Figure 1.
EBNA3A repressed genes are significantly enriched for genes controlled by PcG proteins in diverse tissues and frequently form contiguous gene clusters in the human genome.
(A) Overlap of ENCODE H3K27me3 ChIP-Seq data in wt LCLs (GM12878) with NM genes represented on Affymetrix HG-U133A 2.0 microarrays used in our previous expression profiling study. Enrichment for H3K27me3 was calculated for 125 cellular genes repressed by EBNA3A ≥2.0-fold (p≤0.05) when compared to the total number of genes analyzed on the array. Enrichment is given as odds ratio and was tested for statistical significance using Fisher's exact test. (B) H3K27me3 marks at EBNA3A repressed genes are not only present in LCLs. The 125 EBNA3A repressed genes were analyzed for H3K27me3 occupancy in 7 EBV negative cell lines of various tissue origin using ENCODE data. Odds ratios were calculated with respect to the H3K27me3 status of the total number of genes analyzed on HG-U133A 2.0 arrays in the individual cell lines and tested for statistical significance with Fisher's exact test. (C) EBNA3A repressed genes are committed to PcG control in various cell lines. The 89 EBNA3A repressed genes that scored positive for H3K27me3 in wt LCLs were analyzed for the proportion of genes that are also silenced by H3K27me3 in EBV negative cell lines. Results indicate the percentage of EBNA3A targets detected as PcG-silenced in LCLs only or in addition in some of the 7 analyzed cell lines. (D) EBNA3A repressed genes frequently form co-regulated gene clusters in the human genome. 220 EBNA3A repressed genes were filtered from previous expression profiling data by applying the indicated thresholds for fold change (fc), present call (pc), and p-value criteria and analyzed for their position within the human genome. 47 genes could be assigned to groups of 2, 3, and 4 gene clusters. The expected frequency of clustered genes in the same number of randomly selected genes was calculated for comparison. Error bars indicate 90% confidence intervals (binomial test). Please note that genes in groups overlap, e.g., a cluster of 3 contiguous genes includes the clusters of 2 contiguous genes.
Figure 2.
CXCL10 and CXCL9 reside within a PcG-controlled chromatin domain of 118 kb and are rapidly repressed upon EBNA3A expression.
(A) Schematic representation of a genomic region on human chromosome 4 showing the location of the EBNA3A repressed genes CXCL10 and CXCL9 as well as flanking genes and the H3K27me3 coverage in wt LCLs (GM12878) according to ENCODE data. CXCL10 and 9 comprise a region of 22 kb, which is embedded in an H3K27me3 positive domain of 118 kb. CXCL11 and ART3 also reside within this domain but are neither expressed in wt nor EBNA3A negative LCLs, while SDAD1 and NUP54 show similar expression levels irrespective of the EBNA3A status. Dotted lines demarcate an alternative TSS of ART3, which is not used in LCLs. (B) Validation of differential CXCL10 and 9 expression in wt and EBNA3A negative LCLs derived from 5 unrelated B cell donors. Transcripts of CXCL10 and 9 were quantified by qPCR in triplicate cDNA preparations from LCLs established by infection of B cells with EBVwt or either EBV-E3AmtA (D1, D4, D5) or EBV-E3AmtB (D2, D3). Data were normalized to 18S rRNA levels and are given as mean ± standard deviation (SD). (C) Western blot analysis of EBNA3A expression in ΔE3A-LCLdoxE3A cells prior to and 24 or 48 h post treatment with 100 ng/ml Dox. Protein extracts of the parental EBNA3A negative LCL (D2 E3AmtB 3) and a corresponding wt LCL (D2 wt 1) served as a negative and positive control, respectively. GAPDH immunodetection was used as loading control. Protein band intensities were quantified by densitometry. EBNA3A protein levels were normalized to GAPDH and are given as x-fold expression relative to the expression level in the corresponding wt LCL. (D) Flow cytometric analysis of NGFR expression in ΔE3A-LCLdoxE3A cells prior to and 24 h post treatment with 100 ng/ml Dox. Staining of cells with isotype-matched nonspecific antibodies served as a negative control. (E) EBNA3A induction in conditional LCLs rapidly down-regulates CXCL10 and 9 expression. ΔE3A-LCLdoxE3A cells were induced for EBNA3A (E3A) expression by treatment with 100 ng/ml Dox for 24 or 48 h or left untreated. For metabolic labeling of nascent RNA, cells were cultured in the presence of 4sU for 1 h prior to harvesting. CXCL10 and 9 transcripts in total RNA were quantified by qPCR, normalized to total 18S rRNA levels, and are given as mean ± SD of two biological replicates analyzed in triplicates. (F) CXCL10 and 9 repression by EBNA3A is achieved by reduction of de novo transcription. Nascent RNA was isolated from total RNA prepared in (E). Nascent CXCL10 and 9 transcripts were quantified by qPCR, normalized to nascent 18S rRNA levels, and are given as mean ± SD of two biological replicates analyzed in triplicates.
Figure 3.
H3K27me3 marks are elevated throughout the CXCL10/9 domain in EBNA3A positive LCLs.
(A) Schematic representation of the CXCL10 and 9 encompassing domain indicating the positions of primer pairs A-T used for qPCR quantification of ChIPed DNA relative to the TSS of the analyzed genes. (B–G) ChIP analysis of established wt and EBNA3A negative LCLs (D2 wt 1 and D2 E3AmtB 3) showing the abundance of (B) Pol II, (C) H3ac, (D) H3K4me3, (E) H3K27me3, (F) SUZ12, and (G) EZH2. Bars indicate the enrichment of Pol II, of histone modifications and of PRC2 subunits at the individual loci as assessed by qPCR with primer pairs A-T. Primer pairs for the TSS of GAPDH (ctrlac) and a pericentromeric region on chromosome 1 (ctrlsi) were included as a control for active and silenced chromatin, respectively. Bar height was calculated as percentage of ChIPed DNA recovered from input DNA, after subtraction of values from negative control IgG precipitation. Data are representative of three independent experiments. Error bars indicate SD of triplicate qPCR reactions (with exception of data in panel G, which are given as mean ± range of two independent experiments).
Figure 4.
Transcriptional down-regulation precedes the gain of repressive H3K27me3 chromatin marks.
ChIP analysis of ΔE3A-LCLdoxE3A cells showing the abundance of (A) Pol II, (B) H3ac, (C) H3K4me3, and (D) H3K27me3 across the CXCL10 locus (primer pairs H-J, see Figure 3A) prior to and 24 h post EBNA3A induction with 100 ng/ml Dox. Primer pair S was used as a control. Bars were calculated and displayed as in Figure 3. ChIP analyses of a wt LCL were included for comparison.
Figure 5.
Maintenance of the Polycomb signature depends on EBNA3A.
(A) ChIP analysis of ΔE3A-LCLdoxE3A cells showing the occupancy of Pol II, H3K4me3, and H3K27me3 at the CXCL10 locus (primer pairs I and J, see Figure 3A) prior to, after 2 weeks of EBNA3A expression, and 2 weeks after EBNA3A shut-off. Results were calculated and displayed as in Figure 3. Primer pair S was included as a control. (B) qPCR quantification of EBNA3A transcripts in RNA extracts prepared at the indicated points in time. Results were normalized to 18S rRNA levels and are given as mean ± SD of triplicate qPCR reactions. (C) Western blot analysis of protein extracts using α-EBNA3A antibody. GAPDH detection was used as a loading control. Protein band intensities were quantified by densitometry. EBNA3A protein levels were normalized to GAPDH and are indicated as the percentage of EBNA3A protein remaining after Dox withdrawal relative to the expression level seen before Dox withdrawal. (D) qPCR analysis of CXCL10 and 9 re-expression upon EBNA3A shut-off using the same RNA extracts as in panel (B). Results were normalized to 18S rRNA levels and are given as mean ± SD of triplicate qPCR reactions.
Figure 6.
EBNA3A impairs CXCL10 and 9 induction by IFNγ via a CBF1-dependent mechanism.
(A) Analysis of CXCL10 and 9 repression by EBNA3A in DG75 wtdoxE3A and DG75 kodoxE3A cell lines. EBNA3A expression was induced with 100 ng/ml Dox for 24 h either post (upper panels) or prior to (lower panels) IFNγ treatment for 6 h. CXCL10 and 9 transcripts were quantified by qPCR prior to and post IFNγ or Dox treatment and normalized to 18S rRNA levels. X-fold repression and induction values are given as mean ± SD of two biological replicates analyzed in triplicates. (B) qPCR quantification of CXCL10 and 9 transcripts prior to and post IFNγ treatment of a wt and an EBNA3A negative LCL established from the same B cell donor. Transcript levels were normalized to 18S rRNA levels and are shown as mean ± SD of three independent experiments.
Figure 7.
EBNA3A directly targets intergenic enhancers between CXCL10 and 9 that are also bound by CBF1 and EBNA2.
(A) Close-up of enhancer regions R1–R3 which are clustered within an intergenic 6 kb region located between CXCL10 and 9. R1–R3 are bound by CBF1 and EBNA2 in LCLs according to published ChIP-seq results [19], which are displayed as raw read data for EBNA2, CBF1, and input DNA duplicates. The depicted region was additionally analyzed for CBF1 consensus binding sites [100] and aligned with ENCODE DNase-seq data, ChIP-seq data for H3K4me1, H3K27ac, p300, and Pol II, as well as strong enhancer annotations revealed by chromatin state segmentation. All displayed ENCODE data were generated with wt LCLs (GM12878). Black lines demarcate region R1, R2, and R3. (B) ChIP analysis with α-HA antibody showing the binding of HA-tagged EBNA3A to regions R1–R3 24 h post HA-EBNA3A induction with 100 ng/ml Dox in ΔE3A-LCLdoxHA-E3A cells. Results were either calculated as described in Figure 3 (left panel) or displayed as fold enrichment of α-HA precipitated DNA over negative control IgG precipitation (right panel). Primer pair Q (see Figure 3A) shows neither CBF1 nor EBNA2 binding and was used as a negative control. (C) ChIP analysis of EBNA2 occupancy at regions R1–R3 prior to and 24 h post HA-EBNA3A induction with 100 ng/ml Dox in ΔE3A-LCLdoxHA-E3A cells. Results were calculated and displayed as in Figure 3. Primer pair Q was used as a negative control. (D) ChIP analysis of EBNA2 occupancy at regions R1–R3 in established wt and EBNA3A negative LCLs. Results are shown as mean ± SD of two independent experiments analyzed in duplicates. Primer pair Q was used as a negative control.
Figure 8.
EBNA3A binding to intergenic enhancers reduces enhancer activity.
ChIP analysis of (A) histone H3, (B) H3K4me1, (C) H3K27ac, and (D) Pol II occupancy at regions R1–R3 prior to and 24 h post HA-EBNA3A induction with 100 ng/ml Dox in ΔE3A-LCLdoxHA-E3A cells. Data are shown as mean ± SD of three independent experiments analyzed in duplicates. Results for H3K4me1 and H3K27ac were normalized to total histone H3 levels to account for the low nucleosomal occupancy at regions R1–R3. Asterisks indicate the p-value as determined by Student's t-test. Primer pair Q was used as a negative control.
Figure 9.
A 2-step model for EBNA3A's mode of action.
(A) EBNA3A displaces the transactivator EBNA2 from CBF1 occupied intergenic enhancers. Reduction of EBNA2 triggered enhancer activity by EBNA3A binding causes a rapid transcriptional shut-down of adjacent CXCL10 and 9 genes. In the absence of EBNA2, however, EBNA3A acts by its intrinsic repressor activity, rendering CXCL10 and 9 refractory to IFNγ-mediated induction. (B) The transcriptionally repressed state of CXCL10 and 9 is subsequently fixed on the chromatin level by PcG proteins. PRC2-catalyzed H3K27me3 marks spread in a domain-wide fashion, potentially starting from remote enhancers. The gain of H3K27me3 levels to full range is a slow process that requires a time period of at least 14 days. When EBNA3A expression is discontinued, PcG repression is reversed and re-expression of distal genes is permitted (blue stars: H3K27ac; red hexagons: H3K27me3).