Global DNA signatures of temperature and nutrient limitation in prokaryotes

Abstract

Microorganisms adapt to their environment through changes in both their genes and overall genome composition. However, identifying universal principles of genomic adaptation remains challenging because different environments present multiple, overlapping selective pressures. To overcome this challenge, we analyzed DNA composition patterns across diverse environments using machine learning. By examining tetranucleotide frequencies from 1,112 marine and soil metagenomic samples, we discovered that environmental temperature can be accurately predicted from DNA composition alone (R2=0.82). This temperature signal remained robust even when analyzing individual bacterial phyla and classes, suggesting a fundamental adaptive response. To examine this adaptation mechanism, we analyzed GC content relationships with temperature and found opposing relationships in different environments: positive correlations in soil but negative correlations in marine samples. We show that this inverse relationship in marine environments is driven by nutrient availability, as GC content increases with nutrient levels while nutrients decrease with temperature in marine ecosystems. Despite these contrasting GC patterns, we identified specific tetranucleotides composed of equal numbers of GC and AT bases (50% GC content) that showed consistent temperature correlations across all environments. These findings reveal a complex interplay between temperature adaptation and nutrient limitation in shaping microbial genomes.