Supraphysiologic levels of the AML1-ETO isoform AE9a are essential for transformation

Significance

The AE9a protein (alternative splicing at exon 9) is often used to model t(8;21) leukemia. Our study demonstrates that increased oncogene dosage is a critical parameter of AE9a transformation, likely as a result of impaired transcriptional regulation of AML1-ETO target genes. This insight could assist in identifying those downstream genes most critical for t(8;21)-associated transformation.

Abstract

Chromosomal translocation 8;21 is found in 40% of the FAB M2 subtype of acute myeloid leukemia (AML). The resultant in-frame fusion protein AML1-ETO (AE) acts as an initiating oncogene for leukemia development. AE immortalizes human CD34⁺ cord blood cells in long-term culture. We assessed the transforming properties of the alternatively spliced AE isoform AE9a (or alternative splicing at exon 9), which is fully transforming in a murine retroviral model, in human cord blood cells. Full activity was realized only upon increased fusion protein expression. This effect was recapitulated in the AE9a murine AML model. Cotransduction of AE and AE9a resulted in a strong selective pressure for AE-expressing cells. In the context of AE, AE9a did not show selection for increased expression, affirming observations of human t(8;21) patient samples where full-length AE is the dominant protein detected. Mechanistically, AE9a showed defective transcriptional regulation of AE target genes that was partially corrected at high expression. Together, these results bring an additional perspective to our understanding of AE function and highlight the contribution of oncogene expression level in t(8;21) experimental models.

The t(8;21)(q22;q22) chromosomal translocation comprises the N-terminal DNA binding domain of AML1 (RUNX1) and nearly the entire ETO (RUNX1T1) gene, forming the fusion protein AML1-ETO (AE) (1, 2). Conditional expression and transduction–transplantation approaches have demonstrated that expression of AE provides self-renewal signaling to hematopoietic stem/progenitor cells (HSPCs) but does not induce transformation in the absence of additional cooperating events (3–7). Several such cooperating events have been identified, including overexpression of WT1, mutant c-KIT, TEL-PDGFRB, FLT3-ITD, loss of p21, and treatment with the DNA-damaging agent ENU (6, 8–13). Conversely, C-terminal truncation of AE through frameshift mutation (AML1-ETOtr) or alternative splicing at exon 9 (AE9a) leads to acute myeloid leukemia (AML) transformation of murine HSPCs (14, 15). The AE9a transcript is expressed in ∼70% of t(8;21)⁺ AML patients, along with full-length AE (14, 16). AE9a lacks the conserved ETO domains NHR3/4, whose interaction with corepressor proteins such as N-CoR/SMRT and HDACs is important for transcriptional repression (17–20). Despite maintaining some corepressor interaction, AE9a has greatly diminished N-CoR and SMRT interaction and is a much less potent transcriptional repressor than full-length AE (17, 19–22). The AML1-ETOtr mutant showed altered regulation of cell cycle proteins, thereby providing a proliferative advantage in K562 cells relative to AE (15). Recently, DeKelver et al. showed that the interaction between N-CoR and the NHR4 domain plays an inhibitory role in leukemia development in the context of full-length AE (23). However, analysis of a small cohort of t(8;21) patient samples did not detect any mutations in the NHR4 domain, indicating that such mutations are probably very infrequent (24). The role of AE9a in the genesis of t(8;21)-associated AML is currently unclear.

Previously, we showed that expression of AE in human CD34⁺ HSPCs promotes enhanced self-renewal and long-term expansion of CD34⁺ progenitors, making this model a valuable preleukemia model of t(8;21) disease (25–27). We have also established human-based models using the CBFB-MYH11 and MLL-AF9 fusion proteins (28–30). These models seem to be faithful representations of human disease at multiple levels, demonstrating a high degree of relevance to primary patient samples in a tractable model system. In the present study, we examined the functional consequences of AE9a expression in comparison with full-length AE in human HSPCs. Our data identify oncogene expression level as a critical parameter for AE9a function.

Results

AE9a Promotes Self-Renewal of Human CD34⁺ Hematopoietic Cells.

We examined the function of the AML1-ETO isoform AE9a (and the AE9aΔ2 mutant as a negative control) in human CD34⁺ umbilical cord blood (UCB) cells (Fig. 1A). Cells were transduced with MSCV-HA-AE-IRES-Thy1.1-, AE9a-, or AE9aΔ2-expressing retrovirus and sorted for Thy1.1 expression. AE9a partially recapitulated the effects of AE transcriptional activation and repression although most target genes were regulated to a lesser extent compared with AE, including CEBPA, SPI1, OGG1, CDKN1A, and MPL, indicating some loss of function for AE9a (Fig. 1B). AE9a extended cell growth and maintained a CD34⁺ population throughout the lifespan of the cultures, similar to the effects of AE (Fig. 1 C and D). AE9a cultures consistently trended toward a higher percentage of CD34-expressing cells over time, unlike AE cultures, which showed relatively constant levels of CD34-expressing cells. However, this selection was not advantageous, given the similar growth kinetics of the cells. As expected, deletion of NHR2 in the context of AE9a resulted in a fully nonfunctional mutant, in line with previous findings (31). AE and AE9a dramatically repressed erythroid colony formation and showed a myeloid colony bias, with significantly fewer colonies formed overall compared with controls (Fig. 1E). CD34-expressing cells were maintained in colony-forming unit (cfu) assays, consistent with colony-replating potential for both AE and AE9a (Fig. 1 E and F).

Fig. 1.

AE9a extends CD34⁺ UCB culture life, retains CD34 progenitor marker expression, and engrafts immunodeficient mice. (A) Schematic of HA-tagged AE, AE9a, or AE9aΔ2. CD34⁺ umbilical cord blood (UCB) cells were transduced with MIT-HA-AE, -AE9a, or -AE9aΔ2, or empty vector and sorted for Thy1.1 surface marker expression. (B–G) Quantitative PCR (QPCR) analysis was performed on week 1 transduced AE and mutant cultures. (B) Representative data from four independent clones are presented. (C) Liquid growth of cultures was assessed weekly. (D) CD34 progenitor marker expression was determined by flow cytometry. Combination of three independent experiments (cell lines) is shown (*P < 0.05). (E) Primary (5,000 cells) colony-forming ability was assessed on CD34⁺ UCB cells transduced and sorted as above. Flow cytometry was performed (E, Right). CD33 and CD235a were monitored for myeloid vs. red cell bias. CD34 and Thy1.1 expression were also shown. (F) Secondary colony-forming ability was performed with each sample. Representative results from five independent experiments are shown for C–F. (G) Established long-term AE and AE9a cultures were injected (intrafemorally) into NSGS mice. After 12 wk, BM aspirations were performed and analyzed for engraftment. Representative flow plots are shown.

The ability of the AE9a mutant to transform murine HSPC to AML prompted examination of the engraftment and transformation potential of human cells expressing AE9a in comparison with full-length AE. Cells were transplanted by intrafemoral injection into immunodeficient NOD/SCID IL2RG^−/− (NSG) mice transgenic for the human cytokine stem cell factor, granulocyte–macrophage colony-stimulating factor (GM-CSF), and interleukin-3 (NSGS) as previously described (32). AE and AE9a showed comparable engraftment levels at 12 wk, with high variability and no apparent differences among the tested cultures (Fig. 1G). Despite producing a robust AML in a murine model, AE9a was not able to generate leukemia in the human xenotransplant model under the conditions tested (Fig. 1G).

AE9a Selects for High Expression in Long-Term Human Cultures.

In weekly monitoring of in vitro cultures by flow cytometry using the Thy1.1 surface protein as a surrogate marker for AE or mutant expression, we noticed a reproducible increase in AE9a expression over time (Fig. 2 A and B). This intensity shift was evident in both the CD34-negative and -positive cell fractions. Elevated levels of mutant protein expression were confirmed by immunoblotting (Fig. 2C). Oncogene dosage has previously been shown to play a key role in leukemia development, as demonstrated for the MLL-AF9 fusion and for c-myc (33, 34). Xenotransplant assays were performed to determine whether selection for increased AE9a expression was also evident in vivo. AE or AE9a in vitro cultures with similar levels of fusion protein expression (as shown by Thy1.1 staining) (Fig. 2D, Top) were injected into NSGS mice. Eight weeks postinjection, bone marrow (BM) cells were analyzed for human CD45 and Thy1.1 expression. AE9a cells showed a 10-fold increased Thy1.1 expression relative to AE cells (Fig. 2D, Bottom), demonstrating robust selection for high expression in an in vivo setting.

Fig. 2.

AE9a selects for high expression in long-term cultures. (A) AE and AE9a cultures were sorted as in Fig. 1. Week 1 and week 13 Thy1.1 expression is shown. Representative data from at least five experiments are shown. (B) Mean fluorescence intensity (MFI) of AE9a Thy1.1 expression is shown relative to AE Thy1.1. (C) Western blot analysis of CD34⁺ fraction of long-term cultures (week 18) confirms higher expression in AE9a cells compared with AE. (D) Week 6 AE and AE9a cultures were injected intrafemorally into sublethally irradiated NSGS mice. Mice were killed 8 wk after injection and analyzed by flow cytometry for intensity of AE and AE9a.

The accumulation of cells expressing high levels of AE9a could be due to a number of factors. Protein stability was assessed using the protein synthesis inhibitor cycloheximide (Fig. S1A). Protein turnover for AE9a was actually more rapid than for full-length AE, suggesting that increased protein stability was not contributing to high AE9a expression. Given the decreased ability of AE9a to efficiently repress AE target genes, it was possible that AE9a does not repress expression from the murine stem cell virus (MSCV) promoter [which contains a core binding factor (CBF) recognition site] whereas AE does. However, AE and AE9a showed similar activity on an MSCV luciferase reporter, suggesting that hyperactivation of the MSCV promoter by AE9a is unlikely to be driving the increased AE9a expression (Fig. S1B).

Fig. S1.

Protein stability and MSCV promoter activation for AE9a. (A) Protein stability was assessed upon cyclohexamide treatment of AE or AE9a cultures, followed by Western analysis. Densitometry of the western analyses are shown (Bottom). Data represent three independent experiments. (B) The 293t cells were transfected with AE or AE9a expression plasmid along with an MSCV-luciferase reporter and tk-renilla reporter for normalization. A combination of four independent experiments are shown.

High AE9a Expression Is Essential for Establishment of Long-Term Human Cultures.

To determine whether high expression is necessary for AE9a function, human UCB cells transduced with vector, AE, or AE9a were sorted for expression of the vector-encoded Thy1.1 surface marker immediately after transduction. Fractions with the lowest (30% positive) and highest (30% positive) expression were sorted, and cell growth was monitored on a weekly basis (Fig. 3 A and B). AE-low cells showed robust growth, and expression analysis at week 5 showed that AE-low cultures reached a median level of AE expression, based on both AE protein and Thy1.1 expression (Fig. 3A and Fig. S2). AE-high cells initially showed a lag in proliferation (Fig. 3B, Inset), possibly due to an up-regulation of the p21 cell cycle inhibitor (Fig. 3A), but recovered growth potential after 2–3 wk, presumably due to outgrowth of clones expressing lower levels of AE (Fig. 3 A and B). In contrast, AE9a-high cells had excellent proliferative ability but, interestingly, maintained or even increased AE9a levels over time (Fig. 3 A and B and Fig. S2). Strikingly, AE9a-low cells showed no enhanced proliferation relative to control cells and stopped growing by 6 wk in vitro (Fig. 3B), demonstrating that increased protein levels of AE9a are essential for this function in human UCB cells. A similar result was obtained in methylcellulose cfu assays, with AE9a-high cells consistently producing greater colony numbers upon replating than AE9a-low cells, in contrast to effects seen with full-length AE (Fig. 3C). These data are consistent with the observations from long-term liquid culture indicating that AE-low and AE9a-high cells have selective advantages for growth and suggest that high expression of AE9a is necessary for these functions.

Fig. 3.

Sorting for low AE9a expression prevents establishment of long-term cultures. Human CD34⁺ cord blood cells were transduced with MIT, AE, or AE9a retrovirus and sorted for 30% low and 30% high Thy1.1-positive cells. (A) Western analysis comparing AE low/high and AE9a low/high samples at time of sort with week 5 samples from liquid culture. Anti-RUNX1, anti-p21, and anti–β-actin are shown. (B) Sorted cells were grown in liquid culture medium, and weekly cell counts were performed. The Inset highlights the initial slow growth of AE-high cultures compared with AE-low. (C) CD34⁺ cells transduced and sorted as in A were plated in methylcellulose colony-forming media at 5,000 cells per plate (primary) or 20,000 cells per plate (secondary). After 2 wk, total colony number was counted. Representative plating is shown. A combination of five independent experiments with three replicates each were performed.

Fig. S2.

Flow cytometric analysis of sorted samples. (Top) Flow analysis of initial sort comparing mean fluorescence intensity (MFI) of low and high populations. (Bottom) Week 5 comparison of MFI for Thy1.1⁺ cells showing higher intensity within the AE9a high population. (Right) Relative MFI is shown with AE low-sorted cells set to “1”.

Selective Pressure for High Expression of AE9a in a Murine Leukemia Model.

Murine fetal liver (FL) cells transduced with AE9a and transplanted into recipient mice develop a robust AML (14). To further investigate the effects of high versus low AE9a expression in this model, E14.5 FL cells were transduced with AE or AE9a [using an internal ribosome entry site (IRES)-GFP to monitor expression] and sorted for 30% high and 30% low expression based on GFP (Fig. 4A). Sorted cells were plated in methylcellulose, and colony numbers were determined upon weekly replating. AE-low and AE-high cells were equally efficient in serial replatings (Fig. 4B). In contrast, whereas AE9a-high cells displayed a sustained replating ability, AE9a-low cells showed significantly diminished replating after week 3. AE- or AE9a-transduced FL cells were transplanted into lethally irradiated mice, and peripheral blood (PB) was analyzed at weeks 0, 4, and 10 for mean fluorescence intensity (MFI) of GFP expression. Although initial expression levels of AE and AE9a were comparable, expression of AE9a was significantly higher at week 10 relative to both AE levels at week 10 and initial AE9a levels (Fig. 4C). These data are consistent with results obtained in the human UCB model wherein AE9a selects for high expression in vitro and in vivo.

Fig. 4.

AE9a-high expression facilitates murine replating ability and leukemia development. (A) Embryonic day (E) 13.5–14.5 fetal liver cells were harvested from C57BL/6 mice and transduced with MIG-AE or MIG-AE9a retrovirus. Transduced fetal liver cells were sorted for 30% low and 30% high GFP-positive cells to isolate low and high AE and AE9a expression. (B) Low- and high-sorted cells were plated in methylcellulose colony-forming media. Up to seven serial platings were performed. Combined data from three independent experiments are shown (*P < 0.05). (C) MIG-AE or MIG-AE9a transduced fetal liver cells were transplanted into lethally irradiated BoyJ recipient mice. GFP intensity from initial transduction and from peripheral blood at weeks 4 and 10 are shown. The number in each graph is the mean fluorescence intensity (MFI) value for GFP. (D) Fetal liver cells were transduced and transplanted as in C. AE9a cells were expanded for 2–3 wk in vivo, followed by BM harvest. BM cells were sorted and gated for c-Kit expression, followed by low and high GFP expression. A limited number of GFP-positive cells (1 × 10⁵) were transplanted into sublethally irradiated BoyJ recipient mice. Survival analysis is indicated; P value = 0.02. (E) Representative flow from AE9a low leukemic BM shows comparable MFI (GFP) to an AE9a-high leukemic BM profile. (F) Western analysis of a subset of AE9a-low and -high leukemic mice (Top). Anti-ETO and anti–β-actin are shown. Relative intensity is shown (Bottom) (G, Left) Western analysis of rare AE leukemic mice generated in our studies caused truncation of full-length AE. Anti-runx1 and anti-actin are shown. (Right) Survival curve of AE transplanted mice (48 total mice).

To evaluate effects of AE9a expression levels on AML development, AE9a-transduced FL cells were expanded in vivo for 2–3 wk. Subsequently, BM was harvested, and c-Kit–positive cells were sorted for low (30%) and high (30%) GFP expression (Fig. S3A). GFP-positive cells (1 × 10⁵) were transplanted into sublethally irradiated mice. Survival analysis showed a marked latency shift in mice harboring AE9a-low cells compared with AE9a-high cells, with an overall difference in penetrance as well (Fig. 4D). Importantly, GFP expression and AE9a protein analysis indicated a selection for high expression even in AE9a-low–expressing tumors (Fig. 4 E and F and Fig. S3B). Based on Southern blot analysis, the number of viral integrants did not differ between clones sorted for initial low or high expression of AE9a, with between one and four integrants detected in each clone (Fig. S4). Thus, the high levels of AE9a protein are not directly correlating with integrant copy number, indicating that cell clones have other mechanisms than copy number integration to attain high AE9a levels. Because AE9a is attenuated for transcriptional function relative to AE, as shown in Fig. 1, we tested the expression level differences of defined AE target genes in murine cells sorted for high and low expression of AE9a. Genes were selected based on previously proposed significance as being dysregulated in, and direct targets for, both mouse (AE9a) and human t(8;21) leukemia (35). There was a significant difference in the expression level of multiple AE9a target genes between these two groups, suggesting that one or more AE9a target genes are insufficiently regulated in the AE9a-low cells (Fig. S5 A–D). Furthermore, we generated an AE gene signature using the sorted murine cells expressing AE in comparison with sorted MIG control cells. Unsupervised cluster analysis showed that AE9a-high samples closely resembled the AE samples whereas AE9a-low samples were more heterogenous and less similar to the AE samples (Fig. S5E). Together, these data demonstrate that high expression is a prerequisite to facilitate AE9a AML formation in a murine transduction/transplantation assay and implicate the transcriptional regulation of AE target genes as a possible driving force behind the increase in AE9a expression.

Fig. S3.

AE9a expression in leukemia samples. (A) Schematic of experimental setup for in vivo AE9a sorted high and low expression comparison. (B) Western analysis of a subset of AE9a low and high leukemic mice (Top). Anti-ETO and anti-β-actin are shown. Relative intensity is shown (Bottom).

Fig. S4.

Integration analysis of AE9a low and high leukemic samples. Genomic DNA was isolated from AE9a low and high leukemic spleen cells, digested with BamHI, and hybridized. Integration number was determined by Southern analysis using a radiolabeled GFP specific probe.

Fig. S5.

Expression level comparison of common AE9a target genes. Murine fetal liver cells were harvested and transduced with AE9a-IRES-GFP. Cells were transplanted into mice and allowed to engraft for 2–3 wk. Post in vivo cell expansion, mice were killed and sorted based on 30% low and 30% high GFP expression. RNA was generated from each sample and run using an Affymetrix mouse gene array chip. Results were analyzed using Qlucore Omics Explorer. *P ≤ 0.05, **P < 0.01, ***P < 0.001. Previously described genes observed to be (A) up-regulated in human and mouse t8;21 leukemia, (B) down-regulated in human and mouse t8;21 leukemia, (C) up-regulated in AE9a mouse leukemia from ChIP, and (D) down-regulated in AE9a mouse leukemia from ChIP are shown. (E) Heatmap using an AE gene signature to compare AE9a-low and AE9a-high clustering.

A number of mice that received cells transduced by full-length AE also developed AML (4 out of 48 mice developed AML) (Fig. 4G, Right) that was indistinguishable from AML found in AE9a mice although most mice did not develop any disease, as has been shown by multiple laboratories (7, 36). Interestingly, three of these AMLs expressed protein that was in the same size range expected for AE9a (Fig. 4G, Left). These proteins were also highly expressed relative to the initial levels of AE present in transduced cells. Upon sequence analysis of the proviral integrants, all three had insertions/mutations that mimic an AE9a-like sequence and would be expected to give rise to a truncated AE protein (Fig. S6). These data indicate that there is a strong selective pressure for high expression of a truncated AE protein in the murine transduction/transplantation assay.

Fig. S6.

Sequence comparison of AE-generated leukemia. RNA was isolated, and cDNA was generated from AE leukemic samples from Fig. 4G. Sequencing was performed on the suspected shortened mouse AE leukemic samples from PCR-amplified, gel-purified products. The AE and AE9a plasmid sequence was used for comparison. **Indicates end of protein translation. Blue letters indicate extraneous nucleotides that are not part of the native transcript.

Full-Length AE Dominates Under Conditions of AE Plus AE9a Coexpression in Murine BM Cells in Vitro.

To identify potential selective pressures upon coexpression of both AE and AE9a, murine BM cells were isolated and transduced with AE, AE9a, or AE plus AE9a. Cells were plated in methylcellulose and monitored weekly by flow cytometry for the percentage of transduced cells (Fig. 5). Despite initial robust transduction efficiency, cells expressing AE9a were gradually outcompeted by cells expressing full-length AE or AE plus AE9a, suggesting functional dominance of full-length AE under these conditions. In vivo analysis of coexpression was also examined. Despite several attempts at recapitulating AE9a leukemia in the context of full-length AE, AE9a alone prevailed as the dominant expressing protein (Fig. S7), implying that the presence of AE under these conditions may be deleterious to leukemia formation.

Fig. 5.

Coexpression of AE and AE9a in murine BM cells necessitates the presence of AE fusion in vitro. Murine BM cells were transduced with MIT-AE, MIG-AE9a, or MIT-AE plus MIG-AE9a. Posttransduction, cells were plated in methylcellulose. Cultures were monitored weekly by flow cytometry to measure the percentage of culture of each population. Cells were replated each week for up to 6 wk. (A) Representative flow from week 1 and week 5 cultures demonstrating a push toward AE containing cells at week 5. (B) Percentages of cultures from AE plus AE9a transduced cells are shown. Representative of two independent experiments.

Fig. S7.

Absence of AE expression in mouse leukemia generated from AE plus AE9a sorted cells. Murine fetal liver cells were harvested from E13.5–14.5 embryos and cultured in StemSpan media with rat rSCF, huIL-6, and mIL-3. Cells were transduced with either MIT-AE, MIG-AE9a, or MIT-AE plus MIG-AE9a. Cells were expanded in vivo for 2–3 wk, harvested, and sorted for ∼30% low and ∼30% high GFP expression. Then, ∼100,000 cells were transplanted into sublethally irradiated BoyJ mice. (A) Leukemic samples were analyzed via Western analysis for AE and AE9a expression with anti-Runx1. (B) The Kaplan–Meier curve is shown.

Coexpression of AE and AE9a in Human Cells Results in AE Dominance and Relieves Selective Pressure for Increased AE9a Expression.

The differential selective pressures observed in the murine system prompted investigation of AE and AE9a coexpression studies in the human model. Human CD34⁺ UCB cells were transduced with AE, AE9a, or AE plus AE9a. Transduction frequencies at week 2 of culture ranged from 20% to 50% for each sample (Fig. 6 A and B). AE and AE9a single-transduced cultures showed a gradual selection for the transduced cells, with nearly 100% purity being reached at weeks 6–8 (Fig. 6B). Similar to the results obtained in the murine studies, coexpression in the human system led to an accumulation of AE single-transduced and AE plus AE9a double-transduced cells, with loss of AE9a single-transduced cells over time. Western blot analysis of cells from week 14 cultures showed increased AE9a expression in the single-transduced population in comparison with single-transduced AE cells, similar to our previous findings (Fig. 6C). However, in the context of transduction with full-length AE, AE9a levels were substantially reduced relative to full-length AE, with no sign of selective pressure for high expression. Expression analysis was also performed on double-positive AEplusAE9a sorted cells. Again, lower levels of AE9a were present in the double-positive population compared with AE9a single-transduced cultures (Fig. S8A). Furthermore, although AE9a single-transduced cultures selected for high percentages of CD34-expressing cells over time (Figs. 2B and 6D), AEplusAE9a cells displayed a phenotype resembling that of AE, with much lower percentages of CD34 surface expression (Fig. 6D). To determine the relevance of these findings to t(8;21) AML patient samples, we screened a subset of t(8;21)-positive samples for AE/AE9a protein expression by Western blot (23 patient samples in all). Analysis revealed a dominant band within the size range of full-length AE, with no clear evidence of an AE9a band, similar to results observed with the Kasumi-1 t(8;21) cell line (Fig. 6E and Fig. S8 B and C). Our result that none of the 23 randomly selected t(8;21) patient samples has detectable 9a protein predicts that between 0% and 14.3% of t(8;21) patients express 9a protein (Wilson method with 95% confidence), which is significantly lower than the reported percentage (∼70%) of patients harboring the 9a transcript (binomial exact test P value < 0.05) (14, 16). Collectively, these data underscore the significance of oncogene expression to disease phenotype and suggest that, whereas AE9a is a potent oncogene at high expression levels in murine transduction/transplantation assays, its contribution to t(8;21) disease requires careful verification.

Fig. 6.

Coexpression of AE and AE9a relieves AE9a selective pressure in human cells. CD34⁺ cells were transduced with MIT vector, MIG-AE, MIT-AE9a, or MIG-AE plus MIT-AE9a. (A) Representative flow cytometry at weeks 2, 4, and 14 showing expression of AE (GFP) and/or AE9a (Thy1.1). (B) Cultures were monitored on a weekly basis for GFP and Thy1.1 expression. Percentages of culture are shown for MIT, AE, and AE9a single transductions (Top) and AE plus AE9a (Bottom). Representative of two independent experiments. (C) Week 14 AE, AE9a, and AE plus AE9a bulk cultures were subjected to Western blot analysis for anti-runx1 and anti-actin. (D) CD34 expression was monitored on AE, AE9a, and AE plus AE9a transduced cultures. (E) AML patient samples were analyzed for AE and AE9a expression using an anti-runx1 antibody (Left). The t(8;21) samples are designated. (Right) AE/AE9a ratio of band intensity is shown.

Fig. S8.

AE9a does not select for high expression in the context of AE in vitro. (A) AE, AE9a, or AE plus AE9a week 10 cultures were sorted for a pure population of single- (AE or AE9a) or double-positive (AE plus AE9a) cells. Western analysis was performed using anti-runx1 and anti-actin. (B) AML patient samples were analyzed for AE and AE9a expression using an anti-runx1 antibody. The t(8;21) samples are designated. (C) Western analysis of representative AE and AE9a cell lines compared with a subset of t8;21 samples and the Kasumi-1 cell line.

Discussion

Although an increasing number of studies have used the AE9a oncogene as representative of t(8;21) human leukemia, much remains to be understood regarding the functional differences with full-length AE. Interestingly, AE9a showed a partial deficiency of function in the human preleukemia model, including aberrant activation and repression of select target genes. When an AE-like effect was observed, as in the case of long-term liquid cultures, increased oncogene expression was necessary for this effect. These data bring important insight into the pressures driving transformation in these model systems and raise questions as to the proper interpretation of the data relative to mechanisms of transformation operative in human AML patient samples.

Oncogene expression level has been previously identified as a critical parameter in modulating leukemia development. The MLL-AF9 fusion gene produces very distinct results depending on the expression level of the fusion protein. Both low expression (under the control of the endogenous promoter) and high expression (retroviral transduction) of MLL-AF9 led to leukemia development when targeting the stem cell population, but only high levels of oncogene expression drove leukemia development upon targeting the granulocyte-macrophage committed progenitor (33). In addition, a recent study showed the frequency of leukemia-initiating cells in acute lymphocytic leukemia correlated with increased expression of the oncogene c-myc (34). Our study supports the hypothesis that AE9a preferentially requires selection for high-expressing clones to promote long-term growth and leukemia development whereas low expression is insufficient to elicit a transforming effect. These data are in sharp contrast to results observed with full-length AE, wherein high expression in human cells is inhibitory to cell growth and low expression is preferred. These differences may relate to the differential regulation of target genes, in particular the repression of CEBPA and SPI1 and up-regulation of CDKN1A, the p21 cell cycle inhibitor. A study from Alcalay and coworkers showed significantly reduced recruitment of AE9a to target promoters relative to AE in U937 cells engineered to express one or the other oncogene, supporting our data showing a need for increased expression to elicit an AE-like transcriptional regulation (37). These data could indicate that many of the targets and signaling pathways are in common between AE and AE9a and that increased expression of AE9a is needed to mirror the same transcriptional changes as AE. Our gene expression data comparing AE9a-high and AE9a-low cells would support this idea.

Despite the absence of strong selective pressure for high AE expression in our mouse and human models, clinical evidence supports the importance for high AE levels in t(8;21) leukemia. Disease recurrence and therapeutic response both correlate with transcript levels of the t(8;21) oncogene. At diagnosis, low levels of transcript (on a per-cell basis) trend toward better overall and event-free survival (38). Furthermore, disease relapse is significantly correlated with higher transcript levels (39–42). Similarly, high transcript levels of AE9a also correlated with t(8;21) disease relapse (16, 43). Jiao et al. (16) showed a strong positive correlation between c-Kit expression and high levels of AE9a transcript. Although these studies are strictly based on RNA transcript levels by quantitative RT-PCR, our data showing the importance of high AE9a protein expression in human and mouse model systems support the findings of these clinical studies. It is possible that signals from cooperating events play a role in how AE or AE9a affect downstream signals. A number of cooperating oncogenes have been shown to promote leukemic transformation with full-length AE (6, 8–13). A recent study demonstrated that various c-Kit–activating mutations cooperate with AEtr (the truncated form of AE, similar to AE9a) to initiate an oncogenic-like phenotype in human mobilized PB CD34⁺ cells (44). Interestingly, expression of AEtr alone had minimal effects in their study, indicating possible experimental differences compared with the present study. Along with cell source differences, the use of GM-CSF and serum in the study by Wichmann et al. (44) could have contributed to the observed decline in CD34⁺ cells. Furthermore, it is important to note the limited positive selection elicited by AEtr in their experiments, which may be attributable to low initial transduction frequency and potentially low levels of AEtr expression. Future studies should include examination of cooperating events in the presence of both AE and AE9a.

It is apparent that selective pressure for increased AE9a expression is occurring as a result of necessity. Exactly how this high oncogene expression is achieved in these model systems is unclear. Although we have ruled out increased protein stability and integration as driving factors, it is possible that the specific chromatin geography at integration sites could allow for increased expression. In addition, although we show that AE9a does not positively regulate the MSCV promoter in transient transfection assays, this finding does not rule out that, in the context of intact chromatin, AE is a repressor of the MSCV promoter whereas AE9a has an attenuated function. Interestingly, in the presence of AE expression in both human and murine cells in vitro, high AE9a expression is not positively selected, and, in patient samples, we are unable to readily detect AE9a protein even in cells that express transcript. From these data, we conclude that the transformation elicited by AE9a in murine myeloid cells in vivo is particular to this approach and may not reflect conditions found in t(8;21)⁺ human cells. This conclusion may reflect the relative ease with which murine cells are transformed compared with human cells or may be due to differential activation of signals between the two species in response to AE9a expression. Future studies are needed to explore these hypotheses.

What, then, is the contribution of AE9a to t(8;21) AML? It is still possible that there exist a group of patients whose blast cells express high levels of AE9a protein. Alternatively, the rapid turnover of the AE9a protein could contribute to the apparent absence in patient samples, leaving open the possibility that AE9a is functionally contributing to t(8;21) disease but at much lower expression levels and in ways that cannot be evaluated in model systems. Upon reaching supraphysiologic levels of expression as in the murine and human experimental systems, AE9a seems to resemble full-length AE signaling to a significant extent, with an additional gain-of-function in murine cells in vivo that has not yet been fully characterized. Thus, independent of its role in t(8;21) human disease, AE9a may serve as a valuable model that could potentially provide insight into the molecular mechanisms of AE leukemia. However, care must be taken in interpreting and extrapolating results from this model to t(8;21) human leukemia.

Our studies identified expression level as a critical factor for AE9a leukemia development. Full-length AE may also require such selection for leukemia development but only rarely reached high levels to promote leukemia in our model systems. It will be of interest to examine AE protein levels in murine leukemia samples generated in the context of cooperating events. Further studies will require focusing on factors that control AE expression level and examining the possibility of targeting the signaling cascades that respond to increased expression.

Materials and Methods

Retroviral Plasmids and Transduction.

pMSCV-HA-AE-IRES-Thy1.1 (MIT-AE) and pMSCV-Flag-AE-IRES-GFP were described (45, 46). pMSCV-Flag-AE9a-IRES-GFP was from Dong Er Zhang, University of California, San Diego, La Jolla, CA. AE9a cDNA was cloned into the BamHI and EcoRI sites of MIT-AE. UCB preparation, CD34⁺ isolation, and virus production were as described (28).

Colony-Forming Unit Assays.

Transduced cells were sorted for total or 30% low and 30% high Thy1.1 or GFP expression. Thy1.1-positive cells were plated in MethoCult 4100 (Stem Cell Technologies). Colony assays were as described with the addition of THPO (20 ng/mL) (28). For fetal liver cells, 1,000 GFP⁺ cells were plated in MethoCult GF 3434 (Stem Cell Technologies). Colonies were counted 7 d later. Mouse colony-forming assays were performed in triplicate with 2–3 repeats.

Xenotransplant Assays.

NOD/SCID/IL2rg^−/− (NSG) mice transgenic for the human cytokines stem cell factor (SCF), GM-CSF, and IL-3 (NSGS) (32) were injected intrafemorally with AE or AE9a cultures. Mice were killed and analyzed by flow cytometry. Leukemia was characterized by lethargy, scruffy coat, occasional limb paralysis, and the necessity to sacrifice the animal due to illness, as well as a blast population of greater than 20% in the bone marrow. All antibodies used were from BD Biosciences.

Fetal Liver Experiments.

Transduced fetal liver cells were injected i.v. into BoyJ recipient mice. Cells were expanded 2–3 wk in vivo, mice were killed, and bone marrow (BM) was harvested. c-Kit⁺ cells were sorted (BD FACS Aria) for 30% low and 30% high GFP-expressing populations. Then, 1 × 10⁵ sorted c-Kit⁺/GFP⁺ cells were injected into sublethally irradiated recipient mice.

AML Patient Samples.

Patient samples were used under Cincinnati Children’s Hospital Medical Center (CCHMC) institutional review board (IRB)-approved protocol “Tissue Repository (Tumors & Vascular Anomalies).” Informed consent was obtained from all subjects whose samples were included here.

SI Materials and Methods

Flow Cytometry.

Transduced human cells were stained using anti-CD90.1-biotin antibody (Thyl.1), followed by streptavidin-PE and hCD34 or hCD45 antibodies. Cells were stained for 30 min at 4 °C, washed in PBS plus 2% (vol/vol) FBS, and analyzed on a FACS Canto analyzer (BD).

AML Patient Samples.

Patient samples were collected under IRB-approved protocols. Informed consent was obtained from all subjects participating in these protocols.

Immunoblotting.

Immunoblotting was carried out as described (28).

Murine Fetal Liver Transduction and Transplantation.

Embryonic day 13.5–14.5 fetal liver cells were harvested from C57BL/6 females. Cells were cultured in StemSpan media with 50 ng/mL rat recombinant SCF, 10 ng/mL human IL-6, and 10 ng/mL murine IL-3. After overnight stimulation, cells were collected and transduced with two rounds of retrovirus (MIT-AE, MIG-AE9a, or MIT-AE plus MIG-AE9a). Posttransduction, cells were collected, counted, and transplanted into lethally irradiated BoyJ recipient mice. Cells were expanded 2–3 wk in vivo, mice were killed, and bone marrow (BM) was harvested. c-Kit⁺ cells were sorted (BD FACS Aria) for 30% low and 30% high GFP-expressing populations. Then, 1 × 10⁵ sorted c-Kit⁺/GFP⁺ cells were injected into sublethally irradiated recipient mice. For the AE plus AE9a in vivo experiment, cells were sorted for AE plus AE9a-low (30%) and AE plus AE9a-high (30%).

Xenotransplant Assays.

NOD/SCID/IL2rg^−/− (NSG) mice transgenic for human cytokines SCF, GM-CSF, and IL-3 (NSGS) (32) were injected intrafemorally with established AE or AE9a cultures. Mice were killed and phenotypically analyzed by flow cytometry. All antibodies used were from BD Biosciences.

Antibodies.

Anti-RUNX1 was from Cell Signaling Technologies. Anti-p21 was from Santa Cruz Biotechnology. Anti-ETO was a gift from Inge Olsson, Lund University, Lund, Sweden.

QPCR.

Week 1 cultures were subjected to RNA isolation (RNeasy kit; Qiagen), followed by cDNA synthesis with MulLV RT and random hexamers (Applied Biosystems). QPCR was carried out using SYBR green for product detection on an Eppendorf mastercycler realplex machine.

Southern Blot.

Genomic DNA was isolated from AE9a leukemic mouse spleen cells using DNeasy Blood & Tissue kit (Qiagen). DNA was digested with BamHI and hybridized. A probe directed against GFP was used to determine plasmid integration as described (29).

Reporter Analysis.

The 293t cultures were transfected with the indicated expression plasmids and analyzed for reporter activity using the dual reporter assay system (Promega).

Murine AE9a Low/High Affymetrix Analysis.

Cells were prepared as for murine fetal liver experiments and sorted for 30% low and 30% high GFP⁺ populations. Six AE9a-low and six AE9a-high samples were compared. RNA was isolated via the Qiagen RNeasy RNA isolation kit and submitted to the CCHMC Microarray Core. Samples were hybridized to an Affymetrix Mouse Gene 1.0 ST chip. Data were analyzed using Qlucore Omics Explorer software. Selected gene comparisons were chosen based on the publication by Lo et al. (35). For heatmap generation, CHP files were imported into Qlucore Omics Explorer v3.1 for analysis with embedded statistics functions. AE signature genes were generated by comparing AE-high sorted cells with MIG; 928 genes in total with fold change (up or down) >2 and P value < 0.01 (two-sided t test) were selected as the AE signature. To evaluate how this signature was represented in AE9a, unsupervised clustering analysis was performed on all samples based on the gene expression of the 928 signature genes, and the heatmap was generated.

Sequencing of AE Leukemia Samples.

RNA was isolated from AE leukemic samples (RNeasy kit; Qiagen) and subjected to cDNA synthesis with MulLV RT and random hexamers (Applied Biosystems). cDNA was amplified using a forward AE primer (5′-ACTAAGGCGGTGTCAAGAAG) and a reverse IRES primer (5′-AAGCGGCTTCGGCCAGTAAC) or AE9a-specific reverse primer (5′-GCTGGGCAGCACCTAGGTATGGCC). The amplified product was submitted to the CCHMC DNA sequencing core for sequence analysis.

PCR Screening of t8;21 Patient Samples for AE and AE9a Transcript.

Standard PCR analysis was performed on cDNA generated from t8;21 patient samples. Common (AE and AE9a) forward and AE reverse primers were obtained from Yan et al. (14) as follows: AE and AE9a, F-GAGGGAAAAGCTTCACTCTG; AE, R-TCGGGTGAAATGTCATTGCC; AE9a, R-CCTCATATGACCCAGGACAG. PCR conditions are available upon request.

Statistics and Animal Models.

The Student’s t test was used for all statistical analyses. Numbers of animals were chosen to give sufficient confidence in endpoint readouts with statistical analysis. For animal studies, at least two independent experiments were performed, with six mice for each group. Animals were age-matched, and treatments were distributed among multiple cages to provide randomization. No blinding was performed for animal studies. All animal studies were carried out under an approved Institutional Animal Care and Use Committee (IACUC) protocol.