CCCTC-Binding Factor Translates Interleukin 2- and α-Ketoglutarate-Sensitive Metabolic Changes in T Cells into Context-Dependent Gene Programs

Abstract

Despite considerable research connecting cellular metabolism with differentiation decisions, the underlying mechanisms that translate metabolite-sensitive activities into unique gene programs are still unclear. We found that aspects of the interleukin-2 (IL-2)-sensitive effector gene program in CD4⁺ and CD8⁺ T cells in type 1 conditions (Th1) were regulated by glutamine and alpha-ketoglutarate (αKG)-induced events, in part through changes in DNA and histone methylation states. We further identified a mechanism by which IL-2- and αKG-sensitive metabolic changes regulated the association of CCCTC-binding factor (CTCF) with select genomic sites. αKG-sensitive CTCF sites were often associated with loci containing IL-2- and αKG-sensitive genome organization patterns and gene expression in T cells. IL-2- and αKG-sensitive CTCF sites in T cells were also associated with genes from developmental pathways that had αKG-sensitive expression in embryonic stem cells. The data collectively support a mechanism wherein CTCF serves to translate αKG-sensitive metabolic changes into context-dependent differentiation gene programs.

Figure 1. αKG-induced metabolic changes regulate select genes in the IL-2-sensitive pathway

(A, B) Transcripts from (A) primary CD4⁺ T cells or (B) CD8⁺ T cells polarized in type 1 conditions and exposed to high IL-2 (black bars), low IL-2 (white bars), or low IL-2 with αKG (grey bars) for two days were measured by qRT-PCR and/or RNA-seq (see also Fig. S2). (C) Flow cytometry, (D) ELISA, and (E, F) western blot analyses of protein expression levels for CXCR3, IFNγ, TCF1 and HK2 are shown for primary CD4⁺ T cells maintained as indicated in (A). The n= at least (A, C, E, F) 3, (B) 4, or (D) 6 independent biological replicates for each gene/protein. (A, B, D) Error bars represent the standard error of the mean (SEM) and an unpaired student t-test was performed with p values indicated (* ≤0.05, **≤0.001, and ***≤0.0001). See also Fig. S1, S2.

Figure 2. Gln- and α KG-sensitive events regulate a portion of the IL-2-sensitive pathway

(A, B) Venn diagrams comparing genes that were induced (left) or inhibited (right) by at least 2-fold in high IL-2 (blue circles) or the addition of αKG to low IL-2 (yellow circles) as compared to low IL-2 conditions in (A) CD4⁺ or (B) CD8⁺ T cells. (C, D) FDR q values for the enrichment of select pathways from a GSEA examining genes from (C) CD4⁺ or (D) CD8⁺ T cells that were induced 2-fold in high IL-2 relative to low IL-2 (black bars) or low IL-2 with αKG relative to low IL-2 (hatched bars). (E) Flow cytometry analysis of E450 dilution in primary CD4⁺ T cells polarized in Th1 conditions and exposed to high IL-2 (dark blue), low IL-2 (light blue), or low IL-2 with αKG (purple) for two days. Data are representative of 3 independent biological replicates. (F) qRT-PCR analysis of transcript expression from primary CD4⁺ T cells polarized in Th1 conditions that were maintained in media with Gln and in high IL-2 conditions (black bar), low IL-2 (white bar), or low IL-2 with αKG (dark grey bar) or were maintained without Gln and in high IL-2 (light grey) or high IL-2 with αKG (medium grey) for two days. The n= at least 4 independent biological replicates for each gene. Error bars represent SEM and an unpaired student t-test was performed with p values indicated (* ≤0.05, **≤0.001, and ***≤0.0001). (G) The frequency of CD45.1⁺OTII cells in the recipients of control (Ctrl) and DON treated cells. (H, J) The frequency and (I, K) number of CD45.1⁺OTII cells with a (H, I) Teff phenotype (Bcl6^loCXCR5^loTbet^hiBlimp1^hi) or (J, K) Tfh cell phenotype (Bcl6^hiCXCR5^hiTbet^loBlimp1^lo) in recipients receiving control or DON treated cells. Data are shown as the mean ± SD (n=5–6 mice) and are representative of two independent experiments. P values were determined using a two-tailed Student´s t-test (* ≤0.05, **≤0.005). (L) Heatmap displaying Z-scores for genes that were induced (top) or inhibited (bottom) by at least 2-fold in comparison to cells maintained without Gln. Conditions are shown above each lane (see STAR methods for detailed description of design). See also Fig. S1–3.

Figure 3. IL-2 and αKG-sensitive events are related to H3K27me3 and DNA methylation states

(A) GSEA of select chemical and genetic perturbations datasets examining genes induced in the low IL-2 with αKG compared to the low IL-2 condition in CD4⁺ Th1 (light grey) or CD8⁺ Tc1 (dark grey) cells. (B) Histograms representing differential H3K27me3 ChIP-seq peak read depth enrichment between high IL-2 compared to low IL-2 conditions (left) or low IL-2 with αKG relative to low IL-2 conditions (right). The distribution of the log ratio of peak read enrichment is displayed. (C, D) qRT-PCR analysis of transcripts from primary CD4⁺ Th1 cells exposed to (C) high IL-2 (black bars), low IL-2 (white bars), high IL-2 with DON (light grey bars) or high IL-2 with GSK-J4 (dark grey bars) or (D) high IL-2 (black bars), low IL-2 (white bars), low IL-2 with αKG (dark grey bars), or low IL-2 with 5-azacytidine (light grey bars). (C, D) The n= at least (C) 4 except high IL-2 with DON n= 3 or (D) 3 independent biological replicates for each gene. Error bars represent SEM and an unpaired student t-test was performed with p values indicated (* ≤0.05, **≤0.001, and ***≤0.0001). See also Fig. S3.

Figure 4. IL-2- and αKG-sensitive events impact CTCF

(A) A heatmap representing genome-wide CTCF enriched ChIP-seq peaks from biological replicates of primary CD4⁺ T cells polarized in Th1 conditions that were exposed to high environmental IL-2, low IL-2, or low IL-2 with αKG conditions for two days. (B) Histograms representing differential CTCF ChIP-seq peak read depth enrichment between high IL-2 compared to low IL-2 conditions (left) or low IL-2 with αKG relative to low IL-2 conditions (right). The distribution of the log ratio of peak read enrichment is displayed. (C) CTCF-ChIP-seq tracks from two biological replicates of CD4⁺ Th1 cells exposed to high IL-2, low IL-2, or low IL-2 with αKG displayed with the UCSC genome browser. Differential CTCF peaks are highlighted with a blue arrow. (D, E) qRT-PCR analysis of transcripts from primary CD4⁺ T cells polarized in Th1 conditions and transfected with a control siRNA to GFP (black bar), (D) an siRNA to Ctcf (grey bar) or (E) an siRNA to Rad21 (dark grey bar). The control GFP siRNA is the same for Figure S3E and 4D. The n= at least 3 independent biological replicates for each gene. Error bars represent SEM and an unpaired student t-test was performed with p values indicated (* ≤0.05, **≤0.001, and ***≤0.0001). See also Fig. S4–7.

Figure 5. IL-2- and αKG-sensitive CTCF association correlates with changes in genomic organization

(A–D) In situ Hi-C or CTCF ChIP-seq experiments were performed with primary CD4⁺ T cells polarized in Th1 conditions and maintained in high IL-2, low IL-2, or low IL-2 with αKG for two days. (A,B) UCSC genome browser tracks displaying a Hi-C (A) PC1 analysis to define transcriptionally permissive (PC1 positive; black) versus inert/or transcriptionally repressive (PC1 negative; grey) genomic compartments, (B) an analysis of outer TAD boundaries and (A,B) ChIP-seq CTCF tracks as described in 4C. (C, D) Circos plots for genomic regions surrounding Hlx (Chr1:184,000,000–187,200,000) and Cxcr4 (Chr1:128,800,000–131,040,000) indicating the probability for genomic interactions from the in situ Hi-C analysis. The minimum probability of interaction shown is a p value of ≤0.0001 with the increased weight of a line indicating a higher significance for the interaction (lower p value). (E) UCSC genome browser tracks displaying CTCF ChIP-seq from Th1 cells exposed to high IL-2, low IL-2, or low IL-2 with αKG, and an ATAC-seq analysis of NP-specific CD8⁺ T cells isolated 9 days after influenza infection that were sorted into CD25^hi or CD25^lo populations. Regions displayed were associated with the interaction loops identified in the circos plots in (C, D). Blue arrows indicate IL-2- and αKG-sensitive CTCF sites and yellow arrows indicate ATAC-seq changes. See also Fig. S4–6.

Figure 6. IL-2- and αKG-sensitive events inhibit DNA methylation within CTCF peaks found in proximity to Hk2 and Sell

(A) CTCF-ChIP-seq tracks from two biological replicates of CD4⁺ Th1 cells exposed to high IL-2, low IL-2, or low IL-2 with αKG displayed with the UCSC genome browser. The differential CTCF peaks analyzed in (B) and (C) are highlighted with a blue arrow above each track. (B, C) Bisulfite sequencing analysis of DNA from primary CD4⁺ T cells polarized in Th1 conditions and maintained in high IL-2, low IL-2, or low IL-2 with αKG for two days. The CpG sites within IL-2- and αKG-sensitive CTCF peaks in proximity to (B) Hk2 and (C) Sell as indicated in (A) were monitored. Open circles represent the average of clones with no methylation whereas closed circles represent the average of methylated clones for an experiment. Results are representative of two independent biological replicates (see also Fig. S4E).

Figure 7. A subset of genes with αKG-sensitive CTCF peaks have αKG-inducible gene expression in ES cells

(A, B) UCSC genome browser tracks representing Hi-C PC1 analysis, CTCF ChIP-seq, or H3K27Ac ChIP-seq peaks for CD4⁺ Th1 cells exposed to high IL-2, low IL-2, or low IL-2 with αKG. Displayed are regions surrounding the (A) Pdgfra gene or (B) genes from differentiation pathways associated with ES and T cells. (C) Graphs representing the FPKM values from the RNA-seq analyses of CD4⁺ Th1 cells, E14Tg2a cells, and CD8⁺ Tc1 cells with 3, 4, and 4 independent biological replicates respectively. See also Fig. S7.