Inducible chromatin priming is associated with the establishment of immunological memory in T cells

Abstract

Immunological memory is a defining feature of vertebrate physiology, allowing rapid responses to repeat infections. However, the molecular mechanisms required for its establishment and maintenance remain poorly understood. Here, we demonstrated that the first steps in the acquisition of T-cell memory occurred during the initial activation phase of naïve T cells by an antigenic stimulus. This event initiated extensive chromatin remodeling that reprogrammed immune response genes toward a stably maintained primed state, prior to terminal differentiation. Activation induced the transcription factors NFAT and AP-1 which created thousands of new DNase I-hypersensitive sites (DHSs), enabling ETS-1 and RUNX1 recruitment to previously inaccessible sites. Significantly, these DHSs remained stable long after activation ceased, were preserved following replication, and were maintained in memory-phenotype cells. We show that primed DHSs maintain regions of active chromatin in the vicinity of inducible genes and enhancers that regulate immune responses. We suggest that this priming mechanism may contribute to immunological memory in T cells by facilitating the induction of nearby inducible regulatory elements in previously activated T cells.

Keywords: chromatin; epigenetics; gene regulation; immunity; memory T cell.

Figure 1. Previously activated T cells stably maintain an extensive array of DHSs at the IL3/CSF2 locus

1. A

   The 130‐kb human IL3/CSF2 BAC transgene. The insulator (Ins) and enhancer elements (E) are shown as boxes.

2. B

   Steps in the route to T blast cell transformation and re‐activation. Purified CD4⁺ or CD8⁺ T cells were activated with 2 μg/ml ConA for 40 h and then maintained in IL‐2 as T_B. Cells were re‐stimulated with 20 ng/ml PMA and 2 μM calcium ionophore (PMA/I).

3. C

   Inducible mRNA expression levels in CD4 T_N and T_B stimulated with PMA/I for the indicated times. mRNA levels were expressed relative to the levels of beta‐2 microglobulin (B2m) with SEM. The number of replicates (n) for each is shown underneath.

4. D

   UCSC genome browser shot of the human IL3/CSF2 locus showing DNase‐Seq and ChIP‐Seq in CD4 T_N and T_B with (red) and without stimulation (black) with PMA/I for 2 h, plus the ENCODE Jurkat T‐cell DNase‐Seq data (Thurman et al, 2012). Black arrows represent stable DHSs and red arrows are inducible DHSs, with the distances in kilobases of the DHS from either the IL3 or CSF2 promoters.

5. E

   Human CSF2 mRNA expression in CD4 T_N and CD4 T_M stimulated for 2 h with PMA/I, expressed as in (C).

6. F

   Southern blot DNA hybridization analyses of DHSs in human T_N and T_M and C42 T_B.

7. G, H

   Luciferase reporter gene assays in stimulated Jurkat T cells transfected with the pXPG plasmid containing the CSF2 (G) or IL3 (H) promoter alone or in combination with the indicated pDHS and enhancer DNA regions as defined in (D). The IL3 −4.1/1.5 construct contains a contiguous region spanning the IL3 promoter from −4.3 kb to +50 bp. The number of replicates (n) is shown below, and the error bars indicate SEM (G) or SD (H).

Figure EV1. Comparisons of mouse T_N and T_B TF mRNA and chromatin profiles

1. PCR analyses of mRNA expression of NFAT and AP‐1 family transcription factors in CD4 T_N and T_B stimulated with PMA/I for the times indicated. Expression levels are normalized to the levels of B2m. Values represent the mean and SEM of 3 to 10 replicates, with a median of 5 replicates for each value shown.

2. UCSC genome browser shot of a 900‐kb region of mouse chr11 showing DNase I‐Seq for CD4 and CD8 T_N, T_B, and T_M plus published datasets for H3K27me3 in CD4 and CD8 T_N and Th2 cells. T_N(2) and T_B(2) represent biological replicates.

Figure 2. Impaired induction of IL3 gene expression and enhancer DHS formation following deletion of a pDHS in human Jurkat T cells

1. Map of the region upstream of IL3 gene spanning the −34‐kb pDHS and −37‐kb inducible enhancer, together with the locations of the guide RNAs used to delete the −34‐kb pDHS and the PCR primers used to detect the deletion. On the right is a PCR analysis confirming deletion of the −34‐kb pDHS on both alleles in 2 out of 4 clones selected for the analyses shown below.

2. Average IL3 (upper) and JUN (lower) mRNA expression in the −34‐kb^−/− clones A and B compared to the WT clones A and B stimulated for 2, 4, and 8 h with PMA/I. mRNA levels were normalized first to GAPDH and then to the level of gene expression in untransfected Jurkat T cells. Values represent the average of two −34‐kb^−/− and two WT clones from two independent experiments (n = 4) with SD.

3. IL3 and JUN mRNA expression levels after 2 h of stimulation with PMA/I normalized as in (B). The standard error is shown from five independent experiments.

4. Deletion of the −34‐kb pDHS impairs induction of the iDHS at the −37‐kb inducible enhancer. The −34‐kb^−/− clones A and B, the WT clones A and B, and untransfected Jurkat T cells were stimulated with PMA/I for 3 h. A range of DNase I concentrations were used to determine the chromatin accessibility of the −37‐kb iDHS in two independent clones, with values expressed relative to normal unstimulated Jurkat cells. Increased accessibility was detected by a reduction in signal detected by qPCR. The active TBP promoter and an inactive region on Chr18 are used as controls. Independent experiments for the −34‐kb^−/− and WT clones A and B compared to the untransfected Jurkat T cells are shown in the upper and lower panels, respectively.

Figure 3. Genomewide mapping identifies a class of DHSs restricted to previously activated T cells

1. Density maps depicting all DNase‐Seq peaks in the order of increasing DNase‐Seq tag count signal for CD4 T_M compared to T_N. On the right are the locations of the defined subset of 2,882 pDHSs and the log2 T_M/T_N fold change in expression of the closest gene to the corresponding DHS.

2. Density maps for all DNase‐Seq and ChIP‐Seq peaks shown in order of increasing DNase‐Seq tag count signal for CD4 T_B compared to T_N. The T_N H3K27ac track is from published data (Lara‐Astiaso et al, 2014).

3. Average DHS signal at 2,882 pDHSs and 2,882 invariant DHSs in CD4 T_N, T_B, and T_M. The locations of the 2,882 pDHSs are indicated in (A) and (B).

4. Pie chart showing the genomic distribution of pDHSs.

5. Average H3K4me2, H3K27ac, and BRD4 signal at 2,882 pDHSs.

6. Plot of the log2 fold change in gene expression following 2 h PMA/I for CD4 T_M compared to T_N. 1,895 genes (black) have a log2 fold change of 1 or above in T_M but not T_N.

7. Distance and gene expression analyses. P‐value represents χ² significance against randomly expected number of pDHSs within 25 kb of a TSS (method described in Appendix).

Figure EV2. CD4 and CD8 T cells share a common set of pDHSs

1. A

   Density maps representing the DNase‐Seq peaks in the cell types indicated at the top, shown in the order of increasing DNase‐Seq tag count signal for the CD4 T_B compared to CD4 T_N. T_B(2) and T_N(2) are biological replicates.

2. B

   Average DNase I profile at the pDHSs in CD4 T_B(2) and T_N(2), CD8 T_B and T_N, plus the average H3K27me3 profiles for publically available CD4 and CD8 datasets in T_N T_M, effector T cells (T_E), and Th2 cells.

3. C, D

   Boxplots of log2 mRNA expression fold change (C) and absolute expression levels (D) of the 1,895 T_M‐specific genes in CD4 T_N, CD4 T_B and CD4 T_M. Boxes represent the first and third quartile, respectively. Bottom and top whiskers represent the first and third quartile minus and plus 1.5 times the interquartile range.

4. E

   Gene ontology for the 1,895 T_M‐specific genes.

5. F

   Mean cumulative mRNA array values for two alternatively spliced forms of the gene encoding NFATc1. Values are based on 4 separate micro‐array values and are shown with SD.

6. G

   Hierarchical correlation clustering of mRNA levels for the top 1% of genes with the highest variance of mRNA expression between populations of CD4 and CD8 T_{N ,} T_B, and T_M. Treatment with PMA/I is indicated by a “+” sign. Pearson correlations are shown according to the color scale (top left). B, N, and + (right) indicate the dominant groups resulting from clustering.

Figure 4. iDHSs lie close to pDHSs and are associated with inducible genes

1. A

   Density maps identifying iDHSs and showing DNase‐Seq and H3K4me2 ChIP‐Seq peaks in order of increasing DNase‐Seq tag count signal for CD4 T_B ⁺ compared to T_B cells. Also depicted are the locations of 6,823 major DHSs that are 5.5‐fold induced and a subset of 1,217 of these iDHSs that are 11‐fold induced. On the right is the log2 T_B ⁺/T_B fold change in expression of the closest gene to the corresponding DHS.

2. B

   Average DNase I profiles of the 1,217 iDHSs in T_N (+/− PMA/I) and T_B (+/− PMA/I) (left), and in T_M (+/− PMA/I) (right).

3. C

   Average DNase I and H3K27ac profiles of the 1,049 dDHSs in T_N and T_B (+/− PMA/I) which are fourfold diminished after stimulation.

4. D

   The genomic distribution of the 1,217 iDHSs.

5. E–H

   Barplots showing the number of T_M‐specific genes with an iDHS within 150 kb of the TSS (E); the median distances between the TSSs and the closest iDHS of the T_M‐specific genes grouped according to the fold induction in T_M after 2 h plus PMA/I compared to T_M for genes which had a TSS < 1 Mb from an iDHS (F); the number of T_M‐specific genes which have an iDHS within 150 kb of the 683 pDHSs (G); and the median distances from the closest pDHS to the closest iDHS grouped according to the fold induction in T_M after 2 h plus PMA/I compared to T_M (H). P‐values represent either χ² significance against randomly expected number of DHSs within 25 kb (F and H) or t‐test significance against equally sized random DHSs (E and G). The methods used to calculate P‐values are described in the Appendix.

6. I

   UCSC genome browser shot of the Th2 Il4/Il13/Rad50 locus showing DNaseI‐Seq and ChIP‐Seq. Black and red arrows represent pDHSs and iDHSs, respectively. The values above the arrows indicate the distance in kb from the Il4 promoter.

Figure EV3. Properties in inducible DHSs

1. Average H3K4me2, H3K27ac, and BRD4 signals at the 1,217 iDHSs in CD4 T_B and T_B ⁺, plus H3K4me2 and H3K27ac for T_N.

2. Gene ontology for the 187 T_M‐specific genes located within 25 kb of both a pDHS and an iDHS.

3. DNase I‐Seq and ChIP‐Seq at the Il10 locus in CD4 T_{N ,} T_{M ,} T_{B ,} T_B ⁺, and Th2 cells and CD8 T_N and T_B.

4. Il4 and Il10 mRNA expression in CD4 T_N and CD4 T_M stimulated with PMA/I for the times indicated. Relative mRNA values are expressed as in Fig 1C, with SEM. The number of replicates for each (n) is shown underneath.

Figure 5. pDHSs bind constitutively expressed transcription factors

1. De novo motifs enriched within 2,882 pDHSs determined using HOMER.

2. Motif distribution in all DHSs ordered by increasing DNase‐Seq tag count signal for CD4 T_B relative to T_N as in Fig 3B.

3. DNase‐Seq and RUNX1 and ETS‐1 ChIP‐Seq density maps showing the binding at the DHSs ordered as for (B), with average profiles of RUNX1 and ETS‐1 binding to the pDHSs in T_B compared to T_N shown below.

4. JUNB ChIP‐Seq in T_B and T_B ⁺ at the DHSs defined in T_N and T_B and ordered as in (B) with average profiles shown below.

5. Patterns of RUNX1, ETS‐1, and JUNB binding at RUNX, ETS, and AP‐1 motifs at the Trpm6 locus.

Figure 6. Co‐association of ETS‐1 and RUNX1 in pDHSs, and the role of RUNX1 in establishing priming

1. Hierarchical clustering of motif co‐association enrichments in pDHSs. Z‐scores represent enrichment of observed versus background co‐associations computed in 1,000 randomly selected, chromatin‐accessible regions.

2. Log2 mRNA expression levels of Runx1, Ets1, and B2m in untreated and PMA/I‐treated T_{N ,} T_{M ,} and T_B from the microarray analysis.

Figure 7. Motifs for constitutive TFs are footprinted in vivo in pDHSs

1. Enriched motifs defined by HOMER using a de novo motif search of the digital DNase I FPs identified in the pDHSs in T_B.

2. DNase I cleavage patterns in T_B from the FPs determined by Wellington at the pDHSs centered on the motif named at the top and ordered according to increasing FP occupancy score. Left: Cuts are shown within a 200‐bp window with positive (red) and negative (green) strand imbalances in DNase I cuts. Right: Average profiles of the actual DNase I cuts at footprinted motifs within the pDHSs, with upper strand DNA cuts shown in red and lower strand cuts in blue.

3. Average profiles of the DNase I cuts at footprinted motifs within the pDHSs in T_M determined as in (B).

4. Example of FP patterns and motifs at the −3.7‐kb and −35‐kb pDHSs at the Ccl1 locus in T_B.

5. mRNA array values for Ccl1 expression.

6. Luciferase reporter gene assays in stimulated Jurkat T cells performed as in Fig 1H of the IL3 promoter alone or in combination with the Ccl1 −3.7‐kb or −35‐kb DHSs, with SD. Values are expressed as the mean with the number of replicates for each (n) shown underneath.

7. Footprinting of AP‐1 sites in T_B ⁺. DNase I upper and lower strand cleavage patterns were calculated as in (B) (left) plus the average DNase I profiles (right) for all AP‐1 motifs within the subset of 2,882 defined pDHSs in T_B cells before and after stimulation, ranked in order of decreasing FP probability score.

8. Distribution of the FP probability scores for the data shown in (G).

Figure EV4. DNase I footprints of TFs in T_B

1. Left: DNase I cleavage strand imbalance patterns within the footprints identified by Wellington at the pDHSs centered on the motif named at the top and ordered according to increasing FP occupancy score. Relative levels of DNase I cuts are shown within a 200‐bp window with upper strand DNA cuts shown in red and lower strand cuts in green. Right: Average profiles of the DNase I cuts at the different motifs within the pDHSs, upper strand DNA cuts shown in red and lower strand cuts in blue.

2. Results of the HOMER de novo motif search of pDHS digitally footprinted regions in CD4 T_M.

3. ETS and RUNX motif‐containing footprints (right) within DHSs in T_M sorted by T_M/T_N fold change (left).

Figure 8. Inducible DHSs bind inducible and constitutively expressed transcription factors

1. De novo motif search of the 1,217 iDHSs using HOMER.

2. Motif distributions in the DHSs ordered by increasing DNase‐Seq tag count signal for CD4 T_B ⁺ cells compared to T_B cells as in Fig 4A.

3. RUNX1 and JUNB ChIP‐Seq density maps depicting binding at all DHSs ordered as in (B).

4. RUNX1 and JUNB ChIP‐Seq profiles for DHSs at the Il10 locus in T_B and T_B ⁺.

5. DNase I cleavage strand imbalance patterns displayed as in Fig 7G for footprinted TF motifs in T_B ⁺ at the iDHSs.

6. Example of FP patterns and motifs at the −15‐kb and −35‐kb iDHSs in T_B ⁺ cells at the Ccl1 locus.

Figure EV5. Co‐association of inducible TFs in iDHSs

1. Hierarchical clustering of motif co‐association enrichments in iDHSs. Z‐scores represent enrichment of observed versus background co‐associations computed in 1,000 randomly selected, chromatin‐accessible regions. Z‐score scale as in Fig 6A.

2. UCSC genome browser shot of the Il4, Il13, and Rad50 loci showing DNase‐Seq and ChIP‐Seq for T_B and T_B ⁺.

3. De novo motifs identified by HOMER within iDHS digital FPs.

4. Average profiles of DNase I cuts at motifs within the iDHSs. Upper strand DNA cuts are shown in red and lower strand cuts in blue.

Figure 9. Composition and properties of pDHSs and iDHSs

1. Motif counts for abundant TF binding sites at the specific DHSs in T_M, T_B, and T_B ⁺.

2. Total motif counts for 5 inducible motifs (AP‐1, NFAT, EGR, NF‐κB, and CREB/ATF) and 5 constitutive motifs (ETS, RUNX, KLF, GATA, and E‐box) in the specific DHSs in T_M, T_B, and T_B ⁺. The motifs used here are defined in Dataset EV5.

3. Overlaps between ETS‐1 and RUNX1 ChIP peaks and the 2,882 pDHSs in T_N compared to T_B.

4. Mechanisms of pDHS and iDHS regulation in T cells.