Staphylococcus aureus Genomes Harbor Only MpsAB-Like Bicarbonate Transporter but Not Carbonic Anhydrase as Dissolved Inorganic Carbon Supply System

ABSTRACT In recent years, it became apparent that not only autotrophic but also most other bacteria require CO2 or bicarbonate for growth. Two systems are available for the acquisition of dissolved inorganic carbon supply (DICS): the cytoplasmic localized carbonic anhydrase (CA) and the more recently described bicarbonate transporter MpsAB (membrane potential generating system). In the pathogenic species Staphylococcus aureus, there are contradictions in the literature regarding the presence of a CA or MpsAB. Here, we address these contradictions in detail. We could demonstrate by careful BLASTp analyses with 259 finished and 4,590 unfinished S. aureus genomes that S. aureus does not contain CA and that the bicarbonate transporter MpsAB is the only DICS system in this species. This finding is further supported by two further pieces of evidence: (i) mpsAB deletion mutants in four different S. aureus strains failed to grow under atmospheric air, which should not be the case if they possess CAs, since we have previously shown that both CA and MpsAB can substitute for each other, and (ii) S. aureus is completely resistant to CA inhibitors, whereas Staphylococcus carnosus, which has been shown to have only CA, was inhibited by ethoxyzolamide (EZA). Taken together, we demonstrate beyond doubt that the species S. aureus possesses only the bicarbonate transporter MpsAB as its sole DICS system. IMPORTANCE The discrepancies in the current literature and even in NCBI database, which listed some protein sequences annotated as Staphylococcus aureus carbonic anhydrase (CA), are misleading. One of the existing problems in publicly available sequence databases is the presence of incorrectly annotated genes, especially if they originated from unfinished genomes. Here, we demonstrate that some of these unfinished genomes are of poor quality and should be interpreted with caution. In the present study, we aimed to address these discrepancies and correct the current literature about S. aureus CA, considering the medical relevance of S. aureus. If left unchecked, these misleading studies and wrongly annotated genes might lead to a continual propagation of wrong annotation and, consequently, wrong interpretations and wasted time. In addition, we also show that bicarbonate transporter MpsAB-harboring bacteria are resistant to CA inhibitor, suggesting that pathogens possessing both MpsAB and CA are not treatable with CA inhibitors.

That being said, HCO 3 2 is also equally as important even for nonautotrophic bacteria due to the fact that many metabolic pathways require either HCO 3 2 or CO 2 as the substrates or as products of metabolism (3,4). In this regard, these bacteria utilize the enzyme carbonic anhydrase (CA) as a dissolved inorganic carbon supply (DICS) system (3,5) and the more recently described bicarbonate transporter MpsAB (membrane potential generating system) in Staphylococcus aureus (4). MpsAB is present not only in autotrophic bacteria such as Hydrogenovibrio crunogenus, Nitrobacter winogradskyi (6), and Halothiobacillus neapolitanus (7) but also in many nonautotrophic bacteria, like some strains of Bacillus subtilis, Legionella pneumophila, and Vibrio cholerae (6,8). MpsAB works alone and/or together with CA function to supply bicarbonate for anaplerotic reactions. Although both systems are interchangeable, they rarely coexist in a given species (4).
In our previous work, we showed that Staphylococcus carnosus harbors only a CA gene and confirmed that the protein is functional (8). As there was no S. carnosus-specific CA homolog present in S. aureus, we deduced that MpsAB functions as the sole CO 2 /bicarbonate concentration system in S. aureus. Moreover, MpsAB outperforms CA, and the former has an advantage in species where CO 2 diffusion is impeded, for example in mucus biofilm-forming bacteria. As such, our findings are in contradiction with other studies about the presence of CA in S. aureus.
Since there are several publications in which an S. aureus-specific CA has been described and also studied (9)(10)(11)(12), we investigated the question of whether a CA actually exists in S. aureus. Using BLASTp analyses, phenotypic characterization of mpsAB mutants, and the resistance studies to CA inhibitors, we demonstrate that there is no CA present in S. aureus.

RESULTS
All finished S. aureus genomes contain no CA-related Pfam motifs. To enable a quick and systematic search for the presence of CA in S. aureus, we screened for the occurrence of protein families (Pfam) motifs (PFam00484, PFam00194, and PFam10563 for prokaryotic-type CAs, eukaryotic-type CAs, and putative CA-like domain, respectively) using the database from Integrated Microbial Genomes and Microbiomes (IMG/M) (13). We used this database instead of NCBI because it is more organized to perform searches and the exact strains from finished genomes, permanent drafts, or drafts could be selected. All 259 finished sequenced S. aureus strains do not contain any of the three CA-related Pfam motifs.
To demonstrate the reliability of Pfam motifs, we performed protein-protein Basic Local Alignment Search Tool search (BLASTp) of the protein sequence of an experimentally confirmed CA from Staphylococcus carnosus (8) against fully sequenced (finished) representative strains from the genus Staphylococcus. The presence of CAs based on Pfam motif correlated with the high percentage of protein identity from S. carnosus CA (Table 1). No significant protein identity was detected when there was no Pfam motif present, such as in S. aureus, Staphylococcus haemolyticus, and Staphylococcus lugdunensis. In addition, a search in all the 259 finished S. aureus genomes in IMG/G and also AureoWiki (14), which is manually curated, revealed that no protein is annotated as CA or putative CA.
BLASTp showed no protein similarity of a-, b-, andc-CAs in S. aureus. Given that S. aureus and S. carnosus are from the same genus, they should share protein homology and more similarity with each other than with any bacteria from other genera. Therefore, the protein sequence of S. carnosus CA, which is from the class of b-CAs, was subjected to BLASTp search against all finished S. aureus genomes in IMG/M, but no similarity was found. As not all the microbial genomes might be integrated in IMG/M yet, we also performed the same BLASTp against S. aureus (taxonomy ID [taxid]: 1280) in NCBI database. We found two hits: NCBI accession numbers SPZ78436.1 and SPZ78435.1 ( Table 2). As both the proteins are found in only one strain and based on the data in Table 2, most likely the genomes samples sequenced belonged to other staphylococcal species or the genes were wrongly annotated. Therefore, we concluded that there is no b-CA in S. aureus.
Since the different classes of CAs have independent evolutionary origins (3), we also searched for the presence of aand g-CAs in S. aureus. We selected some bacteria whose CAs were experimentally proven, and these protein sequences were subjected to BLASTp search in S. aureus, as well as two CA-harboring species, S. carnosus and Staphylococcus pseudintermedius, as controls (Table 3 and 4). As shown in Table 3, there was no similarity among these CAs with S. aureus, S. carnosus, and S. pseudintermedius except for two cases. First, BLASTp of two human CAs resulted in two hits with proteins annotated as S. aureus CA ( Table 3). Considering that they are found in only two unfinished S. aureus genomes (Table 3) and the errors observed in these sequences (Table 2), the genome samples were most likely contaminated. For the same reason as that mentioned previously, we deduced that there are no aandg-CAs in S. aureus (Table 3 and 4).
BLASTp of all proteins annotated as S. aureus CAs revealed errors in permanent draft genomes. To further confirm that there is indeed no CA present in S. aureus, we searched the NCBI database for all the proteins annotated as S. aureus CA and performed an extensive BLASTp search. All 259 finished genomes showed no identity at all against the seven CAs listed in Table 5, except for two cases where low identities were found in in proteins annotated as acetyltransferase, galactoside O-acetyltransferase, sulfate permease, or hypothetical protein but not as CA. As with the BLASTp of a-, b-, and g-CAs above, almost all of the CA similarities found were in unfinished genomes, suggesting that these genomes are often unreliable and should be interpreted with caution. To prove our point, we extended the same BLASTp search in 4,590 unfinished genomes. Results similar to those found with finished genomes were found, and all other hits were found in assemblies which were marked as contaminated by NCBI (Table 5). One particular strain, C0673, showed multiple hits for NGG14433.1, NGB42162.1, SPZ78435.1, SPZ78436.1, and WP_094666538.1. According to NCBI, C0673 is an unfinished genome with 89 contigs where the taxonomy check is inconclusive. Although this strain is annotated as S. aureus C0673 in NCBI database, it is highly questionable. Thus, we downloaded the genome sequence and checked it against public databases for molecular typing and microbial genome   Table 2 for comment on this protein a Identity refers to shared identical residues with each of the CA proteins (UniProt ID) from selected bacteria and the indicated Staphylococcus species using BLASTp. diversity (PubMLST) (15). According to PubMLST, the predicted taxa for C0673 are actually 83% Staphylococcus sciuri, now known as Mammaliicoccus sciuri. Using the IMG/M database, pairwise average nucleotide identity (ANI) with two finished S. sciuri genomes revealed that C0673 has 97% nucleotide identity with S. sciuri SNUDS-18 and 96% nucleotide identity with S. sciuri FDAARGOS_285. C0673 is wrongly annotated as S. aureus in NCBI database, which gave us false-positive hits in our BLASTp because S. sciuri but not S. aureus has a CA as stated in Table 1 and our previous work (8).
All protein sequences annotated as S. aureus CAs in NCBI are not from S. aureus. Given the observation that not a single strain out of 4,849 S. aureus genomes has a reasonable protein identity with any of the sequences annotated as S. aureus CA, we proceeded to examine the authenticity of these sequences. All of the nine sequences listed in Table 2 originated from unfinished genomes, and most of them contain many contigs, indicating these genomes are of low quality (16). When these sequences were subjected to BLASTp search, they showed only one hit against their own sequences and the rest were from either other staphylococcal species or other microorganisms, or even human. This clearly suggests that these genome assemblies were contaminated or contain sequencing errors and therefore are not accurate and should be corrected.
Deletion of mpsAB in four different backgrounds of S. aureus causes severe growth defect in atmospheric conditions. In our previous study, we demonstrated that deletion of mpsAB in two different S. aureus backgrounds, SA113 and HG001 (both are methicillin-susceptible S. aureus), could not grow under normal atmospheric conditions, indicating that there is no functional CA (4,17). Here, we deleted mpsAB in two more S. aureus strains, JE2 and MW2, which are methicillin resistant (MRSA). Like with SA113 and HG001, the MRSA deletion mutants could not grow under atmospheric air, indicating that MpsAB is the only DICS system (Fig. S1).
MpsAB-harboring strains are resistant to CA inhibitors. CA inhibitors, especially sulfonamides, are able to effectively inhibit most of the CAs and consequently hinder the bacterial growth (18,19). With regard to this, we tested eight such inhibitors to provide further evidence that CA does not present in S. aureus. Acetazolamide (AZA),  Table 5 for comments regarding these protein sequences. These strains have the same similarity as all the finished genomes. As there only a few strains out of 4,590 permanent draft sequences with unspecific and low identity, the origins of each of these strains were not examined. C0673 is wrongly annotated as S. aureus in NCBI database.  (20)(21)(22). S0859 is an N-cyanosulphonamide synthetic compound reported to be a selective inhibitor of sodium-bicarbonate cotransporters (NBC, SLC4) in mammalian heart (23). The chemical structures are provided in Figure S2. At the highest concentration tested (1,000 mM), all the compounds did not inhibit MpsAB-harboring S. aureus and S. epidermidis as well as strains where CAs were deleted and complemented with MpsAB instead, including S. carnosus carrying plasmid containing mpsABC (Table 6). In CA-harboring strains, only EZA showed an MIC of 64 mM against S. carnosus, which was increased to 250 mM when CA was overexpressed in S. carnosus TM300 (pRB473 can) ( Table 6). To verify that activity of EZA is mediated through the inhibition of CA, we repeated the MIC determinations in both normal atmospheric air and 5% CO 2 conditions. EZA was inactive against these strains when incubated in the presence of 5% CO 2 compared to atmospheric air, while there was no difference in S. aureus ( Fig. 1; Table S3). Vancomycin and oxacillin were used as a control and, as expected, displayed no difference in MIC in both conditions. Collectively, these results suggest that the target for EZA is most likely the intracellular CA, which is not present in S. aureus and S. epidermidis.

DISCUSSION
The discrepancies in the current literature regarding the presence of CA in S. aureus are substantial to warrant a comprehensive study to correct them, especially given that S. aureus is a clinically important pathogen. The first publication was in 1990 when Nafi et al. used a protein-binding monospecific antibody prepared against purified Neisseria sicca CA by immunoblotting method and also determined CA activity in cell extracts of various bacteria to screen for the presence of CA (18). Although CA activity was not detected in S. aureus, there was a positive reaction in the immunoblot, suggesting a reaction with a CA-like protein. In 1999, Smith et al. reported a molecular mass of 23 kDa in immunoblot with antisera raised against b-CA from Methanobacterium thermoautotrophicum DH and also some CA activity in S. aureus cells extract (24). Detection of target proteins by immunoreactivity alone is highly questionable in S. aureus because of its two IgG-binding proteins.
In 2015, Capasso and Supuran reported that the genome of S. aureus encodes only for g-CA, but no other information or citation was given to support this statement (9). In the following year, the same authors stated that S. aureus has a g-CA, referring to protein EVX10196.1, which was used to build a CA phylogenetic tree (10). In NCBI database, EVX10196.1 is annotated as 2,3,4,5-tetrahydropyridine-2,6-dicarboxylate N-acetyltransferase from S. aureus M20916, which is an unfinished genome with 67 contigs. This 239-amino-acid protein is listed as nonessential by AureoWiki and is annotated as dapD,  Dcan (pRB473 mpsABC) .1,000 .1,000 .1,000 .1,000 .1,000 .1,000 .1,000 .1 which is part of an operon consisting of six genes involved in the biosynthesis of lysine (25). In S. aureus, lysine is an important amino acid, as it is needed not only as a building block for proteins but also as a component of the cell wall peptidoglycan. Therefore, EVX10196.1 is not a CA. Last year, the same group which reported earlier that S. aureus encodes only g-CA now presented the production, kinetics, and inhibitory characterization of b-CA from the S. aureus (11). The CA gene was obtained from UniProt ID EZX15767 and was synthesized to produce a recombinant protein in Escherichia coli. A search in UniProt revealed that this protein is encoded by a gene V070_02709 from the S. aureus strain C0673. From our results above, C0673 is in fact S. sciuri and not S. aureus, and hence the CA activity described was actually from S. sciuri. A very recent publication from the same group followed up on the study by reporting its inhibition profile of S. aureus CA with anions and other small molecules (12). The same recombinant protein described earlier was used in this study, meaning that the CA inhibition referred to S. sciuri and not S. aureus. Our bioinformatics analyses have clearly shown that S. aureus CAs are wrongly annotated as such, while in fact they are not present in S. aureus (Table 1 to 5). The absence of a CA in S. aureus is also supported by the deletion of mpsABC in four different S. aureus backgrounds (4, 17) (Fig. S1) and the fact that S. aureus is resistant to CA inhibitor EZA whereas S. carnosus, which has been shown to possess only CA, was inhibited by EZA ( Fig. 1; Table 6; Table S6). The MIC values also imply that EZA is specific only for CA but not bicarbonate transporters (Table 6), which further complicates the treatment of pathogens such as S. aureus and S. epidermidis. The MIC values for CA-possessing S. carnosus and S. pseudintermedius in our study (64 to 250 mM) were comparable to those of Helicobacter S. carnosus TM300, and S. carnosus TM300 (pRB473-can), in which the CA was overexpressed, and S. pseudintermedius ED99. Paper disks impregnated with 10 ml of EZA, oxacillin (OXA), and vancomycin (VAN) as positive controls at concentrations of 1 mM each and appropriate concentration of DMSO as negative control were incubated at 37°C overnight in atmospheric and CO 2 conditions. serially diluted (from the highest concentration of 1 mM to the lowest concentration of 2 mM) with 50 ml of cationic adjusted Müller Hinton broth (MHB) in 96-well microtiter plates. Equal volumes of bacterial inoculum (1 Â 10 6 ) were added and the plates were incubated at 37°C with continuous shaking for 24 h in atmospheric air (Table 6) and, if necessary, in 5% CO 2 conditions (Table S6). The MIC was determined as the lowest concentration that completely inhibited visible growth of the bacteria and also confirmed with a TECAN Reader (Infinite M200). Antibiotics vancomycin and oxacillin were used as standard antibiotic controls, while positive controls referred to the bacterial cells treated with DMSO or methanol at a concentration equivalent to the highest concentration used to dissolve the CA inhibitors. MHB alone was used as negative control. The MIC determinations were performed in three independent biological replicates with three technical replicates each.
For visual representation of the semiquantitative results on agar, four strains that were inhibited by EZA were used (Fig. 1). MHB agar plates were swabbed with bacterial inoculum adjusted to an OD 578 of 0.1. Disks made of filter paper were impregnated with 10 ml of 1 mM EZA, vancomycin, and oxacillin (positive controls) and DMSO at appropriate concentration (negative control) before being placed on the agar. The agar plates were incubated overnight at 37°C in atmospheric air and 5% CO 2 conditions. Data availability. The main data supporting the findings of this work are available within the article and in the Supplemental Material or from the corresponding author upon reasonable request.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.4 MB.