Methylation and silencing of miRNA-124 by EVI1 and self-renewal exhaustion of hematopoietic stem cells in murine myelodysplastic syndrome

Abstract

By expressing EVI1 in murine bone marrow (BM), we previously described a myelodysplastic syndrome (MDS) model characterized by pancytopenia, dysmegakaryopoiesis, dyserythropoiesis, and BM failure. The mice invariably died 11–14 months after BM transplantation (BMT). Here, we show that a double point mutant EVI1-(1+6Mut), unable to bind Gata1, abrogates the onset of MDS in the mouse and re-establishes normal megakaryopoiesis, erythropoiesis, BM function, and peripheral blood profiles. These normal features were maintained in the reconstituted mice until the study was ended at 21 months after BMT. We also report that EVI1 deregulates several genes that control cell division and cell self-renewal. In striking contrast, these genes are normalized in the presence of the EVI1 mutant. Moreover, EVI1, but not the EVI1 mutant, seemingly deregulates these cellular processes by altering miRNA expression. In particular, the silencing of miRNA-124 by DNA methylation is associated with EVI1 expression, but not that of the EVI1 mutant, and appears to play a key role in the up-regulation of cell division in murine BM cells and in the hematopoietic cell line 32Dcl3. The results presented here demonstrate that EVI1 induces MDS in the mouse through two major pathways, both of which require the interaction of EVI1 with other factors: one, results from EVI1–Gata1 interaction, which deregulates erythropoiesis and leads to fatal anemia, whereas the other occurs by interaction of EVI1 with unidentified factors causing perturbation of the cell cycle and self-renewal, as a consequence of silencing miRNA-124 by EVI1 and, ultimately, ensues in BM failure.

The inappropriate activation of EVI1 in 10–15% of myelodysplastic syndrome (MDS) patients is associated with megakaryocytic and erythroid dysplasia, refractory anemia unresponsive to erythropoietin (EPO) administration and bone marrow (BM) failure (1). By expressing EVI1 in murine BM cells, we generated a mouse model of MDS. The reconstituted mice showed dysplastic erythropoiesis and megakaryopoiesis, progressive pancytopenia, severe anemia, and BM failure leading to their death at 11–14 months after BM transplantation (BMT), confirming the association between EVI1 and MDS in the mouse (2). However, in contrast to the majority of EVI1-positive MDS patients, the disease in the mice never progressed to acute myeloid leukemia (AML). We proposed that the primary causes of death in the EVI1-positive mice were severe anemia, associated with loss of response to Epo, and BM failure. Using in vitro assays, we later determined that the transcription factor Gata1, required for activation of EpoR and c-Mpl (3), was functionally impaired by inappropriate interaction with EVI1, but not by EVI1-(1+6Mut), a point mutant of EVI1 that does not recognize Gata1 (4). The study reported here had two major objectives. First, we set out to determine whether the disruption in vivo of the EVI1–Gata1 interaction alleviates MDS in the mouse. We compared two groups of EVI1- and EVI1-(1+6Mut)-positive mice and showed that the point mutant EVI1 group of mice, in which the EVI1–Gata1 interaction is disrupted, exhibited normoblastic erythropoiesis. In addition, the point mutant EVI1 did not induce cytopenia and BM failure in the recipient mice, which like the control animals, appeared healthy and survived until the experiment was terminated at 21 months after BMT. The second objective was to identify the mechanism(s) by which EVI1 expression leads to BM failure. We used a candidate gene approach and gene expression arrays for these studies. The results revealed that the expression of EVI1 leads to the down-regulation of miRNA-124 expression. Bisulfite DNA sequencing demonstrated that silencing of miRNA-124 is caused by CpG island methylation associated with EVI1, but not with the mutant EVI1-(1+6Mut). We further show that repression of miRNA-124, in turn, ostensibly up-regulates genes involved in cell cycling, especially Cyclin D3, and of genes linked to self-renewal, most notably, Bmi1. Importantly, our results suggest that EVI1 plays a dual role in the pathogenesis of MDS in a murine model: one, through its interaction with Gata1, resulting in defective erythropoiesis, and the other, by interaction with as yet unknown factors leading to repression of miRNA-124, causing deregulation of cell cycling and self-renewal, ultimately producing BM failure.

Results

Disruption of Two Zn Finger Motifs Eliminates EVI1-Positive MDS in the Mouse.

Previously, we established that EVI1-(1+6Mut) is unable to interact with Gata1 in murine BM cells in vitro (4). To determine whether EVI1-(1+6Mut) could reduce the severity of MDS in vivo, we expressed EVI1-(1+6Mut), EVI1, or the retroviral vector in lineage negative (Lin-) murine BM cells by infection, and after BMT the animals’ phenotype and survival were compared with those of control mice as described (2). Fluorescent activated cell sorting (FACS) with the Ly5.2+ marker showed that the engraftment of C57BL/Ly5.2 donor cells in C57BL/Ly5.1 recipient mice ranged between 67% and 78%. The expression of EVI1 or EVI1-(1+6Mut) in the recipient mice was confirmed by Western blotting (Fig. S1A). To confirm the reproducibility of our data, we carried out two separate BMT experiments 9 months apart with a total number of 9 EVI1, 10 EVI1-(1+6Mut), and 4 control mice. Because the two experiments had identical results, the data are combined and discussed together. To follow the progress of the disease, periodic peripheral blood (PB) counts were performed. As reported (2), at 9–10 months after BMT, all of the EVI1-positive mice developed mild anemia and cytopenias that became very severe, resulting in death ≈3–4 months later (Fig. S1 B and C). The PB smears of the moribund EVI1-positive mice showed cytopenias and anisopoikilocytosis (Fig. S1D). In contrast, at the time when all of the EVI1-positive mice had succumbed to their disease, the EVI1-(1+6Mut)-positive mice appeared healthy (Fig. S1C). There was no sign of cytopenias as indicated by ranges of red blood cells (RBCs), white blood cells (WBCs), and platelets (Fig. S1B), and no cytologic abnormalities in either the WBCs or RBCs were detected in the PB smears at 13 months and at 21 months after BMT (Fig. S1D). Of the 10 EVI1-(1+6Mut)-positive animals, only one mouse showed signs of cytopenia that remained mild and did not affect survival.

EVI1-(1+6Mut) Does Not Induce Significant Pathologic Abnormalities in Hematopoietic Organs.

Relative to control mice, the splenic red pulp of the EVI1-positive mice was infiltrated by extensive extramedullary hematopoiesis often with a megakaryocytic prominence (Fig. S2 A and A-1), and the BM showed erythroid hyperplasia, dyserythropoiesis, and increased numbers of hypolobated dyspoietic megakaryocytes as is often seen in the BM of EVI1-positive MDS patients (5) (Fig. S2 D and H). No increase in blasts was morphologically detected in the BM of the EVI1-positive mice. At the time when the last three surviving EVI1-positive mice had to be killed because of their disease, we also killed three healthy EVI1-(1+6Mut)-positive animals for comparison. The dysplastic features of the EVI1 mice were absent in the EVI1-(1+6Mut)-positive animals, which, like the control mice, had only sparse extramedullary hematopoiesis in the spleen (Fig. S2 B and C, and enlarged B-1 and C-1), suggesting that in this organ, the disruption of EVI1–Gata1 interaction offsets most of the EVI1-induced morphologic abnormalities. Furthermore, there was no morphologic evidence of dysplasia in the BM of the EVI1-(1+6Mut) mice (Fig. S2 E and K). Even though there was still morphologic evidence of a mild erythroid hyperplasia, the overall features of their BMs were morphologically similar to controls (Fig. S2 G and L). To quantify the erythroid cell population, we dissected the spleens and BM of the EVI1 and EVI1-(1+6Mut) and analyzed the cells by FACS using the erythroid-specific marker, Ter119. In agreement with our previous report (2), the number of EVI1-positive Ter119+ cells was 3-fold higher than in the control cells. In the EVI1-(1+6Mut) mice, the number of Ter119+ cells was intermediate between the EVI1 animals and the control mice (Fig. S3), concordant with our morphologic impressions of a mild erythroid hyperplasia. As shown in Table S1, the mild erythroid hyperplasia persisted in the EVI1-(1+6Mut) mice until the experiment was terminated at 21 months after BMT.

EVI1-(1+6Mut)-Positive BM Cells Respond to Epo and GM-CSF.

To determine whether EVI1-(1+6Mut)-positive BM progenitors differentiate in vitro like normal cells, we purified Lin- cells from the three groups of mice 12 months after BMT, plated equal number of cells in semisolid medium containing Epo, and scored the colonies 7 days later (Fig. 1A). The results confirm that the EVI1-positive cells are unable to generate colonies in response to Epo (dark gray bar). In contrast, EVI1-(1+6Mut)-positive cells (black bars) respond to Epo and form colonies that are comparable in number and size with those of age-matched control cells (light gray bars) not only 12 months after BMT but also at the time the study was terminated 21 months after BMT. Because by this latter time point, all of the EVI1-positive mice had succumbed to their disease, the comparison is limited to EVI1-(1+6Mut)-positive and control progenitors. Cytospin preparations of the EVI1-(1+6Mut) colonies (Fig. 1B, lane 3) showed progressive differentiation similar to controls (Fig. 1B, lane 1) and none of the dyserythropoietic features previously observed in the EVI1-positive cells (Fig. 1B, lane 2). We quantified the expression of two Gata1-dependent genes, EpoR and c-Mpl, in the BM of 6 EVI1-(1+6Mut), 7 EVI1, and 4 control mice. The results summarized in Fig. 1C confirm that these genes are down-regulated in the EVI1 mice (lane 2). In contrast, their expression in the EVI1-(1+6Mut) mice (lane 3) either at 12 months (lane 3, empty circles) or at 21 months after BMT (lane 3, black circles) is not significantly different from that of the control animals (lane 1). We reported (2) that at time of death the EVI1-positive BM cells were unable to respond to G-CSF and GM-CSF. To determine whether BM cells of EVI1-(1+6Mut)-positive mice respond to these cytokines, we isolated Lin- cells from EVI1, EVI1-(1+6Mut), and vector mice 12 months after BMT, plated equal number of cells, and compared their ability to respond to GM-CSF. As shown in Fig. 1D, the cells isolated from EVI1-(1+6Mut) mice 12 months after BMT form colonies comparable with controls in response to GM-CSF. In contrast, cells isolated from moribund EVI1 mice form a significantly lower number of very small colonies (Fig. 1D). When the study was terminated at 21 months after BMT, the EVI1-(1+6Mut) cells maintained colony forming potential (Fig. 1D), albeit lower than control cells. These in vivo results confirm the repressive role played by EVI1 on the function of Gata1, leading to the development of fatal anemia in the EVI1 mice. Because it is known that Gata1 primarily regulates erythroid and megakaryocytic differentiation rather than proliferation of myeloid lineages, the finding of an absence of BM failure in the mutant EVI1 mice suggested to us that the general inability to respond to growth factors and the resulting BM failure in EVI1 mice could be due to defects in pathways regulating cell cycling and/or self-renewal unrelated to Gata1.

Fig. 1.

EVI1-(1+6Mut)-BM cells have a normal response to Epo and GM-CSF. (A) Equal numbers of Lin- cells isolated from EVI1-, EVI1-(1+6Mut)-, and control mice were plated in methylcellulose in presence of Epo. EVI1 blocks the response to this cytokine and very few small colonies are generated (lane 2). In contrast, the response to Epo in EVI1-(1+6Mut)-positive BM cells isolated 12 (lane 3) and 21 months after BMT (lane 5) is >80% of the control response (lanes 1 and 4). The number of colonies is given as a percent of the control colonies taken arbitrarily as 100 and represents the average of four different experiments. (B) Cell cytospins after Epo stimulation. Dysplastic aspects including trinucleation are evident in EVI1-positive cells (Middle). The dysplastic elements are not observed in EVI1-(1+6Mut)-positive cells (Bottom) and control cells (Top). (C) Expression quantification by Q-PCR of EpoR (Upper) and c-Mpl (Lower) reveals their repression in 7 EVI1-positive mice (lane 2) compared with 4 controls (lane 1). EpoR and c-Mpl are not repressed in EVI1-(1+6Mut)-positive mice 12 months (solid circles) or 21 months after BMT (empty circles). The results are plotted as a ratio between EpoR or c-Mpl and Abl1 multiplied by 100. (D) Lin- cells as described in A were plated in presence of GM-CSF.

EVI1 Up-Regulates Genes Associated with Cell Division, DNA Replication, and Accelerates S₁ Phase Entry by Shortening of G₁ Phase.

To determine whether EVI1 alters cell cycle dynamics, we examined the number of cells/colonies formed in vitro from Lin-cells infected with EVI1, EVI1-(1+6Mut), or empty retroviral vector at time of BMT. We found that each EVI1-positive colony (Fig. 2A, dark gray bars) contains approximately twice as many cells as the control (Fig. 2A, light gray bars) or the EVI1-(1+6Mut)-colonies (Fig. 2A, black bars). This increase in cell number could be explained by a repression of apoptosis or an up-regulation of cell division. To distinguish between these two effects, we used flow cytometry to quantify the number of cells in G₁, S, or G₂/M phase. The results (Fig. 2B) show that EVI1 reduces the numbers of cells in G₁ by 15% (from 60 to 45%) and increases the number of cells in S phase by 13% (from 30 to 43%) and in G₂/M phase by 2% (from 9 to 11%) relative to EVI1-(1+6Mut) and control cells. No significant difference was noted between the EVI1-(1+6Mut) cells and the control cells. To identify genes affected by EVI1 in primary cells, we used the Oligo GEArray, which examines 112 genes associated with the regulation of G₁, S, and G₂/M. Because there was no significant difference in the proliferation rate between the control cells and the EVI1-(1+6Mut) cells (Fig. 2 A and B), the comparison was limited to EVI1 and control cells. Total RNA from EVI1-positive or control BM cells was isolated, purified and labeled, and the RNA probes were hybridized to the microarrays. Nineteen of the 112 genes in the microarray were up-regulated in EVI1-positive BM progenitors compared with the control cells (Table S2). These genes could be clearly grouped in sets with specific functions. Cyclins and CDKs were up-regulated, as well as genes involved in DNA replication (MCM complex genes), or in mitotic spindle checkpoints. Such up-regulation was not observed in the BM of moribund EVI1-positive mice (Table S3). The gene that appeared to be most strongly deregulated was Cyclin D3, and its up-regulation was confirmed by Western blot (Fig. 3A Left, lane 2). In contrast to EVI1-positive cells, there was no increase of Cyclin D3 level in EVI1-(1+6Mut) cells (Fig. 3A Left, lane 3). These latter two results were confirmed in the hematopoietic cell line 32Dcl3. It has been reported that EVI1 moderately increases production of Cyclin A and Cdk2 in the fibroblast rat cell line Rat 1 (6). Although we detected the increase of Cdk2, no significant change in Cyclin A was observed in BM cells. Overall, our results indicate that Cyclin D3 is the gene most affected by EVI1 in primary BM cells and, as would be expected, Rb was hyperphosphorylated after expression of EVI1 compared with control cells (Fig. 3B, lane 4).

Fig. 2.

EVI1 accelerates cell division by shortening of G₁ phase. (A) Number of cells per colony generated by Lin- cells cultured in GM-CSF for 7 days. EVI1-positive colonies (dark gray bar) contain a higher number of cells (25,000) than control colonies (12,500 cells, light gray bar) or EVI1-(1+6Mut) colonies (14,000 cells, black bar). (B) Cell cycle analyses show a 15% reduction in the number of cells in G₁ and an increase of the number of cells in S and G₂/M phase (respectively 13% and 2%) when EVI1 is expressed (dark gray bars). The percentage of EVI1-(1+6Mut)-positive cells in G₁, S, and G₂/M phase (black bars) is not significantly different from the controls (light gray bars).

Fig. 3.

EVI1 up-regulates proteins that regulate cell cycling of BM cells. (A) Western blot analysis of control primary BM cells (lane 1), EVI1-positive cells (lane 2), or EVI1-(1+6Mut)-positive cells (lane 3) shows an increase in the level of Cyclin D3 and Cdk2. (B) Western blotting with Rb antibody reveals a higher level of phosphorylated Rb when EVI1 is expressed (lane 4) compared with vector-transfected control cells (lane 2). Lane 1 and 3 indicate the size of dephosphorylated Rb, taken as marker. V and E indicate vector cells and EVI1 cells, respectively.

EVI1 Up-Regulates Genes Associated with Self-Renewal at Time of BMT but Not at Time of Death.

To determine whether EVI1 deregulates hematopoietic stem cell (HSC)/early progenitor cell self-renewal, we quantified the expression of Bmi1, Gata2, Hoxb4, and Gfi-1 by quantitative real-time PCR (Q-PCR) at time of BMT. These genes have been described as positive regulators of self-renewal (7–10). The results (Fig. 4A, lane 2 for Bmi1 and Table S4 for Gata2, Hoxb4, and Gfi-1) show that the transcription of most of these genes is significantly increased in BM cells that express EVI1. The up-regulation of Bmi1 was confirmed by Western blotting. In contrast, the expression of this gene in EVI1-(1+6Mut)-positive cells was comparable with that of the control cells (Fig. 4A, lanes 1 and 3). We thought that the inability of EVI1-positive cells to form colonies at time of death could be related to a loss of self-renewal. Because sorted KSL cells were too scarce in the BM of the moribund animals, we were forced to use the remaining small amounts of total BM cells. Consequently, our analysis was limited to one gene and, because of the restricted amounts of cells, could only be followed by one technique, Q-PCR. We chose to analyze Bmi1 because this gene is perhaps the most critical regulator of self-renewal, proliferation, and senescence in BM (7) and in other tissues (11–13). The results (Fig. 4B) show that the expression of this gene was constant in the EVI1-(1+6Mut) mice at 12 months (black circles) and at 21 months (white circles) after BMT, which was similar to age-matched controls (gray circles). In contrast, Bmi1 was strongly reduced in the EVI1 mice at time of death (Fig. 4B, gray squares, lane 2), suggesting that at time of death there is a significant decrease in the number of HSC in these mice compared with age-matched EVI1-(1+6Mut) mice and control mice.

Fig. 4.

Increase of Bmi1 in EVI1-positive BM cells at time of BMT but significant reduction of Bmi1 levels occurring at time of death. (A) Q-PCR expression analysis shows that Bmi1 is strongly up-regulated in EVI1-positive BM cells at time of BMT (dark gray squares). At this time, Bmi1 is expressed approximately at the same level in vector mice (light gray circles) and EVI1-(1+6Mut)-positive mice (black circles). (B) At time of death, EVI1-positive BM cells express a significantly lower level of Bmi1. In EVI1-(1+6Mut)-positive mice, the expression of this gene is similar to the vector (light gray circles) both at 12 months (black circles) and 21 months (empty circles) after BMT.

EVI1 Deregulates miRNAs in Primary BM Cells.

MicroRNAs are a new class of gene regulators with pleiotropic actions. They are small noncoding RNAs, 19–25 nucleotides (nt) in length, that up- and down-regulate gene expression during cell development, differentiation, and apoptosis, and several miRNAs control the progression of the cell cycle. Because of the large number of genes deregulated by EVI1 in BM cells (Table S2), we thought that EVI1 might alter the expression of miRNAs. To prove or disprove our hypothesis, we used the TaqMan Rodent MicroRNA Array A (Applied Biosystems). The screening results identified several miRNAs that were up-regulated or down-regulated by EVI1 in BM cells at time of BMT (Table S5). Among them, miRNA-124 was repressed ≈19-fold by EVI1 in BM cells. We used Q-PCR to verify the results in primary BM cells, as well as in the 32Dcl3 cell line. The results confirm that the expression of miRNA-124 is repressed by EVI1 almost 8-fold in BM cells (Fig. 5A, left side, dark gray bar). Strikingly, EVI1-(1+6Mut), which does not alter cell division, did not have any significant effect on miRNA-124 expression (Fig. 5A, left side, black bar). Down-regulation of miRNA-124 by EVI1 was also observed in 32Dcl3 cells (Fig. 5A, right side). We note here that the expression of miRNA-124 has been associated with the regulation of cell cycling (14) and differentiation (15). It was reported that 32Dcl3 cells expressing EVI1 do not respond to G-CSF and fail to survive in culture (16). To determine whether the ectopic expression of miRNA-124 in EVI1-positive 32Dcl3 cells can re-establish G-CSF response and survival, we cloned miRNA-124, generated two independent clones of EVI1-positive/miRNA-124-positive 32Dcl3 cells, and confirmed the expression of EVI1 in one clone by Western blotting (Fig. 5B). The survival and response to G-CSF of the clones were analyzed. As shown in Fig. 5 C and D, the expression of miRNA-124 is sufficient to re-establish G-CSF response in EVI1-positive 32Dcl3.

Fig. 5.

EVI1 represses miRNA-124. Ectopic expression of miRNA-124 in EVI1-positive 32Dcl3 cells allows G-CSF-dependent differentiation. (A) Q-PCR expression of miRNA-124 is taken arbitrarily as 1 for vector-infected BM and 32Dcl3 cells (light gray bars). The expression of miRNA-124 is drastically reduced when EVI1 (dark gray bars) but not when EVI1-(1+6Mut) (black bars) is expressed. (B) Western blotting of 32Dcl3 cells confirms the expression of EVI1 in the infected cells. (C) 32Dcl3 cells infected with the empty vector (orange line) differentiate in response to G-CSF. When EVI1 is expressed, the cells are unable to respond and die (brown line). Two independent clones of cells expressing both EVI1 and miRNA-124 survive and differentiate in G-CSF (blue and gray lines). (D) Differentiation occurs in the presence of G-CSF in 32Dcl3 cells infected with EVI1 and miRNA-124 (clone#1/clone#2), whereas EVI1-positive 32Dcl3 cells treated with G-CSF undergo cell death. Vector-transfected 32Dcl3 cells are shown in Upper Left as a comparison.

Ectopic Expression of miRNA-124 in EVI1-Positive BM Cells Abrogates the Up-Regulation of Cell Division and Self-Renewal.

Consecutive plating assays carried out by our group have shown that EVI1 enhances self-renewal and enables primary hematopoietic cells to form colonies in vitro for a time twice as long as the control cells before terminal differentiation. These results are murine-strain independent (17) and are not observed when EVI1-(1+6 Mut) is expressed in the cells (Fig. S4). To evaluate the effect of miRNA-124 in the differentiation of murine BM cells, Lin- cells were infected with the empty vector, EVI1, EVI1-(1+6Mut), or both EVI1 and miRNA-124 and cultured in vitro. The results (Fig. 6 and Fig. S4) verified that when EVI1 is expressed the number of colonies and the replating potential are doubled compared with the empty vector (Fig. 6A, dark gray bars) and confirmed that EVI1-(1+6Mut), which does not induce MDS in the mice, is unable to prolong colony growth (black bars). Most importantly, we found that coexpression of miRNA-124 with EVI1 (Fig. 6A, white bars) normalizes the replating potential and the differentiation of the BM cells both at 14 days and 21 days of culture. We have established here that miRNA-124 is associated with up-regulation of Cyclin D3 and Bmi1 in EVI1-positive cells at time of transplantation (Figs. 3 and 4). To determine whether miRNA-124 can normalize the expression of these two genes, primary Lin- cells were infected with the empty vector, EVI1 alone, EVI1 with a second vector expressing miRNA-124 or EVI1-(1+6Mut). The results (Fig. 7) reveal that Cyclin D3 accumulates to similar levels in vector cells and in cells that express EVI1-(1+6Mut) (lanes 1 and 3), whereas EVI1 increases Cyclin D3 and Bmi1 (lane 2). Most importantly, their expression is reduced after miRNA-124 is coexpressed with EVI1 (lane 4). These data strongly suggest that a functional regulatory link exists between the cell division and self-renewal pathways deregulated by EVI1, the down-regulation of miRNA-124, and the up-regulation of Cyclin D3 and Bmi1.

Fig. 6.

Ectopic expression of miRNA-124 normalizes cell replication, colony formation, and differentiation of EVI1-positive BM cells. (A) The enhanced replating potential and colony formation of EVI1-positive BM cells are reduced after ectopic expression of miRNA-124 (white bars). Vector (light gray bars) and EVI1-(1+6Mut) (black bars) cells are shown for comparison. (B) Phenotype of cells after 14 and 21 days in methylcellulose culture shows a more differentiated morphologic state in EVI1+miRNA-124 infected BM cells than in those solely infected with EVI1. BM cells containing EVI1+miRNA-124 are morphologically comparable with those cells infected with empty vector or EVI1-(1+6 Mut).

Fig. 7.

miRNA-124 represses the up-regulation of Bmi1 and Cyclin D3 in EVI1-positive BM cells. Western blot analysis of BM cells infected with a retrovirus expressing EVI1 alone (lane 2) or EVI1 and miRNA-124 (lane 4) demonstrates that miRNA-124 significantly reduces the expression of Bmi1 and Cycin D3 to the level observed in control (lane 1) and EVI1-(1+6Mut) (lane 3) cells.

EVI1, but Not by EVI1-(1+6Mut), Induces a Striking Increase of CpG Methylation in the Promoter of miRNA-124-3.

Several groups have shown that miRNAs are altered in human malignancies and can function as tumor suppressor genes or oncogenes through expression regulation of their target genes (15). Some miRNAs harboring CpG islands undergo methylation-mediated silencing, and there is recent evidence supporting a role for miRNA's as targets of aberrant mechanisms of DNA hypermethylation (15). We thought that the down-regulation of miRNA-124 could be associated with inappropriate DNA methylation. There are three known genes of miRNA-124 that are closely related homologs between man and mouse. In the mouse, they are located on chromosome bands 2H4, 3A1, and 14D1. Analysis of the three genes indicated that miRNA-124-3 gene harbors the largest embedded number of CpG dinucleotides, 24 CpG within a stretch of 190 base pair (bp) including 9 bp of the stem-loop start site (Fig. 8A). We used bisulfite DNA sequencing to determine whether EVI1 alters the CpG methylation within this region in Lin- BM cells cultured in semisolid medium for 2 weeks. The methylation of the 24 CpGs, summarized in Fig. 8B, shows that the number of methylated CpG (black boxes) was low for empty vector and EVI1-(1+6Mut) cells (Fig. 8B Middle and Lower). In contrast, there was a dramatic increase in CpG methylation when EVI1 was expressed (Fig. 8B Upper). The methylation was mostly centered around two clusters, which are designated by clusters I and II (Fig. 8B). The nt sequence of the clusters includes nt −140/−123 for cluster I (Fig. 8A, green block), and nt −40/−15 for cluster II (Fig. 8A, purple block). The nt designations are arbitrarily numbered with respect to the stem-loop start site (Fig. 8A, black arrow).

Fig. 8.

EVI1 induces miRNA-124-3 methylation in murine BM cells. (A) DNA sequence of 190 bp analyzed by bisulfite sequencing. I (green) and II (purple) indicate the regions containing the clusters of methylated CpG (bold italics). The black arrow shows the starting site of the stem-loop (sequence in blue). (B) There are 24 CpG dinucleotides within the sequenced stretch of DNA indicated by black or empty cells. Each single CpG is either methylated (black box) or unmethylated (empty box). Two clusters (I and II) of significant methylation are observed when EVI1 is expressed (Top). In contrast, control cells expressing the empty vector (Middle) or the EVI1-(1+6Mut) (Bottom) have limited numbers of methylated CpGs.

Discussion

This work is based on and expands two reports that we published showing that inappropriate EVI1 expression in the mouse was associated with a fatal MDS-like disease (2) and that the interaction between EVI1 and Gata1 disrupts erythropoiesis in vitro (4). Here, we set out to determine whether the disruption of EVI1–Gata1 interaction could alleviate the MDS phenotype in vivo. Our serial complete blood counts, morphologic, cell cycle, and plating studies clearly indicate that the disruption of EVI1–Gata1 interaction blocks the onset of defects associated with development of MDS in vivo. The complete disappearance of MDS-like features, especially cytopenia and BM failure, in our Gata1-independent EVI1 mutant mice was unexpected and raised the second question that we specifically addressed here: the nature of the key players disrupting cell-sustaining pathways in EVI1-positive cells, ultimately leading to BM failure. EVI1 is a very complex protein and little is known about its normal functions. It is required for embryonic cell growth and development and its targeted disruption affects the development of virtually all of the developing systems leading to the death of a significantly hypocellular embryo at day 8 (18). The lack of a well-defined phenotype in embryos and the widespread hypocellularity in multiple embryonic organs suggested that EVI1 probably does not have tissue-specific functions, but rather that EVI1 could be required for cell proliferation at least during embryogenesis (18). This role of EVI1 as a positive regulator of cell growth is confirmed in the studies reported here with BM cells in which there a sustained up-regulation of cell proliferation after EVI1 expression. In a murine system, EVI1 does not however induce transformation in vitro and all of the BM cells eventually differentiate, albeit several weeks later than the control cells. Therefore, to support a prolonged ability to generate robust colonies, EVI1 must be able to skew symmetric divisions of a HSC/progenitor cell in favor of asymmetric divisions that preserve a large pool of HSC/progenitor cells. As we show here, these effects are triggered by the ability of EVI1 to up-regulate genes that control cell cycling and self-renewal. The actions of EVI1 on the primary cells used for BMT are strikingly different from those observed in mice a year later, at time of their death, in which ineffective hematopoiesis arising from impaired progenitor responsiveness to normal signals promotes an accelerated loss of progenitors and their progeny due to apoptosis (2). What is startling is not our finding that with time a protein in a BM cell can induce apoptosis, but that when first expressed in the primary cell—at time of BMT—this same protein has opposite effects, which are a stronger response to cytokines in vitro, an enhanced self-renewal, and an accelerated cell division. These contradictory responses of EVI1, which are supported by the results of flow cytometry and quantification of genes associated with the cell cycle, are limited and reflect intrinsic properties of the primary cells as shown with parallel replating assays of C57BL/6J and DBA2 cells. In the absence of transformation in vivo, the dual, but opposite nature of EVI1 can be explained by its ability to promote faster cell divisions of a primary HSC with the consequence of exhausting a finite self-renewal potential over a given period. This latter explanation is in fact supported by our observations of the lowering of Bmi1 expression in the older BM of the moribund EVI1-positive mice compared with age-matched control animals. Of course, one could argue that the significant decrease of Bmi1 actually reflects the number of HSC rather than its level of expression in each cell. This argument is valid and true, as demonstrated by the very low colony-forming potential of EVI1-positive BM cells around the time of death; however, at the same time, it also supports our suggestion that by accelerating cell division, EVI1 depletes the number of HSC with time. The hypothesis we propose to explain BM failure in EVI1-positive mice is based on our data showing that EVI1 up-regulates cell division not only in differentiating cells but also in their progenitors. This effect, when occurring in an HSC, favors self-renewal and leads to the early depletion of the self-renewal potential of the EVI1-positive HCS. Direct evidence to support this hypothesis in EVI1-positive MDS patients is lacking, however three separate groups found that in mice (19), in monkeys (20), and in two young adults with X-CGD treated by gene therapy (21), the insertion of a retrovirus in the MDS1/EVI1 locus leads to the emergence of dominant hematopoietic clones with improved self-renewal and proliferation. More recently, one of the three groups (21, 22) showed that the two patients in whom the retrovirus had integrated in the EVI1 locus developed BM failure and MDS ≈2 years after the gene therapy. Taken together, these reports would indicate that the mechanism we propose could very likely apply to MDS patients.

A question that remains to be answered is how EVI1 can control so many different genes/pathways. The data presented in this report suggest that EVI1 modulates the expression of several miRNAs, and, among them, the down-regulation of miRNA-124 appears to be critical. As reported by others, we find that the down-regulation of miRNA-124 occurs by methylation of CpG dinucleotides upstream of its transcription's starting site. The role of miRNA-124 is confirmed by its coexpression in primary cells and in the 32Dcl3 cell line, in which the effects of EVI1 on self-renewal, cell cycling, and survival are reversed. The association between the expression of EVI1 and repression of miRNA-124 is clearly supported by the CpG methylation results that we provide. It is important to note that the mutant EVI1-(1+6Mut), which neither induces MDS in mice nor affects self-renewal or cell cycling, has virtually no effect on the expression/methylation of miRNA-124. Therefore, it is likely that an interaction could occur between an unknown factor and zinc finger motifs 1 and 6 of EVI1. At this time, one can only speculate on the mechanism by which EVI1 activates CpG methylation, and one such speculation, is that EVI1 could affect the activity or expression of DNA methylases directly or indirectly.

In conclusion, the data presented here suggest that EVI1 is capable of inducing an MDS-like disease in the mouse by interfering with two distinct pathways regulated by separate factors. The first pathway leads to a severe impairment of erythropoiesis by interaction with and functional inhibition of Gata1, whereas the other one produces alterations of cell cycling and self-renewal by interfering with miRNA-124 expression as a result of interaction with factors that are unknown at this time. There is another aspect of our results that perhaps has an important consequence from a clinical perspective. Both pathways altered by EVI1 require the integrity of two zinc finger motifs, and our data show that their disruption is sufficient to virtually abrogate the features of MDS in the mice. Given the fact that patients suffering from MDS, AML, and chronic myelogenous leukemia in blast crisis (CML-BC) with increased EVI1 gene expression have dismal clinical outcomes, screening of small molecule libraries for potential compounds that interfere with the activity of the two zinc finger motifs may provide valuable drug lead discoveries for the treatment of these diseases.

Material and Methods

Cloning.

The MSCV-EVI1 and the MSCV-EVI1(1+6Mut) have been described (4). To express miR-124, a 136-bp PCR fragment including miR-124 was prepared by PCR with primers 5′-GGAAGATCTCCTTCCTTCTTCCTTCCTCA-3′ and 5′-CCCCAAGCTTCCTCGTGGACCCAAGGTG-3′ and cloned into the pSuper.retro.puro vector (OligoEngine).

Cell Culture and BM Transplantation.

The isolation and infection of Lin- cells and the reconstitution of irradiated syngeneic recipient mice have been described (2, 4). All of the animal studies were performed in accordance with the guidelines of the Animal Care Committee of the University of Illinois at Chicago.

PB Cells Count and Histological Examination.

Starting at 4 months after BMT, the PB counts of recipient mice were recorded with a Coulter Counter Z1 (Beckman Coulter) every 2 weeks for the first 6 months and then once a week until the animals were killed as described (2). PB and BM smears and organ preparations were carried out according to described methods (2).

Rb Phopshorylation.

Immunocomplexes on beads were washed and equilibrated with 100 mM Mes with 1 mM PMSF. Each sample was divided in half (control and experimental sample). Potato acid phosphatase (0.5 U, Boehringer) was added and incubated for 15 min at 37 °C.

Flow Cytometry Assays and Cell Cycle Profiling.

Cells were collected from BM or spleen cells and analyzed with a FACS Vantage flow cytometer (BD Biosciences) as described (2). Cell cycle profiling has been reported (2).

Q-PCR.

Total RNA extracted from BM cells infected with EVI1, EVI1-(1+6Mut), or the empty vector was transcribed into cDNA and analyzed with an iCycler-iQ version 3.0 software (Bio-Rad) as described (2).

Oligo GEArray.

We used the mouse cell cycle Oligo-GEArray (SABiosciences) according to the manufacturer's instructions.

Colony-Formation Assay.

Lin- BM cells were isolated and infected and the colonies were cultured and counted as described (4).

DNA Bisulfite Sequencing.

Lin- BM cells infected with EVI1, EVI1-(1+6Mut), or the empty vector were cultured in vitro for 14 days. Genomic DNA was converted with EpiTect Bisulfite Kit (Qiagen). An aliquot of bisulfite-converted DNA was amplified with methylation-insensitive primers: forward: 5′-GTAGGAAYGTTTYGAGGGATTTGTT-3′; reverse: 5′-CTCTCTTAACATTCACCRCRTACCTTA-3′, cloned into pCR4-TOPO vector (Invitrogen), and sequenced.