Salmonella Phage Genomes Show Distinct Codon Usage and GC Patterns

A comparative genomic analysis of 82 Salmonella bacteriophages has identified distinct molecular signatures separating temperate and virulent phage lifestyles. Published in Frontiers in Microbiology, the study examined 41 temperate and 41 virulent phage genomes, comprising 7,759 protein-coding genes, to elucidate differences in genome architecture, nucleotide composition, and predicted translational adaptation. The research team from the University of Georgia in Tbilisi utilized genome-aware statistical methods to account for the non-independence of genes within individual phage genomes, providing a rigorous framework for comparing these two distinct biological strategies.

The analysis revealed that virulent phages, which directly lyse host cells, generally possessed larger and more variable genomes than their temperate counterparts. The mean genome size for virulent phages was 80,784 base pairs, compared to 44,433 base pairs for temperate phages. Furthermore, virulent phages exhibited greater broad-scale genomic disconnectedness, with 81.7% of within-group genome pairs lacking significant BLAST similarity, whereas only 74.0% of temperate phage pairs showed this lack of similarity. This suggests that the virulent phage dataset contained a higher degree of genetic heterogeneity at the broad genomic level.

Nucleotide composition emerged as a major source of differentiation between the two lifestyle groups. Temperate phages displayed significantly higher genome-average overall GC content and GC3 content, which refers to the proportion of guanine and cytosine nucleotides at the third codon position. The difference in overall GC content was 4.484 percentage points higher in temperate phages, while the GC3 difference was 7.706 percentage points. These compositional differences were statistically robust, with false-discovery-rate adjusted q-values of 0.00013 for both metrics, indicating that the third codon position contributes substantially to the observed synonymous codon preferences.

Conversely, virulent phages exhibited higher values for predicted translational adaptation indices. The genome-average Codon Adaptation Index (CAI-2) was 0.0131 higher in virulent phages, and the Index of Translation Elongation (ITE) was 0.0321 higher. These associations persisted after adjusting for nucleotide composition and coding-sequence length, suggesting that the codon usage patterns in virulent phages are not solely driven by their lower GC content. However, the effective number of codons (Nc), a measure of codon usage bias, did not show a statistically significant difference between the two groups, indicating that stronger codon bias cannot be definitively attributed to the virulent lifestyle in this dataset.

Multivariate analyses further supported the differentiation between the phage groups. Genome-equal PERMANOVA detected significant differences across the complete molecular dataset and all five examined gene functional categories. The greatest effects were observed in DNA repair/recombination, lysis-associated, and structural genes. Principal Component Analysis (PCA) of the gene-level profiles showed that the first principal component was primarily associated with GC composition, where temperate phages scored higher, while the second component was associated with CAI-2, ITE, and sequence size, where virulent phages scored higher. This two-component structure highlights the complementary nature of the compositional and codon-adaptation signatures.

Relative Synonymous Codon Usage (RSCU) analyses provided codon-specific evidence of these differences. Of the 59 informative synonymous codons examined, 23 showed significant, large-effect differences between the two lifestyle groups. Twelve of these codons had higher mean RSCU values in virulent phages, while 11 showed higher values in temperate phages. A notable example was the opposing preference for cysteine codons: UGU showed higher usage in virulent phages, whereas UGC was more prevalent in temperate phages. This pattern is consistent with the contrasting GC3 content observed between the groups, illustrating how nucleotide composition influences specific codon choices.

The study employed several statistical strategies to ensure that the findings were not artifacts of gene-level dependencies. By using whole-genome permutations and genome-cluster bootstrapping, the researchers treated individual phage genomes as the units of biological inference rather than individual genes. This approach is critical because genes within the same genome are not independent observations. Sensitivity analyses, including relatedness-thinning procedures to account for closely related genomes, confirmed that the principal molecular contrasts were robust to the representation of similar phage lineages.

Correlation analyses revealed that the positive association between CAI-2 and ITE remained strong in both phage lifestyles even after adjusting for GC content. The partial Spearman correlation coefficients were 0.771 for temperate phages and 0.776 for virulent phages, with confidence intervals excluding zero. This suggests that the relationship between these translational indices is not merely a byproduct of shared nucleotide composition. However, the authors caution that these computational indices are predictions based on reference codon preferences and do not directly measure translational efficiency during infection.

Limitations of the study include the taxonomic imbalance between the two phage lifestyle groups and the lack of direct experimental measurements of gene expression and translation. The phage groups had limited taxonomic overlap, with only Sarkviridae and Drexlerviridae containing representatives of both lifestyles. This taxonomic disparity means that the observed associations cannot be exclusively attributed to phage lifestyle, as they may also reflect lineage-dependent effects. Additionally, the selection of genomes was informed by BLAST-based relatedness to ensure diversity, which may have influenced the genomic heterogeneity represented in the final dataset.

The findings provide a detailed molecular characterization of Salmonella phage lifestyle-associated signatures. While the study does not establish causal mechanisms or therapeutic suitability, it identifies complementary compositional and codon-adaptation patterns that may inform future investigations into phage biology. The distinct genomic architectures and predicted translational profiles of temperate and virulent phages offer a foundation for understanding how these viruses adapt to their respective ecological niches within the Salmonella host. Future research involving larger, taxonomically balanced datasets and experimental validation of translation dynamics will be necessary to fully elucidate the evolutionary significance of these molecular differences.

Initial screening of National Center for Biotechnology Information records evaluated 63 candidate temperate and 67 candidate virulent phages. Researchers excluded 20 temperate candidates whose lifestyles lacked experimental confirmation in peer-reviewed studies, alongside two others due to missing annotations or partial sequences. Ten virulent phage candidates were eliminated because of deficient genome annotations. Morphological assignments showed that virulent phages predominantly displayed siphovirus-like and myovirus-like structures, commonly classified within Sarkviridae, Ackermannviridae, Demerecviridae, and Straboviridae. In contrast, temperate phages featured a higher proportion of podovirus-like morphotypes, frequently grouping into Peduoviridae, Casjensviridae, or Zimmerviridae. Although all examined viruses infect Salmonella enterica, their host profiles encompassed multiple serovars, led by Typhimurium, Enteritidis, and Gallinarum. A subset of virulent phages even exhibited host ranges extending beyond Salmonella to infect species such as Escherichia coli, Shigella sonnei, and Klebsiella pneumoniae.

To describe divergence patterns without assuming phylogenetic trees, the authors formulated a combined genomic distance metric incorporating reciprocal nucleotide identity and query coverage. Pairs lacking detectable similarity received the theoretical upper bound distance of the square root of two. When comparing only alignable genome pairs that possessed detectable similarity, temperate phages demonstrated greater sequence divergence, showing a mean genomic distance of 0.741 and a median of 0.949. Virulent phage pairs with detectable alignments exhibited markedly lower divergence, with a mean distance of 0.486 and a median of 0.226. Furthermore, formal non-parametric testing confirmed substantial separation in overall genome dimensions between the two lifestyles. A Kruskal-Wallis rank-sum test returned a test statistic of 17.846 with one degree of freedom, yielding an eta squared effect size of 0.211 and an epsilon squared of 0.208. These values indicate that approximately 21 percent of the rank-based variation in phage genome size was directly associated with lifestyle category.

The investigation analyzed 2,869 genes across temperate phages and 4,890 genes across virulent phages. Virulent phage coding sequences averaged 624.16 nucleotides in length with a median of 414 nucleotides, producing larger predicted proteins averaging 23,270 daltons. In comparison, temperate phage coding sequences averaged 530.05 nucleotides with a median of 345 nucleotides and predicted proteins averaging 19,560 daltons. Temperate phages maintained significantly higher GC proportions across all positional codon measurements. Mean first-position GC content reached 54.43 percent in temperate phages versus 51.37 percent in virulent phages, while second-position GC content was 40.93 percent compared to 36.53 percent. Genome-equal permutation tests confirmed that differences across first, second, and combined first-and-second codon positions remained statistically significant. In addition, virulent phage genes exhibited a slightly elevated mean predicted protein energetic cost of 24.06 compared to 23.57 in temperate phages, which the researchers noted represents a secondary characteristic linked to expanded sequence size and amino acid composition rather than an independent energetic adaptation.

To isolate translational signals from nucleotide biases, the researchers evaluated linear regression models adjusted for overall GC, GC3, and coding sequence length, employing cluster corrections across the 82 phage genomes. Membership in the virulent lifestyle remained significantly associated with elevated Codon Adaptation Index values and higher translation elongation scores, yielding adjusted regression coefficients of positive 0.0217 and positive 0.0236, respectively. The analysis of multivariate dispersion through permutation testing revealed that only the regulatory gene category exhibited a statistically significant difference in dispersion between lifestyles, with an adjusted q-value of 0.0006. In contrast, dispersion tests for structural, lysis, replication, and DNA repair categories showed no significant differences, suggesting that centroid shifts accounted for differentiation in those modules. Furthermore, specific codon usage patterns extended beyond cysteine pairings. Virulent phages demonstrated significantly higher relative synonymous codon usage for codons including AGU, GGU, GUA, CGU, ACU, and CAA. Temperate phages exhibited pronounced preferences for synonymous GC-rich codons such as AGC, CAG, CUG, and GGC, underscoring the pervasive influence of nucleotide architecture on viral translational adaptation.

Subject of Research: Comparative genomics of Salmonella temperate and virulent phages

Article Title: Genome architecture, GC3 content, codon usage, and predicted translational adaptation of Salmonella temperate and virulent phages

Article References: Taktakishvili, K., Janelidze, D., Gamkrelidze, A., Kobakhidze, S., Elbakidze, T., & Kotetishvili, M. (2026). Genome architecture, GC3 content, codon usage, and predicted translational adaptation of Salmonella temperate and virulent phages. Frontiers in Microbiology, 17, Article 1927342. https://doi.org/10.3389/fmicb.2026.1927342

Image Credits: AI Generated

DOI: 10.3389/fmicb.2026.1927342

Keywords: Salmonella phages, bacteriophage lifestyle, genome architecture, codon usage bias, GC3 content, translational adaptation, Codon Adaptation Index, Index of Translation Elongation, relative synonymous codon usage, PERMANOVA, principal component analysis, phage evolution, genomic heterogeneity, temperate phages, virulent phages, molecular signatures, statistical genomics, Frontiers in Microbiology, University of Georgia, protein-coding genes