What do unidirectional transcriptional regulatory elements reveal about enhancer evolution and CTCF positioning?

From a binary marker to a directional spectrum

Earlier work established divergent transcription as a robust marker of active TREs, with enhancer RNAs arising bidirectionally and CTCF/cohesin acting as barriers against antisense initiation at divergent promoters [1][2]. A 2024 plant-versus-animal comparison reinforced this as a metazoan-specific pattern: human and Drosophila promoters show bimodal histone modification around the TSS consistent with bidirectional initiation, whereas Arabidopsis, soybean, and rice show unidirectional PTM enrichment [3]. The new study replaces the binary divergent/unidirectional split with a continuous directionality index derived from plus- versus minus-strand PRO-cap reads, revealing a spectrum in which some divergent elements are strongly strand-skewed [1]. This matters because it shows that 'divergent' is not a uniform class and that strand imbalance is graded, not categorical.

Unidirectional elements survive mappability and depth controls

The most obvious objection to unidirectional elements is that the missing second peak reflects poor mappability or insufficient sequencing depth. The authors tested both: mappability was higher than random genomic background with no strand bias, and downsampling PRO-cap from ~45 million to 1 million uniquely mapped reads showed that below ~10 million reads many divergent elements are misclassified as unidirectional [1]. A ProCapNet deep learning model trained on sequence confirmed that ~20 million reads are needed for stable binary classification [1]. These controls are what convert unidirectional elements from a possible artifact into a defensible biological category, and they set a practical sequencing-depth floor for anyone attempting similar classifications.

Weaker chromatin, stronger CTCF: a distinct architecture

Unidirectional elements show significantly lower PRO-cap, DNase-seq, and H3K27ac signals than divergent counterparts, and they occupy a smaller functional region [1]. Across 310 ENCODE ChIP-seq targets, divergent elements are enriched for most proteins including Pol II and general transcription factors, whereas CTCF and cohesin components RAD21 and SMC3 bind more strongly at unidirectional elements [1]. Position-resolved profiling identified four binding clusters at divergent proximal elements, including repressor-like clusters on the lower-transcription strand, while unidirectional elements show a compressed architecture with cluster 2 proteins aligned to the PRO-cap side [1]. Core promoter content, however, is similar: TATA-box, Inr, and DPR motifs appear at comparable or slightly higher proportions at unidirectional TSSs and maximum TSSs of divergent elements than at minimum TSSs [1]. The interpretation is that the difference between the two classes lies in overall architecture and factor positioning, not in the sequence grammar of individual initiation units.

Younger sequence, weaker constraint, and a possible evolutionary trajectory

Using syntenic block aging across the UCSC 100-way alignment, the authors found that both classes are older than random genomic background, but unidirectional elements are slightly younger than divergent ones, and within both classes the higher-PRO-cap side tends to be older [1]. PhyloP conservation peaks at the open-chromatin center for both classes, but the unidirectional peak is narrower, matching the smaller functional footprint seen in ChIP-seq [1]. CDTS, which reflects ongoing purifying selection in the human lineage, is lower on the higher-PRO-cap, older side, indicating stronger intolerance to variation there [1]. The authors interpret this as a possible evolutionary progression from unidirectional to divergent distal elements, but they explicitly note that sequence age provides only an upper bound on when regulatory activity arose and that future PRO-cap data from diverse species would be needed for conclusive evidence [1]. This is a hypothesis about directionality evolution, not a demonstrated trajectory.

CTCF and ZNF143 as position-dependent brakes and tuners

The mechanistic core of the paper comes from degron systems coupled to PRO-cap. CTCF degradation upregulated opposite-strand transcription at 3–5% of unidirectional elements, supporting a TF-blocking model in which CTCF restricts bidirectional activity at a subset of elements [1]. ZNF143 perturbation produced position-dependent effects: deleting motifs downstream of the minimum TSS changed transcription on that side, while deleting motifs upstream of the maximum TSS changed transcription on the other side [1]. This dual, position-dependent behavior aligns with an independent report of spatial grammar for NRF1, YY1, and NFY using csRNA-seq, Start-seq, and TSS-MPRA, though the authors argue their PRO-cap and PINTS pipeline offers better sensitivity and TSS resolution than the earlier tools [1]. ZNF143's link to CTCF-anchored chromatin loops had been noted previously [4], but the new work adds directionality as a functional readout of that relationship.

Where the conclusions stop

Several boundaries limit how far these findings travel. The analysis is based on mammalian genomes and cell lines, primarily K562 for the discovery work, so generalization to non-mammalian systems is not supported; the plant literature shows fundamentally different directionality rules at TSSs and CREs [3]. The authors acknowledge that stringent filtering excluded unidirectional elements not overlapping DNase peaks and elements with other TREs within 500 bp, which may remove biologically meaningful cases [1]. In fully adjusted regression models that included PRO-cap total counts, GC content, and mappability, directionality classification remained significant for DNase, H3K27ac, and target-gene number, but phyloP and CDTS associations were not robust to adjustment, and the authors caution that PRO-cap count may act as a partial mediator [1]. The CTCF and ZNF143 effects are perturbation-based and in cell lines, not in vivo developmental or organismal contexts. The broader literature on unidirectional systems in materials science and engineering [5][6][7][8][9][10][11] is methodologically unrelated and does not inform these biological claims.

About These Sources

This research page is built on 11 peer-reviewed studies — published from 2015 to 2026, 6 from 2024 or later, collectively cited 486 times — selected as the most relevant from 11 studies that passed quality screening, drawn from 48 papers retrieved from a database of over 500 million.

Sources used in this answer