HDAC5 controls the functions of Foxp3⁺ T-regulatory and CD8⁺ T cells

Abstract

Histone/protein deacetylases (HDACs) are frequently upregulated in human malignancies and have therefore become therapeutic targets in cancer therapy. However, inhibiting certain HDAC isoforms can have pro-tolerogenic effects on the immune system, which could make it easier for tumor cells to evade the host immune system. Therefore, a better understanding of how each HDAC isoform affects immune biology is needed to develop targeted cancer therapy. Here, we studied the immune phenotype of HDAC5^−/− mice on a C57BL/6 background. While HDAC5^−/− mice replicate at expected Mendelian ratios and do not develop overt autoimmune disease, their T-regulatory (Treg) cells show reduced suppressive function in vitro and in vivo. Likewise, CD4⁺ T-cells lacking HDAC5 convert poorly to Tregs under appropriately polarizing conditions. HDAC5^−/− Tregs show increased acetylation of Foxo1, which is deacetylated by HDAC5 and important for maintaining the Treg cell phenotype. To test if this attenuated Treg formation and suppressive function translated into improved anti-cancer immunity, we inoculated HDAC5^−/− mice and littermate controls with a lung adenocarcinoma cell line. Cumulatively, lack of HDAC5 did not lead to better anti-cancer immunity. We found that CD8⁺ T cells missing HDAC5 had a reduced ability to produce the cytokine, IFN-γ, in vitro and in vivo, which may offset the benefit of weakened Treg function and formation. Taken together, targeting HDAC5 weakens suppressive function and de-novo induction of Tregs, but also reduces the ability of CD8⁺ T cells to produce IFN-γ.

Introduction

Cancers are commonly associated with epigenetic changes, affecting gene expression of oncogenes and tumor suppressors, and altering immune function.¹ Loss of histone-4 acetylation at K16 and K20 has been identified in a wide variety of human cancers, implying involvement of control of histone acetylation in malignancy.^{1, 2} Histone (and non-histone) acetylation is a reversible post-translational modification of lysine residues that neutralizes the otherwise positive charge of a lysine’s ε-amino group. Acetylation is promoted by histone/protein acetylase (HAT) enzymes, and reversed by histone/protein deacetylase (HDAC) enzymes.³ Several HDAC isoforms are overexpressed in human cancers, and HDAC inhibitors have therefore been used as antineoplastic agents in clinical practice and in ongoing clinical trials.^{1, 4, 5} The effects of HDAC inhibitors include alterations in the control of apoptosis and cell growth, cell differentiation, angiogenesis, and effects on tumor immunity.⁴ The latter is of particular interest, as several pan-HDAC inhibitors, such as Trichostatin A or Vorinostat, have been found to attenuate immune responses and favor T-regulatory (Treg) cell-mediated immunosuppression.^{6, 7} Tregs are an important T cell subset capable of decreasing immune responses ⁸. Tregs are characterized by their expression of the transcription factor, Forkhead box P3 (Foxp3), which plays a key role in their development and function.^{9, 10} Augmenting Treg function and formation is desirable to mitigate autoimmune disease or prevent allograft rejection.¹¹ However, in cancer biology, Tregs play a very different role and may help tumors escape host anti-tumor immunity.¹² Thus, in cancer treatment, unintentional Treg augmentation by HDAC inhibition needs to be considered an adverse effect.¹³ Tregs can be recruited to tumor sites or induced de-novo from conventional T cells, resulting in interference with anti-tumor immune responses.¹⁴ Depleting Tregs, or impairing their function or formation, is a promising therapeutic approach to increase anti-tumor immunity.^{15, 16}

Treg function can be altered through modulation of Foxp3, which is subject to post-translational control by HAT and HDAC enzymes.^{17, 18} Acetylation of Foxp3 improves DNA binding and suppressive ability ¹⁹, and reduces the turnover of Foxp3 through ubiquitination and proteasomal degradation.^{20, 21} Inhibition of the HAT, p300, reduces Foxp3 acetylation, weakens Treg function and iTreg formation, and improves anti-tumor immunity.²² Conversely, targeting specific HDACs that deacetylate Foxp3, such as HDAC9, HDAC6, or Sirtuin-1, increases Foxp3 acetylation and strengthens Foxp3⁺ Treg cell function.^{7, 18, 23–25}

To reconcile potential conflicts between anti-neoplastic effects versus undesired augmentation of Treg function and formation, and to minimize the adverse effects of global pan-HDAC inhibition, it will be necessary to develop isoform-selective HDAC inhibitors. To that end, it is important to study the properties of each individual HDAC in cancer biology and immunity. Here, we investigated HDAC5, a class IIa HDAC, and explored its role in T cell function. In contrast to other class IIa HDACs, we found that loss of HDAC5 weakened Treg function and de-novo induced Treg (iTreg) formation. Therefore, we investigated the utility of targeting HDAC5 to augment anti-cancer immunity. Our results are important for the development of future HDAC inhibitor-based therapies.

Materials and Methods

Animals and cardiac allografting

We purchased BALB/c (H-2^d), C57BL/6 (H-2^b), Thy1.1 (H-2^b), and B6/Rag1^−/− (H-2^b) mice from The Jackson Laboratory. HDAC5^−/− mice²⁶ were a gift from Eric Olson to WWH. Mice housed under specific pathogen-free conditions were studied using protocols approved by the Institutional Animal Care and Use Committees of the Children’s Hospital of Philadelphia and the University of Pennsylvania (13-000561 and 14-001047). We transplanted BALB/c hearts into B6/Rag1^−/− recipients.⁷ Allograft recipients received adoptively transferred intravenously with 1 × 10⁶ WT T-effector cells plus 5 × 10⁵ Treg cells (WT or HDAC5^−/−), and allograft survival was monitored daily by palpation.

Antibodies, media and small molecules

For flow cytometry, we purchased mAbs to murine CD4 (APC-eFluor 780, PE), CD8 (Pe-Cy7), CD25 (APC), CD62L (APC-Cy7), CD44 (PE) and Foxp3 (APC) from eBioscience, as well as CD4 (Pacific Blue) and Ki67 (PerCP-Cy5.5) from BD Pharmingen. Additional antibodies used for flow cytometry included CD8 (APC), CD4 (FITC), IL-2 (PE), IFN-γ (PE-Cy7), CD25 (APC-Cy7) from Biolegend, and CD69 (PE-Cy5) as well as CD44 (Pacific Blue) from BD Biosciences. We used the Aqua LIVE⁄DEAD® staining kit (Life Technologies) to detect dead cells. E7 tetramer PE-conjugated antibodies were provided by Yvonne Patterson.²⁷ For immunoblotting, we purchased antibodies to acetylated Foxo1 (Ac-FKHR Antibody, D-19) from Santa Cruz, and HDAC2, HDAC5, Foxo1 (C29H4), and Phospho-FoxO1 (Thr24)/Foxo3a (Thr32) and β–actin from Cell Signaling. We also purchased Foxp3 (eBioscience) and phosphorylated HDAC2 (phospho S394, Abcam) antibodies. For T cell culture medium, we used RPMI-1640 supplemented with 10% FBS, penicillin (100 U/mL), streptomycin (100 μg/mL) and 55 nM β2-mercaptoethanol. LMK235 was synthesized by the TK group and initially by Linda Marek as previously published ²⁸, dissolved in dimethylsulfoxide (DMSO) and diluted in tissue culture media. Equally diluted DMSO alone was used as a control.

Cell isolation and flow cytometry

Spleen and peripheral lymph nodes were harvested and processed to prepare single cell suspensions of lymphocytes. We used magnetic beads (Miltenyi Biotec) for isolation of conventional T cells (Tconv, CD4⁺CD25⁻), Treg cells (CD4⁺CD25⁺), and antigen presenting cells (CD90.2⁻). We confirmed equal purity post-isolation by flow cytometry. For Foxp3 staining, surface marker-stained cells were fixed, permeabilized, and labeled with anti-Foxp3 mAb.²⁹ Flow cytometry data was captured using Cyan ADP Flow Cytometer (Dako) and analyzed using FlowJo 9.5.3 (Treestar) software.

T cell assays

For Treg suppression assays, purified Tconv were labeled with carboxyfluorescein succinimidyl ester (CFSE, Molecular Probes) and stimulated with irradiated antigen presenting cells plus CD3ε mAb (1 μg/ml, BD Pharmingen). After 72 h, the extent of proliferation of T-effector cells was quantitated by flow cytometric analysis of CFSE dilution. Suppressive Treg function from multiple independent Treg suppression assays was calculated by an area-under-curve method as previously described.^{6, 30} For in vivo Treg suppression assays, we adoptively transferred CD90.1⁺CD4⁺CD25⁻ Tconv with or without WT or HDAC5^−/− Tregs (CD90.2⁺CD4⁺CD25⁺) into B6/RAG1^−/− mice at a 4:1 ratio.⁷ Lymphocytes were isolated after one week, and expansion of CD90.1⁺CD4⁺ T cells was quantified by flow cytometry. To compare independent in vivo Treg suppression assays, we normalized CD90.1⁺CD4⁺ cell counts to the average of the Thy1.1 without Tregs group of that experiment. For conversion to Foxp3⁺ Tregs, Tconv cells were incubated for 4 days with CD3ε/CD28 mAb beads, plus TGF–β (3 ng/ml) and IL-2 (25 U/ml), and analyzed by flow cytometry for Foxp3⁺ iTreg.²³ For conversion to Th17 cells, we used a protocol by Thomas et al.³¹ Briefly, CD8⁺ cells were depleted using Miltenyi CD8 microbeads, and the remaining CD4⁺ T cells and antigen presenting cells were cultured with soluble CD3ε and CD28 mAb (1 μg/ml each) for four days in the presence of anti-IL-4 and anti-IFN-γ mAbs (20 μg/ml), TGFβ (1 ng/ml), and IL-6 (10 ng/ml). For intracellular IL-17 staining, cells were stimulated with 30 ng/ml PMA and 1 μM ionomycin (Sigma Aldrich) for 5 hours in the presence of GolgiStop reagent (BD Biosciences). For assessment of in vitro cytokine production, we incubated freshly isolated CD4⁺CD25⁻ Tconv from WT and HDAC5^−/− mice overnight (37º C, 5% CO₂) in 24 well plates pre–coated with CD3ε and CD28 mAbs (2 μg/mL). In the morning, PMA/ionomycin and GolgiStop were added to reach final concentrations of 50 ng/mL PMA, 1 μM ionomycin, and 0.67 μL GolgiStop per mL of medium. Cells were then incubated for 5 additional hours (37º C, 5% CO₂), and harvested for flow cytometry. We used the Fixation/Permeabilization Buffer set from eBioscience for intranuclear staining.³²

RNA Isolation, qPCR and Western Blots

RNA was extracted using RNeasy kits (Qiagen), and RNA integrity and quantity were analyzed by photometry (DU640; Beckman-Coulter). Revere transcription, qPCR, and Western blotting were performed as reported.^{25, 33} Primers were purchased from Applied Biosystems.

Tumor studies

We used a HPV-16 immortalized TC1 murine lung adenocarcinoma model ²⁷ to assess tumor immunity, as previously described.²² All tumor experiments with HDAC5^−/− mice used littermate controls. Briefly, TC1 cells were grown in RPMI, 10% FBS, 2 mM glutamine, and 5 μg/mL penicillin and streptomycin. Each mouse was shaved on its right flank and injected subcutaneously with 1.2 × 10⁶ TC1 tumor cells. Tumor volume was determined by calculating (3.14 × long axis × 1^st short axis × 2^nd short axis) ÷ 6, as previously reported.³⁴ After tumor retrieval and weight measurement, tumor tissues were cut into small fragments, followed by digestion for one hour at 37 °C with a cocktail containing collagenases type I (10 U/100 mL, Worthington), II (10 U/100 mL, Worthington) and IV (10 U/100 mL, Worthington), DNase I (5 U/100 mL, Worthington), and elastase (5 U/100 mL, Worthington) in L15-medium (Leibovitz). Single cell suspensions from spleens and tumors were then blocked with anti-Fc receptor antibody (eBioscience) for 30 min before flow cytometric analysis. For cytokine re-stimulation, splenic and tumor single cell suspensions were re-stimulated for 4–6 hours in 6 well plates pre–coated with CD3ε and CD28 mAbs (2 μg/mL) or with PMA/ionomycin (30 ng/mL, 1 μM) in presence of Golgi Stop (0.7 μl/mL) and characterized for intracellular cytokine staining as outlined by BD Biosciences.

Data analysis

Data were analyzed using GraphPad Prism 5.0d software. Normally distributed data were displayed as mean ± standard error of the mean. Non-normally distributed data were with median and interquartile range. Measurements between two groups were performed with an unpaired Student-t test if normally distributed, or Mann-Whitney U test if otherwise. For paired samples, we used a paired Student-t test or Wilcoxon matched-pairs signed rank test, depending on whether or not data were normally distributed, respectively. Survival was assessed using a log-rank (Mantel-Cox) test.

Results

Loss of HDAC5 impairs Treg function

Mice deficient in HDAC5 developed normal populations of CD4⁺ and CD8⁺ T cells (Suplemental figure S1a and b). They also reproduced at expected Mendelian ratios, formed normally sized lymphoid tissues, and did not develop spontaneous illness over 18 months of observation. Subsets of activated CD4⁺ and CD8⁺ T cells, as well as CD4⁺Foxp3⁺ Treg cells, were comparable to those of wild type (WT) control mice (Supplemental figure S1c–e). However, upon bead isolation of CD4⁺CD25⁺ Treg (with equal Foxp3⁺ purity), we observed that HDAC5^−/− Tregs had less suppressive function in vitro (Figure 1a). We noted weaker HDAC5^−/− Treg function with every pairing of WT or HDAC5^−/− APC and proliferating T-effector (Teff) cells in five independent experiments (Figure 1b). In vivo, adoptive transfer of HDAC5^−/− Treg into B6/Rag1^−/− mice showed only a limited ability to suppress the homeostatic proliferation of CD90.1⁺ T-effector cells that were co-transferred at a 4:1 T-effector:Treg ratio (Figure 1c). Lastly, we further confirmed diminished HDAC5^−/− Treg function in allograft studies. B6/Rag1^−/− (H-2^b) recipients of BALB/c cardiac allografts (H-2^d) were adoptively transferred with autologous 5 × 10⁵ Tregs of either WT or HDAC5^−/− origin, plus 1 × 10⁶ WT Tconv cells. While WT Treg were able to prevent rejection of the mismatched cardiac allograft at a 2:1 Tconv to Treg ratio, Treg cells deficient in HDAC5 were unable to maintain the allografts (Figure 1d). In summary, mice deficient in HDAC5 appeared normal, but their Treg cells showed impaired suppressive function both in vitro and in vivo.

Figure 1.

Treg lacking HDAC5 have weak suppressive function. (a, b) Comparison of the ability of HDAC5^−/− versus WT Tregs to suppress proliferation of CFSE-labeled WT or HDAC5^−/− T-effector (Teff) cells in vitro, stimulated with anti-CD3ε mAb and irradiated CD90.2⁻ antigen presenting cells (APC) from either WT or HDAC5^−/− mice. (a) Representative flow cytometry and (b) cumulative analysis showing HDAC5^−/− Treg have weaker suppressive Treg function versus results for WT Treg. Data shown as mean ±SEM, Student’s t-test of 17 observations pooled from five independent experiments. (c) Homeostatic proliferation showed weakened suppressive function of Treg with HDAC5 deletion in vivo (6 mice per group pooled from two independent experiments, Student’s t-test with * indicating P<0.05, ** P<0.01, and *** indicating P<0.001, respectively). (d) Cardiac allograft survival of BALB/c hearts transplanted to B6/Rag1^−/− recipients adoptively transferred with 1 × 10⁶ WT Tconv, and either 5 × 10⁵ WT or HDAC5^−/− Treg. Treg lacking HDAC5 are unable to prevent rejection (Mantel-Cox test, 7/group).

Targeting HDAC5 weakens iTreg formation

In contrast to Treg, HDAC5^−/− Tconv showed no obvious difference in function. Proliferation in response to CD3ε/CD28 stimulation was similar to WT controls, regardless of APC origin (Figure 2a and b). We observed that proliferating T-effector cells lacking HDAC5 were less responsive to Treg mediated suppression (Figure 2c), however, the difference was very modest. Likewise, cytokine production by CD4⁺ Tconv in response to PMA/ionomycin stimulation was similar (Figure 2d and e). To assess HDAC5^−/− Tconv proliferation in vivo, we adoptively transferred 5 × 10⁵ CD4⁺CD25⁻ Tconv from WT and HDAC5^−/− mice into B6/Rag1^−/− recipients intravenously and observed the mice for 42 days in a colitis model.³² No differences in weight were apparent (Figure 2f). B6/Rag1^−/− recipients of HDAC5^−/− Tconv showed a trend to higher CD4 cell proliferation, but the difference was not significant (Figure 2g). Both WT and HDAC5^−/− adoptively transferred CD4 T cells had equal fractions of Ki67, suggestive of comparable proliferation at the end of the experiment (Figure 2h).

Figure 2.

HDAC5 deletion does not affect CD4⁺CD25⁻ T cell proliferation and cytokine production. (a) CFSE-labeled HDAC5^−/− T-effector cells (Teff) respond to co-stimulation equally compared to WT control. Cells stimulated with anti-CD3ε, anti-CD28 mAb and irradiated antigen presenting cells (APC) from either WT or HDAC5^−/− mice. Data representative of two independent experiments. (b, c): Comparison of HDAC5^−/− versus WT APC and CD4⁺CD25⁻ T cells from the Treg suppression assays in Figure 1a, b. Data pooled from 16 observations in four independent experiments. (b) Lack of HDAC5 did not affect co-stimulation from irradiated APC. (c) Proliferating HDAC5^−/− Teff were 8.1 ±2% more resistant to Treg suppression. (d, e): Cytokine production after PMA/ionomycin stimulation is not different between HDAC5^−/− and WT CD4⁺ Tconv. Data pooled from three independent experiments. (f-h): In vivo Tconv proliferation. 1 × 10⁶ WT or HDAC5 were adoptively transferred into B6/Rag1^−/− recipients, who did not develop differences in weight (f), CD4⁺ T-cell counts in their spleens (g), or Ki67⁺ cells among their CD4⁺ T cells (h). Abbreviations: WT, wild type; mLN, mesenteric lymph nodes.

In one aspect, CD4⁺ Tconv cells lacking HDAC5 were very different from WT controls. When exposed to polarizing conditions to promote iTreg development, HDAC5^−/− Tconv showed markedly decreased conversion rates (Figure 3a). This finding was persistent over five independent experiments (Figure 3b). Likewise, use of LMK235, a class IIa HDAC inhibitor with low nanomolar IC₅₀ against HDAC5 ²⁸, incrementally impaired iTreg conversion (Figure 3c). However, this may not necessarily prove definitive, as LMK235 can also limit cellular proliferation, at least at the 100 nM level (Figure 3d). By contrast, HDAC5^−/− T-effector cells did not have any altered potential to convert to a Th17 phenotype (Figure 3e). In summary, targeting of HDAC5 impairs the ability of T-effector cells to convert into iTreg cells.

Figure 3.

Targeting HDAC5 impairs de-novo Treg induction without affecting Th17 cell conversion. (a, b) HDAC5^−/− and WT Tconv cells were subjected to polarizing conditions to form induced Treg (iTreg) using TGF-β, IL-2 and anti-CD3ε/CD28 beads. (a) Representative flow cytometry and (b) cumulative data show marked impairment of HDAC5^−/− T-effector cells to form iTreg cells. (c) Use of an class IIa HDAC inhibitor (LMK235), produced a similar effect, although reduced Foxp3 formation may to some extend also be explained by impaired T-effector cell proliferation (d). (e) Polarization of HDAC5^−/− CD4⁺ cells to a Th17 phenotype was unaltered and equal to WT control. Data representative of two independent experiments.

Loss of HDAC5 leads to decreased Foxp3 protein in Tregs

Our data show that mice lacking HDAC5 have weaker Treg, and their Tconv have diminished capacity to convert into iTreg cells. We questioned if key Treg genes such as Foxp3, and/or class IIa HDAC-dependent proteins could be affected by the loss of HDAC5, and thus provide a molecular mechanism to explain these data. We found that HDAC5^−/− Treg showed no significant differences in cytotoxic T-lymphocyte-associated protein (CTLA)-4 or Foxp3 gene expression, although a trend towards lower Foxp3 mRNA expression was noted (Figure 5a). At the protein level, Foxp3 protein was persistently reduced in HDAC5^−/− Treg (Figure 5b and c). Given previous data on class IIa HDAC deficient mice, especially HDAC9 ⁷, the findings of decreased Treg function, iTreg conversion, and Foxp3 protein expression in HDAC5^−/− mice were somewhat surprising. We hypothesized that specific client proteins and genes regulated by HDAC5, such as Glut4 or of Forkhead box-O1 (Foxo1), might be responsible. While Glut4 was undetectable in Tregs or Tconv, we found Foxo1 to be slightly more phosphorylated (T24), in HDAC5^−/− versus WT Treg cells (Figure 4d and e). Foxo1 acetylation (K259, K262 and K271) trended slightly higher, but the difference was very subtle and not persistent (n=6, Figure 4d and e). In Tconv, Foxo1 acetylation and phosphorylation was not affected by lack of HDAC5 (Supplemental figure S2). Taken together, these data show that Treg lacking HDAC5 have less Foxp3 protein and only slightly more phosphorylated Foxo1 protein without significant changes in Foxo1 acetylation nor Foxp3 or Ifng mRNA expression.

Figure 5.

HDAC5 deletion does not convey improved anti-tumor immunity. (a) TC1 murine lung adenocarcinoma volume curves and (b) final tumor weights at the end of the experiment show no difference between 13 HDAC5^−/− and 14 littermate control mice pooled from two independent experiments. (c) Tumor specific CD8⁺ T cells retrieved from the TC1 tumor are reduced among all tumor-infiltrating lymphocytes (TIL). (d) Representative and (e) cumulative data showing CD8⁺ T cells retrieved from the TC1 tumors produce less IFNγ upon anti-CD3ε/CD28 mAb stimulation. (f) Western blot of TC1 tumors retrieved from HDAC5^−/− and littermate control mice showing equal Foxp3 protein. (c-f) Data obtained from TC1 tumors of eight WT and three HDAC5^−/− mice.

Figure 4.

Loss of HDAC5 reduces Foxp3 protein expression. (a) Quantitative PCR shows a trend towards lower Foxp3 mRNA expression without reaching significance. Data displayed as median with range (3/group). (b, c) Western blot showing a trend to reduced Foxp3 expression in HDAC5^−/− Treg. (b) Representative and (c) densitometry data pooled from four independent experiments (paired t-test, 4/group). (d) Western blot showing FOXO1 protein expression in WT and HDAC5^−/− Treg cells. HDAC5^−/− Treg exhibit equal FOXO1 acetylation (K259, K262 and K271) and slightly increased phosphorylation (T24). (e) Densitometry pooled from six independent experiments and shown with median and interquartile range. P-values indicate Wilcoxon matched-pairs signed rank test.

Loss of HDAC5 does not affect anti-cancer immunity

The combination of overall healthy mice without overt autoimmune disease, yet diminished Treg function and poor iTreg conversion, combined with potential direct anti-neoplastic effects of HDAC5 targeting ²⁸, suggested HDAC5 might be an interesting target for tumor immunotherapy. We tested this by injecting 1 × 10⁶ TC1 tumor cells (murine lung adenocarcinoma ²⁷, immortalized with HPV) into HDAC5^−/− or littermate controls, and following the mice for tumor growth. Cumulatively, loss of HDAC5 did not confer enhanced anti-tumor immunity. Both tumor volume and tumor weight collected at the end of the experiment showed no differences between HDAC5^−/− and littermate control groups (Figure 5a and b). Upon flow cytometry, we noted that HDAC5^−/− mice had less tumor-specific infiltrating CD8⁺ T cells directed against the E7 antigen (Figure 5c). Furthermore, tumor-infiltrating HDAC5^−/− CD8⁺ T cells subjected to CD3ε/CD28 stimulation produced less IFN-γ compared to their WT counterparts (Figure 5d and e). Tumor tissues retrieved at the end of the experiment showed comparable Foxp3 protein expression (Figure 5f). Likewise, draining lymphnodes and spleens recovered from the tumor bearing mice showed no difference in CD4⁺Foxp3⁺ Treg by flow cytometry (Supplemental figure S3). In summary, deletion of HDAC5 does not confer improved anti-tumor immunity in the TC1 model. Loss of HDAC5 may interfere with CD8⁺ T cell recruitment or function.

HDAC5 deletion diminishes IFN-γ production in CD8 T cells

The decreased IFN-γ expression by tumor infiltrating CD8⁺ T cells suggested that deletion of HDAC5 might impair IFN-γ production in that cell subset. To investigate, we isolated CD8⁺ T cells from normal HDAC5^−/− mice, and stimulated them with CD3ε/CD28 monoclonal antibodies, as well as with PMI/ionomycin. Similarly to the findings on tumor infiltrating lymphocytes, HDAC5^−/− CD8⁺ T cells produced significantly less IFN-γ (Figure 6a and b). We explored HDAC2 S394 phosphorylation as a signaling pathways potentially affected by loss of HDAC5, but found that this was unaffected by HDAC5 deletion (Figure 6c). In summary, loss of HDAC5 interferes with the ability of CD8⁺ T cells to produce IFN-γ in response to CD3ε/CD28 or PMA/ionomycin stimulation.

Figure 6.

HDAC5 deletion impairs CD8⁺ T-cell IFN-γ production. (a) CD8⁺ T cells from HDAC5^−/− (H5) mice produced less IFN-γ when stimulated with PMA/ionomycin. (b) Cumulative data showing reduced IFN-γ production in HDAC5^−/− CD8⁺ T-cells normalized to each individual WT control. Data pooled from three independent experiments. (c) Assessment of HDAC2 phosphorylation (S394). CD8⁺ and CD4⁺ T-cells show equal phosphorylation between HDAC5^−/− and WT mice. CD8⁺ and CD4⁺ Tconv cell stimulation was achieved via anti-CD3ε/CD28 co-stimulation for 2 h.

Discussion

The finding that targeting of HDAC5 decreased Treg suppressive function and impaired iTreg formation was initially surprising. When we started our investigation, we expected, based upon previous studies of other class IIa HDACs, that HDAC5^−/− mice would have improved Treg function and iTreg formation.^{11, 23} Considering that we found the opposite, and combined with the observation that HDAC5^−/− Treg had less Foxp3 protein expression than WT Treg cells, we concluded that HDAC5 and HDAC9 may well have different roles in Foxp3⁺ Treg cells.

One potential difference might relate to the biology of Forkhead box-O1 protein. HDAC5, together with HDAC4, is known to control the deacetylation of Foxo1, specifically at lysine residues K259, K262, and K271.³⁵ Foxo1 acetylated at these residues is transcriptionally inactive, becomes phosphorylated, and is exported out of the nucleus.³⁶ Foxo1 is important in Treg biology and maintaining a Treg phenotype ³⁷, in part by suppressing Ifng mRNA expression in Treg.³⁸ T cells with estrogen receptor-cre mediated deletion of Foxo1 fail to form Foxp3⁺ iTregs under polarizing conditions, somewhat reminiscent of the HDAC5^−/− Tconv described in this work.³⁹ However, in our study, we did not observe any significant increase Foxo1 acetylation, nor in Ifng mRNA transcription in HDAC5^−/− Tregs, and only a slight increase in Foxo1 phosphorylation. Foxp3 gene transcription in Tregs showed a trend to be reduced in HDAC5^−/− Tregs, but this did not reach statistical significance. Therefore, it seems unlikely that changes in Foxo1 acetylation due to the absence of HDAC5 are a sufficient explanation for the observed Treg phenotype.

Since HDAC5^−/− mice had weak Tregs and a limited capability to induce iTreg cells, while at the same time exhibiting relatively normal CD4 Tconv function and no overt autoimmune disease, we hypothesized that HDAC5 might be a promising target for enhancing anti-tumor immunity. In addition, there are data showing potential direct anti-neoplastic effects of HDAC5 targeting, including use of the HDAC5 inhibitor, LMK235, to reduce chemotherapy resistance in vitro.²⁸ However, in our TC1 tumor model, HDAC5^−/− mice did not exhibit any improved anti-tumor immunity. We did notice that TC1 tumors in HDAC5^−/− had weaker CD8 T-cells that produced less IFN-γ, and formed fewer tumor antigen-specific infiltrating CD8⁺ T cells. Furthermore, decreased IFN-γ production was also shown in CD8 T cells isolated from healthy HDAC5^−/− mice. We speculate that the reduced suppressive function of HDAC5^−/− Treg cells may be offset by impaired CD8⁺ T cell responses. Both Treg and CD8⁺ T cells are key regulators of host anti-tumor immune responses.⁴⁰

The observation that HDAC5^−/− CD8⁺ T cells produced less IFN-γ opened additional possibilities for identifying joint mechanisms affecting both CD8 and Treg cells, and, to explain the discrepancy between HDAC5^−/− compared to HDAC9^−/− phenotypes. For example, HDAC2 has been reported to be deacetylated at K75 by HDAC5, but not HDAC9.⁴¹ Once deacetylated, HDAC2 becomes accessible to Casein-kinase 2α1 mediated S394 phosphorylation, which activates its deacetylase activity.^{41, 42} This is relevant to CD8 T-cell function as HDAC2 has been shown to repress Runx3 ⁴³, which is important for CD8 T-cell function.⁴⁴ It is therefore possible that overactive HDAC2 due to loss of HDAC5 deacetylation might impair the development of cytotoxic CD8⁺ T cells. Moreover, the HDAC2-Runx axis would be a very fitting explanation for the Treg data, given that Runx1 and Runx3 are important for Treg function and iTreg cell development.^{45, 46} Finally, this hypothesis could also reconcile the discordant Treg function and iTreg formation data between HDAC5^−/− compared to HDAC9^−/− mice, since only HDAC5, but not HDAC9, deacetylates HDAC2 at K75.⁴¹ However, in our hands, HDAC5^−/− and control mice did not display differential S394 phosphorylation of HDAC2. Both CD8 and CD4 T-cells, as well as Treg, showed equal HDAC2 phosphorylation at baseline and under stimulation. Thus another mechanism may be involved, and further studies on how HDAC5 affects CD8⁺ T cell IFN-γ production will be needed.

Importantly, our findings do not rule out pharmacologic HDAC5 targeting as a therapeutic strategy. On the contrary, in contrast to other class IIa HDACs, the loss of HDAC5 did not strengthen, but rather weaken Tregs, which remains an important distinction to other HDACs. However, we propose that any evaluation of a future HDAC5 inhibitor for anti-neoplastic therapy in vivo should also include assessment of effects on anti-tumor immunity in general and CD8 T cell function in particular.

In summary, loss of HDAC5 weakens Treg suppressive function and iTreg formation, as well as IFN-γ production in CD8⁺ T cells. Mice lacking HDAC5 do not develop spontaneous illness and do not have enhanced anti-tumor immunity.

What’s new?

Current histone/protein deacetylase (HDAC) inhibitors in cancer therapy have significant toxicity and risk unwanted potentiation of regulatory T cells (Treg) in the tumor microenvironment, which can help tumors escape the immune response. We found that, in contrast to most other HDAC isoforms, HDAC5 deletion weakened Treg function in vitro and in vivo, but also limited IFN-γ production by CD8 T cells. These findings are important for the development of isoform-selective HDAC inhibitors in cancer treatment.