Cancer Cells Become Susceptible to Natural Killer Cell Killing after Exposure to Histone Deacetylase Inhibitors Due to Glycogen Synthase Kinase-3–Dependent Expression of MHC Class I–Related Chain A and B

Introduction

Attention has recently focused on the use of histone deacetylase (HDAC) inhibitors as anticancer agents. Major classes of HDAC inhibitory compounds include short chain fatty acids (e.g., phenylbutyrate), derivatives of hydroxamic acid [e.g., PXD101, suberoylanilide hydroxamic acid (SAHA), trichostatin A], the benzamide derivative MS-275, and the naturally occurring depsipeptide FR901228 ( 1, 2). Clinical studies have shown that pharmacologic relevant levels of HDAC inhibitors can be achieved safely in humans and that treatment of cancer is possible ( 3, 4).

It is presently believed that the therapeutic efficacy of HDAC inhibitors is linked to induction of apoptosis and growth arrest of cancer cells. The ability to arrest proliferation is critically dependent on induction of the cyclin-dependent kinase inhibitor p21 ( 5, 6) whereas the induction of apoptosis has been attributed to disruption of the cellular redox state by a less understood mechanism, likely involving mitochondrial membrane disruption ( 7, 8). Furthermore, the induction of apoptosis has recently been linked to induction of 15-lipoxygenase-1 ( 9).

Judging by their name, HDAC inhibitors act to increase acetylation of histones, leading to uncoiling of chromatin and thereby presumably regulating a variety of genes implicated in apoptosis and growth inhibition. Although this may certainly be correct, numerous other proteins can be regulated by acetylation including p53, hypoxia-inducible factor 1, Ku-70, p300/CREB binding protein–associated factor, and GATA-1 ( 10– 12), implying that increased acetylation of histones may not be the sole action of HDAC inhibitors.

Natural killer (NK) group 2, member D (NKG2D) is an activating receptor that is expressed on NK cells, CD8 T cells, and γδ T cells ( 13). Naïve CD4 T cells do not express NKG2D; however, CD4⁺/CD28⁻ autoaggressive T cells from patients with rheumatoid arthritis have recently been shown to express NKG2D through an inflammatory cytokine-dependent pathway ( 14).

NKG2D recognizes MHC class I–related chain (MIC) A, MICB, and UL16-binding proteins (ULBP) 1 to 4 in humans ( 13, 15). MICA/B are generally not expressed in normal tissue, with exception of intestinal epithelial cells ( 16, 17); NKG2D ligands are, however, constitutively expressed on some tumor cells and are also induced by cellular stress and viral/bacterial infections ( 16, 18– 20). Tumor cells expressing ligands for NKG2D can become susceptible to NK cell killing even if the tumor cells have normal MHC class I expression ( 13, 21), suggesting that sufficiently high levels of NKG2D stimulation can override the inhibitory signaling provided by MHC class I recognition. Mice lack MIC genes but express the ULBP homologous proteins retinoic acid early inducible-1 (RAE-1), H60, and murine ULBP-like transcript 1, which are ligands for murine NKG2D ( 22– 25). Diefenbach et al. ( 23) have shown that RAE-1-transfected cancer cells can be eliminated in vivo by NK cell and CD8 T-cell activities, which also induced adaptive immunity to the parental tumor cells without RAE-1 expression. Girardi et al. ( 26) showed that mice lacking γδ T cells are highly susceptible to epithelial malignancies and that γδ T cells could kill skin carcinoma cells by a NKG2D-dependent mechanism in vitro. These experiments strongly support a role of NKG2D/NKG2D ligand interaction in tumor rejection and surveillance.

It is thus not surprising that tumor cells and virus-infected cells have evolved mechanisms to obstruct the NKG2D/NKG2D ligand stimulatory axis. Some tumor cells are known to constitutively express MICA on the cell surface; however, these cells also shed MICA in a soluble form possibly by proteolysis ( 27, 28). Besides hindering recognition of the MICA-expressing tumor cells, the soluble MICA leads to down-regulation of NKG2D expression on CD8 T cells, NK cells, and γδ T cells ( 29), thereby further inhibiting generation of antitumor immunity.

Two genes encode the closely related glycogen synthase kinase-3 (GSK-3) proteins, GSK-3α and GSK-3β, with the β form often shown to be the most important regulatory enzyme. GSK-3 activity is involved in diverse cellular actions ranging from regulation of β-catenin (Wnt) signal pathway to stimulation of differentiation, cellular stress, and apoptosis ( 30– 32). GSK-3 is a constitutive active serine-threonine kinase that mainly targets proteins that are previously phosphorylated. GSK-3 is negatively regulated by serine phosphorylation of the NH₂-terminal domain, which makes an autoinhibitory loop interaction with the substrate binding site, hindering binding of otherwise potential prephosphorylated substrates ( 32). The negatively acting NH₂-terminal phosphorylation of GSK-3 is mainly governed by growth factor–induced phosphatidylinositol 3-kinase/AKT activity. However, regulation of GSK-3 is more complicated; for example, β-catenin can be targeted to GSK-3 by Axin, which directly interacts with GSK-3 through a different binding site ( 32). GSK-3 activity has been linked to various pathologic conditions including Alzheimer's disease, diabetes type II, cancer, chronic inflammatory diseases, and psychiatric disorders ( 30– 32).

In addition to the previously described growth arrest and apoptosis, our study proposes a new mechanism for the anticancer effect of HDAC inhibitors, whereby cancer cells become targets for NKG2D-expressing cells through expression of MICA/B. Besides enhanced killing, this kind of recognition has the potential to induce lasting acquired immunity to cancer.

Materials and Methods

Cells and materials. Jurkat E6-1 cells were from American Type Culture Collection; JTag-9, Jurkat cells stably transfected with large T antigen from SV40 virus, was kindly provided by Carsten Geisler (University of Copenhagen, Denmark). MCF-7 was kindly provided by Henrik Leffers (The State Hospital, Copenhagen, Denmark). HeLa, Daudi, Aml193, Arh77, DOHH-2, Cem, Granta, U266, K562, HT29, and DLD-1 cell lines were kindly provided by Jesper Jurlander (The State Hospital). All cell clones were grown in RPMI 1640 supplemented with 10% fetal bovine serum, 2 mmol/L glutamine, 2 mmol/L penicillin, and streptomycin.

Cells isolated from human blood were grown in medium supplemented with 200 units/mL interleukin 2 (IL-2; Boehringer Mannheim Gmbh, Mannheim, Germany). Buffy coats from healthy human volunteer donors were obtained from the State Hospital. Peripheral blood mononuclear cells (PBMC) were obtained using lymphocyte separation medium (ICN Pharmaceuticals, Inc., Costa Mesa, CA) according to the description of the manufacturer. Activated T-cell populations were obtained by ionomycin (1 μmol/L)/phorbol 12-myristate 13-acetate (PMA; 10 ng/mL) stimulation of peripheral blood lymphocyte (PBL) cultures for 2 days. NK cells were purified from PBMCs by a negative selection kit from Dynal Biotech (Oslo, Norway) according to the description of the manufacturer. Anti-MICA/B monoclonal antibody (mAb), 6D4, was kindly provided by Thomas Spies (Fred Hutchinson Cancer Research Center, Seattle, WA); it can now be obtained from Becton Dickinson (Franklin Lakes, NJ). Anti-MICA/B, anti-MICB, and NKG2D-Fc mAbs were from R&D Systems (Minneapolis, MN). Murine immunoglobulin G (IgG) 2a control antibody was from DakoCytomation (Glostrup, Denmark). Human IgG control was from Bethyl Inc. (Montgomery, TX). Monoclonal CD11a-PE was from Leinco Technologies (St. Louis, MO). Monoclonal CD3-FITC, clone UCHT1, was from DakoCytomation and monoclonal CD19-PerCP-Cy5.5 was from BD Biosciences (Franklin Lakes, NJ). Polyclonal anti–phospho-GSK-3α/β(Ser21/9) was from Cell Signaling Technology, Inc. (Beverly, MA). Monoclonal anti–GSK-3α/β and polyclonal antihuman β-catenin were from BioSource (Camarillo, CA). Polyclonal anti–topoisomerase I was from Abcam (Cambridge, United Kingdom). Polyclonal anti–extracellular signal-regulated kinase 2 (ERK2) was from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). N-Acetyl-cysteine was from Sigma-Aldrich (St. Louis, MO). ZVAD-fmk was from Calbiochem (San Diego, CA). Kenpaullone was from Alexis Biochemicals (San Diego, CA) and LiCl was from Honeywell Riedel-de Haën (Seelze, Germany). Trichostatin A and SAHA were from Alexis. FR901228 was kindly provided by National Cancer Institute (Bethesda, MD). PXD101 was kindly provided by TopoTarget (Copenhagen, Denmark).

Flow cytometry. For surface staining, cells were stained with the appropriate antibodies or NKG2D/Fc for 30 minutes at 4°C, washed, and occasionally stained with secondary polyclonal rabbit anti-mouse immunoglobulin/FITC or polyclonal rabbit anti-human IgG/FITC (DakoCytomation) for 30 minutes at 4°C. For Annexin V/propidium iodide staining, we used the Annexin V-FITC apoptosis detection kit I from BD PharMingen (San Diego, CA) according to the description of the manufacturer. Data acquisition and flow cytometric analysis were done on a BD FACSCalibur using CellQuest software. Results are always showing Fsc/Ssc on a linear scale and fluorescence on a log₁₀ scale.

Reverse transcription-PCR analysis. RNA was isolated from Jurkat cells using RNeasy mini kit (Qiagen, Hilden, Germany) and reverse transcribed using SuperScript III reverse transcriptase enzyme (Invitrogen, Carlsbad, CA). PCR was done using standard conditions. MICA primer sequences were MICA_523-F, 5′-GCCATGAACGTCAGGAATTT-3′, and MICA_760-R, 5′-GACGCCAGCTCAGTGTGATA-3′. The housekeeping gene ribosomal protein, large, P0 (RPLP0) was used as a loading control. RPLP0 primer sequences were RPLP0-F, 5′-GCTTCCTGGAGGGTGTCC-3′, and RPLP0-R, 5′-GGACTCGTTTGTACCCGTTG-3′.

Cell lysis and Western blot analysis. Cytoplasmic extracts: Cells were solubilized in lysis buffer [1% NP40, 20 mmol/L Tris-HCl (pH 8.0), 140 mmol/L NaCl, 10% Glycerol, 1 mmol/L phenylmethylsulfonyl fluoride, 1 mmol/L Na₃VO₄, 10 mmol/L NaF, 1 mmol/L IAA, 5 mmol/L EDTA, 7.5 μg/mL aprotinin] for 30 minutes at 4° C and cell lysates were clarified by centrifugation at 13,000 rpm for 10 minutes. Occasionally nuclear and cytoplasmic proteins were separated by Nuclear Extraction Kit (Active Motif, Carlsbad, CA) according to the protocol of the manufacturer. Samples were mixed with Laemmli sample buffer (2% SDS) and boiled for 5 minutes. Proteins were resolved in SDS-polyacrylamide gels and transferred to nitrocellulose. Proteins were visualized by the indicated primary antibody and horseradish peroxidase–conjugated secondary antibody followed by enhanced chemiluminescence detection.

Glycogen synthase kinase-3β kinase assay. For kinase assays, GSK-3β was immunoprecipitated from cytoplasmatic extracts obtained from 2 × 10⁷ cells using a Cytoplasmatic and Nuclear Extraction Kit (Active Motif).

The cell lysates were precleared by incubation with 55 μL protein A-Sepharose beads for 2 hours and precipitated with 2 μg GSK-3β mAb (BD Transduction Laboratories, Franklin Lakes, NJ). After 2 hours of incubation at 4°C, 55 μL of protein-A-Sepharose beads were added and the lysates were incubated at 4°C overnight. The immunocomplexes were washed five times with lysis buffer, once with kinase buffer (20 mmol/L MOPS, 10 mmol/L Mg-acetate, 1 mmol/L EDTA, 1 mmol/L DTT), and resuspended in 50 μL kinase buffer containing 0.2 μg/μL phospho-glycogen synthase peptide-2 (Upstate Biotechnology, Charlottesville, VA), 90 μmol/L ATP, 13.5 mmol/L MgCl, and 10 μmol/L [γ-³²P]ATP. After 30 minutes of incubation at 37°C, supernatants were transferred to Unifilter Microplates (Whatman, Inc., Middlesex, United Kingdom) and washed thrice in 0.75% phospho-acid. DMSO as a vehicle control did not affect the GSK-3β kinase assay. Disc-bound radioactivity was quantified by liquid scintillation counting.

Transient transfections with small interfering RNA. Jurkat Tag cells were transiently transfected with the Nucleofector kit (Amaxa, Koeln, Germany) according to the protocol of the manufacturer. In brief, 1.5 × 10⁶ cells were resuspended in 100 μL Cell Line Nucleofector Solution V, mixed with 0.8 nmol/L of small interfering RNA (siRNA), and pulsed using the Nucleofector program G-10. After 24 hours, the cells were retransfected with the same type and amount of siRNA.

Cells were stained for flow cytometry or lysed for Western blot analysis 48 hours after the first transfection. Half of the samples used for flow cytometry had been incubated with 20 ng/mL FR901228 in the last 18 hours before staining. The siRNA was purchased from MWG-Biotech AG (Ebersberg, Germany). siRNA target sequences were GSK-3β, 5′-GUUAGCAGAGACAAGGACG-3′, and GSK-3α, 5′-GUGAUUGGCAAUGGCUCAU-3′.

Cytolytic assay. Cells were incubated with or without 20 ng/mL FR901228 and labeled with 50 μCi Na₂⁵¹CrO₄ pr. 10⁶ cells for 1 hour. The cells were then incubated with freshly isolated NK cells or IL-2-activated PBMCs for 4 hours at various effector/target ratios starting at 18 hours after FR901228 addition.

In some cases, anti-MICA/B (R&D Systems), NKG2D-Fc, or control antibody was added to the target cells 30 minutes before the coculture to a final concentration of 10 μg/mL.

The supernatant was subjected to gamma counting. The maximum and spontaneous release was defined as counts from samples incubated with 5% Triton X-100 or medium alone. Cytolytic activity was calculated as % lysis = (experimental release − spontaneous release) / (maximum release − spontaneous release) × 100. Each experiment was done in triplicates and with PBMCs from at least three different donors.

Results

Histone deacetylase inhibitors induce MICA/B expression on Jurkat T cells. Using gene chip screening of Jurkat T cells, we found that the HDAC inhibitor FR901228 increased MICA expression by a factor of ∼7.

Jurkat T cells had a low basal level of MICA surface expression that was potently induced by 18 hours of treatment with FR901228. MICA was detected by the MICA-01 and 03 allotype–specific mAb AMO1 (data not shown) and with the pan-reactive anti-MICA/B mAb 6D4, recognizing both MICA and MICB ( Fig. 1A ). Staining with an antibody specifically recognizing MICB showed that MICB was also induced (data not shown). In contrast, FR901228 treatment of Jurkat cells for 18 hours did not up-regulate surface expression of LFA-1 (CD11a), CD3, HLA-A, B, and C (MHC class I), HLA-DP, HLA-DQ, HLA-DR (MHC class II), HLA-G1, or HLA-E ( Fig. 1A and data not shown). Thus, FR901228 treatment of Jurkat cells induces both MICA and MICB expressions; MICA and MICB are homologous proteins that likely bind to NKG2D in a similar fashion.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

MICA/B expression by Jurkat T cells after exposure to HDAC inhibitors. A, Jurkat T cells were treated with or without 20 ng/mL FR901228 for 18 hours and stained with anti-MICA/B mAb 6D4, anti-LFA-1 (CD11a) mAb, anti–MHC class I (HLA-A, B, and C) mAb, or NKG2D-Fc. IgG2a is the isotype control for the anti-MICA/B mAb 6D4; the gates are set according to the relevant isotype control for each antibody. Flow cytometry results are shown for viable cells gated from Fsc/Ssc-scatter plots. B, top, Jurkat T cells were treated for the indicated time with 20 ng/mL FR901228 and stained with anti-MICA/B mAb 6D4 or subjected to reverse transcription-PCR analysis for MICA or RPLP0 mRNA expression (middle). Bottom, Jurkat T cells were treated with the indicated amount of FR901228 for 18 hours and stained with anti-MICA/B mAb 6D4. C, NKG2D-Fc (top) or 6D4 (bottom) staining of Jurkat T cells after treatment with the indicated amount of FR901228 (ng/mL), trichostatin A (TSA; ng/mL), SAHA (μmol/L), or PXD101 for 18 hours. D, MICA/B staining of CD3 T cells or CD19 B cells gated from a resting PBMC population or MICA/B staining of blasts from 2-day ionomycin/PMA stimulated PBL cultures after treatment with the indicated amount of FR901228 (ng/mL) for 18 hours; gates were set according to isotype control stainings. The bar graphs show percentage of MICA/B-expressing cells with background of control-treated cells subtracted. Representative of at least three independent.

To investigate if the induced MICA/B proteins can bind to the native NKG2D receptor, Jurkat T cells were treated with or without FR901228 and labeled with recombinant NKG2D-Fc. Figure 1A shows that recombinant human NKG2D-Fc binds strongly to Jurkat T cells, with a similar profile as the 6D4 antibody. It should be noticed that NKG2D, besides MICA and MICB, also binds ULBP1 to 4 on human cells. The regulation of ULBP proteins has not been addressed in the current study.

Surface expression of MICA/B could be detected after 6 hours and was increased to a plateau after 14 to 18 hours ( Fig. 1B). This was correlated by an increase in MICA mRNA levels ( Fig. 1B), suggesting that FR901228 affects MICA expression at the transcriptional level; however, we cannot rule out that the effect stems from mRNA stabilization. The decline in MICA/B surface expression after 20 hours is most likely a result of early initiation of apoptosis as the antioxidant N-acetyl-cysteine, which potently inhibits HDAC inhibitor–mediated apoptosis, prevents the decline in MICA/B expression (compare Figs. 1B and 2A ). A titration showed that 2 ng/mL of FR901228 induced near maximal MICA/B expression ( Fig. 1B). Generally, ∼60% to 90% of Jurkat T cells became MICA/B positive after FR901228 treatment; however, percent positive cells and the staining intensity varied but were never absent (over 50 experiments).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Inhibition of apoptosis does not abrogate FR901228-induced MICA/B expression. A, anti-MICA/B mAb 6D4 staining of Jurkat T cells treated for 26 hours without or with 20 or 100 ng/mL FR901228 in the presence or absence of 10 mmol/L N-acetyl-cysteine. MICA/B expression is shown for the viable cells; the gate can be seen in the Fsc/Ssc plots. B, anti-MICA/B mAb 6D4 staining of Jurkat T cells treated for 18 hours with or without 20 ng/mL FR901228 in the presence or absence of 50 or 20 μmol/L ZVAD-fmk. MICA/B expression is shown for gated viable cells. The percentages of viable cells after 48 hours are shown on the top of each bar; viability was measured by exclusion of 7-aminoactinomycin D. C, Annexin V, propidium iodide, and anti-MICA/B mAb 6D4 staining of Jurkat cells treated with 20 ng/mL FR901228 for 22 hours. The cutoff for MICA/B staining is set according to control-treated cells (not shown). D, anti-MICA/B mAb 6D4 staining of Jurkat T cells treated for 18 hours without or with the indicated amounts of vinblastine, thapsigargin, tunicamycin, staurosporine, or FR901228. Inset, detailed example with vinblastine. Representative of at least three independent experiments.

Trichostatin A, SAHA, and PXD101, three structurally different HDAC inhibitors led to increased staining detected by both anti-MICA/B antibody and NKG2D-Fc ( Fig. 1C and data not shown). This strongly suggests that HDAC inhibitor activity, and not a secondary effect, is responsible for the increased MICA/B expression.

As expected, FR901228 treatment did not significantly induce MICA/B expression on resting CD3 T cells and CD19 B cells from peripheral blood ( Fig. 1D), suggesting that HDAC inhibitors do not regulate MICA/B expression on healthy resting cells. Activated T cells have previously been shown to constitutively express MICA/B ( 33, 34). To test if MICA/B could be further up-regulated in these cells, we treated resting PBLs with ionomycin/PMA for 2 days. As expected, activated T cells had an increased basal level of MICA/B expression; interestingly, this could be further augmented by FR901228 treatment ( Fig. 1D). This suggests that the ability to up-regulate MICA/B expression after HDAC inhibitor exposure depends on the activation state of the cell. MICA/B induction is thus not entirely specific to transformed cells; however, it is noteworthy that primary resting blood cells remain MICA/B negative after HDAC inhibitor exposure.

In conclusion, HDAC inhibitors induce MICA/B expression on Jurkat T cells and further increase MICA/B expression on activated T cells.

MHC class I–related chain A and B expression is not dependent on apoptosis induction by histone deacetylase inhibitor treatment. MICA expression can be regulated by oxidative stress ( 35) and several studies have shown that exposure to HDAC inhibitors leads to production of reactive oxygen species ( 36, 37); it was thus our first hypothesis that HDAC inhibitors induced MICA/B through an oxidative stress mechanism. As shown in Fig. 2A, this does not seem to be the case. The antioxidant N-acetyl-cysteine could indeed inhibit FR901228-induced apoptosis in Jurkat T cells; however, this did not affect the intensity of MICA/B expression and actually increased the number of positive cells. The broad specific caspase inhibitor ZVAD-fmk also protected cells from FR901228-induced apoptosis without inhibiting MICA/B expression ( Fig. 2B). To examine MICA/B expression on viable versus apoptotic cells, we made a costaining of Annexin V, propidium iodide, and MICA/B after FR901228 treatment. As shown in Fig. 2C, only the viable cells (Annexin V negative) up-regulated MICA/B whereas apoptotic cells (Annexin V positive/propidium iodide negative) did not up-regulate MICA/B ( Fig. 2C).

These different data thus strongly suggest that FR901228-induced MICA/B is caused by a separate signal pathway that is not linked to the stress of dying cells.

Vinblastine, staurosporine, tunicamycin, and thapsigargin all potently induced apoptosis in Jurkat T cells without affecting MICA/B expression ( Fig. 2D, detailed example with vinblastine shown in inset), which is in agreement with the notion that HDAC inhibitor–induced MICA/B is a specific event not linked to induction of cell death.

In conclusion, the results suggest that a distinct signal pathway, which is not linked to oxidative stress or execution of cell death, leads to MICA/B expression after HDAC inhibitor exposure.

Histone deacetylase inhibitors induce glycogen synthase kinase-3 activity that is essential for MHC class I–related chain A and B expression. Because MICA expression is generally linked to cellular stress, infection, and transformation, we hypothesized that a stress-regulated enzyme might be involved. By screening inhibitors of stress-related enzymes, we found that LiCl and kenpaullone, two structurally different inhibitors of GSK-3 kinase activity ( 38), potently inhibited FR901228-induced MICA/B expression in a concentration-dependent manner ( Fig. 3A ). Furthermore, the GSK-3 inhibitor SB216763 also effectively inhibited FR901228-induced MICA/B expression (data not shown). LiCl or kenpaullone alone did not change the low basal level of MICA/B expression ( Fig. 3A and data not shown). To confirm the involvement of GSK-3, we transiently transfected Jurkat T cells with siRNA specific for GSK-3α or GSK-3β. Western blotting after transfection with siRNA against GSK-3α or GSK-3β showed that the targeted isoform of GSK-3 was specifically down-modulated by the relevant siRNA; however, the expression of especially GSK-3β was not completely blocked ( Fig. 3B). As further shown in Fig. 3B, down-modulation of GSK-3α or GSK-3β inhibited FR901228-induced MICA/B surface expression; simultaneous down-regulation of both isoforms led to the most potent reduction in MICA/B up-regulation. Interestingly, reduction of GSK-3α or GSK-3β led to an increase in the basal level of MICA/B expression, suggesting that basal and HDAC inhibitor–induced MICA/B expressions are not regulated by similar pathways. As control, we examined the high basal level of LFA-1 on Jurkat cells. This was not affected by GSK-3α or GSK-3β knockdown either with or without FR901228 treatment (data not shown).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

FR901228-induced MICA/B expression is dependent on GSK-3 activity. A, anti-MICA/B mAb 6D4 staining of Jurkat T cells treated for 18 hours with or without 20 ng/mL FR901228 in the presence or absence of the indicated amount of LiCl or kenpaullone, added 1 hour before FR901228 addition. MICA/B expression is shown for gated viable cells with the respective background subtracted. Detailed example is shown for control- and 50 mmol/L LiCl–treated cells. B, Jurkat T cells were transiently transfected with the indicated siRNA. Top, Western blot of GSK-3α, GSK-3β, and ERK of cell lysates from siRNA-transfected and control-treated cells. Bottom, anti-MICA/B mAb 6D4 staining of transfected cells after treatment for 18 hours with or without 20 ng/mL FR901228. C, Jurkat T cells were stimulated for the indicated time with or without 20 ng/mL FR901228, cells were subsequently lysed, and GSK-3β was immunoprecipitated; equal amounts of GSK-3β were used for enzyme assay using radioactive ATP and phospho-glycogen synthase peptide-2 as substrate. D, Jurkat T cells were stimulated for the indicated time with or without 20 ng/mL FR901228, nuclear lysates were subsequently Western blotted for the phospho-GSK-3β(Ser9), and total GSK-3β and topoisomerase I (nuclear loading control) or whole-cell lysates were blotted for β-catenin and ERK (loading control). Representative of at least three independent experiments.

In conclusion, these different results give compelling evidence that GSK-3 activity is critically involved in HDAC inhibitor–mediated MICA/B expression.

The GSK-3 enzyme is a constitutively active serine-threonine kinase; it is thus possible that MICA/B expression is dependent on its constitutive activity and that HDAC inhibitors do not actively regulate GSK-3 activity. This does not seem to be the case. FR901228 treatment induced a moderate increase in GSK-3β activity, measured by an in vitro kinase assay ( Fig. 3C). The GSK-3β activity could be inhibited by 50 mmol/L LiCl (data not shown). In agreement with this, we observed a decrease in the Ser9 phosphorylation of the nuclear GSK-3β ( Fig. 3D), an action that is critically linked to activation of GSK-3β.

One of the most well-defined targets of the GSK-3 kinase is β-catenin, which is tagged for ubiquitin/proteasome–mediated degradation by GSK-3-mediated serine phosphorylation. Stability of β-catenin is critical for the Wnt signal pathway and this signal pathway is therefore deactivated by GSK-3 activity. Based on these considerations, we hypothesized that HDAC inhibitor treatment might down-modulate β-catenin expression. As shown in Fig. 3D, β-catenin expression was down-modulated after exposure to FR901228 for 8 hours and more pronounced after 22 hours. Besides correlating with the induced GSK-3 activity, this also shows that HDAC inhibitors potently deactivate the β-catenin/Wnt signal pathway, which is strongly involved in several forms of cancer development.

Multiple types of cancer cells express MHC class I–related chain A and B through a glycogen synthase kinase-3–dependent mechanism after histone deacetylase inhibitor treatment. A panel of cancer cells was tested for MICA/B expression after FR901228 treatment. As shown in Fig. 4A and B , Daudi (B-cell acute lymphoblastic leukemia), Aml193 [acute myelogenous leukemia (AML)], Arh77 (multiple myeloma), DOHH-2 (malignant non-Hodgkin's lymphoma), Cem (T-cell acute lymphoblastic leukemia), Granta (mantle cell lymphoma), U266 (multiple myeloma), MCF-7 (epithelial breast adenocarcinoma), HeLa (epithelial cervix adenocarcinoma), HT29 (epithelial colorectal adenocarcinoma), and DLD-1 (epithelial colorectal adenocarcinoma) all increased MICA/B expression after FR901228 treatment for 18 hours; however, K562 (AML; Fig. 4A) and U937 (data not shown) did not respond to FR901228 treatment by increased MICA/B expression. The increased MICA/B expression was not equal in the different lines tested; a minor increase was observed in MCF-7 and HeLa ( Fig. 4A and B); an intermediate increase was observed in Daudi, Aml193, DOHH2, Cem, U266, and DLD-1; and a robust increase was observed in Granta, Arh77, and HT29 ( Fig. 4A). There was no strict correlation with the type of cancer and the degree of MICA/B induction [compare, e.g., multiple myeloma lines U266 and Arh77 ( Fig. 4A) or T-cell acute lymphoblastic leukemia cell lines Jurkat and Cem ( Figs. 1A and 4A)].

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Various cancer cells express MICA/B after FR901228 treatment. A, the indicated cells were treated with or without 20 ng/mL of FR901228 for 18 hours and subsequently stained with anti-MICA/B mAb 6D4; the gate is set arbitrarily to enlighten the difference in staining intensity. The percentages of dead cells based on a Fsc/Ssc plot is indicated on the right. B, HeLa cells were treated and stained as in (A); however, cells were incubated with or without 50 mmol/L LiCl 1 hour before FR901228 addition and control IgG2a stain is included. Representative of at least three independent experiments.

LiCl and kenpaullone could, to various extents, inhibit FR901228-induced MICA/B in all the cells tested ( Fig. 4B and data not shown), strongly suggesting that GSK-3 activity is generally involved.

It has previously been described that HeLa cells have constitutive expression of MICA, which we indeed also observed. Interestingly, however, the constitutive MICA/B expression in HeLa cells was not affected by GSK-3 inhibition, which nonetheless inhibited the FR901228 superinduction of MICA/B ( Fig. 4B, note that the gate in Fig. 4B is set to show FR901228-superinduced MICA/B and thus not with respect to the isotype control; all HeLa cells are 100% positive for MICA/B). Even high concentrations of LiCl (50 mmol/L) and prolonged exposure (72 hours) did not affect the constitutive MICA expression on HeLa cells (data not shown). This suggests that a separate molecular signal pathway, distant from the one leading to constitutive expression, regulates MICA/B after exposure to HDAC inhibitors.

FR901228-induced MHC class I–related chain A and B is functionally active. Although recombinant NKG2D-Fc binds to HDAC inhibitor–induced MICA/B, ensuring its functional ability, it was still important to verify the functionality of MICA/B and further investigate if the increased MICA/B expression could enhance NK cell recognition and destruction of cancer cells.

We assessed Arh77 and Granta as they expressed high amounts of MICA/B after FR901228 exposure without being drastically affected in viability within the time frame of the assay. Furthermore, both cell types are relatively resistant to NK cell–mediated lysis, ensuring a “therapeutic window.”

Arh77 or Granta was incubated with or without 20 ng/mL FR901228 for 18 hours and then subjected to a standard 4-hour ⁵¹Cr release assay with nonactivated freshly purified NK cells from human blood as effector cells. As expected, untreated Arh77 or Granta cells were largely resistant to NK cell–mediated lysis; however, FR901228 treatment of either Granta or Arh77 cells led to a strong augmentation of sensitivity and killing by freshly isolated NK cells ( Fig. 5A ). FR901228 also potently enhanced the susceptibility of Granta and Arh77 cells to activated NK cells from 4-day IL-2-activated PBMC cultures ( Fig. 5B). Inclusion of a recombinant human NKG2D-Fc or blocking MICA/B antibody led to a partial inhibition of NK activity towards FR901228-treated Granta cells; this, however, did not affect NK activity toward vehicle-treated cells ( Fig. 5C). These results clearly suggest that HDAC inhibitor–induced MICA/B expression is functionally active and can target cancer cells for NK cell–mediated lysis. Furthermore, the very potent augmentation of FR901228-treated cells towards NK cell lysis, which is only partially affected by inhibition of NKG2D/ligand interaction, suggests that additional molecules regulating NK cell activity may be modulated by HDAC inhibitors. MHC class I expression on Granta and Arh77 cells was not significantly changed by FR901228 treatment (data not shown).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

FR901228-induced MICA/B are functionally active. A, Granta or Arh77 cells were treated with or without 20 ng/mL FR901228 for 18 hours, labeled with ⁵¹Cr, and cocultured for 4 hours with freshly purified NK cells from healthy donors at the indicated effector/target ratio. B, Granta or Arh77 cells were treated with or without 20 ng/mL FR901228 for 18 hours, labeled with ⁵¹Cr, and cocultured for 4 hours with 4-day IL-2 activated PBMC cultures at the indicated effector/target ratio. C, Granta cells were pretreated with human IgG control, NKG2D-Fc, IgG2a control, or blocking MICA/B antibody (all 10 μg/mL) before cytotoxic assay with freshly isolated NK cells as described in (A) at an effector/target ratio of 10:1. Experiments were done in triplicates. Columns, mean % lysis; bars, SE. Representative of at least three independent experiments.

Discussion

HDAC inhibitors are potent inducers of apoptosis and growth inhibition in a variety of transformed cells in vitro and in vivo; on the other hand, they are relatively nontoxic to normal cells and they are thus recognized as good candidates for novel anticancer therapy ( 39).

Cytotoxic killing of cancer cells by radiation or chemotherapy has been the prevailing treatment for cancer. The major drawback is lack of complete eradication, leading to selection of resistant cancer cells. A combinatory treatment targeting the cancer cells in different ways is probably a more fruitful approach, much similar to treatment of serious bacterial infections with a mixture of antibiotics. One arm of this approach has been to boost the immune system to improve recognition and destruction of cancer. Current attempts are unspecific activation by IL-2 or pan-T-cell activation by CD3- and CD28-coupled stimulatory beads, or induction of specific anticancer immunity by dendritic cell–based vaccines ( 40– 42).

Fine-tuning of signals from activating and inhibitory receptors regulates NK cell activity and it is conceivable that NK activity is dependent on the relative number of positive and negative interactions. MHC class I–bearing cells may indeed be killed by NK cells when NKG2D ligands are up-regulated on the cell surface ( 13, 21). Our study suggests that, in addition to the direct cytotoxic effect, HDAC inhibitors also improve the immunorecognition of cancer cells by inducing MICA/B. This brings an interesting new dimension to the usage of HDAC inhibitors as anticancer agents, particularly given that MICA/B are induced in the surviving fraction of cells.

There was no strict correlation between the main type of cancer and the ability of HDAC inhibitors to induce MICA/B. This is in line with several reports showing that constitutive expression of MICA/B can be detected on a range of different cancer types ( 16, 19, 27, 43). In addition, the HDAC inhibitor–mediated induction was not restricted to tumor cells already showing high constitutive MICA/B expression. The critical factor(s) determining whether a cancer cell responds to HDAC inhibitor treatment by up-regulating MICA/B is still unknown. Of interest, the viability of several of the cancer cells tested was not significantly affected by FR901228 treatment for 18 hours; however, the main portion of cells shifted from negative to MICA/B positive. It should be noted that our study only includes established cancer cell lines and that the effect on primary tumor cells requires further investigation.

We were not able to completely prevent NK cell–mediated lysis of FR901228-treated cells by incubation with blocking MICA/B antibody or NKG2D-Fc (see Fig. 5C) and although it is difficult to rule out that this is due to incomplete blocking of MICA/B, it is possible that other NK activating factors are induced by HDAC inhibitor treatment. In this respect, Insinga et al. ( 44) and Nebbioso et al. ( 45) recently showed that HDAC inhibitor killing of human leukemia cells in vivo, at least in part, is dependent on induction of tumor necrosis factor (TNF)–related apoptosis-inducing ligand and its ligands death receptor 4 (DR4) and DR5. NK cells are known to express functional TNF-related apoptosis-inducing ligand ( 46– 48) and it is indeed possible that a part of the enhanced NK lysis we observe after FR901228 treatment is caused by induction of DR4 or DR5 on the target cells. In any case, we find that MICA/B are expressed on various cancer cells after HDAC inhibitor treatment; it should be noticed that NKG2D and TNF-related apoptosis-inducing ligand are not solely expressed by the same effector cells nor are they mediating the same intracellular signals.

We have previously shown that the HDAC inhibitor FR901228 suppresses de novo T-cell activation in vitro without being cytotoxic to naïve T cells ( 49). This may potentially interfere with the generation of antitumor immunity after FR901228 treatment although the prompt clearance from serum and lack of severe inhibition of the immune system in clinical studies imply that a therapeutic window may exist. The increased MICA/B expression on the activated T cells after HDAC inhibitor treatment could obviously facilitate NK cell–mediated killing. However, it has recently been recognized that T cells, NK cells, and dendritic cells make a delicate cross talk during the primary activation of the acquired immune system involving MICA/B ( 50, 51). MICA/B expression on activated T cells may therefore not necessarily function as a tag for destruction, although this requires further investigation.

MICA/B were not increased on resting T and B cells after HDAC inhibitor treatment and it is unlikely that aberrant MICA expression leads to autoimmunity in vivo as this would have been recognized during clinical trials.

One fundamental question needs to be asked: Why is MICA/B induced on cancer cells and activated T cells after HDAC inhibitor treatment and not on normal healthy resting cells?

It is well described that transformation can lead to MICA expression although the mechanism involved is currently unknown ( 19, 24). Our hypothesis is that some cancer cells manage to hold MICA/B expression in check as part of their successful in vivo transformation. Significant portions of the MICA/B expression program could then still be in place and may keep the MICA/B genes in an unfolded state that can be targeted by HDAC inhibitor treatment. A more rigid understanding of the complex signal pathways regulating MICA/B is therefore of particular interest.

As both activated T cells and cancer cells are characterized by proliferation, it could be hypothesized that MICA/B expression requires progression through the cell cycle. However, this seems not to be the case because (a) HDAC inhibitors are potent inducers of cell cycle arrest and yet do induce MICA/B expression; and (b) the DNA polymerase inhibitor Aphidicolin, which causes G₁ arrest, does not affect FR901228-induced MICA/B expression on Jurkat T cells. ⁴

On the other hand, a link between viability and MICA/B induction seems to exist. We observe that inhibition of HDAC inhibitor–induced apoptosis either does not affect or actually increases the percentage of MICA/B-expressing cells whereas enhancement of the apoptotic process limits the induction. Thus, although MICA/B are known to be stress inducible, the HDAC inhibitor regulation is likely not a side effect of the stress caused by the apoptotic process. From a therapeutic point of view, it is also noteworthy that the MICA/B up-regulation occurs specifically on the surviving fraction of cells.

Groh et al. ( 16) was the first to characterize MICA/B as stress-inducible proteins after observing heat shock–mediated increase of MICA and de novo MICB expressions in HeLa cells. By homology searches, they found a heat shock protein 70 heat shock element (HSEHSP) in both MIC promoters, which is likely involved in heat shock induced MICA/B expression. It is doubtful that a similar mechanism is responsible for HDAC inhibitor–induced MICA/B because binding of heat shock factor-1 to HSEHSP is potently inhibited by GSK-3β activity ( 52, 53).

Oxidative stress has also been reported to increase MICA/B expression in human colon carcinoma ( 35); again, it is very unlikely that a similar mechanism is involved the current observations as the antioxidant N-acetyl-cysteine actually potentiates HDAC inhibitor induction of MICA/B.

Based on GSK-3 specific inhibitors and siRNA directed against GSK-3α and GSK-3β, we have found strong evidence that GSK-3 activity is critically involved in FR901228 regulation of MICA/B in Jurkat T cells. The siRNAs against the two isoforms were not equally effective in down-modulating specific protein levels, making it difficult to determine whether one of the isoforms plays a dominant role.

We also find that FR901228 increases GSK-3β activity in the Jurkat T cells, this was detected by a reduction in Ser9 phosphorylation of GSK-3β and an increase in GSK-3β activity measured by an in vitro kinase assay. Furthermore we observe reduced levels of β-catenin, a well-known substrate of GSK-3, after prolonged exposure to FR901228 — a reduction that can be inhibited by cotreatment with LiCl. ⁴

The increase in GSK-3β activity measured by the kinase assay is relatively weak and also transient, which is not consistent with the sustained hypophosphorylation of GSK-3β on Ser9 and the longer-term effects on β-catenin and MICA/B levels. The reason for this discrepancy is not clear. Other studies assessing GSK-3β activity by in vitro kinase assay have reported a similar somewhat limited activation ( 54, 55); however, we cannot exclude that our in vitro kinase assay might be too crude to reveal smaller, yet important, changes in the activity of GSK-3β. It should be noted that we have not investigated the kinase activity of GSK-3α, which may function in addition to GSK-3β. Moreover, several mechanisms play a part in controlling the actions of GSK-3, including serine and tyrosine phosphorylation, protein complex formation, and subcellular distribution ( 31, 32, 56), adding complexity to the regulation.

There are a multitude of signaling pathways integrated by GSK-3 that may regulate HDAC inhibitor–induced MICA/B expression and these are currently being investigated in our laboratory. It is still not clear how proximal GSK-3 is situated in the signal pathway regulating MICA/B. Interestingly, GSK-3 activity does not seem to be required for the constitutive MICA/B expression observed on certain tumor cells ( Fig. 4B), suggesting that the constitutive and HDAC inhibitor–induced MICA/B expressions are regulated by different mechanisms. HDAC inhibitors can still superinduce MICA/B by a GSK-3-dependent mechanism, suggesting that both signal pathways can occur simultaneously.

GSK-3 has been shown to be required for proliferation or to exert proapoptotic functions in different contexts ( 57, 58). In our study, GSK-3 inhibition does not seem to prevent MICA/B up-regulation simply by enhancing the apoptotic process. LiCl actually inhibits FR901228-induced apoptosis in Jurkat T cells as assessed by Annexin V staining. ⁴ These results are in agreement with a recent article showing that GSK-3 activity correlates with increased susceptibility to FR901228-induced apoptosis in lung adenocarcinoma cell lines ( 59).

Viral infections with cytomegalovirus and hepatitis C virus have been shown to hamper NKG2D ligand expression by the infected cells, which may be of critical importance for establishment of the chronic infections ( 60– 64). Recently, FR901228 was shown to increase survival in a mouse xenograft model of human EBV-positive lymphoblastoid cell lines ( 65). It will thus be of considerable interest to study the potential of HDAC inhibitors to activate MICA/B expression in virus-infected cells.