Activated Ca²⁺/Calmodulin-Dependent Protein Kinase IIγ Is a Critical Regulator of Myeloid Leukemia Cell Proliferation

Ca²⁺ signaling is an important component of signal transduction pathways regulating B and T lymphocyte proliferation, but the functional role of Ca²⁺ signaling in regulating myeloid leukemia cell proliferation has been largely unexplored. We observe that the activated (autophosphorylated) Ca²⁺/calmodulin-dependent protein kinase IIγ (CaMKIIγ) is invariably present in myeloid leukemia cell lines as well as in the majority of primary acute myelogenous leukemia patient samples. In contrast, myeloid leukemia cells induced to terminally differentiate or undergo growth arrest display a marked reduction in this CaMKIIγ autophosphorylation. In cells harboring the bcr-abl oncogene, the activation (autophosphorylation) of CaMKIIγ is regulated by this oncogene. Moreover, inhibition of CaMKIIγ activity with pharmacologic agents, dominant-negative constructs, or short hairpin RNAs inhibits the proliferation of myeloid leukemia cells, and this is associated with the inactivation/down-regulation of multiple critical signal transduction networks involving the mitogen-activated protein kinase, Janus-activated kinase/signal transducers and activators of transcription (Jak/Stat), and glycogen synthase kinase (GSK3β)/β-catenin pathways. In myeloid leukemia cells, CaMKIIγ directly phosphorylates Stat3 and enhances its transcriptional activity. Thus, CaMKIIγ is a critical regulator of multiple signaling networks regulating the proliferation of myeloid leukemia cells. Inhibiting CaMKIIγ may represent a novel approach in the targeted therapy of myeloid leukemia. [Cancer Res 2008;68(10):3733–42]

The myeloid leukemias, including acute myelogenous leukemia (AML) and chronic myelogenous leukemia (CML), are neoplastic disorders characterized by the aberrant proliferation and differentiation of myeloid precursor cells. AML is a remarkably heterogenous disease and is associated with multiple different types of genetic mutations. These mutations are generally of two functional types, including those which (1) interfere with the differentiation of myeloid precursors and (2) dysregulate the proliferation and viability of these myeloid precursors (1). For example, AML-specific chromosome translocations and mutations involving the retinoic acid receptor α (RARα; ref. 2), AML-1 (3), and C/EBPα (4) genes interfere with the differentiation of myeloid precursors, whereas mutations involving tyrosine kinase receptors such as c-kit (5) and flt-3 (6) dysregulate their proliferation and viability. Other specific signal transduction pathways that are aberrantly activated in AML include those involving Ras/Raf/mitogen-activated protein kinase (MAPK; ref. 7), Stat 3 (8), SRC-related kinases (9), and phosphatidylinositol 3-kinase (PI3K)/Akt (10). Similarly, in CML, the oncogenic bcr-abl tyrosine kinase that is generated by the Philadelphia chromosome is associated with the activation of multiple downstream signal transduction pathways, including Stat5 (11), MAPK (12), PI3K (13), and SRC family kinases (14).

The Ca²⁺/calmodulin-dependent protein kinases (CaMK) are ubiquitously expressed multifunctional serine/threonine protein kinases that function through Ca²⁺ signaling to regulate the development and activity of many different cell types (15). CaMKs include CaMKI, CaMKII, and CaMKIV, each of which has multiple isoforms. A critical feature shared by members of the CaMK family involves the regulation of their enzymatic activity by changes in intracellular Ca²⁺ concentration. Ca²⁺ binding to the ubiquitously expressed calmodulin triggers a conformational change that enhances its binding to the different CaMKs. This binding of Ca²⁺/calmodulin in turn triggers a conformational change in the CaMKs leading to activation of these enzymes.

The most widely studied of these Ca²⁺-regulated enzymes is CaMKII, which consists of four homologous but distinct genes (CaMKIIα, CaMKIIβ, CaMKIIγ, and CaMKIIδ). One of the critical functional characteristics of CaMKII is its distinct holoenzyme structure consisting of 12 identical subunits (16). Activation of individual subunits of this dodecameric holoenzyme by Ca²⁺/calmodulin leads to the phosphorylation of adjacent enzyme subunits at Thr²⁸⁶ (for the α isoform) or at Thr²⁸⁷ (for the β, γ, and δ isoforms; ref. 17). This CaMKII autophosphorylation leads to increased enzymatic activity and markedly diminishes the requirement for Ca²⁺/calmodulin binding to maintain enzymatic activity (18). Thus, the autophosphorylated CaMKII is autonomous and relatively independent of Ca²⁺ concentration and Ca²⁺/calmodulin regulation.

CaMKII comprises approximately 1% to 2% of total brain protein, and previous studies have identified an important role of this enzyme in regulating neuronal cell development and activity (19). Other studies using pharmacologic inhibitors of the CaMKs have indicated a role for CaMKII in regulating the cell cycle progression of cultured NIH3T3 fibroblasts (20, 21). In contrast, studies determining the role of CaMKII in regulating hematopoiesis have been relatively limited. In T lymphocytes, CaMKII is involved in regulating cytokine expression (22) and T-cell receptor–triggered cell proliferation (23) as well as the generation of memory T cells (24). In previous studies, we have observed that CaMKIIγ is the CaMKII isoform that is preferentially expressed in myeloid cells, and this enzyme inhibits the retinoic acid–induced differentiation of myeloid leukemia cell lines by phosphorylating and inhibiting the transcriptional activity of the RARα (25). In these previous efforts, we observed that CaMKII inhibitors enhance the differentiation of certain myeloid leukemia cell lines, but this effect was only noted in leukemias that were responsive to retinoic acid (25). In the present study, we extend these studies beyond the retinoic acid–responsive myeloid leukemias and describe a critical and central role of CaMKIIγ in regulating the proliferation of a wide variety of myeloid leukemia cells. We observe that autophosphorylated CaMKIIγ is commonly present in myeloid leukemias. Inhibitors of CaMKIIγ, including pharmacologic agents and dominant-negative constructs, inhibit myeloid leukemia cell proliferation, and this is associated with the inactivation of multiple signal transduction pathways, including p44/42 MAPK, Stat3/Stat5, and glycogen synthase kinase 3β (GSK3β)/β-catenin. Moreover, we observe that Stat3 is directly phosphorylated by CaMKII at Ser⁷²⁷. These observations highlight the importance of CaMKII in regulating the proliferation of myeloid leukemia cells and suggest that inhibiting CaMKIIγ activity may be of significant value in the targeted therapy of myeloid leukemia.

Cell lines. All leukemia cell lines are cultured at 37°C in RPMI 1640 supplemented with 5% heat-inactivated FCS.

Antibodies and Western blots. Protein extracts, immunoprecipitations, and Western blots were performed as previously detailed (26). Antibodies against CaMKI, CaMKIIγ, CaMKIV, c-myc, Stat3, Stat5, and calmodulin as well as the phosphospecific CaMKII (Thr²⁸⁶ and Thr²⁸⁷) and phosphospecific CaMKI (Thr¹⁷⁷) antibodies were from Santa Cruz Biotechnology. Antibodies against phosphorylated Stat1 (Ser⁷²⁷), phosphorylated Stat3 (Ser⁷²⁷ and Tyr⁷⁰⁵), phosphorylated Stat5 (Tyr⁶⁹⁴), phosphorylated p44/42 MAPK (Thr²⁰²/Tyr²⁰⁴), p44/42 MAPK, phosphorylated p38 MAPK (Thr¹⁸⁰/Tyr¹⁸²), and phosphorylated GSK3β (Ser⁹) were from Cell Signaling Technology. Antibodies against Stat1, GSK3β, β-catenin, protein phosphatase 2A catalytic subunit (PP2Ac), and cyclin D1 were from BD Biosciences.

Chemicals. Different inhibitors of specific signal transduction pathways, including AG490, JAK 3 inhibitor II, PP1, PD98059, UO 126, SB 203580, wortmannin, LY 294002, c-Jun NH₂-terminal kinase (JNK) inhibitor II, Raf kinase inhibitor I, pho kinase inhibitor, RO-31-8220, GÖ 6983, cucurbitacin I, AMP kinase inhibitor compound C, KN62, KN93, and STO609 were all from Calbiochem. All-trans retinoic acid (ATRA), 12-O-tetradecanoylphorbol-13-acetate (TPA), arsenic trioxide, and SB216367 were from Sigma.

Patient AML and normal CD34⁺ samples. Patient AML samples were obtained from the Children's Oncology Group bank of AML cells. Ficoll gradient centrifugation was used to isolate mononuclear cells from bone marrow aspirates from newly diagnosed patients with AML. These cells consisted of >80% blasts by morphology and were stored frozen under liquid nitrogen. Normal CD34⁺ cells were immunopurified from granulocyte colony-stimulating factor–mobilized, leukophoresed peripheral blood stem cell donors. Appropriate Institutional Review Board approval was obtained for using these patient and normal donor samples, which are anonymized such that individuals cannot be identified.

Glutathione S-transferase fusion proteins. The coding region of the human Stat3α gene was amplified by reverse transcription-PCR (RT-PCR) and subcloned into the glutathione S-transferase (GST) fusion protein expression vector pGEX5x.1. The GST-Stat3 fusion protein was purified from bacteria extracts using glutathione 4B-Sepharose (Amersham Biosciences).

Expression vectors. A cDNA harboring the coding sequence of amino acids 1 to 290 of CaMKIIγ was cloned into the LXSN retroviral expression vector (M28248). Using site-directed mutagenesis, we mutated Lys⁴³ of this construct to methionine to generate a kinase-deficient CaMKII mutant labeled kdCaMKII.

CaMKIIγ cDNA amplification and sequencing. Total cellular RNA from the leukemia cell lines and normal tissue was isolated by Trizol reagent (Invitrogen). First-strand cDNA was made by the superscript amplification system (Invitrogen) with the oligo(dT)_12-18 primer. To determine the expression of the different CaMKIIγ isoforms, we performed RT-PCR using primers flanking the variable region of CaMKIIγ, including 5′-CGTCAGGAGACTGTGGAGTGT (sense) and 5′-TCACTGCAGCGGTGCGGCAGG (antisense).

To amplify and sequence the complete CaMKIIγ coding sequence, the cDNA from the HL60 and NB4 cell lines was amplified in separate reactions with two different primer pairs flanking the NH₂-terminal and COOH-terminal regions of the CaMKIIγ coding sequence. The sequence of these primers as well as the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) control primers are available on request. The PCR products were isolated and subcloned into the TA cloning vector for DNA sequencing.

Transient transfections and luciferase assays. K562 cells were transiently transfected using electroporation. Cells (5 × 10⁶) were electroporated at 500 μF, 250 V, and cultured for 24 h under standard conditions. NIH3T3 cells were transiently transfected using Lipofectamine 2000 (Invitrogen). Stat3 transcriptional activity was determined in K562 cells transfected with the pSIE-tk-LUC luciferase reporter, which harbors a Stat3-binding site (CCTTAATCCTTCTGGGAATTCTGGCT). The β-gal reporter driven by the β-actin promoter was used as an internal control for transfection efficiency.

CaMKIIγ RNA interference constructs. Three LXSN-based plasmids harboring three different synthetic human CaMKIIγ oligos designed to generate short hairpin RNA (shRNA) hairpins were constructed as previously detailed (25). We synthesized and inserted an additional oligo generating a shRNA corresponding to human CaMKIIγ (NM_172171) coding region (954–976) into the pLL3.7 lentiviral vector plasmid, which harbors a green fluorescent protein (GFP) marker (27). These four plasmids were simultaneously electroporated into K562 cells, which were selected in G418 (800 μg/mL) for 10 d. GFP⁺ cells from this culture were then fluorescence-activated cell sorting enriched and cell lysates were subjected to Western blot analysis.

Construction and use of doxycycline-inducible expression vectors. A full-length CaMKIIγ cDNA was mutated at Lys⁴³ to methionine to render it kinase deficient. It was then FLAG tagged at the NH₂ terminus and cloned into the pTRE-Tight expression vector, which harbors a tet-responsive element.¹

This plasmid was electroporated into K562 cells together with the Clontech pTet-On Advanced plasmid, which codes for a doxycycline-regulated transactivator protein. The electroporated cells were selected in G418 (800 μg/mL) for 10 d. Cell lysates for Western blot analysis were obtained from these transfected cells as well as from the same cells treated for 2 d with doxycycline (1 μg/mL).

Activated CaMKIIγ is frequently present in myeloid leukemia cell lines and primary AML cells. We previously observed that CaMKIIγ is the CaMKII isoform that is predominantly expressed in myeloid cells (25). The binding of Ca²⁺/calmodulin to the inactive CaMKIIγ triggers autophosphorylation of this enzyme at Thr²⁸⁷, leading to autonomous enzyme activity that is relatively independent of further Ca²⁺/calmodulin regulation (17, 28, 29). This autophosphorylated CaMKIIγ can be detected using a CaMKII Thr²⁸⁷ phosphospecific antibody. Using this antibody on Western blots, we observed that the autophosphorylated (activated) CaMKIIγ was invariably present in different human leukemia cell lines (Fig. 1A). To further investigate the role of CaMKIIγ in the pathogenesis of human leukemia, the present study focuses on the myeloid leukemias, and we observe that the activated, autophosphorylated CaMKIIγ is also present at varying levels in the majority (11 of 16) of primary patient AML samples (Fig. 1B). In contrast, Western blots did not reveal any detectable autophosphorylated CaMKIIγ in three different normal peripheral WBC samples (data not shown).

Because alternative splicing within a variable region of CaMKIIγ generates several different isoforms (30), we determined whether the myeloid leukemia cells exhibiting CaMKIIγ autophosphorylation expressed any unusual CaMKIIγ isoform(s). A RT-PCR analysis using primers flanking this alternatively spliced CaMKIIγ variable region (Fig. 1C) indicated that, in normal CD34⁺ cells as well as in leukemia cells, the previously described CaMKIIγ1 and CaMKIIγ3 isoforms were predominantly and commonly expressed (Fig. 1C). In addition, we determined whether the activation (autophosphorylation) of CaMKIIγ observed in the myeloid leukemia cell lines was associated with any particular mutations within the CaMKIIγ coding region in these cells. We PCR amplified and sequenced the entire CaMKIIγ coding region cDNA from the HL60 and NB4 myeloid cell lines. The CaMKIIγ cDNA coding sequence from both these cell lines matched the Genbank CaMKIIγ1 and CaMKIIγ3 sequences. Thus, the CaMKIIγ activation exhibited by the myeloid leukemia cells is not associated with the expression of any aberrantly spliced CaMKIIγ isoform(s) nor with any particular CaMKIIγ coding region mutation(s).

Terminal differentiation/apoptosis of myeloid leukemia cell lines is associated with decreased CaMKIIγ activation. ATRA induces the terminal granulocytic differentiation of certain myeloid leukemia cell lines, including HL60 (31) and NB4 (32), and ATRA also induces terminal monocytic differentiation of the U937 cell line (33). We used Western blots to compare CaMKIIγ expression in these cells at different times after this ATRA-induced terminal differentiation. We observed in all three of these cell lines that ATRA-induced terminal differentiation was associated with a marked decrease in the levels of activated CaMKIIγ, whereas total CaMKIIγ levels remain essentially unchanged (Fig. 2A). This decrease in activated CaMKIIγ was most prominent 3 to 5 days after differentiation induction, which, as previously observed (31, 32), is the time that these cells undergo terminal differentiation and cease to proliferate. This ATRA-induced decrease in activated CaMKIIγ was not associated with any decrease in cellular calmodulin levels (data not shown). This decreased CaMKIIγ activation was ATRA dose dependent (Fig. 2B,, rows 1–3) and seemed to closely correlate with differentiation induction because a similar decrease was not observed in ATRA-treated HL60R cells, which harbor an inactivating point mutation in RARα that renders them unresponsive to ATRA-induced granulocytic differentiation (Fig. 2B,, row 4; ref. 34). Similarly, other myeloid leukemia cell lines that are insensitive to ATRA-induced differentiation, including K562, KG1, THP1, and KCL22, did not exhibit any decrease in activated CaMKIIγ expression following ATRA induction (Fig. 2B,, rows 5–8). A similar marked decrease in activated CaMKIIγ was also observed in HL60 cells treated with TPA, which induces terminal macrophage-like differentiation of these cells (Fig. 2C; ref. 35). Similarly, we observe that the terminal differentiation/apoptosis induced by arsenic trioxide in promyelocytic leukemia cell lines, such as NB4 (36), is accompanied by a down-regulation in phosphorylated CaMKII (Fig. 2D). Thus, the terminal differentiation of myeloid leukemia cells is associated with a marked down-regulation of the autophosphorylated CaMKIIγ.

Oncogenic bcr-abl regulates CaMKIIγ activation. In contrast to its lack of response to ATRA, the K562 cell line, derived from a patient with CML, undergoes proliferative arrest following exposure to imatinib (Gleevec), which is a potent inhibitor of the tyrosine kinase activity of the bcr-abl oncogene. We observed that this imatinib-induced proliferation arrest is accompanied by a rapid, marked decrease in autophosphorylated CaMKIIγ, whereas there was no significant change in the total levels of CaMKIIγ following this imatinib exposure (Fig. 3A).

The marked down-regulation of CaMKII activation (autophosphorylation) by the bcr-abl inhibitor imatinib observed in K562 cells suggests that the activation of CaMKIIγ in these CML cells requires bcr-abl activity. To confirm this, we used the TonB210.1 cell line, which is a Baf3-derived line whose proliferation/viability is dependent on either interleukin-3 (IL-3) alone or the tet-regulated expression of bcr-abl (tet-on; ref. 37). We observe in the TonB210.1 cells that the decreased bcr-abl expression associated with tet withdrawal is also accompanied by a marked and relatively rapid (within 4 h) reduction in the levels of activated (autophosphorylated) CaMKII, whereas total CaMKIIγ levels remain unchanged (Fig. 3B). To determine if enhanced bcr-abl expression can initiate CaMKII activation, we first deprived the IL-3–dependent TonB210.1 cells of IL-3 for 24 h, which induced cell proliferative arrest associated with decreased autophosphorylated CaMKIIγ (Fig. 3C,, lane 1). Treatment of these cells with tet restores CaMKII autophosphorylation (Fig. 3C,, compare lanes 1 and 2), and this phosphorylation was inhibited in a dose-dependent manner by the CaMK chemical inhibitors KN62 and KN93 (Fig. 3C,, lanes 3–6), which interfere with the binding of the Ca²⁺/calmodulin complex to the CaMKs (38). In contrast, this bcr-abl–mediated enhancement of CaMKIIγ phosphorylation was not inhibited by the CaMK kinase (CaMKK) chemical inhibitor STO609 (Fig. 3C , lanes 7 and 8). Taken together, these observations with the K562 and TonB210.1 cells indicate that oncogenic bcr-abl directly or indirectly activates CaMKIIγ and that this bcr-abl–induced CaMKIIγ activation is dependent on Ca²⁺/calmodulin binding to CaMKIIγ.

Effect of inhibitors of different signal transduction pathways on CaMKIIγ activation. The above observations indicate that CaMKII activation is downstream of bcr-abl, but we do not know whether this oncogene directly or indirectly (through one or more of its downstream effectors) mediates this activation. The molecular basis for bcr-abl–mediated cell transformation is complex and involves several specific downstream signal transduction pathways activated by bcr-abl, including Stat5 (11), Ras/Raf1/MAPK (12), PI3K (13), and SRC family kinases (14). Indeed, the cooperative activation of these pathways may be necessary for full bcr-abl–mediated transformation (39). We used specific chemical inhibitors to determine whether any of these signal transduction pathways downstream of bcr-abl might be involved in the bcr-abl–mediated activation of CaMKIIγ in K562 cells. In contrast with the Gleevec-induced down-regulation of CaMKIIγ activation (Fig. 3D,, lane 17), none of these chemical inhibitors reduced CaMKIIγ activation, including those that target JAK/Stat (Fig. 3D , lanes 3, 4, and 16), PI3K (lanes 9 and 10), p38 (lane 8), JNK (lane 11), Ras/Raf/MAPK/extracellular signal-regulated kinase (ERK) kinase/ERK (lanes 6, 7, and 12), protein kinase C (lanes 14 and 15), AMP kinase (lane 18), and SRC family kinases (lane 5). Thus, the bcr-abl–mediated activation of CaMKIIγ does not seem to involve any of the signal transduction pathways that are known downstream effectors of bcr-abl.

Inhibitors of CaMKII inhibit the proliferation of myeloid leukemia cell lines. To determine whether inhibiting CaMKIIγ activity might have any effect on myeloid leukemia cell proliferation/viability, we treated K562 cells with the CaMK inhibitor KN93. We found that KN93 inhibited the proliferation of K562 cells in a dose-dependent manner (Fig. 4A,, i). In contrast, STO609, a potent small-molecule inhibitor of CaMKK (40), did not significantly inhibit K562 proliferation nor did KN92, an inactive structural analogue of KN62 (Fig. 4A,, ii). The KN93-induced inhibition of K562 cell proliferation was accompanied by a reduction in the levels of the activated (autophosphorylated) CaMKIIγ (Fig. 5A,, row 1). KN93 did not affect K562 viability, and no effect on the morphologic differentiation of these cells was observed. A similar KN93-mediated inhibition of cell proliferation without any significant effect on cell viability or morphologic differentiation was also noted in other cultured myeloid leukemia cell lines, including KG1, KCL22, THP1, and Kasumi (Fig. 4B).

To further assess the role of CaMKIIγ in regulating myeloid leukemia cell proliferation, we transiently transfected K562 cells with an expression vector harboring a truncated CaMKIIγ construct mutated at Lys⁴³, a conserved residue near the ATP-binding site of this enzyme. This mutation results in a “kinase-dead” enzyme (41), and expression of this mutant inhibits endogenous CaMKII activity in a dominant-negative manner (42). We observed a marked reduction in K562 colony formation in cells transfected with this mutated CaMKIIγ compared with cells transfected with vector alone (Fig. 4C). Thus, both pharmacologic and dominant-negative construct inhibition of CaMKII results in inhibition of myeloid leukemia cell proliferation.

Pharmacologic, dominant-negative or shRNA-mediated inhibition of CaMKIIγ is associated with the down-regulation of multiple signal transduction pathways. We used phosphospecific antibodies to determine whether the KN93-induced inhibition of K562 cell proliferation (Fig. 4A,, i) was associated with any changes in specific phosphoprotein signal transduction pathways. As noted above, KN93 treatment of K562 cells is associated with a time-dependent reduction in CaMKIIγ phosphorylation (Fig. 5A,, row 1). This reduction was associated with a decrease in the activation (phosphorylation) of several different critical components of signal transduction pathways, including phosphorylated p44/42 MAPK (Fig. 5A,, row 3), p38 (row 5), phosphorylated Stat3 (Tyr⁷⁰⁵ and Ser⁷²⁷; rows 10 and 11), and phosphorylated Stat5 (Tyr⁶⁹⁴; row 13). These changes induced by KN93 were not accompanied by any change in the total levels of these phosphorylated proteins (Fig. 5A). These observations indicate that, in K562 myeloid cells, activated CaMKIIγ is involved in regulating the activation (phosphorylation) of p44/42 MAPK, Stat3/Stat5, and p38.

We also observe that the KN93-induced down-regulation of CaMKII activation is associated with a marked reduction in GSK3β Ser⁹ phosphorylation (Fig. 5A,, row 6). This is of particular interest because GSK3β, a serine/threonine kinase, phosphorylates and promotes the degradation of β-catenin, a key component of the Wnt signaling pathway, which is frequently activated in myeloid leukemia (43, 44). Because phosphorylation of GSK3β at Ser⁹ inhibits the activity of this enzyme (45), we would predict that the KN93-induced reduction in GSK3β Ser⁹ phosphorylation would lead to enhanced GSK3β enzyme activity associated with enhanced degradation of β-catenin. Indeed, a marked reduction in the expression of β-catenin is observed in the KN93-treated K562 cells (Fig. 5B,, row 3). Consistent with this, in the KN93-induced cells, we also observe a reduction of c-myc and cyclin D1 protein levels, both of whose expression is normally up-regulated by β-catenin (Fig. 5B , rows 4 and 5; refs. 46, 47).

KN93 inhibits CaMKII activity by blocking its interaction with the Ca²⁺/calmodulin complex. However, this compound can potentially inhibit other enzymes activated by Ca²⁺/calmodulin, including CaMKI, CaMKIV, CKLiK, as well as the CaMKKs. Nevertheless, several lines of evidence indicate that the effects of KN93 are mediated through CaMKII inhibition rather than through inhibiting any of these other CaMKs. First, CaMKIV is not expressed in K562 cells (Supplementary Fig. S1A), and the CaMKI expressed by K562 is primarily in the inactive (unphosphorylated) form (Supplementary Fig. S1B). Second, there is little if any CKLiK (CaMKIδ) expression in K562 cells compared with other cell types (Supplementary Fig. S1C). Moreover, STO609, a potent small-molecule inhibitor of the CaMKKs, exhibits minimal effects on K562 proliferation (Fig. 4A,, ii), indicating that neither the CaMKKs nor their downstream substrate CaMKI is the target of KN93. In addition, we transduced a FLAG-tagged, Lys⁴³-mutated, dominant-negative CaMKIIγ under control of a doxycycline-inducible promoter [designated kdCaMKIIγ (tet-on)] into K562 cells. Treatment of these cells with doxycycline up-regulated expression of this construct (Fig. 5D,, row 1, lane 4), and this induced expression was temporally associated with inhibition of K562 proliferation (Fig. 5C,, i) together with down-regulation of the activation (phosphorylation) of the MAPK, Stat3/Stat5, and β-catenin pathways (Fig. 5D,, compare lanes 3 and 4). Finally, we used a small interfering RNA approach to knock down the expression of CaMKIIγ in K562 cells. The shRNA-transduced K562 cells with reduced expression of CaMKIIγ (Fig. 5D,, row 2, compare lanes 5 and 6) exhibited reduced cell proliferation (Fig. 5C,, ii), and this is associated with down-regulation of the MAPK, JAK/Stat, and GSK3β/β-catenin pathways (Fig. 5D , compare lanes 5 and 6). Thus, the effects of KN93 on K562 cells are likely mediated through its inhibitory activity on CaMKIIγ rather than through any of the other CaMKs.

CaMKIIγ directly phosphorylates and activates Stat3 in myeloid leukemia cells. In addition to tyrosine phosphorylation of Stat3 at Tyr⁷⁰⁵, Stat3 Ser⁷²⁷ phosphorylation is required for full Stat3 transcriptional activity (48). The activated Stat3 has oncogenic activity (49), and Stat3 activation is frequently observed in myeloid leukemia cell lines (Fig. 6A,, i) as well as in myeloid leukemia patient samples (Fig. 6A,, ii; ref. 8). Our observation that the CaMKII inhibitor KN93 induces a decrease in Stat3 Ser⁷²⁷ phosphorylation (Fig. 5A,, row 11) suggests that Stat3 may be a direct substrate of CaMKIIγ in myeloid leukemia cells. Consistent with this, we observe that 9 of 11 of the primary AML samples that exhibit CaMKIIγ activation (Fig. 1B) also exhibit Stat3 Ser⁷²⁷ phosphorylation, whereas none of the 5 primary AML samples that were negative for CaMKIIγ activation exhibits Stat3 Ser⁷²⁷ phosphorylation (Fig. 6A,, ii). In addition, we observe in the TonB210 cells that the tet-induced bcr-abl induction is associated with a marked increase in Stat3 Ser⁷²⁷ phosphorylation without any change in total Stat3 levels (Fig. 6B,, compare lanes 1 and 2). Moreover, this enhanced Stat3 Ser⁷²⁷ phosphorylation is blocked by KN62 or KN93 (Fig. 6B,, lanes 3 and 4), indicating that a CaMK is involved in this bcr-abl–induced Stat3 Ser⁷²⁷ phosphorylation. We also observe that cucurbitacin I, a small-molecule Stat3 inhibitor, will, as expected, inhibit Stat3-responsive reporter activity, and this reporter activity is also inhibited by KN93 in a dose-dependent manner (Fig. 6C). These latter observations indicate a functional interaction between Stat3 and CaMKII in myeloid leukemia cells.

To further assess the CaMKIIγ/Stat3 interaction, we observed that in K562 cells Stat3 coimmunoprecipitates with CaMKIIγ (Fig. 6D,, i). Moreover, CaMKIIγ phosphorylates Stat3 at Ser⁷²⁷ both in vitro (Fig. 6D,, ii) and in vivo (Fig. 6D,, iii). Together, these observations suggest that Stat3 is likely a direct substrate of CaMKIIγ in myeloid leukemia cells and that the KN93-induced inhibition of K562 cell proliferation (Fig. 4A) may be mediated, at least in part, through its inhibition of Stat3 Ser⁷²⁷ phosphorylation.

The CaMKs are major targets of Ca²⁺-regulated physiologic events, and the role of a particular CaMK (i.e., CaMKII) in regulating both neuronal and cardiac muscle cell development and activity has been well established. Our present study defines a novel and critical role for the activated (autophosphorylated) CaMKII (CaMKIIγ) in regulating the proliferation of myeloid leukemia cells. Indeed, we observe that relatively high amounts of the autophosphorylated CaMKIIγ are invariably present in myeloid leukemia cell lines, and this activation is also observed in the majority of primary patient AML samples. In different in vitro models of myeloid leukemia cell terminal differentiation/growth arrest including those triggered by ATRA, TPA, arsenic trioxide, or Gleevec, the loss of proliferation of the induced leukemia cells is accompanied by a marked reduction in CaMKIIγ activation. This marked down-regulation of CaMKIIγ activation likely directly contributes to the loss of leukemia cell proliferative capacity because we observe that both pharmacologic, dominant negative and shRNA-mediated inhibition of CaMKII expression/activity inhibits myeloid leukemia cell proliferation. Moreover, we observe that CaMKIIγ directly or indirectly regulates multiple signaling pathways previously implicated in leukemia cell proliferation, including the MAPK, JAK/Stat, and GSK3β/β-catenin pathways.

What are the molecular events that trigger and maintain CaMKIIγ constitutive activation (autophosphorylation) in the myeloid leukemia cell lines and primary patient AML samples? The myeloid leukemias are remarkably heterogenous, and diverse molecular events may trigger CaMKIIγ activation in these leukemias. Our observations in cell lines driven by oncogenic bcr-abl indicate that CaMKII activation is downstream of bcr-abl but is not regulated by other bcr-abl–activated signal transduction events involving the MAPK, JAK/Stat, SRC-related kinase, or PI3K pathways (Fig. 3D). The primary trigger of CaMKII activation is its binding to Ca²⁺/calmodulin, whose levels are regulated by intracellular Ca²⁺ concentration. Our observations in TonB210.1 cells suggest that the bcr-abl–mediated CaMKII activation is dependent, at least in part, on Ca²⁺/calmodulin binding to CaMKII because this bcr-abl–induced activation is inhibited by KN62 or KN93, which blocks Ca²⁺/calmodulin binding to CaMKII (Fig. 3C). This suggests that bcr-abl might regulate intracellular Ca²⁺/calmodulin levels, and this would likely involve the regulation by this oncogene of intracellular Ca²⁺ concentration. Nevertheless, how bcr-abl might regulate intracellular Ca²⁺ levels and Ca²⁺/calmodulin activity is currently unknown.

Closely related to the question of how CaMKIIγ is activated in myeloid leukemia cells is how this enzyme becomes inactivated (dephosphorylated) during terminal myeloid differentiation/growth arrest (Fig. 2). Indeed, we observe a prominent decrease in CaMKIIγ activation during ATRA-induced terminal granulocytic differentiation of myeloid leukemia cells and also note that normal granulocytes harbor relatively low levels of the activated CaMKIIγ. This terminal myeloid differentiation is not associated with any decrease in total cell calmodulin concentration to account for the reduced CaMKIIγ activity. One possibility to explain this down-regulation of CaMKIIγ autophosphorylation is that myeloid differentiation is accompanied by changes in Ca²⁺ channel activity with an associated decrease in intracellular Ca²⁺ concentration that reduces the Ca²⁺/calmodulin activation of CaMKII. Another possibility is that an increase in phosphatase activity that targets the activated (autophosphorylated) CaMKIIγ accompanies myeloid differentiation. Consistent with this latter possibility, we have observed that PP2A, a phosphatase known to dephosphorylate CaMKII in neuronal cells (50), coimmunoprecipitates with CaMKIIγ in myeloid cells (data not shown). However, the levels of PP2A do not seem to change during myeloid leukemia cell differentiation (data not shown), and whether the phosphatase activity that specifically targets the activated (autophosphorylated) CaMKIIγ is developmentally regulated during myeloid differentiation is presently unknown.

What are the critical substrates of CaMKIIγ that are involved in regulating myeloid leukemia cell proliferation? In HeLa cells, CaMKII phosphorylation of the cdc25 phosphatase is associated with cdc2 activation and enhanced G₂-M transition (51). Moreover, the localization of CaMKII to the mammalian midbody also suggests a role for this enzyme in regulating cytokinesis (52). We have previously observed that CaMKIIγ directly phosphorylates and inhibits the transcriptional activity of RARα, which is an important regulator of the terminal differentiation of certain myeloid leukemia cells (25). Our present study indicates that Stat3, which exhibits oncogenic activity (49), is directly phosphorylated by CaMKIIγ, and this phosphorylation enhances Stat3 transcriptional activity (Fig. 6). We also observe that CaMKIIγ directly or indirectly promotes the inhibitory phosphorylation of GSK3β, an enzyme that normally down-regulates the expression of cell growth/proliferation factors, including β-catenin, c-myc, and cyclin D1 (Fig. 5B). Thus, CaMKII activity seems to orchestrate a complex interacting network of molecular events that are involved in regulating the proliferation of myeloid leukemia cells.

Our observation that CaMKIIγ is an important regulator of multiple phosphoprotein signaling networks regulating myeloid leukemia cell proliferation suggests that targeting this enzyme might be of therapeutic benefit in human myeloid leukemia. Indeed, we reproducibly observe an inhibition of myeloid leukemia cell proliferation triggered by KN93, a small-molecule inhibitor of CaMKII. However, the CaMKII pharmacologic inhibitors have significant limitations. These compounds inhibit CaMKII activity by interfering with their binding to the Ca²⁺/calmodulin complex. However, once CaMKII has become activated (autophosphorylated), its dependence on further Ca²⁺/calmodulin binding is markedly diminished (18). Thus, these compounds exert considerably less inhibition on CaMKII once it has become activated (autophosphorylated). An “ideal” CaMKIIγ inhibitor would be a small molecule, which would directly inhibit the activated CaMKIIγ by directly targeting its ATP-binding pocket. Such a CaMKIIγ inhibitor has not been identified, but we would predict that it would exert much greater activity in inhibiting leukemia cell proliferation than KN93 and might be of significant value in the targeted therapy of myeloid leukemia.

Grant support: NIH grant RO1 CA118971 (S.J. Collins) and Leukemia and Lymphoma Society grant 6144.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank LeMoyne Mueller for excellent technical assistance.