Exploiting rRNA operon copy number to investigate bacterial reproductive strategies

Abstract

The potential for rapid reproduction is a hallmark of microbial life, but microbes in nature must also survive and compete when growth is constrained by resource availability. Successful reproduction requires different strategies when resources are scarce and when they are abundant^1,2, but a systematic framework for predicting these reproductive strategies in bacteria has not been available. Here, we show that the number of ribosomal RNA operons (rrn) in bacterial genomes predicts two important components of reproduction-growth rate and growth efficiency-which are favoured under contrasting regimes of resource availability^3,4. We find that the maximum reproductive rate of bacteria doubles with a doubling of rrn copy number, and the efficiency of carbon use is inversely related to maximal growth rate and rrn copy number. We also identify a feasible explanation for these patterns: the rate and yield of protein synthesis mirror the overall pattern in maximum growth rate and growth efficiency. Furthermore, comparative analysis of genomes from 1,167 bacterial species reveals that rrn copy number predicts traits associated with resource availability, including chemotaxis and genome streamlining. Genome-wide patterns of orthologous gene content covary with rrn copy number, suggesting convergent evolution in response to resource availability. Our findings imply that basic cellular processes adapt in contrasting ways to long-term differences in resource availability. They also establish a basis for predicting changes in bacterial community composition in response to resource perturbations using rrn copy number measurements⁵ or inferences^6,7.

Figure 1. Maximum growth rate is related to a bacterium's rrn copy number (a, n=184) and carbon use efficiency (b, n=8)

Non-phylogenetic OLS regression (solid lines with 95% CI) reveals that these traits are correlated (a, p < 0.001; b, p = 0.037) and phylogenetic regression (dashed lines) demonstrates that the relationships can't be explained by shared ancestry (a, p < 0.001; b, p = 0.036). Mean carbon use efficiency is plotted in panel b from two independent flasks, i.e. biological replicates. Species represented in panel b are: Sphingopyxis alaskensis RB2256 (▲), Acidobacteriaceae sp. TAA166 (□), Rhodospirillaceae sp. PX3.14 (●), Pseudomonas sp. HF3 (■), Arthrobacter sp. EC5 (▼), Escherichia coli K12 MG1655 (◆), Bacillus subtilis Marburg ATCC 6051 (○), Vibrio natriegens ATCC 14048 (◇).

Figure 2. Protein synthesis yield (a, n=8) and rate (b, n=10) are correlated with log₂-rrn copy number, but in opposite directions

Non-phylogenetic OLS regression (solid lines with 95% CI) reveals these traits are correlated and phylogenetic regression (dashed line) demonstrates evolutionary history is not responsible for the relationship (Table 1, Supplementary Table 1). Mean protein yield and mean translational power are plotted, with error bars representing standard error of three independent flasks, i.e. biological replicates. Species represented in panel a are: Sphingopyxis alaskensis RB2256 (▲), Acidobacteriaceae sp. TAA166 (□), Rhodospirillaceae sp. PX3.14 (●), Pseudomonas sp. HF3 (■), Arthrobacter sp. EC5 (▼), Escherichia coli K12 MG1655 (◆), Bacillus subtilis Marburg ATCC 6051 (○), Vibrio natriegens ATCC 14048 (◇). Species in panel b are the same as panel a, with the following changes and additions: Escherichia coli B REL607 (◆), Comamonadaceae sp. HS5 (), Mycobacterium sp. PX3.15 (×), Sphingobacteriaceae sp. LC9 (), Oxalobacteriaceae sp. EC4 (), Sphingobacteriaceae sp. EC2 (+).

Figure 3. Phylogenetic principal component analysis (pPCA) of genome content of 1,167 unique bacterial species

KEGG modules (a) or orthologs (b) datasets were analyzed separately. pPCA genome scores were regressed against log₂-rrn and slopes of phylogenetic linear regressions are reported below their corresponding analysis (c and d; slope estimate plotted with its 95% confidence interval and *indicates slope p < 0.05). pPCA axes displayed in a and b have the highest magnitude slopes from the first six pPCA axes.