Blimp-1 and c-Maf regulate  Il10 and negatively regulate common and unique proinflammatory gene networks in IL-12 plus IL-27-driven T helper-1 cells

Version Changes

Revised. Amendments from Version 1

Summary of major differences to the new submission in answer to Reviewers’ comments: Introduction has been modified to clarify the questions being addressed in our study, in context of the published literature. We have now rearranged and expanded our text to address this. The transcription factors that regulate the differentiation and function of Th1 cells producing Ifng are well established, including STAT molecules and T-bet, however, whether these transcription factors regulate Il10 expression as part of the differentiation pathway of Th1 cells has been difficult to decipher (Gabrysova et al., 2014, new reference number to be applied). Whether transcription factors such as Prdm1 or Maf, which positively induce Il10expression, do so as part of the differentiation pathway of Th1 cells, or alternatively, simultaneously negatively regulate proinflammatory cytokines in Th1 cells, therefore not contributing to Th1 differentiation but instead reinforcing a regulated Th1 response, is unclear (Gabrysova et al., 2014; Gabrysova et al., 2018, new reference numbers to be applied). At the end of the paragraph where we introduce the publication of Kuchroo et al., 2020, (original reference 13) which showed that Blimp-1 and c-Maf regulation Il10 expression in IL-27-differentiated “Tr1” cells, we have added another sentence. The role of Prdm1 and Maf in regulating Il10 and proinflammatory cytokines in IL-10-producing Th1 cells, however, has not been reported. In the Results, we have simplified Figures 2 and 5, and clarified the text, pointing out that no Supplementary Figures can be included. In the Discussion, we have now indicated that the effects of transcription factors on regulation of Il10 expression are undoubtedly context-specific and may vary according to the T effector cell response. Finally, in the Discussion, we have added a paragraph to reference the reported roles of Blimp-1 and c-Maf in regulation of Il10 expression in human T cells.

Abstract

Background

CD4 ⁺ Th1 cells producing IFN-γ are required to eradicate intracellular pathogens, however if uncontrolled these cells can cause immunopathology. The cytokine IL-10 is produced by multiple immune cells including Th1 cells during infection and regulates the immune response to minimise collateral host damage. In this study we aimed to elucidate the transcriptional network of genes controlling the expression of Il10 and proinflammatory cytokines, including Ifng in Th1 cells differentiated from mouse naive CD4 ⁺ T cells.

Methods

We applied computational analysis of gene regulation derived from temporal profiling of gene expression clusters obtained from bulk RNA sequencing (RNA-seq) of flow cytometry sorted naïve CD4 ⁺ T cells from mouse spleens differentiated in vitro into Th1 effector cells with IL-12 and IL-27 to produce Ifng and Il10, compared to IL-27 alone which express Il10 only , or IL-12 alone which express Ifng and no Il10, or medium control driven-CD4 ⁺ T cells which do not express effector cytokines . Data were integrated with analysis of active genomic regions from these T cells using an assay for transposase-accessible chromatin with sequencing (ATAC)-seq, integrated with literature derived-Chromatin-immunoprecipitation (ChIP)-seq data and the RNA-seq data, to elucidate the transcriptional network of genes controlling expression of Il10 and pro-inflammatory effector genes in Th1 cells. The co-dominant role for the transcription factors, Prdm1 (encoding Blimp-1) and Maf (encoding c-Maf) , in cytokine gene regulation in Th1 cells, was confirmed using T cells obtained from mice with T-cell specific deletion of these transcription factors.

Results

We show that the transcription factors Blimp-1 and c-Maf each have unique and common effects on cytokine gene regulation and not only co-operate to induce Il10 gene expression in IL-12 plus IL-27 differentiated mouse Th1 cells, but additionally directly negatively regulate key proinflammatory cytokines including Ifng, thus providing mechanisms for reinforcement of regulated Th1 cell responses.

Conclusions

These data show that Blimp-1 and c-Maf positively and negatively regulate a network of both unique and common anti-inflammatory and pro-inflammatory genes to reinforce a Th1 response in mice that will eradicate pathogens with minimum immunopathology.

Introduction

CD4+ T helper 1 (Th1) cells are critical in controlling infection by production of the cytokine IFN-γ, which upregulates the expression of MHC-class II molecules on antigen-presenting cells (APC) thus enhancing their capacity to present antigen to activate CD4 ⁺ T cells, and also activates macrophages to kill intracellular pathogens ¹ . However, uncontrolled Th1 responses can cause immunopathology ² . IL-10 is a regulatory cytokine that has been widely shown to limit immunopathology particularly during infection and intestinal responses to pathobionts ^{3– 6} and mutations in IL-10 or the IL-10R result in inflammatory bowel disease (IBD) in humans ^{7, 8} . Most cells of the immune system can produce IL-10 to limit over-exuberant immune responses and pathology ^{5, 6, 9} . Th1 cells have been shown to be the critical source of IL-10 to limit responses to pathogens such as Toxoplasma gondii ¹⁰ and Leishmania major ¹¹ and thus avoid immunopathology. The cytokine IL-27, which has been reported to regulate the immune response by multiple mechanisms ^{12– 14} , promotes IL-10 production by effector Th1 CD4 ⁺ T cells in vivo in response to the malaria parasites Plasmodium chabaudi ¹⁵ and L. major ¹⁶ infections, providing a critical mechanism for protection from severe immunopathology. The transcription factors that regulate the differentiation and function of Th1 cells producing Ifng are well established, including STAT molecules and T-bet, however, whether these transcription factors regulate Il10 expression as part of the differentiation pathway of Th1 cells has been difficult to decipher ⁹ . Whether transcription factors such as Prdm1 or Maf, which positively induce Il10 expression, do so as part of the differentiation pathway of Th1 cells, or alternatively, simultaneously negatively regulate proinflammatory cytokines in Th1 cells, therefore not contributing to Th1 differentiation but instead reinforcing a regulated Th1 response, is unclear ^{9, 17} .

Several transcription factors have been shown to regulate IL-10 in T cells ^{13, 17} . Both common and cell-specific transcriptional mechanisms are in place to tightly regulate the expression of Il10 and proinflammatory gene expression in T cells to ensure a controlled immune response to pathogens and/or other insults ^{5, 6, 18– 20} . Since transcription factors have multiple gene targets this raises the question as to whether known transcription factors that positively regulate Il10 may simultaneously negatively regulate proinflammatory cytokine expression in T cells, thus driving a controlled response to control immune responses to pathogens and pathobionts to limit host damage. For example, the transcription factor c-Maf has been shown to induce Il10 expression directly across multiple T cell subsets both in vitro and in vivo ^{5, 6, 13, 18– 20} whilst also acting as a negative regulator of Il2 ¹⁷ and Th17 responses ²¹ . Blimp-1, encoded by the Prdm1 gene, has also been shown to induce IL-10 ^{13, 19, 20} , although originally described as a global regulator of T cell homeostasis and differentiation in vivo ^{22– 26} .

Recently Kuchroo et al. ¹³ , systematically identified regulators for Il10 and highlighted Prdm1 and Maf as two central nodes of the Il10 regulatory circuits that cooperatively promoted IL-10 production in ‘Tr1 cells’ differentiated in vitro with IL-27 and through gain-of-function in Th1, Th2, Th17, and Treg cells ¹³ . IL-27-driven ‘Tr1 cells’ lacking both Prdm1 and Maf (DKO) showed an almost complete loss of Il10 expression. Moreover, expression of several transcription factors shown previously to be important for Il10 expression, including Fosl2, Hif1a, Hlx, and Notch1 ²⁷ , was found to be abrogated in IL-27-driven ‘Tr1 cells’ from CD4 ⁺ T cells deficient in both Prdm1 and Maf, showing the latter to be upstream major controllers of Il10 gene regulaton ¹³ . Moreover, DKO ‘Tr1 cells’ showed a unique reduction in chromatin accessibility in co-inhibitory receptor gene loci such as Ctla4, Pdcd1 (PD-1), Tigit, Havcr2 (Tim-3), suggesting that Prdm1 and Maf have complementary but indispensable roles in regulating Tr1 at the transcriptional level and reinforcing the expression of negative immune regulators. The role of Prdm1 and Maf in regulating Il10 and proinflammatory cytokines in IL-10-producing Th1 cells, however, has as yet not been reported ^{9, 17} .

In this study we applied computational analysis of gene regulation derived from temporal profiling of gene expression clusters integrated with analysis of active genomic regions in CD4 ⁺ Th1 effector cells differentiated with IL-27 plus IL-12, which express Il10 together with proinflammatory cytokines. Gene expression and active genomic region analysis data was compared to T cells differentiated in IL-27 alone (named “Tr1 cells” ¹³ ), which express Il10 but no proinflammatory cytokines , or conversely IL-12 alone, which express proinflammatory cytokines but not Il10, all compared to medium control, which do not express cytokines. The aim was to elucidate the transcriptional network of genes controlling expression of Il10 and pro-inflammatory effector genes in Th1 cells, and identify transcription factors that not only induced Il10, and additionally negatively regulated Th1 proinflammatory gene expression, and were therefore not part of the Th1 differentiation pathway. We show that the transcription factors Blimp-1 and c-Maf each have unique and common effects on cytokine gene regulation and not only co-operate to induce Il10 gene expression in IL-12 plus IL-27 differentiated Th1 cells, but additionally directly negatively regulate key proinflammatory cytokines including Ifng, thus providing mechanisms for reinforcing the regulation of Th1 cell responses. Thus, Blimp-1 and c-Maf positively and negatively regulate a network of both unique and common anti-inflammatory and pro-inflammatory genes to reinforce a Th1 response that will allow eradication of pathogens with minimum immunopathology.

Methods

Animals

Mice were bred and maintained under specific pathogen free conditions in accordance with the Home Office UK Animals (Scientific Procedures) Act 1986. Age-matched male or female mice were used for experiments. Maf ^fl/fl mice were provided by M. Sieweke and C. Birchmeier (Max Delbrück Centre for Molecular Medicine, Germany) ²⁸ and backcrossed to C57BL/6J for 10 generations and then crossed to Cd4 ^Cre mice to generate Maf ^fl/fl Cd4 ^Cre mice as described in 17. Prdm1 ^fl/fl mice were purchased from the Jackson Laboratory (Stock Number 008100) ²⁹ , and further backcrossed to C57BL/6J for four generations and then crossed to Cd4 ^Cre mice to generate Prdm1 ^fl/fl Cd4 ^Cre mice. Prdm1 ^fl/fl Maf ^fl/fl Cd4 ^Cre and Prdm1 ^fl/fl Maf ^fl/fl control mice were generated in-house by crossing Maf ^fl/fl Cd4 ^Cre with Prdm1 ^fl/fl Cd4 ^Cre mice. All mouse breeding was performed under strict care and husbandry ensuring no discomfort to the mice. Breeding was carried out in accordance with UK Home Office regulations, under Project License, O’Garra P5AF488B4, 30 Apr 18 | Amended: 10 Jan 20 | Expired: 29 Apr 23, and were approved by The Francis Crick Institute Ethical Review Panel before each submission to the UK Home Office. This study adhered to the ARRIVE guidelines ³⁰ .

Naïve CD4 ⁺ T cell sorting and in vitro helper T cell differentiation

For each experiment, spleens were obtained after humane killing of mice by either placing the animal into a secure chamber and filling it gradually with carbon dioxide until the animal was unconscious and until death was confirmed and then followed by cervical dislocation; or by cervical dislocation of the neck, depending on how many mice needed to be humanely killed. Spleens from a total of 5–10 age-matched mice per genotype were homogenized and incubated with unconjugated rat anti-mouse antibodies against B220 (RA3.6B2, DNAX), MHC class-II (M5/114, eBioscience) and CD8 (C291.2.43, DNAX). CD4 ⁺ T cells were then negatively enriched using magnetic beads (BioMag, Qiagen). Live naïve CD4 ⁺CD62L ⁺CD44 ^loCD25 ^- T cells were then sorted to over 95% purity using the following antibodies: CD4 (RM4-5, e450), CD8 (53-6.7, FITC), CD62L (MEL-14, PE-Cy7), CD44 (IM7, PE) and CD25 (PC61.5, APC) (all from eBioscience); and propidium iodide (final concentration 2µg/ml, Sigma) on either a MoFlo XDP or Influx flow cytometer (Beckman Coulter, Inc.). Sorted naïve CD4 ⁺ T cells were then plated at 500,000 cells/well in flat-bottom 48-well plates and stimulated with plate-bound anti-CD3 (5µg/ml, 2C11, Harlan) and soluble anti-CD28 (2µg/ml, 37.51, Harlan) for up to 4 days in the presence of no polarizing cytokines for Medium control, IL-12 (rmIL-12p70, 5ng/ml, eBioscience), IL-12+IL-27 (rmIL-12p70, 5ng/ml, eBioscience; rmIL-27, 25ng/ml, R&D); or IL-27 (rmIL-27, 25ng/ml, R&D). All cells were cultured in conditioned RPMI (BE12-702F, Lonza) supplemented with 10% (v/v) heat-inactivated FCS (Gibco), 100U/ml Pen-Strep, 2mM L-glutamine, 1mM Sodium pyruvate, 10mM HEPES (all Lonza) and 0.05mM 2-Mercaptoethanol (Sigma) in a humid incubator at 37°C with 5% carbon dioxide. For each experiment three wells were differentiated per condition to give technical triplicates from the pool of naïve CD4 ⁺ T cells per genotype.

Quantitative RT-PCR

RNA was extracted from in vitro differentiated T-helper cells using the QIAShredder and RNeasy Mini Kit, or RNeasy Micro kit, both with on-column DNase digestion, according to the manufacturer’s instructions (Qiagen). Eluted RNA was then reverse transcribed using a High Capacity cDNA Reverse Transcription kit (Applied Biosystems) plus RNasin (Promega) according to the manufacturer’s instructions, followed by RNaseH (Promega) treatment for 30 min at 37°C. High-Capacity cDNA Reverse Transcription kit (Applied Biosystems) was used to convert RNA into cDNA. Samples were incubated for: 10 min 25°C, 2 hr 37°C, 5 min 85°C in a thermal cycler (Vertiti Thermo Cycler, Applied Biosystems). Residual RNA was digested by RHase H incubation (final concentration 0.03U/μl, Invitrogen) for 30 min at 37°C. Reverse transcribed cDNA was then diluted to 5ng/μl using nuclease-free water (Ambion) and stored at -80°C. TaqMan™ Assay system (Applied Biosystems) was used for RT-qPCR-analysis, reaction mix/primer probes are summarised below. Reactions were carried out in 96-well plates (Applied Biosystems) on either a 7900HT or QuantStudio 3 RT-qPCR machine (Applied Biosystems). For each experiment, RT-qPCR was always performed to confirm the deletion of either Maf, Prdm1 or both. All genes were analysed relative to the housekeeping gene hypoxanthine phosphoribosyltransferase 1 (encoded by the Hprt) gene. Delta Ct (ΔCt) was calculated by taking the difference between the Ct value of the gene of interest and the Ct for Hprt in a given sample, which was then inputted into the following equation (1.8^- ΔCt)*10^5 to give relative gene expression. For consistency, in the Applied Biosystems qPCR software the Ct threshold was manually set to 0.25 with automatic baseline threshold activated for all experiments and primer probes. cDNA was then analysed for the expression of specific genes on a 7900HT ABI, QS3 or QS5 real-time PCR system, using the TaqMan Universal Master Mix II – no UNG and the following TaqMan mouse probes (all from Applied Biosystems): Il10 (mm00439616_m1), Ifng (mm01168134_m1), Tbx21 (mm00450960_m1), Hprt (mm03024075_m1). All expression levels were normalised to the internal housekeeping gene Hprt and calculated as 1.8 ^{-(Ct Hprt−Ct gene)} x10 ⁵. Reactions are run on a Thermal Cycler, Veriti model (Applied Biosystems).

Statistical analysis

All figure legends show the number of independent biological experiments performed for each analysis and replicates. For PCR analysis, two-tailed unpaired t-test with 95% confidence interval was used for statistical analysis. All statistical analysis, apart from the sequencing data analysis was carried out with GraphPad Prism 8 (RRID:SCR_002798) software (GraphPad, USA) (*=p≤0.05; **=p≤ 0.01; ***=p≤ 0.001, ****=p≤0.0001). Analyses for sequencing data were performed with R Project for Statistical Computing (RRID:SCR_001905) version 3.6.1 and Bioconductor (RRID:SCR_006442) version 3.9 unless otherwise stated. Error bars and n values used are described in the figure legends.

RNA-seq of in vitro differentiated T-helper cells

RNA was extracted using the QIAShredder and RNeasy Mini Kit, or RNeasy Micro kit, both with on-column DNase digestion, according to the manufacturer’s instructions (Qiagen). RNA-seq libraries were made with total RNA equally pooled from the technical triplicate wells within an independent biological experiment, using the Illumina Stranded TruSeq Library preparation kit V2 and unique multiplexing indexes, according to the manufacturer’s instructions (Illumina). All libraries were then sequenced using the HiSeq 4000 system (Illumina) with paired-end read lengths of 100bp and at least 25 million reads per sample.

ATAC-seq of in vitro differentiated T-helper cells

ATAC-seq samples from in vitro differentiated T-helper cells were prepared as outlined in 31. For each sample, 50,000 cells were lysed in cold lysis buffer containing 10mM Tris-HCl, pH 7.4, 10mM NaCl, 3mM MgCl2, 0.1% Nonidet™ P40 substitute (all Sigma) and the nuclei incubated for 2 hours at 37°C with 50μl of TDE1/TD transposase reaction mix (Illumina). Tagmented DNA was then purified using the MinElute kit (Qiagen) and amplified under standard ATAC PCR conditions: 72°C for 5 min; 98°C for 30s and thermocycling at 98°C for 10s, 63°C for 30s and 72°C for 1 min for 12 cycles. Each 50μl PCR reaction consisted of: 10μl Tagmented DNA, 10μl water, 25μl NEBNext High-Fidelity 2x PCR Master Mix (NEB), 2.5μl Nextera XT V2 i5 primer and 2.5μl Nextera XT V2 i7 primer (Illumina). NexteraXT V2 primers (Illumina) were used to allow larger scale multiplexing, these sequences were ordered directly from Sigma (0.2 scale, cartridge) and diluted to 100μM with 10mM Tris-EDTA buffer, pH8 (Sigma) and then to 25μM with DEPC-treated water (Ambion) for use in the reaction. Following amplification, ATAC-seq libraries were cleaned up using 90μl of AMPure XP beads (Beckman Coulter) and two 80% Ethanol washes whilst being placed on a magnetic plate stand, before being eluted in 1mM (0.1x) Tris-EDTA buffer, pH8 (Sigma) diluted with DEPC-treated water (Ambion). ATAC-seq libraries were then checked on the TapeStation/BioAnalyser (Agilent) before being sequenced on the HiSeq 4000 system (Illumina), with paired-end read lengths of 50bp and at least 50–80 million uniquely mapped reads per sample.

RNA-seq: pre-processing and quality control ( in vitro CD4+ T cell datasets)

Paired end RNA-seq reads were quality controlled and adapters were trimmed using skewer (RRID:SCR_001151) software version 0.2.2 ³² with the following parameters: "-m pe -q 26 -Q 28 -e -l 30 -L 100", specifying the relevant adapter sequences. Reads were then aligned to mm10 genome and the GENCODE reference transcriptome version M22 using STAR (RRID:SCR_004463) software version 2.7.1 ³³ , excluding multi-mapping reads by setting the parameter “outFilterMultimapNmax” to 1. In order to increase read mapping to novel junctions the parameter “twopassMode” was set to "Basic". Raw gene counts were retrieved using QoRTs software version 1.1.8 ³⁴ , specifying the “stranded” parameter for the in vitro CD4 ⁺ T cell datasets due to the nature of the library preparation. Normalized read counts were retrieved using DESeq2 (RRID:SCR_015687) version 1.24.0 ³⁵ and rlog transformed in order to visualize gene quantifications.

RNA-seq kinetics analysis of in vitro CD4+ T cells

Differentially expressed genes (DEG) at any given time point for IL-12, IL-12+IL-27, and IL-27 compared to medium control were obtained using DESeq2 (fold change >=1.5 and Benjamini-Hochberg adjusted p value<0.05), resulting in a total of 2,300 DEG. Next, the gene expression values of these DEG were subjected to k-means clustering using a k=9; where the optimal k was obtained using the R library “factoextra” (RRID:SCR_016692) (factoextra 2017). The expression values of the 2,300 DEG were standardized per gene (row z score) and plotted in a heatmap. The mean expression and 90% c.i. were obtained for these clusters and plotted.

Transcription factors correlating with Il10 expression

A gene was considered to be coding for a transcription factor if it was present in at least two of the following references ^{36– 38} , or Ingenuity Pathway Analysis (RRID:SCR_008653) (IPA) database (genes annotated as either "transcription regulator" or "ligand-dependent nuclear receptor") (QIAGEN Redwood City, www.qiagen.com/ingenuity).

Genes encoding for transcription factors that have expression patterns correlating with Il10 gene expression were analysed (absolute Pearson's correlation >0.7) from Days 1 to 4 in all conditions. A linear regression model was fitted for the top 9 transcription factors positively correlating with Il10 expression.

Genome-wide differential footprint detection with BaGFoot software

To identify potential transcription factors underlying the gene expression changes occurring between Day 2 and Day 3, we applied the BaGFoot software on our ATAC-seq data ³⁹ , using all ATAC-seq peaks identified in each treatment condition at Day 2 and Day 3. BaGFoot predicts these changes by searching for TF-binding motif matches in regions with altered ATAC-seq insertion patterns between the two days. We used 318 motifs of class A and B quality in the HOCOMOCO (RRID:SCR_005409) database v11 ⁴⁰ . BaGFoot does not consider replicates for the analysis, thus we performed two pair-wise comparisons (using all biological replicates) for each condition and calculated the average changes in accessibility and footprint-depth. Results are displayed as bagplots, using a fence of factor 2. We identified TFs with potentially altered binding between Day 2 and Day 3 by identifying the outliers of the multivariate distribution, as assessed by the Mahalanobis distance of each TF to the multivariate distribution ³⁹ . The statistical significance of these distances was tested using a Chi-square distribution followed by a Benjamini-Hochberg correction for multiple-testing, as recommended by BaGFoot, and shown as tables.

Differential gene expression analysis of in vitro CD4 ⁺ T cells: Cd4 ^Cre-mediated deletion of transcription factors versus floxed controls

The DEG of CD4 ⁺ T cells with Cd4 ^Cre-mediated deletion of Prdm1, Maf, or both Prdm1 and Maf against their corresponding floxed controls were obtained for all conditions (Medium, IL-12, IL-12+IL-27, and IL-27) using DeSeq2 with the default thresholds (Benjamini-Hochberg adjusted p value<0.1), this resulted in 208, 561, and 802 for cells with Cd4 ^Cre-mediated deletion of Prdm1, Maf, and both Prdm1 and Maf, respective.

Singular Value Decomposition analysis and biological interpretation

Singular Value Decomposition (SVD) analysis was performed as in 17 on each set of samples that shared a genetic background: a) Prdm1 ^fl/fl Cd4 ^Cre was analysed with Prdm1 ^fl/fl, b) Maf ^fl/fl Cd4 ^Cre with Maf ^fl/fl, and c) Prdm1 ^fl/fl Maf ^fl/fl Cd4 ^Cre with Prdm1 ^fl/fl Maf ^fl/fl. Prior to the SVD analysis, the rlog-normalized gene counts were "centered" by subtracting the mean expression per gene. The average values of the right-singular vectors, which relate the association of each sample to a component, were plotted as bar-plots.

For each right-singular vector three linear models were fitted: 1) a full linear model containing the variables for the differentiation condition and the genotype, 2) a reduced model containing only the differentiation condition, and 3) a reduced model containing only the genotype. In order to identify the association of each component with the condition and/or genotype the Akaike Information Criterion (AIC) score was calculated and an analysis of variance (ANOVA) with a Chi-squared-test was performed between the full model (1) versus the reduced ones (2 and 3). The component capturing the Cd4 ^Cre-mediated transcription factors deletion was chosen using the following criteria:

• a) The component where the AIC of the reduced genotype model is lower than AIC of the full mode (line-plot)
• b) The component with the smallest p value in the ANOVA (heatmap). Only statistically significant values (BH adjusted p value <0.05) were plotted for visualisation.
• c) The component in which the average values of the right-singular vectors diverge in sign for Cd4 ^Cre-mediated transcription factors deletion vs. floxed controls shown as histograms.

The left-singular vectors, which relate the contribution of a gene to a component, were segregated between positive and negative values, and each set was subjected to k-means clustering ( k=2). The genes belonging to the most positive and negative clusters were selected for further examination and they are referred in the text as the "SVD components associated with Cd4 ^Cre-mediated deletion".

For each genetic background the standardised expression values (row z score) of genes belonging to the component capturing the Cd4 ^Cre-mediated transcription factors deletion are shown in the heatmaps.

To further dissect the genes mostly affected by the Cd4 ^Cre-mediated transcription factors deletion a matrix containing the fold-changes of the three genetic backgrounds ( Prdm1, Maf, and Prdm1xMaf) was created and subjected to K means clustering (k=7).

Data annotation

Gene Ontology (GO) ^{41, 42} (RRID:SCR_002811) enrichment was assessed using the R package “topGO” (RRID:SCR_014798) ⁴³ . The top 100 GO terms of "biological processes'' were further synthesized using REViGO (RRID:SCR_005825) ⁴⁴ with allowed similarity=0.4. The top 10 GO terms were shown for annotation.

ATAC-seq: pre-processing and quality control of in vitro CD4+ T cells

Paired end ATAC-seq reads were quality controlled and adapters were trimmed using Skewer software version 0.2.2 ³² with the following parameters: "-m pe -q 26 -Q 30 -e -l 30 -L 50", specifying "CTGTCTCTTATACAC" as reference adapter sequence to remove. Quality controlled reads were then aligned to mm10 genome using BWA-MEM (RRID:SCR_010910) ⁴⁵ with (Picard toolkit 2018 (RRID:SCR_006525)) and SAMtools (RRID:SCR_002105) 1.3.1 ⁴⁶ was used to discard discordant alignments and/or with low mapping qualities (mapQ<30). In order to account for transposase insertion, reads were shifted +4bp in the forward and -5bp in the reverse strand; moreover, read-pairs that spanned >99bp were excluded from further analyses as they would span nucleosomes ³¹ .

Identification of open chromatin sites

MACS2 (version 2.1.1) (RRID:SCR_013291) was used to identify ATAC-seq peaks using the following parameters: "parameters --keep-dup all --nomodel --shift -100 --extsize 200; q-value < 0.01", in order to identify enrichment of Tn5 cutting sites ⁴⁷ .

Differentially accessible site detection in vitro CD4+ T cells

In order to identify open chromatin sites that differed in accessibility between Cd4 ^Cre-mediated transcription factor deletion and floxed controls, DiffBind (RRID:SCR_012918) software version 2.0.2 ⁴⁸ was used with the following parameters: "dba.count:minOverlap=0, score= DBA_SCORE_RPKM, bRemoveDuplicates=FALSE, bUseSummarizeOverlaps= TRUE; dba.analyze: method=DBA_DESEQ2, bFullLibrarySize=T" for each condition and genetic background. An ATAC-seq peak was considered to represent remodelled chromatin if the absolute fold-change>1.5 and FDR<0.05.

Identification of c-Maf and Blimp-1 putative binding sites

c-Maf ChIP-seq raw fastq files were obtained from GSE40918 ²¹ and Blimp-1 ChIP-seq raw fastq files were obtained from GSE79339 ⁴⁹ and GSE66069 ⁵⁰ . Trimmomatic version 0.36 was used for quality control and trim adapter sequences using the following parameters: "HEADCROP:2 TRAILING:25 MINLEN:26" ⁵¹ . Trimmed reads were aligned to mouse genome mm10 with Bowtie (RRID:SCR_005476) 1.1.2 ⁵² with the parameters: "y -m2 --best --strata -S". MACS2 2.1.1 was used with default parameters to identify ChIP-seq peaks, and peaks with a q-value<0.01 were defined as statistically significant binding sites. For each transcription factor, a final peak set was generated from the union of the statistically significant binding sites identified in each biological replicate (c-Maf) or GEO dataset (Blimp-1). This resulted in 45,727 ChIP-seq binding sites for c-Maf and 16,893 binding sites for Blimp-1. CRUNCH suite ⁵³ was used to infer the c-Maf motif, as the source dataset provided biological replicates and a suitable input control. On the other hand, the motif of Blimp-1 was taken from HOCOMOCO database v11 ⁴⁰ as two distinct ChIP-seq datasets were used, neither with biological replicates, thus not suitable for analysis with CRUNCH. These identified motifs were used as input for FIMO software ⁵⁴ the sequences underlying the ATAC-seq peaks were scanned for motif-matches in order to identify further putative binding sites of c-Maf and Blimp-1.

Visualisation of genome browser tracks

"bamCoverage" from DeepTools (RRID:SCR_016366) 2.4.2 was used to normalize ATAC-seq data to RPKMs and the R package "ggbio" (RRID:SCR_003313) was used to visualize the genome browser tracks ⁵⁵ . The CNS sites marked have the following coordinates ( Table 1).

Table 1. Coordinates correspond to mm10 genome.

Gene	Chromosome	start	end	CNS	Citation
Il10	1	130999687	130999976	-20	19
Il10	1	131010529	131010907	-9	19
Il10	1	131015400	131015593	-4.5	19
Il10	1	131019438	131019696	-0.5	19
Il10	1	131025957	131026170	6.45	57
Prdm1	10	44444389	44444891	14	19
Prdm1	10	44459445	44459716	-1	19
Prdm1	10	44459873	44460126	-1.5	19
Prdm1	10	44460425	44460609	2	19
Ifng	10	118435228	118435850	-6	58
Ifng	10	118419035	118419610	-22	58
Ifng	10	118406839	118407520	-34	58
Ifng	10	118458481	118459017	18	58
Ifng	10	118460275	118460853	20	58
Ifng	10	118470550	118471067	29	58
Maf	8	115707132	115707487	-0.5	59
Maf	8	115707694	115708065	-1	59

Integration of ATAC-seq, ChIP-seq, motifs, and RNA-seq from in vitro CD4+ T cells. ChIP-seq-identified binding sites of c-Maf and Blimp-1 were filtered using the ATAC-seq peaks in order to obtain those binding sites that were biologically relevant to our in vitro CD4 ⁺ T cells datasets. At this stage, each ATAC-seq had assigned an overlapping ChIP-seq peak, a motif, or none. A gene was assigned to an ATAC-seq peak based on distance proximity, if the peak was within +/- 3kb of the gene body coordinates, using the R package “ChIPSeeker” (RRID:SCR_021322) ⁵⁶ . Thus, a gene was a direct target of c-Maf and/or Blimp-1 if said gene was annotated with an ATAC-seq peak containing a putative binding site from either transcription factor (identified by motif or ChIP-seq). To quantitatively rank abundance of c-Maf or Blimp-1 binding sites in each gene, we applied the same approach as before in 17, additionally the likelihood of c-Maf or Blimp-1 to regulate a gene was calculated using the Binding and Expression Target Analysis (BETA) (RRID:SCR_005396) software, with the following parameters: "plus -g mm10 --da 0.5 --df 1 -c 1".

Direct targets of c-Maf and/or Blimp-1 are depicted in the gene regulatory networks. For all networks, Maf and Prdm1 nodes were added, in order to visualize target genes of Blimp-1 and c-Maf. These gene regulatory networks show the integration of all these "omic" datasets and were generated using the "igraph" R package ⁶⁰ . Each node represents a gene, and the size of a node represents the left-singular vectors obtained from the SVD analysis, thus relating the effect on expression the Cd4 ^Cre-mediated deletion of Prdm1 and/or Maf had on a gene. The edges depict the relationship of a target gene with c-Maf and Blimp-1, the thickness of the edge shows the likelihood of a gene being a target of either c-Maf or Blimp-1 as assessed by the BETA software. The colouring of the edge shows if the target gene has a c-Maf (green), Blimp-1 (pink), or c-Maf and Blimp-1 (blue) binding site assigned. For visualization purposes, the size of the maximum left-singular vector was fixed to be equal to the 2 ^nd highest; additionally, the labels of gene names were only added for the 50 most affected genes according to the SVD analysis. Scores used to generate the networks are available in Supplementary Table 7 as Underlying data ³⁰ .

Results

Expression of Prdm1 and Maf correlates with Il10 expression in Th1 cells

We previously showed that T cell-specific deletion of Maf resulted in the maximal reduction of IL-10 production by Th1 cells in vivo as compared to other T cell subsets ¹⁷ . However, since this effect was incomplete, we set out to identify additional transcription factors that positively regulate Il10 in Th1 cells. To achieve this, we first analysed changes in temporal gene expression in vitro in naïve CD4 ⁺ T cells stimulated with anti-CD3 and anti-CD28 as described in the Methods, and differentiated these over time with IL-27 plus IL-12 into CD4 ⁺ Th1 effector cells which express Il10 together with proinflammatory cytokines such as Ifng, and compared to T cells differentiated in IL-27 alone (named “Tr1 cells” ¹³ ), which express Il10 but no proinflammatory cytokines , or conversely IL-12 alone, which express proinflammatory cytokines including Ifng but not Il10, all compared to medium control, which do not express cytokines. The combination of IL-12 and IL-27 has been shown in vivo to be required for maximal levels of IFN-γ and IL-10 production by CD4 ⁺ Th1 cells ^{12, 15, 16} . Cells were cultured under these different conditions for 4 days and sampled at each time point for RNA-based next-generation sequencing (RNA-seq). Cells clustered distinctly for the most part according to time point of differentiation ( Figure 1a, Figure 1b), with the principal component 1 separating days 1 and 2 from days 3 and 4 ( Figure 1b; Supplementary Table 1 in Underlying data ³⁰ ).

Figure 1. Temporal gene regulation by IL-27 in CD4 ⁺ T cells is accompanied by clusters of candidate transcription factors that correlate with Il10 expression, including Blimp-1 and c-Maf.

RNA-seq analysis of flow cytometry sorted naïve CD4 ⁺ T cells activated in vitro with anti-CD3 and anti-CD28 antibodies and differentiated in the presence of Medium (no cytokines), IL-12, IL-12+IL-27, or IL-27 from Day 1 to Day 4. a, Principal component analysis (PCA) showing the two most dominant variables, time point (PC1: Day 1-2 versus Day 3-4) and differentiation condition (PC2: presence versus absence of IL-27 during differentiation). b, Unsupervised hierarchical clustering of a pair-wise Spearman correlation of samples encompassing CD4 ⁺ T cells differentiated in vitro in the presence of Medium, IL-12, IL-12+IL-27, or IL-27 from Day 1 to Day 4. c, Heatmap visualization of differentially expressed genes per condition compared to Medium (fold change >=1.5 and BH adjusted p value<0.05), partitioned into 9 clusters using k-means clustering. The values above the heatmap between parentheses show the number of differentially expressed genes at each time point. d, Gene expression profiles depicting the mean gene expression ±90% confidence intervals (c.i.) across all genes for clusters 3, 2 and 6 accompanied by mean gene expression for representative cytokines Il10, Ifng and Il21 respectively. e, All transcription factors annotated in the mouse genome positively correlating (Pearson’s r>0.7) and negatively correlating (Pearson’s r<0.7) with Il10 expression in in vitro CD4 ⁺ T cells differentiated in the presence of Medium, IL-12, IL-12+IL-27, or IL-27 from Day 1 to Day 4. Data from n=2 biological replicates.

Co-regulated clusters of gene expression were revealed using k-means clustering ( Figure 1c; Supplementary Table 2 in Underlying data ³⁰ ). Clusters 3, 2 and 6 each expressed effector T cell cytokines genes, including Il10, Ifng and Il21, respectively ( Figure 1c). Expression of Il10 (Cluster 3) was seen in IL-12+IL-27 or IL-27 driven T cells, peaking at days 3 and 4 of culture, although other genes within this Cluster 3 were also increased similarly in Th1 cells driven by IL-12 alone ( Figure 1c and Figure 1d). By contrast, expression of Ifng within Cluster 2, was maximal in IL-12+IL-27 driven Th1 cells from days 2–4 of culture. Ifng showed delayed induction in IL-12 alone driven Th1 cells, and a small increase in IL-27 alone driven T cells by day 2, which then decreased with time to the levels seen in medium control cultures ( Figure 1c and d). Expression of Ifng clustered with expression of the transcription factor Jun ( Figure 1c). Collectively the expression of the genes in Cluster 6 was induced by IL-12+IL-27 and IL-27 alone and to a lesser extent in IL-12 driven Th1 cells and was maximal from days 1–4 of culture ( Figure 1c and d). Expression of Il21 was also observed within this Cluster 6 and was maximally induced in IL-12 + IL-27 and IL-27 alone driven T cells by day 1 to day 4 of culture, however IL-12 driven Th1 cells only started to express Il21 by days 3 and 4 of culture as compared to medium controls ( Figure 1c and d). Our findings suggest a role for IL-21 in autocrine expansion of T effectors rather than as a regulator of Il10 as has been previously suggested ^{61, 62} , since it is produced maximally by Th1 cells, which do not produce IL-10.

Il10 expression clustered with the transcription factors Nfatc2, Hif1a and Nfil3 (Cluster 3), while Il21 expression clustered with the transcription factors Prdm1, Maf and Batf (Cluster 6), all of which were highly expressed upon culture with IL-12+IL-27 or IL-27 alone ( Figure 1c). Regardless, the transcription factors showing the highest positive correlation with Il10 expression were first Prdm1, then Id2, Asb2, Hlx, Nfatc2 and Maf ( Figure 1e), in keeping with the literature ¹³ . Of transcription factors previously reported to regulate Il10 expression, only expression of Prdm1 and Maf, was significantly increased under IL-12+IL-27 and IL-27 alone conditions only, reaching maximal levels when Il10 expression was observed at days 3 and 4 and showing the strongest correlation with Il10 expression, while their expression was not observed under IL-12 or medium alone conditions ( Figure 2a). Importantly expression of Batf, previously suggested to regulate Il10 in Th2 cells ⁶³ , which was not revealed as correlating with Il10 expression in the analysis in Figure 1e, did not correlate with Il10 expression under IL-12+IL-27 and IL-27 alone conditions, but rather was maximally expressed on days 1 and 2, rapidly diminishing by days 3 and 4, while it’s expression increased with time in IL-12 alone differentiated Th1 cells which did not express Il10 ( Figure 2b), therefore suggesting a broader function than the regulation of Il10 ( Figure 2b). Other transcription factors that have been associated with Il10 gene expression including Hif1a and Nfil3 ^{13, 17} , which although expressed under IL-12+IL-27 and IL-27 alone conditions, showed less of a correlation with Il10 expression, and were found to increase with time in Th1 cells differentiated with IL-12 only which do not express Il10, therefore again suggesting a broader role for these transcription factors in Th1 differentiation ( Figure 2b).

Figure 2. Blimp-1 and c-Maf strongly positively correlate with IL-27 induced Il10 expression in differentiating naïve CD4 ⁺ T cells.

RNA-seq analysis of flow cytometry sorted naïve CD4 ⁺ T cells activated in vitro with anti-CD3 and anti-CD28 antibodies and differentiated in the presence of Medium (no cytokines), IL-12, IL-12+IL-27, or IL-27 from Day 1 to Day 4. a, Mean gene expression of Prdm1 and Maf, both transcription factors positively correlating with Il10 expression, accompanied by linear regression of Prdm1 and Maf against Il10 expression across all conditions and timepoints. b, Mean gene expression profiles (top panel) of transcription factors previously associated with Il10 expression. Linear regression (bottom panel) of these transcription factors against Il10 expression across all conditions and timepoints.

To further investigate global changes in transcriptional activity in CD4 ⁺ naïve T cells cultured as above, we used the assay for transposase-accessible chromatin plus sequencing (ATAC-seq) to reveal functionally active genomic regions at days 2 and 3, timepoints, which marked key transcriptional changes during the differentiation of CD4 ⁺ naïve T cells into Th1 cells (cultured as in Figure 1a). The ‘bivariate genomic footprinting’ (BaGFoot) software ³⁹ was applied to the ATAC-seq data, to detect global changes in transcription factor binding activity (genome-wide) occurring between day 2 and day 3 under the different conditions. Differences in binding activity are assessed by BaGFoot software by quantifying the differences in Tn5 transposition within a transcription factor motif, by measuring the ‘footprint depth’ ( Figure 3, y axis) and ‘flanking accessibility’ ( Figure 3, x axis) and comparing these metrics between timepoints. A transcription factor bound to chromatin has a high footprint depth and a high flanking accessibility.

Figure 3. Blimp-1 and c-Maf have increased differential binding between Day 2 and Day 3 of culture only under IL-12+IL+27 and IL-27 cytokine driving conditions.

BaGFoot analysis of transcription factors with putative genome-wide changes in chromatin binding between Day 2 and Day 3, as assessed by Tn5 insertion patterns (obtained with ATAC-seq) in in vitro activated and differentiated CD4 ⁺ T cells in the presence of Medium (no cytokines), IL-12, IL-12+IL-27, or IL-27. Presented as the change in change in the ‘flanking accessibility’ of motifs (ΔFA) plotted against the ‘footprint depth’ (–ΔFPD); wedges along axes indicate direction and degree of change in transcription factor binding between Day 2 and Day3. Dark shading in the “bagplots” indicates a region with no change in transcription factor binding patterns, and light shading indicates a region in which most non-significant minor changes in binding occur. Transcription factors with significant change in binding are found as outliers outside the “bagplot”. A table of P values is provided for the outliers of the “bagplot” and indicates the statistical confidence assigned to the differential binding between Day 2 and Day 3 of a transcription factor in each CD4 ⁺ T cell differentiation condition.

Only IL-27 and IL-12+IL-27 cultured T cells showed increased transcriptional activity for Prdm1 (higher activity at day 2 vs day 3) and Maf (higher activity at day 3 vs day 2), reinforcing a role for these transcription factors in regulating transcriptional programs in IL-12+IL-27 and IL-27 cultured T cells. On the other hand, the AP-1 family members, e.g. Jun and Fos, showed increased transcriptional activity at day 3 across all cytokine differentiation conditions including medium control ( Figure 3) suggesting a broader role in T helper cell activation/differentiation. Batf transcriptional activity was only evident in cells cultured in medium alone or IL-12; and Stat 3, 4 and 5 transcriptional activity was only detected under IL-12 conditions, which were not accompanied by Il10 gene expression, again implicating these transcription factors in broader roles in Th1 cell activation/differentiation. Detection of STAT activity and a transcriptome more similar to T cells cultured in medium or in IL-12 alone at days 1 and 2 ( Figure 1a), suggests that Th1 cells cultured with IL-12, which do not express Il10 were temporally and qualitatively different with respect to global transcriptional activity, to Th1 cells cultured with IL-12+IL-27 and IL-27 alone, where both conditions lead to Il10 expression, although proinflammatory cytokine expression was only observed in IL-12+IL-27 driven Th1 cells.

Reduction or abrogation of Il10 expression in IL-12+IL-27 and IL-27-driven CD4+ T cells upon deletion of Prdm1, Maf or both Prdm1 and Maf, is accompanied by increased Ifng expression

Since expression of Il10 in IL-12+IL-27 and IL-27-driven CD4 ⁺ T cells appeared to correlate strongly with expression of Prdm1 as well as Maf over time, while Ifng expression appeared to be reduced at peak times under these conditions, we wished to determine the requirement of Prdm1 and/or Maf in the regulation of both cytokines. To address this, naïve CD4 ⁺ T cells from Prdm1 ^fl/fl Cd4 ^Cre, Maf ^fl/fl Cd4 ^Cre, Prdm1 ^fl/fl Maf ^fl/fl Cd4 ^Cre and respective floxed control mice, were differentiated into Th1 cells, with IL-12 or IL-12+IL-27, and IL-27 alone or medium controls (as in Figure 1) and expression of Il10, Ifng and Tbx21 was first assessed by RT-PCR ( Figure 4). Th1 cells differentiated with IL-12+IL-27 and IL-27-driven T cells showed significant levels of Il10 expression, which was diminished in the absence of Prdm1 or Maf ( Figure 4a–c). Effects on T cells differentiated with IL-27 alone are in keeping with findings of Zhang et al., in ‘Tr1 cells’ ¹³ . In addition to the effects that we observed on Il10 expression in the IL-12+IL-27-driven Th1 cells, the absence of Prdm1, Maf or both transcription factors, resulted in an increase in Ifng expression, while the expression of the Th1/IFN-specific transcription factor Tbx21 ⁶⁴ was not significantly affected ( Figure 4a–c).

Figure 4. Blimp-1 and c-Maf promote the expression Il10, whilst repressing Ifng in CD4 ⁺ T cells without affecting differentiation capacity in presence of IL-12, IL-12+IL-27, or IL-27.

Naïve CD4 ⁺ T cells activated in vitro with anti-CD3 and anti-CD28 antibodies and differentiated in the presence of Medium (no cytokines), IL-12, IL-12+IL-27, or IL-27 to Day 3. a– c, Real-time quantitative PCR analysis of Il10, Ifng and Tbx21 expression upon Cd4 ^Cre-mediated deletion of a, both Prdm1 and Maf, and either b, Prdm1 or c, Maf alone. Graphs show mean of technical triplicates ±SD. Data representative of n=3 independent biological experiments.

Deciphering the transcriptional programs regulated by Blimp-1 and c-Maf in CD4 ⁺ T cells

To determine the role of Blimp-1 and c-Maf on the regulation of cytokine gene networks in an unbiased fashion, naïve CD4 ⁺ T cells from Prdm1 ^fl/fl Cd4 ^Cre, Maf ^fl/fl Cd4 ^Cre, Prdm1 ^fl/fl Maf ^fl/fl Cd4 ^Cre and respective floxed control mice, were differentiated with IL-12 or IL-12+IL-27, and IL-27 or medium controls as described for Figure 1 and Figure 4. Cells were sampled at day 3 and processed for RNA-seq, which was then subjected to bioinformatics analyses ( Figure 5; Supplementary Table 4 in Underlying data ³⁰ ). Hierarchical clustering of a Pearson’s correlation analyses in each genotype revealed that the greatest variations in gene expression were cytokine driven, while differences resulting from transcription factor deletion were difficult to discern ( Figure 5a–c). Variance captured by singular-value-decomposition (SVD) components ( Figure 5d-f) allowed clustering of differential gene expression according to the cytokine-driven conditions and additionally the effects of transcription factor deletion, supported by biological pathway analysis ( Figure 5 and Figure 6; Supplementary Table 5 in Underlying data ³⁰ ). The variance explained by the SVD components capturing the Cd4 ^Cre-mediated transcription factors deletion were: 1.77% for the Prdm1 ^fl/fl Cd4 ^Cre (Component 7); 1.62% for the Maf ^fl/fl Cd4 ^Cre (Component 6); and 3.45% for the Prdm1 ^fl/fl Maf ^fl/fl Cd4 ^Cre (Component 5) ( Figure 5d–f). It is of note, that while deletion of Prdm1 and Maf in CD4 ⁺ T cells demonstrated the effects of Blimp-1 and c-Maf in regulating Il10 and negatively regulating common and unique proinflammatory gene networks in IL-12+IL-27-driven Th1 cells and to a lesser extent in IL-27 driven “Tr1” cells, these effects were less pronounced than gene expression changes resulting from TCR-stimulation and culture in the cytokines, IL-12, IL-12+IL-27, or IL-27 ( Figure 5d–f, Component 1, which constitute 58-69% of the response) and ( Figure 5d–f, Component 2 or Component 3, which constitute 14–16% or 4–8% of the response, respectively).

Figure 5. Deciphering the transcriptional programs regulated by Blimp-1 and c-Maf in CD4 ⁺ T cells.

RNA-seq analysis of CD4+ T cells differentiated in vitro with Medium, IL-12, IL-12+IL-27, or IL-27 on Day 3 with Cd4 ^Cre-mediated transcription factor deletion. a– c, Clustering of Spearman correlation showing the effect of deletion of a, Prdm1, b, Maf, or c, both Prdm1 and Maf on gene expression across all conditions. d– f, Singular value decomposition (SVD) analysis identifying changes in gene expression upon deletion of d, Prdm1, e, Maf, or f, both Prdm1 and Maf. Histograms depict percentage of variance explained for the first ten SVD-components. To statistically determine the variables associated with each component, the Akaike information criterion (AIC) score (line-plot) and ANOVA p values ( X ² test; heatmap) were calculated. g and j Heatmap of fold-changes for genes within the SVD components associated with Cd4 ^Cre-mediated transcription factor deletion in cells differentiated with g, IL-12+IL-27 and j, IL-27 alone. Fold-change values were subjected to k-means clustering to unbiasedly identify genes most affected. h- i and k- l. Heatmap of genes most affected by Cd4 ^Cre-mediated transcription factor deletion in h- i IL-12+IL-27 and k- l, IL-27 alone differentiated cells (extracted from heatmaps in g and j). Data from n=3–4 biological replicates.

Figure 6. Deciphering the transcriptional programs regulated by Blimp-1 and c-Maf in in vitro differentiated CD4+ T cells from dominant IL-27 and IL-12 driven transcriptional changes.

a, c, e, For each SVD component associated with Cd4 ^Cre-mediated deletion of either a, Prdm1, c, Maf, or e, both Prdm1 and Maf contributing genes were partitioned into “positively associated” (black outline) or “negatively associated” (orange outline) with the SVD component and their expression values visualized in a heatmap. b, d, f, Gene ontology biological pathways enriched within the SVD component associated with Cd4 ^Cre-mediated deletion of either b, Prdm1, d, Maf, or f, both Prdm1 and Maf. Data from n=3–4 biological replicates.

Heatmap values of fold-changes for genes within the combined SVD components associated with Cd4 ^Cre-mediated transcription factor deletion in cells differentiated with IL-12+IL-27 or IL-27 were subjected to k-means clustering to identify clusters of genes that were most affected by the different transcription factors in an unbiased fashion ( Figure 5g and j) Supplementary Table 6 in Underlying data ³⁰ ). The average fold-changes for each cluster showed k-means Cluster 2 to be the most decreased, and Cluster 1 the most increased in Prdm1 ^fl/fl Cd4 ^Cre and Prdm1 ^fl/fl Maf ^fl/fl Cd4 ^Cre, but less so in Maf ^fl/fl Cd4 ^Cre T cells, as compared to floxed controls, when differentiated with IL-12+IL-27 ( Figure 5g-i). Cluster 2 contained the genes most decreased in expression including Il10, Lars2 and Id2 ( Figure 5h). Conversely, Cluster 1, contained a large number of proinflammatory effector genes that were the most increased in expression including, ll3, Penk, Eomes, Il23r, Cd200, Ifng, Il2 and Csf2 ( Figure 5i). When differentiated with IL-27 alone the average fold-changes in the T cells for each cluster showed k-means Cluster 5 to be the most decreased, and Cluster 1 the most increased, in all transcription factor deficient T cells including Prdm1 ^fl/fl Cd4 ^Cre, Maf ^fl/fl Cd4 ^Cre and Prdm1 ^fl/fl Maf ^fl/fl Cd4 ^Cre, as compared to floxed controls ( Figure 5j-l). Cluster 5, containing the genes most decreased in expression included Il10, Timp1, Pdpn, Rorc and Hlx ( Figure 5k). Conversely, Cluster 1, containing proinflammatory effector genes that were the most increased in expression included, Eomes, Ifng, Il2, Csf2 and Penk ( Figure 5l), suggesting that in addition to upregulating Il10, Blimp-1 (Prdm1) and c-Maf may directly repress proinflammatory genes to reinforce a suppressive response.

GO enrichment analysis was also applied to each SVD component associated with Cd4 ^Cre-mediated deletion of either Prdm1, Maf, or both Prdm1 and Maf ( Figure 6a–f). Contributing genes were partitioned into “positively associated” (black outline) or “negatively associated” (orange outline) with the SVD component and their expression values visualized in a heatmap ( Figure 6a, 6c and 6e) and annotated using the biological processes within the GO database ( Figure 6b, 6d and 6f). An increase in expression of Ifng and Th1-associated and T cell activation-associated pathways was observed across all cytokine-driven conditions even in the absence of detectable Il10 expression, such as in Th1 cells driven by IL-12 alone, and was the most pronounced in Prdm1 ^fl/fl Cd4 ^Cre and Prdm1 ^fl/fl Maf ^fl/fl Cd4 ^Cre but less so in Maf ^fl/fl Cd4 ^Cre T cells, as compared to floxed controls ( Figure 6a–f; black outline). Conversely, accompanying the decrease in Il10 expression in IL-12+IL-27 and/or IL-27-driven cultures, decreased expression of Lars2, Id2, Hlx, Timp1, Pdpn and Rorc, was observed in Prdm1 ^fl/fl Cd4 ^Cre, Maf ^fl/fl Cd4 ^Cre and Prdm1 ^fl/fl Maf ^fl/fl Cd4 ^Cre T cells, as compared to floxed controls, but Lars2 and Id2 were reduced to a lesser degree in Maf ^fl/fl Cd4 ^Cre T cells in IL-12+IL-27-driven cultures ( Figure 5g–h); Supplementary Tables 5 and 6 ³⁰ ), suggesting that Blimp-1 and c-Maf regulate common and distinct genes/pathways to enforce a regulated immune response. These data support the alternative k-means clustering analysis approach ( Figure 5g–l) described above.

Prdm1 and Maf have complementary roles in regulating Il10 and Ifng expression in IL-12, IL-12+IL-27 and IL-27-driven T cells

To identify the molecular mechanisms whereby Blimp-1 and c-Maf affected gene regulation in in vitro differentiated CD4 ⁺ T cells from the different CD4-specific transcription factor deleted mice, we used the assay for transposase-accessible chromatin plus sequencing (ATAC-seq) to reveal changes in functionally active genomic regions across each cytokine-driven condition. Consistent with the RNA-seq profile ( Figure 5a–c), ATAC-seq revealed that the cytokines added during culture was the most dominant variable shaping the open chromatin landscape of in vitro differentiated CD4 ⁺ T cells ( Figure 7a–c). Specifically, the principal component 1 (explaining 42–56% of the variance) segregated in vitro differentiated CD4 ⁺ T cells in the presence of IL-12 or medium from those differentiated in the presence of IL-12+IL-27 or IL-27. Moreover, major chromatin remodelling occurring between Day 2 and Day 3 was observed in all in vitro differentiated CD4 ⁺ T cells in the presence of cytokines. However, no major changes in accessibility were observed upon Cd4 ^Cre-mediated deletion of Prdm1, Maf, or both Prdm1 and Maf ( Figure 7a–f). Together these results highlight a role for IL-27 (including IL-27 plus IL-12) in influencing the open chromatin landscape of in vitro differentiated CD4 ⁺ T cells and suggests that c-Maf and Blimp-1 do not drive chromatin remodelling in order to perform their gene regulation functions, as has been shown for c-Maf ²¹ . It would appear from our findings that cytokines are the main drivers of changes in chromatin remodelling while changes in chromatin remodelling are not observed in the absence of Prdm1, Maf or both transcription factors ( Figure 7a-f). This indicates that Blimp-1 and c-Maf do not themselves induce chromatin remodelling in these differentiating Th1 cells, as previously reported for c-Maf in Th17 cells in vitro ²¹ and as we have reported for c-Maf in ex-vivo T cells from infection models in other contexts ¹⁷ .

Figure 7. ATAC-seq reveals chromatin remodelling occurring between Day 2 and Day 3 in in vitro differentiated CD4 ⁺ T cells in the presence of IL-27, but no major changes in accessibility upon Cd4 ^Cre-mediated deletion of Prdm1, Maf, or both Prdm1 and Maf.

ATAC-seq analysis of CD4 ⁺ T cells differentiated in vitro in the presence of Medium, IL-12, IL-12+IL-27, or IL-27 on Day 2 and Day 3 with Cd4 ^Cre-mediated deletion of either Prdm1, Maf, or both Prdm1 and Maf, and corresponding floxed controls. a– c, PCA plots showing PC1, explaining cell differentiation in the presence of IL-27, versus PC2, explaining transition from Day 2 to Day 3. d– f, Unsupervised hierarchical clustering of a pair-wise Spearman correlation of read coverages underlying ATAC-seq peaks called in in vitro differentiated CD4+ T cells in the presence of Medium, IL-12, IL-12+IL-27, or IL-27 from Day 2 to Day 3.

Analysis of public c-Maf and Blimp-1 ChIP-seq datasets against our ATAC-seq data from T cells differentiated with IL-27+IL-12, IL-27 or IL-12, as compared to medium control, confirmed binding of Blimp-1 and c-Maf at accessible chromatin regions within the Il10 locus ( Figure 8a). These findings further support the hypothesis that Blimp-1 and c-Maf may be critical regulators of IL-10 in multiple settings including Th1 cells differentiated with IL-12+IL-27 in addition to in “Tr1” cells differentiated with IL-27 as reported ¹³ . We additionally show here that both transcription factors also bind accessible chromatin within the Ifng locus ( Figure 8b). Thus, binding of both Blimp-1 and c-Maf to the Il10 and Ifng loci indicate that Blimp-1 and c-Maf are direct positive regulators of Il10, and direct negative regulators of Ifng ( Figure 8a and b). Blimp-1 and c-Maf showed binding to the Il10 and Ifng locus at distinct sites, which suggests complementary action to induce Il10 while reducing Ifng expression ( Figure 8a and b). Distinct c-Maf and Blimp-1 Ifng binding sites were also observed by ATAC-seq analysis and these increased in T cells differentiated with IL-12 and IL-12+IL-27 and to a lesser extent in IL-27-driven T cells ( Figure 8b). These accessible binding sites were not affected by T -cell specific depletion of Prdm-1 and Maf, or both transcription factors, as compared to controls.

Figure 8. Integration of multiomic datasets identifies Il10 and Ifng as targets for reciprocal regulation by Blimp-1 and c-Maf.

Genome browser tracks of ATAC-seq data for a, Il10 and b, Ifng from Day 3 CD4 ⁺ T cells differentiated in the presence of Medium, IL-12, IL-12+IL-27, and IL-27. ChIP-seq tracks for c-Maf (green, GSE40918) and Blimp-1 (red, GSE79339) are shown. CNS reported in the literature for Il10 and Ifng are highlighted in grey shading and labelled in the bottom track. Data from n=2–3 biological replicates.

Gene regulatory networks derived from multiomic data integration highlight shared and unique targets of Blimp-1 and c-Maf

We further integrated our RNA-seq data and ATAC-seq data with data obtained by analysis of public Blimp-1 and c-Maf ChIP-seq datasets ^{21, 49} and public motif data to identify genes that were targets of Blimp-1 and c-Maf (as depicted in Figure 9 schematic). These results were confirmed by BETA software, which integrates ChIP-seq analyses and gene-expression data to identify target genes ( Figure 9). Gene regulatory networks derived from the multiomic data integration highlighted unique and shared targets between Blimp-1 and c-Maf affected by Cd4 ^Cre-mediated deletion of Prdm1, Maf or both Prdm1 and Maf ( Figure 10 and Figure 11). Specific clusters from the RNA expression data ( Figure 5g and 5j) were chosen for our depiction as gene regulatory networks in Figure 10 and Figure 11, because these clusters contained direct targets of Blimp-1 and/or c-Maf that were the most affected at the RNA expression level upon Cd4Cre-mediated deletion of Prdm1, Maf, or both Prdm1 and Maf eg. for IL-12+IL-27 found in Clusters 2,1,7 in Figure 5g; and for IL-27 alone found in Clusters 5, 7, 1 in Figure 5j.

Figure 9. Schematic of framework applied to integrate multiomic datasets.

Framework schematic for the identification of putative direct target genes of Blimp-1 and c-Maf in CD4+ T cells differentiated in vitro in the presence of Medium, IL-12, IL-12+IL-27, or IL-27 on Day 3, with replicates indicated. For each condition, Blimp-1 (GSE79339 and GSE66069) and c-Maf (GSE40918) ChIP-seq peaks and their corresponding motifs were filtered with and associated to the ATAC-seq peaks to identify the biologically relevant binding sites of Blimp-1 and c-Maf. Each ATAC-seq peak was associated to a gene based on distance proximity, thus, allowing the association of changes in the transcriptome with the binding of Blimp-1 and/or c-Maf (see Methods). Data from n=2–4 biological replicates.

Figure 10. Gene regulatory networks derived from multiomic data integration highlight a majority of shared targets between Blimp-1 and c-Maf affected by Cd4 ^Cre-mediated deletion of Prdm1, Maf, or both Prdm1 and Maf.

RNA-seq, ATAC-seq, ChIP-seq, and motifs were integrated (see Methods and Figure 9) to derive gene regulatory networks of the direct targets of Blimp-1 and/or c-Maf. Direct targets of Blimp-1 and/or c-Maf that were most affected upon Cd4 ^Cre-mediated deletion of a, Prdm1, b, Maf, or c, both Prdm1 and Maf with IL-12+IL-27 (Clusters 2,1,7 from Figure 5j) and IL-27 alone (Clusters 5, 7, 1 from Figure 5m) differentiated CD4+ T cells in vitro. In the networks, the nodes correspond to genes affected upon Cd4 ^Cre-mediated deletion of the transcription factors. The node size reflects the contribution of a gene to the SVD component associated with the deletion of the transcription factors. The node colouring represents the fold change of the Cd4 ^Cre-mediated knockouts compared to the floxed control gene expression. The edge colour depicts if Blimp-1 (pink), c-Maf (green), or both Blimp-1 and c-Maf (blue) have binding sites assigned to the target genes; whilst the thickness of the edge shows the likelihood of c-Maf and/or Blimp-1 regulating a gene according to the BETA software. Data incorporates the RNA-seq and ATAC-seq biological replicates outlined in Figure 9 schematic. Data from n=2–4 biological replicates.

Figure 11. Gene regulatory networks derived from multiomic data integration point to known and new shared targets between Blimp-1 and c-Maf.

Gene regulatory networks of the direct targets of Blimp-1 and/or c-Maf were derived, as in Figure 10, for clusters that contained genes that had been previously been reported to be regulated by c-Maf and Blimp-1 and were affected upon Cd4 ^Cre-mediated deletion of a, Prdm1, b, Maf, or c, both Prdm1 and Maf with IL-12+IL-27 (Clusters 5 from Figure 5g) and IL-27 alone (Clusters 6 from Figure 5j) differentiated CD4+ T cells in vitro.

Direct targets of Blimp-1 and/or c-Maf that were the most affected upon Cd4 ^Cre-mediated deletion of Prdm1, Maf, or both Prdm1 and Maf with IL-12+IL-27 were found in Clusters 2,1,7 from Figure 5g and for IL-27 alone in Clusters 5, 7, 1 from Figure 5j, and are depicted as gene regulatory networks in Figure 10a, 10c and 10e and Figure 10b, 10d and 10f respectively. In the networks shown in Figure 10a, 10c and 10e (IL-12+IL-27) and in Figure 10b, 10d and 10e (IL-27 alone), the nodes correspond to genes affected upon Cd4 ^Cre-mediated deletion of the transcription factors, with the node size reflecting the contribution of a gene to the SVD component associated with the deletion of the transcription factors; the node colouring represents the fold change of the Cd4 ^Cre-mediated knockouts compared to the floxed control gene expression (red up, blue down-regulated). The edge colour depicts if Blimp-1 (pink), c-Maf (green), or both Blimp-1 and c-Maf (blue) have binding sites assigned to the target genes; whilst the thickness of the edge shows the likelihood of c-Maf and/or Blimp-1 regulating a gene according to the BETA software. Data incorporates the RNA-seq and ATAC-seq biological replicates as outlined in the Figure 9 schematic.

In IL-12+IL-27-differentiated Th1 cells, shared targets included Il10, Id2, Ccl3, Ccl4, Ccl5, which were reduced upon Cd4 ^Cre-mediated deletion of Prdm1, both Prdm1 and Maf and to a lesser extent Maf, as compared to floxed controls ( Figure 10a, 10c and 10e; Supplementary Table 7 ³⁰ ). Both Prdm1 and Maf were shown to be direct targets of each other and were reduced in the absence of the reciprocal transcription factor ( Figure 10a, 10c and 10e; Supplementary Table 7 ³⁰ ). Blimp-1 and/or c-Maf targets of genes which were upregulated in Th1 cells differentiated with IL-12+IL-27 upon Cd4 ^Cre-mediated deletion of both Prdm1 and Maf, but to a lesser extent upon Prdm1, and to a far lesser degree upon Maf deletion, as compared to floxed controls, included Ifng, Il2, Il3, Tcf7, Tnfsf4, Cd80, Cd83, Eomes, Serpine1/2, Penk, Cd200, Il23r ( Figure 10a, 10c and 10e; Supplementary Table 7 ³⁰ ). Under these differentiation conditions unique direct targets of c-Maf included Il3 and Stra6 (encoding the Vitamin A Receptor) and of Blimp-1 included Penk, and these genes were the most upregulated in the Th1 cells upon Cd4 ^Cre-mediated deletion of either Prdm1 or both Prdm1 and Maf, and to a lesser extent upon Maf deletion, as compared to floxed controls ( Figure 10a, 10c and 10e; Supplementary Table 7 ³⁰ ). In T cells differentiated with IL-27 alone, shared targets again included Il10, but also Timp1, Pdpn, Rorc, and these were reduced upon Cd4 ^Cre-mediated deletion of both Prdm1 and Maf and either Prdm1 or Maf alone, as compared to floxed controls ( Figure 10b, 10d and 10f). Shared targets of Blimp-1 and c-Maf which were upregulated in IL-27 differentiated T cells upon Cd4 ^Cre-mediated deletion of either Prdm1 or both Prdm1 and Maf, and to a much lesser extent in Maf, as compared to floxed controls included Il2, Ifng, Tnfsf4 and Eomes ( Figure 10b, 10d and 10f; Supplementary Table 7 ³⁰ ). Unique targets of Blimp-1 were revealed in IL-27 differentiated T cells and included Hlx, which was reduced and Penk which was increased upon Cd4 ^Cre-mediated deletion of Prdm1, as compared to floxed controls ( Figure 10b, 10d and 10f; Supplementary Table 7 ³⁰ ). Unique targets of c-Maf included ll3 which was increased upon Cd4 ^Cre-mediated deletion of both Prdm1 and Maf and either Prdm1 or Maf alone ( Figure 10b, 10d and 10f; Supplementary Table 7 ³⁰ ). Under IL-27 differentiation conditions both Prdm1 and Maf were also shown to be direct targets of each other and were reduced in the absence of the reciprocal transcription factor ( Figure 10b, 10d and 10f; Supplementary Table 7 ³⁰ ).

Additionally, further network analysis of the IL-12+IL-27 CD4 T differentiated cell data ( Figure 11a, 11c and 11e; Supplementary Table 7 ³⁰ ) was applied to “Cluster 5” (from Figure 5g) with addition of the Prdm1 and Maf genes and to IL-27 CD4 T differentiated cell data ( Figure 11b, 11d and 11f; Supplementary Table 7 ³⁰ ) to “Cluster 6” of Figure 5j again with addition of the Prdm1 and Maf genes. Although weakly affected target genes were observed in these clusters that were affected by Cd4 ^Cre-mediated deletion of Prdm1, both Prdm1 and Maf as compared to floxed controls, these clusters included Tigit, Lag3 in keeping with previous reports ^{13, 65} .

In summary we show that c-Maf and Blimp-1 are direct targets of each other, which regulate each other to directly induce Il10 gene expression in Th1 cells differentiated in IL-12+IL-27 ( Figure 12). We show that Il10 is a direct target of both transcription factors in Th1 cells as previously reported for ‘Tr1 cells’ differentiated with IL-27 alone ¹³ . We additionally show that both c-Maf and Blimp-1 can directly bind to the proinflammatory cytokine loci Ifng, Il2 and Id2 in IL-12+IL-27 differentiated Th1 cells and negatively regulate their expression, thus enforcing a controlled Th1 effector response. Moreover, c-Maf binds and positively regulates Stra6 and binds and negatively regulates Il3; while Blimp-1 binds and positively regulates Id2 and binds and negatively regulates Cd200 and Eomes ( Figure 12) in Th1 cells differentiated with IL-12 plus IL-27. Importantly, these genes are most strongly affected by Cd4 ^Cre-mediated deletion of both Prdm1 and Maf.

Figure 12. Comprehensive transcriptomic analysis reveals that Blimp-1 and c-Maf regulate Il10, cross-regulate each other, but also negatively regulate common and unique proinflammatory gene networks.

Summary schematic of some of the findings herein presented in regards of the genes regulated by c-Maf and Blimp-1. The vast majority of direct targets detected were shared between c-Maf and Blimp-1, however they were affected to different extents upon the Cd4 ^Cre-mediated deletion of Prdm1, Maf, or both Prdm1 and Maf.

Our study shows that the transcription factors Blimp-1 and c-Maf are co-dominant positive regulators of Il10 in IL-12+IL-27-driven Th1 cells as was recently reported for IL-27-driven “Tr1” cells ¹³ . We additionally show that both Blimp-1 and c-Maf also negatively co-regulate common and unique proinflammatory gene networks in both IL-12 plus IL-27 differentiated Th1 cells and IL-27-differentiated “Tr1”cells. These data demonstrate that together Blimp-1 and c-Maf control a network of genes, specifically inducing Il10 expression, while negatively regulating proinflammatory molecules, to ensure a tightly regulated IL-12-driven Th1 effector responses to limit host damage.

Discussion

Using computational inference of gene regulation derived from temporal gene cluster profiling and analysis of active genomic regions in Th1 cells differentiated with IL-12 and IL-27 or IL-12 alone, which produce proinflammatory cytokines but differ with respect to Il10 expression, we show that Blimp-1 and c-Maf are co-dominant transcriptional regulators of Il10 gene expression. We confirm these findings using T-cell specific deletion of these transcription factors and show that both transcription factors additionally negatively regulate a network of proinflammatory effector genes in Th1 cells through their indirect and direct action on shared and distinct effector target genes, thus reinforcing a controlled Th1 cell response.

Temporal profiling of gene expression has been reported to facilitate the development of regulatory transcriptional networks dictating the differentiation of naïve CD4 ⁺ T cells in Th17 cells ^{21, 66} , and Th2 cells ⁶⁷ , leading to the discovery of key regulators of activation and differentiation and reinforcing the general principles for T helper cell differentiation. We herein applied clustering algorithms and temporal profiling to gene expression data together with analysis of active genomic regions to reveal the transcriptional networks regulating Il10 expression against that of the Th1-specific proinflammatory cytokine Ifng. Data from these analyses were compared across Th1 cells differentiated with IL-12 plus IL-27, which expressed high levels of both Ifng and Il10, and Th1 cells differentiated with IL-12 alone, which expressed high levels of Ifng but no Il10, alongside T cells differentiated with IL-27 alone, often referred to as ‘Tr1 cells’ ¹³ , which expressed Il10 but little to no Ifng and control CD4 ⁺ T cells cultured in medium alone where Il10 or effector cytokine expression is below the level of detection. This comparative analysis of Il10 expression against Ifng and other effector cytokines allowed us to identify transcription factors predicted to be positive regulators of Il10. Additionally, this computational analysis allowed us not only to identify putative co-dominant transcription factors regulating Il10 but additionally to determine any effects on Ifng and effector cytokine gene expression and transcription factors associated with Th1 cell differentiation. Th1 cells producing IFN-γ and IL-10 have been shown to be required for control of Th1 cell responses in vivo during chronic infection with intracellular pathogens to inhibit collateral host damage ^{10, 11, 15, 16} , whereas Th1 cells producing only IFN-γ have been associated with acute infection ^{68, 69} . CD4 ⁺ T cells producing IL-10 only in the absence of proinflammatory cytokines have been described ^{70, 71} and often referred to as ‘Tr1 cells’ ^{13, 72} , which in vivo have been reported in the intestine ^{70, 73, 74} although it is unclear whether they are effector Th cells which have diminished their proinflammatory cytokine expression whilst maintaining IL-10 production in the intestine ⁷³ or under certain metabolic conditions ⁷⁵ . Therefore, a detailed knowledge of transcriptional regulation of Il10 expression and effector cytokines such as Ifng and accompanying Th1 effector cytokines may provide therapeutic avenues to target inflammatory disease.

Analysis of transcription factors revealed that Prdm1 showed the most significant positive correlation with Il10 expression followed by Id2, Asb2, Hlx, Nfatc2 and Maf, some of which we and others have previously reported as regulators of Il10 ^{9, 13, 19, 20, 69} . Other proposed transcription factors reported to regulate Il10 expression in T cells, such as Hif1a and Nfil3 ^{9, 13, 20, 69} showed only a slight correlation with Il10 expression in Th1 cells differentiated with IL-12 plus IL-27 and were also expressed in Th1 cells differentiated with IL-12 alone, which did not express Il10, suggesting that they are not directly involved in Il10 gene expression, but may play a broader role in the differentiation of Th1 cells. Likewise, the transcription factor Batf, which has been previously reported to regulate Il10 expression in differentiating Th2 cells ⁶³ was actually down-regulated in both IL-12 plus IL-27 and IL-27 alone differentiation conditions, whilst increasing under IL-12 alone differentiation conditions, and showed very poor correlation with Il10 expression. Thus, as discussed earlier for IL-21, it is possible that a number of these transcription factors are required for the differentiation and possibly proliferation of T helper cells rather than for direct positive regulation of Il10 gene expression. Alternatively, the capacity of these transcription factors to regulate Il10 may be cell and/or context dependent, for example regulating Il10 in some effector T cells and not others, as previously discussed ^{9, 17} .

Analysis of transcriptional activity from ATAC-seq data, using BaGFoot software ³⁹ revealed differential transcriptional activity for the transcription factors Prdm1 and Maf between day 2 and day 3 in IL-12 plus IL-27 differentiated Th1 cells and IL-27 differentiated T cells, which both express Il10. These transcription factors were not found to be significantly active between day 2 and day 3 in IL-12 differentiated Th1 cells or medium control cultured CD4 ⁺ T cells, which do not express Il10, suggesting that these transcription factors may be co-dominant regulators of Il10. By contrast, the transcriptional activity of the AP-1 family member, Batf, was only evident in cells cultured in medium alone or IL-12, which do not express Il10. Other AP-1 family members Jun and Fos showed increased transcriptional activity across all conditions including in IL-12-driven Th1 cells producing Ifng and no Il10, as well as in Il10 expressing cells. This suggests that Batf and other AP-1 family members may be pioneer factors involved in Th cell differentiation, as previously suggested ⁷⁶ , rather than major regulators of Il10 ⁶³ and endorses the role of Jun and Fos as enhancers of Il10 gene regulation ^{6, 57} . Increased transcriptional activity of Stat 3, 4 and 5 was most pronounced under IL-12 conditions in Th1 cells expressing Ifng but not Il10. Transcriptional activity of Bhlhe40, a known negative regulator of IL-10 ¹⁷ and which we have previously shown to be negatively regulated by c-Maf, was found to be increased in cells cultured in IL-12, IL-12+IL-27 and IL-27, supporting its role as a regulator of Il10 ¹⁷ . Our findings suggesting that Prdm1 and Maf are co-dominant regulators of Il10 in IL-12 plus IL-27 differentiated Th1 effector cells are in keeping with the report from Kuchroo et al. ¹³ , who recently computed a transcriptional network induced by IL-27 in CD4 ⁺ T cells, termed ‘Tr1 cells’, expressing Il10, but little to no Ifng. Hence Prdm1 and Maf not only promote Il10 expression in a T cell regulatory setting such as ‘Tr1 cells’ as reported, but additionally appear to be co-dominant regulators of Il10 in an effector Th1 setting accompanying high levels of Ifng. Indeed, specific deletion of Prdm1, Maf and the combination of both these transcription factors in IL-12 plus IL-27 differentiated Th1 effector cells expressing Ifng, confirmed their co-dominant role in regulating Il10 gene expression in these pro-inflammatory cells. Thus, Prdm1 and Maf are not only central hubs in regulating the expression of Il10 in IL-27 differentiated ‘Tr1 cells’ where they were confirmed to control a regulatory circuit of multiple other transcriptional modulators using Prdm1/Maf DKO ‘Tr1 cells’ ¹³ , but as we now show, also regulate Il10 expression in a proinflammatory Th1 effector setting.

Commitment of T helper cells to specific subsets requires induction of master transcription factors that induce specific transcriptional programs that direct a specific T cell subset towards terminal differentiation while restricting the fates of other T cell subsets ⁷⁷ . Thus, we questioned whether Prdm1 and Maf may be part of the network for Th1 cell differentiation with Il10 expression accompanying terminal differentiation of these cells to provide feedback regulation, or alternatively contribute to a regulated Th1 response by antagonising the expression of Ifng and other proinflammatory molecules. The absence of Prdm1 and Maf in IL-12 plus IL-27 differentiated Th1 cells actually resulted in an increase in Ifng expression, showing that while Prdm1 and Maf synergistically promote Il10 expression, they negatively regulate the expression of the effector cell programme, reflected by increased expression of Ifng, thus controlling Th1 effector responses. It is likely that this co-dominant transcriptional regulation by Prdm1 and Maf is in place to ensure a controlled Th1 response against chronic infection with intracellular pathogens to minimise accompanying pathology. The negative regulation of Ifng that we observed was most pronounced in Th1 cells differentiated in IL-12 plus IL-27. This was in contrast to the discussion from Kuchroo et al. ¹³ , in IL-27-only driven IL-10 producing ‘Tr1 cells’, which expressed minimal to no Ifng, that although Prdm1 and Maf synergistically promoted IL-10 production, they did not inhibit production of T helper cell signature cytokines. This may reflect the differential effects on distinct T cell subsets. However, our findings using in-depth clustering of RNA-seq data demonstrated that T-cell specific deletion of Prdm1, Maf, or both transcription factors, led to an increase in several proinflammatory genes in both IL-27 differentiated T cells as well as IL-12 plus IL-27 differentiated Th1 cells, although to a much larger extent (expression and number of genes) in the Th1 cells. Although we found that the absence of Prdm1, Maf or both transcription factors resulted in an increase in Ifng expression, the expression of the Th1/IFN-specific transcription factor Tbx21 ⁶⁴ was not significantly affected, suggesting that Blimp-1 and c-Maf may potentially have direct effects on the Ifng gene itself. This was supported by combining ATAC-seq and ChiP-seq data, which clearly revealed both unique and overlapping Blimp-1 and c-Maf sites not only in the Il10 locus, but additionally in the Ifng locus. Moreover, Th1 cells differentiated in IL-12 (together with IL-27 or not) resulted in increased chromatin accessibility in the Ifng locus possibly enhancing the regulatory action of Prdm1 and Maf.

Relevant to our study, in recent years Blimp-1 and c-Maf have been associated with transcriptional signatures from human disease such as colitis and rheumatoid arthritis ^{78, 79} . The conserved nature of these transcription factors between mouse and humans suggests similar transcriptional mechanisms for cytokine gene regulation operate in mouse and humans. In support of this, SNPs in Blimp-1 have been associated with elevated IFN-g expression in colitis patients ⁷⁹ and Blimp-1 has been shown to bind conserved CNS sites in human and mouse in the Ifng/IFNG and Tbx21/TBX21 loci in T cells and NK cells ^{21, 80} . Likewise, c-Maf has been shown to be regulated by the cholesterol pathway and regulate IL10 expression by human Th1 cells ⁷⁴ in keeping with regulation of Il10 expression in mouse Th1 cells as we report herein. Very recently both Blimp-1 and c-Maf have been reported to co-regulate CD4 ⁺ T cell derived IL-10 in Crohn’s patients ⁸¹ and identified in Th1/Tr1 cells from malaria patients ^{82, 83} . However, the exact dynamic role of both Blimp-1 and c-Maf play in the regulation of the expression of cytokines, and other important inflammatory genes needs further research.

Additionally, gene regulatory networks derived from multiomic data integration highlighted a large number of shared and some unique targets of Blimp-1 and c-Maf which showed positive and negative regulatory effects on gene expression. Accompanying these dominant effects upon T cell-specific deletion of Prdm1 and Maf on Il10 expression in IL-12 plus IL-27 driven Th1 cells, was a decrease in other genes including Id2, Lars2 and Tigit, whilst in IL-27 driven T cells the decrease in Il10 was accompanied by a decrease in expression of genes including Timp1, Hlx, Tigit, Rorc, in keeping with the reports from Kuchroo on Il10-only producing ‘Tr1 cells’ ¹³ , while deletion of both Prdm1 and Maf in IL-12 plus IL-27 differentiated Th1 cells led to an increased number as well as level of proinflammatory gene expression including, Ifng, Il23r, Eomes, Il2,Il3, Penk and Cd200 and others, this was mostly less marked in the IL-27 alone differentiated cells, although many were found to be shared targets of both Prdm1 and Maf. A limitation of this study is firstly that the regulation of Il10 versus proinflammatory cytokines by Blimp-1 and c-Maf was firstly only demonstrated at a transcriptional level and consequent effects on protein production were not the focus of the study. Secondly, the effects of T cell-specific deletion of Prdm1 and Maf on Il10 and proinflammatory cytokine expression was only investigated in vitro. Our continuing studies will further address the role of Blimp-1 and c-Maf in regulation of cytokine responses in vivo, and the physiological consequences of T cell-specific deletion of these transcription factors in response to pathogens and/or pathobionts.

In summary, we have shown that Prdm1 and Maf are co-dominant transcription factors that induce Il10 gene expression, together with a cluster of genes including other transcription factors and co-inhibitory receptor genes, indicating their role in establishing an immunoregulatory gene programme in T cells. In addition, our findings show that both Prdm1 and Maf also negatively regulate a number of proinflammatory genes including Ifng, Il23r, Eomes, Il2, Il3, Penk and Cd200 and others, most strongly in Th1 cells, demonstrating their major role in controlling Th1 responses to allow eradication of pathogens with minimum pathology.

Funding Statement

This work was supported by The Francis Crick Institute which receives its core funding from Cancer Research UK (FC001126), the UK Medical Research Council (FC001126), and the Wellcome Trust (FC001126); this work was also supported by Wellcome (215628, <a href=https://doi.org/10.35802/215628>https://doi.org/10.35802/215628</a>), a Wellcome Investigator award awarded to Anne O’Garra, which this work will support with respect to annotation of transcriptional analysis. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data availability

Underlying data

The materials, data and any associated protocols that support the findings of this study are available from the corresponding author upon request. The RNA-seq and ATAC-seq datasets have been deposited in the NCBI Gene Expression Omnibus (GEO) database with the primary accession number GSE197789. Publicly available datasets used in this study include GSE40918, GSE79339, and GSE66069.

GEO: Blimp-1 and c-Maf regulate Il10 and negatively regulate common and unique proinflammatory gene networks in IL-12 plus IL-27-driven T helper-1 cells [Mus musculus (house mouse)]. Accession number GSE197789; https://identifiers.org/geo:GSE197789 ⁸⁴

GEO: A validated regulatory network for Th17 cell specification [Mus musculus (house mouse)]. Accession number GSE40918; https://identifiers.org/geo:GSE40918 ²¹

GEO: Hobit and Blimp1 instruct a universal transcriptional program of tissue-residency in lymphocytes [Mus musculus (house mouse)]. Accession number GSE79339; https://identifiers.org/geo:GSE79339 ⁴⁹

GEO: Analysis of Blimp-1 and Irf-1 genomic binding in wild type and Prdm1/Blimp-1 mutant embryonic gut [Mus musculus (house mouse)]. Accession number GSE66069; https://identifiers.org/geo:GSE66069 ⁵⁰

Figshare: Blimp-1 and c-Maf regulate Il10 and negatively regulate common and unique proinflammatory gene networks in IL-12 plus IL-27-driven T helper-1 cells. https://doi.org/10.6084/m9.figshare.23592393 ³⁰

This project contains the following underlying data:

• - SupplementaryTable1_KineticsRawGeneCounts.xlsx
• - SupplementaryTable2_KineticsDifferentialGeneExpression_KmeansCluster.xlsx
• - SupplementaryTable3_TF_Cytokines_GeneLists.xlsx
• - SupplementaryTable4_CD4creDeletion_FloxCtrls_RawNormGeneCounts
• - SupplementaryTable5_SVDresults.xlsx
• - SupplementaryTable6_SVD_CD4CreDeletionComponent_KmeansClusterFoldChanges.xlsx
• - SupplementaryTable7_DirectTargets_Blimp1-cMaf_NetworkValues.xlsx

Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).

Reporting guidelines

Repository: ARRIVE checklist for ‘Blimp-1 and c-Maf regulate Il10 and negatively regulate common and unique proinflammatory gene networks in IL-12 plus IL-27-driven T helper-1 cells’. https://doi.org/10.6084/m9.figshare.23592393 ³⁰

Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).