SOCfinder: a genomic tool for identifying social genes in bacteria

Abstract

Bacteria cooperate by working collaboratively to defend their colonies, share nutrients, and resist antibiotics. Nevertheless, our understanding of these remarkable behaviours primarily comes from studying a few well-characterized species. Consequently, there is a significant gap in our understanding of microbial social traits, particularly in natural environments. To address this gap, we can use bioinformatic tools to identify genes that control cooperative or otherwise social traits. Existing tools address this challenge through two approaches. One approach is to identify genes that encode extracellular proteins, which can provide benefits to neighbouring cells. An alternative approach is to predict gene function using annotation tools. However, these tools have several limitations. Not all extracellular proteins are cooperative, and not all cooperative behaviours are controlled by extracellular proteins. Furthermore, existing functional annotation methods frequently miss known cooperative genes. We introduce SOCfinder as a new tool to find bacterial genes that control cooperative or otherwise social traits. SOCfinder combines information from several methods, considering if a gene is likely to [1] code for an extracellular protein [2], have a cooperative functional annotation, or [3] be part of the biosynthesis of a cooperative secondary metabolite. We use data on two extensively-studied species (P. aeruginosa and B. subtilis) to show that SOCfinder is better at finding known cooperative genes than existing tools. We also use theory from population genetics to identify a signature of kin selection in SOCfinder cooperative genes, which is lacking in genes identified by existing tools. SOCfinder opens up a number of exciting directions for future research, and is available to download from https://github.com/lauriebelch/SOCfinder.

Keywords: cooperation; cooperative gene; kin selection; relatedness; social gene.

Fig. 1.

Overview of SOCfinder. We input a genome sequence, and cooperative genes are found based on three modules: [1] Extracellular genes [2]. Genes annotated with functions known to be cooperative, based on sequence similarity [3]. Genes for secondary metabolites that are known to be cooperative. We output a list of genes for cooperative traits for each module, and a final list that combines all three.

Fig. 2.

Categorisation of cooperative and private behaviours in bacteria. Cooperative behaviours are involved in the production and secretion of molecules that provide benefits that can be shared with other cells. Private behaviours give fitness benefits only to the individual expressing the gene.

Fig. 3.

Principles of existing methods to find cooperative genes in genomes. We can look for: (a) Genes that have been shown to be cooperative in lab experiments (artisan). (b) Extracellular proteins that are secreted from the cell. (c) Genes that are annotated with functions that we know are cooperative, based on sequence similarity to proteins of known function. (d) Genes that are significantly upregulated when individuals are cooperating (transcriptome). (e) Genes for the biosynthesis of secondary metabolites that are known to be cooperative. A table of specific tools that can be used to find cooperative genes according to these principles is in Supplement S2.

Fig. 4.

Flow diagram of the blast process for finding cooperative genes. Gram-positive and Gram-negative genomes are run against their own databases of high-confidence non-extracellular proteins (database one or two), but both are run against the same databases of higher- and lower-confidence extracellular proteins (databases three and four). Full information of the databases, as well as the definition of a significant match are found in (Tables 1–3).

Fig. 5.

SOCfinder on 1301 genomes of 51 species. The x-axis shows the proportion of the genes in a genome that are categorised by SOCfinder as cooperative. For each species, a point represents the proportion for one genome, and the bar represents the median proportion.

Fig. 6.

(a and b) Number of genes captured by each method. (c and d) Percentage of artisanal cooperative genes captured by each method. The left panels (a and c) are for P. aeruginosa , and the right panels (b and d) are for B. subtilis .

Fig. 7.

Overlap between methods to find cooperative genes. The top Venn diagram is for P. aeruginosa , and the bottom Venn diagram is for B. subtilis . The red circle is genes categorised as cooperative by the Artisanal approach. The blue circle is genes categorised as cooperative by SOCfinder. The yellow circle is genes categorised as cooperative (extracellular) by PSORTb.

Fig. 8.

Nucleotide polymorphism for private (gold) and cooperative (blue) quorum-sensing controlled genes. The top three graphs (a–c) show P. aeruginosa , and the bottom three graphs (d–f) show B. subtilis . The left graphs (a and d) show cooperative genes identified by SOCfinder. The middle graphs (b and e) show cooperative genes identified by PanSort. The right graphs (c and f) show cooperative genes identified by PSORTb. For each graph, the dotted line shows the background level of nucleotide polymorphism for a set of private genes. The black line and * shows a significant difference between cooperative and private genes.