Functional analysis of natural variation in the RNA-binding protein CsrA across the bacterial domain predicts...

Bacteria employ sophisticated post-transcriptional regulatory mechanisms to adapt to environmental changes. Carbon storage regulator A (CsrA), a highly conserved RNA-binding protein, serves as a critical post-transcriptional regulator by typically recognizing GGA-containing hairpin loops in target mRNAs and repressing translation. However, how this conserved regulator evolved diverse species-specific regulatory networks remains unclear. We developed Swarm-seq, a high-throughput platform assessing CsrA homologs across the bacterial domain for regulating flagella-dependent swarming in Bacillus subtilis. Testing over five-hundred codon-optimized csrA homologs revealed functional divergence, partitioning CsrAs into two broad classes. Class I (CsrAHp, CsrASm, RsmNPa) strongly inhibited swarming, while Class II (CsrAEc, RsmAPa) failed despite sequence conservation. This differential activity occurred despite canonical GGA motifs in flagellin (hag) transcript, suggesting evolutionary plasticity in RNA-binding specificity beyond motif recognition. Leveraging this dataset, we trained machine learning algorithms to predict CsrA functionality, experimentally validating Bdellovibrio bacteriovorus CsrA (CsrABb) and Pseudomonas putida RsmA (RsmAPp) as Class I. Our findings establish Swarm-seq as a powerful platform for characterizing CsrA homologs from genetically intractable or unculturable bacteria and demonstrate the potential for machine learning-guided discovery of functional regulatory proteins, providing insights into post-transcriptional regulatory network evolution.