Identification of a transcription factor network regulating anti-TNF mediated IL10 expression in human CD4+ T cells

Abstract CD4+ T cells are key players in immune-mediated inflammatory diseases (IMIDs) through the production of inflammatory mediators including tumour necrosis factor (TNF). Anti-TNF therapy has revolutionized the treatment of several IMIDs and we previously demonstrated that in vitro treatment of human CD4+ T cells with anti-TNF promotes anti-inflammatory IL-10 expression in multiple subpopulations of CD4+ T cells. Here we investigated the transcriptional mechanisms underlying the IL-10 induction by TNF-blockade in CD4+ T cells, isolated from PBMCs of healthy volunteers, stimulated in vitro for 3 days with anti-CD3/CD28 mAb in the absence or presence of anti-TNF. After culture, CD45RA+ cells were depleted before performing gene expression profiling and chromatin accessibility analysis. Gene expression analysis of CD45RA-CD4+ T cells showed a distinct anti-TNF specific gene signature of 183 genes (q-value < 0.05). Pathway enrichment analysis of differentially expressed genes revealed multiple pathways related to cytokine signalling and regulation of cytokine production; in particular, IL10 was the most upregulated gene by anti-TNF, while the proinflammatory cytokines and chemokines IFNG, IL9, IL22, and CXCL10 were significantly downregulated (q-value < 0.05). Transcription factor motif analysis at the differentially open chromatin regions, after anti-TNF treatment, revealed 58 transcription factor motifs enriched at the IL10 locus. We identified seven transcription factor candidates for the anti-TNF mediated regulation of IL-10, which were either differentially expressed or whose locus was differentially accessible upon anti-TNF treatment. Correlation analysis between the expression of these transcription factors and IL10 suggests a role for MAF, PRDM1, and/or EOMES in regulating IL10 expression in CD4+ T cells upon anti-TNF treatment.


Introduction
Immune-mediated inflammatory diseases (IMIDs) is a term used to represent a clinically diverse group of diseases that can affect multiple tissues including bone and joint (rheumatoid arthritis and spondyloarthritis), skin (psoriasis and atopic dermatitis), bowel (inflammatory bowel disease), lung (asthma), and central nervous system (multiple sclerosis).These conditions are currently incurable, reduce quality of life and are associated with increased mortality.
CD4+ T cells play a key role in the initiation and/or perpetuation of chronic inflammatory disorders, with different CD4+ T cell subsets being pathogenic in several disease settings.Activated CD4+ T cells contribute to the pathogenesis of IMIDs via effector functions including activation of APCs [1], influencing Ig class switching [2] and by producing multiple inflammatory cytokines including IFNγ, IL-17, and tumour necrosis factor (TNF) [3,4].
Targeting pathogenic disease mechanisms via cytokine blockade has revolutionized the treatment of IMIDs, particularly following the pioneering application of TNF inhibitors in the treatment of rheumatoid arthritis (RA) [5].The application of TNF blockade is now a standard of care for several IMIDs including Crohn's disease, psoriasis and spondyloarthritis [5].Currently, there are multiple biologics that inhibit or modulate the effects of TNF [6], with adalimumab (ADA), a fully human anti-TNF monoclonal antibody, being one of them.
Anti-TNF biologics can neutralize membrane-bound and soluble TNF, although to different extents; some inhibitors can also induce reverse signalling via membrane-bound TNF.Furthermore, the Fc-regions of anti-TNF mAbs can mediate antibody-dependent cellular cytotoxicity and complementdependent cytotoxicity [6] or promote direct interaction between monocytes and CD4+ T cells [7].Thus, TNF inhibitors can act on multiple levels by exerting distinct biological effects on immune cells.
In addition, a connection between anti-TNF treatment and IL-10 production has been proposed in multiple IMIDs [8]: patients with juvenile idiopathic arthritis [9], psoriasis [10], RA [11], or Crohn's disease [12] were shown to have higher serum or cellular levels of IL-10 after treatment with anti-TNF.The source of IL-10 upon TNF blockade is not always clear in these studies as several immune cell types from the innate (macrophages and dendritic cells) or adaptive (T cells) systems produce IL-10 [13].Indeed, anti-TNF has been shown to modulate innate immune cell function [14] and induce a macrophage regulatory phenotype characterized by an increased production of IL-10 [15].TNF blockade can also affect the interaction between innate cells and T cells, leading to the expansion of suppressive Tregs [7] or priming CD4+ T cells for IL-10 production [8,16].In vitro induction or maintenance of IL-10 production in CD4+ T cells upon TNF inhibition does not require the presence of naïve CD4+ T cells [16] and can also be independent of APC interaction as we have shown previously [16][17][18].
IL-10 production by most human CD4+ T cell subsets requires ERK activation [19], which suggests a common molecular mechanism within the different subsets; however, IL10 expression also depends on binding of other transcription factor including the transcription factors specific protein 1 [20], SP3 [21], CCAAT/enhancer binding protein-β (C/ EBPβ) [22], IFN-regulatory factor 1, and STAT3 [23].The list of genes involved in the regulation of IL-10 in human CD4+ T cells is expanding, with MAF suggested as a universal transcription factor for the regulation of IL-10 production [19,[24][25][26].MAF can act synergistically or interact with other transcription factors including aryl hydrocarbon recepto [27] and PR domain zinc finger protein 1 (BLIMP-1) [28] leading to a positive regulation of IL-10 production.BLIMP-1, which is encoded by the PRDM1 gene, has also been shown to positively regulate IL-10 production in CD4+ T cells independently of MAF [29][30][31].Regulation of IL-10 production by CD4+ T cells is thus not linear and relies on the interplay of transcription factors.
The aim of this study was to further elucidate the molecular mechanism by which anti-TNF regulates IL-10 expression in CD4+ T cells.We investigated the transcriptomic and chromatin changes in human CD4+ T cells following treatment with anti-TNF mAb.This led to the identification of a transcription factor module consisting of MAF, PRDM1, and EOMES, which may underlie the modulation of IL-10 expression by TNF inhibition.

Cell isolation
Peripheral blood samples were obtained from healthy adult volunteers.Peripheral blood mononuclear cells were isolated by density gradient centrifugation using Lymphoprep (Axis-Shield, Oslo, Norway).CD4+ T cells were isolated by magnetic-activated cell sorting (MACS) using the CD4+ T cell Isolation Kit II (Miltenyi Biotec, Bergisch-Gladbach, Germany).Average purities were 98% for CD4+ T cells.After 3 days of culture, CD4+ T cells were depleted of CD45RA+ cells using CD45RA MicroBeads (Miltenyi Biotec).The study was approved by the Bromley Research Ethics Committee (06/Q0705/20), and written informed consent was obtained from all participants.

RNA-seq
RNA was isolated from 500 000 CD4+ CD45RA-cells after 3 days of stimulation in the presence or absence of 1 ug/ml adalimumab.Total RNA (RNA integrity number > 9) was used to construct RNA sequencing libraries.Starting with 100 ng total RNA, cDNA synthesis and amplification were performed using a NEBNext Ultra II Directional RNA Library Prep Kit for Illumina (NEB, MA, USA) following the manufacturer's protocol.Dual-indexed libraries were prepared using the NEBNext ® Multiplex Oligos for Illumina ® (Dual Index Primers Set 1) following the manufacturer's protocol.The libraries were sequenced at Genewiz Ltd (Genewiz, Germany) on an Illumina NovaSeq 2 × 150 bp at an average of 35 M reads per sample.The reads were mapped against the genome using STAR on default parameters against the human hg38 reference genome release 32.Differential gene expression analysis (q-value < 0.05) was performed using DESeq2.Functional enrichment analysis of differential gene lists to identify significantly enriched biological functions and pathways was performed using gprofiler [32] using standard settings.Raw data can be accessed at GSE216688.

ATAC-seq
Nuclei from 50 000 CD4+ CD45RA-cells stimulated for 3 days in the presence or absence of ADA were isolated and ATAC-seq was performed according to a published protocol [33].Paired-end libraries (50 cycles) were prepared according to the ATAC-seq protocol (see above).To obtain the open chromatin regions, reads were aligned to hg38 using Bowtie v2.

Statistical analysis
Statistical testing was performed with GraphPad Prism 9.0 (GraphPad, San Diego, CA, USA).Data sets were tested using the appropriate non-parametric test as indicated in figure legends.The correlation matrix between the expression of IL10 and the seven transcription factor candidates reports the two-tailed Spearman r correlation computed for every combination.P-values or q-values (P-values corrected for multiple testing, as specified in figure legends) < 0.05 were considered statistically significant.

In vitro treatment with anti-TNF confers a distinct gene signature onto memory CD4+ T cells characterized by differential expression of multiple cytokines
We previously demonstrated that in vitro anti-TNF treatment led to a significant increase in IL-10 production 72 h post stimulation both at RNA and protein levels [11,[16][17][18].To elucidate the molecular mechanism behind the anti-TNF-mediated induction of IL-10, we performed combined RNA-seq and ATAC-seq.We focused on memory CD4+ T cells as we identified that IL-10 production is restricted to that population (Fig. 1A).CD4+ T cells were stimulated with anti-CD3 and anti-CD28 (aCD3/CD28) mAb for 72 h in the absence or presence of 1 μg/ml of the anti-TNF drug adalimumab (ADA), without PMA and ionomycin restimulation.We then enriched for memory CD4+ T cells by depleting CD45RA+ cells, followed by isolation of RNA and nuclei to perform paired RNA and ATAC-sequencing (Supplementary Fig. S1).We confirmed that there was an anti-TNF-mediated increase in IL-10 production (using PMA and ionomycin restimulation) by the CD4+ CD45RA-T cells, defined from now on as memory T cells, which we used as input for sequencing (Fig. 1B and C).
Principal component analysis (PCA) of the RNA-seq data showed that donor and gender were the dominant sources of variation in the dataset, with anti-TNF treatment having a relatively subtle effect (Fig. 2A).Differential gene expression analysis revealed that anti-TNF treatment conferred a unique gene signature in memory CD4+ T cells, with 67 upregulated and 115 downregulated genes (q-value < 0.05).Notably, IL10 was found to be the top upregulated gene by adalimumab (Fig. 2B and C and Table 1).Functional enrichment analysis of the differentially expressed genes revealed enrichment for multiple gene sets associated with cytokine signalling and regulation of cytokine production pathways (Fig. 2D).In particular, we found that CD4+ T cells that were stimulated in the presence of anti-TNF for 72 h showed a significant decrease in the expression of the inflammatory cytokines and chemokines IFNG, IL9, IL22, and CXCL10, in parallel to the significant increase in IL10 expression (Fig. 2E).Analysis of gene changes at 24 h post anti-TNF treatment showed a similar significant decrease in the inflammatory cytokine and chemokine genes IFNG, IL9, IL22, and CXCL10, as well as  in IL17F and IL5, with an increase in IL2, but not IL10 expression ( Supplementary Fig. S2A and B), consistent with our previous data that showed that the differential expression of IL-10 became manifest at later timepoints only [16,18].Of interest, we found that BHLHE40, a MAF repressor that when deficient in CD4 Th1 cells leads to decreased IFN-γ and increased IL-10 [35], as well as its antisense transcript BHLHE40-AS1, which has been suggested as a modulator of a proinflammatory cytokine signature [36], were significantly downregulated at 72 h, but not 24 h, post anti-TNF treatment ( Supplementary Fig. S2C).Taken together, these data indicate that in vitro anti-TNF treatment alters the cytokine gene profile of memory CD4+ T cells leading to a downregulation of inflammatory cytokine gene expression at both 24 h and 72 h, and a differential upregulation of immunomodulatory IL10 at 72 h.

Anti-TNF treatment changes chromatin accessibility revealing a putative transcription factor network controlling IL-10 expression
To identify how anti-TNF treatment mechanistically alters the regulome of CD4+ T cells leading to increased IL-10 expression at 72 h, we performed Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) on the same samples.We identified 152 595 peaks (i.e.regions of accessible chromatin) within our samples.PCA of accessible chromatin regions shared between donors showed a similar result to our RNA-seq analysis with donor and gender being the largest sources of variation in the dataset, followed by anti-TNF treatment (Fig. 3A).Despite this variability, we identified 2,113 differential peaks (Poisson P-value < 0.0001 and fold change ≥ 1.5) in anti-TNF treated samples compared to no drug control (heatmap shown in Supplementary Fig. S3A).
We performed motif enrichment analysis (P-value < 0.05) of known transcription factors at differentially accessible (DA) open chromatin regions (OCRs) in anti-TNF-treated cells at 72 h post stimulation in each donor independently.This analysis identified 76 enriched motifs upon anti-TNF treatment that were shared by all donors (Supplementary Fig. S3B and Table 2).Since our interest was to determine which transcription factors may regulate IL-10 production upon anti-TNF treatment, we focussed our analysis on the shared transcription factor motifs that mapped to the IL10 locus, which reduced our candidate pool down to 58 transcription factors and fusion proteins (included by the motif finding software available from ChIP-seq data) (Fig. 3B and Table 3).We further refined our candidate list by focussing on the common transcription factors mapping at the IL10 locus whose expression we could measure in our RNA-seq data, which therefore excluded fusion proteins (Fig. 3C).Finally, to identify potential anti-TNF dependent IL-10 regulators, we Table 1: list of differentially expressed (DE) genes between no ADA and ADA-treated cells at 72 h post stimulation (q-value < 0.05) Identification of a transcription factor network regulating induced IL-10, 2024, Vol.assessed which common transcription factors binding at the IL10 locus were either differentially expressed (DE genes) or had their gene locus differentially accessible (DA OCRs) after anti-TNF treatment (Fig. 3D).This analysis revealed seven transcription factors mapping to the IL10 locus that were either differentially expressed (upregulated or downregulated) and/or differentially accessible upon treatment with anti-TNF.Of these seven, the transcription factor MAF, which was shown previously to be a transcriptional regulator of IL10, was both differentially upregulated and differentially accessible upon anti-TNF treatment.In addition, KLF5, EOMES and PRDM1 were found to be differentially accessible upon ADA treatment, whilst KLF2 was differentially upregulated (Fig. 3E and F).We also identified ELK1 and HLF as being differentially downregulated after anti-TNF treatment, suggesting a potential inhibitory role (Fig. 3E).When we evaluated these transcription factors at 24 h we found only KLF2 differentially upregulated after anti-TNF, however, both loci for MAF and PRDM1 were differentially accessible in all donors ( Supplementary Fig. S3C and D).
We next performed correlation and linear regression analysis of gene expression at 72 h posttreatment with anti-TNF and found a highly positive correlation between MAF, PRDM1, or EOMES with IL10, as well as a positive correlation between these three transcription factors (Fig. 4A and Supplementary Fig. S4) suggesting potential synergy; in contrast, either no or a negative correlation was found between KLF2, KLF5, ELK1, and HLF with IL10 (Fig. 4A and Supplementary Fig S4).Furthermore, we found that the motifs for AP-1/MAF, PRDM1, and EOMES (Fig. 4B) could bind at multiple OCRs either at distinct or shared sites on the IL10 locus (Fig. 4C).Taken together, these data show that anti-TNF leads to chromatin remodelling in CD4+ T cells whilst transcription factor motif analysis at the differential OCRs after treatment with anti-TNF indicates a putative transcription factor network that can regulate IL10 expression.

Discussion
Since our initial observations that anti-TNF treatment led to an increase in anti-inflammatory IL-10 in T cells from healthy donors and in patients with inflammatory arthritis [11], we have worked towards understanding the molecular mechanism underlying the anti-TNF dependent regulation of IL-10.Our previous work showed that anti-TNF treatment in vitro maintained a long-term increased production of IL-10 in multiple T-cell populations [16,17], which could indicate transcriptional regulation.We initially hypothesized that the transcription factor IKZF3 (encoding Aiolos), which is upregulated upon anti-TNF treatment [11], had a direct role in regulating the anti-TNF effect on IL-10.Detailed analysis however demonstrated that IKZF3/Aiolos is associated with, but not sufficient to drive IL-10 [18].Thus, in this report, we sought to uncover the mechanism underlying the maintenance of the IL-10 program in CD4+ T cells by anti-TNF.
Despite a high level of donor and gender variation, we were able to identify anti-TNF-dependent transcriptional changes in bulk memory CD4+ T cells.Reassuringly, the most upregulated gene in our dataset was IL10.Furthermore, we found a number of pro-inflammatory cytokines and chemokines significantly downregulated upon anti-TNF treatment.In particular, IFNG was the most highly expressed inflammatory cytokine that was significantly downregulated in all donors upon anti-TNF treatment.It should be noted that we previously did not observe such a strong effect of anti-TNF on IFN-γ at the protein level [17]; this may be due to the effect of PMA and ionomycin restimulation in those assays, which has been shown to increase IFN-γ production [37].
In our attempt to gain mechanistic insight into the molecular regulation of IL10 by anti-TNF, we performed motif discovery analysis at differential OCRs upon treatment with anti-TNF to define a putative transcription factor network.Of our final seven transcription factor candidates we focussed on those that were either differentially regulated, or whose loci were more accessible after anti-TNF and those whose expression positively correlated with IL10.
Perhaps unsurprisingly, the prime candidate we found for the anti-TNF-dependent regulation of IL-10 was MAF.Multiple studies have demonstrated its correlation with IL10 regulation either directly or synergistically with other transcription factors, including PRDM1 [25,26].PRDM1, which encodes the transcriptional regulator BLIMP-1, has been described itself as a requirement for IL-10 production in multiple cell populations including mucosal Tregs [30], cytotoxic T lymphocytes [38] and IL-10-producing effector Th cells [28].
There is less evidence of EOMES directly regulating IL-10, although it has been described as a lineage-defining transcription factor in human IL-10-producing Tr1-like cells either independently [39] or synergistically with Blimp-1 [40].Indeed, our own data showed that not only the expression of MAF, PRDM1, and EOMES was positively correlated with IL10 expression, but also with each other, suggesting that IL-10 regulation could be the result of the synergistic effect of multiple transcription factors.Interestingly, and in line with our findings, a recent in vitro study using mouse CD4+ T cells showed that c-Maf and BLIMP-1 not only work synergistically but can also target each other and act as positive modulators of Il10 gene expression, while at the same time acting as negative regulators of Ifng by directly binding to the loci encoding these cytokines in Th1 effector cells upon differentiation with IL-12 and IL-27 [41].While there are differences between the mouse and human systems, the conserved nature of these transcription factors would suggest a similarity in their transcriptional regulation of cytokine production.
Our data showed gender as a potential source of variation in our transcriptomic and regulome data, upon treatment with anti-TNF.There has been an increased interest in the differences found in immune responses and immune homeostasis between genders, which has been partially correlated with disease susceptibility and response to treatment [42].This highlights the importance of considering gender when studying immunological responses.It would be interesting and relevant to explore how gender affects the T-cell response to activation, both in the presence or absence of anti-TNF and whether there is a direct effect on IL-10 regulation, production or even a variation in the kinetics of its modulation.
In conclusion, the work presented here reveals changes caused by anti-TNF treatment in CD4+ memory T cells at both transcriptional and regulome levels.Our data support the concept that anti-TNF alters the immune response of memory T cells by negatively modulating inflammatory mediators, while positively modulating anti-inflammatory IL-10.We provide evidence for a putative transcription factor module regulating IL10 expression upon anti-TNF treatment consisting of MAF, which showed the strongest evidence in our data, together with PRDM1 and EOMES.These findings reveal potential targets for novel therapeutic strategies that could aim at stabilizing the immune-modulating effect of anti-TNF treatment on human CD4+ T cells.

Figure 1 :
Figure 1: IL-10 production is restricted to memory T cells and significantly increased by anti-TNF at 72 h.(A) Representative flow cytometry plot showing frequencies of IL-10+ cells among CD4+ T cells (left) and frequencies of CD45RA+ and CD45RO+ cells within CD4+ IL-10+ cells (right) ex vivo or 72 h post stimulation with aCD3/CD28 mAb, in the absence or presence of ADA.(B) Representative flow cytometry plot showing frequencies of IL-10 producing cells in CD4+ CD45RA-T cells ex-vivo (ex vivo) or 72 h post stimulation with aCD3/CD28 mAb, in the absence or presence of adalimumab (ADA).(C) Cumulative plots showing the frequencies of populations from (B) ex vivo or 72 h post stimulation with aCD3/CD28 mAb, in the absence (filled symbol) or presence (open symbol) of ADA in the cells that were taken forward for RNAseq and ATACseq (Wilcoxon matched-pairs signed rank test, n = 4).

Figure 2 :
Figure 2: anti-TNF treatment confers a specific gene signature characterized by increased IL10 and decreased inflammatory cytokine genes.(A) PCA plots of gene expression showing clustering of samples based on donor, gender and anti-TNF treatment.(B) Heatmap of differentially expressed (DE) genes between ADA and no ADA-treated cells at 72 h post stimulation (q-value < 0.05).(C) Volcano plot showing differentially expressed genes (coloured symbols) between no ADA and ADA-treated cells at 72 h.Threshold lines at q-value < 0.05 and +1.5 and -1.5-fold change.(D) Top 10 significant biological functions and pathways (q-value < 0.05) from functional enrichment analysis of DE genes with a 1.5-fold change cutoff.(E) Cumulative plot showing normalized read counts of DE cytokine and chemokine genes at 72 h post stimulation with aCD3/CD28 mAb, in the absence (filled symbol) or presence (open symbol) of ADA.

3 , No. 1 7Figure 3 :
Figure 3: anti-TNF treatment modulates chromatin accessibility and reveals a putative transcription factor network regulating IL10.(A) PCA plots of chromatin accessibility showing clustering of samples based on donor, gender, and anti-TNF treatment.(B) Venn diagram showing the overlap of common transcription factor motifs from Fig. S3B significantly enriched (P < 0.05) in ADA-treated cells at 72 h post stimulation with transcription factor motifs mapping at the IL10 locus.(C) Heatmap showing normalized read counts (log2 transformed) of the common motifs mapped at the IL10 locus (intersection from B) in CD45RA-T cells 72 h post stimulation with aCD3/CD28 mAb, in the absence (left panel) or presence (right panel) of ADA.(D) Venn diagram showing the overlap of common transcription factor motifs significantly enriched and mapping at the IL10 locus, the common differentially accessible (DA) OCRs and differentially expressed (DE) genes from ADA-treated cells at 72 h post stimulation.(E) Cumulative plot showing normalized read counts of highlighted genes in D at 72 h post stimulation with aCD3/CD28 mAb, in the absence (filled symbol) or presence (open symbol) of ADA.(F) Cumulative plot showing from each donor (round symbols) averaged normalized peak intensities ('tags') per 10 million reads (RP10M) for differential OCRs at the gene loci from D 72 h post stimulation with aCD3/CD28 mAb, in the absence or presence of ADA.

Figure 4 :
Figure 4: MAF, PRDM1, and EOMES form a putative transcription factor module regulating IL10 expression upon anti-TNF treatment.(A) Heatmap showing Spearman correlations of IL10 gene expression and transcription factors from Fig. 3E at 72 h posttreatment with anti-TNF.(B) DNA binding motif sequence logo and q-value (Benjamini) for MAF, PRDM1, and EOMES from motif enrichment analysis at differentially accessible peaks from ATAC-seq of CD4+ CD45RA-T cells 72 h post stimulation.(C) Representative example of OCR at the IL10 locus 72 h post stimulation with aCD3/CD28 mAb, in the absence (top track) or presence (bottom track) of ADA; layered H3K27Ac track (ENCODE) added for reference.OCRs corresponding to ATAC peaks are highlighted.Differential peaks (indicated by asterisks) in ADA treated T cells (Poisson P-value = 0.0001 and fold change = 1.5 cutoffs) and mapping of MAF, PRDM1 and EOMES binding motifs enriched at differential peaks are shown.

Table 2 :
list of common TFs motifs significantly enriched (P < 0.05) at differentially accessible OCRs in ADA-treated cells at 72 h post stimulation

Table 3 .
list of common TFs motifs significantly enriched (P < 0.05) in treated cells at 72 h post stimulation and mapping at the IL10 locus