HDAC3 Is Required for the Downregulation of RORγt during Thymocyte Positive Selection

The generation of immune cells is dependent on epigenetic machinery that controls gene expression at each developmental stage. Lysine acetylation is a well-known posttranslational modification added by histone acetyltransferases or removed by histone deacetylases (HDACs) to regulate chromatin structure and gene expression. When histone acetyltransferases add acetyl groups, this neutralizes the positively charged lysine, leading to the loosening of DNA around histones and the occurrence of gene transcription (1). Removal of acetyl groups by HDACs leads to chromatin condensation and gene repression. There are 18 HDAC enzymes grouped into four classes (I–IV) based on their domain organization and function (1). Depending on the HDAC, they are ubiquitously expressed or tissue specific.

HDAC3 belongs to the class I HDAC family (1, 2). Traditionally, HDAC3 assembles with the nuclear receptor corepressor and silencing mediator of retinoic and thyroid receptors (3–5). Together with different transcription factors, HDAC3 functions as the catalytic component of nuclear receptor corepressor/silencing mediator of retinoic and thyroid receptors complexes to deacetylate histones at specific promoters to mediate gene silencing (6). Somatic deletion of HDAC3 leads to embryonic lethality, whereas tissue-specific deletion leads to hypertrophy in liver and heart (7), failure of hematopoietic stem cell maintenance (8), defects in peripheral T cell maturation (9), invariant NKT cell development (10), and regulatory T cell (Treg) dysfunction (11).

In this study, we examined the role of HDAC3 in early T cell development (T cell development reviewed in Refs. 12, 13). Briefly, T cell precursors from the bone marrow migrate to the thymus where they commit to the T cell lineage at the double-negative (DN; CD4⁻CD8⁻) stage DN2. DN3 thymocytes undergo TCRβ rearrangement and β-selection to test for proper TCRβ rearrangement and expression. Post–β-selection thymocytes (DN3b, DN4, and immature single-positive [ISP]) proliferate before transitioning to the double-positive (DP; CD4⁺CD8⁺) stage. After rearranging their TCRα-chain, DP thymocytes that recognize self-MHC weakly are positively selected and transition to the SP stage (14). Thymocytes that fail to engage with MHC die by neglect (15). Thymocytes that express a TCR with strong affinity to self-peptide presented by MHC die by negative selection (14). Positively selecting thymocytes undergo CD4-versus-CD8 lineage commitment (16), maturation (17, 18), and exit the thymus to become peripheral CD4 and CD8 T cells.

In the present study, we demonstrated that HDAC3 is required for positive selection. HDAC3-deficient DP thymocytes fail to upregulate Bcl-2, leading to enhanced apoptosis at the SP stage. The enhanced apoptosis was not due to a defect in TCR signaling or enhanced negative selection, but due to the failure to downregulate retinoic acid–related orphan receptor (ROR) γt during positive selection. Knocking out RORγt largely rescued this defect in HDAC3-deficient thymocytes. Thus, HDAC3 is essential for the downregulation of RORγt during positive selection.

HDAC3 fl/fl mice (19) and IL-7Rα transgenic (Tg) mice (20) were previously described. Human Bcl-2 Tg mice were generated by S. Korsmeyer (Dana–Farber Cancer Institute, Boston, MA) (21) and provided by A. Singer (National Institutes of Health, Bethesda, MD). Bcl-x_L Tg mice (22), RORγt-knockout (KO) mice (23), and CD2-icre mice (24) were purchased from The Jackson laboratory. OT-II mice (25) were purchased from Taconic. Mice were housed in barrier facilities and experiments were performed at the Mayo Clinic with the approval of the Institutional Animal Care and Use Committee. All mice were analyzed between the ages of 6 and 14 wk. All genetically modified mice were examined with either littermate or age-matched controls, which may include floxed only mice (no Cre), CD2-icre, or WT mice, as no differences were observed between these mice. For convenience, the control mice in each experiment are termed wild-type (WT) but may represent either floxed only, CD2-icre, or WT mice.

FACS analysis was performed on an LSR II flow cytometer (BD Biosciences) or Attune NxT flow cytometer (Thermo Fisher), and all experiments were analyzed using FlowJo (Tree Star). Cytoplasmic and nuclear proteins were examined via intracellular flow cytometry. Thymocytes or mesenteric lymphocytes were labeled with surface markers before being fixed and permeabilized with a Foxp3/Transcription Factor Staining Buffer Set (for nuclear protein staining; eBioscience) or an intracellular fixation and permeabilization buffer kit (for cytoplasmic protein staining; eBioscience). All analyses included size exclusion (forward scatter [FSC] area/side scatter [SSC] area), doublet exclusion (both FSC height/FSC width and SSC height/SSC width), and dead cell exclusion (fixable viability dye; eBioscience). All other reagents for flow cytometry were purchased from BD Biosciences, eBioscience, BioLegend, Tonbo Biosciences, or Abcam.

Cell cycle analysis was performed using a Click-iT EdU imaging kit (Molecular Probes) and Vybrant DyeCycle Violet stain (Molecular Probes). EdU (10 μM) was added to freshly isolated thymocytes in complete media (RPMI 1640 media with 10% FCS with penicillin/streptomycin/glutamine) for 2 h at 37°C. Cells were then harvested for surface and intracellular EdU staining per the manufacturer’s guidelines. Vybrant DyeCycle Violet was then used to stain for DNA content in HBSS for 30 min at 37°C.

Mixed bone marrow chimeras (BMCs) were generated by i.v. injecting 4 × 10⁶ cells from either mixes of WT (CD45.1^−/−)/B6.SJL (CD45.1^+/+) or CD2-icre HDAC3 conditional KO (cKO) (CD45.1^−/−)/B6.SJL (CD45.1^+/+) mice at a 50:50 ratio into lethally irradiated congenic B6.SJL (CD45.1^+/+) recipients. Recipient mice received the antibiotic enrofloxacin in their drinking water for 3 wk and were analyzed after 9 wk.

Single-cell suspensions of total thymocytes were incubated at 37°C for 0, 2, 4, 8, 16, and 24 h in complete culture medium (RPMI 1640, 10% FCS, penicillin/streptomycin/glutamine) before surface staining for thymocyte populations as well as with annexin V (BD Biosciences) and fixable viability dye (eBioscience). After staining at each time point, cells were fixed with the IC fixation buffer from the intracellular fixation and permeabilization buffer kit (eBioscience) to allow FACS analysis of all time points simultaneously.

Chromatin immunoprecipitation (ChIP) was performed on total thymocytes from WT or HDAC3-cKO mice using the SimpleChIP enzymatic chromatin IP kit (Cell Signaling Technology, catalog no. 9002) per the manufacturer’s guidelines. Chromatin was incubated with α-acetyl histone 3 Ab (Millipore) and control Abs (Cell Signaling Technology, catalog no. 9002) overnight at 4°C with rotation. Extracted DNA was analyzed by real-time PCR using previously published primer sequences spanning the RORC promoter (26).

For cytokine analysis, isolated lymphocytes from mesenteric lymph nodes (mLNs) were stimulated or left unstimulated overnight in complete culture media (RPMI 1640) with 200 nM PMA and 1 μM ionomycin. Three hours after stimulation, brefeldin A/monensin (500×) from BD Biosciences was added to stop secretion of cytokines for intracellular flow cytometry analysis. Cells were harvested the following day for intracellular cytokine staining.

Mouse colon was harvested and fixed in 10% formalin for 24 h at room temperature and then stored in 70% ethanol at 4°C until processing by paraffin embedding, sectioning, and standard H&E staining techniques. Images were viewed on a Leica DMI3000 B microscope at ×10 and ×20 magnification, captured using the Leica EC3 camera, and processed with the Leica Application Suite EZ software.

Data comparing absolute cell number, percentage of annexin V⁺ cells, and ChIP between experimental groups were analyzed using a two-tailed unpaired Student t test. Analysis of percentage of cells in S phase was determined by a two-tailed paired Student t test. Ex vivo viability time course assay was analyzed using two-way ANOVA using GraphPad Prism.

To investigate the role HDAC3 plays in T cell development, we generated CD2-icre HDAC3^fl/fl mice (afterward referred to as HDAC3-cKO). CD2-icre expresses Cre in a lymphoid progenitor (27), resulting in Cre-mediated deletion of HDAC3, affecting development of lymphocytes early in both T cell and B cell lineages. However, in this study, we focus only on T cell development. We used flow cytometry to identify at which stage of thymic development elimination of HDAC3 protein was complete. HDAC3 protein expression was reduced at the DN3 stage and eliminated in DN4 thymocytes onward, equivalent to an isotype control (Supplemental Fig. 1A).

T cell development was severely blocked in HDAC3-cKO mice as compared with WT mice. There was a 2-fold decrease in total thymocyte cell numbers in HDAC3-cKO mice (Fig. 1A). HDAC3-cKO mice had an increased frequency of DN cells and a decreased frequency of CD4SP thymocytes as compared with WT mice (Fig. 1B). The CD4⁻CD8⁺ gate contains both CD8SP and ISP thymocytes, which can be distinguished by TCRβ expression (28). Almost all CD8SP thymocytes in HDAC3-cKO mice were TCRβ⁻ ISPs (Fig. 1C). Examining absolute cell numbers, there was a small increase in DN cells, a decrease in DP cells, and a severe defect in TCRβ⁺ SP thymocytes compared with WT mice (Fig. 1D).

HDAC3-cKO mice exhibited an increase in DN thymocytes and a decrease in DP thymocytes as compared with WT mice. Therefore, we analyzed DN thymocytes to determine whether there was a defect in β-selection or proliferation. After excluding lin⁺CD3⁺ cells, c-Kit and CD25 delineates early thymic progenitors (c-Kit⁺CD25⁻), DN2 (c-Kit⁺CD25⁺), DN3 (c-Kit⁻CD25⁺), and DN4 (c-Kit⁻CD25⁻) thymocytes. An increased frequency of DN3 cells and a decreased frequency of DN2 and DN4 cells were observed in HDAC3-cKO mice compared with WT mice (Fig. 1E). However, absolute cell numbers revealed a 2-fold increase in only DN3 cells in HDAC3-cKO mice (Fig. 1F). β-Selection occurs between DN3a and DN3b stages, with cells increasing CD27 and FSC after β-selection (DN3b) (29). The frequency of DN3b cells was not altered in HDAC3-cKO mice (Fig. 1G), demonstrating that HDAC3 is not required for thymocyte β-selection. HDAC3 is also required for DNA replication in hematopoietic progenitor cells (8). To determine whether the loss of HDAC3 is also required for DNA replication after β-selection, we analyzed proliferation by examining EdU incorporation. Cell cycle stages were distinguished using Vybrant DyeCycle Violet (DNA content) and EdU (G₀/G₁, EdU⁻Violet^lo; S phase, EdU⁺; G₂/M, EdU⁻Violet^hi). Compared to WT thymocytes, HDAC3-deficient thymocytes showed a significant increase in the percentage of DN3 cells in S phase (Fig. 1H, left column, 1I) but no change in DN4 thymocytes (Fig. 1H, right column, 1I). Increased proliferation of DN3 thymocytes could explain the increase in cell number of DN3 thymocytes in HDAC3-cKO mice. Thus, DN thymocytes maintain the capability to proliferate after β-selection in the absence of HDAC3. Additionally, no changes in viability were observed in DN3 and DN4 cells from WT and HDAC3-cKO mice (data not shown). Therefore, HDAC3-cKO mice do not exhibit a block in T cell development at the DN stage.

HDAC3-cKO mice showed a dramatic decrease in both the frequency and number of TCRβ⁺ CD4SP and TCRβ⁺ CD8SP thymocytes (Fig. 1A, 1B), suggesting that HDAC3 is required for positive selection. To analyze this, the TCRβ-versus-CD69 profile was examined in these mice. Thymocytes that successfully rearrange their TCRα-chain and express TCR on the surface but have not yet received a signal through their TCR are TCRβ^intCD69⁻ (immature). TCR signaling induces CD69 expression (TCRβ^intCD69⁺; selecting) and successful positive selection leads to TCRβ upregulation (TCRβ⁺CD69⁺; post–positive selection) (30). Maturation of SP thymocytes leads to downregulation of CD69 in preparation for egress (TCRβ⁺CD69⁻; mature). Compared to WT mice, HDAC3-deficient thymocytes exhibited a reduced frequency and cell number of TCRβ^intCD69⁺ thymocytes, followed by a severe reduction in the frequency and cell number of postselection (TCRβ⁺CD69⁺) and mature thymocytes (TCRβ⁺CD69⁻) (Fig. 2A, 2B). Similar defects are observed using a CD24-versus-TCRβ profile, where there was a dramatic decrease in the frequency of mature CD24^int/loTCRβ⁺ thymocytes (Fig. 2A). Although CD24 expression is lower in DP thymocytes from HDAC3-cKO mice, it would not contribute to the block in positive selection observed in these mice, as T cell development is normal in CD24-KO mice (31). Therefore, HDAC3 has a unique and critical role during positive selection that is not compensated for by any other coexpressed HDAC family member. A positive selection block was not observed in CD4-cre HDAC3-cKO mice (32), which may be attributed to HDAC3 protein expression in DP thymocytes (Supplemental Fig. 1B). HDAC3 has a preference for acetylated histone H3 Lysine-9 (H3K9Ac) (33, 34), an epigenetic mark associated with active gene expression. Using an Ab specific for H3K9Ac in WT mice, global H3K9Ac was reduced as cells progressed from positive selection through maturation (Fig. 2C). However, in the absence of HDAC3, this mark remained high (Fig. 2C). Similarly, H3K9Ac decreases from DP to SP thymocytes in WT mice but was maintained at high levels in HDAC3-deficient SP thymocytes (Fig. 2D). Therefore, in the absence of HDAC3, downregulation of H3K9Ac during thymocyte development and maturation does not occur. Examination of semimature SP thymocytes (TCRβ⁺CD24⁺) in HDAC3-cKO mice revealed a failure to upregulate TCRβ, Egr2, and CCR7 as compared with WT mice (Fig. 2E). Semimature SP thymocytes were examined, as a large fraction of mature SP thymocytes (TCRβ⁺CD24^lo) in HDAC3-cKO mice were recirculating T cells that did not express Rag1-GFP (35) (Supplemental Fig. 2). We also examined whether the lack of SP thymocytes was due to increased negative selection. Enhanced negative selection would result in increased levels of Nur77, Bim, and Helios. However, no increase in Nur77, Bim, or Helios was observed in DP and semimature CD4SP thymocytes from HDAC3-cKO mice (Fig. 2F). Therefore, HDAC3 is required for thymocyte positive selection.

To determine whether the block in positive selection was cell intrinsic, we generated mixed BMCs with a 50:50 combination of CD45.1⁻ WT or HDAC3-cKO bone marrow cells with CD45.1⁺ B6.SJL bone marrow cells. The mice were analyzed 9–11 wk after transfer. Positive selection was normal in WT mice compared with B6.SJL mice (Fig. 2G). However, HDAC3-cKO mice still exhibited a positive selection defect that could not be rescued by the presence of WT B6.SJL cells (Fig. 2G). Additionally, ∼99.9% of T cells in the periphery from HDAC3-cKO/B6.SJL chimeric mice were CD45.1⁺, consistent with the severe block in T cell development (data not shown). Therefore, the HDAC3-dependent block in thymocyte positive selection is cell intrinsic.

To confirm a defect in positive selection, we generated OT-II–HDAC3-cKO mice. The OT-II TCR transgene recognizes OVA presented by I-A^b, leading to the generation of CD4SP but not CD8SP thymocytes (25). In contrast to OT-II thymocytes, OT-II–expressing HDAC3-cKO thymocytes failed to successfully navigate positive selection even though CD69 was efficiently induced (Fig. 3A). With OT-II and OT-II–HDAC3-cKO mice not being in a Rag-KO background, examination of Vβ5 and Vα2 expression to detect the OT-II transgene in SP thymocytes revealed very few Vβ5^hiVα2^hi CD4SP thymocytes in OT-II–HDAC3-cKO mice compared with OT-II mice (Fig. 3B, 3C), whereas the number of Vβ5^hiVα2^hi CD8SP thymocytes was equivalent (Fig. 3C). Furthermore, very few peripheral OT-II T cells were present in OT-II–HDAC3-cKO mice (Fig. 3D). Therefore, expression of an OT-II TCR transgene does not rescue positive selection in HDAC3-cKO mice.

Successful positive selection leads to the induction of Bcl-2 expression and SP thymocyte survival (36). To examine whether apoptosis contributed to the lack of SP thymocytes, cell death was analyzed. CD4SP and CD8SP thymocytes from HDAC3-cKO mice exhibited an ∼6-fold increase in annexin V staining as compared with WT mice, indicative of enhanced apoptosis (Fig. 4A, 4B) independent of negative selection (Fig. 2E). The viability of HDAC3-deficient thymocytes was examined ex vivo to determine whether the enhanced cell death was cell intrinsic. After 24 h in culture without stimulation, 76% of WT TCRβ⁺ CD4SP thymocytes were viable whereas only 44% of HDAC3-deficient TCRβ⁺ CD4SP thymocytes were viable (Fig. 4C,). TCRβ⁺ CD8SP thymocytes from HDAC3-cKO mice exhibited a similar decrease in viability (data not shown). Bcl-2 is critical for SP thymocyte survival and is upregulated after positive selection (36). IL-7Rα is also expressed after positive selection and is important for homeostasis in the periphery (37). Whereas WT TCRβ^loCD69⁺ DP thymocytes undergoing positive selection and TCRβ⁺CD69⁺ DP thymocytes that recently completed positive selection upregulated IL-7Rα and Bcl-2 (Fig. 4D), neither IL-7Rα or Bcl-2 expression was induced in the absence of HDAC3 (Fig. 4D). To determine the relative contribution of IL-7Rα or Bcl-2 expression to the lack of CD4SP and CD8SP thymocytes in HDAC3-cKO mice, we crossed HDAC3-cKO mice to IL-7Rα Tg mice or Bcl-2 Tg mice. Transgenic expression of IL-7Rα did not alter the reduced SP thymocyte frequency (Fig. 4E) or cell number (Fig. 4F) observed in HDAC3-cKO mice. Similarly, HDAC3-deficient thymocytes with the Bcl-2 transgene (Bcl-2 Tg–HDAC3-cKO) did not exhibit restored SP thymocyte development (Fig. 4G). Whereas total SP thymocyte numbers in Bcl-2 Tg–HDAC3-cKO mice were increased as compared with HDAC3-cKO mice (Fig. 4H), they showed a similar fold decrease when compared with Bcl-2 Tg mice. Given that a large fold reduction in SP thymocytes from WT to HDAC3-cKO mice was maintained with introduction of a Bcl-2 transgene (10-fold for Bcl-2 Tg to Bcl-2 Tg–HDAC3-cKO mice), it can be concluded that Bcl-2 expression does not rescue the block in T cell development in HDAC3-cKO mice. Thus, the defects observed in HDAC3-cKO mice are not simply due to altered survival due to the inability to induce expression of IL-7Rα or Bcl-2 during positive selection.

Consistent with the role of HDAC3 in extinguishing rather than inducing gene expression, our focus shifted to genes required to be downregulated during positive selection. Expression of the orphan nuclear receptor RORγt is high in DP thymocytes and then decreases upon positive selection (38). Constitutive expression of RORγt using a transgene (RORγt Tg) causes thymocyte development to be arrested at the DP stage due to a block in positive selection, with decreased thymic cellularity and lower surface TCRβ expression on the few SP thymocytes that develop (39). This phenocopy between RORγt Tg mice and HDAC3-cKO mice led us to examine whether RORγt expression was altered in HDAC3-cKO mice. As expected, RORγt was downregulated in selecting (TCRβ^loCD69⁺) and postselection (TCRβ⁺CD69⁺) DP thymocytes in WT mice (Fig. 5A). In contrast, HDAC3-deficient selecting (TCRβ^loCD69⁺) and postselection (TCRβ⁺CD69⁺) DP thymocytes maintained high levels of RORγt expression (Fig. 5A). This high level of RORγt was also maintained at the semimature SP stage (Fig. 5B). We examined whether histone acetylation was enhanced at the RORC promoter using a series of previously published primers for ChIP (26). H3 acetylation was significantly increased (p = 0.034–0.049) at three different sites upstream of exon 1, although histone acetylation at sites that overlap with exon 1 was similar and only minor differences were observed at sites further upstream (Fig. 5C). As RORγt expression is primarily regulated at the transcriptional level at this stage (39), increased acetylation at the RORC promoter in HDAC3-cKO mice demonstrates that deacetylation of its promoter and thus the transcriptional downregulation of RORγt during positive selection depends on HDAC3, which is not compensated by any other HDAC coexpressed during T cell development.

To test whether the failure to downregulate RORγt in HDAC3-cKO mice is responsible for the block in positive selection, we interbred HDAC3-cKO mice with RORγt-KO mice to correct for the failure to downregulate RORγt upon positive selection. However, RORγt-KO mice also have a block in T cell development, as RORγt expression is required for DP thymocyte survival through regulating Bcl-x_L expression (40, 41). Transgenic expression of Bcl-x_L rescues the DP thymocyte survival defect and consequently the block in T cell development in RORγt-KO mice (41). Therefore, we generated RORγt-KO Bcl-x_L Tg HDAC3-cKO mice (hereafter referred to as RB3 mice) to determine whether the deletion of RORγt (in conjunction with constitutive Bcl-x_L expression) would restore positive selection in HDAC3-cKO mice. Examination of RB3 mice revealed that thymic cellularity was restored to WT and control RORγt Bcl-x_L Tg levels (Fig. 6A), whereas interbreeding Bcl-x_L Tg or RORγt-KO mice separately to HDAC3-cKO mice did not restore thymic development (Supplemental Fig. 3A) or thymic cellularity (Fig. 6A). In fact, HDAC3-cKO combined with RORγt deficiency exacerbated the defect in thymic development as compared with each single KO alone, leading to a block at the ISP stage (Supplemental Fig. 3A). However, whereas thymic cellularity in RB3 mice was similar to WT and control RORγt-KO Bcl-x_L Tg mice (Fig. 6A), the CD4-versus-CD8 profile was aberrant (Fig. 6B). Both RORγt-KO Bcl-x_L Tg and RB3 mice exhibited a reduced frequency of DP thymocytes and enhanced frequency of CD4⁻CD8⁺ cells compared with WT and HDAC3-cKO mice (Fig. 6B). This pattern suggests that the CD4⁻CD8⁺ gate includes DP thymocytes that lack CD4. Hence, we used the Immunological Genome Project Consortium (http://www.immgen.org) to pinpoint surface markers that distinguish DP thymocytes from ISPs. We identified CCR9 as a suitable target, where ISPs express CCR9 at a much lower level than DP thymocytes, and confirmed this differential expression in WT mice by flow cytometry (Fig. 6C). Using DP expression of CCR9 as a positive control, the TCRβ⁻ CD4⁻CD8⁺ population in RORγt-KO Bcl-x_L Tg and RB3 mice was bimodal, containing both CCR9⁻ and CCR9⁺ populations (Fig. 6C, Supplemental Fig. 3B). This demonstrates that CCR9⁺ TCRβ⁻ CD4⁻CD8⁺ thymocytes likely are CD4⁻ DP thymocytes. CD4⁻CD8⁺ thymocytes that express TCRβ also contained a DP signature in RORγt-KO Bcl-x_L Tg and RB3 mice because most TCRβ⁺ CD4⁻CD8⁺ thymocytes in these mice express CXCR4, characteristic of DP thymocytes, opposed to positively selected CD8SP thymocytes that do not express CXCR4 (Fig. 6D,Supplemental Fig. 3C). To identify positively selected CD8SP thymocytes within the TCRβ⁺ CD4⁻CD8⁺ cells, we examined Runx3 expression because this transcription factor is required for CD8-lineage differentiation after positive selection (42). Although all WT CD8SP thymocytes were Runx3⁺, most TCRβ⁺ CD4⁻CD8⁺ cells were Runx3⁻ in RORγt-KO Bcl-x_L Tg and RB3 mice (Fig. 6E), consistent with this population being CD4⁻ DP thymocytes. Interestingly, this population is also present in HDAC3-cKO (Fig. 6D, 6E), Bcl-x_L Tg HDAC3-cKO, and RORγt-KO HDAC3-cKO mice (Supplemental Fig. 3D). As a result, we used Runx3 expression to quantify positively selected CD8SP thymocytes, along with TCRβ⁺ CD4SP thymocytes, to determine whether deletion of RORγt (in conjunction with Bcl-x_L expression) rescued positive selection in HDAC3-cKO mice. CD8SP (TCRβ⁺, Runx3⁺) cell number was restored to WT levels in RB3 mice, but there was no rescue of CD8SP thymocytes in Bcl-x_L Tg HDAC3-cKO or RORγt-KO Bcl-x_L Tg mice (Fig. 6F). Additionally, we performed a phenotypic analysis on CD8SP thymocytes (TCRβ⁺, Runx3⁺) from RB3 mice to determine whether their expression profile was similar to WT CD8SP thymocytes. As WT DP thymocytes transition to the SP stage, TCRβ, CCR7, and MHC class I were upregulated, and CXCR4, CCR9, and CD24 were downregulated (Fig. 6G). These markers were expressed similarly on CD8SP thymocytes between WT and RB3 mice (Fig. 6G), demonstrating that CD8SP thymocytes from RB3 mice underwent an equivalent change in gene expression compared with normal CD8SP cells. Therefore, generation of CD8SP thymocytes demonstrates that deletion of RORγt rescues positive selection in HDAC3-cKO mice, but this rescue was not observed in CD4SP thymocytes.

The profound alteration in the CD8SP/CD4SP ratio indicates that HDAC3 may also play an important role the development of CD4SP thymocytes. To investigate this, expression levels of transcription factors required for CD4-lineage development were examined. We examined Gata3, Tox, and ThPOK expression in TCRβ^intCD69⁺ (selecting) and TCRβ⁺CD69⁺ (postselection) populations, as these transcription factors are required for CD4SP development during these stages. Gata3 and Tox are upregulated after DP thymocytes receive a TCR signal (43, 44), and WT thymocytes showed high levels of Gata3 and Tox within the TCRβ^intCD69⁺ population (Fig. 7). However, examination of RB3 mice revealed decreased expression of Gata3 and Tox in TCRβ^intCD69⁺ thymocytes (Fig. 7). ThPOK is the master transcription factor for CD4SP thymocyte differentiation (45, 46) and expression initiates in WT TCRβ^intCD69⁺ thymocytes (Fig. 7). Gata3 and Tox are required to induce ThPOK expression for CD4SP development (44, 47). Consequently, ThPOK was very low in TCRβ^intCD69⁺ and TCRβ⁺CD69⁺ thymocytes RB3 mice compared with WT mice (Fig. 6G), reflective of the few number of CD4SP thymocytes in RB3 mice. Therefore, transcription factors required for CD4SP development are decreased in RB3 mice.

In the course of our studies, HDAC3-cKO mice developed rectal prolapse as early as 8 wk of age. Because HDAC3-deficient DP thymocytes undergoing positive selection failed to downregulate RORγt (Fig. 5A), the few T cells produced in these mice might skew toward the RORγt⁺ Th17 lineage and lead to the development of inflammatory bowel disease (IBD). Analysis of spleens in HDAC3-cKO mice revealed very few T cells that displayed primarily a memory phenotype compared with WT mice (Supplemental Fig. 4A, 4B), indicative of homeostatic proliferation. Flow cytometry confirms that most peripheral T cells did not express HDAC3 in HDAC3-cKO mice and are not cells that escaped deletion (Supplemental Fig. 4C). Consistent with CD4-cre HDAC3-cKO mice (9), naive T cells in CD2-icre HDAC3-cKO mice had decreased CD55 expression (Supplemental Fig. 4D), reflective of a block in peripheral T cell maturation. Analysis of mLNs also revealed an overall decrease in T cell cellularity, but the proportion of CD4⁺ T cells in the mLN compared with WT mice was similar (Fig. 8A, 8B), as HDAC3 is also required for B cell development (R.L. Philips and V.S. Shapiro, unpublished results). Although Foxp3⁺ Tregs were present, there was a substantial increase in the frequency of RORγt⁺ Th17 cells within the Foxp3⁻ conventional CD4⁺ T cell pool (Fig. 8A). A similar increase in RORγt⁺ Th17 cells was also present in 3-wk-old mice (data not shown). Comparing absolute numbers of mesenteric Foxp3⁺ Tregs and RORγt⁺ Th17 cells, Tregs vastly outnumbered RORγt⁺ Th17 cells in WT mice, whereas there were slightly more RORγt⁺ Th17 cells than Tregs in HDAC3-cKO mice (Fig. 8C). These RORγt⁺ CD4⁺ T cells were functional Th17 cells, as they produced IL-17 upon stimulation in culture (Fig. 8D). An enhanced frequency of RORγt⁺ Th17 cells was also present in the spleen of HDAC3-cKO mice, but analysis of cellularity revealed equivalent cell numbers between WT and HDAC3-cKO mice (Supplemental Fig. 4E, 4F). Whereas HDAC3-cKO mice show enhanced Th17 differentiation in mLNs, there were equivalent frequencies of RORγt⁺ Tregs (Fig. 8G). RORγt⁺ Tregs have been shown to be important in constraining immunoinflammatory responses in the gut (48), and therefore an overall decrease in RORγt⁺ Treg number (Fig. 8G) along with a decrease in total Tregs and an increase in Th17 cell number (altered Treg/Th17 ratio) are consistent with the development of IBD. Examination of colon length revealed that HDAC3-cKO mice had a significantly shorter colon compared with WT mice (Fig. 8H). Additionally, H&E staining of the colon revealed thickening and remodeling of the lamina propria as well as immune cell infiltration (Fig. 8H). Reduced colon length and abnormal gut tissue both indicate IBD; therefore, the absence of HDAC3 leads to the generation of a high proportion of RORγt⁺ Th17 cells in the periphery and the development of IBD.

Lymphocyte development is tightly regulated, requiring successful transit of cells through several developmental stages and checkpoints prior to their release into the circulation. Each checkpoint involves induction or repression of a particular subset of genes, the disruption of which causes developmental arrest. We investigated the role of HDAC3 in T cell development using CD2-icre cKO (HDAC3-cKO) mice. Although T cells coexpress several HDAC family members during development (1, 49, 50), HDAC3 has a unique role, as conditional deletion of HDAC3 led to a block in T cell development at the DP stage due to an inability to undergo positive selection. The block in T cell development could not be rescued by an OT-II TCR transgene. Successful positive selection requires downregulation of RORγt, as mice with constitutive expression of RORγt have a similar block in T cell development at positive selection (39). In HDAC3-cKO mice, RORγt was not downregulated upon TCR stimulation at the DP stage, demonstrating that the block in positive selection may be due to an inability to downregulate RORγt. Consistent with this, we observed enhanced RORC promoter acetylation in the absence of HDAC3, indicating that HDAC3 may directly deacetylate histones at the RORC promoter to inhibit expression. RORγt also controls expression of Bcl-x_L at the DP stage to regulate DP survival (40, 41). Therefore, to determine whether sustained expression of RORγt was responsible for the block in positive selection, RB3 mice were generated, which largely restored positive selection, leading to restoration of CD8SP thymocyte (TCRβ⁺, Runx3⁺) numbers. However, CD4SP thymocytes were not generated, indicating a potential role for HDAC3 in lineage choice during T cell development. Therefore, HDAC3 is required at the DP stage for downregulation of RORγt during positive selection.

A recent study by Stengel et al. (51) also found that HDAC3 was required for thymocyte positive selection. However various aspects are different from our work, including that the block in positive selection was due to a defect in gene regulation that could be bypassed by a strong TCR signaling with an OT-II TCR transgene (51). Differences observed between both studies may stem from the mouse models used, where Stengal et al. used an Lck-cre for HDAC3 elimination compared with CD2-icre in our work. Lck-cre deletes as late as in the DN4 stage (52). Although an LSL-GFP reporter was used to indicate Cre-induced deletion, the timing of the loss of HDAC3 protein expression was not examined. We demonstrate that HDAC3 protein was almost entirely eliminated by the DN3 stage in CD2-icre HDAC3-cKO mice and is absent in subsequent stages. Therefore, this could explain why CD2-icre HDAC3-cKO mice revealed a more dramatic block in thymocyte positive selection than did Lck-cre HDAC3-cKO mice. Additionally, as TCR transgenes are typically expressed early, at the DN stage, the rescue of T cell development in OT-II Lck-cre HDAC3-cKO mice may reflect positive selection prior to Cre deletion and loss of HDAC3 expression. Our results clearly demonstrate that the block in positive selection cannot be rescued by OT-II transgene in the absence of HDAC3 protein.

Stengel et al. (51) used microarray and RNA sequencing to determine how gene regulation is disrupted in Lck-cre HDAC3-cKO mice and identified hundreds of genes that are differentially expressed. Based on differential gene expression, they hypothesized that TCR signaling is altered, leading to changes in positive selection. However, we demonstrate that the key gene downstream of HDAC3 required for positive selection is RORγt, as protein levels were not reduced with positive selection and CD8SP thymocyte development was restored to WT levels in RB3 mice. RORC mRNA levels were also increased in DP thymocytes from Lck-cre HDAC3-cKO mice examined by Stengel et al. (51). As few CD4SP thymocytes developed in RB3 mice, HDAC3 might play an additional role in thymocyte development during CD4-lineage choice. This decision is initiated after DP thymocytes receive a positive selection signal. After positive selection, thymocytes downregulate CD8 to become intermediate (CD4⁺CD8^lo) thymocytes that assess TCR signal persistence, as CD4 stabilizes the TCR interaction with MHC class II–restricted thymocytes and they become CD4SP thymocytes [reviewed in Singer et al. (16)]. This process is dependent on expression of Gata3, Tox, and ThPOK, as knockout models targeting each of these transcription factors leads to very few CD4SP thymocytes (44–47). With MHC class I–restricted thymocytes, the downregulation of CD8 in intermediate thymocytes destabilizes the TCR–MHC interaction, leading to termination of the TCR signal. IL-7R signaling then induces expression of Runx3 to suppress ThPOK and CD4 expression to become CD8SP thymocytes (53). CD69⁺ thymocytes in RB3 mice that had recently received a TCR signal showed decreased levels of Gata3, Tox, and ThPOK, demonstrating that the transcriptional program for the CD4-lineage was not induced. One possibility is that the lack of CD4 expression on DP cells in RB3 mice alters this test of MHC restriction. The absence of CD4 expression in DP thymocytes in RB3 as well as in RORγt-KO Bcl-x_L Tg and HDAC3-cKO mice suggests that HDAC3 may work in concert with RORγt to maintain CD4 expression in DP thymocytes because the effect seen in HDAC3-cKO mice and RORγt-KO Bcl-x_L Tg mice is augmented in RB3 mice, which is under investigation.

Current and previous work demonstrates HDAC3 to be required for multiple stages in T cell development, including positive selection, peripheral T cell maturation, and the generation of Tregs (9, 11, 51). Utilizing different mouse models that delete HDAC3 at different stages in development (CD2-icre versus CD4-cre, Supplemental Fig. 1), we have shown HDAC3 to be required for thymocyte positive selection and peripheral T cell maturation (9). Positive selection and T cell maturation (both thymic maturation and postthymic maturation) are distinct but equally important processes required to generate the peripheral naive T cell pool [reviewed by Hogquist et al. (18)]. As shown by Hsu et al. (9), many markers associated with thymic maturation were not altered in SP thymocytes from CD4-cre HDAC3-cKO mice, as measured by changes in CD69, CD62L, CD24, CCR4, CCR7, and CCR9 in Rag1-GFP⁺ SP thymocytes. Thymic egress, examined by Rag1-GFP and S1P1 expression, was also not altered in CD4-cre HDAC3-cKO mice (9). However, in the periphery, there were very few naive CD4 and CD8 T cells, and most naive T cells present were Rag1-GFP⁺ recent thymic emigrants. These HDAC3-deficient recent thymic emigrants were functionally immature and were targeted for elimination by complement in the periphery (9). Thus, HDAC3 is additionally required for postthymic T cell maturation as well as thymocyte positive selection. The peripheral T cell maturation defect was also observed in naive T cells from CD2-icre HDAC3-cKO mice (Supplemental Fig. 4), as demonstrated by low expression of CD45RB and CD55. Although positive selection of CD8 SP thymocytes was restored, RB3 mice also had a severe reduction in the number of splenic naive CD8 T cells (Supplemental Fig. 4G), consistent with a block in peripheral T cell maturation. As positive selection occurred normally in CD4-cre HDAC3-cKO mice (9) and RORγt was appropriately developmentally downregulated during positive selection (data not shown), peripheral T cell maturation in CD4-cre HDAC3-cKO mice is a process independent of altered RORγt expression. Therefore, RORγt deficiency/Bcl-x_L overexpression, which restores positive selection and the generation of CD8 SP thymocytes in CD2-icre HDAC3 cKO mice, would not be expected to rescue the defect in peripheral T cell maturation in the absence of HDAC3. Therefore, although HDAC3 is required to downregulate RORγt during positive selection, RB3 mice have few peripheral naive T cells, consistent with the previously characterized block in peripheral T cell maturation (9).

Although there is redundancy and overlap between HDAC family members in regulating histone deacetylation, H3K9Ac deaceylation was dependent on HDAC3 in this model. H3K9 modifications are associated with activation and repression, with H3K9Ac associated with gene activation (54) and di- or trimethylation of H3K9 associated with repression leading to condensed and transcriptionally inactive heterochromatin (55). H3K9 monomethylation is enriched at actively expressed genes, indicating it is not a repressive mark (55). Using flow cytometry, global H3K9Ac was analyzed during T cell development. H3K9Ac decreases as thymocytes progress through positive selection to become mature T cells. However, in the absence of HDAC3, downregulation of H3K9Ac remained high, demonstrating that the deacetylation of these targets is dependent on HDAC3. The identification of these targets and the role they have in regulating T cell development, lineage commitment, and maturation remain to be determined.

HDAC3-cKO mice developed rectal prolapse/IBD starting at 2 mo of age. Examination of T cell populations demonstrated that there were few peripheral Tregs in these mice, consistent with the recent work of Wang et al. (11) determining that HDAC3 is required for induced Treg development and function in CD4-cre HDAC3-cKO mice. However, a substantial number and proportion of conventional T cells in CD2-icre HDAC3-cKO mice were Th17 T cells that expressed RORγt and produced IL-17 upon stimulation. As a result, the altered Treg/Th17 ratio likely contributes to IBD development. Additionally, young mice (3 wk old) also showed an increased frequency of RORγt⁺ Th17 cells before any signs of IBD appear, suggesting that this enhancement is not due to impaired Treg function but a failure to regulate RORγt expression by HDAC3. However, the frequency of RORγt⁺ Tregs in HDAC3-cKO mice was similar to WT levels, suggesting that additional mechanisms are required for regulating RORγt expression in Tregs. Therefore, HDAC3 is required for regulating RORγt expression in the thymus and periphery.

Currently, HDAC inhibitors are in clinical trials for a variety of cancers. However, our study introduces the possibility of an unwanted side effect where inhibitors of HDAC3 would be predicted to block lymphocyte output, thus decreasing entry of new naive T cells into the peripheral T cell pool. As tumor-reactive T cells are often tolerized, cancer immunotherapy depends on a diverse naive T cell pool for efficacy. Thus, therapies that target HDAC3 that interfere with the ability of the immune system to replenish the T cell pool with naive tumor reactive lymphocytes may not be optimal.

We thank Dr. Al Singer for IL-7Rα Tg mice, Dr. Nubuo Sakaguchi for Rag1-GFP mice, and Dr. Scott Hiebert for HDAC3 floxed mice. We also thank Dr. Michael Shapiro, Dr. Fan-Chi Hsu, Puspa Thapa, and Barsha Dash for thoughtful discussions and critical reading of the manuscript.

The authors have no financial conflicts of interest.