ERG+85-associated TF network dynamics and if so, by what mechanism, we focused on studying ERG and USP9X factors. USP9X is a deubiquitinase enzyme that can regulate stability of numerous cancer-related proteins in solid cancers, including ERG (40), MCL1 (68), and β-catenin (69). In agreement with higher expression of USP9X in ERG+85High fraction of Jurkat cells (Fig. 6B), we also revealed a higher expression of USP9X in AML samples characterized by elevated ERG levels (Fig. 6C). These findings point to a previously unrecognized positive correlation between the expressions of these two factors in human leukemia.
To provide insights into the potential involvement of USP9X in human leukemogenesis, we investigated changes in the USP9X expression in leukemia samples with established disease parameters. Using TCGA, we uncovered that high USP9X levels in patients with AML are associated with lower overall survival, higher relapse rate, and poor cytogenetics (Fig. 6D; Supplementary Fig. S6A–S6G).
To investigate the potential transcriptional regulation of USP9X gene by ERG, we ectopically expressed ERG cDNA in ERG+85Neg Jurkat cells. We observed USP9X mRNA upregulation in ERG-overexpressing cells as well as gradual tagBFP induction that becomes pronounced only after 10 days (Fig. 6E and F).
Furthermore, we revealed enrichment of ERG transcription factor in the USP9X promoter region in several AML cell lines and CD34+ cord blood cells by analyzing published ChIP- seq datasets (Fig. 6G). Interestingly, human HSCs expressed elevated USP9X levels relatively to the more differentiated subtypes (Supplementary Fig. S6H). A more comprehensive inspection revealed a canonical ETS TF DNA–binding motif in the USP9X promoter indicating a possible ERG-USP9X–positive feedback loop (Supplementary Fig. S6I).
Collectively, our results highlight the existence of the genetic link between USP9X, ERG, and ERG+85-associated stemness network and point to the functional involvement of USP9X in human leukemia.
USP9X inhibition impairs ERG+85 transcriptional network and inhibits leukemia cells’ functionality
To validate whether alteration in USP9X levels would affect ERG expression and ERG+85-tagBFP reporter activity, USP9X expression was downregulated using RNA interference approach (Fig. 7A; Supplementary Fig. S7A and S7B). USP9X knockdown by two different shRNAs (shUSP9X #2 and shUSP9X #3) led to the reduction in ERG+85 reporter activity coupled with a decrease in ERG protein levels (Fig. 7B). These results confirm that endogenous USP9X regulates ERG protein and uncovers high sensitivity of the ERG transcriptional network to changes in one of its components.
Functional interrogation of the ERG/USP9X feedback loop. A, Jurkat cells were infected with the indicated viruses expressing USP9X or scramble shRNAs and then selected with puromycin (3 μg/mL) for 3 days. At day 5 after infection, cells were analyzed using qRT-PCR for USP9X mRNA reduction. n = 3 independent experiments. B, ERG+85 reporter activity and ERG protein expression analyses upon USP9X knockdown. The histograms shown are representative of three independent experiments. C, Differential sensitivity of tagBFPNeg and tagBFPHigh cells to USP9X inhibitor –WP1130. Cells were treated for 3 days with the indicated concentrations of WP1130 and analyzed for viability using FACS. Dye exclusion assay, n = 3 independent experiments. D, Susceptibility of primary AML samples to WP1130 treatment. Cells were treated for 24–72 hours with the indicated concentrations of USP9X inhibitor and then cell number and viability were determined by flow cytometry. E, Ubiquitination status of ERG after WP1130 treatment. ELF153 cells were treated with WP1130 in the presence of MG132. Immunoprecipitation was performed using anti-ERG antibody and immunoblotting was performed using anti-ubiquitin antibody. Input ERG protein is shown in the bottom image. F, Effect of WP1130 on the invasive phenotype of tagBFPHigh and tagBFPNeg subpopulations in ELF153 cells. Mean ± SD of three experiments. G, Western blot analysis of ERG protein level after treatment (16 hours) with the indicated concentrations of WP1130. H, Schematic depiction of the proposed ERG- and USP9X-positive feedback loop relationships. *, P < 0.05; **, P < 0.01; ****, P < 0.0001; ns, nonsignificant.
Functional interrogation of the ERG/USP9X feedback loop. A, Jurkat cells were infected with the indicated viruses expressing USP9X or scramble shRNAs and then selected with puromycin (3 μg/mL) for 3 days. At day 5 after infection, cells were analyzed using qRT-PCR for USP9X mRNA reduction. n = 3 independent experiments. B, ERG+85 reporter activity and ERG protein expression analyses upon USP9X knockdown. The histograms shown are representative of three independent experiments. C, Differential sensitivity of tagBFPNeg and tagBFPHigh cells to USP9X inhibitor –WP1130. Cells were treated for 3 days with the indicated concentrations of WP1130 and analyzed for viability using FACS. Dye exclusion assay, n = 3 independent experiments. D, Susceptibility of primary AML samples to WP1130 treatment. Cells were treated for 24–72 hours with the indicated concentrations of USP9X inhibitor and then cell number and viability were determined by flow cytometry. E, Ubiquitination status of ERG after WP1130 treatment. ELF153 cells were treated with WP1130 in the presence of MG132. Immunoprecipitation was performed using anti-ERG antibody and immunoblotting was performed using anti-ubiquitin antibody. Input ERG protein is shown in the bottom image. F, Effect of WP1130 on the invasive phenotype of tagBFPHigh and tagBFPNeg subpopulations in ELF153 cells. Mean ± SD of three experiments. G, Western blot analysis of ERG protein level after treatment (16 hours) with the indicated concentrations of WP1130. H, Schematic depiction of the proposed ERG- and USP9X-positive feedback loop relationships. *, P < 0.05; **, P < 0.01; ****, P < 0.0001; ns, nonsignificant.
To further substantiate the role of USP9X in regulation of leukemia cell properties, we utilized two partially selective USP9X deubiquitinase inhibitors WP1130 (70) and G9 (71) combined with leukemia lines differing in their ERG+85 reporter activity (Supplementary Fig. S7C). WP1130 as well as G9 treatment induced a dose-dependent decline in viability in all cell lines; however, leukemia lines that transactivated ERG+85 reporter (e.g., TEX, ME1 and ELF153) were more sensitive to USP9X inhibitor–mediated growth inhibition than cell lines lacking ERG+85 transactivation (e.g., K562, THP1, AML193; Fig. 7C; Supplementary Fig. S7D). To test whether MCL1 plays a major role in the preferential sensitivity of the ERG+85High cells to the USP9X inhibitor, we treated the same cell line panel with the MCL1-specific inhibitor S63845 (72). We found no correlation between the ERG+85 reporter activity and viability decline upon treatment with S63845 (Supplementary Fig. S7E), suggesting that MCL1 inhibition solely could not explain the ERG+85 activity–dependent sensitivity of cells to USP9X targeting. To extend our findings, we treated patient-derived AML samples containing ERG+85-positive fraction with WP1130 (Fig. 7D). Variable degree of growth inhibition in four of five samples was detected.
To test directly whether USP9X inhibition affects the ubiquitination status of the endogenous ERG TF in leukemia, we treated leukemia cells with WP1130 and performed ERG protein immunoprecipitation followed by immunoblotting with an antibody against ubiquitin. Indeed, treatment of ELF153 cells with WP1130 resulted in the time-dependent increase in the level of ubiquitinated endogenous ERG protein (Fig. 7E; Supplementary Fig. S8B).
We have previously shown that only ERG+85High cells exhibit strong invasive properties in vitro and in vivo. To examine whether this invasive potential can be affected by the novel ERG+85 regulator USP9X, we performed the invasion assay in the presence of WP1130. As can be seen in Fig. 7F, WP1130 exposure led to inhibition of tagBFPHigh cells’ invasion.
In agreement with our findings in the growth inhibition assays, we observed degradation of ERG protein upon WP1130 treatment in sensitive, but not resistant samples (Fig. 7G; Supplementary Fig. S8C).
To summarize, ERG regulates USP9X transcription by binding to its promoter while USP9X can stabilize ERG via deubiquitination. Interference with this feed-forward regulation via USP9X targeting can inhibit growth and invasion potential of the ERG+85-positive leukemias (Fig. 7H).
Discussion
Advanced tools suitable for separating functionally distinct subpopulations of leukemia cells are obligatory for investigation and efficient targeting of LSCs. Moreover, functional characterization of LSC subpopulations can benefit from a real-time cell tracking tool to conduct diverse in vitro and in vivo studies. While in our recent study (17) we described the bioinformatics and experimental strategy to identify one such tool as a marker for stemness state (ERG+85 enhancer) and developed a fluorescent lentiviral reporter that can accurately recapitulate this enhancer endogenous activity, in this study, we advanced this reporter system for functional and molecular dissection of leukemia cell heterogeneity. Furthermore, we identified the USP9X as ERG-inducible deubiquitinase with a role in leukemia cells’ invasion and growth.
Cell-to-cell functional heterogeneity has being recognized as one of the driving forces for leukemia therapy failure and disease relapse (3, 4, 31, 73). So far, cell characterization strategies based on metabolic state labeling (74), dye efflux (75), or differential cell surface markers (73) provided solid support for the existence of the functionally distinct leukemia subpopulations in the same sample. Cell-to-cell heterogeneous activity of the ERG+85 reporter, as imaged by our system, allows dissection of the functional heterogeneity according to the stemness program–related activity. Indeed, elevated expression of normal HSC-associated transcripts in human leukemias constitutes an adverse predictor of patient survival independent of other genetic and clinical criteria (30, 76). Reduced apoptosis induction, superior chemo- and radioresistance of tagBFPHigh subpopulation, as revealed in this work, coupled with marked enrichment for the relapse-associated gene signature point to the association between the two characteristics. Moreover, adapting of this reporter labeling strategy to the high-throughput drug screen holds the potential for the early identification of small molecules capable targeting relapse-initiating cells.
Using this reporter in multiple established and primary AML samples, we showed that only ERG+85High leukemic cells are capable of giving rise to ERG+85Neg cells and sustain multiple rounds or replating. These findings suggestive of the extensive regenerative and self-renewal potential of the ERG+85High fraction agree with our previous findings with normal cord blood–derived CD34+ cells in which high ERG+85 activity characterize cells with the most immature (CD34+D38−) phenotype (17). The lack of noticeable transition from ERG+85Neg cells toward ERG+85High cells in spite of prolonged in vitro or in vivo growth is surprising, giving the stochastic-state transitions previously reported in breast cancer (77). This finding may indicate that ERG+85 enhancer–based reporter enables identification of the more stable cell states that could not be identified by approaches utilizing a limited predefined set of cell surface markers or transient/metabolic labeling. Indeed, only by using an extensive repertoire of cell surface markers, distinct leukemia stem cells lineages were identified in a specific murine leukemia model driven by MLL-AF9 (78).
In addition, our findings shed light on the understanding of LSC phenotypes and permit their analysis in situ. ERG+85 reporter discovered at least two distinct self-renewing and leukemia-initiating lineages in Jurkat cells that differ in their functional capabilities. Boosted migratory/invasive activity in vitro combined with the capacity to repopulate distant bone marrow territories in vivo suggests that ERG+85High cells can have important clinical implications. For example, exploration of various bone marrow niches can endow selective protection on ERG+85High cells from chemotherapy or targeted therapy (54, 79). Preferential expression of genes related to EMT and activated β-catenin by ERG+85High cells, as discovered by our current transcriptomics analysis of Jurkat subsets, can provide potential clues on regulators implicated in invasive, migratory, and chemo-resistance phenotypes (57, 80). In addition, high expression of ALCAM (CD166) in ERG+85High cells, which is required for efficient migration and homing of HSCs, might also yield therapeutic opportunity to target these cells (81, 82).
Positive feedback loop between ERG TF and its deubiquitinase USP9X, which we revealed here, exemplifies a distinctive utilization of the differential ERG+85 reporter activity for reconstructing gene-regulatory networks implicated in leukemia cells’ functional heterogeneity. It can provide plausible molecular mechanism pertaining increased radiation/chemotherapy tolerance and activation of β-catenin pathway as we observed in ERG+85High cells. Indeed, elevated USP9X can stabilize antiapoptosis regulator MCL1 (68, 83) and β-catenin (69). USP9X's ability to reinforce ERG+85 stemness program can provide molecular support to our bioinformatics analysis that demonstrated USP9X elevation in the poor prognosis and relapsed AML samples. Importantly, we demonstrated that leukemia lines and patient AMLs are variably sensitive to WP1130 treatment. These findings advocate the usage of USP9X inhibitors (40, 84) as a promising approach in targeting leukemias that are dependent on ERG/USP9X axis.
In conclusion, our results validate a novel experimental tool to dissect leukemia cells' functional heterogeneity. We believe that insights obtained with this approach can assist in developing personalized treatments aimed to target stem-like subpopulations in human leukemia.
Disclosure of Potential Conflicts of Interest
S. Izraeli is a consultant/advisory board member at SIGHTDX. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
Conception and design: N. Aqaqe, M. Yassin, A.A. Yassin, S. Izraeli, M. Milyavsky
Development of methodology: N. Aqaqe, M. Yassin, A.A. Yassin, S. Izraeli, M. Milyavsky
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Yassin, A.A. Yassin, N. Ershaid, E. Kugler, E.R. Lechman, O.I. Gan, A. Mitchell, M. Milyavsky
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Aqaqe, M. Yassin, A.A. Yassin, M. Milyavsky
Writing, review, and/or revision of the manuscript: N. Aqaqe, M. Yassin, E. Kugler, S. Izraeli, M. Milyavsky
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N. Aqaqe, M. Yassin, N. Ershaid, C. Katz-Even, A. Zipin-Roitman, A. Mitchell, M. Milyavsky
Study supervision: E.R. Lechman, J.E. Dick, S. Izraeli, M. Milyavsky
Acknowledgments
The authors thank Dr. J.E. Pimanda (Prince of Wales Clinical School, University of New South Wales Sydney, Sydney, Australia) for providing the PGL2-ERG+85 luciferase vector, Dr. N. Donato (University of Michigan, Ann Arbor, MI) for providing G9, Dr. L. Broday (Tel Aviv University, Tel Aviv, Israel) for providing reagents and assistance with ubiquitination assay, Dr. L. Shlush (Weizmann Institute of Science, Rehovot, Israel) for providing AML samples. This work was partially supported by Israel Science Foundation (ISF 1512/14 to M. Milyavsky), Varda and Boaz Dotan Research Center in Hemato-Oncology (to M. Milyavsky and S. Izraeli), and Israel Cancer Research Fund (RCDA 14-171 to M. Milyavsky). N. Aqaqe is a recipient of a PhD scholarship from State of Israel Ministry of Science, Technology and Space. M. Yassin is a recipient of Israel Council for Higher Education PhD scholarship for minorities. This work was performed in partial fulfillment of the requirements for a PhD degree of Muhammad Yassin, Nasma Aqaqe, and Eitan Kugler, the Dr. Miriam and Sheldon G. Adelson Graduate School of Medicine, Sackler Faculty of Medicine, Tel Aviv University, Israel.
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
Supplementary figure legends and full list of materials and methods references
Supplementary Table 1
Supplementary Table 2
Supplementary Table 3