AML1/RUNX1 increases during G1 to S cell cycle progression independent of cytokine-dependent phosphorylation and induces cyclin D3 gene expression.

AML1/RUNX1, a member of the core binding factor (CBF) family stimulates myelopoiesis and lymphopoiesis by activating lineage-specific genes. In addition, AML1 induces S phase entry in 32Dcl3 myeloid or Ba/F3 lymphoid cells via transactivation. We now found that AML1 levels are regulated during the cell cycle. 32Dcl3 and Ba/F3 cell cycle fractions were prepared using elutriation. Western blotting and a gel shift/supershift assay demonstrated that endogenous CBF DNA binding and AML1 levels were increased 2-4-fold in S and G(2)/M phase cells compared with G(1) cells. In addition, G(1) arrest induced by mimosine reduced AML1 protein levels. In contrast, AML1 RNA did not vary during cell cycle progression relative to actin RNA. Analysis of exogenous Myc-AML1 or AML1-ER demonstrated a significant reduction in G(1) phase cells, whereas levels of exogenous DNA binding domain alone were constant, lending support to the conclusion that regulation of AML1 protein stability contributes to cell cycle variation in endogenous AML1. However, cytokine-dependent AML1 phosphorylation was independent of cell cycle phase, and an AML1 mutant lacking two ERK phosphorylation sites was still cell cycle-regulated. Inhibition of AML1 activity with the CBFbeta-SMMHC or AML1-ETO oncoproteins reduced cyclin D3 RNA expression, and AML1 bound and activated the cyclin D3 promoter. Signals stimulating G(1) to S cell cycle progression or entry into the cell cycle in immature hematopoietic cells might do so in part by inducing AML1 expression, and mutations altering pathways regulating variation in AML1 stability potentially contribute to leukemic transformation.

Core binding factor (CBF) 1 is a family of transcription factors containing one of three CBF␣ subunits, RUNX1/AML1, AML2, or AML3, and a CBF␤ subunit (1). The CBF␣ subunits bind a common DNA consensus site via the N-terminal Runt domain, which also mediates heterodimerization with CBF␤ (2, 3).
CBF␤ does not contact DNA, but increases the DNA affinity of the ␣-subunits (4,5). CBF activities are reduced in 30% of acute myeloid leukemia cases, due either to AML1 point mutations, AML1 gene deletion, or, most often, chromosomal abnormalities involving genes encoding AML1 or CBF␤ (1).
Based on the conclusion that AML1 stimulates the G 1 to S transition, we sought to determine whether AML1 levels are themselves regulated during the cell cycle. 32Dcl3 and Ba/F3 cell cycle fractions were prepared using elutriation, a method that avoids the cytotoxicity of chemical synchronizing agents. Endogenous or exogenous AML1 levels increased during G 1 to S cell cycle progression, as did AML1 DNA binding activity. An AML1 variant lacking two ERK phosphorylation sites retained cell cycle variation, and IL-3 induced phosphorylation of endog-enous AML1 independent of cell cycle phase. In addition, inhibition of AML1 repressed cyclin D3 mRNA expression, and AML1 bound and activated the cyclin D3 gene.
Plasmids-The human AML1B cDNA, the longest splice variant of AML1 (32), was ligated into the polylinker of pMTCB6 (18) to generate pMT-AML1. pMT-Myc-AML1 was generated by inserting an oligonucleotide encoding Kozak's rules for translation initiation followed by a methionine and a Myc tag upstream of the AML1 cDNA. The oligonucleotide sense strand, with a KpnI-compatible 3Ј-end was 5Ј-CATCGA-TGCCGCCACCATGGAGCAGAAGCTCATCTCCGAGGAAGATCTCA-GTAC. The sequence encoding the initiating Met and 11-residue Myc tag is underlined. Ligation into the polylinker upstream of the AML1 cDNA also inserted five additional residues, STEFA, between the tag and the AML1 cDNA. pMT-Myc-AML1-(1-217) was generated by 3Јdeletion to a SmaI site, with insertion of a stop codon. A HindIII/BssHII segment of the AML1(S276A/S293A) cDNA, kindly provided by H. Hirai, was used to replace a similar segment in pMT-Myc-AML1 to generate pMT-Myc-AML1(S276A/S293A). pBabePuro-AML1-ER and pBabePuro-AML1 (86 -217)-ER have been described (25). The numbering of these mutants reflects the AML1B sequence (32). pD3(Ϫ447)-Luc was prepared by ligating an XhoI/BamHI fragment from pD3GH (33) into p19LUC. pD3(Ϫ447)mAML1-LUC was prepared by replacing the AML1 site at Ϫ383, 5Ј-GACCACA-3Ј, with 5Ј-GCTAGCA-3Ј, by ligating two set of annealed oligonucleotides between the XhoI site at Ϫ447 and the SfoI site at Ϫ333. The ligation product was verified by DNA sequencing.
Elutriation and Fluorescence-activated Cell Sorting (FACS) Analysis-100 -200 ϫ 10 6 cells were washed and resuspended in 10 ml of elutriation medium (150 mM NaCl, 167 mM glucose, 0.3 mM EDTA, 1% bovine serum albumin). The cells were then injected into a JE-6B rotor and spun at 2000 rpm at 25°C. The flow rate of elutriation medium was increased gradually and cell cycle fractions collected. Fractionation was monitored in each experiment by incubation of 5 ϫ 10 5 cells with 11 g/ml Hoechst 33258 dye in 0.7% Nonidet P-40, 4.7% formaldehyde followed by FACS analysis.
Western Blotting, Gel Shift Assay, and Northern Blotting-Total cellular protein extracts were prepared using Laemmli sample buffer and subjected to SDS-polyacrylamide gel electrophoresis and Western blotting as described (18). Extract corresponding to 1.5 ϫ 10 6 cells were loaded in each lane. AML1 antiserum was kindly provided by H. Drabkin, CBF␤ antiserum was kindly provided by N. Speck, and ER␣ and D3 antisera (MC-ER and C-16, Santa Cruz Biotechnology), c-Myc monoclonal antibody (9E10, Roche Applied Science), and cyclin E antiserum (Upstate Biotechnology) were obtained commercially. Equivalent amounts of extract were electrophoresed and subjected to Coomassie Blue dye staining to assess total protein content of each sample.
Total cellular RNAs were prepared using Trizol reagent (Invitrogen) or RNAeasy (Qiagen) per the manufacturer's instructions. The RNAs, 10 g per lane, were subjected to Northern blotting as described (31). Murine cyclin D1, D2, and D3 cDNAs were kindly provided by C. Sherr and the murine AML1 cDNA by Y. Ito (2). Densitometric analysis was carried out with the NIH Image 1.62 program.

RESULTS
Endogenous CBF DNA Binding and AML1 Increase During G 1 to S Progression-We employed counterflow elutriation to fractionate 32Dcl3 and Ba/F3 cells into cell cycle fractions based on density. In this procedure, the flow of the elutriation media opposes the centrifugal force on the cells, establishing an equilibrium. As the flow rate is increased, denser cells come into position to exit the elutriation chamber. Use of elutriation allowed us to study cells in active, physiologic cell cycle, which did not prove possible by synchronizing cells followed by release of the chemical block as 32Dcl3 cells treated in this manner underwent apoptosis rather than continued cell cycle progression (not shown). A typical cell cycle fractionation profile and the proportion of cells in each fraction are shown ( Fig.  1). Fr 18 -20 contains G 1 phase cells, Fr 21-24 contains late G 1 /early S cells, Fr 25-30 contains mainly S phase cells, and Fr 31-34 mainly G 2 /M phase cells. Cell cycle fractionation was monitored in each experiment. In some experiments the cells began to elute at 20 ml/min rather than 18 ml/min, as indicated in each figure.
AML1 is expressed entirely in the cell nucleus (34). We first employed a gel shift assay to assess CBF DNA binding activity during the cell cycle, utilizing an equivalent amount of nuclear protein per sample (Fig. 2, A and B, left panels). Based on densitometric analysis, CBF DNA binding activity increased almost 4-fold as 32Dcl3 or Ba/F3 cells progressed from G 1 to S. Of note, micrograms of nuclear protein per cell were not significantly different between the G 1 and S phase samples. This pattern of CBF gel shift activity was reproduced in four separate experiments. To confirm that the indicated bands contain members of the CBF family, the peak fractions for each cell line were subjected to competition with increasing amounts of unlabelled wild-type CBF oligonucleotide or with an oligonucleotide carrying clustered point mutations in the CBF binding site (Fig. 2C). Specific competition was observed with the CBF complexes. To resolve AML1-specific DNA from among the CBF binding activity, equal amounts of the 32Dcl3 nuclear extracts were subjected to a supershift assay with normal rabbit serum, AML1 antiserum, or AML1 antiserum in the presence of its specific peptide ( Fig. 2A, right panel). A specific supershift band was obtained from each fraction (arrow) which increased 2.3-fold as cells progressed from G 1 to S. Gel shift assay indicated that Ba/F3 cells have substantially less overall CBF DNA binding activity than 32Dcl3 cells (Fig. 2B), perhaps accounting for our inability to detect a supershifted AML1 species with the Ba/F3 extracts, although the antibody did reduce the intensity of the CBF bands (Fig. 2B, right panel).
We next prepared total cellular proteins from 32Dcl3 and Ba/F3 cell cycle fractions and analyzed these for AML1 expression by Western blotting (Fig. 3, A and B). A slowly migrating nonspecific band was evident in the 32Dcl3 samples, and Coomassie Blue dye-stained protein fractions serve as a further control for protein loading. Densitometric analysis indicated that total cellular proteins increased 2-fold during the cell cycle, consistent with doubling of cell size. AML1 levels increased 2.7-or 2.6-fold from G 1 (Fr 18 -19) to S (Fr 28 -29) or G 2 /M (Fr 32-33) in the 32Dcl3 fractions, relative to total cellular proteins, and increased ϳ1.8or 1.9-fold in the same Ba/F3 fractions. Additional repetitions of this analysis with 32Dcl3 cells demonstrated a 3.0-or 3.5-fold increase (for an average of 3.0-fold) in AML1 expression during the cell cycle (see Fig. 8B). This degree of cell cycle variation is consistent with the EMSA studies for 32Dcl3 cells, but the increase in CBF DNA binding was greater than that for AML1 in Ba/F3 cells, perhaps reflecting a greater contribution of AML2 and AML3 to CBF cell cycle variation in these cells.
To determine whether the variation in AML1 protein levels during the cell cycle is paralleled by variation in AML1 RNA levels, total cellular RNAs were prepared from 32Dcl3 and Ba/F3 cell cycle fractions and 10 g of each sample was subjected to Northern blotting (Fig. 4). AML1 mRNA increased only 1.3-fold in both Ba/F3 cells and 32Dcl3 cells from early G 1 to late S phase, relative to ␤-actin mRNA. Of note, RNA per cell increased 2-fold, on average, during the cell cycle. Thus, relative to total RNA, AML1 RNA increased 2.6-fold in each line.
Comparison to ␤-actin may be more relevant, as total RNA largely represents ribosomal RNA.
AML1 Protein Levels Are Reduced in Cells Arrested in G 1 -Decreased expression of AML1 in early G 1 suggests that cells arrested in G 1 would have reduced levels of AML1 compared with proliferating cells. 32Dcl3 or Ba/F3 cells were exposed to mimosine, an agent, which induces the p27 cyclin-dependent kinase in hematopoietic cells (35). Mimosine was employed at 0.2 or 0.4 mM for 6 h, as exposure for longer times led to substantial apoptosis. At 0.4 mM, the G 1 /S ratio increased from 1.4 to 2.2 in Ba/F3 cells and from 1.1 to 1.6 in 32Dcl3 cells, with minimal apoptosis (Fig. 5, A and B). Total cellular proteins from these cultures were analyzed for endogenous AML1 expression by Western blotting (Fig. 5B). Exposure to 0.4 mM mimosine for 6 h reduced endogenous AML1 levels 1.7-fold in Ba/F3 cells and 2.5-fold in 32Dcl3 cells, relative to total protein content. For comparison, 32Dcl3 cells were transferred from IL-3 to G-CSF, which induces differentiation concomitant with a potent G 1 arrest (31). The G 1 /S ratio increased from 1.1 to 5.0 after 1 day (D1) and to 8.9 after 2 days (D2), without evident apoptosis (Fig. 5, A and B). AML1 levels were reduced 3-fold after 1 day and more than 10-fold after 2 days in G-CSF.
To determine whether reduced AML1 levels associated with G 1 arrest are due to reduced AML1 RNA levels, total cellular RNAs were prepared from 32Dcl3 cells cultured in IL-3 or in G-CSF for 1 or 2 days and from 32Dcl3 or Ba/F3 cells cultured in 0.4 mM mimosine for 6 h. These RNAs were subjected to Northern blotting for AML1 and ␤-actin (Fig. 5C). Relative to ␤-actin, AML1 RNA increased in 32Dcl3 cells in response to exposure to mimosine for 6 h or to G-CSF for 2 days and only decreased 10% in Ba/F3 cells in response to mimosine.
Exogenous AML1 Levels Also Increase During G 1 to S Progression-Based on our finding that endogenous AML1 levels are regulated during the cell cycle, we predicted that exogenous AML1 protein levels might be controlled similarly. To clearly distinguish exogenous AML1, we positioned a 17-residue tag, MEQKLISEEDLASTEFA (the Myc epitope is underlined), at FIG. 2. CBF and AML1 DNA binding activity increases during G 1 to S progression. A, nuclear extracts, 12 g of total protein per lane, from the indicated 32Dcl3 cell cycle fractions were subjected to gel shift assay with a radiolabeled AML1 binding site from the myeloperoxidase gene, WTMPO, either alone (left panel) or in the presence of 1 l of normal rabbit serum (R), AML1 antiserum (Ab), or AML1 antiserum, and 4 g of its specific peptide (P) (right panel). The relative intensity of the CBF shift or the AML1 supershift bands between fractions is shown. B, nuclear extracts from the indicated Ba/F3 fractions were subjected to gel shift assay similarly. The right panel was exposed for a shorter time. C, 32Dcl3 fraction 25-28 and Ba/F3 fraction 24 -25 were subjected to gel shift assay in the absence of competitor (Ϫ) or in the presence of 10-, 50-, and 200-fold excess of unlabelled wild-type (WT) or mutant (Mut) competitor. The two panels were exposed for similar times.  5. Endogenous AML1 levels are reduced by G 1 cell cycle arrest. A, cell cycle profiles of 32Dcl3 or Ba/F3 cells exposed to 0.4 mM mimosine (mim) for 6 h or of 32Dcl3 cells transferred from IL-3 to G-CSF for 1 or 2 days (D1, D2). B, total cellular proteins prepared from Ba/F3 or 32Dcl3 cells exposed to the indicated doses of mimosine for the indicated times were subjected to Western blotting for AML1 and to Coomassie Blue dye staining (left panels). Similar analysis was carried out on 32Dcl3 cells cultured in IL-3 or after transfer to G-CSF for 1 or 2 days (right panels). The G 1 /S ratio and the relative expression of AML1, corrected for loading, is shown for each sample. C, total cellular RNAs prepared from 32Dcl3 cells cultured in IL-3 or in G-CSF for 1 or 2 days and from 32Dcl3 and Ba/F3 cells cultured in 0.4 mM mimosine for 6 h were subjected to Northern blotting for murine AML1 and ␤-actin. Expression of AML1 RNA relative to ␤-actin in each sample is shown. the N terminus of the 480-residue AML1 cDNA just downstream of a segment containing Kozak's rules for optimum translational initiation. The myc-AML1 cDNA, lacking endogenous AML1 5Ј-or 3Ј-untranslated regions (UTRs), was then positioned downstream of the zinc-responsive metallothionein promoter in pMTCB6 to generate pMT-myc-AML1. This plasmid also contains the G418-resistance gene linked to the SV40 promoter. pMT-myc-AML1 was linearized and electroporated into 32Dcl3 cells. Two G418-resistant subclones expressing myc-AML1 in the presence of zinc chloride were identified by Western blotting. When a total cellular extract from zinctreated 32D-Myc-AML1 cells was analyzed by Western blotting using an AML1 antibody, two doublets were detected, with the lower doublet co-migrating with AML1 detected from parental 32Dcl3 cells (Fig. 6A, left panel). Presence of a doublet may reflect AML1 phosphorylation (36). The increased size of exogenous myc-AML1 is due to the 17-residue Myc tag and to the fact that exogenous human AML1 contains 27 residues at it N terminus lacking in the endogenous murine protein. Of note, the remainder of human and murine AML1 are 97% identical (32). We observed that myc-AML1 was expressed at ϳ3-fold higher level than endogenous AML1. Use of the zinc-responsive MT promoter not only avoids generation of cell lines perturbed during isolation by high basal activity of AML1 but also allows control of myc-AML1 expression via use of suboptimal zinc concentrations (Fig. 6A, right panel). Even in the absence of zinc, basal expression of myc-AML1, especially the lower band of the doublet, is evident. 75 M zinc chloride led to expression 2-3-fold below that detected in cells exposed to 100 M zinc, presumably similar to endogenous AML1 levels. 32D-Myc-AML1 cells were therefore exposed to 75 M zinc and subjected to elutriation. Cell cycle fractions were analyzed by Western blotting for myc-AML1, cyclin D3, and cyclin E (Fig. 6B). Compared with early G 1 cells (Fr 18 -20), Myc-AML1 levels increased 2.3-fold by mid-S phase (Fr 27-28) and 3.0-fold by late-S/G 2 /M phase (Fr 29 -30). Cyclin D3 and E levels increased similarly. Analysis of a second 32Dcl3 subclone demonstrated a 3.4-fold increase in exogenous Myc-AML1 during cell cycle progression (not shown). Northern blot analysis indicated that AML1 RNA increased 1.8-fold relative to ␤-actin, consistent with both transcriptional and post-transcriptional regulation of AML1 during G 1 to S progression (Fig. 6C).
To evaluate the effect of cell cycle phase on the expression of exogenous AML1 by an independent strategy we utilized 32Dcl3 cells expressing AML1-ER from the MMLV retroviral LTR in pBabePuro (20). Western blot analysis of cell cycle fractions demonstrated that AML1-ER increased 1.9 -2.3-fold by S phase (Fr 25-28 and Fr 29 -31) and 4.5-fold by G 2 /M phase (Fr 32-34) relative to CBF␤ levels, which remained constant throughout the cell cycle (Fig. 6D). Cyclin D3 levels again increased during G 1 to S progression. Thus, exogenous AML1 levels varied similar to endogenous AML1 during the cell cycle FIG. 6. Exogenous AML1 protein levels increase during the cell cycle. A, total cellular proteins prepared from 1 ϫ 10 6 32Dcl3 (32D) cells or from 32D-Myc-AML1 cells exposed to 100 M zinc chloride for 16 h were subjected to Western blotting using an AML1 antiserum (left panel). Extracts from 32D-Myc-AML1 cells exposed to the indicated concentrations of zinc chloride (M) for 16 h were subjected to Western blotting using a Myc tag antibody. B, total cellular proteins prepared from the indicated 32D-Myc-AML1 cell cycle fractions, obtained after culture in 75 M zinc chloride for 16 h, were subjected to Western blotting for myc-AML1 (using the Myc tag antibody), cyclin D3, or cyclin E. Coomassie Blue dye staining of these fractions is also shown (bottom panel). C, total cellular RNAs prepared from the indicated 32D-Myc-AML1 cell cycle fractions were subjected to Northern blotting for Myc-AML1 (using a human AML1 probe) and ␤-actin. D, total cellular proteins prepared from the indicated 32D-AML1ER cell cycle fractions, obtained after culture for 16 h in 200 nM 4HT, were subjected to Western blotting for AML1-ER (using ER antiserum), CBF␤, and cylcin D3. Myc-AML1 or AML1-ER expression in each sample, corrected for loading, is shown in parts B, C, and D.
despite lack of the AML1 promoter, expected to control transcription, or the AML1 5Ј-and 3Ј-untranslated RNA segments, expected to control translation efficiency and RNA stability.
Exogenous AML1 DNA Binding Domain Does Not Vary during the Cell Cycle-To begin to map the domain of AML1 responsible for its cell cycle variation, 32Dcl3 lines stably expressing either Myc-AML1-(1-217) from the MT promoter, retaining only the DNA binding domain and 87 N-terminal residues, or AML1-(86 -217)-ER from the pBabePuro retroviral vector, retaining only the DNA binding domain, were subjected to elutriation and Western blot analysis (Fig. 7, A and B). In contrast to Myc-AML1 or AML1-ER, which increased 3.0-or 4.5-fold, Myc-AML1-(1-217) increased 1.5-fold and AML1-(86 -217)-ER increased 1.6-fold during the cell cycle. Analysis of a second subclone expressing AML1-(86 -217)-ER gave similar results (not shown).
Phosphorylation of AML1 by ERK Does Not Mediate Its Cell Cycle Variation-Serines 276 and 293 are targets of cytokinemediated, ERK-dependent phosphorylation in hematopoietic cells (36). In particular, these investigators found that exogenous HA-AML1 was expressed as four closely spaced bands in Ba/F3 cells, that the upper two bands were absent after withdrawal of IL-3 and serum, and that mutation of serines 276 and 293 to alanine prevented the appearance of the upper two bands. To determine whether cytokine-dependent phosphorylation of endogenous AML1 varied during the cell cycle, we first withdrew 32Dcl3 cells from serum and IL-3 for 3 h by transferring them at room temperature to the same buffer used for elutriation. Complete media was then added, the cells were returned to the 37°C, 5% C0 2 incubator, and protein samples were prepared at various time intervals. These samples were subjected to Western blotting using AML1 antiserum (Fig. 8A). Before starvation, the majority of AML1 was in the upper band of a doublet, whereas after starvation the majority of AML1 was in the lower band. By 60 or 120 min the two bands were of equal intensity, and after 240 min the upper band again had a greater intensity. After serum and IL-3 withdrawal the cells potentially enter a temporary state not fully responsive to IL-3. Having found that 4 h are required to recover cytokine-depend-ent AML1 phosphorylation, we then sought to determine whether this varied among cell cycle fractions. 32Dcl3 cell cycle fractions were returned to IL-3 for 4 or 6 h and assayed for AML1 by Western blotting (Fig. 8B). Each fraction developed a doublet band indicative of phosphorylation, but the levels of AML1 after exposure to serum and IL-3 continued to vary during the cell cycle. In particular, AML1 levels increasing 1.4or 3.6-fold in the 0 h fractions, comparing the earliest with the latest cell cycle fractions, and 4.6-or 3.3-fold in the samples exposed to IL-3 (for an average of 3.0-and 3.5-fold increase during the cell cycle in the two experiments). Thus, cytokinedependent phosphorylation occurs during each cell cycle stage and does not lead to a rapid stabilization of AML1.
To further examine the role of ERK-dependent AML1 phosphorylation, 32Dcl3 lines stably expressing exogenous Myc-AML1(S276A/S293A) from the MT promoter were developed. This AML1 mutant increased more than 3-fold during cell cycle progression (Fig. 8B), similar data were obtained when this experiment was repeated. Thus, cytokine-mediated phosphorylation of AML1 does not account for the increase in endogenous or exogenous AML1 observed during G 1 to S progression.
The Cyclin D3 Promoter Is Activated by AML1-Cyclin D3 mRNA levels did not vary, relative to ␤-actin, in 32Dcl3 or Ba/F3 cell cycle fractions (Fig. 4). However, E2F1:DP-1, a heterodimeric factor with peak activity in late G 1 /early S phase, potently activated the murine cyclin D3 promoter, 2 as predicted from the presence of an E2F consensus site (33). Therefore, we considered the possibility that the peak in CBF DNA binding activity in late G 1 /early S might be designed similarly to enable maximal cyclin D3 expression when needed in hematopoietic cells. We first examined Ba/F3 cells expressing dominant inhibitory AML1 oncoproteins (17,18), CBF␤-SMMHC (INV) or AML1-ETO (A-E), from the MT promoter to assess the consequences of CBF inhibition on cyclin D1, D2, and D3 mRNA expression (Fig. 9A). Ba/F3 lines were employed rather than 32Dcl3 lines for this purpose as the MT promoter is leakier in 32Dcl3 cells, making it difficult to maintain highlevel expression of these cell cycle-inhibitory oncoproteins. Cyclin D1 expression was not detected, as has been described previously for hematopoietic cell lines (37). Zinc did not affect cyclin D2 or D3 mRNA expression in parental Ba/F3 cells, but their expression was inhibited by both the INV or A-E proteins by 7 h, with D3 expression being affected the greatest. We selected candidate AML1 binding sites from the murine D1, D2, and D3 promoters based on the AML1 consensus, 5Ј-PuAC-CPuCA-3Ј, and assessed their affinity for AML1 (Fig. 9B). In this assay, a strong AML1 site from the MPO promoter, WTMPO, was radiolabeled and competed with 5-, 10-, and 25-fold excess of unlabelled WTMPO, a version of this oligonucleotide mutant in the AML1 site, MutMPO, and each of the oligonucleotides derived from the cyclin D promoter regions. The best available site from the D1 promoter region, 5Ј-AAC-CACC-3Ј, did not compete, from the D2 promoter region, 5Ј-AACCACA-3Ј, competed modestly, and from the D3 promoter, 5Ј-GACCACA-3Ј, competed to an even greater extent, almost as well as the WTMPO oligonucleotide. Differences in the affinities of the D2-541 and D3-383 sites may reflect affects of flanking sequences. To determine whether AML1 can transactivate the murine cyclin D3 promoter, a 447 bp promoter fragment, including the transcription start site was linked to luciferase and co-transfected with pCMV-AML1 and pCMV-CBF␤ in 293T cells (Fig. 9C). AML1 activated the cyclin D3 promoter 2.3-fold in this context in multiple repetitions, and mutation of the AML1 binding site at Ϫ383 reduced this activation to 2 F. Bernardin-Fried and A. D. Friedman, unpublished data. 1.6-fold, suggesting a role for this binding site, although the differences in activity did not reach statistical significance. DISCUSSION A key finding of this study is that endogenous CBF DNA binding and AML1 protein levels reproducibly increase 3-fold, when normalized to total protein content, during the G 1 to S cell cycle transition. In addition physiologic levels of exogenous AML1, expressed inducibly to avoid biasing cell cycle distribution, increased similarly, supporting the additional conclusion that AML1 protein stability is regulated during the cell cycle. Reduced expression of AML1 in G 1 -arrested cells is consistent with this conclusion. Regulation of AML1 transcription may also occur during cell cycle progression. These findings were confirmed in two growth factor-dependent hematopoietic cell lines. We have focused our studies on hematopoietic cells as these are most physiologically relevant in view of the finding that AML1-null mice specifically lack hematopoiesis (6,7). To put the degree of AML1 variation in perspective, mdm2 levels increase 4-fold and cyclin B1 varies 10-fold during cell cycle progression in HeLa cells (38). AML1 levels remained high in G 2 /M phase compared with S phase cells, whereas CBF and AML1 DNA binding was reduced, suggesting that AML1 DNA binding activity is itself regulated during S to G 2 /M progression.
Several lines of experimentation using the 32Dcl3 and Ba/F3 cell lines had demonstrated that gene activation by AML1 stimulates their G 1 to S progression. In particular, transrepression of AML1 target genes slowed G 1 progression, an effect overcome by co-expression of regulators known to stimulate S phase entry, and exogenous AML1 accelerated the G 1 to S transition, dependent upon the AML1 transactivation domain (18 -20, 23, 25). It is therefore not surprising that hematopoietic cells require increased AML1 during late G 1 and have developed mechanisms to reduce AML1 during the G 1 arrest associated with differentiation. As differentiation is a multistep process, the reduced AML1 protein levels evident after G-CSF treatment may reflect mechanisms other than those utilized to regulate AML1 during the cell cycle.
It is intriguing that hematopoietic cells predominantly express cyclin D2 and D3, rather than D1, and that we have identified cis elements within the D2 and D3 promoters which bind AML1. On its own, AML1 only activated the D3 promoter 2.3-fold, perhaps reflecting the need to cooperate with other proteins to contribute most effectively. And mutation of the AML1 binding site at Ϫ383 only partially abrogated this activation, perhaps indicating contribution from additional AML1 binding sites. AML2 or AML3 may also contribute to regulation of the cyclin D3 promoter, as supershifting with AML1 antiserum did not eliminate cell cycle variation of residual CBF DNA binding activity ( Fig. 2A), perhaps accounting in part for our inability thus far to detect interaction of AML1 with the cyclin D3 promoter by chromosomal immunoprecipitation. In- FIG. 8. Cytokine-dependent phosphorylation does not regulate AML1 expression during the cell cycle. A, 32Dcl3 cells were serum-and IL-3starved for 3 h at room temperature in elutriation medium, and then returned to IMDM with fetal bovine serum and IL-3 at 37°C, 5% CO 2 for the indicated number of minutes. Total cellular proteins were then subjected to Western blotting for AML1. An extract from cells collected before starvation (pre) was also analyzed. Upper and lower bands in the AML1 doublet are indicated by arrows. B, the indicated 32Dcl3 cell cycle fractions were returned to IMDM/FBS/IL-3 for 4 h (left panels) or 6 h (right panels). Total cellular protein samples prepared prior to addition of growth medium (0 h) and after addition for the indicated times were subjected to Western blotting for AML1 and to Coomassie Blue staining. Expression of AML1 in each sample, corrected for loading, is shown below, with the 0 h samples being analyzed separately from the 4 or 6 h samples (ϩIL3). As there was not sufficient sample available for two of the Coomassie lanes, AML1 was normalized to the paired 0 or 6 h Coomassie signal, in view of the equality of the slowly migrating nonspecific band in the same paired fractions. C, total cellular proteins from the indicated 32D-Myc-AML1(S276A/ S293A) cell cycle fractions were subjected to Western blotting using Myc tag antibody. Coomassie Blue dye staining of these fractions is also shown. Myc-AML1(S276A/S293A) expression in each sample, corrected for loading, is indicated below.
terestingly, the C-terminal regions of AML2 or AML3 when knocked into the AML1 locus allow nearly normal in vivo hematopoiesis (39). Future experiments will further characterize regulation of the cyclin D3 gene by CBF family members.
As AML1 is expressed in pluripotent and lineage-restricted hematopoietic stem cells (40, 41), we speculate that stimulation of AML1 protein stability or RNA expression also plays a role in the entry of these cells into cell cycle from G 0 , as required to maintain hematopoietic homeostasis or during cytokine-mediated stress responses. In this regard, both c-Myc and B-Myb RNA levels increase when cells transition from G 0 to G 1 , and during active proliferation B-Myb protein levels are maximal during S phase due to cyclinA/cdk2-mediated phosphorylation (42)(43)(44). Similarly, cyclin D levels increase upon cytokine-mediated cell cycle entry (45). Although Ba/F3 and 32Dcl3 cells divide rapidly, the reduced levels of cyclin D3 we detected in early cell fractions may indicate that after mitosis these lines enter a G 0 -like state and are transiently resistant to cytokine signaling. We recently found that CBF␤-SMMHC inhibits G 1 to S progression in bone marrow-derived myeloid progenitors, just as it does 32Dcl3 and Ba/F3 cells (46), and AML1-ETO inhibits the proliferation of human and murine progenitors (47,48). In addition, reduced megakaryocyte ploidy in adult AML1(ϩ/Ϫ) mice may reflect inhibition of cell cycle progression (49). Future experiments will further address the role of AML1 in regulation of the cell cycle and cell cycle entry in hematopoietic stem cells.
Several mechanisms might operate to control AML1 protein stability. One possibility is that phosphorylation of AML1 by cdks such as cdk2, cdk4, or cdk6 increases its stability during late G 1 . AML1 contains three potential cdk2 phosphorylation sites ((S/T)(P/X)(K/R)) that are conserved in the murine and human proteins and 12 of the less well defined cdk4/6 phosphorylation sites ((S/T)P). Direct phosphorylation by cdk2 regulates cell cycle specific activity of the MEF transcription factor in COS cells (50). Cyclin E variation during the cell cycle depends upon ubiquitination by a specific ubiquitin ligase, which results in proteosome targeting and degradation (51). AML1 levels in 32Dcl3 cells were found to be increased by the MIG132 proteosome inhibitor (52). We replicated this finding using 32D-Myc-AML1 cells (not shown), although significant cellular toxicity ensued precluding cell cycle fractionation. Protein-protein interaction studies might identify a specific ubiquitin ligase or another protein, which stabilizes AML1 in late G 1 . Our deletional analysis of myc-AML1 and AML1-ER suggests that residues C-terminal to the DNA binding domain are involved in the regulation of AML1 stability. Of note, cyclin D3 interacts with AML1 via this region (53), potentially recruiting cdk4/6. We found that cytokine-dependent phosphorylation of serines 276 and 293 by ERK did not vary during the cell cycle nor alter AML1 stability. Future experiments will investigate whether additional AML1 phosphorylations or protein interactions control its stability during cell cycle progression or in response to cell cycle entry or exit. In this regard it is likely insufficient to characterize the ubiquitination or phosphorylation of a particular AML1 residue during the cell cycle as the unstable modified or unmodified form may be rapidly degraded. Rather, AML1 mutants resistant to modification might be identified which remain at high, constant levels throughout the cell cycle.
In a subset of acute myeloid leukemias, AML1 activities are reduced due to mutation or expression of dominant oncoproteins, potentially favoring transformation due to inhibition of differentiation and/or reduced apoptosis (13,54,55). Perhaps additional leukemias harbor mutations, which reduce AML1 level via protein modification, using the same biochemical pathway employed to reduce AML1 expression in G 1 phase cells. In these leukemias, additional mutations stimulating G 1 progression may be required (26,27). Other acute leukemias harbor increased numbers of AML1 genes due to gene amplification or to trisomy 21 (56 -59). In these leukemias, AML1 may facilitate transformation by stimulating cell cycle progression. Perhaps additional leukemias harbor alterations, which increase AML1 levels indirectly. Investigating the mechanisms underlying AML1 cell cycle variation might therefore uncover novel pathways contributing to leukemogenesis.