GSK-3 Inhibition Sensitizes Acute Myeloid Leukemia Cells to 1,25D-Mediated Differentiation

Novel therapeutic approaches are urgently needed for acute myeloid leukemia (AML). Although AML is the most common form of acute leukemia in adults, only 20% of patients over the age of 56 years survive 2 years (1). Despite the poor outcomes, there has not been a change in standard AML therapy in over the age of 40 years. The standard AML chemotherapeutics, typically an anthracycline and cytarabine, target rapidly dividing cells instead of the primary pathophysiology of AML, a maturation arrest. Through targeting the differentiation arrest, myeloid leukemia cells can be forced to mature into cells that lose their proliferative capacity without the necessity for overt cytotoxicity. By targeting the maturation arrest instead of rapidly dividing cells, a more efficacious and less toxic therapeutic regimen may be developed.

All-trans retinoic acid (ATRA) is an AML differentiation agent that has revolutionized the management of an uncommon subtype of AML, acute promyelocytic leukemia (APL). Regimens containing ATRA lead to the long-term survival and presumed cure of the majority of APL patients (2). Unfortunately, ATRA has not been found to be clinically useful for non-APL leukemia. Because of the success of ATRA in a subtype of AML, we and others have searched for approaches to enhance ATRA's activity in non-APL leukemia (3). For example, inhibiting the kinase glycogen synthase kinase 3 (GSK3) was found to lead to a dramatic enhancement of ATRA-mediated AML differentiation due to the removal of GSK3-mediated inhibitory phosphorylation of ATRA's receptor (4, 5). GSK3 is a constitutively active serine/threonine kinase, that is important in signaling pathways involved in the regulation of cell fate, protein synthesis, glycogen metabolism, cell mobility, proliferation, and survival (6–8). The strategy of combining GSK3 inhibition and ATRA for non-APL leukemia is currently under clinical investigation (clinicaltrials.gov NCT01820624).

Similar to ATRA, 1,25-dihydroxyvitamin D3 (1,25D) and related analogues have widely been considered promising therapy for cancer, particularly AML, due to their ability to differentiate leukemic cells. Despite the high promise of 1,25D in preclinical studies, clinical trials have been disappointing likely due to dose-limiting hypercalcemia that occurs at relatively low doses (9). Because of the role of GSK3 in ATRA differentiation, we investigated the effects of GSK3 inhibition on 1,25D differentiation. Surprisingly, we found that GSK3 is also able to regulate 1,25D-mediated signaling but in a very different manner as ATRA signaling. GSK3 inhibition leads to the hyperphosphorylation of the vitamin D receptor (VDR), leading to increased differentiation. As specific GSK3 inhibitors have started to enter clinical use for cancer, our novel approach to differentiate AML cells using low doses of 1,25D with GSK3 inhibition has high clinical potential.

SB415286 (SB) was obtained from Tocris Bioscience. Lithium, NBT, propidium iodide, 1,25D, SP600125, and noble agar were from Sigma-Aldrich. SB was used at 7.5 μmol/L, lithium at 10 mmol/L, and 1,25D at 5 nmol/L (low dose) or 25 nmol/L (high dose or unspecified). p-Serine, p-c-jun, actin, p-JNK, GSK3, p-MEK1/2, and p-p38 antibodies were obtained from Cell Signaling Technology. SRC3, cyclin A, and VDR antibodies were from Santa Cruz Biotechnology. P-208 VDR antibody was from Abcam. OCI-AML3 cells were obtained from DSMZ and 293T, HL-60, THP-1, Monomac-3, U937, and HeLa cells were obtained from ATCC (all cells were purchased in 2010). Upon receiving the cell lines, frozen stocks were prepared within 1 to 2 passages and new stocks were thawed frequently to keep the original condition. The cell lines were passaged for less than 6 months after receipt or resuscitation. They were also routinely authenticated on the basis of growth rate, morphology, and viability and were frequently confirmed to be mycoplasma-free.

Primary AML human bone marrow cells were obtained from the CWRU Hematopoietic Stem cell core facility. The AML samples that had fewer than 80% leukemic cells were purified by flow sorting (FACSAria) using CD34⁺ antibody (BD Biosciences). Guava ViaCount was obtained from Merck Millipore and used according to the manufacturer's directions.

Cells were cultured in RPMI1640 media (Hyclone) with 10% serum, 100 U/mL penicillin, and 100 μg/mL streptomycin.

NBT reduction assay and Wright–Giemsa staining of cytospin preparations was performed in the same manner as described earlier (10). CD11b-PE and CD14-FITC (Biolegend) antibodies were used to measure surface marker expression on a Beckman Coulter FC500 cytometer.

Colony assays were performed as described previously (10).

Cells were fixed in 90% methanol, treated with RNase A (Sigma-Aldrich), stained with propidium iodide (50 μg/mL), and analyzed by flow cytometry.

293T cells were transfected with the indicated constructs or empty vector pLKO (Sigma-Aldrich) and the packaging plasmids pCMVΔR8.74 and pMD.G using lipofectamine (Invitrogen). The GSK3α and GSK3β1 knockdown cells were described previously (4). GSK3β2 knockdown cells were made using a shRNA from Sigma (TRCN0000040000) and confirmed for knockdown (Supplementary Fig. S1). AML cells were infected with the viral supernatant concentrated overnight in PEG in the presence of polybrene and stable cell lines were generated by selection with puromycin (2 μg/mL). AML cells were also directly transfected with xtremegene HP (Roche).

Western blot analyses and immunoprecipitations were performed on cells treated as indicated as described previously (10). The bands were quantitated using ImageJ software and normalized on the basis of the loading control (except for the immunoprecipitations).

HeLa cells were transfected with VDR and RXR. After 72 hours, cells were treated as indicated and nuclear lysates were prepared. Three milligrams of protein was mixed with double-stranded oligonucleotide (5′-AGCTCAGGTCAAGGAGGTCAG-3′) labeled by T4 PNK (NEB), 4 μL electrophoretic mobility shift assay (EMSA) buffer (50 mmol/L HEPES, 375 mmol/L KCl, 12.5 mmol/L MgCl₂, 0.5 mmol/L EDTA, 5 mmol/L DTT, 15% Ficoll), 1 μL of poly dI/dC (1 mg/mL), 1 μL BSA (10 mg/mL) in a total volume of 20 μL for 20 minutes at room temperature. Samples were run on a nondenaturing PAGE gel, dried, and exposed to X-ray film.

OCI and HL-60 cells were transfected with VDRE-luciferase [kindly provided by, Dr. Paul MacDonald Case Western Reserve University (CWRU); Cleveland, Ohio] and Renilla-luciferase and treated as indicated. The cells were lysed with the Bright-Glo Reagent (Promega) and the signal was detected using a Spectramax L luminometer (Molecular Devices).

HeLa, OCI, and HL-60 cells were transfected with Gal4-SRC3, VP16-VDR, and MH100×4-tk-luc plasmids (kindly provided by Noa Noy). Forty-eight hours later, cells were treated for 16 hours and luminescence was measured using the Bright-Glo reagent as described above.

Six-week-old female NSG mice (n = 5 per group) were injected intravenously with 10 × 10⁶ primary human AML cells. Drug treatment was started 3 days after cell injection. Twenty nanograms 1,25-D, or vehicle (20 μL of DMSO and 80 μL of water) were injected intraperitoneally five times a week for 3 weeks. Lithium chow (0.4% lithium carbonate from Teklad) was fed to mice for the time that the mice received the other drugs. The CWRU Animal Research Committee approved all of the animal work used in this study.

HL-60 cells were treated with SB and 1,25D alone or in combination in triplicate for 24 hours and total RNA were isolated using TRIzol (Invitrogen). The cRNA was prepared using the TargetAmp-Nano Kit (Epicenter) and hybridized to the human HT-12 Expression BeadChip (Illumina). The data were normalized and analyzed using the Genomestudio software (Illumina). The pathway analysis was performed using Ingenuity Pathway Analysis software (Ingenuity) for genes with a greater than 1.5-fold change in expression. The microarray data can be found at ArrayExpress Accession# E-MTAB-3196.

Total RNA was isolated from cells treated with SB, 1,25D, or a combination for 24 hours using TRIzol Reagent (Invitrogen). RNA was transcribed into cDNA using the Enhanced Avian RT First Strand Synthesis Kit (Sigma). Relative quantitative RT-PCR was performed in triplicate using the FastStart SYBR Green Master (Roche Diagnostics) on an Applied Biosystems 7500 Fast Real-Time PCR System (Applied Biosystems) by normalizing the results to actin. Primers used for confirmation of microarray data are listed in Supplementary Fig. S2 and were purchased from Sigma.

P values were generated using an unpaired t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001. Error bars depict SD and the results are an average of at least three independent experiments. Unless otherwise noted in the figures, all P values represent the comparison of the combination treatment to treatment with 1,25D and GSK3 inhibitor alone.

In order for 1,25D to be a successful clinical differentiation agent, it is necessary to increase its activity at clinically achievable doses. As GSK3 inhibition has been shown to enhance differentiation effects of ATRA, we investigated the combination of GSK3 inhibition with 1,25D. We found that GSK3 inhibition can significantly sensitize AML cells to differentiate in response to low doses of 1,25D that alone have only weak differentiation inducing effects. Importantly, the levels of 1,25D we utilized fall within the range that have been reported to be achievable in human trials for cancer therapy (11–13). Several of these clinical trials have reported C_max concentrations in the 25 to 30 nmol/L range that corresponds to the high range used in our studies.

Initially, the differentiation effects were assessed using the NBT reduction assy. The NBT assay is a highly specific and commonly used method to quantitate myeloid differentiation. It measures functional differentiation by detecting the respiratory burst capacity, a process that only occurs in differentiated cells (14–18). For example, in HL-60 cells (AML M2 subtype), the combination regimen led to 92% differentiation as opposed to 30% with a low dose of 1,25D alone and 9% with GSK3 inhibition alone using SB (Fig. 1A). In addition to SB, a well-characterized specific and potent GSK3 inhibitor, we also observed similar findings with lithium, which is an FDA-approved GSK3 inhibitor (Fig. 1A).

In addition to HL-60 cells, we observed a similar ability of GSK3 inhibition to enhance 1,25D-mediated differentiation in other AML cell lines such as OCI-AML3 (AML M4 subtype) and Monomac3 (AML M5 subtype; Fig. 1B). Besides assessing differentiation using the NBT reduction assay, we have confirmed AML differentiation by measuring the upregulation of CD14 and CD11b surface expression, commonly used markers for AML differentiation (Fig. 1C–E). For example, CD11b expression in HL-60 cells is 86% with the combination of 1,25D and GSK3 inhibition whereas only 32% and 23% with 1,25D and GSK3 inhibition alone, respectively. Although CD11b is upregulated during granulocytic differentiation, CD14 induction suggests these cells are specifically undergoing monocytic differentiation.

Morphologic assessment of AML cells treated with the combination regimen demonstrates monocytic differentiation as can be seen from increased cytoplasm and altered nuclear morphology including nuclear indentation (Fig. 1F). In addition to AML cell lines, GSK3 inhibition is also able to lead to evidence of differentiation in primary human AML cells (Fig. 1G). For example, 6 of 7 primary AML samples tested showed an enhancement of differentiation when 1,25D was combined with a GSK3 inhibitor. Of note, the sample that did not exhibit a differentiation response to the combination regimen also did not show any differentiation response to 1,25D alone. Though much larger studies are necessary to define specific AML subtypes that would benefit most from this regimen, the samples tested are derived from AML patients representing a variety of AML subtypes and genetic abnormalities (Supplementary Fig. S3). Despite the heterogeneity in AML samples, the majority of patients showed a response to the 1,25D and GSK3 inhibitor regimen.

In addition to differentiation, the combination regimen leads to an enhancement in AML growth inhibition over treatment with either agent alone (Fig. 2A). Consistent with the observed growth inhibition, the combination of 1,25D and GSK3 inhibition leads to effects on the cell-cycle status of leukemic cells. For example, though only modest effects are seen after a low dose of 1,25D or GSK3 inhibition, the combination treatment leads to a decrease in cells in the S-phase and a concomitant increase in cells in the G₀–G₁ phase of the cell cycle (Table 1). For example, the combination leads to 68.0% of cells in the G₀–G₁ phase as compared with 61.6% and 55.7% in the SB- and 1,25D-treated groups, respectively. Consistent with the growth inhibition, the expression of the cell cycle–related protein cyclin A is downregulated more by cotreatment of 1,25D and GSK3 inhibition as opposed to single-agent treatment (Fig. 2B).

As the primary goal of AML differentiation therapy is to permanently prevent the growth of AML cells, colony assays were performed to test for irreversible growth arrest after limited treatment. For this assay, AML cells were exposed to drug for 3 days, the drug was washed off and an equal number of viable cells were plated in soft agar. A dramatic reduction in the number of colonies formed was observed with the combination of 1,25D and GSK3 inhibition as opposed to single-agent treatment in a wide variety of AML cell lines (Fig. 2C and D). For example, in OCI cells, the combination regimen led to only 9.3% of the number of colonies of the vehicle-treated cells as compared with 72.9% and 95.3% in the SB- and 1,25D-treated samples, respectively. These results support the combination regimen leads to significant irreversible growth arrest in viable cells.

Besides cell-based studies, GSK3 inhibition also enhances the ability of 1,25D to demonstrate efficacy in a mouse model system of circulating human AML (Fig. 2E). The primary human AML cells used for this study are derived from a patient with relapsed, treatment refractory non-APL leukemia. Although utilizing clinically achievable levels of either a GSK3 inhibitor (lithium) or 1,25D demonstrate some efficacy, the combination regimen led to prolonged survival. Of note, our mouse studies using lithium demonstrated efficacy with circulating lithium levels (∼0.7 mmol/L; ref. 19) that are well within the normal therapeutic range of lithium for bipolar disease (0.6–1.2 mmol/L; ref. 20). This study further suggests that GSK3 inhibition using an already FDA-approved agent is a promising strategy to achieve a long held goal in AML therapy of enhancing the clinical activity of 1,25D in non-APL leukemia.

Though we utilized well-established GSK3 inhibitors, we also performed genetic experiments to validate the role of GSK3 in AML differentiation to account for potential off-target effects. Utilizing GSK3β knockdown cells, we observed the GSK3β-deficient AML cells are sensitized to 1,25D-mediated differentiation reinforcing that GSK3 does play an important role in 1,25D-mediated differentiation (Fig. 3A).

To elucidate how GSK3 is regulating 1,25D-mediated differentiation, we assessed whether GSK3 impacted the transcriptional activity of the 1,25D receptor (VDR) as GSK3 is a kinase that is known to modulate the activity of several nuclear receptors (21, 22). Although GSK3 inhibition with lithium or SB did not lead to significant effects on VDR transcriptional activity, the combination of GSK3 inhibition and 1,25D led to a marked increase in VDR transcriptional activity (Fig. 3B). In addition to chemical GSK3 inhibitors, cells with stable knockdown of GSK3β, but not GSK3α, exhibited a higher level of VDR transcriptional activity after 1,25D treatment as compared with parental cells (Fig. 3C).

To further explore how GSK3 modulates VDR transcriptional activity, we assessed whether GSK3 could directly bind to VDR through coimmunoprecipitations or phosphorylate VDR through in vitro kinase assays. We did not detect any direct interactions with VDR or the ability of GSK3 to phosphorylate VDR (data not shown). In addition, we assessed whether GSK3 impacted the ability of VDR to bind DNA, which is a necessary prerequisite to activating transcription. Through the use of gel shift studies, GSK3 inhibition was not found to contribute to the ability of VDR to bind DNA (Fig. 4A).

As VDR transcriptional activity is largely controlled by its interactions with coregulators, we assessed the impact of GSK3 on the interaction of VDR with a key coactivator SRC3. Utilizing both a mammalian two-hybrid assay as well as coimmunoprecipitation studies, the combination of 1,25D and GSK3 inhibition was found to increase the association of SRC3 with VDR providing a mechanistic insight into how GSK3 enhances VDR activity (Fig. 4B–D).

As the association of VDR with its coregulators is often controlled by changes in phosphorylation of VDR, we explored whether the combination of GSK3 inhibition and 1,25D could lead to changes in the phosphorylation of VDR. Though low dose 1,25D and SB did not lead to significant changes, we observed an induction of serine phosphorylation of VDR upon combination treatment (Fig. 4E). Though VDR is known to have multiple phosphorylation sites, serine-208 is known to be a key site that is phosphorylated that impacts VDR transcriptional activity, but not DNA binding (23). Therefore, we tested specifically for changes in serine-208 phosphorylation. The combination of 1,25D and GSK3 inhibition leads to induction of serine-208 phosphorylation on VDR providing further insight into how this regimen enhances VDR activity and subsequent differentiation (Fig. 4F).

Consistent with our biologic findings, we identified downstream AML differentiation and growth inhibition pathways can be significantly modulated by the combination of 1,25D and GSK3 inhibition. To investigate these pathways, we performed a gene microarray study to look at early expression changes after drug treatment. From this study, significantly more genes were modulated at least 1.5-fold in the combination group (179 genes) as compared with 1,25D alone (67 genes) and GSK3 inhibition alone (22 genes) after 24 hours of treatment (Fig. 5A). As expected, the array revealed changes in expression of a number of well-characterized genes involved in AML differentiation and growth inhibition (e.g., CEBPB, CEBPE, MYC, CD11b, CD14, and BTG1; Supplementary Fig. S4). The array also revealed a stronger regulation of numerous well-established VDR target genes with the combination regimen as opposed to single-agent treatment. We validated the enhanced upregulation of several established VDR target genes such as CYP24A, HBEGF, and FBP1 by real-time PCR (Fig. 5B; ref. 24).

Pathway analysis was performed to elucidate signaling pathways that may be responsible for the enhanced differentiation in the combination regimen. As expected, 1,25D alone led to the activation of known 1,25D-dependent pathways such as ERK (25). Interestingly, pathway analysis showed that the JNK signaling pathway was only significantly activated by the combination treatment and not low doses of either agent alone (Supplementary Fig. S5: as seen by JNK, MAPK8 activation, and SP600125 inhibition).

To assess the role of JNK in the differentiation response to GSK3 inhibition and 1,25D, we assessed for the activation of JNK as well as the other MAPK proteins, p38 and MEK1/2 by Western analysis. Consistent with the pathway analysis, we observed higher activation of JNK, but not p38 or MEK1/2, after combination treatment as opposed to single treatment using SB and 1,25D (Fig. 5C and Supplementary Fig. SS6). To explore the impact of JNK signaling on AML differentiation in response to GSK3 inhibition and 1,25D, we impaired JNK activation in leukemia cells. After JNK inhibition, the differentiation effects of the combination of 1,25D and GSK3 inhibition were impaired as measured using both NBT reduction and CD14 induction, suggesting that JNK plays an important role in this differentiation process (Fig. 5C). In addition to JNK activation itself, we also observed a dramatic increase in the phosphorylation of the JNK target protein c-jun in the combination treated cells as opposed to leukemia cells treated with the single agents (Fig. 5D).

Overall these studies reveal a model through which GSK3 can modulate differentiation induced by 1,25D (Fig. 6). As GSK3 is a constitutively active kinase, it suppresses VDR activity under basal conditions, which limits the ability of 1,25D to induce differentiation. The suppression of VDR activity is due to a modulation of its phosphorylation, which likely impacts its interaction with the coactivator SRC3. Upon GSK3 inhibition, VDR phosphorylation is induced and SRC3 is recruited allowing enhanced transcriptional activation and differentiation induction. As the combination regimen allows the use of low doses of 1,25D, the combination of GSK3 inhibition and 1,25D is a promising strategy for future clinical testing.

AML differentiation therapy can result in cells that are irreversibly growth arrested and eventually die without the necessity for overt cytotoxicity. We and others have been actively searching for novel differentiation therapies to attempt to improve the prognosis of AML patients (3, 26, 27). In this study, we identified a novel therapeutic approach for AML differentiation and elucidated its mechanisms of action. We demonstrate that GSK3 is a negative regulator of 1,25D-mediated AML differentiation. Inhibition of GSK3 through pharmacologic or genetic approaches was found to significantly enhance AML differentiation in a variety of AML cell lines and patient samples. In addition, the combination of 1,25D and GSK3 inhibition was found to enhance growth inhibition and dramatically reduce the ability of AML cells to form colonies in semisolid media. Finally, this combination regimen showed promise in an animal model of circulating AML using clinically achievable doses of lithium.

Mechanistic studies have provided significant insights into how GSK3 regulates 1,25D signaling. In particular, GSK3 inhibition leads to an enhanced phosphorylation of VDR. Presently, it is not clear how GSK3 inhibition leads to VDR phosphorylation but it is likely due to GSK3-dependent regulation of an intermediate kinases. Several kinases in addition to JNK have been shown to modulate VDR phosphorylation. This enhanced VDR phosphorylation leads to the recruitment of the coactivator SRC-3 and increases the transcriptional activity of VDR and differentiation induction. Besides VDR transcriptional activity, pathway analysis also revealed that activation of JNK signaling also plays a role in the potentiation of 1,25D signaling by GSK3 inhibition. The role of JNK signaling in AML differentiation in response to 1,25D alone has been well established (28). However, this work illustrates the role of JNK in modulating the combination effects of GSK3 inhibition and 1,25D.

Besides identifying a novel therapeutic strategy with clinical potential, our study has revealed a novel role for GSK3 in 1,25D receptor signaling. Interestingly, GSK3 is also known to function as either a positive or negative regulator of other steroid/nuclear receptors such as the estrogen, androgen, and retinoic acid receptors (4, 21, 22). It will be interesting to see whether GSK3 inhibition can enhance biologic effects of 1,25D signaling in other contexts beyond leukemia.

GSK3 is also an interesting target as our studies as well as work from others have demonstrated that a wide variety of AML cells from diverse subtypes with different genetic abnormalities tend all to be sensitive to GSK3 inhibition (5, 29–31). One major advantage of targeting GSK3 is that GSK3 inhibition preferentially inhibits the growth of AML, but not normal hematopoietic progenitor cells. This activity is in stark contrast to the standard AML chemotherapeutics that often lead to significant bone marrow toxicities. Though further studies are necessary, it is known that the dysregulation of GSK3 in AML (as well as other cancers) is common and does not appear to depend on particular genetic changes (5, 29–33). In addition, it has been reported that vitamin D signaling is intact in AML across subtypes therefore providing a specific strategy that may be more universal than many targeted approaches (34).

In conclusion, we have identified that GSK3 is a novel regulator of 1,25D signaling and that inhibition of GSK3 in combination with 1,25D in AML is a promising strategy for clinical testing.

No potential conflicts of interest were disclosed.

Conception and design: K. Gupta, D.N. Wald

Development of methodology: K. Gupta, J. Ignatz-Hoover, S. Moreton, D.N. Wald

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Gupta, T. Stefan, S. Moreton, G. Parizher, D.N. Wald

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Gupta, T. Stefan, G. Parizher, D.N. Wald

Writing, review, and/or revision of the manuscript: K. Gupta, Y. Saunthararajah, D.N. Wald

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Gupta, S. Moreton, D.N. Wald

Study supervision: D.N. Wald

This work was supported by grants from the Leukemia and Lymphoma Society and American Cancer Society (D.N. Wald).

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.