Forced Expression of Cyclin-Dependent Kinase 6 Confers Resistance of Pro-B Acute Lymphocytic Leukemia to Gleevec Treatment

Abstract

The gene encoding c-ABL, a nonreceptor protein tyrosine kinase, is involved in a chromosomal translocation resulting in expression of a BCR-Abl fusion protein that causes most chronic myelogenous and some acute lymphocytic leukemias (CML and ALL) in humans. The Abelson murine leukemia virus (A-MuLV) expresses an alternative form of c-Abl, v-Abl, that transforms murine pro-B cells, resulting in acute leukemia and providing an experimental model for human disease. Gleevec (STI571) inhibits the Abl kinase and has shown great utility against CML and ALL in humans, although its usefulness is limited by acquired resistance. Since STI571 is active against A-MuLV-transformed cells in vitro, we performed a retroviral cDNA library screen for genes that confer resistance to apoptosis induced by STI571. We found that forced expression of Cdk6 promotes continued cell division and decreased apoptosis of leukemic cells. We then determined that the transcription factor E2A negatively regulates Cdk6 transcription in leukemic pro-B cells and that the v-Abl kinase stimulates Cdk6 expression via an extracellular signal-regulated kinase 1-dependent pathway. Finally, we show that the cyclin-dependent kinase 4 and 6 (CDK4/6) inhibitor PD0332991 can act synergistically with STI571 to enhance leukemic cell death, suggesting a potential role for CDK6 inhibitors in the treatment of STI571-resistant CML or ALL.

INTRODUCTION

The c-Abl gene encodes a nonreceptor protein tyrosine kinase that is necessary for normal hematopoiesis and neurogenesis in mice (18, 37, 42). In humans, it is involved in a 9;22 chromosomal translocation, the Philadelphia (Ph) chromosome, that is associated with the vast majority of cases of chronic myelogenous leukemia (CML) and a fraction of acute lymphocytic leukemia (ALL) cases (44). The resultant oncogenic BCR-Abl fusion protein is a constitutively active kinase. Similarly, v-Abl, the product of a fusion between retroviral gag and c-Abl, is the transforming component of Abelson murine leukemia virus (A-MuLV), a replication-defective retrovirus that causes pro-B cell leukemia in young mice (33, 34). In vitro, A-MuLV infection transforms progenitor B cells, resulting in leukemic cell lines arrested between the pro-B and large pre-B stages of development (4, 17). Transformation by A-MuLV begins with a proliferative phase, followed by an apoptotic crisis during which cells acquire mutations in p53 pathway genes, including p19^ARF (32). The resultant leukemic cells depend upon the constitutive kinase activity of v-Abl to promote cytokine (interleukin-7 [IL-7])-independent growth and survival and to block further differentiation (29).

STI571, also known as Imatinib mesylate or Gleevec, is a small-molecule inhibitor that is highly specific for Abl family kinases and is an effective treatment for patients with CML. As with all single chemotherapeutic agents, however, CML patients develop resistance to STI571. The most common cause of resistance is point mutations in the BCR-Abl that prevent binding of the drug (28, 39). Other mechanisms of resistance include amplification of BCR-Abl, mutations in drug transporter genes, and activated expression of another oncogene, LYN (28). This paper reports our attempt to identify additional genes that synergize with ABL in leukemic transformation and whose misregulation may contribute to resistance to STI571.

Treatment of A-MuLV-transformed leukemic cell lines with STI571 results in G₁ cell cycle arrest, developmental progression to a pre-B-cell-like state, and ultimately apoptotic cell death (29). Initial studies using a temperature-sensitive mutant, v-Abl, demonstrated that the shift to nonpermissive temperature resulted in the induction of the Rag genes and an increase in germ line kappa transcription (4), and studies from our lab using STI571 to inactivate v-Abl kinase showed similar results (29). DNA microarray analyses revealed that upon inactivation of v-Abl, several genes associated with pre-B-cell differentiation, such as Spi-B and IRF-4, as well as tumor suppressor genes, such as Ku70, BRCA1, and RB, are induced. Conversely, several proto-oncogenes, such as c-Myc, N-myc1, and Lyl1, are repressed upon inactivation of v-Abl. These results further confirm that v-Abl signaling results in both proliferation and a block in pre-B-cell differentiation.

v-Abl activates various signaling pathways, and dissecting the relative importance of each for transformation has proven difficult. Several reports have shown that p53 inactivation (40, 41) and the downregulation of both products of the Ink4a/Arf locus, p19^ARF and p16^NKk4A, are important drivers of efficient v-Abl transformation (32, 35). Pathways and downstream substrates involved in maintaining cells in the transformed state are less well understood. In the current report, we utilized STI571 to inactivate v-Abl in an attempt to elucidate the growth and survival signals activated by v-Abl that maintain cells in a transformed state. We hypothesize that identifying pathways downstream of v-Abl in leukemic transformation may lead to the elucidation of novel mechanisms of STI571 resistance in leukemic patients and of mechanisms regulating pro- to pre-B-cell differentiation.

MATERIALS AND METHODS

Cells culture and chemical inhibitors.

A-MuLV cell lines 7G-S, PD31, and 220-8 and Ph⁺ human ALL cell lines NALM1 and BV173 were cultured in RPMI medium supplemented with 5% heat-inactivated fetal calf serum, penicillin (100 μg/ml), streptomycin (100 μg/ml), and 2-mercaptoethanol (50 μM). Phoenix cells were cultured in Dulbecco's modified Eagle's medium (DMEM) containing 5% heat-inactivated fetal calf serum, penicillin, and streptomycin. For cells infected with retroviral plasmids, expressing fusion proteins containing the ligand binding domain of the estrogen receptor, 1 μM 4-hydroxtamoxifen, or ethanol were used to induce protein activity. The following chemical inhibitors were used: STI571 (Novartis), 5 μM phosphatidyl inositol 3-kinase [PI(3)K] inhibitor Ly294002 (Calbiochem), 2 μM Akt inhibitor VIII (Calbiochem), 5 μM Jun N-terminal kinase (JNK) SP600125 (Calbiochem), 1 μM p38 SB203580 (Calbiochem), and PD0332991 (obtained from Pfizer). Kill curve assays were performed either by manual counting using trypan blue exclusion or by counting live gated cells from 100 μl of culture using a flow cytometer. All cell counting experiments were performed in triplicate.

Quantitative PCR.

RNAs were isolated using TRIzol reagent (Invitrogen), and cDNAs were prepared using Moloney MLV reverse transcriptase (Invitrogen) according to the manufacturer's protocol. Quantitative reverse transcription-PCRs (RT-PCRs) were performed with JumpStart Taq (Sigma) and EvaGreen (Biotium) and using an ABI 7300 thermocycler (Applied Biosysems). The amplification programs were as follows: 95°C for 5 min; 95°C for 15 s; and 60°C for 20 s, 72°C for 30 s (data collected) for 40 cycles. The melting curves were as follows: 95°C for 20 s, 60°C for 15 s, and up to 95°C for 20 s with a 19-min ramping time. Primers used in this study were as follows: HPRT, 5′-CTGGTGAAAAGGACCTCTCG and 3′-TGAAGTACTCATTATAGTCAAGGGCA; CDK6, 5′-GGCGTACCCACAGAAACCATA and 3′-AGGTAAGGGCCATCTGAAAACT; CDK4, 5′-ATGGCTGCCACTCGATATGAA and 3′-TCCTCCATTAGGAACTCTCACAC; E12, E47, PAX5, and RAG2 primers were previously described (3).

ChIP.

Chromatin immunoprecipitation (ChIP) was conducted as previously described (21). Fifty million HF4 cells (gift from Y. Zhuang, Duke University), which are v-Abl-transformed E2A His-Flag-tagged cells, were treated with 1 μM STI571 for 16 h or left untreated. Each immunoprecipitation mixture was incubated with 6 μg of Flag antibody (F1804; Sigma-Aldrich) or mouse IgG (Santa Cruz Biotechnology). Recovered DNA was resuspended in 250 μl Tris-EDTA (TE) and analyzed by quantitative PCR. Input samples represented 1% of total DNA, and percent input was calculated as the enriched/input ratio. Primers used for PCR analyses of ChIPs were as follows: 5′ region A, GCACGACACTACTCCCCTTC; 3′ region A, ATGGCAAGCTTAGTGGGAGA; 5′ region D, GAAAAGAAAGGAAGCAATTTCC; 3′ region D, GGGGCTCCTAGAACCCTGTA; 5′ region EX1, GAGTGCAGACCAGTGAGGAG; 3′ region EX1, GGGGTGCTCGAAGGTCTC. Primers for CD19 and mb1 were described previously (20).

Immunoprecipitation and immunoblot analysis.

Whole-cell extracts were prepared from 220-8 and 7G-S cells treated with 2 μM STI571 or left untreated. A total of 30 to 50 μg of cell extracts, as determined by Bradford assay, was separated using SDS-PAGE, transferred onto an Immobilon-FL membrane (Millipore), and incubated with anti-CDK6 (CP06; Calbiochem), anti-CDK4 (sc-260; Santa Cruz Biotechnology), antiactin (sc-615; Santa Cruz Biotechnology), or anti-ID2 (sc-489; Santa Cruz Biotechnology). For the IP mixture, 500 μg of whole-cell extracts was precleared and then incubated with the ID2 antibody for 3 h at 4°C. Subsequently, 30 μl of protein A/G-Sepharose (Santa Cruz Biotechnology) beads was added and incubated for an additional hour at 4°C. IP mixtures were washed four times with radioimmunoprecipitation assay (RIPA) buffer and eluted by boiling the beads in SDS sample buffer.

Electrophoretic mobility shift assay (EMSA).

Oligonucleotides obtained from Elim (Hayward, CA) were annealed, labeled, and used as probes. Each probe was designed so that the binding site was in the center of the 24- to 28-nucleotide duplex. Binding reaction mixtures contained 0.025 pmol ³²P-labeled probe and 5 to 10 μg of nuclear extract from 220-8 pro-B cells. Native polyacrylamide gels were dried and analyzed using a phosphorimager and ImageQuant software (Amersham).

Retroviral cDNA library screen.

A bone marrow pro- and pre-B-cell retroviral cDNA library consisting of five separate pools, each containing 6 × 10⁵ cDNAs (described in reference 3), was packaged in Phoenix cells by using calcium-phosphate transfection. Two days posttransfection, the supernatants were harvested and used to infect 3 × 10⁶ 7G-S cells per pool. Beginning 2 days postinfection, each pool of cells was treated with 5 μM STI571 for 2 days and then grown in RPMI medium without STI571 for another 5 days. Dead cells were removed by sedimentation through Ficoll-Hypaque (GE Life Sciences), and genomic DNA was harvested from the remaining live cells. Retroviral cDNA inserts were recovered from this genomic DNA by PCR using primers specific to the retroviral vector and the Advantage 2 PCR enzyme system (BD Biosciences). These amplified inserts were purified, digested, and then recloned into the CMSCV-IRES-Thy1.1 retroviral vector, resulting in a simplified library pool. Each pool of enriched cDNA was then used to infect the parental 7G-S cells, followed by another round of selection as described above. A total of three rounds of selection were performed. Following the third round, individual amplified cDNA inserts were separated on an agarose gel, recloned, and subjected to DNA sequence analysis.

Cell cycle analyses.

Cells (0.5 × 10⁶) were harvested, washed once in phosphate-buffered saline (PBS), and resuspended in 500 μl of PBS. The resuspended cells were gently vortexed at low speed while adding 500 μl of 100% ethanol dropwise at about 1 drop per second to permeabilize and fix the cells, which were then incubated on ice for at least 30 min. The cells were spun down at 720 × g for 5 min, resuspended in 500 μl of PBS containing 100 μg/ml RNase and 50 μg/ml propidium iodide, and incubated at room temperature for 20 min prior to analysis by flow cytometry using FlowJo software.

To measure active DNA synthesis, bromodeoxyuridine (BrdU) was added to cultures for 10 min prior to cell harvest. Cells were then washed in PBS and permeabilized with ethanol for 30 min prior to fixation in 1% paraformaldehyde and 0.5% Tween 20 for 30 min. Following incubation, the cells were stained with fluorescein isothiocyanate-conjugated antibody to BrdU and 7-amino actinomycin D (7AAD; Sigma-Aldrich) and then analyzed by flow cytometry.

Retroviral infections.

Both CMSCV-IRES-Thy1.1 and MSCV-IRES-CD2 mir30 retroviral plasmids were previously described (3). CMSCV-IRES-Thy1.1-CDK6 was constructed by inserting a PCR-amplified and sequence-verified murine CDK6 cDNA into the SfiI site of CMSCV-IRES-Thy1.1. shRNA sequences in the context of mir30 were obtained from Open Biosystems. The shRNA sense strand for targeting Cdk6 was CGCACATTGAATTAAATAGCATA and for Erk-1 it was AACCCTGGAAGCCATGAGAGAT. Oligonucleotides were obtained from Invitrogen and were cloned into MSCV-IRES-CD2 mir30 as previously described (3). Phoenix cells (at 75 to 80% confluence) were cotransfected with 10 μg of the retroviral plasmid of interest (described above) and 3 μg of vesicular stomatitis virus G protein by using calcium phosphate precipitation. Retroviral supernatants were collected 2 days posttransfection and used to infect A-MuLV-transformed cells. The following antibodies were used for flow cytometry to track infected cells: phycoerythrin (PE)-Thy1.1 (OX-7; Pharmingen), PE-hCD4 (RPA-T4; eBiosciences), and PE-hCD2 (RPA-2.10; eBiosciences).

RESULTS

cDNA library screen for genes that confer resistance to STI571.

An A-MuLv-transformed pro-B-cell leukemia cell line, 7G-S, was used in a retroviral cDNA library screen for genes that enhance leukemic cell survival upon pharmacologic inactivation of v-Abl. Culture of 7G-S for 2 days in 5 μM STI571 results in approximately 1.8% cell survival compared to untreated cells (data not shown), representing the background of the screen. We used a cDNA library generated from wild-type murine bone marrow pro- and pre-B-cell mRNA, reverse transcribed and cloned into an murine stem cell virus (MSCV)-based Thy1.1-marked retroviral vector (3). Library pools were packaged into retroviral particles and used to transduce 7G-S cells. The infected cells were cultured in the presence of STI571 for 2 days and then allowed to recover for an additional 5 days in the absence of STI571. Genomic DNA was purified from surviving cells, and proviral cDNA inserts were recovered by PCR amplification and then cloned back into the original retroviral vector, generating an enriched library. Two additional rounds of retroviral infection and selection were performed to further enrich for genes that confer resistance to cell death in leukemic cells treated with STI571 before subjecting the resultant individual cDNA inserts to DNA sequence analysis.

Nine unique genes were identified after three rounds of retroviral infection and selection. c-Myc, a gene known to be induced by v-Abl, was identified in three separate library pools (46) (see Table S1 in the supplemental material). One cDNA identified in the screen encodes the cyclin-dependent kinase 6 gene (Cdk6), which was selected in two independent pools. To confirm that CDK6 overexpression confers resistance to STI571, a wild-type Cdk6 cDNA was cloned into a retroviral vector also expressing a Thy1.1 cDNA, used to mark transduced cells. Retrovirus containing either wild-type Cdk6 or a control vector lacking this cDNA insert was used to infect several A-MuLV-transformed cell lines (Fig. 1A; see also Fig. S1 in the supplemental material). The heterogeneous populations of untransduced and transduced cells containing wild-type Cdk6 expressed 1.75- to 2.5-fold over endogenous levels (data not shown), and the empty vector control was cultured in the presence of a low dose of STI571 for 7 days. The fraction of cells expressing Cdk6 increased from 16.9% in untreated cells to 52.5% after treatment with STI571, as indicated by THY1.1 expression (Fig. 1A). This survival advantage was not apparent in cells expressing the empty vector control, in which the Thy1.1-expressing cells remained virtually unchanged: 9.7% in untreated cells compared to 9.5% in STI571-treated cells (Fig. 1A). Additionally, cell numbers continued to increase after 48 h in CDK6-expressing cells, while the empty vector control cells persisted at about the same number after treatment with STI571 (Fig. 1B). Without STI571, both Cdk6- and empty vector-transduced cells proliferated similarly. We conclude from these experiments that forced expression of exogenous CDK6 in A-MuLV-transformed cells promotes survival and proliferation in the presence of STI571.

Fig. 1.

Exogenous CDK6 expression renders cells partially resistant to STI571. (A) 7G-S cells, transduced with either an empty vector control retrovirus or one expressing a Cdk6 cDNA, were cultured for 7 days in the presence or absence of 0.2 μM STI571. Flow cytometry data are shown, comparing untreated and STI571-treated cells for surface levels of Thy.1.1, a marker of retroviral transduction. Numbers represent the percentages of Thy1.1-expressing cells in cultures treated with STI571; values in parentheses are the percentages of Thy1.1 cells in cultures not treated with STI571. Data are representative of three similar experiments. (B) Sorted 7G-S cells expressing either empty vector control or Cdk6 were cultured in the presence or absence of 0.2 μM STI571 for the indicated durations, and live cells were enumerated in a trypan blue exclusion assay. Data are representative of at least three independent experiments.

Regulation of cell cycle proteins by STI571.

Previous studies showed that STI571 arrests cell cycle progression in v-Abl-transformed cells (29). Since CDK6 is a known positive regulator of cell cycle activity, we asked whether STI571 treatment alters endogenous CDK6 expression in untransduced cells. We found that both transcript and protein levels decreased upon STI571 treatment, 3.5-fold and 2-fold at 24 h, respectively (Fig. 2A and B). This is in contrast to the effect of STI571 on the protein levels of CDK4, a closely related family member, which remained unchanged in A-MuLV-transformed cells treated with STI571 (Fig. 2B). In addition, 7G-S cells transduced with a retrovirus expressing a CD2-tagged shRNA against endogenous CDK6 showed a loss in the fraction of infected cells over time (see Fig. S2 in the supplemental material). These results indicate that CDK6 is required in A-MuLV-transformed pro-B cells and is necessary for transformed cell growth and survival. Interestingly, we found that the decrease in CDK6 levels in STI571-treated cells is accompanied by considerable decreases in expression of the genes encoding heterodimeric kinase partners CyclinD2 (25-fold) and CyclinD3 (2-fold) (Fig. 2C). We also detected a corresponding increase in the INK4 inhibitors that compete with cyclins for CDK4/6 binding. The transcript levels of both p18^INK4 and p19^INK4 increased by approximately 8-fold upon STI571 treatment (Fig. 2C), confirming observations made previously in a microarray study (29). Thus, decreases in CDK6 and D-type cyclin expression and increases in CDK inhibitor levels likely contribute to the G₁ cell cycle arrest observed in STI571-treated leukemic cells.

Fig. 2.

Cdk6 transcript and protein levels decrease with the inhibition of v-Abl. (A) 7G-S cells were cultured overnight in 2 μM STI571 or left untreated, and RNA was harvested and analyzed by quantitative RT-PCR for Cdk6 mRNA with the results normalized to hprt. Error bars correspond to the standard deviations of triplicate samples. (B) 7G-S cells were cultured in the presence of 2 μM STI571 for the indicated durations followed by whole-cell lysis. Lysates were analyzed by immunoblotting with antibodies to CDK4, CDK6, and actin. Relative CDK6 levels normalized to actin are indicated as determined by quantitative imaging. (C) Graph of quantitative RT-PCR analyses of Ccnd2, Ccnd3, Cdkn2c, and Cdkn2a mRNA levels relative to the hprt control, using samples from the experiment summarized for panel A. Error bars correspond to the standard deviations of triplicate samples. Data are representative of at least three independent experiments.

CDK6 promotes cell cycle progression.

Since v-Abl-transformed cell lines expressing exogenous CDK6 outcompete untransduced cells in the presence of STI571, we proceeded to ask whether this was due to enhanced proliferation. An analysis of cell cycle distribution using propidium iodide staining and flow cytometry revealed a higher percentage of Cdk6-expressing cells than control cells continuing to cycle after treatment with STI571 (Fig. 3A). CDK6 overexpression did not affect the distribution of cells between the G₁, S, and G₂/M stages in the absence of STI571 (Fig. 3A). As early as 8 h after treatment with STI571, empty vector control cells began to decrease in the percentage of S/G₂ cells, while cells expressing CDK6 did not show a decrease until 24 h after treatment. The decrease in S/G₂ cells was associated with an increase in G₁-phase cells, indicating a cell cycle blockade at the G₁-to-S-phase transition, which is delayed in cells expressing exogenous CDK6. These results were confirmed by pulse-labeling cells with BrdU and then analyzing its incorporation into newly synthesized DNA by flow cytometry. CDK6-expressing cells treated with STI571 continued to progress into the S-phase, while similarly treated empty vector control cells showed a lower percentage of S-phase entry (Fig. 3B). In the absence of STI571, both empty vector control- and CDK6-expressing cells displayed similar fractions of cells entering S phase. This indicates that CDK6 is not limiting for cell cycle progression in the absence of STI571 treatment, but it becomes limiting thereafter.

Fig. 3.

Exogenous Cdk6 expression drives cell cycle progression and prevents apoptosis of STI571-treated A-MuLV-transformed cells. (A) Sorted 7G-S cells expressing either the empty vector control (upper) or Cdk6 cDNA (lower) were cultured in 0.2 μM STI571. Cells were harvested and permeabilized at the indicated times, and DNA content was assayed using propidium iodide, detected by flow cytometry. Histograms of the cell cycle (DNA content) profiles at the indicated times are shown. A graphical representation of the percentages of cells in S/G₂/M is shown on the right. These data are representative of three independent experiments. (B) BrdU incorporation after 10 min of labeling in cells from the experiment described for panel A, cultured in STI571 for 14 h. Flow cytometry was performed using 7AAD and antibodies to BrdU to determine the percentages of cycling cells. The bar graph represents the percentages of cells in each stage, G₁, S, and G₂/M, in the indicated samples. The data are representative of at least three similar experiments.

CDK6 partially inhibits genes important in the differentiation of cells induced by STI571.

v-Abl transformation is associated with a block in B-cell developmental progression. We showed previously that treatment of A-MuLV-transformed pro-B cells with STI571 releases cells from this block and results in their progression from a pro-B-cell into a pre-B-cell-like state (29). In many systems, proliferation and differentiation are closely linked. For example, the overexpression of CDK6 has recently been shown to block the differentiation of erythroleukemia cells (5, 26). To test whether CDK6 expression can similarly block STI571-induced differentiation, cells expressing CDK6 and empty vector control cells were treated with STI571 and then analyzed for the expression of differentiation-associated genes previously shown to be induced by STI571 treatment of A-MuLV cells (29). As anticipated, in the empty vector control cells, STI571 treatment resulted in induced expression of rag2 (67-fold), kappa germ line transcript (kGT; 21-fold), E12 (8-fold), and Pax5 (17-fold). However, in cells expressing CDK6, the level of induction was lower: rag2, 23-fold; kGT, 14-fold; E12, 2-fold; Pax5, 8-fold (Fig. 4A). In three independent experiments, the overall difference in the levels of E12, kGT, Pax5, and Rag2 induction in STI571-treated cells expressing CDK6 compared to empty vector control cells averaged 30%, 70%, 50%, and 40%, respectively (Fig. 4B). This demonstrates that forced CDK6 expression partially inhibits STI571-mediated induction of a set of genes known to be associated with B-cell differentiation.

Fig. 4.

Exogenous Cdk6 partially blocks the STI571-induced activation of genes involved in pro- to pre-B-cell differentiation. (A) Sorted 7G-S cells expressing the empty vector control or Cdk6 cDNA were cultured overnight in 2 μM STI571or left untreated. RNA was harvested, and quantitative RT-PCR was performed for E12, kappa germ line transcript, Pax5, and Rag2. All values were normalized to hprt expression levels. The number above each pair of STI571-treated and untreated samples reflects the fold difference, and the difference between control and Cdk6-transfected cells was statistically significant based on a two-tailed Student's t test, as indicated. (B) Each transcript is represented as the percentage of control induction, determined by dividing the normalized transcript level in cells expressing Cdk6 cDNA by the normalized transcript level in cells expressing the empty vector control. This is representative of three independent experiments.

E2A represses Cdk6 transcript levels.

We next explored how the inhibition of v-Abl activity might result in a decrease in Cdk6 expression. Previous work comparing E2A-deficient and -sufficient thymocytes identified Cdk6 as a negatively regulated target of E2A activity (36). We performed a ChIP-chip analysis to look at targets of E2A binding in A-MuLV-transformed cells and identified a peak over the first exon of Cdk6 (see Fig. S3 in the supplemental material). E2A encodes two closely related transcription factors, E12 and E47. To test whether E2A represses Cdk6 in A-MuLV-transformed cells, we took advantage of an E47-estrogen receptor (ER) fusion protein whose activity can be regulated by the addition of tamoxifen (45). As shown in Fig. 5A, the addition of tamoxifen to E47-ER-transduced A-MuLV cells decreased the transcript level of Cdk6 by 3-fold while having no effect on the transcript level of Cdk4.

Fig. 5.

E47 directly binds to the Cdk6 gene and represses its transcription. (A) Sorted PD31 cells transduced with the ER-E47 retroviral vector were cultured with ethanol (vehicle control) or tamoxifen (4-OH) for 6 h. RNA was harvested, and Cdk6 and Cdk4 mRNA levels were measured by quantitative RT-PCR normalized to hprt. Error bars indicate the standard deviations of triplicate samples. (B) Nuclear extracts from 220-8 cells that were cultured for the indicated times or for 24 h (+) in 2 μM STI571 or in the absence of STI571 (-) were incubated with a ³²P-labeled oligonucleotide containing the E2A binding site μE5 from the IgH intronic enhancer and analyzed in an EMSA. A dried gel phosphorimage is shown. Lanes labeled α-Ig and α-E47 contain control and E47-specific antisera. The arrow indicates the specific E47-DNA complex. (C) A 3-kb region encompassing the first exon of Cdk6 (upper). The black box represents the first exon, and the arrow indicates the transcriptional start site. The letters indicate putative E2A binding sites. The alignment of putative E2A binding sites in the murine Cdk6 promoter and first exon to the E-box consensus binding site are indicated by the box (lower). (D) Results of an EMSA performed as described above, using nuclear extracts from 220-8 cells cultured in the presence (+) or absence (-) of 2 μM STI571 and a 24- to 30-bp radiolabeled probe representing each of the putative E2A binding sites compared to the μE5 probe. The arrow indicates the specific E2A-DNA complex. (E) HF4 cells (expressing Flag-tagged endogenous E2A) cultured in the presence or absence of 1 μM STI571 were subjected to chromatin immunoprecipitation with anti-Flag or control IgG antibody. Precipitates were analyzed by quantitative PCR using primer pairs encompassing region A, region D, region EX1, Cd19 (negative control), and Cd79a (positive control). The results are presented as enrichment over input. (F) 7G-S cells were cultured in the presence (+) or absence (-) of 2 μM STI571 for 16 h and then analyzed by quantitative RT-PCR for E47, E12, Id2, and Id3 mRNA. All values were normalized to hprt levels, and error bars represent standard deviations of triplicates. (G) 7G-S cells were cultured in the presence of 2 μM STI571 for 16 h followed by whole-cell lysis. Lysates were analyzed by immunoblotting with antibodies to Id2 and tubulin.

To test the effect of STI571 treatment on E2A DNA binding activity in A-MuLV cells, we performed an EMSA with nuclear extracts from control and STI571-treated A-MuLV cells and a probe containing a known E2A target site, μE5 from the IgH intronic enhancer (22, 38, 43). We observed a DNA-protein complex in A-MuLV-transformed cells that was increased in extracts from cells treated with STI571 and was eliminated upon incubation with an antibody specific to E47 but not those exposed to an IgG control (Fig. 5B). A time course experiment showed that this complex increases within 2 h of treatment with STI571 (Fig. 5B). The equivalence of nuclear extract concentration and activity was demonstrated by performing an EMSA using a probe containing an OCT-1 binding site as a control (see Fig. S4 in the supplemental material). These data led us to the hypothesis that in A-MuLV-transformed cells, E2A may be repressing Cdk6 transcription. In an attempt to localize relevant E2A binding sites within the Cdk6 promoter, we analyzed a 2-kb region upstream and 1-kb region downstream of the Cdk6 first exon as well as the region identified from our ChIP-chip experiment, focusing on several computationally predicted binding sites (Fig. 5C). We performed EMSAs to determine if these sites can bind E2A in vitro. Endogenous E2A from nuclear extracts of untreated and STI571-treated A-MuLV-transformed cells and each probe containing putative E2A binding sites, A to G, EX1 and EX2, were incubated and resolved on native polyacrylamide gels. The data in Fig. 5D show that E2A activity increased in the presence of STI571 when incubated with sequences A, D, and EX1, demonstrating that E2A can bind these three sites in vitro. ChIP was also performed to assay for binding of E2A to these three promoter region sites in vivo in chromatin isolated from A-MuLV-transformed cells after 0 and 16 h of culture in the presence of STI571 (10). Primers were designed to amplify regions encompassing region A, region D, region EX1, Mb-1 (positive control), and Cd19 (negative control). As expected, low levels of binding to Cd19 were detected in all samples, while the mb-1 site was enriched and increased upon treatment with STI571. We observed increased E2A binding to region D but not A or EX1 in the Cdk6 promoter in STI571-treated cells (Fig. 5E). This demonstrates that, in vivo, E2A binds to region D.

E2A binds DNA as a homo- or heterodimer with other E-proteins, and a family of inhibitors called the ID proteins competes for dimerization and blocks E-protein function (15, 31). The E2A gene encodes two subtly different proteins, E12 and E47. Both proteins are important in driving B-cell development, and their levels increase with developmental progression (31). To explain the increase in E2A binding activity upon STI571 treatment of A-MuLV cells, we predicted that E2A expression would increase while ID expression would decrease, resulting in E2A repression of Cdk6 transcription in STI571-treated cells. To test this hypothesis, we measured E2A (both E12 and E47) and ID transcript levels in untreated and STI571-treated A-MuLV-transformed cells. While the levels of E12 and E47 increased 3-fold and 8-fold, respectively, upon STI571 treatment (Fig. 5F), we found Id3 transcripts decreased by 3-fold but Id2 transcripts increased by 4-fold (Fig. 5F), as previously demonstrated in a microarray analysis published by our lab (29). To determine if the protein level of ID2 follows that of the transcript level in treated A-MuLV-transformed cells, we performed immunoblot assays, which showed that the ID2 protein level increases similarly to the transcript level in STI571-treated cells (Fig. 5G). It has been demonstrated that ID2 and ID3 phosphorylation can interfere with binding to E47 (6, 14), so we asked whether ID2 is phosphorylated in A-MuLV-transformed cells upon treatment with STI571. We observed that ID2 serine phosphorylation is not present in STI571-treated A-MuLV cells (data not shown).

Upstream signaling pathway regulating CDK6.

The signaling pathway regulating Cdk6 expression in A-MuLV cells has not been identified, although several such pathways are known to be activated by v-Abl. In particular, both the PI(3)K and mitogen-activated protein kinase (MAPK) pathways regulate leukemic cell growth, survival, and proliferation (16, 19). To explore the involvement of v-Abl-activated signaling pathways in the regulation of Cdk6 transcript levels in A-MuLV-transformed cells, we treated 7G-S cells with a panel of chemical inhibitors targeting the PI(3)K and MAPK pathways and then measured Cdk6 transcript levels. These inhibitors included Ly294002 [PI(3)K inhibitor], Akt inhibitor VIII (Akt), SP600125 (Jnk), SB203580 (p38), and U0126 (MEK1/2) (Fig. 6A). The only inhibitor that resulted in a decrease of Cdk6 transcript level was U0126, which targets MEK1/2. Since E2A represses Cdk6 transcription, the effect of U0126 on E2A activity was determined. However, E2A transcription and DNA binding activity did not change in A-MuLV-transformed cells treated with U0126 (data not shown).

Fig. 6.

The MEK-ERK pathway regulates Cdk6 transcription. (A) 7G-S cells were cultured with the indicated chemical inhibitors, dimethyl sulfoxide (DMSO; vehicle control), or STI571 for either 16 h (upper) or for the indicated times (lower) and then analyzed by quantitative RT-PCR for Cdk6 and hprt mRNA levels. All values were normalized to hprt levels, and error bars represent standard deviations of triplicates. (B) Immunoblot of cells sorted 5 days postinfection with Erk1 shRNA or empty vector control. ERK1/ERK2 antibody was used to determine knockdown efficiency. Antibody to tubulin was used as a loading control. (C) Sorted 7G-S cells expressing the empty vector control and ERK1 shRNA were assayed for Cdk6 transcripts by quantitative RT-PCR normalized to hprt.

The downstream targets of MEK1/2 are extracellular signal-regulated kinase 1/2 (ERK1/2). To test whether the regulation of CDK6 expression goes through ERK1 or ERK2, we first generated shRNA to knock down expression of ERK1. 7G-S cells were infected with retrovirus expressing hCD2 as a marker for infection and with an shRNA directed to ERK1. An immunoblot assay using sorted cells infected with the ERK1 shRNA showed a decrease in ERK1 but not ERK2 expression, demonstrating the effectiveness of the shRNA (Fig. 6B). To determine if inhibiting ERK1 results in a decrease in Cdk6 transcript level, we sorted for hCD2-positive cells 5 days after infection and measured Cdk6 transcript levels in transduced and control cells. As shown in Fig. 6C, the level of Cdk6 was lower in A-MuLV-transformed cells infected with the shRNA to ERK1 than in the empty vector control group. These data suggest that in A-MuLV-transformed pro-B cells, v-Abl drives Cdk6 transcription via an MEK1- and ERK1-dependent pathway.

PD0332991 increases sensitivity of CDK6-expressing cells and Ph⁺ cells to STI571.

Our data suggest that CDK6 when misregulated might cause resistance of leukemic cells to STI571 treatment. To test the possible therapeutic significance of this discovery for CML patients whose leukemias are resistant to STI571 treatment but may still depend upon CDK6 activity, we asked whether inhibition of CDK6 in 7G-S cells overexpressing this protein could render cells more sensitive to STI571-induced G₁ arrest. There are several CDK inhibitors available, and each targets multiple CDKs (24). The inhibitor PD0332991, currently in phase I and II clinical trials (as per clinicaltrials.gov) for several different types of cancer, targets only CDK4 and CDK6 (9). PD0332991 and STI571 were used alone or in combination to treat cells expressing empty vector control or Cdk6. As expected, the control cells were sensitive to either inhibitor alone and showed a synergistic effect of treatment with the combination, as reflected in the decreased percentage of G₁ cells compared to untreated cells (Fig. 7A). Cells overexpressing Cdk6 and treated in the same fashion resulted in an average 6.2%, 6.2%, and 18.9% decrease in the S/G₂ population, respectively, compared to untreated cells.

Fig. 7.

Synergy between STI571 and PD0332991 in causing G₁ cell cycle arrest of v-Abl-transformed cells. (A) Sorted 7G-S cells transduced with either an empty vector control or retrovirus expressing Cdk6 were cultured with STI571 (50 nM) and PD0332991 (25 nM) alone or in combination for 16 h. Cells were harvested and analyzed for DNA content by using propidium iodide staining and flow cytometry. The effect of each inhibitor is shown as the difference in the percentage of S-G2-phase cells in untreated compared to drug-treated cultures. The number below each bar indicates the concentration (in nM). The significance differences in values as determined by a two-tailed Student's t test are indicated. (B) NALM1 cells were cultured in the presence of STI571 (5 μM) and PD0332991 (50 nM) alone or in combination for 48 h. Triplicate culture samples were harvested, and live cells were enumerated by flow cytometry at 8, 24, 32, and 48 h. The cell number is displayed as the change compared to time zero. The difference between untreated and combined treatment cultures at 48 h was statistically significant based on a two-tailed Student's t test, as indicated.

To test the potential clinical relevance of these observations, Ph⁺ cell lines derived from ALL patients were tested for their sensitivity to combination treatment with STI571 and PD0332991. NALM1 and BV173, in which CDKN2A is partially or entirely deleted, were treated with PD0332991, STI571, or a combination of the two drugs (8). After 48 h of treatment with STI571, PD0332991, or the combination, NALM1 cells showed reductions of 24%, 30%, and 56% in cell number, respectively, compared to untreated cells (Fig. 7B). A similar reduction was observed for BV173 (see Fig. S5 in the supplemental material). This suggests that in ALL and CML patients, combined treatment with Abl and CDK6 inhibitors might prove more effective than either inhibitor alone and that CDK6 inhibitors might prove useful in some patients who have developed STI571 resistance.

DISCUSSION

The majority of CML patients exhibiting resistance to STI571 possess mutations in the ATP binding pocket of the BCR-Abl protein that prevent binding of STI571 (28). We have taken an unbiased genetic approach to identify potential downstream targets of v-Abl that when overexpressed render transformed cells resistant to STI571 and thus may explain a fraction of the remaining drug-resistant disease. In addition, this model system provides an opportunity to assess mechanisms that can inhibit pro- to pre-B-cell differentiation. Our studies identified CDK6 as a downstream target of v-Abl regulation and a potential candidate for STI571 resistance in CML patients. Inactivation of v-Abl leads to the downregulation of CDK6, and we found that this repression is directly mediated by E47 and depends on ERK1. When combined with STI571, a small-molecule inhibitor, PD0332991, which targets only CDK4 and CDK6, showed synergy in causing G₁ cell cycle arrest in A-MuLV-transformed cells and in cells overexpressing CDK6. PD0332991 is a potential therapy for patients resistant to STI571 who exhibit increased CDK6 levels.

CDK6 and CDK4.

The G₁ cyclin-dependent kinases CDK4 and CDK6 share 71% amino acid homology and are thought to be very similar in function (12). There is growing evidence, however, that important differences exist between CDK4 and CDK6, including subcellular localization (7, 11, 23), distinct responses to the cell cycle inhibitor p21 (30), and differences in amplification in various types of tumors (24). Most recently, a role for CDK6 but not CDK4 was observed in the differentiation of a variety of cell types (7, 26, 27). Similarly, in our genetic screen, we identified Cdk6 but not Cdk4. As shown in Fig. 1A, CDK6 overexpression renders cells relatively resistant to STI571-induced cell cycle arrest. CDK4, when expressed at the same level as CDK6, can also render cells resistant (data not shown); however, as shown in Fig. 1B, CDK4 protein levels in A-MuLV cells were not affected by STI571 treatment. This implies that CDK4 may be irrelevant for cell cycle progression in A-MuLV-transformed cells.

A new pathway for the control of proliferation and apoptosis in leukemic pro-B cells.

The cytokine IL-7 is critical for pro-B-cell survival and proliferation. However, the transformation of pro-B cells by A-MuLV relieves this dependence and allows the cells to proliferate without IL-7. Inactivation of v-Abl by STI571 induces differentiation to a pre-B cell-like state, as characterized by the cessation of cell cycle, RAG induction, and light chain gene rearrangement (29). These events are very similar to the transition from pro- to pre-B cells, and therefore v-Abl in many ways is able to mimic all the signals required for proliferation and maintenance in the pro-B-cell state. In this study, we have described a pathway in which leukemic pro-B cells continue cycling by keeping E2A levels and activity low and thus maintaining high CDK6 levels high. Upon inactivation of v-Abl, E2A activity increases, which directly represses Cdk6 transcription causing cell cycle arrest. Similarly, during the pro- to pre-B-cell transition in vivo, cells exit the cell cycle and E2A activity increases to target Igκ transcription and recombination. In pre-B cells, one pathway coordinating cell cycle exit depends upon the pre-BCR-mediated activation of the RAS-MEK-ERK pathway (25). The RAS-MEK-ERK pathway decreases CyclinD3 transcription and induces E2A activity. We also showed that CyclinD3 transcripts decrease upon inactivation of v-Abl, and perhaps in pre-B cells CDK6 levels are also repressed by E2A, as in leukemic cells. However, in contrast to previous findings that during the pro- to pre-B-cell transition, RAS-MEK-ERK pathway activity drives cell cycle exit and induces E2A activity, in leukemic cells the RAS-MEK-ERK pathway is active and is inactivated upon inhibition of v-Abl. We showed that in leukemic cells, the RAS-MEK-ERK pathway maintains CDK6 levels and that a MEK inhibitor or Erk-1 shRNA decreases CDK6 transcript levels but does not have any effect on E2A level or activity. This difference in the effects of the RAS-MEK-ERK pathway between pre-B and leukemic cells may be explained by the lack of pre-BCR expression on A-MuLV-transformed cells (7). The absence of any effects on E2A in A-MuLV-transformed cells in the presence of a MEK inhibitor suggests that two different pathways are regulating the induction and the repression of Cdk6 transcription. Interestingly, it has been observed that ERK2 and not ERK1 mediates resistance to STI571 in BCR-Abl leukemic cells (1). This demonstrates that the Erk pathway drives proliferation in Abl-transformed cells, but our finding shows that ERK1 is important in v-Abl-transformed cells.

E-protein DNA binding activity, including that of E47, is modulated by the presence or absence of the ID inhibitors. Most important in B cells are ID2 and ID3. While Id3 transcript levels do decrease in A-MuLV-transformed cells treated with STI571, Id2 transcript levels increase. The increase of Id2 levels in STI571-treated A-MuLV-transformed cells is consistent with a previous microarray study performed in our lab (29). Several posttranslational modifications have been described that may explain the increase of both E2A DNA binding activity and ID2 levels. Specifically, serine phosphorylation prevents the binding of ID2 and ID3 to E2A (6, 14). We were unable to confirm that this is the case for ID2 in leukemic cells, however. It is possible that a yet-to-be-identified posttranslational modification of E2A prevents binding to IDs or that the increase in E2A protein level is greater than the net increase in ID protein levels, allowing the binding activity of E2A to increase in the presence of STI571.

Schwartz et al. identified Cdk6 among a small set of genes downregulated in E47 as sufficient compared to E47-deficient thymomas (36). The inverse relationship between between E47 and CDK6 expression levels in A-MuLV-transformed cells supports the notion that E47 represses Cdk6. In a ChIP assay for E2A analyzed on a genomic tiling array, we detected E2A occupancy over the promoter of the Cdk6 locus in A-MuLV-transformed cells. Using conventional ChIP, with subsequent analysis by quantitative PCR, we showed that upon STI571 treatment of A-MuLV-transformed cells, the increase in E2A binding at the promoter was modest. Although some level of E2A binding was observed in the absence of STI571, it was not sufficient to fully repress Cdk6. At the Cdk6 promoter, it is possible that once cells are treated with STI571, E2A either binds de novo or that cofactors are induced and recruited to regions where E2A is present to functionally repress transcription.

CDK6 and cancer.

Cell cycle progression is highly regulated at several checkpoints to ensure proper DNA synthesis and chromosome segregation. It has become increasingly clear that the deregulation of CDK pathways causes unscheduled proliferation that contributes to oncogenic transformation. Misregulation of CDK4 or CDK6 and their cell cycle inhibitors in particular are associated with many types of cancer (24). Our study shows that in pro-B leukemic cells, CDK6 is the dominant CDK regulating the cell cycle. Several microarray studies, using primary mononuclear cells or cell lines responsive or resistant to STI571, were compared to acquire a gene expression signature of cells resistant to STI571 (13). None of these studies identified CDK6 as being upregulated in resistant cells. However, these gene expression signature studies also failed to define gene expression patterns that could definitively identify patients that may be resistant to STI571, suggesting that a wide diversity of genes can be misregulated to render cells resistant to this drug. Interestingly, a study exploring the epigenetic alterations of microRNAs in ALL identified a tumor suppressor microRNA, miR-124a, that is hypermethylated at its promoter and thus downregulated in ALL (2). It was demonstrated that miR-124a targets Cdk6 expression. In mice, downregulation of miR-124a induced upregulation of its target, Cdk6, which contributed to abnormal proliferation of ALL cells. The epigenetic downregulation of miR-124a was also shown to confer a poor prognosis in ALL patients. This reinforces the importance of CDK6 misregulation in human cancers and potentially in CML patients.

STI571 and PD0332991.

Misregulated CDKs, cyclins, or cell cycle inhibitors can result in alterations in cell cycle activity, and they have all been linked to various cancers (24). Whether a mutation affects cyclins or their inhibitors, the end effect is ultimately hyperactivity of CDKs. The role of CDK6 revealed in our studies has important therapeutic implications for treating Ph⁺ ALL and CML patients resistant to STI571 and perhaps for other cancers as well. We tested this hypothesis by taking A-MuLV-transformed cells overexpressing CDK6, which are not as responsive as control A-MuLV-transformed cells to STI571, and treating the cells with both STI571 and PD0332991 to explore the potential of combination treatment. The observation that inhibiting v-Abl and CDK4/6 led to a synergistic G₁ cell cycle arrest suggests that this combination of inhibitors may be useful in a clinical setting. This synergistic effect suggests that an additional pathway converges with the v-Abl pathway in regulating CDK6 levels. The combination of inhibitors was also tested in Ph⁺ ALL cell lines in which the CDK6 inhibitor Cdkn2a was partially or entirely deleted, resulting in constitutively enhanced CDK6 activity. The combination of PD0332991 and STI571 showed increased inhibitory effects in cell proliferation compared to STI571 or PD0332991 alone. These data demonstrate that targeting CDK6 activity directly can cause some Abl-transformed cells to be more sensitive to STI571. Since it is becoming clear that CDK4 and CDK6 have distinct functions in different cell types, it would be useful to develop inhibitors directed specifically to each CDK. In addition, we suggest that it would be worthwhile to screen CML patients with STI571-resistant disease but normal levels of unmutated BCR-Abl for misregulated CDK6 expression.