Thiohalobacter thiocyanaticus and Guyparkeria sp. strain SCN-R1 were both grown in 200 ml mineral medium with 20 mM thiosulfate as substrate. The medium contained 1 M NaCl, 3 g l^-1 potassium phosphate buffer, pH 7, 1 mM MgCl₂, 1 ml l^-1 of trace metal solution (Pfennig and Lippert, 1966), 10 μg l^-1 of vitamin B₁₂ and 4 mM NH₄Cl. To prevent acidification of the medium during thiosulfate oxidation, 40 mM of filter-sterilized NaHCO₃ solution was added. The cultures were incubated in 1 L flasks closed with a rubber septum on a rotary shaker at 140 rpm and 30°C. After complete consumption of thiosulfate (monitored by acidic 0.01 N J₂ titration), the cells were harvested by centrifugation, washed once in 1 M NaCl and stored at -80°C for DNA extraction.

Comparative denaturing gel electrophoresis with fractions of cell extract from cells of two halophilic SOB species grown with either thiosulfate or thiocyanate as electron donor were performed according to Laemmli protocol using a gradient of PAGE from 5 to 15% (Laemmli, 1970).

High molecular weight DNA was extracted by using the DNeasy UltraClean Microbial Kit (Qiagen). The extracted DNA was sequenced by the commercial company BaseClear (Leiden, Netherlands). The genome of Guyparkeria sp. strain SCN-R1 was sequenced with Illumina HiSeq, while the genome of Thiohalobacter thiocyanaticus was sequenced with a combination of Illumina HiSeq and PacBio Sequel. De novo assembly of the Illumina HiSeq paired-end sequence reads was performed with SPAdes. SSPACE Premium Scaffolder version 2.3 (Boetzer et al., 2011) was used to place the assembled contigs into scaffolds. Gapped regions within the scaffolds were (partially) closed using GapFiller version 1.10 (Boetzer and Pirovano, 2012). De novo assembly of the PacBio reads was performed using HINGE (Kamath et al., 2017). Mis-assemblies and nucleotide disagreements between the assembled sequences and the Illumina data were corrected with Pilon version 1.21 (Walker et al., 2014). The completeness of the genomes was determined with CheckM (Parks et al., 2014). The assembled genomes were first annotated in RAST (Overbeek et al., 2014¹) and subsequently in Genbank. An overview of the locus tags, annotation, cellular location, presence of transmembrane helices, etcetera of translated a sequences of Guyparkeria sp. SCN-R1 have been given in Supplementary Table S1, and that of Thiohalobacter thiocyanaticus Hrh1^T in Supplementary Table S2. The genome sequences have been deposited at Genbank under accession number QZMT00000000 for Guyparkeria sp. SCN-R1 and QZMU00000000 for Thiohalobacter thiocyanaticus HRh1^T.

Search of TcDH and other genes in bacterial genomes was performed using Tblastn algorithm in standalone BLAST v. 2.7.1 (Camacho et al., 2009) using a single genome sequence as database. Search of homologs was performed by online BLAST using BLASTp and PHI-BLAST algorithms (2017) with default settings (NCBI Resource Coordinators, 2017). InterProScan online tool and CDD/SPARCLE database was used for hypothetical protein function (Marchler-Bauer et al., 2017). Secondary structure prediction was performed with JPred4 (Drozdetskiy et al., 2015) with default settings. Jobs were submitted via JalView v. 2.10.3 (Waterhouse et al., 2009).

Transmembrane helix predictions: TMHMM2² – one of the recommended predicting tools according to Punta et al. (2007). Signal peptide prediction was performed with signalP v. 4.1 (Petersen et al., 2011).

Multiple alignments of protein sequences: Blastp search results with low e-values (<10^-5) and reasonable sequence identity (>30%) was used as inputs for ProbCons algorithm (Do et al., 2005).

For global pairwise alignment Needleman-Wunch algorithm on NCBI website³ was used with default parameters.

Comparative analysis of the ATPase sequences was done with the online program www.phylogeny.fr using the Advanced Mode. The sequence logo’s was created with the web-based application WebLogo (Crooks et al., 2004).

For a whole genome comparison of Thiohalobacter thiocyanaticus HRh1^T with a closely related Thiohalobacter strain FOKN1, the two genomes were aligned by using Mauve 2.3.1 (Darling et al., 2004) plugin inside Geneious 7.1 software (Biomatters Ltd., New Zealand). The two scaffolds of HRh1^T and the chromosome of FOKN1 genomes were previously rearranged in order to obtain the correct alignment (Consensus 1 of HRh1^T was reverse-complemented before concatenate with HRh1^T Consensus 0; FOKN1 was reverse-complemented; both residue numbering changed to start with the chromosome replication initiator DnaA gene). Average nucleotide identity (ANI) was calculated using the BLAST (ANIb) and MUMmer (ANIm) algorithms performed by the JSpecies online tool (Richter and Rosselló-Móra, 2009⁴), while digital DNA-DNA hybridization (dDDH) was performed by Genome-to-Genome Distance Calculator 2.1 online submission form (GGDC, Meier-Kolthoff et al., 2013⁵).

Phylogenetic analysis of 16S rRNA sequences was done in ARB (Westram et al., 2011) using the SILVA database (Quast et al., 2013). Briefly, the 16S rRNA sequences of Thiohalobacter thiocyanaticus HRh1^T and Guyparkeria sp. SCN-R1 were aligned using SINA (Pruesse et al., 2012). The aligned sequences were imported into ARB and added to an existing tree using the parsimony approach. Subsequently, a neighbor joining tree was calculated using 16S rRNA sequences of related SOB.

SoeA-like sequences were retrieved from the NCBI protein database, by using protein BLAST. Sequences were aligned using T-COFFEE (Di Tommaso et al., 2011) and the most likely amino acid substitution matrix was determined using protest 3 LG model (Abascal et al., 2005; Le and Gascuel, 2008) with gamma-distributed rates and empirical frequencies (model parameter –m PROTGAMMALGF in RAxML) for all alignments (Stamatakis, 2014, 2015). Maximum likelihood trees were calculated using RAxML 8.2.12 using the rapid bootstrap analysis algorithm. The tree was visualized in MEGA7.

Table 1 gives an overview of the general genome features of Guyparkeria sp. SCN-R1 and Thiohalobacter thiocyanaticus HRh1^T. The genome of Guyparkeria sp. SCN-R1 consists of 68 scaffolds with a total size of 2.4 Mbp. CheckM analysis showed that it was 100% complete. The genome of Thiohalobacter thiocyanaticus HRh1^T, which was sequenced with PaqBio and Illumina, consists of two scaffolds with a total size of 3.3 Mbp. CheckM analysis showed a completeness of 97%. Supplementary Tables S1, S2 show a selection of the most important functional proteins found in two genomes.

The genomes of both organisms contain genes encoding sulfide dehydrogenase of the Fcc (flavocytochrome c) type, although in both cases the cytochrome c subunit has only one heme c motif instead of the customary two motifs. Such modification had also been noticed for haloalkaliphilic SOB species of the genus Thioalkalivibrio (Berben et al., 2017a,b). The catalytic flavin subunit of Thiohalobacter is only distantly related to that of Thioalkalivibrio, while that of Guyparkeria is close to the one present in Halothiobacillus neapolitanus. Genes for the alternative sulfide dehydrogenase – sulfide quinone reductase (SQR) were not found in both genomes.

Both organisms have the Sox system, but in a different assemblage. The genome of Guyparkeria encodes a complete Sox pathway consisting of soxABXYZCD, allowing the oxidation of both sulfur atoms of thiosulfate completely to sulfate, while Thiohalobacter has an incomplete Sox system, i.e., lacking the soxCD sulfur dehydrogenase, which is more typical for the gammaproteobacterial SOB. Although at optimal growth conditions with thiosulfate as substrate HRh1^T did not produce any visible zero-valent sulfur particles, extracellular sulfur started to accumulate in cultures when pH dropped below 6 and at oxygen limitation. Also, young colonies on thiosulfate agar were always silver-white due to extracellular sulfur formation and the sulfur disappeared only in a month-old colonies (Sorokin et al., 2010). The zerovalent sulfur is further oxidized to sulfite by the reverse Dsr system. In Thiohalobacter the rDsr system is represented by two gene clusters located in the vicinity of each other: the functional dsrAB (related to both aerobic and anaerobic SOB within the order Chromatiales) and the accessory dsrMKJOP. The dsrEFH, which is needed for the transfer of zerovalent sulfur atom into the cell and normally present in the rDsr system, is absent. Three dsrC/tusE sulfur transferase genes were present elsewhere in the genome of Thiohalobacter. While one of them is definitely a genuine dsrC, important for the rDsr function, the two other copies might fulfill the function of the missing DsrEFH. The indicator of the direct pathway, dsrD (Rabus et al., 2015), is absent in the genome of Thiohalobacter. Therefore, it still can be considered as a variation of the rDSR system.

Further oxidation of sulfite formed by the action of rDSR system in Thiohalobacter can be executed by two pathways. One is traditionally associated with the rDsr and includes adenylyl phosphosulfite reductase (Apr) and sulfate adenylyl transferase (Sat). The gene cluster including putative AprABM and a separate sat gene are present in the Thiohalobacter genome. A second pathway includes direct sulfite oxidation to sulfate by a putative SoeABC quinone-interacting sulfite dehydrogenase which is encoded by an operon whereby the catalytic subunit is annotated as DMSO reductase or formate dehydrogenase. SoeA, DmsA, and FdhA all belong to the Mo oxidoreductase superfamily and, therefore, phylogenetic reconstruction is necessary to predict the most probably function. This done, the translated protein sequence from Thiohalobacter indeed clustered with a large group of putative SoeA present in the genomes of lithoautotrophic SOB (Supplementary Figure S1). However, the role of this gene cluster in sulfite oxidation can only be confirmed by expression analysis. Currently, the only functional evidence on the importance of this alternative type of sulfite dehydrogenase (to its more studied SorAB homologous enzyme which catalytic subunit also belongs to the Mo-supurfamily oxido-reductases and which uses cytochrome c as its immediate e-acceptor) has been provided for anoxygenic SOB Allochromatium vinosum (Dahl et al., 2013), whereby it also coexists with the rDSR system.

Figure 2 gives an overview of the different sulfur oxidation-related genes present in both Thiohalobacter thiocyanaticus HRh1^T and Guyparkeria sp. SCN-R1.

Both organisms encode form IAc RuBisCO (“green form”), indicating the presence of the Calvin-Benson ribulose-bisphosphate carboxylase pathway for autotrophic CO₂ fixation. The large subunit CbbL of Thiohalobacter is closely related to the Thioalkalivibrio CbbL proteins, while the CbbL of Guyparkeria is equally related to both Thioalkalivibrio and Halothiobacillus neapolitanus. The RuBisCO operon of Guyparkeria, in addition, has genes encoding carboxysome shell proteins (Cso), which is a common feature for this phylogenetic radiation of SOB (Yeates et al., 2008).

Another important group of enzymes in CO₂ fixation in autotrophic prokaryotes are carbonic anhydrases (CA), interconverting CO₂ and bicarbonate. Both genomes contain (independently placed from the RuBisCO operon) a γ-CA. In addition, Thiohalobacter also have two different copies of β-CA. Larger assortment of the CA in Thiohalobacter is probably related to the absence of carboxysomes that play an additional role in the carbon fixation efficiency, i.e., the carboxysomal shell protein CsoS3 is functioning as the carboxysome-specific carbon anhydrase (Price et al., 2008). Moreover, each organism has another CA copy, which is associated with the bicarbonate-dependent cyanate hydratase, i.e., an α-CA in Thiohalobacter and a β-CA in Guyparkeria.

Both halophilic SOB encode two types of heme-copper terminal oxidases, i.e., a cytochrome c oxidase of the cbb₃ type, which is regarded as a high O₂ affinity enzyme, and a quinol oxidase of the ba₃ type. In addition, the Thiohalobacter genome also contains an operon encoding for a second quinol oxidase of the bo₃ type, belonging to the copper-free cytochrome b class of oxidases. Membrane spectroscopy showed a clear difference in the heme composition between thiosulfate and thiocyanate-grown cells of Thiohalobacter: in the former hemes c, b, and a₃ types were visible with a domination of the heme c, while in the thiocyanate-grown cells only hemes c and b were detected with a much higher proportion of heme b (Supplementary Figure S2). This might indicate that during growth with thiosulfate at least two oxidases, cbb₃ and ca₃, were expressed, while during growth on thiocyanate only the cbb₃ was expressed.

For ATP synthesis, Guyparkeria genome encodes an H⁺-dependent F₁F₀ ATP synthase, which is normally present in aerobic chemolithotrophic SOB. In contrast, the genome of Thiohalobacter contains two operons encoding different types of ATPases. One is a 4-subunit H⁺-dependent V type ATP-synthase, and another one is an apparent sodium pumping F₁F₀ ATP synthase (Figure 3 and Supplementary Figure S3) (with the typical operon structure atpDCIRBEFAG and the unique Na⁺-binding motif EST-XX-Y, that is normally present as a secondary ATPase to its H⁺-dependent analog functioning for Na⁺ extrusion (Dibrova et al., 2010). Interestingly, however, in Thiohalobacter the Na⁺-ATP synthase partners with a H⁺-V-type ATPase, which is rare in Proteobacteria.

For the Na⁺/H⁺ homeostasis, both SOB encode the multi-sub-unit Na⁺/H⁺ antiporter MnhABCDEFH, although the operon of Thiohalobacter seems lacking the subunit C.

Both organisms belong to moderate halophiles that usually cope with osmotic stress by synthesizing (or importing) organic osmolytes. The genome of Thiohalobacter encodes genes for de novo synthesis of three potential osmolytes, i.e., trehalose, sucrose and glycine betaine. In contrast, Guyparkeria seems to rely on the synthesis of ectoin as the most probable main osmolyte, and sucrose as the secondary compound, which it also can import. Both halophiles can also import glycine betaine, but for this Guyparkeria has a dedicated transporter OpuD, while Thiohalobacter encodes a less specialized ABC-type transporter for proline/glycine betaine.

Because both species are able to oxidize thiocyanate, either thiocyanate hydrolyase or TcDH have to be present. Comparative gel electrophoresis analysis of total protein extracts from cells grown with thiosulfate against thiocyanate revealed clear differences only in case of Thiohalobacter, whose profile on thiocyanate showed two heavily overproduced proteins, one of which had a similar mass to the TcDH protein of Thioalkalivibrio (∼54 kDa) (Supplementary Figure S4). Tblastn was used to search for TcDH and thiocyanate hydrolase sequences in the genomes of Guyparkeria SCN-R and Thiohalobacter thiocyanaticus HRh1^T. Homologs of thiocyanate hydrolase were not found. However, a single copy of a TcDH homolog was found in both genomes. In addition, a short TcDH fragment corresponding to 52 amino acids was detected upstream of the TcDH gene in the genome of Guyparkeria. The fragment is not present in the Genbank annotation and can be detected only by protein sequence search against nucleotide sequence. It seems to be just a small duplicated non-functional DNA fragment from a TcDH gene.

Comparative analysis of the two sequences from halophilic SOB was performed with the TcDH from Thioalkalivibro paradoxus ARh1^T (crystal structure PDB id 5F75) and its various homologs found in the bacterial genomes (sequence identity above 30%). All residues of the active site in TcDH were present in both TcDH homologs of Guyparkeria and Thiohalobacter (Figure 4). These residues either coordinate copper ions and water molecules in the central cavity of the 5F75 structure, or form the shape of this cavity. The presence of all these highly conserved and structurally important residues in both sequences provide strong evidence that these genes encode TcDH. In addition to the presence of TcDH sequences, we also found cyanase operons in both genomes. However, for some reasons, the cyanase activity was not measurable in cell homogenates of both bacteria (Sorokin et al., 2010, 2014).

The genome region around the TcDH genes was analyzed and manually annotated using various bioinformatic tools (Figure 5). To determine conserved genes for both bacteria each gene in this region was aligned against the other genome. Six genes upstream of the TcDH genes are common to both organisms: they have sequence identity of around 50%, and the same order (Figure 5).

A gene cluster upstream of the TcDH gene in Guyparkeria and Thiohalobacter encode proteins belonging to thioredoxin superfamily. However a conservative COX motif, which is critical for redoxin activity (Meyer et al., 2008), is absent in both sequences. Multiple alignment (Supplementary Figure S5) of these two sequences and their homologs have shown that a part of this superfamily members lack a COX motif or any other cysteines. As a consequence, they lack oxidoreductase activity, which, however, does not affect their ability to bind glutathione (Nardini et al., 2004). A predicted function of proteins incapable of glutathione oxidation is glutathione transport to a protein partner (Nardini et al., 2004). Multiple alignment also shows an appearance of COX motif in several sequences (black boxes), while the overall sequence in this region remains conserved (Supplementary Figure S3). A number of these COX-missing proteins was structurally characterized (Gaudet et al., 1996; Wang et al., 1998; Liepinsh et al., 2001) and have been shown to preserve a conserved tertiary structure of the thioredoxin-like domain. So, we can assume that the tertiary structure of all these sequences is similar and is suitable for the sulfur atom binding/carrying.

In another study, a complex of a thioredoxin-like protein with its protein partner was crystallized (Gaudet et al., 1996), which shows the ability of this family members to form a complex with other proteins. Interestingly, the partner of thioredoxin-like protein in that case (PDB id 2TRC) has exactly the same fold as TcDH – a 7 bladed beta-propeller. Since TcDH catalyzes sulfane atom oxidation of NCS^- to sulfur and NCO^-, it needs an acceptor for the zero-valent sulfur atom. Thus, the protein encoded by the first gene upstream to TcDH might fulfill the role of accepting sulfur from TcDH and transporting it to enzymes involved in further oxidation of zero-valent sulfur atom to sulfate.

Recently, a genome announcement has been published (Oshiki et al., 2017) and a genome deposited in the GenBank (assembly GCF_002356355) of a bacterium named as “Thiohalobacter thiocyanaticus” strain FOKN1 which has the 16S rRNA gene identity of 99.1% to the type strain HRh1^T. While such values in general are typical of closely related prokaryotes, the publishing of an organism with the same species name can not be considered as legal before a full polyphasic taxonomic description is provided. Furthermore, a whole genome comparison using Mauve showed that despite the high 16S rRNA gene identity and the highly similar structure of operons and RNAs, the nucleotide pairwise identity of common regions was only 57% after alignment. This and the calculated ANI and DDH values between the two strains clearly demonstrate (Supplementary Figure S7) that strain FOKN1 represents a different genetic species and is wrongly announced as “Thiohalobacter thiocyanaticus.”

Therefore, here we provide only a brief comparison of the sulfur oxidation related genes present in the genome of this bacterium, since none of the genomic information is yet backed up by the published phenotypical data. Strain FOKN1 has a very similar genomic content related to its sulfur oxidation pathways with the type strain of Thiohalobacter thiocyanaticus HRh1^T, including the following: FccAB type of sulfide dehydrogenase (FOKN1_0725/0726); an incomplete Sox pathway including SoxBAZYX (FOKN1_1148-1152), followed by sulfur transferases rhodanese/TusA (FOKN1_1154-1155); reversed DSR pathway for sulfur oxidation with a duplication of the key functional genes dsrAB (the only significant difference from the type strain), including DsrA1B1 (FOKN1_1306/1307), DsrC1C2A2B2 (FOKN1_1941/1943/1945/1946), sulfur relay proteins TusECB-DsrC (FOKN1_1947-1950), and Dsr accesory proteins DsrMKJPO (FOKN1_ FOKN1_1951-1955). For sulfite oxidation it has AprMBA (FOKN1_1661-1663)/Sat (FOKN1_1877) and a similar to HRh1^T putative SoeCBA (FOKN1_0490-0492).