α-ketoglutarate-dependent KDM6 histone demethylases regulate Interferon Stimulated Gene expression in Lupus

Abstract

Objective

Investigate the hypothesis that interferon (IFN) stimulated gene (ISG) expression in systemic lupus erythematosus (SLE) monocytes is linked to changes in metabolic reprogramming and epigenetic regulation of ISG expression.

Methods

Monocytes from healthy volunteers and SLE patients at baseline or following IFNα treatment were analyzed by extracellular flux analysis, proteomics, metabolomics, chromatin immunoprecipitation and gene expression. The histone demethylases KDM6A/B were inhibited using GSK-J4. GSK-J4 was tested in pristane and resiquimod (R848) models of IFN-driven SLE.

Results

SLE monocytes had enhanced rates of glycolysis and oxidative phosphorylation compared to healthy control (HC) monocytes, as well as increased levels of isocitrate dehydrogenase (IDH2) and its product, α-ketoglutarate (α-KG). As α-KG is a required cofactor for histone demethylases KDM6A and KDM6B, we hypothesized that IFNα may be driving ‘trained immune’ responses through altering histone methylation. IFNα priming (day 1) resulted in a sustained increase in the expression of ISGs in primed cells (day 5) and enhanced expression on restimulation with IFNα. Importantly decreased H3K27 trimethylation was observed at the promoters of ISGs following IFNα priming. Finally, GSK-J4 (KDM6A/B inhibitor) resulted in decreased ISG expression in SLE patient monocytes, as well as reduced autoantibody production, ISG expression and kidney pathology in R848-treated Balb/c mice.

Conclusion

Our study suggests chronic IFNα exposure alters epigenetic regulation of ISG expression in SLE monocytes via changes in immunometabolism, a mechanism reflecting trained immunity to type I IFN. Importantly, it opens the possibility that targeting histone modifying enzymes such as KDM6A/B may reduce IFN responses in SLE.

Graphical Abstract

Introduction

In recent years it has become evident that innate immune cells (monocytes, macrophages and natural killer cells) can display long term changes in their functional programs after challenge with an infectious agent [1]. Termed ‘trained immunity’ or ‘innate immune memory’ these changes lead to increased expression of inflammatory mediators on subsequent infections with the microbial pathogen, or even cross-pathogen enhanced responses. Mechanistically epigenetic reprogramming is at the center of trained immunity, with histone modifications being shown to be central to this phenomenon, altering chromatin structure at genes regulated during these reprogramming events [2, 3]. Trained immunity is also characterized with a shift in cellular metabolism from oxidative phosphorylation to glycolysis [4, 5]. Although an important immune response to pathogens, trained immunity to endogenous ligands when inappropriately induced can potentially drive autoinflammatory or autoimmune reactions such as those observed in systemic lupus erythematosus (SLE) [1].

Our understanding regarding the pathogenesis of SLE has advanced greatly in the last 20 years, primarily driven by important insights into the role of the innate immune system in initiating and maintaining the autoimmune response. Abnormalities in monocytes and macrophages such as enhanced antigen presentation, inflammatory cytokine production and decreased clearance of dead and dying cells are all known to be associated with SLE, as are altered function and phenotype of dendritic cells and neutrophils [6–8]. In addition, nucleic acid sensing and type I interferons (IFN) have emerged as important drivers of pathobiology in SLE [9, 10]. Numerous analyses including bulk transcriptomic analyses of peripheral blood mononuclear cells have found positive correlations between disease activity, organ involvement and IFN stimulated genes (ISGs) [11, 12]. Indeed, measurement of the expression levels of ISGs such as IFI27, IFI44, IFI44L, RSAD2, IRF7 and ISG15 have been used to assign an IFN score, which not only differentiates SLE patients from healthy controls and rheumatoid arthritis (RA) patients but also correlates with disease activity [12, 13].

Changes in metabolic pathways not only support the biosynthetic and bioenergetic needs of a cell, but also play a critical role in controlling the fate or function of immune cells, including myeloid cells, through functional reprograming [14]. Naïve T cells and M2 macrophages, for example, utilize energetically efficient processes and the generation of ATP via oxidative phosphorylation (OXPHOS). Activated T cells, dendritic cells and M1 macrophages on the other hand shift towards glycolysis in order to generate intermediates for proliferating or active cells [15, 16]. Similar to other immune stimuli, anti-viral responses and chronic IFN exposure have been shown to drive changes in immunometabolism [17]. Plasmacytoid dendritic cells (pDCs), for example, upregulate both glycolysis and oxidative phosphorylation in response to type I IFNs in order to fuel the energetic demands of cytokine production [18]. Changes in immunometabolism also fuel or drive immunopathology in autoimmune and inflammatory disease [19]. In both mouse models of SLE and peripheral blood from SLE patients, CD4⁺ T cells show both enhanced glycolysis and OXPHOS [19, 20] and the use of drugs which can modify these pathways (such as metformin and rapamycin) reduces pathology in mouse models of SLE and results in decreased cytokine secretion by CD4⁺ T cells from SLE patients [21, 22].

Cellular metabolism has not only been correlated with cell function but is also an important regulator of epigenetic reprogramming that underlies trained immunity. Metabolites or intermediates on the tricarboxylic acid (TCA) cycle such as succinate, acetyl Co-A, fumarate and alpha-ketoglutaric acid (α-KG) can directly regulate epigenetic enzymes, including histone and DNA modifying enzymes [23]. α-KG, produced by oxidative decarboxylation of isocitrate by the rate limiting enzyme isocitrate dehydrogenase 2 (IDH2) or by deamination of glutamine, is a required cofactor for Jumonji C domain containing histone demethylases (KDM2–7), as well as DNA demethylases (TET1–3) [24]. The ubiquitously expressed enzymes KDM6A and 6B, demethylate histone 3 lysine 27 trimethylation (H3K27me3) marks on histones, leading to removal of these repressive marks and activation of gene expression [25]. Thus, availability of α-KG can modulate cellular fate and differentiation by regulating chromatin remodeling.

This study investigates a novel role for altered metabolism in SLE monocytes in contributing to enhanced IFN-driven gene expression changes through epigenetic reprograming. We demonstrate that SLE monocytes show enhanced rates of glycolysis and oxidative phosphorylation, which correlates with expression of ISGs. We find that IDH2 levels are increased in SLE monocytes and correspond to increased levels of intracellular α-KG and that IDH2 is an IFN-stimulated gene. In keeping with α-KG being an obligate cofactor for KDM enzymes, co-stimulation of monocytes with a cell permeable analogue of α-KG enhances ISG expression, suggesting α-KG ‘primes’ the IFN response through epigenetic reprograming of ISG promoters. Chromatin immunoprecipitation (ChIP)-PCR analysis of ISG promoters demonstrates that H3K4me3 ‘permissive’ marks are increased while H3K27me3 ‘inhibitory’ marks are decreased at ISG promoters. Inhibition of KDM6A/B, both H3K27me3 demethylases, reverses ISG expression in response to IFN priming but more importantly, reduces ISG expression in SLE patient monocytes or in the pristane model of SLE. In addition, we observed inhibition of both KDM6A/B effectively reduces disease progression in the resiquimod model of SLE. Thus, our study shows for the first time a novel link between metabolic and epigenetic reprogramming playing a fundamental and potentially targetable role in ISG expression in SLE monocytes.

Materials and Methods

Reagents

GSK-J4 (Sigma-Aldrich, SML0701) was used at 25μM unless noted otherwise. Recombinant human IFNα (Biolegend, 592704) was used at 1000 IU/ml. Dimethyl α-ketoglutarate (Sigma-Aldrich, 349631) was used at 2μM unless noted otherwise. AGI-6780 (Sigma-Aldrich, 1432660-47-3) was used at 10μM.

Patient Samples

All SLE patients (n=28) (as per ACR diagnostic criteria) were recruited from Cedars-Sinai Medical Center, CA, USA. Age- and sex-matched healthy donors with no history of autoimmune diseases or treatment with immunosuppressive agents were included. All participants provided informed written consent and the study received prior approval from the institutional ethics review board (IRB protocol. 19627). Patients were all female and aged between 23 and 63, with a median age of 43 (as shown in Supplemental table 1). Disease duration ranged from newly diagnosed to 35 years. Systemic Lupus Erythematosus Disease Activity Index (SLEDAI) scores ranged from 0–12 at the time of enrollment (Supplemental table 1).

Isolation and Culture of PBMCs and Monocytes

Peripheral blood mononuclear cells (PBMCs) were separated from whole blood by density-gradient centrifugation with Ficoll-Paque Plus (GE Healthcare). CD14⁺ monocytes were purified from freshly isolated PBMCs by positive selection using magnetic anti-CD14⁺ beads (STEMCELL Technologies) according to manufacturers’ protocol. Purified monocytes were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum and 100μg/ml of penicillin/streptomycin.

Extraction of proteins and metabolites from cells

Cell pellets containing 10 million purified macrophages were treated with 500μl cold 40% acetonitrile, 40% methanol, and 20% water. Then, samples were vortexed vigorously for 15 minutes at 4°C, spun at 10,000×g for 10 minutes at 4°C, and 450μl of supernatant was removed and dried. To resuspend, first 20 μl of methanol was added followed by vortexing, and finally 80μl of water was added along with a final vigorous vortex before injection for liquid chromatography mass spectrometry (LCMS).

Animal Studies

All animal experiments were conducted in accordance with the guidelines and approved protocols (IACUC #010058) of the Cedars-Sinai Medical Center Institutional Animal Care and Use Committee. Cedars-Sinai Medical Center is fully accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC International) and adheres to all applicable laws governing the use of laboratory animals. The laboratory animals are maintained in compliance with the relevant sections of the Animal Welfare Act and the guidelines outlined in the DHHS publication, Guide for the Care and Use of Laboratory Animals. Wild-type C57BL/6 and Balb/c female mice were purchased from Jackson Laboratory and all animals used were between 6 and 12 weeks old.

Pristane model:

6–8-week-old C57BL/6 mice (Untreated, Pristane treated, pristane+GSK-J4 treated; n=7 per group) were injected intraperitoneally with 0.5ml of phosphate buffered saline (PBS) or pristane (2,6,10,14-Tetramethylpentadecane (TMPD), Sigma, P1403) and sacrificed on Day 7. One day prior to sacrificing mice were treated with 10mg/kg GSK-J4 by peritoneal injection. Monocytes were isolated post peritoneal lavage with 10ml PBS by CD11b⁺ magnetic bead separation (StemCell Technologies). RNA was extracted from the peritoneal monocytes using Triazole, and quantitative PCR (qPCR) was performed.

Resiquimod (R848) model:

For the treatment group, resiquimod (1.5mg/kg body weight) dissolved in ethanol and acetone (1:3) was administered topically on the right ear of the mice three times a week for 5 weeks. The vehicle group received an ethanol and acetone (1:3) mixture. One group of resiquimod mice received additional GSK-J4 treatment for 4 weeks. The GSK-4 injections started after 7 days of resiquimod treatment, GSK-J4 (1mg/kg body weight) was administered through intraperitoneal (I.P.) injections, three times a week. After 5 weeks, all mice were sacrificed, and their kidneys, spleens, and blood were collected and processed for further experiments. The spleen weight was measured to calculate the spleen/body weight ratio for each mouse from each group.

Mouse serum isolation and ELISA

Mouse serum was isolated from the collected mouse blood on the same day of sacrificing by allowing the blood to clot at room temperature for 2–4 hours, followed by centrifugation at 2000g for 15 minutes to separate the serum. The levels of mouse serum IgG subtypes (IgG1 and IgG2a) were measured using the IgG1 mouse uncoated ELISA kit (Invitrogen, 885041022) and the IgG2a mouse ELISA kit (Invitrogen, EMIGG2A), respectively. The quantification of anti-dsDNA (total IgG) based on O.D at 450nm level was performed using commercially available kits (MyBiosource.com, MBS701186). Mouse serum creatinine level was measured using the mouse creatinine assay kit (Crystal Chem, 80350) with appropriate standards. All ELISAs were conducted following the manufacturer’s protocols.

Tissue processing and Immuno-histochemistry

Mouse kidneys were isolated and preserved in 10% buffered formalin for 24–48 hours. Following sectioning, tissue sections were stained with PAS and imaged using a microscope (Echo Revolve). Glomerular size was determined by calculating the radius (r) of each glomerulus in ImageJ software and area calculated from at least 40 glomeruli from each of the different mouse groups [26]. Glomerular cellularity was determined by counting total nuclear cells in each glomerulus using light microscopy [27].

Tissue processing and immuno-fluorescence

Mouse kidneys were isolated and preserved in 4% paraformaldehyde for 24 hours and subsequently transferred to 1×phosphate-buffered saline (PBS). Cryosections were prepared from kidney-containing OCT blocks. The kidney cryosections were stained with mouse Alexa 488-tagged anti-IgG antibody (Thermo-fisher Scientific, A-21202) to compare IgG deposition among different mouse groups. Complement protein 3 was identified in different groups using mouse anti-C3 antibody as the primary antibody (Abcam, ab11862), followed by rabbit anti-Rat Alexa 488 as the secondary antibody (Thermo-fisher Scientific, A-21210) and images were taken using a fluorescence microscope (Echo Revolve).

Cell Lines and Culture

THP-1 cells were purchased from the American Type Culture Collection (Manassas, VA) and cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum and 100μg/ml of penicillin/streptomycin.

Real-time Quantitative Polymerase Chain Reaction (RT-qPCR)

RNA was extracted from cell cultures using TRIzol reagent (Sigma-Aldrich) according to manufacturer’s protocol. Reverse transcription and quantitative real-time PCR data was calculated by ΔΔCt method and represented relative to 18s rRNA expression. Primers used in this study are as follows:

Human primer sequences:

• 18S-F: GCTTAATTTGACTCAACACGGGA

• 18S-R: AGCTATCAATCTGTCAATCCTGTC

• IFI27- F: TGCTCTCACCTCATCAGCAGT

• IFI27-R: CACAACTCCTCCAATCACAACT

• IFI44L-F: TCTGCCATTTATGTTGTGTGACA

• IFI44L-R: CAGGTGTAATTGGTTTACGGGAA

• IFI44-F: GGTGGGCACTAATACAACTGG

• IFI44-R: CACACAGAATAAACGGCAGGTA

• IDH2-F: CGCCACTATGCCGACAAAAG

• IDH2-R: ACTGCCAGATAATACGGGTCA

• GLS-F: TACAGGATTGCGAACGTCTG

• GLS-R: TTCCTCCAGACTGCTTTTTAGC

• GLS2-F: CTCTGCAGAAGGCCCTGA

• GLS2-R: GACAGCATGCCACTTTCTGA

• ISG15-F: CGCAGATCACCCAGAAGATCG

• ISG15-R: TTCGTCGCATTTGTCCACCA

• CXCL10-F: GGTGAGAAGAGATGTCTGAATCC

• CXCL10-R: GTCCATCCTTGGAAGCACTGCA

• TNFa-F: GGCGTGGAGCTGAGAGATAAC

• TNFa-R: GGTGTGGGTGAGGAGCACAT

• HIF1A-F:ATCCATGTGACCATGAGGAAATG

• HIF1A-R: TCGGCTAGTTAGGGTACACTTC

Murine primer sequences:

• Ifi27-F:GACTCTCCGTGCCATCTACTG

• Ifi27-R:CCTCTATCGCCATATCTGCCAC

• Cxcl10-F: CCAAGTGCTGCCGTCATTTTC

• Cxcl10-R: GGCTCGCAGGGATGATTTCAA

• Rantes-F: GCTGCTTTGCCTACCTCTCC

• Rantes-R: TCGAGTGACAAACACGACTGC

• Isg15-F: GGTGTCCGTGACTAACTCCAT

• Isg15-R: TGGAAAGGGTAAGACCGTCCT

RNA sequencing

Library Preparation:

Cells designated for transcriptomic analysis were harvested in Trizol and total RNA extracted using a QIAGEN RNEasy Mini Kit. Total RNA was then used for library preparation and sequencing. The SMART-Seq V4 Ultra Low RNA Input Kit for Sequencing (Takara Bio USA, Inc., Mountain View, CA) was used for reverse transcription and generation of double stranded cDNA for library preparation using the Nextera XT Library Preparation kit (Illumina, San Diego, CA). Libraries were sequenced on a NextSeq 500 using with a 1×75 bp read length and coverage of ~60M reads/cell.

Bioinformatics and data analysis:

Raw reads obtained from RNA-Seq were aligned to the transcriptome using STAR (version 2.5.0) /RSEM (version 1.2.25) with default parameters, using a custom human GRCh38 transcriptome reference downloaded from https://www.gencodegenes.org, containing all protein coding and long non-coding RNA genes based on human GENCODE version 33 annotation. Expression counts for each gene in all samples were normalized by a modified trimmed mean of the M-values normalization method and the unsupervised principal component analysis (PCA) was performed with DESeq2 Bioconductor package version 1.10.1 in R version 3.6.3. Each gene was fitted into a negative binomial generalized linear model, and the Wald test was applied to assess the differential expressions between two sample groups by DESeq2. Benjamini and Hochberg procedure was applied to adjust for multiple hypothesis testing, and differential expression gene candidates were selected with a false discovery rate less than 0.05.

Oxygen Consumption and Acidification Analysis

Purified monocytes or THP-1 cells were plated at a density of 250,000 cells/well in 50μl and analyzed on using the Seahorse XF^e96 Extracellular Flux Analyzer (Seahorse Biosciences). Final concentrations of inhibitors are as follows: 2mM oligomycin, 2mM FCCP, 0.5mM antimycin A and 0.5mM rotenone.

Western blot analysis

Proteins were separated on a 7–12% Tris-acetate gel (BioRad) and electrophoretically transferred to PVDF membranes and incubated with primary antibodies (alpha-actinin (3134S), IDH2 (56439), Cell signaling Technologies).

Mitoplex

A tier 2 level, targeted proteomic analysis was described previously [28] and was used to quantify mitochondrial proteins between healthy control monocytes and SLE patient monocytes. Raw data were processed using the Skyline software package (Skyline Daily, version 21.1.0.146) and fragment level data were exported as comma-separated values and further processed using a custom R script. Only fragments with 20% coefficient of variation (CV) across the heavy standard peptides were used for quantification. The abundance ratios of endogenous to heavy standard for all quantified fragments for each peptide from each protein were averaged, and peptide level abundance ratios were further averaged to yield a final abundance ratio for each of the proteins monitored. Adjusted p values were calculated using Excel, two tailed t-test was used for determination of significance.

Metabolomic Analyses (dMRM).

Cell metabolite extractions were analyzed with an Agilent 6470A Triple quadrupole mass spectrometer, operating in negative mode, connected to an Agilent 1290 Ultra High-Performance Liquid Chromatography (UHPLC) system (Agilent Technologies) as previously described. The MassHunter Metabolomics dMRM Database and Method was used to scan for up to 219 polar metabolites within each sample (Agilent Technologies). Resulting chromatograms were visualized in Agilent MassHunter Quantitative Analysis for QQQ. The final peaks were manually checked for consistent and proper integration. Statistical analysis was completed in Agilent Mass Profiler Professional (Agilent Technologies).

Flow Cytometry

Surface antibodies such as CD3 (clone OKT3, APC-Fire 750, Biolegend 317352), CD4 (clone SK3, Percp efluro 710, Invitrogen 46004742), CD8 (clone RPA-T8, BV650, BD BioSciences 563821), CD56 (clone NCAM 16.2, BV605, BD Horizon 562779), CD14 (clone M5E2, BUV 805, BD BioSciences 612903), CD16 (clone 3G8, BV786, BD Horizon 563690), CD19 (clone HIB 19, FITC, Invitrogen 11019942) and Siglec 1 (clone 7–239, BV421, BD Biosciences 742991) was used to phenotype PBMCs. In addition, live/dead dye (GDV510, Tonbo 130870T100) was used to detect live cells. Cells were incubated in CD16/32 (Fc block; BD Biosciences) for 15 mins at room temperature (RT) followed by staining with extracellular antibodies for 30mins at RT in Brilliant Stain Buffer (BD 563794). After surface staining, cells were fixed and permeabilized using eBioSciences Foxp3/Transcription Factor Buffer Set (Invitrogen, 552300) and stained with IDH2 antibody (clone EPR7577, PE, Abcam ab212122; dilution 1:200). Histone H3K27me3 was measured using Tri-Methyl-Histone H3 (Lys27) (C36B11) Rabbit mAb with Alexa Fluor^® 488 Conjugate (Cell signaling technology 5499S). Cell fluorescence was acquired on Symphony S5 (BD Biosciences) and analyzed with the FlowJo software version 10 (Treestar).

Trained Immunity

Monocytes (5×10^5 cells/ml) in RPMI and stimulated for 24h with IFNα (1000U/ml) or PBS, washed and allowed to rest for 4–5 days. Cells were washed and plated at 5×10^5 cells/ml and received a second stimulation of IFNα (1000U/ml) or PBS. After 24 hours cells were washed and harvested for RNA extraction.

Chromatin Immunoprecipitation (ChIP)

Fixation with 1% formaldehyde, sonication, and immunoprecipitation were performed with components of the ChIP IT kit according to the manufacturer’s instructions (Active Motif, Carlsbad). Chromatin DNA was immunoprecipitated with 1.5μg of one of the following antibodies: Tri-Methyl Histone K27 (Cell Signaling Technology, 9733S); Tri Methyl K4 (Abcam, ab8580). Normal rabbit IgG from (Cell Signaling Technology, 2729P) served as the control immunoprecipitation (IP) antibody. DNA was extracted using a DNAeasy kit (Qiagen) and analyzed by qPCR by the ΔΔCt method, having normalized samples using input DNA. Data is represented as percentage of input DNA. Primers used for PCR analysis are as follows:

• IFI27_F: CTTCTGGACTGCGCATGAGG

• IFI27_R: CCACCCCGACTGAAGCACTG

• ISG15_F AGGTGTTTCCAGGGTGTTGG

• ISG15_R GATGAGTTCGCTGCCTCTCA

• RSAD2(ISRE)_F: GAGAAACCAGGAATGCTCG

• RSAD2(ISRE)_R: CCTTCTTTGACTAACACTCAGC

• MX1_F: GGGACAGGCATCAACAAAGCC

• MX1_R: GCCCTCTCTTCTTCCAGGCAAC

• OAS1_F: TGAAATTCAGCACTGGGATCAG

• OAS1_R: GGAGGAGCTGTCTTTGCACTTT

Statistical Analysis

All data are expressed as mean ± SD. Statistical differences were measured using either a Student’s paired or unpaired t-test or 2-way analysis of variance (ANOVA) with Tukey’s post-hoc test when appropriate. When the data analyzed was not distributed normally, we used the Mann–Whitney test or Kruskal–Wallis 1-way ANOVA with Tukey’s post-hoc test. Correlation analysis was conducted using Spearman’s rank correlation test. Data analysis was performed using Prism software version 7.0a (GraphPad, San Diego, CA).

Results:

SLE monocytes have increased glycolysis and OXPHOS compared to monocytes from healthy controls.

Recent evidence suggests that type I IFNs rewire cellular metabolism, enhancing glycolysis in dendritic cells and macrophages, and promoting activation through increased oxidative phosphorylation and fatty acid oxidation [18, 29, 30]. Given the role of IFNs in the pathology of SLE, we asked whether monocytes from SLE patients displayed any changes in metabolic activity compared to monocytes from healthy controls. Analyzing potential metabolic changes we found that monocytes from SLE patients have both increased OXPHOS (as measured by oxygen consumption rate (OCR)) and increased glycolysis (detected by increases in extracellular acidification rate (ECAR)) (Figure 1A–B). RNA-sequencing showed that ISGs were the top differentially expressed genes in SLE monocytes compared to those of healthy controls (HC) (Supplemental Figure 1). Paired analysis of SLE monocytes indicated that the higher expression of ISGs such as IFI27 detected in patient monocytes, the greater the changes in metabolic activity observed, suggesting IFNα exposure may be driving the changes in metabolism observed in SLE monocytes (Figure 1C). In line with this, treatment of healthy monocytes with IFNα (1000U/ml) for 24 hours drove similar changes in OCR and ECAR (Figure 1D&E).

Figure 1. SLE Monocytes show enhanced metabolic activity which can be mimicked by IFNα stimulation in healthy control monocytes.

(A) Oxygen consumption rate (OCR) and (B) extracellular acidification rate (ECAR) was compared between SLE and healthy control (HC) monocytes under basal conditions. Data in A&B is representative of at least 10 individual experiments and is presented as mean±S.D. (C) OCR and ECAR at baseline were compared between paired SLE patients with either high or low baseline levels of IFI27. Data is representative of at least 3 individual experiments and is presented as mean±S.D. (D&E) OCR and ECAR measurements were made in healthy control monocytes treated with IFNα (1000 U/ml) for 24 hours (n=11). Data is representative of individual experiments and is presented as mean±S.D. Data in the right-hand panels represents baseline values of paired samples stimulated with IFNα 24 hours (n=11). Statistical significance was determined using student’s (A-C) unpaired and (D) paired t-test *p<0.05, **p <0.01, ***p <0.001, ****p<0.0001; ns = not significant.

SLE patients have altered expression of IDH2.

We next quantified changes in expression of key proteins between HC (n=8) and SLE (n=12) monocytes critical to carbon metabolism and overall mitochondrial function using a quantitative mass spectrometry assay (termed Mitoplex) which quantifies the levels of 37 proteins key to carbon metabolism (Supplemental Figure 2A) [28]. CD14⁺ monocytes were isolated from HC or SLE patients and intracellular proteins extracted and analyzed by mass spectrometry. Isocitrate dehydrogenase 2 (IDH2), a rate limiting enzyme on the tricarboxylic acid (TCA) cycle, was significantly increased in SLE monocytes compared to monocytes from HC (Figure 2A, B). Levels of succinate dehydrogenase and superoxide dismutase also decreased albeit only weakly (supplemental Figure 2B). Parallel analysis of central chain metabolites from the same samples showed α-ketoglutarate levels (α-KG) were significantly increased (Figure 2C), in keeping with the function of IDH2 in converting isocitrate to α-KG on the TCA cycle. When levels of IDH2 were compared in patients with inactive (IAP, n=5, SLEDAI < 4) or moderately active/active (AP, n=7, SLEDAI >4) disease, we found that IDH2 levels were statistically significant only in patients with active disease (Figure 2D) and that stimulation of monocytes with IFNα could drive these increases (Figure 2E & F). α-KG can also be generated through glutaminolysis. However, no differences in expression of glutaminase 1 or 2 (GLS1/2) were detected between HC or SLE patient monocytes by qPCR (Supplemental Figure 3). When IDH2 levels were analyzed in PBMCs from HC and SLE patients by flow cytometry, we observed that the highest expression of IDH2 was in monocytes (Supplemental Figure 4A, B). As expected, SLE monocytes showed higher surface expression of the ISG, Siglec1 (Supplemental Figure 4C) which correlated positively with IDH2 levels (Spearman’s r=0.66, Supplemental Figure 4D). Importantly, treatment of whole blood samples from SLE patients (n=5) who had higher IDH2 expression levels at baseline with the JAK1/2 inhibitor baricitinib [31], decreased IDH2 expression (Supplemental Figure 5).

Figure 2. SLE patients have altered expression of isocitrate dehydrogenase 2 (IDH2) which is driven by IFNα.

(A, B) Targeted proteomic analysis demonstrates significant increase in expression of IDH2 in SLE monocytes (SLE; n=5) compared to monocytes from healthy controls (HC; n=5 (C) Metabolomic analysis demonstrates increased levels of intracellular α-KG in SLE monocytes (n=5) compared to HC monocytes (n=5); (D) Comparative proteomic analysis of IDH2 levels in monocytes from SLE patients with moderate to active disease (AP) (n=7), inactive disease (IAP) (n=5) or healthy controls (n=5); ((E-F) Overnight IFNα treatment of THP1 monocytic cells increased expression of IDH2 mRNA (E) and protein (F) as measured by qPCR and western blotting when compared to non-treated (NT) cells, respectively. Data is representative of >3 individual experiments.; (A, B, C, E, F) Data is presented as mean±S.D. and statistical significance was determined by student’s unpaired t-test and (D) one-way ANOVA. *p<0.05, **p <0.01, ***p <0.001, ****p<0.0001; ns = not significant.

IFN-induced changes in IDH2 and α-KG contribute to chronic ISG expression.

In addition to their known roles in energy production and metabolic processes, TCA cycle intermediates also possess additional functions as signaling molecules and cofactors for epigenetic enzymes [32]. To test whether exogenous α-KG could affect IFN-driven gene expression we incubated monocytes with a cell permeable analogue of α-KG, dimethyl α-KG, followed by stimulation with IFNα overnight (1000U/ml). Pre-treatment of primary monocytes (Figure 3A) or THP1 cells (Supplemental Figure 6A) with α-KG enhanced IFI27 expression. Importantly, inhibition of IDH2 using AGI-6780 resulted in reduced IFNα-driven gene expression, supporting a role for IDH2 in this response (Figure 3B). Succinate is a known competitive inhibitor of α-KG-dependent histone and DNA demethylases [33]. In line with α-KG being able to drive ISG expression we found that succinate inhibited IFNα-induced IFI27 expression (Supplemental Figure 6B). Interestingly, although no changes in GLS1 or 2 were observed in SLE monocytes (Supplemental Figure 3), BPTES, an allosteric inhibitor of glutaminolysis [34], could reverse IFN-induced IFI27 expression (Supplemental Figure 6C), indicating that altering α-KG levels in monocytes either through IDH2 or glutaminolysis can regulate ISG expression.

Figure 3. α-KG and IFNα pretreatment results in enhanced IFN-driven gene expression.

(A) Healthy control monocytes were untreated or pretreated with (A) 2μM cell permeable dimethyl-αKG followed by stimulation overnight with IFNα, as indicated. IFI27 expression was determined by qPCR; (B) THP1 cells were untreated or pretreated with 10μM AGI-6780 followed by stimulation overnight with IFNα, as indicated. IFI27 and ISG15 expression was determined by qPCR; (C) Trained immunity schematic whereby (D) healthy control monocytes are pulsed with PBS or IFNα (1000U/ml) for 24 hours, washed, restimulated with PBS or IFNα (1000U/ml) on day 5 for 24 hours (as indicated) and IFI27 expression determined by qPCR. Data in A, B & D is representative at least n=3 individual experiments and is presented as mean±S.D. Statistical significance was determined using One-Way ANOVA and Tukey’s multiple comparison test. **p <0.01, ***p <0.001, ****p<0.0001, ns = not significant; (E) Gene set enrichment analysis of RNAseq data from THP1 cells pulsed with IFNα for 24hours, washed and then rested for 5 days demonstrate that genes associated with the IFNα response are most highly enriched compared to non-treated control cells (n=2 for each group).

As epigenetic changes have been shown to contribute to trained immunity or innate immune memory responses in response to pathogen or DAMP exposure, we asked whether IFNα could also induce ‘memory’ responses in monocytes. We therefore set up a ‘trained immunity’ model whereby monocytes were pulsed with IFNα (1000U/ml) for 24 hours, washed twice and then rested for 4–5 days (Figure 3C). Analysis of ISG expression in these samples showed that 5 days after pulsing cells with IFNα stimulus, IFI27 levels were markedly elevated compared to unstimulated control or cells stimulated with IFNα overnight (Figure 3D, comparing green bars to black (unstimulated) and pink (IFNα, 24h)). Indeed, restimulation of these IFN-pulsed monocytes resulted in an even greater expression of IFI27 compared to either pulsed cells or cells treated with IFNα overnight (Figure 3D, comparing pink bars to green (IFNα pulsed) or purple (IFNα pulsed + IFNα 24h). We wondered next whether pulsing cells with IFNα could prime responses to other IFN-inducing stimuli. Interestingly, priming cells with IFNα resulted in enhanced responses to the STING ligand cGAMP in primary monocytes (Supplemental Figure 7A). As an important control, only minimal changes to NFκB-dependent genes TNFα or HIF1α were observed in our model of IFNα-trained immunity (Supplemental Figure 7B). RNAseq analysis of THP1s pulsed with IFNα for 24 hours confirmed that IFI27 and other genes associated with the IFNα response were highly upregulated in pulsed THP1s compared to unstimulated cells (Figure 3E and Supplemental Figure 8).

Epigenetic changes regulate ISG expression in monocytes.

Given that α-KG is an obligate cofactor for KMD6A and KDM6B, we assessed whether pulsing cells with IFNα altered the degree of H3K27me3 at ISG promoters (the target for KMD6A and KDM6B). Chromatin immunoprecipitation (ChIP) PCR showed that pulsing THP1 cells with IFNα drove a decrease in H3K27me3 at the promoter of IFI27 (Figure 4A) and additional ISGs (Supplementary Figure 9), in keeping with increased demethylase activity. Interestingly, H3K4me3 marks were increased at the promoter of IFI27 (Figure 4B), in keeping with previous reports that KDM6A/B can also enhance the level of activatory H3K4me3 marks in a demethylase-independent fashion [35]. Intracellular staining for H3K27me3 followed by flow cytometry analysis demonstrated a decrease in the level of H3K27me3 in healthy monocytes (CD14⁺) following overnight treatment with α-KG (Figure 4C), supporting our previous observation of α-KG-mediated increase in ISG expression (Figure 3A). Furthermore, intracellular analysis of H3K27me3 levels in PBMCs revealed that SLE monocytes (shown in Figure 4D) exhibited lower levels of H3K27me3 compared to healthy monocytes (Figure 4E&F).

Figure 4. Epigenetic regulation of ISG expression is altered by IFNα pretreatment and ISG expression can be reversed by inhibiting KDM6A/B in both SLE patient.

(A, B) THP1 cells trained with IFNα (1000U/ml) showed significant (A) decreased H3K27me3 marks and (B) increased H3K4me3 marks at the IFI27 promoter compared to naive THP1 cells; Data shown is from 3 independent experiments and is presented as mean±S.E.M. Statistical significance was determined using One-Way ANOVA and Tukey’s multiple comparison test, ****p<0.0001, ***p<0.001, *p<0.05, ns=not significant. (C) CD14+ Monocytes from healthy control were treated with α-KG (2μM) for 24 hours as indicated, H3K27me3 marks were quantified by Geometric Mean of H3K27me3-Alexa Fluor 488 positive area and are represented as mean±S.D., n=2. Statistical significance was determined using a one-tailed paired Student’s t test *p<0.05. (D-F) Level of H3K27me3 were analyzed in healthy controls (HC, n=4) and SLE (n=6) PBMCs by flow cytometry. (C) UMAP plot showing distribution of monocyte populations and (D) H3K27me3 levels in HC (top panel) and SLE cells (bottom panel). (E) Relative H3K27me3 expression in CD14⁺ positive monocyte population in HC and SLE (third panel).

GSK-J4, an inhibitor of KDM6A/B, reverses IFN-induced ISG expression.

We next assessed whether inhibiting KDM6A/B with GSK-J4, a specific inhibitor of KDM6A/B [36], could reverse trained immune responses to IFNα in monocytes. Having determined that 25μM of GSK-J4 was non-toxic to THP1 cells and could inhibit IFNα driven ISG expression (Supplemental Figure 10) we found that inhibiting KDM6A/B using GSK-J4 reduced ISG expression in HC monocytes pulsed with IFNα (Figure 5A). We next asked whether inhibition of KDM6A/B using GSK-J4 could reverse ISG expression in monocytes from SLE patients with moderate-high disease activity (SLEDAI ≥ 4). Importantly, GSK-J4 reduced baseline IFI27 (compare black and pink bars) and IFNα-driven IFI27 expression (compare green and purple bars) in SLE patient monocytes (Figure 5B). In vivo analysis in the pristane model of IFN-inducible SLE, also demonstrated that GSK-J4 treatment of mice 24 hours prior to analysis resulted in decreased ISG expression in monocytes recovered from the peritoneal cavity (Figure 5C).

Figure 5. ISG expression can be reversed by inhibiting KDM6A/B in both healthy control and SLE monocytes and in pristane-treated C57Bl/6 mice.

(A, B) Monocytes from (A) healthy controls and (B) SLE patients were pretreated with GSK-J4 (25μM) as indicated, followed by overnight treatment with PBS or IFNα (1000U/ml). ISG expression was measured by qPCR. Data is representative of individual experiments (n=5) and is presented as mean±S.D; (C) C57Bl/6 mice (n=5 per group) were treated with pristane (i.p) on day 0, followed by GSK-J4 (10μg/kg) or vehicle control on day 6. Mice were sacrificed on Day 7. qPCR was used to assess ISG expression in monocytes isolated from the peritoneal cavity of pristane-treated mice, treated with or without GSK-J4 (10μg/kg). Data is presented as mean±S.D; Statistical significance was determined using student’s t-test, *p<0.05, ns= not significant. (A-B) Statistical significance was determined using One-Way ANOVA and Tukey’s multiple comparison test or (C) student’s t-test, ****p<0.0001, ***p<0.001, *p<0.05, ns= not significant.

KDM6A/B inhibition reduces inflammation and kidney pathology in vivo.

To assess the effect of GSK-J4 on disease pathology in vivo, we used an inducible model of SLE whereby topical treatment of Balb/c mice with resiquimod (R848), a TLR-7/8 agonist, induces autoimmune disease [37] (treatment outlined in Figure 6A). As expected, resiquimod-treated mice displayed increased splenomegaly (Supplemental Figure 11A), serum creatinine levels (Figure 6B), elevated levels of anti-double stranded DNA antibodies (Figure 6C), and total levels of IgG1 and IgG2a (Supplemental Figure 11B and C, respectively). qPCR analysis of kidney RNA also demonstrated significantly higher expression of interferon-stimulated gene (ISG) in the resiquimod-treated mice compared to the control group (Figure 6D). Although GSK-J4 did not reduce splenomegaly in resiquimod-treated Balb/c mice (Supplemental Figure 11A), inhibition of KDM6A/B in this model resulted in a significant decrease in serum creatinine (Figure 6B), anti-dsDNA (Figure 6C), and IgG1 and IgG2a levels (Supplemental Figure 11B&C, respectively). Additionally, there was a significant reduction in ISG expression in kidney tissue (Figure 6D). Importantly, the deposition of C3 and IgG in kidney glomeruli was decreased, indicating a reduction in the progression of kidney nephritis (Figure 6E). Glomerular size and cellularity were also decreased in GSK-J4 treated mice (Figure 6F and G, respectively).

Figure 6. Inhibition of resiquimod-induced disease progression in Balb/c mice by GSK-J4.

(A) Schematic representation of the 35-day resiquimod/GSK treatment protocol performed on wild-type Balb/c mice (n=10 per group). (B) Measurement of mouse serum creatinine levels in vehicle control, resiquimod, and GSK+resiquimod treated mice. (C) Detection of anti-double stranded DNA (total IgG) through optical density (O.D.) measurement at 450nm in mouse serum from vehicle control, resiquimod, and GSK+resiquimod. (D) Quantification of Ifi27 expression in mouse kidney using qPCR analysis. (E) Immunohistochemical staining for glomerular immune complex and C3 deposits in renal sections from vehicle control, resiquimod, and GSK+resiquimod treated mice groups. PAS, IgG, and C3 staining results for renal sections. Images were taken at 40× (PAS) and 10× (IgG & C3) magnifications. (F) Glomerular size was quantified on at least 40 glomeruli from 3 different mice group. (G) Glomerular cellularity was determined by counting total nuclear cells in each glomerulus using light microscopy. In panels B-D and F-G, data is presented as mean±S.D. Statistical analysis was performed using One-Way ANOVA and Tukey’s multiple comparison test. ****p<0.0001, ***p<0.001, *p<0.05, ns= not significant.

Thus, taken together our results show that α-KG can positively affect ISG expression in monocytes and that IFN-priming of cells results in enhanced accessibility at the promoters of IFI27 and IFI44 as evidenced by reduced H3K27me3 repressive marks. Furthermore, inhibition of the α-KG-dependent histone demethylases KDM6A/6B using GSK-J4 resulted in reversal of ISG expression in both a mouse model of SLE and in SLE patient cells, suggesting histone modifications regulate IFN-inducible gene expression in the context of SLE. Overall, these findings highlight the therapeutic potential of GSK-J4 as shown by its ability to ameliorate pathology in murine SLE, including kidney dysfunction, autoantibody levels, and ISG expression.

Discussion

In this study we provide evidence that type I IFN contributes to enhanced expression of ISGs via epigenetic and metabolic changes in monocytes. We show that SLE monocytes have enhanced rates of glycolysis and oxidative phosphorylation, an effect that is also seen when monocytes from healthy control volunteers were stimulated with IFNα. Proteomic analysis reveals that SLE monocytes have increased levels of isocitrate dehydrogenase (IDH2) and its product, α-KG. In keeping with α−KG as a required cofactor for KDM family of histone demethylases, treatment of monocytes with a cell-permeable analogue of α-KG results in enhanced expression of the ISG IFI27, and decreased H3K27 trimethylation in monocytes from healthy controls, suggesting that increased α-KG in SLE monocytes may be contributing to enhanced histone 3 demethylation and removal of repressive marks. Similar to other examples of innate immune memory, pulsed treatment of cells with IFNα resulted in sustained expression of IFN-stimulated genes, and subsequent re-exposure to IFNα augmented ISG expression. This was accompanied by decreased H3K27 trimethylation (repressive) and increased H3K4 trimethylation (permissive) marks at the promoter of ISGs. Supporting epigenetic regulation of ISG expression in SLE, GSK-J4, a KDM6A/B-selective inhibitor, reversed ISG expression in IFNα-trained monocytes, and reduced constitutive levels of ISGs in SLE monocytes and in monocytes isolated from the peritoneal cavity of pristane-treated mice. Finally, KDM6A/B inhibition reduced inflammation and kidney pathology in a murine model of TLR7/8-induced autoimmunity. These findings underscore the potential of GSK-J4 in attenuating disease progression in SLE by suppressing the aberrant activation of ISGs.

Metabolic reprogramming is a dynamic process, facilitating functional reprogramming of immune cells and innate immune memory in response to detection of pathogens [38]. pDCs for example upregulate both glycolytic and oxidative phosphorylation pathways in response to type I IFNs in order to fuel the energetic demands of cytokine production [18, 30]. Our data is in line with this, showing that both glycolysis and OXPHOS rates are enhanced in IFNα-treated primary monocytes and that IFNα increases the level of IDH2 and α-KG in monocytes. As α-KG is a required cofactor for the histone demethylase family KDM, increased α-KG levels may drive epigenetic changes in response to IFNα exposure. In keeping with this, treatment of monocytes with a cell permeable analogue of α−KG enhanced ISG expression in response to IFNα. Previous reports support a proinflammatory role for IDH2. For example, deletion of IDH2 in mice reduces LPS-induced inflammation in both microglia and acute lung injury in an α−KG-dependent manner [39, 40]. Thus, the changes in IDH2 levels we observe in SLE monocytes may be contributing to altered chromatin accessibility and changes in gene expression. IDH2 is not the only enzyme regulating α-KG levels. Glutaminolysis can also drive increased α-KG. Conversion of glutamine – the most abundant amino acid in the body – to glutamic acid and then to α-KG requires glutamate dehydrogenase and glutaminase 1 and 2 (GLS1/2) [41]. Both GLS1 and 2 have been shown to be altered in CD4⁺ T cells from SLE patients. GLS1 is upregulated in SLE T cells and drives Th17 cell differentiation [42], whereas GLS2 is decreased, leading to enhanced reactive oxygen species (ROS) levels and reduced IL2 production [43]. However, in SLE monocytes we do not observe any increase in either enzyme, whilst our data demonstrates that IDH2 is an IFN-stimulated gene. However, inhibition of GLS1/2 with BPTES results in a decrease in IFN-stimulated gene expression, suggesting that multiple pathways may contribute to altered α-KG levels. Recent evidence also points to other metabolic enzymes, namely fumarate hydratase as being important in IFN induction and perturbed in SLE [44, 45], although the contribution of these specific pathways in epigenetic regulation of ISG promoters remains an ongoing area of investigation. Importantly, we observe that inhibition of IDH2 using AGI-6870 results in a reduction in ISG expression in THP1 cells, supporting a role for IDH2-driven αKG in this response. Interestingly, AGI-6870 has previously been found to inhibit wild-type IDH2-dependent metabolic reprogramming induced by Epstein-Barr virus [46]. ISG loci are known to be highly dynamic in response to IFN treatment. Genome wide studies have also shown that histone H4 acetylation is globally increased in SLE patient monocytes which is accompanied by persistent binding of the transcription factor IRF1 at the promoters for type I IFNs, thus supporting epigenetic rewiring as an important contributor to altered IFN production and responses [47, 48].

α-KG plays critical roles in a range of metabolic and cellular pathways: as an intermediate in the TCA cycle for energy metabolism, as a precursor of glutamine formation for amino acid synthesis, and as a co-factor for α-KG-dependent dioxygenases. Of the latter, the KDM family of histone targeting enzymes, KDM6A and KDM6B are positive regulators of gene expression as they remove H3K27me3 repressive marks at promoters. KDM6A and 6B (also known as UTX and JMJD3, respectively) have been shown to regulate inflammation and inflammatory disease through altering cytokine promoter accessibility [49–51]. Importantly they are upregulated in response to viral infection and contribute to inflammation and immunopathology [52]. Assessing chromatin remodeling at the promoters of ISGs we observed decreased H3K27me3 and increased H3K4me3 marks in IFNα-trained cells. Supporting a role for KDM6A/B in driving epigenetic changes in response to IFNα we found that the KDM6A/B inhibitor GSK-J4 reverses ISG expression in our model of trained immunity and reduces ISG expression in monocytes isolated from the peritoneal cavity of pristane treated mice. KDM6B has recently emerged as an important regulator of proinflammatory genes by demethylating repressive H3K27me3 marks at promoters of inflammatory genes such as IL6 [51, 53] [54]. In addition, fibroblasts from RA patients were also found to have increased expression of KDM6B, and reducing its expression in mice resulted in decreased inflammation in IL1-treated synovial fibroblasts via demethylation of H3K27me3 at target genes. In the same study, inhibiting KDM6B with GSK-J4 also ameliorated disease in the collagen-induced arthritis model [55, 56]. In SLE CD4⁺ T cells, KDM6B binding to the promoters of CD11a and CREMα resulted in decreased levels of H3K27me3 [57, 58]. KDM6A (also termed UTX) is an X-linked gene that has been shown to escape X inactivation [59], which has important implications for SLE. Interestingly it has recently been reported to be critical for transcriptional control of IFNγ in NK cells and thereby contribute to sex differences in NK responses [49]. In line with the importance of these enzymes in inflammation and SLE, our data shows that GSK-J4, a selective inhibitor for KDM6A/B, can reverse ISG expression in models of trained immunity ex vivo and reduce SLE pathology in vivo, using the resiquimod model of IFN-induced SLE. Given the importance of TLR7/8 in driving type I IFNs expression and SLE, the ability of GSK-J4 to reduce autoantibody induction, ISG expression and kidney pathology in this model supports a role for KDM6A/B in regulating IFN-induced responses and immunopathology in SLE. KDM6A/B inhibition using GSK-J4 has previously been reported to target innate immune cells and reduce the inflammatory cytokine storm observed in a murine model of sepsis [60–62]. Furthermore, GSK-J4 protects against central nervous system autoimmune disease in EAE [63]. In addition, GSK-J4 has also been reported to improve the renal damage in diabetic kidney disease model through inhibition of UTX activity [64]. This suggests that GSK-J4 plays a significant role in mitigating inflammation and the pathology associated with these disorders in animal models and underlines the importance of KDM6A/B as pivotal regulators of inflammation.

Overall, our study demonstrates a novel link between changes in metabolic flux and epigenetic regulation of ISG expression in SLE monocytes, a mechanism reflecting innate immune memory or trained immunity to type I IFN. More importantly it reveals KDM6A/B as potential novel therapeutic targets in IFN-driven disease.