Comparison of the protein-coding genomes of three deep-sea, sulfur-oxidising bacteria: “Candidatus Ruthia magnifica”, “Candidatus Vesicomyosocius okutanii” and Thiomicrospira crunogena

Methods

The 4273π variant of the Raspbian Linux operating system [13] was used on a Raspberry Pi computer (Version 1, Model B, Revision 2.0). Genome-wide protein sets for “Candidatus Ruthia magnifica” strain Cm, “Candidatus Vesicomyosocius okutanii” HA and Thiomicrospira crunogena XCL-2 were downloaded, in FASTA format, from the Ensembl genomes database (http://ensemblgenomes.org) [14, 15]. OrthoMCL software (http://orthomcl.org) [9] and MCL [16] were used to delimit protein groups based on sequence similarity.

The OrthoMCL procedure was followed as outlined in the OrthoMCL user guide, with the exception of using the substitution matrix BLOSUM45 for the ‘all-versus-all’ NCBI BLAST [10] and omitting BLAST’s -z parameter; and for our final analysis, the inflation value (I) was set to 1.4 when running MCL.

Results using I = 1.2, 1.3 and 1.4 gave rise to a group which was not present using other values. It had multiple copies of “Candidatus R. magnifica” proteins but only single copies of “Candidatus V. okutanii” and T. crunogena proteins. UniProt suggested that these proteins had the same function (histidinol-phosphate aminotransferases) but increasing the I value split them up into different groups. Hence on biological grounds, I > 1.4 was rejected. I = 1.4 was used for the final analysis presented here, as it gave the strongest restraints on group formation while still maintaining this aminotransferase group (Table 1, Group 27).

Once groups were delimited, a Perl script, modified from [17], was used to count the number of times each species was represented in each protein group [15]. To verify the reliability of the script, our final OrthoMCL groups file was used as input for an independently-written protein-counting script (Kevin Kiesworo, unpublished) and the same counts were obtained. Additionally, the OrthoMCL groups file from a similar study on different taxa (Hannah Currant, unpublished) was used as input for our script, and the same counts were obtained as in that study.

Functions of both the largest and most interesting groups were then inferred by searching for protein accessions in UniProt: group members were searched until a common function was found between at least four proteins, or for smaller groups all members were searched (Table 1).

Results

Similarities between all three species

679 of the 1070 protein groups delimited by OrthoMCL (63.5%) contained a single protein copy in each of the three species. This high degree of similarity could be a consequence of common ancestry, horizontal transfer in a shared habitat, or most likely a mixture of both.

All three species inhabit deep-sea vent environments which are highly variable and have constantly fluctuating factors such as sulfur and carbon concentrations [18]. In order to survive, the organisms must possess methods that allow them to deal with such fluctuations. One shared process, for example, is their ability to oxidise the sulfur supplied by deep-sea vents to fix carbon for use in cellular functions.

Two groups were predicted to contain duplicate protein copies in all three species (Table 1, Groups 9 and 10). These are consistent with duplication in a common ancestor, with subsequent speciations, although our current work does not distinguish this from other possibilities such as horizontal transfer.

All three species have a duplicate copy of elongation factor Tu (Table 1, Group 9). These paralogs are found in all proteobacteria and it has been hypothesised, therefore, that this duplication event preceded the divergence of this phylum [19]. It has been shown that the tuf genes that encode these proteins undergo gene conversion [20] which inhibits any divergence, and therefore sub- or neo-functionalisation, of the two genes [21]. The persistence of the duplicate may therefore indicate high levels of expression. Detection of this known group is promising in regards to the reliability of our methods.

Each species also has a duplicate in the group of nitrogen regulatory proteins (P-II; Table 1, Group 10). One of these copies, in “Candidatus V. okutanii”, is the product of the glnK gene. This gene seems to be commonly duplicated in some sub-divisions of proteobacteria [22] and its evolution seems to be associated with that of the amtB ammonium transporter gene, to which it is physically and functionally linked [23]. Interestingly, these ammonium transporters make up Group 11 (Table 1) and although there are four copies in T. crunogena, the other species have no duplicates. This would be consistent with genome reduction due to a symbiotic lifestyle [2], although our current work cannot distinguish this with certainty.

Unique to T. crunogena

A large number of protein groups (123) are found only in T. crunogena (Table 2). Also, in 10 groups, T. crunogena has multiple protein copies where the other species only have one copy each (Table 1). As the only independent-living species studied, T. crunogena may require a larger number of genes and proteins for survival. The other, symbiotic, organisms can rely on their hosts to provide some essential functions and, therefore, loss of some genes could prove to be energetically favourable [2]. There is also a lower total protein count for these species (976 and 937 protein sequences, compared to 2196 in T. crunogena).

Unique to “Candidatus R. magnifica”

“Candidatus R. magnifica” has one unique group that consists of glycosyl transferases (Table 1, Group 746).

There was also one group delimited that had multiple “Candidatus R. magnifica” proteins and only single proteins from the other species (Table 1, Group 27). This is a group of histidinol-phosphate aminotransferases.

Unique to “Candidatus V. okutanii”

“Candidatus V. okutanii” has no unique groups or paralogs.