Epstein-Barr virus establishes a lifelong latent infection in most humans and is causally linked to several lymphoid and epithelial cancers. A striking feature of that latent state is the sheer abundance of the virus’s noncoding RNAs, particularly the two Epstein-Barr virus-encoded small RNAs known as EBER1 and EBER2, which accumulate to very high copy numbers inside the nuclei of infected cells. A new computational reanalysis of a public multi-omics dataset now suggests that these abundant viral RNAs are associated with a peculiar promoter-level state in EBV-positive B cells: host promoters in physical contact with EBERs remain chromatin-accessible, yet they are markedly less likely to undergo a coordinated recovery of RNA polymerase II looping and transcriptional output when the viral RNAs are depleted.
The study, published in Epigenetics & Chromatin, reanalyzed the publicly available dataset GSE281522, which was generated in EBV-positive GM12878 cells and combines genome-wide RNA-DNA contact mapping, RNA polymerase II ChIA-PET chromatin interaction maps, RNA-seq transcriptome measurements, and ATAC-seq accessibility profiles collected under EBER depletion and matched control conditions. No new experiments were performed; the entire analysis is a promoter-centered integration of existing public data. Because the source dataset does not resolve EBER1 and EBER2 separately, all references to EBER contact and EBER depletion in the work refer to aggregate measurements and a non-isoform-resolved perturbation.
The analytical core of the paper is a four-state classification of promoter behavior after EBER depletion. Each promoter was labeled as showing no response, loop recovery only, expression upregulation only, or the coordinated Recovery + Up state, meaning a concurrent increase in RNA polymerase II-associated promoter looping and gene expression. Loop recovery was called when the log2 fold-change in looping reached at least 1, and expression upregulation when the log2 fold-change in transcripts per million reached at least 1.
The four-state framework was designed to capture something that neither looping nor expression measures alone can: the concordance between architectural and transcriptional response at the same promoter. When the authors examined how transcriptional upregulation alone was distributed across loop-change bins, they found that expression gains occurred across a broad range of loop fold-changes, whereas the coordinated Recovery + Up response was strongly gated by positive loop change, emerging primarily when loop gains reached moderate-to-strong levels. This pattern supported treating Recovery + Up as a distinct biological endpoint rather than a trivial overlap between two marginal responses. The framework also guarded against a simple confound: mean expression in the depletion versus control conditions largely tracked the diagonal for both EBER-contacted and non-contacted promoters, indicating broadly comparable overall expression levels and no obvious global bias that would artificially push one group into fewer upregulated calls.
The central quantitative result was striking in its magnitude. Recovery + Up promoters were markedly less frequent among EBER-contacted genes, at 0.42 percent, than among non-contacted genes, at 2.00 percent, corresponding to an odds ratio of approximately 0.21 and a Fisher’s exact p-value of approximately 6.3 × 10⁻⁵. In other words, EBER contact was not associated with a subtle shift in mean response but with a pronounced reduction in the probability of entering the joint loop-and-expression recovery state. When contacted promoters were stratified into weak and strong contact groups using a median split on an anchor-level contact-intensity metric, a stepwise gradient emerged: Recovery + Up frequency decreased from 2.0 percent in non-contacted promoters to 0.6 percent in weakly contacted and 0.3 percent in strongly contacted promoters. The authors were careful, however, to interpret these grouped comparisons descriptively rather than as definitive evidence for a stable monotonic dose relationship, because additional robustness analyses using continuous-contact models, contacted-only decile-trend models, and contacted-only nonlinear models did not support a consistent within-contact monotonic gradient.
A major strength of the study lies in the breadth of its sensitivity analyses. The depletion of Recovery + Up among EBER-contacted promoters was recalculated across a grid of 49 alternative loop-recovery and transcriptional-upregulation threshold combinations, and all 49 yielded odds ratios below 1, with a median odds ratio of 0.337 and a range of 0.000 to 0.932; 32 of the 49 tests reached nominal significance and 28 survived Benjamini-Hochberg adjustment. The signal also persisted under seven alternative loop-support cutoffs and seven expression filters, with all cutoffs in each family producing odds ratios below 1. Critically, recomputing promoter-contact assignment using transcription start site-centered windows of ± 1 kb, ± 2 kb, and ± 5 kb preserved the depletion, with odds ratios of 0.416, 0.208, and 0.110 respectively. These results indicated that the core depletion signal was not an artifact of a single cutoff set or promoter-window definition.
Beyond the depletion itself, the authors asked whether EBER contact modifies the functional relationship between loop recovery and transcriptional upregulation. Using a Firth bias-reduced logistic regression model that included a LoopRecovery-by-contact interaction and adjusted for baseline expression and baseline loop signal, they found that loop recovery strongly increased the odds of transcriptional upregulation in non-contacted promoters, with an odds ratio of 2.72 and a 95 percent confidence interval of 1.11 to 6.63. In weakly contacted promoters, this loop-expression coupling was near-null, with an odds ratio of 1.02, and it remained attenuated in strongly contacted promoters, with an odds ratio of 1.20. Within the contacted set, Recovery + Up promoters also showed significantly lower EBER contact strength than other contacted promoters, with a Wilcoxon p-value of 0.0041, indicating that coordinated recovery preferentially occurs at the weak-contact end of the contacted distribution.
To rule out the possibility that this phenotype was a generic property of RNA-associated contact maps, the authors constructed promoter-contact summaries for three comparison RNAs present in the same dataset: the host long noncoding RNA MALAT1 and the viral RNAs miR-BART and v-snoRNA1. None of the control RNAs reproduced a uniform EBER-like depletion pattern. In the MALAT1 and miR-BART comparisons, Recovery + Up frequencies were lower in the EBER-only and Both groups than in reference groups, but the adjusted contrasts were not significant, while the v-snoRNA1 comparison showed a distinct pattern in which the v-snoRNA1-only group had a higher Recovery + Up probability than the Neither group. Baseline-adjusted and matched analyses remained directionally consistent with the primary model, although effect sizes were attenuated in matched comparisons, which the authors interpreted cautiously as evidence of baseline-dependent confounding rather than a reversal of the primary signal.
The pathway-level analyses converged on a specific biological context. Promoters were ranked by continuous EBER contact burden, and the ranked list was tested by preranked gene set enrichment analysis; a complementary over-representation analysis was performed among the top 5, 10, and 20 percent of genes ranked by contact burden. Across both approaches, the Hallmark set TNFA_SIGNALING_VIA_NFKB emerged as the most reproducible pathway-level signal, with additional inflammatory or stress-associated pathways showing weaker or more context-dependent enrichment. The authors emphasized that the pathway results should be interpreted narrowly, with TNFA/NFκB signaling rather than inflammatory pathways in general treated as the most reliable context for the constrained promoter state.
Connecting pathway enrichment to individual loci, the authors prioritized accessible constraint-core promoters by EBER contact burden and ATAC support. This identified a set of candidate promoter-hub loci including RIPK2, IL15RA, GCH1, ICAM1, CSF1, SOCS3, IFIH1, TNFAIP3, IFNGR2, BTG1, CCL5, TAP1, NFKBIA, IL12B, IL6ST, JUN, KLF6, NFKB1, MYC, and TNFSF9. In the final candidate inventory, 152 TNFA/NFκB-linked ATAC-accessible genes and 144 multi-inflammatory ATAC-accessible genes were retained, providing a focused set of loci for downstream mechanistic testing. An external cytokine-response analysis using differential expression tables from GM12878 cells treated with IL-15 provided additional context-specific support: constraint-core genes were depleted among strongly upregulated IL-15-responsive genes across multiple fold-change and adjusted p-value thresholds, a result the authors interpreted as condition-specific validation rather than evidence of a universal inducible-response phenotype.
Perhaps the most important interpretive question was whether the constrained class simply reflected inactive or closed chromatin. To address this, the authors intersected promoter windows with internal ATAC-seq peak calls from the same study. The results argued strongly against promoter closure: ATAC peak overlap was detected in any replicate for 89.38 percent of constraint-core promoters versus 68.38 percent of non-core comparator promoters, in a majority of replicates for 88.19 percent versus 63.58 percent, and in all replicates for 87.07 percent versus 59.27 percent. These differences corresponded to odds ratios of 3.89, 4.28, and 4.63 for the three replicate-consistency definitions. Replicate-averaged ATAC signal was also higher in the constraint-core class, with a median of 42.55 compared with 10.43 in comparator promoters, a median difference of 32.11 with an FDR of 5.98 × 10⁻⁵². The constrained promoter class was therefore not explained by lack of chromatin accessibility, and the authors concluded that EBER-linked promoter constraint reflects altered activation competence rather than passive promoter inactivity.
To illustrate the promoter-hub interpretation at the level of individual loci, the authors visualized the GMEB1 neighborhood using EBER RNA-DNA contact arcs and replicate-level anchor-density tracks, and then represented the same locus as force-directed RNA polymerase II promoter-contact networks under control and EBER-depleted conditions. In these networks, nodes represent promoter or loop-anchor regions and edges represent RNAPII-mediated contacts, with annotations distinguishing focal promoters, direct and indirect neighbors, replicate-supported edges, and single-replicate edges. The authors stressed that these panels are topology visualizations of contact graphs and not physical reconstructions of chromatin architecture, and that the locus is shown as an illustrative example rather than the source of the global statistics.
Several limitations frame the conclusions. All analyses are computational reanalyses of public data from a single EBV-positive B-cell system, so generalization across latency states, cellular backgrounds, and perturbation contexts remains untested. The source dataset does not resolve EBER1- and EBER2-specific contacts or selective depletion effects, so the observed phenotypes cannot be attributed to either transcript specifically, even though prior recombinant-virus work has shown that the two EBERs influence partially distinct host gene-expression programs. The ATAC analysis supports accessibility but does not establish that accessibility is causally responsible for the constrained phenotype. The mechanistic explanations the authors propose, involving RNA- and RNA-binding-protein-rich transcription compartments or sequestration of bridging factors that convert loop recovery into productive transcription, remain inferential rather than directly demonstrated, though they generate concrete, testable predictions about EBER-associated ribonucleoproteins and promoter-hub competence.
Subject of Research: EBER-associated RNA-chromatin contacts and constrained RNAPII promoter hubs in EBV-positive B cells
Article Title: EBER-associated promoter contacts mark accessible but transcriptionally constrained RNAPII promoter hubs in EBV-positive B cells
Article References: Thangameeran, S. I. M., Mani, H., & Peng, C.-W. (2026). EBER-associated promoter contacts mark accessible but transcriptionally constrained RNAPII promoter hubs in EBV-positive B cells. Epigenetics & Chromatin, 19(1), Article 42. https://doi.org/10.1186/s13072-026-00688-9
Image Credits: AI Generated
DOI: 10.1186/s13072-026-00688-9
Keywords: Epstein-Barr virus, EBER, noncoding RNA, RNAPII, promoter looping, ChIA-PET, ATAC-seq, GM12878, NF-kappa-B, chromatin architecture, gene regulation, reanalysis
News Source: Juliet Wilcox. (October 30, 2026). Viral RNA contacts leave immune gene promoters accessible yet transcriptionally constrained. Scienmag.
