The way our genes are regulated—switched on or off in different cells—is far more complex than the basic genetic code itself. One reason is that there are over 1,600 proteins called transcription factors (TFs) that often work together to control gene activity. These partnerships between TFs are essential for determining what type of cell something becomes and how that cell performs its specific tasks. A key part of this process involves “enhancers”— the short regions of DNA that act like switches to turn genes on or off. The enhancers respond to signals from multiple TFs.
In this study, we used high-throughput technology CAP-SELEX (consecutive-affinity-purification systematic evolution of ligands by exponential enrichment) [1] to systematically map DNA-guided TF-TF interactions across more than 58,000 human TF pairs, identifying 2,198 interacting pairs [2]. The affinity of individual TFs for composite motif (the motif, found in the complex of at least two TFs bound to the same short DNA) was lower than their individual motifs, indicating a crucial rule of TF cooperativity in vivo. We also observed that these interactions often happen between TFs from different families, creating a complex and cooperative network.
To confirm our findings, we developed a new technique called mixture-SELEX, which successfully validated more than half of the composite DNA sequences we identified. These composite motifs were especially common in DNA regions that are active in specific cell types, are conserved through evolution, and often involve TFs that are active at the same time in development.
We further tested these DNA switches in developing mouse embryos and found that the composite motifs could drive strong and specific gene activity, confirming their importance in real biological settings [2].
To understand how these TFs physically interact with each other and DNA, we used classical 3D structural analysis. We solved 7 structures of TF:TF-DNA complexes using data collected at ESRF, at beamline ID23-1. These structures revealed that the contacts between TFs are often rather weak and most likely formed after TFs bound to their specific DNA sequence. Those contacts are hard to predict, even with modern AI tools like RoseTTAFold and AlphaFold. While these tools can predict the general shapes of TF-DNA complexes, they still struggle to accurately model the conformations of the amino acid side chains forming the contacts [2].
Our study helps explain a long-standing mystery: how similar TFs can lead to very different biological outcomes. It turns out that the way TFs pair up on DNA plays a critical role in determining their function. In summary, we show that TF pairing, guided by specific DNA sequences, is a key mechanism for precise control of gene activity during development.
A. Popov (ESRF), I. Sokolov (Cambridge University), E. Morgunova (Karolinska Institutet)
[1] Jolma A, Yin Y, Nitta KR, Dave K et al. (2015) Nature, 527, 384-388
[2] Xie Z, Sokolov I, Osmala M, Yue X et al. (2025) Nature, 641, 1329-1338
Figure 1: A. Schematic description of the high-throughput CAP-SELEX process. B. Composite motifs enable efficient integration of signals and high cooperativity. C. The indicated FOXK2–ELF1 composite motif detected by CAP-SELEX. D. Composite motifs of PROX–HOX2 drive highly specific expression of lacZ reporter gene in transgenic E11.5 mouse embryos. E. Both RoseTTAFold2NA (red) and AlphaFold3 (blue) fail to correctly position the side chains of the amino acids that form the contacts between the shown cooperative TF pairs.