Microbial communities may be carrying a hidden record of the temperatures and nutrient conditions in the environments where they live. A new study published in Nature Microbiology reports that environmental temperature can be predicted from microbial DNA composition alone, even when samples come from very different ecosystems. Using tetranucleotide frequencies—the relative abundance of four-letter DNA sequences—the researchers achieved a prediction accuracy of (R^2 = 0.75) across 1,235 marine and soil metagenomes. The result suggests that microbial genomes contain broad, ecosystem-spanning signatures associated with temperature, offering a new way to investigate how microscopic life responds to environmental change.
The study, led by T. Antman, O. Lewin-Epstein and T. Yerushalmi, focuses on metagenomes: collections of DNA recovered directly from environmental samples rather than from isolated organisms grown in the laboratory. This approach captures the genetic composition of entire microbial communities, including bacteria and other microorganisms that remain difficult or impossible to culture. Instead of examining individual genes or identifying particular species, the researchers analyzed the overall structure of the DNA sequences present in each sample. Their central question was whether this composition could reveal the temperature of the environment from which the genetic material originated.
The key measurement was tetranucleotide frequency. DNA is built from four bases—adenine, thymine, guanine and cytosine—and a tetranucleotide is a sequence of four consecutive bases, such as GATC or AAGG. There are 256 possible tetranucleotides, and their frequencies can reflect multiple biological and evolutionary processes, including genome composition, mutation patterns, DNA repair, replication, and selection. Because these short sequences occur throughout genomes, their combined distribution can provide a statistical fingerprint of microbial communities. The researchers used these fingerprints as input for a machine-learning model trained to estimate environmental temperature.
Across marine and soil metagenomes, the model captured a strong relationship between DNA composition and temperature. An (R^2) value of 0.75 means that the predictions explained approximately 75 percent of the variation in measured environmental temperatures within the analyzed dataset. Such a result does not mean that temperature is encoded in a single sequence or that the model can identify temperature perfectly in every sample. Rather, it indicates that thousands of small differences in tetranucleotide usage collectively form a reliable signal. The finding is notable because the samples represented fundamentally different ecological settings, with distinct communities, nutrient regimes and evolutionary histories.
The researchers also found that the temperature signal was visible within individual taxa. This observation is important because a pattern found only at the community level could arise simply from changes in which organisms are present. If warm environments contain one set of organisms and cold environments contain another, a model might predict temperature by recognizing taxonomic turnover rather than a common biological response. The presence of the signal within individual groups is consistent with the possibility that temperature-associated DNA composition reflects a more fundamental genomic or physiological pattern. However, the findings do not establish that temperature directly causes every observed sequence bias, and the authors emphasize that environmental variables can interact in complex ways.
One of the study’s most revealing contrasts involved GC content, the proportion of guanine and cytosine bases in DNA. GC content is often used as a broad genomic characteristic, but its relationship with temperature was not consistent across ecosystems. In soil samples, GC content increased with temperature, whereas in marine samples it decreased. This opposite behavior shows why a single measure such as overall GC percentage may be insufficient for detecting universal environmental signatures. A pattern that appears strong in one habitat can be reversed in another when other ecological pressures are taken into account.
The researchers propose that nutrient availability helps explain this divergence. In the marine metagenomes, nutrient levels decreased as temperature increased, while GC content rose with nutrient availability. Under this relationship, warmer marine environments tended to be more nutrient-limited, and nutrient limitation was associated with lower GC content. Soil communities followed a different environmental structure, producing a positive temperature–GC relationship rather than the negative pattern observed in marine samples. The explanation highlights a central challenge in environmental genomics: temperature rarely acts alone. Its influence is entangled with nutrients, productivity, water chemistry, habitat structure and the composition of the microbial community.
To move beyond this ecosystem-specific conflict, the researchers examined individual tetranucleotides. They identified sequences with 50 percent GC content that displayed consistent and robust correlations with temperature across both marine and soil environments. Because these tetranucleotides have the same overall GC proportion, their associations cannot be explained simply by a rise or fall in total GC content. Instead, the precise arrangement of bases appears to matter. These sequence-level patterns may have helped stabilize the machine-learning predictions when samples from contrasting ecosystems were analyzed together. They also point toward a more detailed form of environmental genomic analysis, in which the order of bases—not only their broad chemical categories—contains ecological information.
The findings could have implications for monitoring microbial responses to global change. Microorganisms drive carbon cycling, nutrient transformations and many other processes that influence the functioning of oceans and soils. If their DNA composition shifts systematically with temperature and nutrient limitation, metagenomic surveys might help detect ecological change even when direct environmental measurements are incomplete. The approach could also assist comparisons among remote or difficult-to-sample habitats by using DNA as an indirect indicator of environmental conditions. Yet the study is best understood as evidence of association rather than a universal thermometer embedded in microbial genomes. Future work will need to test the model on new regions, seasons and habitats, determine how long these signatures persist, and separate temperature effects from the many other forces shaping microbial DNA.
By combining large-scale metagenomic data with machine learning, the study reveals that environmental information can be distributed across the architecture of microbial DNA. The strongest signal did not come from a single gene, a single species or a simple GC-content rule, but from the coordinated frequencies of many short DNA sequences. That pattern remained detectable across taxonomic and ecological boundaries, while also exposing the influence of nutrient availability on genome composition. As climate change alters temperatures and resource conditions across marine and terrestrial ecosystems, these DNA signatures may provide a powerful way to track how microbial life is reorganizing—one four-base sequence at a time.
Subject of Research: Microbial DNA signatures associated with environmental temperature and nutrient limitation across marine and soil ecosystems.
Article Title: Global microbial DNA signatures of temperature and nutrient limitation across ecosystems
Article References: Antman, T., Lewin-Epstein, O., Yerushalmi, T. et al. Global microbial DNA signatures of temperature and nutrient limitation across ecosystems. Nat Microbiol (2026). https://doi.org/10.1038/s41564-026-02451-y
Image Credits: AI Generated
DOI: https://doi.org/10.1038/s41564-026-02451-y
Keywords: microbial genomes, metagenomics, tetranucleotide frequencies, environmental temperature, nutrient limitation, GC content, machine learning, marine ecosystems, soil microbiomes, global change
Tags: cross-ecosystem microbial DNA studiesDNA sequence composition and environmental indicatorsecosystem-specific microbial DNA signaturesenvironmental temperature prediction from microbiomesmetagenomic approaches to environmental sciencemetagenomics and ecosystem analysismicrobial DNA analysismicrobial genomics and environmental monitoringmicrobial response to climate changemicrobial signatures of nutrient availabilitynutrient limitation signatures in microbial communitiestetranucleotide frequency in microbial ecology
