The Interaction between NKAP and HDAC3 Is Critical for T Cell Maturation

Immunity and homeostasis depend on αβ T cells that can be broadly divided into conventional cells (CD4⁺ and CD8⁺ T cells), regulatory T cells (Tregs), and invariant NKT (iNKT) cells (1). All three subtypes develop from αβ CD4⁺CD8⁺ double positive (DP) precursors in the thymus. After positive selection, most DP thymocytes become conventional CD4 or CD8 single positive (SP) cells (1). By contrast, thymic Tregs and iNKT cells are agonist selected at the CD4 SP and DP stages, respectively, via strong TCR interactions with cognate self-ligands (1). As positive selection is insufficient for conventional T cell and Treg functional competency, additional terminal maturation steps are required (2, 3).

T cell maturation begins in the thymus and continues in the periphery as recent thymic emigrants (RTEs) transition to mature naive T cells (3). Maturation enables thymic egress and TCR/CD28 stimulation dependent proliferation and cytokine production (3). It also confers long-term survival by protection from death receptor signaling and resistance to complement proteins. In the case of Tregs, maturation facilitates the acquisition of an activated state critical for tissue-specific tolerance (4). The X-linked transcriptional regulator NKAP is indispensable for T cell maturation (5–7). In CD4-cre NKAP conditional knockout (cKO) mice, NKAP deletion at the DP stage impairs long-term persistence of peripheral T cells, although SP thymocyte production and egress are intact (5). Peripheral NKAP-deficient naive T cells are predominantly RTEs and fail to enter the long-lived naive T cell pool. NKAP-deficient RTEs exhibit reduced cytokine production and increased complement deposition compared with wild type (WT) RTEs. Consistently, expression of molecular markers associated with maturation, such as Qa2, CD45RB, and CD55, are reduced.

Similarly, although Treg-specific NKAP deletion (in Foxp3-YFP-cre NKAP cKO mice) does not impede thymic Treg development, it renders Tregs unable to persist and adopt a mature/activated state (7). Foxp3-YFP-cre NKAP cKO mice resemble Foxp3-mutant “scurfy” mice that do not generate Tregs (7, 8). Both develop systemic autoimmunity with dermatitis, lymphocytic infiltration into vital organs, unchecked T cell proliferation, B cell tolerance breakdown, and lethality by 3 wk of age (7, 9–11). Foxp3-YFP-cre NKAP cKO females carry one X^{Foxp3-YFP-cre, NKAP-fl} allele and an X^NKAP-fl allele and are healthy natural chimeras with a mix of NKAP-sufficient and NKAP-deficient Tregs due to random X-inactivation (7). Unlike NKAP-sufficient Tregs, which develop normally and persist, NKAP-deficient Tregs are rapidly eliminated at the RTE stage, revealing a cell-intrinsic survival defect in Foxp3-YFP-cre NKAP cKO female chimeras.

NKAP is a regulator of gene expression but lacks a defined DNA-binding domain and likely operates within larger molecular complexes (12). NKAP’s C-terminal domain associates with histone deacetylase 3 (HDAC3), a class-I HDAC that modifies gene expression by removing acetyl groups from histone and nonhistone proteins. Similar to NKAP-deficient RTEs, HDAC3-deficient RTEs in CD4-cre HDAC3 cKO mice have decreased persistence, impaired cytokine production, increased complement binding, and decreased CD55 expression (13). In contrast to NKAP-deficient T cells, HDAC3-deficient RTEs express normal levels of Qa2 and CD45RB, demonstrating that these markers associated with maturation may not accurately indicate functional maturity (13). Additionally, although Foxp3-YFP-cre HDAC3 cKO mice develop multiorgan autoimmunity, they survive longer than Foxp3-YFP-cre NKAP cKO mice, suggesting a less severe form of disease (7, 14). Finally, although loss of either NKAP or HDAC3 in conventional T cells and Tregs causes extrathymic maturation defects, intrathymic development of iNKT cells is severely curtailed at the DP stage in either CD4-cre NKAP cKO or CD4-cre HDAC3 cKO mice (15).

Given the phenocopy between mouse models with cKO of NKAP or HDAC3, and their known interaction, the importance of NKAP association with HDAC3 was recently examined in hematopoietic stem cells (HSCs) (16). Truncation analysis coupled with alanine scanning identified a single point mutation (Y352A) sufficient to abrogate the association of NKAP with HDAC3. A conditional deletion/re-expression mouse model was used to couple deletion of native NKAP in HSCs with induction of either YFP-tagged WT or Y352A mutant NKAP transgenes [designated YFP-NKAP(WT) or YFP-NKAP(Y352A)]. Induction of YFP-NKAP(WT) but not YFP-NKAP(Y352A) rescued the defects in HSC maintenance and survival resulting from NKAP deficiency, showing that the Y352A mutation impairs the function of NKAP in vivo. Although HSCs required NKAP association with HDAC3, proliferation of mouse embryonic fibroblasts was restored when YFP-NKAP(Y352A) was expressed (16). Therefore, the necessity for NKAP to associate with HDAC3 is context dependent. These observations prompted us to interrogate the requirement of NKAP to associate with HDAC3 in the context of T cells.

In this study, the consequence of Y352A substitution on NKAP’s function was assessed in T cells. Mice expressing either CD4-cre or Foxp3-YFP-cre drove endogenous NKAP deletion and simultaneous induction of either YFP-NKAP(WT) or YFP-NKAP(Y352A). YFP-NKAP(Y352A) expression at the DP stage coupled to endogenous NKAP deletion failed to restore iNKT cell development, phenocopying CD4-cre NKAP cKO and CD4-cre HDAC3 cKO mice. Furthermore, conventional T cells in which YFP-NKAP(Y352A) was substituted for endogenous NKAP showed decreased peripheral persistence, reduced induction of TNF-α after in vitro TCR/CD28 stimulation, and enhanced complement deposition. In addition to conventional T cells, the substitution of endogenous NKAP with YFP-NKAP(Y352A) in Tregs failed to reverse their disappearance at the RTE stage causing severe autoimmunity similar to Foxp3-YFP-cre NKAP cKO mice. As expected, substitution of YFP-NKAP(WT) reversed all effects of NKAP deficiency on iNKT development and conventional as well as Treg maturation. Recent studies indicate that NKAP deletion causes increased lipid peroxidation in naive CD4 T cells compared with WT naive CD4 T cells (17). Lipid peroxidation is a hallmark of ferroptosis, a form of programmed cell death involving iron-dependent generation of reactive oxygen species (18), and is likely the mechanism underlying NKAP-deficient T cell disappearance (17). In this study, we found that HDAC3-deficient CD4 T cells also exhibit enhanced lipid peroxidation compared with WT naive CD4 T cells, indicating that NKAP and HDAC3 work together to prevent ferroptosis. In summary, these findings show that Y352 of NKAP, which is required for HDAC3 binding, is also critical for NKAP’s roles in T cell maturation and iNKT development and long-term T cell persistence.

YFP-NKAP(WT) and YFP-NKAP(Y352A) knock-in mice were previously described (16). These were crossed to mice with a lox-neo-lox tetracycline transactivator [lnl-tTA; The Jackson Laboratory (19)] expression cassette to allow cre-mediated induction of YFP-NKAP when subsequently crossed with CD4-cre NKAP cKO mice (5), CD4-cre mice (20), Foxp3-YFP-cre NKAP cKO mice (7), and Foxp3-YFP-cre mice (21). All mouse work was performed with approval from the Mayo Clinic Institutional Animal Care and Use Committee.

Single-cell suspensions were prepared from thymii, spleens, and lymph nodes. Single-cell suspensions from spleens were treated with ACK lysis buffer (Quality Biological). Prior to surface staining, cells were treated with rat and mouse serum in a 1:1 ratio (Invitrogen) or Fc-Block (TruStain FcX, clone 93; BioLegend) to prevent nonspecific Ab binding. All experiments included fixable viability dye (Ghost Dye Red 780; Tonbo Biosciences) for exclusion of dead cells during analysis. Cell surface markers were detected with Abs conjugated with fluorescent probes from BioLegend, Tonbo Biosciences, and eBioscience (CA): CD4 (BV785, clones GK1.5 or RM4-5), CD8 (BV510, allophycocyanin, or PE, clones 53-6.7 or 2.43), CD44 (V450, clone IM7), CD62L (allophycocyanin or BV510, clone MEL-14), CD55 (PE clone RIKO-3), GITR (PE or allophycocyanin, clone DTA-1), NRP-1 (allophycocyanin, clone 3E12) and CD25 (BV421, clone PC16), TCR-β (allophycocyanin), H2-K^b (allophycocyanin, clone AF6-88.5), CD69 (PE-Cy7, clone H57-597), CCR7 (PE, clone 4B12), and CD24 (Pacific Blue, clone M1/69). Complement deposition was detected using biotinylated Ab for complement C3 (clone RmC11H9; Cedarlane) followed by PE- or allophycocyanin-conjugated streptavidin (eBioscience or Tonbo). For C3 detection, freshly harvested splenocytes were incubated in GVB⁺⁺ buffer (Complement Technology) for 1 h at room temperature, and cells were washed with FACS buffer and stained with appropriate Abs. Cytokine production was detected by overnight incubation of freshly harvested total splenocytes in 24-well plates coated with plate-bound 10 μg/ml anti-CD3 (clone 2C11; Bio X Cell) and soluble 1 μg/ml anti-CD28 (clone 37.51; Bio X Cell) Abs in RPMI media (Corning) supplemented with 10% FBS, penicillin, streptomycin, HEPES, nonessential amino-acids, l-glutamate, and 2-ME. Surface staining was performed followed by fixation and permeabilization with a cytofix/cytoperm kit (Becton Dickinson). Cytokine presence was detected using PE-conjugated anti–TNF-α (clone MP6-XT22; BioLegend). PE- or BV421-conjugated PBS57-loaded or empty CD1d tetramers were generously provided by the National Institutes of Health Tetramer Facility. Intracellular staining for Foxp3 (PE or allophycocyanin, clone 3G3; Tonbo Biosciences) was done after fixing and permeabilizing cells (Foxp3 staining kit; Tonbo Biosciences), using Abs from eBioscience and Tonbo Biosciences. Data collection was done using the Attune NxT flow cytometer (Thermo Fisher Scientific). Data were analyzed using FlowJo (Tree Star) v9.8 or v10.

Abs to dsDNA were detected using an Autoimmune EIA Anti-dsDNA Test (Bio-Rad Laboratories). Serum was diluted 100-fold in PBS with 1% BSA (PBS/BSA). The assay was completed according to the manufacturer’s instructions with the substitution of HRP-coupled goat anti-mouse IgG for the detection Ab (Southern Biotechnology Associates) diluted 2000-fold in PBS/BSA. Absorbance at 450 nM was measured using a Molecular Devices Spectromax microplate reader.

Anti-nuclear Abs (ANA) were detected by applying diluted sera (1:100) to slides coated with Kallestad HEp-2 cells (Bio-Rad Laboratories). The assay was completed according to the manufacturer’s instructions except for the substitution of AF488-coupled goat anti-mouse IgG for the detection Ab (diluted to 4 μg/ml in PBS/BSA; Invitrogen). Each well was examined at room temperature without immersion medium using a Leica DMI3000B fluorescence microscope with a 20× lens (Plan Fluotar type, numerical aperture of 0.4) and a FITC/EGFP filter cube. Images were acquired using a Q-Imaging Q1-Click camera and Q Capture Pro 6 software and saved as TIFF files. Adobe Illustrator was used to prepare the representative field shown.

Livers from WT, Foxp3-YFP-cre NKAP cKO, Foxp3 [WT→cKO], and Foxp3 [MUT→cKO] mice were isolated and kept in 10% formalin for fixation. Fixed samples were processed by paraffin embedding, sectioning and H&E stained as per standard procedures. Sections were viewed on a Leica DMI3000B microscope, at 20× magnification, and captured using the Leica EC3 camera.

Freshly harvested splenocytes were treated with 1 μM BODIPY-C11 (Thermo Fisher Scientific) in RPMI 1640 media (Corning) supplemented with 10% FBS, penicillin, and streptomycin at 37°C for 1 h. Cells were then washed and surface stained for CD4, CD44, CD62L, and Fixable Viability Dye as described above. To examine the impact of ferroptosis inhibitors on lipid peroxidation, total splenocytes were incubated with 100 μM α-tocopherol (MilliporeSigma) and/or 10 μM ferrostatin-1 (MilliporeSigma) for 15 min at 37°C, followed by staining with 1 μM BODIPY-C11 for 1 h. Cells were then washed and surface stained as described above.

All statistics were calculated using GraphPad Prism (GraphPad Software) or Microsoft Excel. Unpaired Student t test, one-way ANOVA, or two-way ANOVA were used as indicated in figure legends to examine significant differences across mice and experimental conditions. The p values < 0.05 were deemed significant and lack of significant differences (p > 0.05) is indicated in the figures.

Similar maturation defects in CD4-cre NKAP cKO and CD4-cre HDAC3 cKO T cells raise the possibility that NKAP and HDAC3 work together to regulate T cell maturation. A single-point mutation, NKAP(Y352A), abrogates NKAP-HDAC3 binding (16). Here, the impact of this mutation on NKAP’s roles in T cell maturation was examined. We used a mouse model for cre-mediated deletion of endogenous NKAP coupled with cre-mediated expression of YFP-NKAP(WT) or YFP-NKAP(Y352A) (16). Mice were engineered to express YFP-NKAP(WT) or YFP-NKAP(Y352A) upon CD4-cre–mediated endogenous NKAP gene deletion [i.e., CD4-cre NKAP cKO LNL-tTA YFP-NKAP(WT) and CD4-cre NKAP cKO LNL-tTA YFP-NKAP(Y352A) mice]. For simplicity, these will be abbreviated CD4-cre [WT→cKO] and CD4-cre [MUT→cKO], respectively. In addition, CD4-cre LNL-tTA YFP-NKAP(WT) and CD4-cre NKAP cKO LNL-tTA YFP-NKAP(Y352A) mice that express either YFP-NKAP(WT) or YFP-NKAP(Y352A) in the presence of endogenous NKAP were generated to account for potential dominant negative effects. These are abbreviated CD4-cre [WT→WT] and CD4-cre [MUT→WT], respectively.

Splenocytes from CD4-cre [WT→cKO], CD4-cre [MUT→cKO], CD4-cre [WT→WT], and CD4-cre [MUT→WT] mice were analyzed and compared with WT and CD4-cre NKAP cKO mice (Fig. 1). Appropriate T cell maturation leads to long-term persistence of conventional T cells in the periphery (3). However, CD4-cre NKAP cKO mice harbor very few peripheral CD8⁺ and CD4⁺ T cells as a result of a block in T cell maturation (5, 6). In CD4-cre [WT→cKO], the frequencies and absolute cell counts of peripheral CD8⁺ and CD4⁺ T cells were restored to WT levels, indicating normal T cell maturation (Fig. 1A, 1C, Supplemental Fig. 1). In contrast to WT mice, CD4-cre [MUT→cKO] mice exhibited reductions in CD8⁺ and CD4⁺ peripheral T cell frequencies, and absolute cell counts and were very similar to CD4-cre NKAP cKO mice, indicating impaired long-term persistence. In particular, there were few naive (CD62L⁺CD44^lo) CD8⁺ and CD4⁺ T cells in both CD4-cre [MUT→cKO] and CD4-cre NKAP cKO mice (Fig. 1A, 1C). Thus, NKAP(Y352A) could not substitute for endogenous NKAP to support T cell maturation, leading to the disappearance of peripheral CD4 and CD8 T cells.

To determine whether the expression of YFP-NKAP(Y352A) exerted a dominant negative effect on peripheral T cell persistence, splenocytes from CD4-cre [WT→WT] and CD4-cre [MUT→WT] were analyzed. Absolute numbers of total CD8⁺ T cells and naive CD8⁺ T cells in CD4-cre [WT→WT] and CD4-cre [MUT→WT] were similar to WT mice (Fig. 1A). However, there was an ∼2-fold reduction in total and naive CD4 T cells in CD4-cre [WT→WT] and CD4-cre [MUT→WT] compared with WT (Fig. 1B). Therefore, YFP-NKAP(WT) or YFP-NKAP(Y352A) had no adverse effects on CD8⁺ frequencies and numbers and had only a partial effect on CD4⁺ T cell numbers. However, as CD4⁺ and CD8⁺ T cell total and naive numbers were very similar between CD4-cre [WT→WT] and CD4-cre [MUT→WT] mice, the 2-fold decrease observed in case of CD4⁺ T cells is likely an outcome of NKAP overexpression rather than mutation at Y352.

The failure of YFP-NKAP(Y352) to rescue peripheral T cell numbers could be a result of a reduction in its expression as compared with YFP-NKAP(WT). Next, YFP expression in CD8⁺ and CD4⁺ naive and memory T cells was examined. YFP was expressed at similar levels between CD8⁺ naive (CD62L⁺CD44^lo) and CD8⁺ memory (CD44⁺) T cells (Fig. 1B) and between CD4⁺ naive (CD62L⁺CD44^lo) and CD4⁺ memory (CD44⁺ CD62L⁻) T cells (Fig. 1D) from CD4-cre [WT→cKO], CD4-cre [WT→WT], and CD4-cre [MUT→WT] mice. By contrast, no YFP expression was observed in memory CD4⁺ or memory CD8⁺ T cells from CD4-cre [MUT→cKO]. Previously, memory T cells from CD4-cre NKAP cKO mice were shown to be NKAP gene nondeleters that underwent homeostatic expansion owing to a sparse T cell niche (5). YFP expression in CD4-cre [MUT→cKO] depends on cre activity, and the lack of YFP expression in CD4⁺ and CD8⁺ memory cells is consistent with expansion of NKAP nondeleting cells that upregulate CD44 and downregulate CD62L. YFP expression in CD4⁺ naive T cells from CD4-cre [WT→cKO], CD4-cre [MUT→cKO], CD4-cre [WT→WT], and CD4-cre [MUT→WT] mice was comparable. Interestingly, there was bimodal expression of YFP-NKAP(Y352A) in naive CD8⁺ cells with a YFP⁺ population and a YFP⁻ population, indicating that a fraction of naive T cells escaped cre-mediated NKAP deletion. Examination of only the YFP⁺CD8⁺ naive fraction shows that the average geometric mean fluorescence intensity (gMFI) of YFP is not reduced in CD4-cre [MUT→cKO] compared with CD4-cre [WT→cKO]. In summary, the failure of YFP-NKAP(Y352A) to substitute for endogenous NKAP is not due to decreased expression as compared with YFP-NKAP(WT).

WT and CD4-cre NKAP cKO mice do not exhibit significant differences in the production of DP and SP thymocytes, and the decrease in peripheral T cells in CD4-cre NKAP cKO mice is due to defects in long-term persistence (5, 6). Splenic T cell profiles of CD4-cre NKAP cKO and CD4-cre [MUT→cKO] mice were virtually indistinguishable, suggesting that YFP-NKAP(Y352A) expression failed to reverse defects in T cell persistence stemming from NKAP deficiency. However, another possibility was YFP-NKAP(Y352A) expression caused a reduction in thymic output. To examine this, frequency and absolute counts of CD4⁺ SP and CD8⁺ SP thymocytes were compared between WT, CD4-cre NKAP cKO, CD4-cre [WT→cKO], CD4-cre [MUT→cKO], CD4-cre [WT→WT], and CD4-cre [MUT→WT] mice (Fig. 2). In agreement with previous data, there was no significant difference in the numbers of CD4⁺ SP thymocytes in WT and CD4-cre NKAP cKO mice and only a small difference in CD8⁺ SP thymocytes (Fig. 2A). CD4-cre [WT→cKO] mice produced normal numbers of SP thymocytes (Fig. 2A). In contrast, CD4-cre [MUT→cKO], CD4-cre [WT→WT], and CD4-cre [MUT→WT] mice had reduced CD4⁺ SP and CD8⁺ SP thymocyte numbers compared with WT. Importantly, SP thymocyte generation in CD4-cre [MUT→cKO] and CD4-cre [MUT→WT] mice were equivalent, although there was decreased peripheral T cell numbers in CD4-cre [MUT→cKO] mice compared with CD4-cre [MUT→WT] mice. Thus, peripheral T cell deficiency in CD4-cre [MUT→cKO] mice is not simply due to decreased thymic output but represents an inability of peripheral T cells to mature and accumulate in the long-term T cell pool.

Surprisingly, CD4-cre [MUT→cKO], CD4-cre [WT→WT], and CD4-cre [MUT→WT] mice showed greater expression of YFP compared with CD4-cre [WT→cKO] (Fig. 2B). This was unexpected, as the mice were identically engineered to express YFP-NKAP(WT) or YFP-NKAP(Y352A) (16). However, the reduction in SP thymocytes in CD4-cre [MUT→cKO], CD4-cre [WT→WT], and CD4-cre [MUT→WT] mice followed enhanced YFP expression, suggesting that overexpression of YFP-NKAP in the thymus in these lines reduced SP generation compared with CD4-cre [WT→cKO] mice. To further examine whether NKAP overexpression reduces the size of SP population in the thymus, YFP-high and YFP-low CD4-cre [WT→WT] as well as CD4-cre [MUT→WT] thymocytes were compared with WT thymocytes expressing native NKAP only to assess the impact on SP thymocyte development and maturation (Supplemental Fig. 2). YFP-high and YFP-low CD4 and CD8 SP thymocytes were generated at similar levels in CD4-cre [WT→WT] and CD4-cre [MUT→WT] (Supplemental Fig. 2A). Maturation is a gradual process accompanied by an increase in the expression of the cell surface protein Qa2 and a decrease in the cell surface protein CD24. Using Qa2 and CD24, maturation was examined to assess whether NKAP transgene expression level may alter maturation. In both CD4-cre [WT→WT] and CD4-cre [MUT→WT] thymocytes, YFP-low CD4⁺Qa2⁺CD24⁻ mature cells were generated at frequencies equivalent to WT, whereas YFP-high CD4⁺Qa2⁺CD24⁻ mature cells were not produced. Thus, high YFP-NKAP expression was detrimental to the generation of mature CD4⁺Qa2⁺CD24⁻ SP thymocytes and the Y352A mutation did not alter this effect. Similarly, the generation of YFP-high CD8⁺Qa2⁺CD24⁻ SP thymocytes was impeded, whereas YFP-low CD8⁺Qa2⁺CD24⁻ cells were generated at levels comparable to WT (data not shown). The process of intrathymic maturation can be demarcated using additional maturation markers (3, 22). SP thymocytes can be divided into CCR7⁻ CD24⁺ or H2-K^b− CD69⁺semimature (SM), CCR7⁺ CD24⁺ or H2-K^b+ CD69⁺ mature 1 (M1), and CCR7⁺ CD24⁻ or H2-K^b+ CD69⁻ mature 2 (M2) thymocytes (3, 22). CD4 and CD8 SM, M1, M2 populations were formed in WT, CD4-cre NKAP cKO, CD4-cre [WT→cKO], and CD4-cre [MUT→cKO] mice (Supplemental Fig. 2B, 2C, 2E, 2F). The examination of YFP expression in CD4 and CD8 SM, M1, M2 populations revealed that M1 and M2 SP thymocytes had lower YFP expression compared with SM SP thymocytes (Supplemental Fig. 2D, 2G). This finding is consistent with the previous observation that high NKAP transgene expression is detrimental for mature SP generation and likely creates a bias for YFP-low cells to enter the periphery. Furthermore, YFP was expressed by all CD8 SP thymocytes, indicating that bimodal expression of YFP starts in the periphery.

Although CD4-cre NKAP cKO and CD4-cre HDAC3 cKO mice do not present blocks in conventional T cell development, they suffer a severe block in iNKT development at the DP stage, leading to the development of very few thymic or peripheral iNKT cells (15). Unlike conventional T cells that have a diverse repertoire of TCR α-chains, iNKT cells express a fixed TCR α-chain and can be identified by CD1d tetramer loaded with the glycolipid PBS-57 (1, 15). iNKT cell development was examined in the thymus in WT, CD4-cre NKAP cKO, CD4-cre [WT→cKO], and CD4-cre [MUT→cKO] mice using CD1d tetramer loaded with the glycolipid PBS-57 (Fig. 2C). Similar to CD4-cre NKAP cKO mice and CD4-cre HDAC3 cKO mice, very few iNKT cells were detected in the thymus of CD4-cre [MUT→cKO] mice, demonstrating the importance of Y352 of NKAP in iNKT development.

During T cell maturation, RTEs gain the ability to produce cytokines (such as TNF-α) upon TCR/CD28 stimulation and acquire complement resistance (3). As the Y352A mutation impairs NKAP’s function in long-term T cell persistence, its effect on TNF-α production was examined next. Total splenocytes from WT, CD4-cre NKAP cKO, CD4-cre [WT→WT], CD4-cre [MUT→cKO], CD4-cre [WT→WT], and CD4-cre [MUT→WT] mice were stimulated by TCR/CD28 activation (Fig. 3A, 3B). A smaller fraction of CD4-cre NKAP cKO CD4⁺CD44⁻ T cells produced TNF-α compared with WT T cells (Fig. 3A). Compared with CD4⁺CD44⁻YFP⁺ cells from CD4-cre [WT→cKO], fewer CD4⁺CD44⁻YFP⁺ cells from CD4-cre [MUT→cKO] expressed TNF-α, indicating a defect in functional T cell maturation (Fig. 3B). The frequency of TNF production was similar in CD4⁺CD44⁻YFP⁺ cells from CD4-cre [WT→WT], CD4-cre [MUT→WT], and CD4-cre [WT→cKO] mice, ruling out the possibility of dominant negative effects of either YFP-NKAP(WT) or YFP-NKAP(Y352A) overexpression on TNF-α production in peripheral T cells (Fig. 3B).

Thymic maturation leads to gradual acquisition of complement resistance (3, 17). NKAP-deficient and HDAC3-deficient peripheral T cells exhibit elevated susceptibility to complement deposition (6, 13). To assess whether substitution of endogenous NKAP with NKAP(Y352A) could restore complement resistance, C3 binding to total CD4 naive T cells from WT, CD4-cre NKAP cKO, CD4-cre [WT→cKO], CD4-cre [MUT→cKO], CD4-cre [WT→WT], and CD4-cre [MUT→WT] was examined (Fig. 3C). C3 binding was enhanced specifically in naive CD4⁺ T cells from CD4-cre [MUT→cKO], similar to that observed in naive CD4⁺ T cells from CD4-cre NKAP cKO mice. Therefore, the Y352A mutant of NKAP cannot substitute for WT NKAP in protecting peripheral T cells from complement deposition.

Maturation is accompanied by increases in expression of cell surface proteins Qa2, CD45RB and CD55 (3, 5, 6, 13). Whether the NKAP(Y352A) mutation altered the ability of naive CD4 T cells to express these maturation markers was examined (Fig. 3D, 3E). As previously shown, CD4⁺ naive T cells from CD4-cre NKAP cKO mice expressed significantly lower levels of Qa2, CD55, and CD45RB as compared with WT (5, 6). In contrast, CD55 and CD45RB were expressed at normal levels in CD4-cre [WT→cKO], CD4-cre [MUT→cKO], CD4-cre [WT→WT], and CD4-cre [MUT→WT] cells. Qa2 was expressed at normal levels in CD4-cre [WT→cKO], CD4-cre [WT→WT] and CD4-cre [MUT→WT] but was only slightly reduced in CD4-cre [MUT→cKO] cells, although this difference was not statistically significant. This is not surprising as HDAC3 deficiency does not lead to decreased Qa2 and CD45RB expression and causes a slight downregulation of CD55 (13). Thus, abrogation of HDAC3 association with NKAP by the Y352A mutation caused a phenotype resembling that of CD4-cre HDAC3 cKO mice, with defects in cytokine production and protection from complement but not reduction in Qa2, CD45RB, and CD55.

Treg-specific NKAP deficiency (Foxp3-YFP-cre NKAP cKO) results in systemic autoimmunity and lethality by 3 wk of age, similar to that observed in scurfy/Foxp3 mutant mice (7). Treg-specific HDAC3 deletion (Foxp3-YFP-cre HDAC3 cKO) also results in autoimmunity, albeit less severe than Treg-specific NKAP deletion, as Foxp3-YFP-cre HDAC3 cKO mice survive slightly longer (7, 14). Both exhibit extensive dermatitis, lymphocytic infiltration into various organs, T cell activation, and proliferation as well as autoantibodies in sera, indicative of Treg-mediated tolerance breakdown. To assess the impact of the NKAP(Y352A) mutation in Treg-mediated tolerance, a Treg-specific conditional deletion/re-expression system was used. Foxp3-YFP-cre NKAP cKO LNL-tTA YFP-NKAP(WT) mice express YFP-NKAP(WT) in Foxp3-YFP-cre NKAP cKO cells, and Foxp3-YFP-cre NKAP cKO LNL-tTA YFP-NKAP(Y352A) mice express YFP-NKAP(Y352A) in Foxp3-YFP-cre NKAP cKO cells. These are abbreviated Foxp3-YFP-cre [WT→cKO] and Foxp3-YFP-cre [MUT→cKO], respectively. To account for possible dominant negative effects, Foxp3-YFP-cre LNL-tTA YFP-NKAP(WT) and Foxp3-YFP-cre LNL-tTA YFP-NKAP(Y352A) that express YFP-NKAP-WT and YFP-NKAP(Y352A) in the presence of endogenous NKAP respectively were used. These are abbreviated as Foxp3-YFP-cre [WT→WT] and Foxp3-YFP-cre [MUT→WT], respectively.

Substitution of endogenous NKAP with YFP-NKAP(WT) in Foxp3-YFP-cre [WT→cKO] mice prevented autoimmunity (Fig. 4A). However, substitution of endogenous NKAP with YFP-NKAP(Y352A) in Foxp3-YFP-cre [MUT→cKO] mice resulted in runted appearance, dermatitis, reduced weight gain, and lethality by 3 wk of age, similar to Foxp3-YFP-cre NKAP cKO and scurfy mice. Total splenic absolute cell numbers were comparable between mice; however, lymph node absolute cell numbers of in Foxp3-YFP-cre NKAP cKO and Foxp3-YFP-cre [MUT→cKO] mice were increased compared with WT or Foxp3-YFP-cre [WT→cKO] mice (Fig. 4B). The thymus atrophies as DP thymocytes undergo increased apoptosis due to rampant inflammation (Fig. 4B) (23). Examination of thymii from 3-wk-old mice showed severe reductions in thymic cellularity in Foxp3-YFP-cre NKAP cKO and Foxp3-YFP-cre [MUT→cKO] mice compared with WT or Foxp3-YFP-cre [WT→cKO] mice. Autoantibodies to nuclear Ags and dsDNA were detectable in the sera of all Foxp3-YFP-cre [MUT→cKO] mice similar to all Foxp3-YFP-cre NKAP cKO mice, indicating a breach in B cell tolerance (Fig. 4C, 4D). Further, histological examination indicated lymphocytic infiltration into Foxp3-YFP-cre [MUT→cKO] livers similar to Foxp3-YFP-cre NKAP cKO livers, whereas Foxp3-YFP-cre [WT→cKO] livers were similar to WT, indicating peripheral tolerance breakdown (Fig. 4E).

Unchecked T cell activation and proliferation is the driving cause of tolerance breakdown (10). In comparison with WT, Foxp3-YFP-cre NKAP cKO and Foxp3-YFP-cre [MUT→cKO] mice exhibited slightly increased CD4 and significantly increased CD8 absolute T cell counts (Supplemental Fig. 3A). In addition, memory/activated CD4⁺CD44⁺CD62L⁻ and CD8⁺CD44⁺CD62L⁻ absolute cell numbers were increased in Foxp3-YFP-cre NKAP cKO and Foxp3-YFP-cre [MUT→cKO] mice (Supplemental Fig. 3B). To exclude the possibility that NKAP(Y352A) exerted dominant negative effects leading to tolerance breakdown, Foxp3-YFP-cre [WT→WT] and Foxp3-YFP-cre [MUT→WT] mice were examined. No aberrant T cell expansion was found, and mice survived to adulthood without any signs of autoimmunity (Supplemental Fig. 3C–E). Therefore, similar to Foxp3-YFP-cre NKAP cKO mice, Foxp3-YFP-cre [MUT→cKO] mice succumb to systemic autoimmunity characterized by B cell and T cell tolerance breakdown.

Treg development is indispensable for ensuring tolerance in humans and mice (24–26). Mature thymic CD25⁺Foxp3⁺ Tregs develop from CD25⁺Foxp3⁻ and CD25⁻Foxp3⁺ precursor populations (27). As rampant autoimmunity leads to thymic involution, 3-wk-old male Foxp3-YFP-cre NKAP cKO mice exhibit diminished overall thymic cellularity and Treg absolute numbers compared with WT mice (7). However, thymic Treg and Treg precursor frequencies in 3-wk-old Foxp3-YFP-cre NKAP cKO mice are similar to WT, indicating normal thymic Treg production. Treg development was examined in male WT, Foxp3-YFP-cre NKAP cKO, Foxp3-YFP-cre [WT→cKO], Foxp3-YFP-cre [MUT→cKO], Foxp3-YFP-cre [WT→WT], and Foxp3-YFP-cre [MUT→WT] mice (Fig. 5A). All mice had similar frequencies of CD25⁺Foxp3⁻ and CD25⁻Foxp3⁺ precursors and CD25⁺Foxp3⁺ Tregs, indicating that thymic Treg production occurs unhindered in Foxp3-YFP-cre [MUT→cKO] mice, and the expression of YFP-NKAP(Y352A) in the presence of endogenous NKAP has no adverse effects on the generation of Tregs.

The underlying cause of autoimmunity in Foxp3-YFP-cre NKAP cKO male mice is disappearance of peripheral Tregs (7). Foxp3 absolute cell numbers and frequencies were assessed in male WT, Foxp3-YFP-cre NKAP cKO, Foxp3-YFP-cre [WT→cKO], and Foxp3-YFP-cre [MUT→cKO] mice. Similar to Foxp3-YFP-cre NKAP cKO mice, Foxp3-YFP-cre [MUT→cKO] mice exhibited a severe decrease in absolute Treg counts in the spleen (Fig. 5B). Examination of Foxp3-YFP-cre [MUT→WT] mice showed no defects in Treg persistence, ruling out dominant negative effects of YFP-NKAP(Y352A) expression. Previously, to bypass any indirect effects of inflammation, Foxp3-YFP-cre NKAP cKO female chimeras were compared with Foxp-3 YFP-cre WT female chimeras (7). Foxp3-YFP-cre WT chimeric females have X^{Foxp3-YFP-cre} and a X^NKAP-fl alleles. Foxp3-YFP-cre NKAP cKO chimeric females have X^{Foxp3-YFP-cre NKAP-fl} and a X^NKAP-fl alleles. Because of random X-chromosome inactivation, both mice generate YFP⁺ (cre⁺) and YFP⁻ (cre⁻) Tregs. However, YFP⁺ Tregs fail to persist in Foxp3-YFP-cre NKAP cKO female chimeras, whereas similar fractions of YFP⁺ and YFP⁻ Tregs are present in Foxp3-YFP-cre WT females (Supplemental Fig. 4A–C). Foxp3-YFP-cre [MUT→cKO] female chimeras had very few YFP⁺ Tregs in the periphery and showed a skewing toward YFP⁻ WT population. Thus, Foxp3-YFP-cre [MUT→cKO] female chimeras are similar to Foxp3-YFP-cre NKAP cKO female chimeras. In contrast, Foxp3-YFP-cre [WT→cKO] female chimeras have YFP⁺ Tregs at frequencies similar to Foxp3-YFP-cre female WT chimeras. In the presence of YFP⁻ WT Tregs, YFP⁺ Foxp3-YFP-cre [WT→WT] Tregs or YFP⁺ Foxp3-YFP-cre [MUT→WT] Tregs did not exhibit any competitive survival disadvantages, demonstrating no adverse effects of YFP-NKAP(Y352A) in the presence of endogenous NKAP on Treg survival. There is considerable overlap between the YFP signal from Foxp3-YFP-cre (cre is fused to YFP) and YFP-NKAP precluding separation of these signals (Supplemental Fig. 4D, 4E). However, a slight increase in YFP gMFI occurs when both Foxp3-YFP-cre and YFP-NKAP transgene are expressed compared with expression of Foxp3-YFP-cre alone. This increase in YFP signal was used to confirm the expression of YFP-NKAP-(WT) and YFP-NKAP-(Y352A) transgenes in chimeric females (Supplemental Fig. 4D, 4E).

Previously, NKAP-deficient YFP⁺ Tregs in Foxp3-YFP-cre NKAP cKO female chimeras were shown to be Nrp-1 low compared with NKAP-sufficient YFP⁺ Tregs from Foxp3-YFP-cre WT female chimeras (7). Nrp-1 low Tregs may represent peripherally derived Tregs or thymically derived RTE Tregs (28). Using RagGFP mice, RTE Tregs were shown to be CD44^lo and CD62L^hi (29). To distinguish between peripherally derived Tregs and RTE Tregs, the expression of CD44, CD62L, and Qa2 by NKAP-deficient YFP⁺ Tregs in Foxp3-YFP-cre NKAP cKO and YFP⁺ Tregs from Foxp3-YFP-cre WT female chimeras was previously examined (7). NKAP-deficient Tregs were found to be Nrp-1^lo CD44^lo Qa2^lo whereas NKAP-sufficient Tregs from Foxp3-YFP-cre WT chimeric females were Nrp-1^hi CD44^hi Qa2^hi (7). Furthermore, NKAP-deficient Tregs had a greater abundance of TCR excision circles, which negatively correlates with the age of T cells (30). These results indicated that NKAP-deficient Tregs disappeared at the RTE stage. As Foxp3-YFP-cre [MUT→cKO] male mice develop autoimmunity at a rate similar to Foxp3-YFP-cre NKAP cKO mice, Foxp3-YFP-cre [MUT→cKO] Tregs likely disappear as RTEs as well. To determine whether abrogation of NKAP’s interaction with HDAC3 leads to Treg disappearance at the RTE stage, expression of Qa2, CD44, and Nrp-1 by Tregs from WT, Foxp3-YFP-cre NKAP cKO, Foxp3-YFP-cre [WT→cKO], and Foxp3-YFP-cre [MUT→cKO] male mice was determined (Fig. 5C). Tregs from 3-wk-old male Foxp3-YFP-cre [MUT→cKO] mice are also Qa2^loCD44^loNrp-1^lo compared with WT and Foxp3-YFP-cre [WT→cKO], indicating that they fail to transition to the more activated and mature state and are eliminated at the RTE stage. To confirm the RTE phenotype of NKAP-deficient Tregs in a nonautoimmune environment, YFP⁺ and YFP⁻ Tregs from Foxp3-YFP-cre, Foxp3-YFP-cre NKAP cKO, Foxp3-YFP-cre [WT→cKO], and Foxp3-YFP-cre [MUT→cKO] female chimeras were examined for expression of CD44 and CD62L (Supplemental Fig. 4F, 4G). Foxp3-YFP-cre NKAP cKO and Foxp3-YFP-cre [MUT→cKO] females harbored very few activated YFP⁺ CD44⁺ Tregs compared with YFP⁻ CD44⁺ Tregs. In contrast, Foxp3-YFP-cre WT and Foxp3-YFP-cre [WT→cKO] females generated YFP⁺ CD44⁺ and YFP⁻ CD44⁺ Tregs. Next, the expression of Qa2 by YFP⁺CD62L⁺ Tregs and YFP⁻CD62L⁺ Tregs from Foxp3-YFP-cre, Foxp3-YFP-cre NKAP cKO, Foxp3-YFP-cre [WT→cKO], and Foxp3-YFP-cre [MUT→cKO] female chimeras was examined (Supplemental Fig. 4H). Qa2 was expressed at low and intermediate levels by YFP⁺CD62L⁺ Tregs from Foxp3-YFP-cre NKAP cKO and Foxp3-YFP-cre [MUT→cKO] chimeric females, respectively, compared with YFP⁻CD62L⁺ Tregs in the same mice. Qa2 expression by YFP⁻CD62L⁺ Tregs and YFP⁺CD62L⁺ Tregs from Foxp3-YFP-cre WT and Foxp3-YFP-cre [WT→cKO] females were comparable. Cumulatively, these results indicate that abrogation of NKAP’s interaction with HDAC3 leads to rapid loss of Tregs at the RTE stage, replicating the phenotype of NKAP deficiency.

Maturing thymocytes gradually gain complement resistance prior to egress from the thymus (17). NKAP deletion results in the inability of peripheral naive T cells to attain protection from complement proteins (6). NKAP-deficient T cells are rapidly opsonized by C3 as well as other complement molecules following thymic egress, implicating complement activation as a possible mechanism driving their disappearance in secondary lymphoid organs (6). In a recent study (17), CD4-cre NKAP cKO mice were crossed with mice also deficient in C1q, MBL1, and MBL2 encoding genes to examine whether combined deletion of the classical (C1q) and lectin pathways (MBL1 and MBL2) could restore NKAP-deficient T cells to normal numbers. Although C3 deposition was prevented in these mice, NKAP-deficient T cells failed to persist in the absence the classical and lectin pathways, indicating that another mechanism mediated T cell clearance. The classical and lectin pathways mediate C3 deposition (31). C3 deficiency did not curb the loss of peripheral T cells in CD4-cre NKAP cKO mice either (17). In addition, there was no evidence of mitochondrial dysfunction or apoptosis in NKAP-deficient T cells (6, 17).

Interestingly, CD4-cre NKAP cKO mice exhibited enhanced lipid peroxidation compared with WT mice (17). Accumulation of lipid peroxides is a defining feature of ferroptosis, an iron-dependent form of programmed cell death. Increased lipid peroxidation in NKAP-deficient naive T cells was detected using BODIPY-C11 581/591, a lipophilic reactive oxygen species sensor that gains fluorescence in the FITC channel following reaction with peroxyl radicals (17, 18, 32, 33). Lipid peroxidation stemming from NKAP deficiency could be inhibited in vitro by treatment of cells with the lipophilic antioxidant α-tocopherol (vitamin E) and/or an inhibitor of ferroptosis, ferrostatin-1, bolstering the idea that NKAP deficiency leads to ferroptosis (17).

To determine whether loss of peripheral T cells in CD4-cre HDAC3 cKO mice occurred by a mechanism similar to CD4-cre NKAP cKO mice, the effect of C3 deficiency was examined (Fig. 6A, 6B). The loss of HDAC3-deficient T cells was not mitigated in C3 KO CD4-cre HDAC3 cKO mice. Next, Ferroptosis was examined using BODIPY-C11 581/591 (Fig. 6C). HDAC3-deficient naive T cells exhibited enhanced lipid peroxidation in CD4-cre HDAC3 cKO as well as C3 KO CD4-cre HDAC3 cKO mice. In vitro, lipid peroxidation of HDAC3-deficient T cells was inhibited by α-tocopherol or ferrostatin-1 or by a combination of both similar to NKAP-deficient T cells (Fig. 6D). Furthermore, compared with WT, C3 KO CD4-cre HDAC3 cKO mice also exhibit lipid peroxidation that can be reversed by α-tocopherol and ferrostatin-1 in combination (Fig. 6E) or individually (data not shown). Overall, our data indicate that susceptibility to complement attack and ferroptosis are regulated by NKAP and HDAC3 in a similar manner, consistent with the evidence that their interaction is important for T cell maturation.

Maturation endows T cells with functional capacity and is an important determinant of immunity and immune system homeostasis. A greater understanding of maturation mechanisms may be harnessed to improve T cell dependent immune responses. Although regulators of early T cell development are very well characterized, only a few regulators of maturation have been identified (2, 3, 5, 13, 22, 34–40). NKAP and HDAC3 have been separately shown to be required for T cell maturation and iNKT development (5, 13, 15). Neither NKAP nor HDAC3 has a defined DNA-binding domain, and they likely associate with chromatin as parts of larger protein complexes. Interestingly, NKAP and HDAC3 physically associate (12), and this interaction can be disrupted by a single amino acid substitution in the NKAP(Y352A) mutant (16). In this study, we report that this mutation in NKAP also impairs its function during T cell maturation and iNKT cell development, providing evidence that the association between NKAP and HDAC3 is critical for T cell maturation and iNKT cell development.

Upon CD4-cre–driven substitution of endogenous NKAP with NKAP(Y352A), a block occurred in intrathymic iNKT cell development, whereas no defects in thymic T cell or Treg development were observed. This phenotype was essentially identical to that presented by CD4-cre NKAP cKO mice. Similarly, CD4-cre or Foxp3-YFP-cre–driven substitution of NKAP(Y352A) for WT NKAP resulted in defects in T cell and Treg maturation similar to those observed in CD4-cre or Foxp3-YFP-cre NKAP cKO mice, respectively. Specifically, in conventional T cells, these defects included a failure in long-term persistence, impaired cytokine production, and a lack of resistance to complement. In Tregs, their disappearance at the RTE stage caused systemic lethal autoimmunity. In control mice, induction of an NKAP(WT) transgene prevented all effects of deletion of the native NKAP gene in each T cell population. Together, these results show that NKAP(Y352A) cannot substitute for WT NKAP during iNKT development in the thymus or during maturation of conventional T cells or Tregs. These results suggest that the ability to associate with HDAC3 is critical for the function of NKAP.

Interestingly, NKAP(Y352A) was able to functionally substitute for WT NKAP during conventional T cell maturation in one respect. Significantly reduced expression of Qa2 and CD45RB, two phenotypic markers associated with maturation, occurs in CD4-cre NKAP cKO RTEs, yet expression of these proteins was restored to nearly normal levels in cells expressing NKAP(Y352A) instead of WT NKAP. Thus, NKAP must influence the expression of these proteins independent of HDAC3. This idea is further supported by the observation that Qa2 and CD45RB expression were not reduced in HDAC3-deficient RTEs despite a general defect in T cell maturation (13). The participation of NKAP in both HDAC3-dependent and independent processes was also demonstrated in a previous study in which NKAP was deleted in multiple hematopoietic population with Mx1-cre along with induction of NKAP(Y352A) (16). Although the mutant was unable to substitute for WT NKAP in the maintenance of the HSC pool, it was able to completely substitute in the erythroid lineage. Furthermore, the Y352A mutation did not cause any reduction in the ability of NKAP to promote proliferation in fibroblasts (16). These results also suggest that NKAP may be present in molecular complexes whose individual components vary depending on the biological context, which will be investigated in the future.

Similar to NKAP, HDAC3 is also critical to prevent lipid peroxidation in naive T cells, and ferroptosis is the likely the cause of disappearance of both NKAP- and HDAC3-deficient T cells. Abrogating the association of NKAP and HDAC3 would likely result in ferroptosis as well. Unfortunately, this cannot be tested in CD4-cre [MUT→cKO] mice because of overlap of BODIPY-C11 581/591 and YFP fluorescence. In vivo, ferroptosis results from complex interactions between several factors including iron accumulation, increased synthesis of acid-phospholipids, depletion of enzymes that reverse lipid peroxidation, and unavailability of antioxidant molecules (18). In T cells, two well characterized inhibitors of ferroptosis are GPX-4 and glutathione (g-l-glutamyl-l-cysteinylglycine or GSH); GPX-4 uses GSH to reduce lipid hydroperoxides to less harmful lipid alcohols, and this process is critical for T cell persistence (41). Both GPX-4 and GSH were previously shown to be present at normal levels in NKAP-deficient CD4 naive T cells, suggesting a GPX-4–independent mechanism of lipid peroxidation (17).

Although the effects of the Y352A substitution on association of NKAP with partners other than HDAC3 cannot be excluded, the findings presented in this article, combined with previous observations of similar requirements for NKAP and HDAC3 during T cell development, strongly motivate future investigations on interplay between the two proteins. For example, the functions of both NKAP and HDAC3 in T cell development earlier than the DP stage have been investigated using cre-mediated deletion at the DN stage (12, 42, 43). Deficiency in either protein resulted in different blocks in T cell development. Whereas the deletion of HDAC3 caused a block at the DP stage, the deletion of NKAP caused a DN3 stage block. Thus, the tools presented in this study hold great potential to further dissect functions of NKAP and HDAC3 in early T cell development and other aspects of T cell biology to identify functions that require their association and functions that do not.

We thank the members of laboratory of V.S.S. for critical reading of the manuscript and Paul Belmonte for technical assistance with lipid peroxidation studies. We thank members of Kay Medina and Hu Zeng laboratories for helpful discussions. We also thank the National Institutes of Health tetramer facility and the Mayo Knockout and Transgenic Core Facility.

The authors have no financial conflicts of interest.