🧬 biology

Codon Usage Bias Shapes CpG Architecture in Millet Coding Sequences

This genome-wide analysis of six millet and related C₄ grass species reveals that synonymous codon usage bias actively shapes CpG dinucleotide architecture in coding sequences, suggesting that codon choice and arrangement preconfigure the structural substrate for gene body methylation and extend the functional role of codon bias beyond translational efficiency to include an epigenetic dimension.

Original authors: Bevin Nishanth J, Suji Somasundaram

Published 2026-07-02

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Bevin Nishanth J, Suji Somasundaram

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Imagine the DNA in a plant's cell as a massive library of instruction manuals. Each manual (a gene) tells the cell how to build a specific protein, like a recipe for a cake.

For a long time, scientists thought that the "spelling" of these recipes didn't matter much, as long as the final cake tasted the same. In genetics, different three-letter "words" (called codons) can spell out the same ingredient (amino acid). This is like having multiple ways to write the word "color" (color vs. colour) or "neighbor" (neighbor vs. neighbour).

This study, looking at six different types of millets (a group of hardy, grass-like crops), discovered that the choice of spelling does matter. It turns out that the way these plants spell their genes isn't just about making proteins; it's also about setting up a hidden "security system" inside the DNA.

Here is a simple breakdown of what they found:

1. The "Security Tags" (CpG Sites)

Inside the DNA, there are specific two-letter combinations called CpG (Cytosine followed by Guanine). Think of these as "security tags" or "sticky notes."

In plants, these tags are often used to mark genes that are important and need to run smoothly without glitches. This marking process is called gene body methylation.
However, if you have too many of these tags, the DNA can get damaged over time (like a sticky note falling off and tearing the paper). So, plants have to be very careful about where they put them.

2. The "Spelling Choice" (Codon Usage Bias)

The researchers found that millet plants are very picky about which "spelling" they use for their genes.

The "Rich" Genes: Some genes are "CpG-rich." These are like the VIP sections of the library. They use specific spellings that naturally create more of those "security tags." The study found that these genes are usually the ones the plant uses the most (like the daily bread recipes). They are also spelled in a very optimized, efficient way.
The "Poor" Genes: Other genes are "CpG-poor." They avoid those specific spellings. These are like the occasional, special-occasion recipes that don't need the same level of constant security.

3. The "Bridge" Effect (Codon Adjacency)

This is the most surprising part of the discovery.

Scientists used to think that security tags (CpG) only appeared inside a single three-letter word.
But this study found that nearly half of these tags are actually formed by the bridge between two words.
Imagine writing a sentence. If you end one word with a "C" and start the next word with a "G," you accidentally (or intentionally) create a "CG" tag right at the boundary.
The millet plants seem to be arranging their words in a specific order so that these bridges form exactly where they are needed. It's like a puzzle where the pieces fit together not just to make a picture, but to create a hidden pattern in the gaps between them.

4. The "Middle" of the Story

When the researchers looked at where these tags were located along the length of the genes, they found a consistent pattern:

The tags are sparse at the very beginning and very end of the gene.
They are dense in the middle.
This is like a book where the cover and the back page are plain, but the middle chapters are filled with highlighted notes. This matches what we know about how plants protect their most important, stable genes.

5. It's Not Just Random Chance

The researchers asked: "Is this just because the plants have a lot of Gs and Cs in their DNA, or is it a deliberate choice?"

They did a math test (called a neutrality plot) and found that about 75–80% of this pattern is due to natural selection, not just random accidents.
This means the plants are actively "choosing" these spellings to build this security architecture. It's a deliberate design, not a happy accident.

The Big Picture

The main takeaway is that the way millet plants spell their genes has a double purpose:

Translation: It helps the cell build proteins efficiently.
Epigenetics: It sets up the physical structure for the "security tags" (methylation) that keep the gene stable and running correctly.

The authors call this an "epigenetic dimension" of codon usage. In simple terms, the plant's spelling choices are pre-programming the DNA's future behavior, ensuring that the most important genes are protected and run smoothly, all by carefully arranging the letters and the spaces between them.

Technical Summary: Codon Usage Bias Shapes CpG Architecture in Millet Coding Sequences

Problem Statement
Synonymous codon usage bias (CUB) is traditionally attributed to translational selection and mutational pressure. However, coding sequences also encode overlapping regulatory layers, including signals for epigenetic modifications. In plants, gene body methylation (gbM) is a conserved epigenetic mark associated with stable, broadly expressed genes, requiring CpG dinucleotides as substrates. While the relationship between CpG density and methylation is established, the extent to which synonymous codon choice and codon arrangement actively shape the "epigenetic dimension" of CpG architecture in plant genomes remains underexplored. Specifically, it is unclear whether codon usage in millets (C₄ grasses) is structured not only for translational efficiency but also to modulate CpG availability for potential gbM.

Methodology
The study performed a genome-wide analysis of six millet and related C₄ grass species: Echinochloa esculenta, Eleusine coracana, Panicum miliaceum, Pennisetum glaucum, Setaria italica, and Sorghum bicolor.

Data Curation: Over 300,000 high-confidence protein-coding sequences (CDS) were extracted from public repositories and filtered for canonical start/stop codons, in-frame integrity, and minimum length (300 bp).
CpG Stratification: Genes were classified into CpG-poor, CpG-moderate, and CpG-rich categories based on quartile thresholds of CpG density (CpG/kb).
Metagene Analysis: CDS were partitioned into 5′, central, and 3′ regions to map positional CpG distribution.
Codon Metrics: The study quantified Effective Number of Codons (ENC), GC content at third positions (GC3s), Codon Adaptation Index (CAI), and CpG observed/expected (O/E) ratios.
Structural Analysis: CpG motifs were categorized by their frame origin: within codons (positions 1–2 or 2–3) or across codon boundaries (inter-codon).
Selection vs. Mutation: Neutrality plots (GC12 vs. GC3s) were used to estimate the relative contributions of mutational bias versus natural selection.
Comparative Analysis: Relative Synonymous Codon Usage (RSCU) and $\Delta$ RSCU were calculated to compare codon preferences between CpG-rich and CpG-poor gene sets.

Key Results

Conserved CpG Architecture: CpG density is heterogeneous but shows a conserved enrichment in the central gene body regions (33–66% of the CDS) across all six species, mirroring known gbM profiles in grasses.
Correlation with Codon Bias: CpG-rich genes exhibit significantly stronger codon bias (lower ENC), higher translational optimization (higher CAI), and elevated GC3s compared to CpG-poor genes ( $p < 2.2 \times 10^{-16}$ ). Strong positive correlations exist between CpG density, GC3s, and CAI.
Independence from GC3s: While GC3s influences CpG density, substantial variation in CpG architecture persists within fixed GC3s bins, indicating that factors beyond simple nucleotide composition (i.e., specific codon choice and order) drive CpG distribution.
Dominance of Inter-codon CpGs: Codon-frame analysis revealed that nearly half (46–49%) of all CpG motifs arise across codon boundaries (inter-codon), highlighting the critical role of codon adjacency in shaping the CpG landscape.
Selection-Driven Patterns: Neutrality plots showed low regression slopes (0.20–0.25), indicating that approximately 75–80% of codon usage variation is driven by selection rather than mutational bias.
Enrichment of CpG-Forming Codons: CpG-rich genes are systematically enriched for CpG-forming codons (e.g., CGC, CCG, ACG, TCG, GCG) and depleted for non-CpG synonymous counterparts, a pattern consistent across all species.

Key Contributions

Epigenetic Dimension of CUB: The study provides evidence that synonymous codon usage in millets encodes an "epigenetic dimension," where codon choice and adjacency preconfigure the structural substrate for gene body methylation.
Codon Adjacency: It quantifies the significant contribution of inter-codon CpG formation, demonstrating that codon order is a major determinant of CpG architecture, not just codon identity.
Cross-Species Conservation: The findings establish that the coupling between codon bias, GC3s, and CpG architecture is a conserved feature across diverse millet and C₄ grass genomes.

Significance and Claims
The authors claim that synonymous codon usage in millet genomes is not solely a result of translational optimization but is also structured to modulate CpG dinucleotide availability. This suggests that coding sequences encode dual layers of information: protein sequence and the potential for epigenetic regulation (gbM). The study posits that CpG-rich, highly biased genes may be selected for stable, low-noise expression profiles facilitated by gbM.

The paper explicitly notes that it does not measure DNA methylation directly; rather, it quantifies the "CpG substrate" that permits gbM. The significance lies in extending the functional role of CUB beyond translation to include the shaping of epigenetic landscapes. The authors suggest that future integration with whole-genome methylome data is necessary to validate whether these CpG-rich architectures correspond to actual gbM states in millets. The work supports the hypothesis that selection acts on codon pairs and CpG-forming dyads to maintain or avoid CpG motifs, influencing chromatin organization and gene regulation.