The genetic code specifies proteins through codons, while a parallel regulatory code specifies where transcription factors (TFs) bind DNA. Stergachis et al. (Science, 2013) reported that these two codes overlap extensively in human genomes: roughly one in seven codons resides within DNase I-detected TF footprints, forming dual-use codons ("duons") whose preferential TF occupancy tracks genome-wide codon bias, whose third positions show excess evolutionary constraint, and 13.5% of GWAS disease- and trait-associated coding variants fall within them. Whether exonic TF binding genuinely directs codon choice and constrains protein evolution — and with what mechanistic roles and disease consequences — remains unresolved and contested. A 2015 reanalysis attributed the constraint signal to base-composition ascertainment bias and found no excess purifying selection from TF occupancy, while functional work in plants, mammals, and cancer genomes increasingly establishes that coding exons can act as bona fide regulatory elements. The open problem is to determine, quantitatively and causally, whether and to what extent TF occupancy within coding sequence shapes synonymous codon choice and coding-sequence evolution in mammals, through which mechanisms, how this force partitions against mutation bias (notably CpG hypermutability and GC-related processes), splice-regulatory and translation-related selection, and whether duon overlap causally contributes to human coding-variant disease risk.

Background

In protein-coding genes, each amino acid is encoded by one of several synonymous codons (nucleotide triplets), and the genome-wide, highly non-random choice among them is called codon usage bias. In organisms with large effective population sizes, codon bias is largely explained by selection for translational efficiency and mRNA stability, but these mechanisms account for only a small fraction of the codon preferences seen in mammals, which nevertheless show signatures of selection. Separately, gene expression is controlled by transcription factors (TFs) — proteins that bind short, specific DNA recognition sequences to activate or repress transcription. Stergachis, Haugen, Shafer, Fu, Vernot, Reynolds, Raubitschek, Ziegler, LeProust, Akey, and Stamatoyannopoulos (Science 342:1367–1372, 2013) mapped in vivo TF occupancy genome-wide using DNase I footprinting, a method in which preferential protection of bound DNA from DNase I cleavage, read out by deep sequencing, pinpoints occupied TF recognition sites at single-base resolution. Across more than 80 human cell types, they found that approximately 15% of protein-coding codons lie within TF footprints, and coined the term "duon" for a codon that simultaneously specifies an amino acid and participates in a TF recognition site. They reported that the codons preferred genome-wide tend to be the ones preferentially occupied by TFs, that synonymous third positions within footprints are under excess evolutionary constraint relative to matched positions outside footprints, and that 13.5% of common disease- and trait-associated single-nucleotide variants from genome-wide association studies (GWAS, which statistically associate common variants with traits across populations) fall within duons. Arginine was a notable exception, illustrating a tension the authors highlighted: CpG dinucleotides (cytosine followed by guanine) are frequent in regulatory DNA and thus favored by TFs, but they mutate at an elevated rate, opposing the preferential use of CpG-containing arginine codons. The paper left open six connected issues: the true census of coding bases under TF occupancy (footprint discovery grows with sequencing depth and showed little saturation across the 81 sampled cell types); the mechanistic roles of exonic TF occupancy (transcriptional regulation, modulation of alternative splicing — the process producing different mRNAs from one gene, translation efficiency, or mRNA stability); the interplay between TF-imposed constraint, CpG mutational dynamics, and codon usage; the causal contribution of duon overlap to disease-associated coding variants (regulatory disruption versus altered protein structure); whether gene-body CpG methylation (methylation of cytosines within transcribed regions, correlated with occupancy in the study) can causally tune opportunistic exonic TF binding in vivo; and quantitative modeling of how TF-imposed codon biases interact with other forces such as transfer-RNA (tRNA) abundances and translational selection. The 2013 claims were soon contested: Xing and He (Molecular Biology and Evolution, 2015) reanalyzed the data and argued the excess constraint on TF-bound codons was an artifact of confounding by base composition — comparing G/C and A/T sites separately, TF-bound and TF-depleted codons showed similar conservation, implying largely no extra purifying selection from TF occupancy, with exonic TF binding mostly passive and evolutionarily neutral — and Agoglia and Fraser (MBE, 2016) found no selective constraint attributable to enhancer activity in three exonic enhancers. Conversely, mechanistic support for dual-function coding has grown: duons in the coding sequence of C4-photosynthesis genes control bundle-sheath-specific expression in plants (Reyna-Llorens et al., PNAS, 2018); codon usage regulates human KRAS expression at both transcriptional and translational levels (Fu, Dang, Counter, and Liu, JBC, 2018); and a genome-scale analysis integrating TF binding, chromatin accessibility, reporter assays, and CRISPR inactivation established coding-exon enhancers as a widespread, cis-regulatory, disease-relevant class of dual-function elements (Mouren, Torres, van Ouwerkerk, Manosalva, Gallardo, Spicuglia, and Ballester, Nature Communications, 2026). Meanwhile, independent work attributes much mammalian synonymous-site constraint to competing forces: exonic splice enhancers (short coding-sequence motifs that recruit splicing factors) under strong selection cover roughly 15–20% of fourfold-degenerate sites (synonymous positions, typically the third codon base, at which all four nucleotides encode the same amino acid) (Savisaar and Hurst, Genome Research, 2018), and a 240-mammal alignment shows 20.8% of such sites constrained with a strong G/C bias, consistent with the "unwanted transcript hypothesis" of selection for GC content and against splicing-disrupting mutations rather than TF binding (Christmas, Dong, Meadows, Kozyrev, and Lindblad-Toh, Nature Communications, 2025). What remains missing is a causal, quantitative partitioning of these forces.

Problem Statement

Determine, with causal and quantitative rigor, whether and to what extent transcription-factor occupancy within mammalian protein-coding sequence — dual-use codons or "duons" (Stergachis et al. 2013) — directs synonymous codon choice, constrains coding-sequence and protein evolution, and contributes to human disease. The source leaves this open across five connected aspects: (1) CENSUS: an estimate of the fraction of coding bases and codons under TF occupancy that quantifies how far the depth- and cell-type-limited DNase I footprint maps of 2013 fall short of saturation of coding TF footprint discovery; (2) CAUSALITY OF CONSTRAINT: whether TF-bound codons genuinely experience purifying selection (the evolutionary removal of deleterious variants) beyond protein-level constraint — explicitly controlling for the base-composition ascertainment bias raised by Xing and He (2015) and for competing forces from splice regulation and GC-related mutation/selection; (3) MECHANISM: which regulatory roles (transcriptional regulation, alternative-splicing modulation, translation efficiency, mRNA stability) exonic TF occupancy actually serves, established causally rather than inferred from correlation alone; (4) INTERPLAY: how TF-imposed preference interacts with CpG hypermutability (notably for arginine codons) and with tRNA abundance and translational selection, in a quantitative model that partitions observed codon usage among these forces, including whether gene-body CpG methylation causally tunes opportunistic exonic binding in vivo; and (5) DISEASE: whether coding-variant pathogenicity at duon-overlapping sites (including GWAS and cancer coding variants) arises from regulatory disruption, altered protein structure, or both, with the regulatory and protein-structural routes distinguished. Each aspect is a measurement of the same underlying object — the superposition of the regulatory code on the genetic code.

Completion standard: the deliverable is a single, unified causal-and-quantitative determination of whether and how TF occupancy within coding sequence shapes synonymous codon choice, coding-sequence evolution, and human disease risk; the five aspects are facets of that one object rather than five separately required results, so a rigorous, well-controlled result on any single aspect is a recognized partial contribution but does not by itself resolve the task.

References

  1. Stergachis AB, Haugen E, Shafer A, Fu W, Vernot B, Reynolds A, Raubitschek A, Ziegler S, LeProust EM, Akey JM, Stamatoyannopoulos JA. Exonic Transcription Factor Binding Directs Codon Choice and Affects Protein Evolution. Science 342(6164):1367-1372 (2013). DOI: 10.1126/science.1243490. https://doi.org/10.1126/science.1243490. PMID 24337295; PMCID PMC3967546.
  2. Xing K, He X. Reassessing the "duon" hypothesis of protein evolution. Molecular Biology and Evolution 32(4):1056-1062 (2015). DOI: 10.1093/molbev/msu409. https://doi.org/10.1093/molbev/msu409
  3. Agoglia RM, Fraser HB. Disentangling Sources of Selection on Exonic Transcriptional Enhancers. Molecular Biology and Evolution 33(2):585-590 (2016). DOI: 10.1093/molbev/msv234. https://doi.org/10.1093/molbev/msv234
  4. Yadav VK, Smith KS, Flinders C, Mumenthaler SM, De S. Significance of duon mutations in cancer genomes. Scientific Reports 6:27437 (2016). DOI: 10.1038/srep27437. https://doi.org/10.1038/srep27437
  5. Reyna-Llorens I, Burgess SJ, Reeves G, Singh P, Stevenson SR, Williams BP, Stanley S, Hibberd JM. Ancient duons may underpin spatial patterning of gene expression in C4 leaves. Proceedings of the National Academy of Sciences USA 115(8):1931-1936 (2018). DOI: 10.1073/pnas.1720576115. https://doi.org/10.1073/pnas.1720576115
  6. Fu J, Dang Y, Counter C, Liu Y. Codon usage regulates human KRAS expression at both transcriptional and translational levels. Journal of Biological Chemistry 293(46):17929-17940 (2018). DOI: 10.1074/jbc.ra118.004908. https://doi.org/10.1074/jbc.ra118.004908
  7. Savisaar R, Hurst LD. Exonic splice regulation imposes strong selection at synonymous sites. Genome Research 28(10):1442-1454 (2018). DOI: 10.1101/gr.233999.117. https://doi.org/10.1101/gr.233999.117
  8. Bergman S, Tuller T. Widespread non-modular overlapping codes in the coding regions. Physical Biology 17(3):031002 (2020). DOI: 10.1088/1478-3975/ab7083. https://doi.org/10.1088/1478-3975/ab7083
  9. Kubota N, Suyama M. Functional variants in hematopoietic transcription factor footprints and their roles in the risk of immune system diseases. bioRxiv (2021). DOI: 10.1101/2021.03.22.436360. https://doi.org/10.1101/2021.03.22.436360 (preprint; not peer reviewed)
  10. Christmas MJ, Dong MX, Meadows JRS, Kozyrev SV, Lindblad-Toh K. Interpreting mammalian synonymous site conservation in light of the unwanted transcript hypothesis. Nature Communications 16:2007 (2025). DOI: 10.1038/s41467-025-57179-w. https://doi.org/10.1038/s41467-025-57179-w
  11. Zhang J, Qian W. Functional synonymous mutations and their evolutionary consequences. Nature Reviews Genetics 26(11):789-804 (2025). DOI: 10.1038/s41576-025-00850-1. https://doi.org/10.1038/s41576-025-00850-1
  12. Mouren JC, Torres M, van Ouwerkerk A, Manosalva I, Gallardo F, Spicuglia S, Ballester B. Exonic enhancers are a widespread class of dual-function regulatory elements. Nature Communications 17:4755 (2026). DOI: 10.1038/s41467-026-71220-6. https://doi.org/10.1038/s41467-026-71220-6