SIKs Regulate HDAC7 Stabilization and Cytokine Recall in Late-Stage T Cell Effector Differentiation

This article is featured in Top Reads, p. 1749

Elucidation of pathways regulating T cell differentiation is important to the advancement of our mechanistic understanding of these processes and their function in various immunological settings, including cancer, autoimmunity, infectious disease, and vaccine development. T cell activation by Ag recognition initiates epigenetic reprogramming to promote transcriptional expression of genes associated with T cell differentiation and the acquisition of effector function (1–4). At the core of these induced genes are hallmark lineage-specifying transcription factors and cytokines that define each effector subset (5–7). Upon differentiation, effector T cells assist in the clearance of the immunological challenge as Ag receptor activation induces T cell degranulation, resulting in the directed release of effector molecules, including lineage-specific cytokines. Following resolution of the immunological insult, effector T cells contract, and long-lived memory is established (8, 9). The dynamic nature of the initial establishment and subsequent restriction of effector functionality highlights temporal differences in regulating T cell differentiation. Accordingly, the need to understand both early and late-stage T cell differentiation processes is of essential importance.

The primary pathways associated with these varied processes have been defined over many years. However, significant interest remains in identifying additional pathways that regulate T cell differentiation to further refine our understanding and ability to modulate T cell functionality. In this study, we report the identification of salt-inducible kinases (SIKs) as previously unappreciated regulators of effector cytokine production in effector CD8⁺ and Th1 cells. We demonstrate that effector T cell SIK activity leads to phosphorylation-dependent stabilization of the class IIa histone deacetylase (HDAC), HDAC7, during the latter stages of differentiation. This enhanced stability of HDAC7 correlates with increased nuclear HDAC7 levels, reduced total and cytokine loci specific histone 3 lysine 27 acetylation (H3K27Ac) levels, and restricted cytokine production. Moreover, inhibition of this pathway, using the kinase inhibitor BioE-1197, durably programs enhanced cytokine functionality by effector T cells.

All procedures were approved by the Johns Hopkins University Institutional Animal Care and Use Committee under protocol numbers M019M71 and M022M271. OT-I (RRID:IMSR_JAX:003831), OT-II (RRID:IMSR_JAX:004194), CD90.1 (RRID:IMSR_JAX:000406), CD45.1 (RRID:IMSR_JAX:002014), and C57BL/6J (RRID:IMSR_JAX:000664) mice were originally acquired from The Jackson Laboratory. P14 TCR-transgenic mice were provided by Dr. David A. Hildeman (Cincinnati Children’s Hospital Medical Center). PAS domain containing serine/threonine kinase (PASK) knockout (KO) mice were a gift from Dr. Jared Rutter (University of Utah School of Medicine) (10, 11). Mice were housed in a specific pathogen-free facility. Six- to 20-wk-old male and female WT C57BL/6J, OT-I, OT-II, P14, and PASK KO mice were used as sources for primary murine CD8⁺ and CD4⁺ T cells for in vitro culture. Six- to 8-wk-old male C57BL/6J mice were used as hosts for adoptive transfer experiments. Males were specifically used based on sex matching to the coadoptively transferred congenic T cells.

BioE-1197 was synthesized using the previously reported method (12). TMP269 (catalog no. HY-18360) and YKL-05-099 (catalog no. HY-101147) were purchased from MedChemExpress. HG-9-91-01 was purchased from Selleck Chemicals (catalog no. S8393). Drugs were solubilized in DMSO (Thermo Fisher Scientific, catalog no. D128-500).

Primary murine T cells were isolated from the spleen and inguinal, axillary, and brachial lymph nodes of mice. Following the generation of a single-cell suspension by dissociation through a 70-µm filter (Fisherbrand, Thermo Fisher Scientific, catalog no. 22-363-548), T cells were enriched using negative isolation. CD8⁺ T cells were isolated using MojoSort mouse CD8 T cell isolation kits (BioLegend, catalog no. 480008) using 8 μl of biotin/Ab mixture and streptavidin beads per mouse. Untouched CD8⁺ T cells were then purified by magnetic separation using LS columns (Miltenyi Biotec, catalog no. 130-042-401). Naive CD4⁺ murine T cells were isolated using naive CD4⁺ T cell isolation kits and LS columns (Miltenyi Biotec, catalog no. 130-104-453) according to the manufacturer’s protocol. T cells were activated and expanded in complete RPMI 1640. Complete RPMI 1640 was prepared by supplementing RPMI 1640 with 10% FBS (Gemini BioProducts, catalog no. 100-106), 2 mM l-glutamine (Thermo Fisher Scientific, catalog no. 25030-081), 10 mM HEPES (Corning, catalog no. 25-060-Cl), 100 U/ml penicillin/streptomycin (Thermo Fisher Scientific, catalog no. 15140-122), 1:100 MEM nonessential amino acids (Thermo Fisher Scientific, catalog no. 11140050), 55 µM 2-ME (Thermo Fisher Scientific, catalog no. 21985023), and 50 µg/ml gentamicin (Quality Biological, catalog no. 120-098-661).

CD8⁺ and CD4⁺ T cells were activated with plate-bound anti-CD3 (5 µg/ml, Bio X Cell, catalog no. BP0001-1) and soluble anti-CD28 (2 µg/ml, Bio X Cell, catalog no. BE0015-1). CD4⁺ T cells were additionally cultured with 10 ng/ml IL-12p70 (PeproTech, catalog no. 210-12), 10 ng/ml IFN-γ (PeproTech, catalog no. 315-05), and 5 μg/ml anti–IL-4 (Bio X Cell, catalog no. BE0045) at the time of activation for proper Th1 skewing. T cells were seeded for activation at a density of 1 × 10⁶ cells/ml. After 48 h, T cells were removed from the activating plate and expanded by 2-fold expansion with IL-2 (10 ng/ml, PeproTech, catalog no. 212-12). These cells were maintained by 2-fold IL-2 expansion every 48 h thereafter, until the experimental endpoint was met. BioE-1197 was administered at the time of activation, 24 h after activation, and at each addition of IL-2 for expansion to maintain pathway inhibition during T cell activation and differentiation. The final concentration of BioE-1197 for each dosing was 50 µM based on previous studies (13, 14). The pan-SIK inhibitors HG-9-91-01 and YKL-05-099 were administered at the same intervals as BioE-1197 at a final concentration of 25 and 270 nM, respectively. The class IIa HDAC inhibitor TMP269 was administered at the time of expansion in IL-2, so as to only inhibit HDAC activity during the time frame in which BioE-1197 diminishes HDAC7 stability, at a concentration of 12.5 µM. Control differentiation conditions for all experiments received an equivalent volume of the vehicle, DMSO.

On day 6 after initial activation, the remaining live T cells were purified by Ficoll density centrifugation (Ficoll-Paque Plus, Cytiva, catalog no. 17144002). Cells were then reactivated by plate-bound anti-CD3 (1 µg/ml) and anti-CD28 (1 µg/ml) at a density of 1 × 10⁶ cells/ml. For intracellular cytokine staining, 1 h after reactivation, monensin and brefeldin A protein transport inhibitors were added at a dilution factor of 1:1000 (BD Biosciences, GolgiStop, catalog no. 554724, GolgiPlug, catalog no. 555029). After 3.5 h of total restimulation, cells were harvested for intracellular cytokine staining and fixed using 1% methanol-free paraformaldehyde (Thermo Fisher Scientific, catalog no. 28906) for 10 min at 37°C. For ELISAs, supernatants were harvested 3.5 h after reactivation and flash-frozen in liquid nitrogen. IFN-γ and TNF-α ELISAs (eBioscience, catalog nos. 88-7314 and 88-7324) were performed per the manufacturer’s protocol.

Single-cell suspensions of splenocytes from congenically marked, littermate P14 TCR transgenic mice were generated by dissociation through a 70-µm filter. CD8⁺ T cells were then activated by culturing whole splenocytes in complete RPMI 1640 at a density of 4 × 10⁶ cells/ml with lymphocytic choriomeningitis virus (LCMV) gp33–41 peptide (1 µg/ml, KAVYNFATM, Johns Hopkins Medical Institute Synthesis and Sequencing Facility). Cells were expanded and dosed according to the procedure described in In vitro T cell culture. On day 6 after activation, live cells were isolated by Ficoll density centrifugation (Ficoll-Paque Plus, Cytiva, catalog no. 17144002). Control differentiated (Thy1.1⁺/Thy1.2⁺) and BioE-1197 differentiated (Thy1.1⁺) CD8⁺ T cells were mixed 1:1 and 1 × 10⁶ cells of each population were coadoptively transferred into WT C57BL/6J (Thy1.2⁺) mice by retro-orbital injection. At 32 d after coadoptive transfer, the mice were challenged with LCMV Armstrong (2 × 10⁵ PFU) by i.p. injection. LCMV Armstrong virus was a gift originally obtained from Susan Kaech (Salk Institute for Biological Studies). At 5–8 d postinfection, splenocytes were harvested and reactivated with gp33–41 peptide for 3.5 h with administration of monensin and brefeldin A (1:1000) protein transport inhibitors for ex vivo restimulation analysis and intracellular cytokine staining.

Viability was assessed by staining with eBioscience fixable viability dye eFluor 780 (Thermo Fisher Scientific, catalog no. 65-0865-14). Abs against the following proteins were purchased: IFN-γ BV786 (BD Biosciences, catalog no. 563773, RRID:AB_2738419), TNF-α BV421 (BD Biosciences, catalog no. 563387, RRID:AB_2738173), CD8 BV605 (BioLegend, catalog no. 100744, RRID:AB_2562609), CD4 BV605 (BioLegend, catalog no. 100548, RRID:AB_2563054), Thy1.1 AF700 (BioLegend, catalog no. 202528, RRID:AB_1626241), Thy1.2 BV510 (BioLegend, catalog no. 140319, RRID:AB_2561395), T-bet eF660 (Thermo Fisher Scientific, catalog no. 50-5825-82, RRID:AB_10596655), Eomes PerCP-eFluor710 (Thermo Fisher Scientific, catalog no. 46-4877-42, RRID:AB_2573759), granzyme B PE/Dazzle 594 (BioLegend, catalog no. 372216, RRID:AB_2728383), perforin PE (BioLegend, catalog no. 154306, RRID:AB_2721639), and H3K27Ac AF488 (Cell Signaling Technology, catalog no. 15485, RRID:AB_2798743). Permeabilization of cells for intracellular staining after fixation was achieved using eBioscience permeabilization buffer (Thermo Fisher Scientific, catalog no. 00-8333-56) per the manufacturer’s instructions. To ensure proper identification of positive signal and quantification, a fluorescence minus one (FMO) condition was prepared and run for every experimental protein of interest across all experimental conditions. All samples were run on a FACSCelesta (BD Biosciences) and analyzed using FlowJo v10.6.2.

Electroporation of Cas9/single-guide RNA (sgRNA) complexes for target gene deletion in naive T cells was completed as previously described (15). Briefly, naive T cells were isolated as described above and 10 × 10⁶ cells were used per sgRNA target. Cas9/sgRNA complexes were prepared in a total volume of 5 μl containing Alt-R S.p. Cas9 nuclease V3 (6 μg, IDT, catalog no. 1081059) and sgRNA (0.3 nmol, Synthego) diluted in RNase-free water. Complexes were allowed to form for 10 min at room temperature. PBS-washed, naive T cells were then pelleted and resuspended in 20 µl of P3 solution (P3 primary cell 4D-Nucleofector electroporation kit, Lonza, catalog no. V4XP-3032), transferred to tubes containing complexed Cas9/sgRNA, and mixed. This solution was then transferred to the electroporation cassette and electroporated with a Lonza 4D-Nucleofector system using the pulse code DS-137. Immediately after electroporation, the cells were supplemented with complete RPMI 1640 and allowed to rest at 37°C for 10 min. Subsequently, the P3 solution was washed out with complete RPMI 1640, after which the cells were pelleted, resuspended in fresh complete RPMI 1640 supplemented with IL-7 (10 ng/ml, PeproTech, catalog no. 217-17), and seeded at a density of 2 × 10⁶ cells/ml. Cells were rested in IL-7 for at least 2 d prior to activation to allow for genomic editing and turnover of the target protein pool. The sgRNAs used in this study were synthesized by Synthego. The sgRNA sequence used for the control sgRNA was 5′-CUCCUUUAGAUCAUUCCUUG-3′ and targets the ROSA locus. The targeted sequence for the liver kinase B1 (LKB1)–specific sgRNA was 5′-CCAGGCCGUCAAUCAGCUGG-3′. The LKB1-specific sgRNA was identified using Synthego’s CRISPR design tool.

Samples were flash-frozen at the time of harvest and then lysed in radioimmunoprecipitation (RIPA) buffer supplemented with 25× protease inhibitor mixture (Sigma-Aldrich, catalog no. 11836145001), 1 μM sodium orthovanadate (MilliporeSigma, catalog no. 450243), 10 μM sodium pyrophosphate (MilliporeSigma, S6422), 50 μM sodium fluoride (MilliporeSigma, catalog no. 7920), 1 mM PMSF (MilliporeSigma, catalog no. 1083709100), and 10 μM β-glycerophosphate (MilliporeSigma, catalog no. G9422). RIPA consists of 100 mM Tris-HCl (pH 8.0) (Quality Biological, catalog no. 351-007-101), 300 mM NaCl (MilliporeSigma, catalog no. S5886), 2% Nonidet P-40 (MilliporeSigma, catalog no. 13021), 1% sodium deoxycholate (MilliporeSigma, catalog no. 30970), and 0.2% SDS (MilliporeSigma, catalog no. L6026). Lysate was clarified and quantified using a Pierce Coomassie Plus (Bradford) assay (Thermo Fisher Scientific, catalog no. 1856209) with a SpectraMax M3 plate reader (Molecular Devices). Equivalent amounts of protein were loaded and separated on 4–12% Bis-Tris mini protein gels (Invitrogen) and then transferred to a polyvinylidene difluoride membrane (Bio-Rad, catalog no. 1620177). Blots were blocked with 5% milk at room temperature, and primary Abs were incubated in 5% BSA overnight at 4°C. Primary Abs used in these studies include antiphosphorylated HDAC4/5/7 (Cell Signaling Technology, catalog no. 3443, RRID:AB_2118723), anti-HDAC7 (Cell Signaling Technology, catalog no. 33418, RRID:AB_2716756), anti-actin (Sigma-Aldrich, catalog no. A2066, RRID:AB_476693), anti-LKB1 (Cell Signaling Technology, catalog no. 3047, RRID:AB_2198327), anti-PASK (Cell Signaling Technology, catalog no. 3086, RRID:AB_2159082), anti–lamin B (Santa Cruz Biotechnology, catalog no. sc-6216, RRID:AB_648156), and anti-IκBα (Cell Signaling Technology, catalog no. 4814, RRID:AB_390781). Secondary Abs for detection include anti-rabbit HRP (Cell Signaling Technology, catalog no. 7074, RRID:AB_2099233), anti-mouse HRP (Cell Signaling Technology, catalog no. 7076, RRID:AB_330924), and anti-goat HRP (Santa Cruz Biotechnology, catalog no. sc-2354, RRID:AB_628490). Signal was detected using SuperSignal West Pico PLUS chemiluminescent substrate or SuperSignal West Femto maximum sensitivity substrate (Thermo Fisher Scientific, catalog nos. 34580 and 34095). Images were captured with the UVP BioSpectrum 500 imaging system or the Bio-Rad ChemiDoc imaging system.

At time of fractionation, the remaining live T cells were purified by Ficoll density centrifugation (Ficoll-Paque Plus, Cytiva, catalog no. 17144002), counted, and an equal cell number was used between treatment groups. Buffer A consists of 10 mM HEPES (Corning, catalog no. 25-060-Cl), 10 mM KCl (Quality Biological, catalog no. 351-044-101), 0.1 mM EGTA (Amresco, catalog no. 0372), and 0.1 mM EDTA (Corning, catalog no. 46-034-Cl) and was prepared and supplemented with 25× protease inhibitor mixture (Sigma-Aldrich, catalog no. 11836145001), 1mM DTT (MilliporeSigma, catalog no. D9779), 1 μM sodium orthovanadate (MilliporeSigma, catalog no. 450243), and 1 mM PMSF (MilliporeSigma, catalog no. 1083709100). Cells were allowed to swell on ice with buffer A for 15 min. An equivalent volume of buffer A, supplemented with 2% Nonidet P-40 (MilliporeSigma, catalog no. 13021), was then added and cells were vortexed vigorously for 30 s. Nuclei were then pelleted at 15,000 rpm for 30 s. The supernatant containing the cytosolic fraction was then transferred to a new tube and the nuclear pellet was lysed in RIPA and subsequently clarified. Equal volumes were loaded between samples, separated on 4–12% Bis-Tris mini protein gels (Invitrogen), and subjected to Western blot. Fractionation was confirmed using the nuclear and cytosolic loading controls lamin B and IκBα, respectively.

Cellular pellets were flash-frozen upon harvest. Pellets were solubilized in TRIzol (Thermo Fisher Scientific, catalog no. 15596018) and RNA was extracted by addition of chloroform (Fisher Scientific, catalog no. C298-500) based on the manufacturer’s protocol. In brief, after mixing by inversion, layers were separated by centrifugation at 4°C. The upper, aqueous layer was collected and RNA was precipitated using isopropanol (Fisher Scientific, catalog no. A416-500) at −20°C overnight. The next day, RNA was pelleted and washed with 70% ethanol twice (Fisher Scientific, catalog no. 2016012). RNA pellets were allowed to air dry for 10 min at room temperature and then were resuspended in water. RNA was quantitated using the SpectraMax M3 (Molecular Devices) plate reader and equivalent amounts of cDNA were generated per condition using ProtoScript II reverse transcriptase (NEB, catalog no. M0368L) and random hexamers (Qiagen, catalog no. 79236) according to the manufacturers’ protocols. qRT-PCR for Ifng (Thermo Fisher Scientific, catalog no. 4331182, TaqMan probe: Mm01168134_m1), Tnf (Thermo Fisher Scientific, catalog no. 4331182, TaqMan probe: Mm00443258_m1), Gzmb (Thermo Fisher Scientific, catalog no. 4331182, TaqMan probe: Mm00442834_m1), Prf1 (Thermo Fisher Scientific, catalog no. 4331182, TaqMan probe: Mm00812512_m1), and Hdac7 (Thermo Fisher Scientific, catalog no. 4331182, TaqMan probe: Mm00469527_m1) were normalized to Rn18s (Thermo Fisher Scientific, catalog no. 4310893E) and run on a StepOnePlus real-time PCR system (Applied Biosystems) using EagleTaq universal master mix (Sigma-Aldrich, catalog no. 7260288190). Fold change differences in target gene expression were calculated by 2^−ΔΔCt.

Day 6 CD8⁺ T cells differentiated in a vehicle control or BioE-1197 were prepared as described above. Following Ficoll density centrifugation, the remaining live cells were fixed with 1% methanol-free paraformaldehyde (Thermo Fisher Scientific, catalog no. 28906) at a density of 1 × 10⁶ cells/ml in complete RPMI 1640 for 10 min at room temperature, while shaking. Fixation was terminated by addition of glycine (MilliporeSigma, catalog no. G7126) to a final concentration of 125 μM. Samples were incubated on ice for 5 min and then pelleted at 2000 rpm and washed twice with cold PBS. Pellets were flash-frozen with liquid N₂ and stored at −80°C.

Chromatin immunoprecipitation sequencing (ChIP-seq) was performed as previously detailed (16). In brief, nuclei were harvested by sequential lysis in lysis buffer (LB)1, LB2, and LB3, which were supplemented with protease inhibitor (Thermo Fisher Scientific, catalog no. 78429; 1× final concentration added right before use). LB1 consists of 50 mM HEPES-KOH (pH 7.5, MilliporeSigma, catalog nos. H4034 and P9333), 140 mM NaCl (MilliporeSigma, catalog no. 7653), 1 mM EDTA (pH 8.0, MilliporeSigma, catalog no. E6758), 10% glycerol (MilliporeSigma, catalog no. G5516), 0.5% Igepal CA-630 (MilliporeSigma, catalog no. I8896), 0.25% Triton X-100 (MilliporeSigma, catalog no. T8787), and KOH to adjust pH to 7.5. LB2 consists of 10 mM Tri-HCl (pH 8.0) (Quality Biological, catalog no. 351-007-101), 200 mM NaCl, 1mM EDTA (pH 8.0), 0.5 mM EGTA, and HCl to adjust pH to 8.0. LB3 consists of 10 mM Tris-HCl (pH 8.0), 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.1% sodium deoxycholate (MilliporeSigma, catalog no. 30970), 0.5% N-lauroylsarcosine (MilliporeSigma, catalog no. L7414), and HCl to adjust pH to 8.0. Cells were resuspended in LB1, rotated for 10 min at 4°C, pelleted at 2000 rpm, resuspended in LB2, and rotated for 5 min at 4°C. Samples were then pelleted at 2000 rpm and resuspended in LB3. Nuclei were then sonicated at 4°C using a Fisher 150E Sonic Dismembrator with the following settings: 50% amplitude, 30 s on/30 s off for 12 min total time. Following sonication, samples were pelleted to remove cellular debris at 20,000 × g for 10 min at 4°C. The supernatant was transferred and the remainder was used for ChIP.

Protein A beads (50 μl) (Thermo Fisher Scientific, catalog no. 10002D) were used per IP and transferred to a 2-ml tube on a magnetic stand. Beads were washed twice with blocking buffer (BB: 0.5% BSA in PBS), then resuspended in 100 µl of BB per IP. For each sample, 3 μg of anti-H3K27Ac Ab (Abcam, catalog no. ab4729, RRID:AB_2118291) was added to the beads, and the mixture was incubated with rotation at room temperature for 1–3 h. The 2-ml tube was then placed on a magnetic rack and washed three times with BB, before resuspending in 50 µl of BB per IP. Then, 50 µl of beads in BB was transferred to each IP sample. Samples were immunoprecipitated overnight at 4°C. The next day, samples were transferred to a 1.5-ml LoBind tube on a magnetic stand, washed six times with RIPA buffer (50 mM HEPES, 500 mM LiCl, 1 mM EDTA, 1% Igepal CA-630, 0.7% sodium deoxycholate, pH to 7.5 using KOH) and one time with TBS buffer (20 mM Tris-HCl [pH 7.5], 150 mM NaCl). ChIP samples were reverse crosslinked overnight in elution buffer (50 mM Tris-HCl [pH 8.0], 10 mM EDTA [pH 8.0], and 1% SDS). DNA was extracted from reverse-crosslinked chromatin samples by incubation with RNase A (Thermo Fisher Scientific, catalog no. EN0531) for 15 min at 37°C and proteinase K (Thermo Fisher Scientific, catalog no. AM2546) for 1 h at 65°C. DNA was subsequently purified using MinElute (Qiagen, catalog no. 28004) purification.

DNA libraries were prepared and amplified using NEBNext Ultra II DNA library prep kit for Illumina (New England Biolabs, catalog no. E7103L) and dual-indexed with NEBNext Multiplex Oligos for Illumina (New England Biolabs, catalog no. E6440S). Final libraries were pooled, quantified with Qubit (Thermo Fisher Scientific), an Agilent Bioanalyzer, and qPCR (Bio-Rad), then sequenced on a NovaSeq SP 100 flow cell (Illumina) using paired-end 2 × 50-bp reads at the Johns Hopkins Single Cell and Transcriptomics Core.

Reads were demultiplexed using bcl2fastq and were aligned to the mm9 genome using Bowtie2 (17). SAMtools (18) was used to filter for mapping quality ≥25, remove singleton reads, convert to a BAM format, remove potential PCR duplicates, and index reads. Then, deepTools (19) was used to perform a running window average (bin size of 10 bp) of BAM files, normalized by RPKM (reads per kilobase of transcript per million mapped reads), and converted into the final bigwig files. These bigwig files were plotted using pyGenomeTracks (20) with standard input parameters for Control and BioE tracks, and the subtraction operation was used for Enrichment tracks. Data files associated with this analysis have been deposited at GEO (GSE242482 https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE242482).

The protein sequences for the kinase domain of PASK and the LKB1-activated members of the AMP-activated protein kinase (AMPK)–related kinase (ARK) family were acquired from UniProt and aligned using T-Coffee (21). The alignment was visualized and colored for percent identity using Jalview (22).

For flow cytometric analysis of in vitro–cultured CD8⁺ or CD4⁺ T cells prior to or during restimulation, representative dot plots and quantification of cytokine, transcription factor, and histone marks were prepared by gating for lymphocytes, single cells, and then either CD8⁺ or CD4⁺. For flow cytometric analysis of in vitro–cultured CD8⁺ or CD4⁺ T cells during differentiation, quantification of H3K27Ac histone marks was achieved by gating for lymphocytes, single cells, live cells, and then either CD8⁺ or CD4⁺. In vivo coadoptively transferred CD8⁺ T cells were identified by gating for lymphocytes, single cells, live cells, CD8⁺, and Thy1.1⁺. Then, vehicle control cells (Thy1.1⁺/1.2⁺) were distinguished from BioE-1197 (Thy1.1⁺)-treated cells by gating for Thy1.2 expression. For all flow cytometry data, gates were set for proteins of interest based on an FMO. For quantification of geometric mean fluorescence intensity (gMFI) values, the FMO for each experimental condition was subtracted from the full stain gMFI values to account for any baseline differences in fluorescence between treatment groups. Flow data for gMFI are presented as a fold change to the average of the control condition for each individual experiment.

Western blot signal intensity was quantified by densitometry with VisionWorks software (Analytik Jena). For each Western blot sample quantified, a background densitometric reading was measured within the lane of that sample and subtracted from the total signal. Phosphorylated HDAC7 and total HDAC7 signals cannot be stripped from the membrane, so they are probed on separate, duplicate Western blots from the same experimental lysate. Western blots are presented as a fold change to the control of each independent experiment. Fold change quantification of phosphorylated and total HDAC7 over time during differentiation in control and BioE-1197 differentiated T cells were calculated to the day 1 control data. This strategy was employed to quantitate change over time, as the signal intensity in some experiments of the naive (day 0) control was too low to be able to accurately calculate a fold change (an extremely small value, or 0).

For all statistical analyses, normality of the collected dataset was evaluated. When normality was confirmed or could not be evaluated due to too small of a sample size, then the appropriate parametric statistical test was used to compare differences between treatment groups. When the normality assumption of a dataset was not met, then the nonparametric statistical test was employed. Experiment-specific statistical tests are described in their respective figure legends. All statistical analyses were conducted using GraphPad Prism version 9.5.0.

We sought to evaluate the role of the PASK/WD repeat domain 5 (WDR5) pathway in regulating T cell differentiation. Mammalian target of rapamycin complex 1 (mTORC1)–dependent activation of PASK has been shown to initiate epigenetic induction of WDR5-mediated effector differentiation programs across multiple cell types (13, 14). The conserved nature of this pathway prompted us to query whether T cells also use PASK to regulate their effector differentiation programs. To this end, CD8⁺ T cells were activated and expanded with IL-2 in either the presence of a vehicle control or the previously described PASK inhibitor, BioE-1197 (23). Effector differentiation was assessed by evaluating cytokine recall potential through the measurement of effector cytokine production upon restimulation, 6 d after initial activation. CD8⁺ T cells differentiated in the presence of BioE-1197 had significantly enhanced IFN-γ and TNF-α effector cytokine production when restimulated compared with control differentiated T cells based on increases in both the percent of cells positive for intracellular cytokine staining and gMFI (Fig. 1A–C). This enhanced cytokine production was further confirmed by assessing cytokine release in response to restimulation by ELISA (Fig. 1D).

Whereas BioE-1197 enhanced cytokine production in CD8⁺ T cells, there was no observed increase in expression of the lineage-specifying transcription factor, T-bet, by either percent or gMFI intensity (Supplemental Fig. 1A–C). Additionally, other effector molecules associated with the CD8⁺ T cell effector differentiation program, granzyme B and perforin, did not exhibit BioE-1197–dependent regulation similar to the cytokine effector molecules. There was no observed difference in granzyme B expression between control and BioE-1197 differentiated CD8⁺ T cells either by percent or gMFI (Supplemental Fig. 1D–F), and although there was a statistically significant increase in perforin expression by percent between control and BioE-1197–treated cells (16.5 versus 21.8%), there was no difference by gMFI and these changes were minimal compared with those observed for cytokine production (Supplemental Fig. 1G–I). Taken together, these data suggest that BioE-1197 was acting to selectively regulate the CD8⁺ T cell effector cytokine program.

Despite the pronounced ability of the PASK inhibitor, BioE-1197, to enhance effector cytokine production, differentiated PASK KO CD8⁺ T cells did not have elevated IFN-γ and TNF-α cytokine production upon restimulation. Moreover, differentiation of PASK KO CD8⁺ T cells in the presence of BioE-1197 still enhanced cytokine production to levels similar to those of wild-type (WT) cells (Fig. 1E, 1F). These data indicated that BioE-1197 enhances CD8⁺ T cell effector cytokine production in a PASK-independent manner, thus illuminating a previously unrecognized mechanism of BioE-1197 involved in regulation of cytokine recall potential in CD8⁺ T cells.

In addition to revealing a previously unappreciated target of BioE-1197, these findings suggested that BioE-1197 produced a potentially durable phenotype, as the cells were washed and no BioE-1197 was added during the restimulation assay to actively increase cytokine production. Although the exact half-life of BioE-1197 is unknown, these data could suggest that the PASK-independent pathway inhibited by BioE-1197 is active during T cell differentiation, rather than during restimulation. Inhibition of this pathway during the course of differentiation could, therefore, result in the stable programming of enhanced cytokine recall, instead of directly enhancing cytokine production during recall. To further assess the sustainability of BioE-1197–enhanced cytokine programming, we generated CD8⁺ lymphocytic choriomeningitis virus (LCMV)–specific P14 T cells treated with either a vehicle or BioE-1197 in vitro and coadoptively transferred these cells, at a 1:1 ratio, into naive WT C57BL/6J hosts. We let these T cells rest in the naive hosts for 32 d and then recalled the T cells with LCMV infection. After challenge, we evaluated cytokine production by ex vivo peptide restimulation assays (Fig. 1G). After this prolonged rest period, BioE-1197 differentiated CD8⁺ T cells still demonstrated enhanced cytokine production compared with their control differentiated counterparts (Fig. 1H, 1I) without any observed differential expression of T-bet or eomes, transcription factors known to regulate effector cytokine production (Supplemental Fig. 1J, 1K). The stable elevation of enhanced cytokine production by BioE-1197 differentiated cells in this park and recall model further supports the hypothesis that the previously unappreciated target of BioE-1197, capable of regulating effector cytokine production, is active during CD8⁺ T cell differentiation. Thus, we sought to identify the unknown pathway inhibited by BioE-1197 during effector differentiation that regulates cytokine recall potential in CD8⁺ T cells.

To elucidate how BioE-1197 stably enhances effector CD8⁺ T cell cytokine production upon restimulation, we sought to identify candidate kinases that could be inhibited by BioE-1197. To this end, we used the protein basic local alignment search tool (BLAST) to query the amino acid sequence corresponding to the kinase domain of PASK to identify similar kinases. This analysis revealed 13 members of the ARK family as potential targets of BioE-1197 (Supplemental Fig. 2). Based on this result, if BioE-1197–enhanced cytokine recall was dependent on ARK family inhibition, it could be due to the selective inhibition of a specific member of this family or the collective inhibition of all/many members of this family. As activation of each of these ARK family members is mediated by LKB1 T-loop phosphorylation (24), we first sought to evaluate whether BioE-1197–enhanced cytokine recall was dependent on ARKs by knocking out the shared upstream activating kinase, LKB1, in naive CD8⁺ T cells. Although prior studies have demonstrated important roles for LKB1 in T cell development, survival, and activation, these studies have not evaluated cytokine recall potential of LKB1-deficient T cells (25–28). Therefore, to knock out LKB1, naive T cells were electroporated with either control or LKB1-specific Cas9/single guide RNA (sgRNA) complexes (15). Cells were then rested in IL-7 to maintain viability of these primary cells while allowing for genomic editing and LKB1 protein turnover prior to activation. Upon activation, control or LKB1 sgRNA-edited T cells were differentiated in a vehicle control or BioE-1197 conditions and cytokine recall was evaluated. Particularly, by culturing LKB1-deficient cells in BioE-1197 differentiation conditions we would be able to assess whether LKB1 was necessary for enhanced effector cytokine production.

In these experiments, LKB1-deficient CD8⁺ T cells exhibited diminished survival compared with control edited CD8⁺ T cells (Fig. 2A, 2B). Although this is a phenotype previously described for LKB1-deficient CD8⁺ T cells (25–27), it is not a phenotype observed with BioE-1197 treatment (Fig. 2A, 2B), suggesting that BioE-1197 is not likely functioning to inhibit all LKB1-activated ARKs. Upon restimulation, vehicle control treated LKB1-deficient CD8⁺ T cells show no increase in IFN-γ or TNF-α cytokine production (Fig. 2C, 2D), suggesting that LKB1 deletion is not sufficient to elevate cytokine. However, BioE-1197–treated LKB1-deficient CD8⁺ T cells cannot elevate cytokine recall to the same extent as control edited CD8⁺ T cells differentiated in BioE-1197 (Fig. 2C, 2D), suggesting that LKB1 is necessary for BioE-1197–enhanced cytokine production. Although not to the same extent as control edited cells, there is a minor increase in cytokine production in LKB1-deficient CD8⁺ T cells differentiated in BioE-1197 compared with vehicle control conditions. This slight increase could be due to incomplete KO of LKB1 using CRISPR/Cas9 in a polyclonal setting, leaving some residual pathway activity to be regulated by BioE-1197 (Fig. 2E). Nonetheless, the diminished ability of BioE-1197 to elevate cytokine production upon restimulation in LKB1-deficient cells compared with control edited cells suggests that BioE-1197–enhanced cytokine recall is highly dependent on LKB1-activated ARK family members.

Given the observed dependence of BioE-1197–enhanced cytokine recall on LKB1-activated ARK family members, we hypothesized that if BioE-1197 is functioning as an ARK family kinase inhibitor, then BioE-1197 differentiated T cells should have diminished phosphorylation of known ARK family substrates. Notably, many members of the LKB1-activated ARK family have been described to phosphorylate class IIa HDACs across numerous cell types (29–35). Given the broad ability of LKB1-activated ARKs to phosphorylate class IIa HDACs, we sought to further confirm that BioE-1197 was functioning to inhibit ARK family members during CD8⁺ T cell differentiation by evaluating class IIa HDAC phosphorylation. To this end, control and LKB1 sgRNA-edited CD8⁺ T cells differentiated in control or BioE-1197 conditions were assessed for phosphorylation of class IIa HDACs on day 6 after activation. Control sgRNA-edited CD8⁺ T cells exhibited distinct phosphorylation of the class IIa HDAC, HDAC7, that was expectedly diminished in LKB1-deficient CD8⁺ T cells. In addition to a reduction in phosphorylated HDAC7, LKB1-deficient CD8⁺ T cells also exhibited a decrease in total HDAC7 (Fig. 2E, 2F). The concurrent reduction in phosphorylated and total HDAC7 protein in LKB1-deficient T cells was not unexpected, as the evaluated HDAC7 phosphorylation site S178 has been described to be critical for the stability of HDAC7 protein (36–38). Therefore, these results indicate that LKB1-deficient CD8⁺ T cells have diminished phosphorylation-dependent stabilization of HDAC7. These experiments additionally demonstrated that BioE-1197 differentiated CD8⁺ T cells exhibit diminished phosphorylation-dependent stabilization of HDAC7, as these cells also have reduced phosphorylated and total HDAC7 (Fig. 2E, 2F). The observed reduction in phosphorylated and total HDAC7 in either LKB1-deficient or BioE-1197 differentiated CD8⁺ T cells could be slightly enhanced by BioE-1197 treatment of LKB1-deficient T cells, although the effect did not reach statistical significance. This could also be explained by incomplete LKB1 deletion using electroporation of CRISPR/Cas9 sgRNA on a polyclonal population, allowing for BioE-1197–mediated inhibition of residual LKB1-dependent pathway activity in the combination treatment (Fig. 2E, 2F). Taken together, the reduced ability of BioE-1197 to elevate cytokine recall in LKB1-deficient T cells and the diminished phosphorylation of the known ARK substrate, HDAC7, further support that BioE-1197 is acting as an inhibitor of ARK family members during T cell differentiation. Additionally, these data reveal the inactivation of ARKs by either LKB1 deletion or BioE-1197 inhibition, during differentiation, leads to inhibition of phosphorylation-dependent stabilization of HDAC7.

Although numerous ARK family members have been described to phosphorylate class IIa HDACs, a previous report has specifically identified the ARK family member, SIK1, as a regulator of HDAC7 phosphorylation-dependent stabilization in cardiomyocytes (36). Given the necessity of LKB1-activated ARKs in BioE-1197 cytokine regulation and the observed inhibition of HDAC7 stability in LKB1-deficient and BioE-1197–treated CD8⁺ T cells, we sought to evaluate whether SIKs were responsible for HDAC7 stability and cytokine recall capacity in differentiating CD8⁺ T cells. To this end, we employed two pan-SIK inhibitors, HG-9-91-01 and YKL-05-099, capable of inhibiting SIK1, SIK2, and SIK3. CD8⁺ T cells differentiated in the presence of either pan-SIK inhibitor reduced phosphorylation-dependent stabilization of HDAC7, as both phosphorylated and total HDAC7 levels were decreased in HG-9-91-01 and YKL-05-099 differentiated effectors compared with control differentiation conditions (Fig. 3A, 3B). Moreover, CD8⁺ T cells differentiated in the presence of pan-SIK inhibition consistently exhibited elevated lineage-specific cytokine production of IFN-γ and TNF-α upon restimulation (Fig. 3C, 3D). These findings indicate that inhibition of ARK family SIKs, during T cell differentiation, is sufficient to reproduce the effects observed with BioE-1197 with respect to HDAC7 stabilization and cytokine recall potential and further support that BioE-1197 is functioning as a selective ARK family inhibitor.

These collective data revealed that LKB1-dependent activation of SIKs regulates HDAC7 phosphorylation-dependent stabilization during T cell differentiation. To further confirm that this is truly differential HDAC7 protein stability, we assessed Hdac7 mRNA transcript levels in CD8⁺ T cells activated and differentiated in either control or BioE-1197 conditions. On day 6 after differentiation, there was no difference in the mRNA expression of Hdac7 reflective of the differences we observed at the protein level (Fig. 4A), further supporting that the differential protein levels of HDAC7 observed between control and BioE-1197 differentiated CD8⁺ T cells is due to differential protein stability.

Given that HDAC7 functions as a transcriptional repressor (39) and BioE-1197 inhibition of phosphorylation-dependent stabilization of HDAC7 results in reduced protein levels of this transcriptional repressor, we hypothesized the differential stability of HDAC7 in control and BioE-1197 differentiated cells may be responsible for differential cytokine production. Specifically, we anticipated that diminished HDAC7 protein levels in BioE-1197 differentiated effector T cells should limit the amount of HDAC7 available in the nucleus of these cells to mediate transcriptional repression, thus enhancing cytokine production. However, although phosphorylation of HDAC7 at S178 is stabilizing for HDAC7 protein, it is also known to disrupt the regular cycling of HDAC7 into and out of the nucleus, as the stabilizing interaction with 14-3-3 occurs within the cytosol (40–42). Therefore, to determine whether HDAC7 stability underlies differential cytokine production, we first assessed whether diminished stability of HDAC7 would result in diminished nuclear availability of HDAC7. Control and BioE-1197 differentiated CD8⁺ T cells were generated and nuclear and cytosolic localization of HDAC7 was assessed after activation by subcellular fractionation and Western blot. This analysis revealed that inhibition of HDAC7 stabilization with BioE-1197 impacts HDAC7 levels both within the cytosolic and nuclear compartments, with BioE-1197–treated effector T cells having lower nuclear levels of HDAC7 compared with control differentiated effector T cells (Fig. 4B). Normalization of HDAC7 levels to subcellular compartment loading controls and calculation of a fold change relative to the control condition, for each compartment, further reveals that BioE-1197 inhibition equivalently reduces both nuclear and cytosolic HDAC7 levels (Fig. 4C). These observations suggest that during T cell differentiation, phosphorylation-dependent stabilization of HDAC7 functions to increase the total amount of HDAC7 in the cell and, therefore, the amount of HDAC7 that can be imported into the nucleus to mediate transcriptional repression.

The diminished nuclear HDAC7 protein levels in BioE-1197 differentiated cells further support the hypothesis that differential HDAC7 stability between control and BioE-1197 differentiated cells could lead to altered HDAC7 transcriptional repression and therefore cytokine regulation. HDAC7 transcriptional repression can be mediated epigenetically by the removal of the activating chromatin transcription mark, H3K27Ac. Therefore, we sought to assess whether differential nuclear HDAC7 levels in control and BioE-1197 differentiated T cells would specifically result in differential total H3K27Ac marks and epigenetic alterations in these conditions. To this end, control and BioE-1197 differentiated effector T cells were assessed for total H3K27Ac levels by flow cytometry on day 6 prior to and during restimulation. Notably, BioE-1197–treated CD8⁺ T cells have elevated H3K27Ac levels at baseline on day 6 compared with control differentiated CD8⁺ T cells, and during restimulation, H3K27Ac levels remain higher in BioE-1197–treated CD8⁺ T cells than control differentiated cells (Fig. 4D). These data indicate that control differentiated CD8⁺ T cells exhibit stabilized HDAC7 that correlates with low H3K27Ac levels. Alternatively, inhibition of HDAC7 stabilization by BioE-1197 during differentiation reduces nuclear HDAC7 levels correlating with increased H3K27Ac activating transcription marks, globally.

Collectively, these data demonstrate that enhanced HDAC7 stability during T cell differentiation correlates with an epigenetic restriction of H3K27Ac marks. These data, thus, support a model whereby T cells use LKB1-activated SIKs to mediate phosphorylation-dependent stabilization of HDAC7 to increase HDAC7 activity by increasing its nuclear availability. BioE-1197 differentiated cells, however, cannot stabilize HDAC7 through LKB1-mediated SIK activity. Therefore, these cells have diminished nuclear HDAC7 levels and elevated H3K27Ac marks. If this model were true and capable of regulating cytokine recall due to an epigenetically altered genomic landscape, one would expect BioE-1197 differentiated effector CD8⁺ T cells to have elevated cytokine transcript abundance upon restimulation in addition to the observed increases in cytokine at the protein level. To confirm this, we activated and differentiated CD8⁺ T cells in control or BioE-1197 differentiation conditions. On day 6 after activation, these cells were restimulated and effector molecule mRNA transcript abundance was assessed. Indeed, BioE-1197 differentiated CD8⁺ T cells have elevated cytokine transcript abundance 2 h after restimulation compared with control differentiated effectors (Fig. 4E), further supporting the proposed model. Importantly, granzyme B and perforin mRNA, which were not impacted by BioE-1197 treatment at the protein level, did not exhibit differences at the transcript level (Fig. 4E). These data indicate that LKB1-dependent SIK activity regulates HDAC7 stability to selectively regulate transcriptional recall of effector cytokine production in differentiating CD8⁺ T cells.

To assess whether the observed transcriptional regulation of effector cytokine production was through direct regulation of H3K27Ac levels within the cytokine gene loci, we performed H3K27Ac ChIP-seq on day 6 after activation for control and BioE-1197 differentiated CD8⁺ T cells. Analysis of the Ifng loci revealed enhanced H3K27Ac marks in BioE-1197 differentiated CD8⁺ T cells compared with control-treated cells. Of note, BioE-1197–treated cells particularly exhibited enhanced H3K27Ac marks across biological replicates within a previously reported, TCR signaling-responsive, conserved noncoding sequence, +40 kb, from the transcription start site denoted by the black bar in Fig. 4F and Supplemental Fig. 3A (43, 44). Additionally, the Tnf loci exhibited enhanced H3K27Ac marks in BioE-1197 differentiated CD8⁺ T cells, also, in a previously reported TCR signaling-responsive conserved noncoding sequence, −9 kb, from the transcription start site denoted by the black bar in Fig. 4G and Supplemental Fig. 3B (45, 46). Importantly, the Tbx21 loci, which showed no differences in expression upon treatment with BioE-1197, exhibited no enhancement of H3K27Ac marks within BioE-1197 differentiated cells (Supplemental Fig. 3C). Moreover, other cytokine loci not associated with the CD8⁺ T cell effector program, including the IL-17A and IL-4 loci, showed no H3K27Ac deposition in both control and BioE-1197 differentiated CD8⁺ T cells (Supplemental Fig. 4). Taken together, these data indicate that stabilization of HDAC7, during T cell differentiation, results in selective reductions of H3K27Ac levels in known, distant regulatory elements of the lineage-specific effector cytokine loci. These reduced H3K27Ac marks, in turn, result in transcriptionally blunted cytokine recall potential.

These data indicated that LKB1-dependent SIK activity induces HDAC7 stability to epigenetically regulate effector cytokine programs in CD8⁺ T cells. Given that this effector cytokine program is shared by CD4⁺ Th1 cells, we sought to determine whether this pathway was also active in Th1 cells. CD4⁺ naive T cells were activated and skewed to a Th1 lineage and treated with a vehicle control or BioE-1197. Th1 cells differentiated in the presence of BioE-1197 also exhibited elevated IFN-γ and TNF-α production in response to restimulation (Fig. 5A–C). Th1-skewed CD4⁺ T cells differentiated in the presence of BioE-1197 exhibited statistically significant increases in the expression of the lineage specifying transcription factor, T-bet, by percent (mean, 97.6 versus 99.1%) and gMFI (mean, 1 versus 1.256), but the magnitude of these differences was quite small and unlikely to explain the differences in observed cytokine production (Fig. 5D–F). Moreover, BioE-1197 differentiated Th1 cells exhibit similarly diminished HDAC7 phosphorylation-dependent stabilization and nuclear availability compared with control differentiated cells (Fig. 5G, 5H), and this Th1 phosphorylation-dependent stabilization of HDAC7 and effector cytokine production is also dependent on ARK family SIKs (Fig. 6).

These experiments reveal that LKB1-activated ARK family SIKs regulate CD8⁺ and Th1 phosphorylation-dependent stabilization of HDAC7 and effector cytokine production. Given the unexpected observations that these differentiating CD8⁺ and Th1 cells engage this pathway to epigenetically constrict cytokine production, we sought to evaluate the kinetics of HDAC7 phosphorylation-dependent stabilization during differentiation to understand why T cells engage this negative regulatory pathway. CD8⁺ and Th1 cells were activated and expanded in either control or BioE-1197 differentiation conditions, and phosphorylation-dependent stabilization of HDAC7 was assessed by Western blot. Phosphorylation-dependent stabilization of HDAC7 in control differentiated CD8⁺ and Th1 cells increased after activation as these cells underwent differentiation, whereas CD8⁺ and Th1 cells differentiated in the presence of BioE-1197 could not undergo phosphorylation-dependent stabilization of HDAC7 (Fig. 7A, 7B). Notably, in both CD8⁺ and Th1 cells, dramatic differences in HDAC7 stability between control and BioE-1197 differentiated cells began to be observed on day 3 after activation, following the expansion of these cells in IL-2. Interestingly, in CD4⁺ Th1 cells there are also subtle differences between control and BioE-1197 conditions for HDAC7 early during differentiation that are not observed in CD8⁺ T cells. Nonetheless, these findings indicate that inhibition of phosphorylation-dependent stabilization during the differentiation process leads to the largest differences in total HDAC7 levels in the later stages of differentiation, during expansion in IL-2, for both CD8⁺ and Th1 cells.

Given that these results revealed HDAC7 stabilization occurs during the late stages of T effector differentiation, we additionally sought to evaluate the kinetics of the global activating transcription mark, H3K27Ac, over time during differentiation. In general, H3K27Ac levels increase over the naive condition in CD8⁺ and Th1 cells during their initial activation but begin to decline as differentiation progresses (Fig. 7C); however, the exact regulation of these marks, on a total scale, exhibits cell type–specific patterns. In CD8⁺ T cells, the initial accumulation of H3K27Ac correlates with low levels of phosphorylated and total HDAC7. Upon day 3, when these CD8⁺ T cells begin to exhibit differential HDAC7 stability, global H3K27Ac levels begin to decline in both control and BioE-1197 differentiated cells. However, in BioE-1197 differentiated CD8⁺ T cells, which have low levels of HDAC7, we observed elevated H3K27Ac marks compared with the control condition during the late stages of differentiation. This general decline over the late stages of differentiation is additionally marked by fluctuations in H3K27Ac levels, with the high points and low points occurring 24 and 48 h after IL-2 exposure, respectively. Alternatively, in CD4⁺ Th1 cells, the initiation of H3K27Ac decline begins earlier. However, during late-stage differentiation, BioE-1197 differentiated cells trend toward higher H3K27Ac marks, although they are not statistically significant. This could be due to fewer genes being regulated by this pathway in CD4⁺ T cells, making resolution of these differences on a total scale more difficult. This is supported by the observation that there does seem to be selectivity in the effector genes regulated by this pathway. Moreover, CD4⁺ Th1 cells do not exhibit the late-stage H3K27Ac fluctuations associated with the CD8⁺ T cells. Nonetheless, in both populations epigenetic restriction, marked by H3K27Ac decline, is associated with late-stage differentiation during IL-2 expansion.

Next, we sought to assess whether this late-stage epigenetic restriction associated with HDAC7 stabilization is responsible for alterations in effector cytokine production in CD8⁺ and Th1 cells. To this end, we hypothesized that if late-stage HDAC7 stabilization occurs to epigenetically regulate cytokine production, then inhibition of HDAC7 deacetylase activity during the late stages of differentiation should be sufficient to elevate cytokine production. To test this hypothesis, we activated and differentiated CD8⁺ and CD4⁺ T cells in the presence of TMP269, a class IIa HDAC inhibitor, given the lack of a HDAC7-specific inhibitor. TMP269 was added at the time of expansion in IL-2, starting at day 2 after activation, so as to specifically inhibit deacetylase activity during the time frame in which control and BioE-1197 differentiated cells have differential HDAC7 stability. Upon restimulation, TMP269 differentiated CD8⁺ and Th1 cells have elevated lineage-specific cytokine production with cytokine levels similar to those of BioE-1197 differentiated effectors (Fig. 8). These observations indicate that inhibition of class IIa HDACs during the late stages of differentiation is sufficient to elevate cytokine recall upon restimulation, suggesting that epigenetic constriction of cytokine genes during late-stage differentiation is active.

Collectively, these observations identify LKB1-dependent SIK activation as a regulator of phosphorylation-dependent stabilization of HDAC7 in late-stage T cell differentiation. This enhanced HDAC7 stability increases nuclear availability of HDAC7 by increasing the amount of HDAC7 available for nuclear import. This enhanced nuclear HDAC7 protein availability leads to a reduction in H3K27Ac marks within effector cytokine loci and, therefore, restricts cytokine recall at the transcriptional level (Fig. 9A). BioE-1197–mediated inhibition of this pathway results in reduced HDAC7 stability and, therefore, diminished nuclear HDAC7 levels, enabling the persistence of H3K27Ac marks within effector cytokine loci. Thus, BioE-1197–mediated inhibition epigenetically equips effector T cells to have enhanced effector cytokine production upon recall (Fig. 9B).

In this study, we report the identification of a previously unappreciated activity of the kinase inhibitor, BioE-1197, in CD8⁺ and Th1 cells. Although previously developed as a PASK inhibitor, we demonstrate that exposure to BioE-1197 in T cells leads to the durable programming of effector CD8⁺ and Th1 cells with enhanced cytokine recall in a PASK-independent manner. The durable nature of this enhanced cytokine production was illustrated by coadoptive transfer park and recall experiments in which even after a 32-d rest in naive hosts, BioE-1197 differentiated CD8⁺ T cells maintain elevated cytokine production upon recall. Although cytokine production is a key feature of T cell effector function, the importance of effector cytokine production, specifically to antitumor immunity, has been recently highlighted. Some preclinical models of immunotherapy have been shown to be reliant on IFN-γ effector cytokine production as opposed to other effector molecules such as perforin or granzymes (47–49). Likewise, IFN-γ signaling signatures have been shown to track with clinical response in patients treated with immunotherapy checkpoint blockade (50). Given these findings, the observed ability of BioE-1197 to enhance effector T cell functionality, specifically by modulating cytokine production, has encouraging immunotherapeutic potential. Particularly, the ability to produce long-term, enhanced cytokine recall following transient differentiation in the presence of BioE-1197 could be of great utility for the generation and expansion of autologous or engineered adoptive cellular immunotherapies.

Given the therapeutic potential of this observed phenotype, we sought to identify the pathway targeted by BioE-1197 in T cells. Using CRISPR/Cas9-mediated LKB1 deletion, we demonstrated that LKB1-activated ARK family members were necessary for the observed BioE-1197 cytokine recall phenotype. Whether all 13 LKB1-activated ARK family kinases are active and have roles in T cell differentiation is unknown. To date, AMPKα1 is the only ARK, demonstrated by T cell–specific deletion, to have important functions in effector T cell activation and metabolism (26, 51, 52). As AMPKα1 deletion can recapitulate some, but not all, phenotypes associated with T cell–specific LKB1 deletion, it is likely that other ARKs are active during T cell differentiation (26). Our studies demonstrated that BioE-1197 was unlikely to be inhibiting all ARKs active during T cell differentiation, as BioE-1197 differentiated CD8⁺ T cells do not exhibit the diminished survival associated with LKB1 deletion. Therapeutically, these data demonstrate a selective advantage of BioE-1197 over targeting upstream molecules, as BioE-1197 can preserve cell viability while enhancing cytokine production. While these analyses revealed dependence of BioE-1197 on some LKB1-activated ARKs, LKB1 deletion was not sufficient to elevate cytokine recall. The observed insufficiency of LKB1 deletion to enhance cytokine recall potential could be overshadowed by the inactivation of other ARK family members during T cell differentiation. This possibility is supported by the observation that pan-SIK inhibition, downstream of LKB1 activation, is sufficient to elevate cytokine production. These collective observations suggest that individual members of the ARK family modulate distinct aspects of T cell differentiation following LKB1-mediated activation as opposed to these kinases functioning redundantly downstream of LKB1, and they identify a need to further evaluate the roles of other ARK family members in effector T cell activation and differentiation. Our data specifically indicate that SIKs are capable of regulating lineage-specific cytokine production, but more work is needed to determine whether this phenotype is dependent on a specific SIK or the collective function of SIKs during differentiation. Previous work has identified SIK2 and SIK3 to be critical regulators of thymocyte development, and SIK-dependent functions in T regulatory cells have been described (53–56). However, to our knowledge, this is the first illustration of SIK-dependent function in effector T cell differentiation.

Deletion of LKB1, pan-SIK inhibition, and BioE-1197 treatment all lead to reductions in phosphorylated and total HDAC7, a known ARK family substrate. SIKs are known to mediate HDAC7 phosphorylation, which regulates HDAC7 stability, and SIK1-mediated phosphorylation-dependent stabilization of HDAC7 has been previously reported (29, 30, 34, 36, 37). These observations suggest that LKB1-mediated SIK activation during T cell differentiation leads to phosphorylation-dependent stabilization of HDAC7. Kinetic analysis of this phosphorylation event throughout differentiation reveals phosphorylation increases after T cell activation. Moreover, inhibition of this phosphorylation event leads to a reduction in HDAC7 stability specifically during the latter part of T cell differentiation. A previous report has documented HDAC7 phosphorylation in CD8⁺ T cells. However, this prior study suggested HDAC7 phosphorylation to be constitutive and independent of TCR activation in CD8⁺ T cells, as there was no further increase in HDAC7 phosphorylation between naive and activated T cells during 30 min (57). The expanded timing of our kinetics analysis, however, reveals phosphorylation of HDAC7 to be observed long after T cell activation, with changes observed during longer time frames, on the order of days, and during differentiation and IL-2 expansion. Moreover, BioE-1197 exposure reveals that diminished HDAC7 phosphorylation particularly impacts HDAC7 stability during the later stages of differentiation, with robust differences becoming evident on day 3 after activation. These observations reveal a timed specificity in regulating HDAC7 stability during late-stage T effector differentiation and identify new potential areas of research to understand what intrinsic or extrinsic signals initiate SIK-mediated stabilization of HDAC7 specifically within this window.

The observed differential stability of HDAC7 in control and BioE-1197 differentiated T cells suggested potential differences in nuclear accumulation of HDAC7, given that increasing the total HDAC7 pool, through phosphorylation-dependent stabilization, would increase the total amount of HDAC7 available for nuclear import. However, phosphorylated HDAC7 is reported to be primarily sequestered in the cytosol, as the stabilizing interaction with 14-3-3 simultaneously blocks its nuclear import signal (36, 37, 40–42). Therefore, although phosphorylation stabilizes the total HDAC7 protein pool, phosphorylated HDAC7 would also be heavily restricted to the cytosol. Experimentally, we were able to show, however, that phosphorylation-dependent stabilization of HDAC7 simultaneously regulates HDAC7 protein levels in both the cytosol and the nucleus. Therefore, although phosphorylation is known to restrict HDAC7 to the cytosol, the simultaneous stabilization of HDAC7 increases the total amount of HDAC7 in the cell and therefore the amount of HDAC7 available for nuclear import, given that phosphorylation of HDAC7 is a reversible phenomenon (58). These data highlight an additional layer of complexity to consider in interpreting the functionality of HDAC7 protein based on phosphorylation status.

Our studies additionally revealed that the stabilization-mediated increase in nuclear HDAC7 correlates with a decline in total H3K27Ac-activating transcription marks within the late stages of T effector differentiation. This observation suggests that as T cells continue to differentiate, their activating epigenetic landscape becomes constrained to tone down effector cytokine production. Administration of BioE-1197 during differentiation was able to increase H3K27Ac levels at a global level in CD8⁺ T cells and, in turn, elevate cytokine production at both the protein and mRNA transcript level. Evaluation of these H3K27Ac marks at the end of differentiation by ChIP-seq revealed that BioE-1197 differentiated CD8⁺ T cells have enhanced H3K27Ac marks specifically within the cytokine loci compared with control differentiated T cells. In both the IFN-γ and TNF-α loci, enrichment of H3K27Ac levels, across biological replicates, in previously described TCR-dependent conserved noncoding sequences was evident. The +40 kb site for IFN-γ has been specifically described to mediate RelA binding in response to restimulation to regulate IFN-γ transcript production (43, 44), whereas the −9 kb site for TNF-α is known to mediate TCR-dependent NFAT binding (45, 46). The observed enrichment of these H3K27Ac marks specifically within these distant regulatory elements, which have known binding sites for TCR-activated transcription factors, further explains how this BioE-1197–mediated phenotype could function independent of lineage-specifying transcription factors known to regulate effector cytokine function. These observations, therefore, additionally raise the question of whether other T effector programs additionally use this pathway to regulate their cytokine programs.

The evident H3K27Ac enrichment within these sites in BioE-1197 differentiated CD8⁺ T cells supports that epigenetic restriction of these distant regulatory elements is ongoing during late-stage T effector differentiation. These data additionally suggest that BioE-1197 can block, to some extent, the observed restriction in effector functionality potentially through epigenetics. Further confirmation of this epigenetic mechanism is rooted in the observed sufficiency of class IIa HDAC inhibition, with TMP269, to elevate lineage-specific cytokine production to the same extent as BioE-1197 administration when only administered during late-stage differentiation. An underlying epigenetic mechanism also explains the durability of BioE-1197–enhanced cytokine recall potential observed in the coadoptive transfer park and recall experiments. Taken together, these data suggest that phosphorylation-dependent stabilization of HDAC7 during late-stage T effector differentiation functions to restrict cytokine recall potential and inhibition of this stabilization, with BioE-1197, epigenetically programs cells for lasting, enhanced cytokine recall.

Previous work in CD8⁺ T cells has demonstrated that overexpression of a triple nonphosphorylatable HDAC7 mutant, compared with untransduced cells, results in a reduction in IFN-γ production (57). However, in our study, drug-mediated inhibition of HDAC7 phosphorylation enhanced IFN-γ production. These results are not incongruent in that retroviral-mediated overexpression can overcome the stability phenotype associated with HDAC7 phosphorylation status by continuously replacing the turned over, nonphosphorylatable HDAC7 population. The turnover of the nonphosphorylatable HDAC7 is supported by a previous report that the nonphosphorylatable mutant is expressed at lower levels than its WT counterpart (59). Therefore, forced overexpression, by replacing the destabilized protein, is able to enhance nuclear accumulation of HDAC7 over the control despite diminished protein stability leading to a reduction in cytokine production. Likewise, in our study, control differentiated cells, in which stabilization leads to increased nuclear HDAC7 total protein, have reduced cytokine production compared with BioE-1197 differentiated cells. Therefore, in both datasets, when HDAC7 nuclear availability is increased there is a correlative decrease in IFN-γ cytokine production.

This study, through evaluation of a previously unreported function of BioE-1197, identifies a previously unappreciated pathway capable of regulating effector cytokine functionality in CD8⁺ and Th1 cells. The data presented in the current study reveal that LKB1-activated SIKs mediate HDAC7 phosphorylation-dependent stabilization and T cell cytokine recall potential. Moreover, inhibition of this pathway leads to the production of effector T cells with stably enhanced cytokine recall potential. The targeted interruption of SIK activity represents a particularly advantageous strategy for epigenetically modulating T cell functionality, as this offers enhanced selectivity over other epigenetic therapies, such as class-specific HDAC inhibitors, which would be active broadly, across many cell types. The work presented in this study defines a previously unrecognized therapeutic target for enhancing T cell functionality, and it illuminates new avenues of research in understanding the processes that regulate late-stage effector T cell differentiation.

C.H.P. and J.D.P. are current employees of Calico Life Sciences LLC. The other authors have no financial conflicts of interest.

We thank Dr. Elizabeth A. Thompson for careful revision and feedback during the preparation of this manuscript, and Drs. Anjana Rao, Patrick G. Hogan, and Srikanth Battu for advice on H3K27Ac ChIP-seq.