Horizontal acquisition of a patchwork Calvin cycle by symbiotic and free-living Campylobacterota (formerly Epsilonproteobacteria)

Abstract

Most autotrophs use the Calvin-Benson-Bassham (CBB) cycle for carbon fixation. In contrast, all currently described autotrophs from the Campylobacterota (previously Epsilonproteobacteria) use the reductive tricarboxylic acid cycle (rTCA) instead. We discovered campylobacterotal epibionts ("Candidatus Thiobarba") of deep-sea mussels that have acquired a complete CBB cycle and may have lost most key genes of the rTCA cycle. Intriguingly, the phylogenies of campylobacterotal CBB cycle genes suggest they were acquired in multiple transfers from Gammaproteobacteria closely related to sulfur-oxidizing endosymbionts associated with the mussels, as well as from Betaproteobacteria. We hypothesize that "Ca. Thiobarba" switched from the rTCA cycle to a fully functional CBB cycle during its evolution, by acquiring genes from multiple sources, including co-occurring symbionts. We also found key CBB cycle genes in free-living Campylobacterota, suggesting that the CBB cycle may be more widespread in this phylum than previously known. Metatranscriptomics and metaproteomics confirmed high expression of CBB cycle genes in mussel-associated "Ca. Thiobarba". Direct stable isotope fingerprinting showed that "Ca. Thiobarba" has typical CBB signatures, suggesting that it uses this cycle for carbon fixation. Our discovery calls into question current assumptions about the distribution of carbon fixation pathways in microbial lineages, and the interpretation of stable isotope measurements in the environment.

Fig. 1

Phylogenomic tree of representatives of Campylobacterota. The 18 single copy genes used in this analysis were chosen based on the AMPHORA2 marker database [90]. Five deltaproteobacterial species were used to root the tree. In blue are genomes with rTCA cycle genes and in orange genomes with CBB cycle genes. The right column indicates GC content of each genome, the dotted line indicates 50% GC content

Fig. 2

“Ca. Thiobarba spp.” share metabolic features of Gammaproteobacteria and Campylobacterota. Figure shows overview of the main metabolic pathways for energy generation and carbon fixation in known chemosynthetic Gammaproteobacteria and Campylobacterota compared with the metabolism of “Ca. Thiobarba spp.”

Fig. 3

Stable carbon isotope values of “Ca. Thiobarba childressi” are consistent with carbon fixation via the CBB cycle. Model of δ¹³C values of deep-sea carbon and the predicted influence of different inorganic fixation pathways on these values. The δ¹³C values of CO₂ originating from ambient seawater are shown in yellow, and the expected δ¹³C values of CO₂ originating from methane oxidation (MOX) are shown in blue. The red line represents the average δ¹³C value measured for “Ca. Thiobarba” peptides using direct Protein-SIF. Reference δ13C values for “Seep methane” and “Seep/ambient CO₂” are based on Macavoy et al. [48] and Sassen et al. [49]. Transformations of δ¹³C values for each metabolic pathway are estimated based on Pearson et al. [47]

Fig. 4

“Ca. Thiobarba” genomes encode a CBB cycle with genes affiliated to at least three phylogenetically distinct classes. The solid arrows indicate enzymatic reactions that are unique to the CBB cycle, while the dashed arrows indicate that the enzymes are also involved in other metabolic pathways. Enzyme names are shown in bold and the colors represent their phylogenetic affiliations

Fig. 5

“Ca. Thiobarba” RuBisCO proteins cluster with gammaproteobacterial sequences. Bayesian inference trees of RuBisCO large (a) and small (b) subunit amino acid sequences under an LG model with Gamma-distributed rates of evolution. Analyses were performed with 6 million generations using two parallel Monte Carlo Markov chains. Sample trees were taken every 25,000 generations. Left arrows indicate truncated tree, tree roots were built from Prochlorococcus and Synechococcus sequences for (a) and Planktothrix and Synechococcus sequences for (b). Full trees are displayed as Figs. S14 and S15

Fig. 6

“Ca. Thiobarba” phosphoribulokinases are loosely affiliated with those from Betaproteobacteria, Alphaproteobacteria, and Verrucomicrobia. Bayesian inference tree of phosphoribulokinase amino acid sequences under an LG model with Gamma-distributed rates of evolution and a proportion of invariant sites. Analyses were performed with 6 million generations using two parallel Monte Carlo Markov chains. Sample trees were taken every 25,000 generations. Left arrow indicates truncated root, the root is built from distant Prochlorococcus and Synechococcus sequences. Full tree is displayed as SI appendix Fig. S23