Obtaining the sequences

To obtain bacterial ACC deaminase (AcdS) and ACC deaminase regulatory protein (AcdR) sequences, BLAST searches were performed in the NCBI databases (www.ncbi.nlm.nih.gov/) using Pseudomonas sp. UW4 acdS and acdR gene, as well as AcdS and AcdR protein sequences as the queries. For fungal ACC deaminase sequence retrieval, BLAST searches were performed in the NCBI database using the Penicillium citrinum AcdS protein sequence as the query. Default BLAST parameters were used when obtaining the sequences.

An NCBI genomic BLAST search (www.ncbi.nlm.nih.gov/sutils/genom_table.cgi) was also performed using Pseudomonas sp. UW4 acdS and AcdS sequences in order to evaluate the presence of ACC deaminase in other completely sequenced organisms. An additional BLAST search was performed in the nematode genomic database (www.nematodes.org) using Pseudomonas sp. UW4 or Penicillium citrinum acdS gene as query.

Moreover, all putative AcdS sequences were analyzed for key protein residues known to be important for ACC deaminase activity, namely Lys51, Ser78, Tyr295, Glu296 and Leu322 [7], [57], [58] using Pseudomonas sp. UW4 as a reference. The AcdS sequences were aligned using MUSCLE [59] and the presence of key amino acid positions were verified. Sequences presenting different amino acids in the above mentioned positions were discarded, as they are likely to represent related PLP dependent enzymes, such as D-cysteine desulfhydrase [58].

Sequence identities and similarities were analyzed using SIAS (http://imed.med.ucm.es/Tools/sias.html) with default parameters.

When available, the genomic regions containing the acdS gene were analyzed in order to identify any patterns present in the acdS gene neighborhood.

Strain information and 16S rRNA gene sequences were obtained via NCBI (http://www.ncbi.nlm.nih.gov), Goldcard (http://www.genomesonline.org/cgi-bin/GOLD/index.cgi) and SILVA (http://www.arb-silva.de), where available.

The accession numbers for sequences used in this study as well as strains descriptions are presented in Tables S1 (Actinobacteria, Deinococcus-Thermus and Firmicutes), S2 (α-Proteobacteria), S3 (β-Proteobacteria), S4 (γ-Proteobacteria), and S5 (Eukaryotes).

ACC deaminase protein sequence analysis and comparison to closely related enzymes

Protein sequence analysis was conducted on AcdS proteins found in completely sequenced representative bacteria. The functional AcdS protein sequences of the Proteobacteria Agrobacterium tumefaciens D3 [60], Azospirillum lipoferum 4B [40], Bradyrhizobium japonicum USDA110 [61], Mesorhizobium loti MAFF303099 [13], Phyllobacterium brassicacearum STM196 [62], Rhizobium leguminosarum 128C53K [12], Sinorhizobium meliloti SM11 [63], Burkholderia phytofirmans PsJN [14], Burkholderia graminis C4D1M [64], Ralstonia solanacearum GMI1000 [42], Variovorax paradoxus 5C2 [22], Pseudomonas sp. UW4 [38], Pseudomonas sp. ACP [1] and the Fungi, Cyberlidnera saturnus [1], Penicillium citrinum [65], Trichoderma asperellum T203 [16], together with the AcdS from Herbaspirillum frinsigense GSF30 [66] and the putative AcdS sequences from Agreia sp. PHSC20C1, Rhodococcus sp. R04 (Actinobacteria), Meiothermus ruber DSM1279 (Deinococcus-Thermus) were used. Sequences were aligned using MUSCLE and the presence of conserved and variable sites was analyzed.

Sequence comparisons were also performed with closely related enzymes. Therefore, D-cysteine desulfhydrase sequences from E. coli [8], as well as the ACC deaminase homologs from Pyrococcus horikoshi [7] and Solanum lycopersicum [58] were used and compared to the various ACC deaminase proteins.

Phylogenetic analysis

The sequences were aligned using MUSCLE and phylograms were constructed in Seaview v.4.2.12 [67] using PhyML [68].

In order to obtain the best substitution model for the construction of the phylogenetic trees, the resulting alignments were analyzed with jModeltest2 [69] and ProtTest [70]. The substitution models were chosen based on minimum BIC (Bayesian Information Criteria) values.

The acdS, acdR and 16S rRNA gene evolutionary history was inferred by using the Maximum Likelihood method based on the GTR model with a discrete Gamma distribution (4 Gamma categories). The AcdS and AcdR phylograms were constructed using the Maximum likelihood method based on the WAG model with a discrete Gamma distribution (4 Gamma categories). Branch support was evaluated using both aLRT (SH like) [71] and bootstrap analysis (100 replicates). Only bootstrap values above 0.75 (75%) are included in the phylograms. The resulting phylogenetic trees were plotted using FigTree v.1.4.1 (http://tree.bio.ed.ac.uk/software/figtree).

Estimates of evolutionary divergence between acdS sequences or 16S rRNA sequences in groups of bacterial strains were computed using MEGA software 6.06 [72]. The number of base substitutions per site from between sequences was calculated and analyses were conducted using the Maximum Composite Likelihood model with 1000 bootstrap replications. The analysis involved 3 nucleotide sequences per group of bacterial species, previously aligned using MUSCLE. Codon positions included were 1st+2nd+3rd+Noncoding. All positions containing gaps and missing data were removed.

ACC deaminase prevalence in completely sequenced organisms

After performing multiple searches in the NCBI database (http://www.ncbi.nlm.nih.gov/sutils/genom_table.cgi) using Pseudomonas sp. UW4 acdS gene as query, it was observed that the acdS gene is not commonly seen in most sequenced organisms. The acdS gene is mainly found in Actinobacteria, members from the Deinococcus-Thermus phylum (Meiothermus), three classes from Proteobacteria (α, β and γ), in various Fungi classes belonging to Ascomycota and Basidiomycota, and in Stramenopiles members.

These results are in agreement with previous reports, which have demonstrated ACC deaminase activity in many Actinobacteria, α, β and γ-Proteobacteria. Remarkably, putative acdS genes were found in Meiothermus, yet, there is no record of ACC deaminase activity in these thermophile strains. Putative acdS genes were also found many in members of Stramenopiles, mostly in Phytophthora. By computational analysis, Singh and Kashyap, [56] suggest that the acdS gene found in Phytophthora sojae encodes a functional ACC deaminase.

Interestingly, despite the known ACC deaminase activity display by bacteria belonging to the Bacteroidetes/Chlorobi or Firmicutes, it was not possible to identify acdS genes in the completely sequenced bacteria belonging to these phyla. In 478 completely sequenced bacteria (accessed in July, 2013) belonging to the Bacteroidetes/Chlorobi, including many Flavobacterium and Chryseobacterium species, the acdS gene is not found. Although candidate acdS genes are identified via BLAST, the active sites contain residues more consistent with D-cysteine desulfhydrase or a related PLP dependent enzyme [58], such as YP_001296100 in which threonines replace residues corresponding to active site residues E296 and L322. ACC deaminase activity has been previously reported to be present in Flavobacterium and Chryseobacterium species, although at very low levels [52], [55], which may represent non-specific activity of D-cysteine desulfhydrase-like enzymes. Similarly, although ACC deaminase activity has been described in many Bacillus and Paenibacillus strains [48]–[51], it was not possible to identify the acdS gene in 271 completely sequenced strains belonging to the Bacilli class (Firmicutes phylum), including many soil and plant associated Bacillus and Paenibacillus species.

It is possible that in these and many other bacterial strains the presence of an acdS gene may be related to a strain's specific feature in which acdS acquisition happened by HGT by result of a co-existence with other ACC deaminase-producing bacteria in environments where ACC deaminase production provides the bacteria with some important advantages. Other possible explanations for this inconsistency may relate to the fact that genome sequencing is biased and the sequenced strains may not be representative of bacteria that interact extensively with plants.

Analysis of ACC deaminase (putative and functional) protein sequences

In the first instance, every sequence used in this study (Table S1–S5) contains the previously described AcdS conserved regions that have been found to be necessary for ACC deaminase activity. Moreover, all bacterial AcdS sequences shared high sequence identity (60 to 100%) to AcdS from Pseudomonas sp. UW4.

When comparing the putative AcdS sequences from Fungi with the functional ACC deaminase from Penicillium citrinum, sequence identities ranged between ∼70 and 99% for the majority of fungal AcdS sequences. Exceptionally, some AcdS sequences from yeasts and some other fungi share only ∼52–55% identity to Penicillium citrinum AcdS. Also, the Stramenopiles members share approximately 60% identity to the Penicillium citrinum AcdS. Interestingly, the AcdS sequences from yeasts, some other Fungi and Stramenopiles share higher identity to Pseudomonas sp. UW4 AcdS sequence (∼70 to 85%), consistent with a relationship with Proteobacteria and the possibility of past horizontal gene transfers. A more detailed description of this issue is presented below.

Protein sequence analysis suggests that the putative acdS genes found in Rhodococcus sp. R04, Agreia sp., PHSC20C1 and Meiothermus ruber DSM1279, encode a true ACC deaminase. By sequence comparison, it was observed that the putative AcdS contain all the conserved features present in all known functional ACC deaminases and not present in the related enzymes (Figure S1). For instance, the putative AcdS sequences contain the important residues E295 and L322 known to be required for ACC deaminase activity [58] and not present in other related enzymes. These results are also supported by the fact that these Actinobacteria and Meiothermus AcdS protein sequences share high identity (70 to 82%) to the functional ACC deaminase from Rhodococcus sp. 4N-4 (partially characterized) [41]. In addition, these sequences show similar sequence identities to other β and γ-Proteobacteria AcdS sequences (∼70%).

ACC deaminase phylogeny: Horizontal gene transfer or vertical transmission?

The comparison between the acdS phylogenetic tree (Fig. 1) and the 16S rRNA based phylogeny (Fig. 2), suggests that ACC deaminase has evolved mainly through vertical transmission with occasional horizontal gene transfer. In the acdS phylogram (Fig. 1), it is observed that closely related strains typically have similar acdS gene sequences. Furthermore, many strains with different origins and isolated from different habitats (Tables S1–S5), but belonging to the same species tend to have similar acdS genes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Phylogram based on the acdS gene.

The evolutionary history was inferred by using the Maximum Likelihood method based on the GTR model. A discrete Gamma distribution was used to model evolutionary rate differences among sites (4 categories). Branch support was evaluated using both aLRT (SH like) and bootstrap analysis (100 replicates). Bootstrap values above 0.75 (75%) are displayed in the phylograms shown next to the branches as *. The analysis involved 335 nucleotide sequences and 931 patterns were found (out of a total of 1155 sites).

https://doi.org/10.1371/journal.pone.0099168.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Phylogram based on 16S rDNA sequences.

The evolutionary history was inferred by using the Maximum Likelihood method based on the GTR model. A discrete Gamma distribution was used to model evolutionary rate differences among sites (4 categories). Branch support was evaluated using both aLRT (SH like) and bootstrap analysis (100 replicates). Bootstrap values above 0.75 (75%) are displayed in the phylograms shown next to the branches as *. The analysis involved 272 nucleotide sequences and 768 patterns were found (out of a total of 1334 sites).

https://doi.org/10.1371/journal.pone.0099168.g002

The presence of the acdS gene in an organism like Meiothermus ruber is also consistent with the vertical transmission of this gene. It is unlikely that this bacterial thermophile (optimum growth at 60°C) isolated from a hot spring has acquired an acdS gene through HGT. This is strongly supported by the acdS gene phylogram (Fig. 1) showing a well bootstrap-supported and unique cluster grouping all Meiothermus acdS sequences distantly from all other acdS genes obtained from different bacterial phyla.

The presence of an acdS gene in the chromosome of the psychrophile marine actinobacterium, Agreia sp. PHSC20C1, (isolated in the Antarctic) and other soil Actinobacteria, is also consistent with the vertical transmission and ancient origin of the acdS gene. In Azorhizobium and Bradyrhizobium strains, the acdS gene is located far away from the “plastic” chromosomal symbiotic island containing the symbiotic genes. If these strains had acquired the acdS gene by HGT it might be expected that it would be present in a region that is more prone to such transfers, such as a symbiotic island or a plasmid.

Blaha et al. [42] and Glick et al. [3] have suggested that that the presence of acdS on plasmids may facilitate the lateral transfer of this gene. On the other hand, the presence of the acdS gene on a plasmid can also account for a different sequence divergence rate. Mobile elements and smaller replicons show higher evolutionary rates when compared to primary chromosomes [73], [74]. By being present on smaller replicons, acdS genes may be subject to different evolutionary rates compared to genes present in primary chromosomes. This may help to explain the acdS gene phylogeny of Burkholderia and Cupriavidus. Instead of clustering together with the other β-Proteobacteria, strains belonging to the Burkholderia and Cupriavidus genus form a separate cluster (Fig. 1). From the available data, most Burkholderia and Cupriavidus strains have the acdS gene present in a second smaller chromosome. Other β-Proteobacteria possesses an acdS gene in the primary chromosome or in plasmids (Table S3). This phenomenon is also observed in Agrobacterium and Rhizobium strains, A. vitis S4 and R. radiobacter K84 which have the acdS gene located in a second chromosome, and therefore, cluster distantly from their Agrobacterium tumefaciens D3 (acdS in plasmid) and Rhizobium (acdS in plasmid) relatives. Thus, there seems to be a connection between acdS phylogenetic distribution, evolution and acdS location in the replicon.

Environmental cues can also lead to different gene mutation rates [75]. Gene loss, acquisition, mutational rates and genome rearrangements may play a crucial role in bacterial adaptation and survival [76], [77]. This is particularly important in organisms living in adverse environments like many of the organisms described here (Tables S1–S5). It is possible that bacteria adapted to different environments may present different acdS divergence rates, thus being responsible for some of the variance in acdS genes in bacteria from the same species. When calculating the 16S rRNA and acdS gene evolutionary distance estimates in specific bacterial species groups it was found that the ratio between 16S rRNA and acdS sequence divergence is not always identical between strains and groups (file S1). For instance, three Burkholderia mallei strains isolated from three different countries show identical 16S rRNA (1200 bp) (d = 0) and identical acdS gene (1019 bp) (d = 0) sequences. In three Burkholderia silvatlantica strains obtained from Brazil, this is not observed; all strains present identical 16S rRNA sequences (1200 bp) (d = 0) but show intraspecific differences in the acdS gene sequences (1019 bp) (d = 0.0059±0.0020), sometimes accounting to up to 5 different nucleotides. Interestingly, all three B. mallei were obtained from human and animal blood and are known pathogens, while the three B. silvatlantica strains were obtained from the rhizosphere of different plants where they act like plant growth-promoting bacteria [64].

Several authors suggested HGT for acdS genes based on results showing a specific relative position of some Pseudomonas (γ-Proteobacteria) strains in the acdS phylogenetic tree [41], [42], [6]. Instead of forming a separate cluster, some Pseudomonas strains clustered together with β-Proteobacteria. In this work, we obtained somewhat similar results. Although members of γ-Proteobacteria group very close to β-Proteobacteria, they form a unique cluster and are not scattered through the phylogenetic tree as observed in previous studies. A very close evolutionary relationship between these two classes has been reported [78]–[81]. In fact, some bacterial strains that belonged to the Pseudomonas genus (γ-Proteobacteria) have been reassigned to the Burkholderia genus (β-Proteobacteria) [82], [83].

Interestingly, the Pseudomonas sp. ACP AcdS sequence shares higher identity (96.7%) with Burkholderia xenovorans LB400 functional ACC deaminase than with Pseudomonas sp. UW4 AcdS (85.3%). This is also observed in the AcdS phylogram, where Pseudomonas sp. ACP groups closer to Burkholderia xenovorans LB400 (Fig. 3). While, Honma and Shimomura [1] tentatively identified Pseudomonas sp. ACP bacterium by phenotypic methods, it is conceivable that Pseudomonas sp. ACP is in fact a Burkholderia strain [64]. If this is in fact the case, then previous studies regarding the phylogeny of acdS may also have been influenced by the confusing relationship between Pseudomonas and Burkholderia. Furthermore, due to the recent divergent evolution and close relationship between γ and β-Proteobacteria, it is very difficult to prove acdS HGT in these classes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Phylogram based on AcdS protein.

The evolutionary history was inferred by using the Maximum Likelihood method based on the WAG model. A discrete Gamma distribution was used to model evolutionary rate differences among sites (4 categories). Branch support was evaluated using both aLRT (SH like) and bootstrap analysis (100 replicates). Bootstrap values above 0.75 (75%) are displayed in the phylograms shown next to the branches as *. The analysis involved 431 amino acid sequences and 386 patterns were found (out of a total of 421 sites). Functional ACC deaminases are shown in bold.

https://doi.org/10.1371/journal.pone.0099168.g003

While less prevalent than previously thought, HGT likely does occur and accounts for a portion of acdS gene evolution. For example, it has been shown that some Mesorhizobium strains may acquire a specific acdS gene by the means of symbiotic island transfer [43]. Nandasena et al., [84] demonstrated that Mesorhizobium opportunistum WSM 2073 acquired a specific symbiotic island when it came in contact with non-endemic populations of M. ciceri bv. biserrulae, thus, gaining the ability to nodulate Biserrula pelecinus. The acdS gene was present within that symbiotic island and was therefore transferred between these strains. Moreover, the acdS gene sequences from those two strains share 100% identity, strongly supporting the idea of a recent transfer event.

Curiously, there are some cases where acdS horizontal transfers seem to have occurred between strains with a more distant evolutionary relationship. This is the case of Pseudomonas isolates GM 18, GM 55, GM 79 and GM 102, which are found to possess acdS genes like those of α-Proteobacteria (Fig. 1). Despite belonging to the γ-Proteobacteria, Vibrio gazogenes ATCC43941 has an acdS gene resembling those of α-Proteobacteria (Fig. 1). Chen et al., (2013) showed that Bacillus cereus AcdSPB4 isolated from the casing soil of Agaricus bisporus possesses an acdS gene highly similar to those of Pseudomonas (Fig. 1) thus, strengthening the idea of acdS horizontal transfer between distantly related strains.

Interestingly, Herbaspirillum seropedicae SmR1, H. frisingense GFS30, H. huttiense subsp. putei AM15032, H. sp. B501, H. sp. GW103, H. sp. YR522, Pseudomonas psychrotolerans L19 and Pseudomonas sp. 313 strains possess acdS genes that are not similar to those found in other bacteria from the same Class or even to other Herbaspirillum and Pseudomonas strains. Instead they form a unique group in the phylogenetic tree (Fig. 1, 3). Furthermore, these strains also possess acdR genes that are frequently found in AcdS+ Proteobacteria. In this scenario it is possible that these strains have horizontally acquired acdS and acdR genes from a different class of bacteria yet to be determined. One may also assume that the putative acdS genes in these strains encode a different type of deaminase, however, ACC deaminase activity has been detected in Herbaspirillum frisingense GFS30 [66]. Moreover, the putative AcdS from H. frisingense GFS30 shows the conserved regions known to be important in functional ACC deaminases (i.e E295, L322) (Fig. S1). Curiously, Herbaspirillum frisingense GFS30 and Herbaspirillum sp. YR522 are both endophytes isolated from Miscanthus and Populus deltoides, respectively.

Similar to what is observed in Bacteria, the AcdS phylogeny in Fungi indicates that closely related strains possess a similar ACC deaminase (Fig. 3). This is consistent with the notion that acdS genes are vertically transmitted in Fungi. However, some fungal strains like Penicillium marneffei and Talaromyces stipitatus (Ascomycota/Eurotiomycetes) are likely to have acquired the acdS from other Fungi belonging to the Sordariamycetes class, suggesting that like in some bacteria, fungal acdS genes may also be acquired by HGT.

In addition, the yeasts Cyberlindnera saturnus, Cyberlindnera jadinii NBRC 0988, Clavispora lusitaniae ATCC 42720 (Ascomycota/Saccharomycetes) and Schizosaccharomyces pombe 972h- (Ascomycota/Schizosaccharomycetes) seem to have acquired an ACC deaminase gene separately from most Ascomycota and presumably from Proteobacteria. ACC deaminase genes like those of Proteobacteria have also been detected in Fungi belonging to different classes such as Punctularia strigosozonata HHB-11173 (Basidiomycota/Agaricomycetes), Fomitopsis pinicola FP-58527 (Basidiomycota/Agaricomycetes), Aureobasidium pullulans AY4 (Ascomycota/Dothideomycetes), Macrophomina phaseolina MS6 (Ascomycota/Dothideomycetes) and Guignardia citricarpa CGMCC3.14348 (Ascomycota/Dothideomycetes).

The Stramenopiles, Phytophthora infestans T30-4, P. ramorum Pr102, P. sojae P6497, P. lateralis, P. kernoviae, P. parasitica, Pseudoperonospora cubensis and Hyaloperonospora arabidopsidis Emoy2 also have ACC deaminase genes most similar to those of Proteobacteria (Fig. 3).

Searches of diverse genomic databases also have revealed the presence of putative acdS genes in other eukaryotic organisms like the nematode Howardula aoronymphium and the fly Drosophila eugracilis. Furthermore, these genes show high similarity to acdS from Proteobacteria (Fig. 3). Some acdS genes are found in bacteria known to be associated with Eukaryotic organisms, for example, Serratia sp. M24T3, isolated from the nematode Bursaphelenchus xylophilus, and Pantoea sp. At-9b, the leaf cutter ant symbiont (Table S4). While it is possible that Howardula aoronymphium and Drosophila eugracilis may have acquired acdS genes from associated bacteria, it is most likely that the presence of acdS in these organisms results from contamination of genomic DNA.

AcdR phylogeny: Have AcdR and AcdS undergone a coupled evolution?

In the study conducted by Prigent-Combaret et al. [40], 45 of 48 studied Proteobacteria were found to possess an LRP homolog (acdR) near the acdS gene. Here, we report the presence of acdR in, at least, 166 of 261 Proteobacteria possessing an acdS gene. Still, it was not possible to obtain the acdR sequence in many (78) organisms and others only have their acdS gene described. The acdR gene was not found at least in 17 acdS+ completely sequenced strains (6 Mesorhizobium strains, 2 Rhizobium strains, Fulvimarina pelagi, 3 root nodule Burkholderia strains containing the 2^nd acdS copy in a plasmid, Halomonas titanicae BH1 and 4 Pseudomonas strains).

Moreover, the acdR gene is found in the opposite direction of the acdS gene in most studied Proteobacteria (data not shown). This is consistent with the previous reports of Grichko and Glick [37], Ma et al. [12] and Prigent-Combaret et al. [40]. This data suggest that acdR is a common mechanism regulating ACC deaminase expression in most Proteobacteria. Moreover, the phylogeny of acdR (Fig. 4) is related to the acdS gene phylogeny (Fig. 1), suggesting that these genes evolved in a similar and dependent manner. Closely related strains have similar acdR genes, as also observed in the acdS phylogram, suggesting that acdR is primarily vertically inherited. In the phylogram based on the acdR gene it is also observed a grouping according to the bacterial Class (taxonomy) and the gene location in the replicon (Example: 2nd chromosome location in Burkholderia and Cupriavidus vs. primary chromosome/plasmid location in other β and γ-Proteobacteria).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Phylogram based on acdR gene.

The evolutionary history was inferred by using the Maximum Likelihood method based on the GTR model. A discrete Gamma distribution was used to model evolutionary rate differences among sites (4 categories). Branch support was evaluated using both aLRT (SH like) and bootstrap analysis (100 replicates). Bootstrap values above 0.75 (75%) are displayed in the phylograms shown next to the branches as *. The analysis involved 166 nucleotide sequences and 509 patterns were found (out of a total of 594 sites).

https://doi.org/10.1371/journal.pone.0099168.g004

Interestingly, there are few cases where it seems that the acdR and acdS are not inherited together or may have undergone genomic rearrangements. While some strains don't have acdR genes in the vicinity of the acdS gene others don't have an acdR gene at all. In Gluconacetobacter xylinus NBRC 3288 there are various regions coding putative LRP in the upstream region of the acdS gene. However, they are not true acdR genes. A sequence sharing high homology to the acdR gene is found far away (aprox. 9 kb) from the acdS gene. This is also observed in Burkholderia xenovorans LB400. In this case, despite the fact that this strain has an acdR gene located far from the acdS gene, it is still able to express ACC deaminase [64]. In Rhizobium leguminosarum bv. viciae 3841 an acdR gene is not found. In Mesorhizobium loti MAFF303099 the acdR gene is also not present, but in this case, the acdS gene transcription is regulated by NifA [13].

It is possible that genome rearrangements or gene insertions in smaller replicons can account for the absence of acdR genes in some acdS+ bacterial strains. The strains Burkholderia sp. CCGE1002, B. phymatum STM815 and B. phenoliruptrix BR3459a (isolated from root nodules) have two copies of the acdS gene, one on the second chromosome and the other on a megaplasmid. The acdS gene copy present on the megaplasmid seems to be the result of acdS gene duplication and later insertion into this smaller replicon. This is consistent with the high identity between the two acdS copies and also the presence of transposase genes in the immediate upstream and downstream regions of the acdS gene. In this case, the acdR gene is not present and may have been lost in this process.

The exceptions notwithstanding, in the majority of cases in Proteobacteria it appears that the evolution of acdS and acdR is coupled. This result is in agreement with previous reports showing that acdR is necessary for optimum ACC deaminase expression [38], [39].

Despite being mostly inherited together, it is observed that these genes may have different evolutionary rates. Thus, for example, compared to the Pseudomonas sp. UW4 AcdR sequence, other AcdR sequences from Proteobacteria show identities ranging from 51% to 87%. This degree of variability is not observed in Proteobacteria AcdS sequences. By coding a regulatory protein it is most likely that the acdR gene is more prone to modifications, thus, allowing fine-tuning of acdS transcription and expression.

Previously, Nikolic et al. [85] stated that “the acdR–acdS gene cluster is rather rare and typically occurs in few α and β-Proteobacterial genera” based on finding the acdR–acdS gene cluster in four α, six β-Proteobacteria and in only three Pseudomonas syringae strains. They concluded that the operon is rather uncommon among γ-Proteobacteria. However, more detailed data presented by Prigent-Combaret et al. [40] and also in this study, supports a widespread occurrence of acdR–acdS gene cluster. The apparent lack of acdR-acdS clusters in Nikolic et al. [85] may be due to the inclusion of putative ACC deaminase sequences that were not confirmed by comparison with conserved protein domains. Thus, sequences coding for D-cysteine desulfhydrases, and possibly other deaminases and aminotransferases, were considered as ACC deaminases, leading to a confusing relationship between acdS and acdR and also the presence of acdS in some bacterial groups.

Other mechanisms regulating ACC deaminase transcription

The expression of ACC deaminase by organisms that don't possess acdR genes indicates that the presence of this regulator is not absolutely necessary for acdS transcription. The presence of CRP and FNR binding sites in the immediate upstream region of the acdS gene in many Proteobacteria [37], [40] suggests that these elements can also account for ACC deaminase expression regulation in some Proteobacteria. It has been demonstrated directly in some instances that FNR as well as CRP regulate acdS transcription [37], [38], [40]. The NifA protein is also a known regulator of ACC deaminase expression in Mesorhizobium loti MAFF303099. In this strain ACC deaminase expression occurs only inside of formed nodules [13], [86]. In addition, a NifA binding site is found in the immediate upstream region of the acdS gene in this and many other Mesorhizobium strains, suggesting that this regulatory mechanism is widespread in this genus [43]. Interestingly, the NifA binding site (5′-TGT-N_9–11-ACA-3′) is quite similar to the CRP binding site (5′-TGTGA-N₆-TCACA-3′).

In many Actinobacteria and in Meiothermus, a gene encoding a protein from the GntR family of transcriptional regulators is found next to the acdS gene. We putatively termed it acd-AR (Actinobacteria) and acd-MR (Meiothermus). When performing BLAST searches using one Acd-AR protein sequence as query, the main hits are always related to other Acd-AR protein sequences found in acdS+ Actinobacteria, suggesting a close relationship between acdS and acd-AR. The same trend is observed in Meiothermus despite the fact Acd-AR shares low identity to Acd-MR. There are no sequences in the database that share a high degree of similarity to Acd-MR. These results are consistent with the possibility that both acd-AR and acd-MR might be involved in the regulation of ACC deaminase expression in these organisms. Curiously, when analyzing the immediate upstream region of the acdS gene in various Actinobacteria it is observed that some strains appear to have no promoter regions (Fig. S2-A). The same is observed in Meiothermus. In these strains the acdS gene forms an operon together with the acd-AR gene and acd-MR gene, respectively.

Interestingly, in Nocardioidaceae Broad-1 a leucine responsive protein is found in the vicinity of the acdS gene, however, it is quite different from the Proteobacteria AcdR protein. Also, in some Actinobacteria and Proteobacteria strains the acdS gene is located near a transcriptional regulator belonging to the LysR family. Moreover, in Saccharopolyspora erythraea NRRL 233 and Streptomyces hygroscopicus ATCC 53653 strains the acdS gene also appears to be part of an operon consisting of a gene encoding a MFS family protein and another gene encoding a M20 peptidase (Fig. S2-B). Interestingly, a LysR transcriptional regulator is also found in the vicinity of the acdS gene in some Proteobacteria like Brenneria sp. EniD312, Burkholderia xenovorans LB400, Dickeya spp. and Pantoea sp. At-9b (Fig. S2-C-H). The presence of peptidase M20 in the vicinity of acdS is also observed in some of these strains.

Further studies are necessary in order to characterize the importance of these regulators in ACC deaminase expression in different organisms.

ACC deaminase origin

To gain additional knowledge regarding the origin and evolution of ACC deaminase multiple searches of the database were conducted; sequences showing a high similarity to different deaminases were obtained and a phylogram was constructed (Fig. 5). In this instance it was observed that ACC deaminase forms a distinct and unique group, where ACC deaminases from different organisms like Bacteria and Fungi cluster together. This is also observed with D-cysteine desulfhydrase, however only a few representatives of the considered “true” D-cysteine desulfhydrases (E.coli D-cysteine desulfhydrase) were obtained. When searching in the database, it was observed that D-cysteine desulfhydrase is an enzyme whose distribution is not widespread and it may also be not nearly as conserved as ACC deaminase. Its presence has been verified mainly in γ-Proteobacteria. Other proteins showing some homology were found in Firmicutes and other α-Proteobacteria, but in those instances showing low identity scores (39%).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Phylogram constructed based on ACC deaminase and related PLP enzymes protein sequences.

The evolutionary history was inferred by using the Maximum Likelihood method based on the WAG model. A discrete Gamma distribution was used to model evolutionary rate differences among sites (4 categories). Branch support was evaluated using both aLRT (SH like) and bootstrap analysis (100 replicates). Bootstrap values above 0.75 (75%) are displayed in the phylograms shown next to the branches as *. The analysis involved 99 aminoacid sequences and 570 patterns were found (out of a total of 594 sites). Sequences used for the construction of this phylogram are described in Table S6.

https://doi.org/10.1371/journal.pone.0099168.g005

Interestingly, D-cysteine desulfhydrase activity has been demonstrated for Solanum lycopersicum [58] and Arabidopsis thaliana [87]. However, these enzymes form a distinct phylogenetic cluster, far away from E. coli and other γ-Proteobacteria D-cysteine desulfhydrase. Controversially, ACC deaminase activity has also been demonstrated for Arabidopsis thaliana. Although Riemenschneider et al. [87] did not detect ACC deaminase activity from the product of Arabidopsis thaliana gene “At1g48420”, McDonell et al. [88] showed that this gene encoded a protein with the ability to breakdown ACC. Moreover, McDonnel et al. [88] suggest that the gene is responsible for regulation of Arabidopsis thaliana endogenous ACC levels. The same authors also suggest that this enzyme may be present in many other plant species. Curiously, Todorovic and Glick [58] did not find ACC deaminase activity in the Solanum lycopersicum At1g48420 protein homolog, sharing 70% identity and clustering together with Arabidopsis thaliana At1g48420 protein (Fig. 5).

Despite showing D-cysteine desulfhydrase and ACC deaminase activity in vitro, it is conceivable that the at1g48420 gene product does not represent a “true” D-cysteine desulfhydrase or ACC deaminase, or at least, is only distantly related to bacterial D-cysteine desulfhydrase and ACC deaminase. The grouping that is observed in Fig. 5 supports this latter conclusion. Instead of clustering with bacterial ACC deaminase or D-cysteine desulfhydrase, the plant protein homologs form a distant and unique cluster with a different phylogenetic background within the broader family of these PLP-dependent enzymes.

Proteins with similar origin and function often tend to be conserved. Thus, if the at1g48420 gene encoded a true ACC deaminase (similar phylogenetic background and specialization towards ACC breakdown), it would likely cluster together with bacterial and fungal ACC deaminases and have similar amino acid residues in specific sites as are present in functional ACC deaminases that are important for ACC breakdown (i.e E295, L322). Nevertheless, it is possible that proteins like At1g48420 evolved and specialized in ACC degradation in a different route than those bacterial and fungal ACC deaminases. In this case, additional studies are necessary to further unveil the characteristics of At1g48420 like proteins.

It is most likely that the ability of at1g48420 gene product to use D-cysteine or ACC as substrates results from the high promiscuity that many deaminases show in cleaving multiple substrates that share similar characteristics. For example, it has been shown that ACC deaminase from Pseudomonas sp. ACP is able to use multiple substrates like D-cysteine and also other D-aminoacids. E. coli and S. thyphimurium D-cysteine desulfhydrases are able to efficiently use β-chloro-D-alanine (β-CDA) and other substrates (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Substrate cleavage abilities of studied ACC deaminase, D-cysteine desulfhydrase and other PLP dependent (ACC deaminase or D-cysteine desulfhydrase homologs) enzymes.

https://doi.org/10.1371/journal.pone.0099168.t001

Intriguingly, Todorovic and Glick [58] demonstrated that mutations in amino acids (E295S/L322T) in Pseudomonas sp. UW4 ACC deaminase lead to the loss of ACC deaminase activity, yet, these mutations conferred an increased D-cysteine desulfhydrase activity to the mutant enzyme. The Km of the double mutant for D-cysteine was much lower than the Km of the native ACC deaminase towards ACC. Moreover, the Km of the double mutant enzyme towards D-cysteine is equivalent to that of a true D-cysteine desulfhydrase. Still, this mutant enzyme shows great inefficiency (K cat = 10.9 min-1) in D-cysteine cleavage. This data shows that small amino acid changes can confer different substrate usage abilities in closely related deaminases.

If ACC deaminase can use multiple substrates, it is possible that in some organisms the production of ACC deaminase can be important for cleavage of such substances, thus, giving these organisms the ability to use other nutrient sources, or to grow under otherwise toxic conditions. This can have major implications in a microorganism's fitness, especially in organisms living under limiting conditions. For example, Soutourina et al. [89] demonstrated that the expression of D-cysteine desulfhydrase by E.coli relieved some of the toxic effects of D-cysteine in bacterial growth. Also, D-cysteine desulfhydrase production allowed the growth of E. coli in a minimal medium containing D-cysteine as the sole sulfur source, demonstrating the importance of D-cysteine desulfhydrase in bacterial growth under sulfate limitation. It is possible that by maintaining a broad ability to cleave ACC-like substrates and some D-aminoacids, ACC deaminase genes were maintained in organisms that live in environments where ACC is not present. On the other hand, the presence of ACC deaminase in organisms that are associated with plants or other ACC-producing organisms, gave them a significant advantage in their ecology so that acdS genes were maintained. This may have led to the significant acdS gene presence in plant associated microorganisms, especially plants grown under perennially stressful conditions [51], and increased acdS gene loss in microorganisms living in environments where ACC is nonexistent.

Overall, it seems that bacterial and fungal ACC deaminases (here considered to be the representatives of true ACC deaminases) belong to a large group of PLP dependent deaminases (including bacterial D-cysteine desulfhydrase) related to tryptophan synthase beta subunit and sharing a common origin. Further, mutations and other evolutionary forces may have lead to some level of substrate specialization. Yet, some conserved features appear to allow these enzymes to be able to utilize a wide range of related substrates. This is exemplified by the data presented in Table 1.

ACC deaminase phylogenetic distribution and evolution

A model for ACC deaminase evolution and phylogenetic distribution is proposed based on the AcdS phylogenetic analysis, AcdS protein sequence analysis, acdS gene location, organism habitat and origin. The evolutionary relationship among Archaea, Bacteria and Eukaryotes was also taken into account when attempting to resolve the evolution of ACC deaminase [90], [91], [78].

From the available sequence data, it would appear that the most ancient point for the origin of ACC deaminase in Bacteria dates to the Actinobacteria or Deinococcus-Thermus. Most Actinobacteria strains investigated (Table S1) possess an acdS gene in their primary and unique chromosome. In another ancient bacterial lineage, the Deinococcus-Thermus, the acdS gene is also found in the chromosome of its representatives Meiothermus, suggesting an acdS chromosomal location in a common ancestor for Bacteria.

In many α-Proteobacteria, including Azorhizobium and Bradyrhizobium, the acdS gene is found in the primary chromosome of these strains. The acdS gene is also found in the primary chromosome of many marine α-Proteobacteria, and in the vinegar isolate Gluconacetobacter xylinus NBRC 3288. Despite the fact that ACC deaminase genes were not yet detected in any δ or ε -Proteobacteria (181 genome sequences available in the database), the presence of acdS genes in α-Proteobacterial strains that live in environments where ACC is not present, suggests that acdS was present in a Proteobacteria ancestor, located in its primary chromosome and likely was acquired by vertical transmission.

Many α-Proteobacteria have acdS genes located on plasmids, symbiotic islands or second chromosomes; this is the case of the studied Rhizobiaceae (Rhizobium, Sinorhizobium, and Agrobacterium), Phyllobacteriaceae (Phyllobacterium and Mesorhizobium) and Azospirillum strains (Table S2). Extensive gene transfer analysis between completely sequenced α-Proteobacteria suggested that secondary chromosomes originated from intragenomic transfers from primary chromosomes to ancestral plasmids [92]. This mechanism may have not only led to the origin of a second chromosome in some α-Proteobacteria but also in other Proteobacteria. In this regard, it is possible that acdS was transferred from the primary chromosome to a plasmid in some α-Proteobacteria. This possibility is consistent with the presence of acdS genes in the plasmids of Rhizobium and Sinorhizobium species and in the second chromosome of Rhizobium radiobacter K84 and Agrobacterium vitis S4. Slater et al. [92] also suggested that some strains like Mesorhizobium could have evolved by plasmid gene integration into the main chromosome. This suggestion is consistent with the observation that the same gene arrangement is found between Mesorhizobium symbiotic islands and some Rhizobium and Sinorhizobium symbiotic plasmids, where the acdS gene is located.

It is likely that intragenomic transfers of acdS genes from primary chromosomes to plasmids may have occurred in members of α-Proteobacteria as well as in β and γ-Proteobacteria. The presence of acdS genes in the second chromosome of Burkholderia and megaplasmids in Ralstonia and some strains of Pseudomonas is consistent with this idea. The occurrence of such phenomena may have led to a dispersal of acdS genes through plasmids that are readily transmissible between closely and more distant related strains. This leads to the puzzling phylogeny of the acdS gene that is observed in bacteria belonging to the same Order or Family (taxonomy) (Fig. 1).

In most Fungi, AcdS sequences share an average sequence identity of ∼50% with Bacterial AcdS. An exception to this case is the AcdS from yeasts, Punctularia strigosozonata HHB-11173, Fomitopsis pinicola FP-58527 (Basidiomycota/Agaricomycetes), Aureobasidium pullulans AY4, Macrophomina phaseolina MS6 and Guignardia citricarpa CGMCC3.14348 (Ascomycota/Dothideomycetes), and Stramenopiles, which appear to have a Bacterial origin. As observed in the AcdS based phylogram (Fig. 3), it seems that Fungal (excluding the above mentioned exceptions) and Bacterial AcdS diverged long ago. At this point it's not possible to corroborate both hypothesis of AcdS monophyletic or paraphyletic origin. Still, protein sequence analysis show some conserved amino acid regions (e.g. His80 and Ala161) in ACC deaminase from Fungi, Actinobacteria, Deinococcus-Thermus, and α-Proteobacteria, suggesting a common origin for acdS in these organisms. Organisms belonging to β- and γ-Proteobacterial classes show different amino acids in the referred positions, suggesting a later divergence from the α-Proteobacteria and the rest of ancient classes. The Fungi grouping closer to Actinobacteria is also observed in the phylogram (Fig. 3) suggesting a common origin for ACC deaminase in these organisms.

Based on the currently available data, we suggest that acdS genes had an ancient origin that may date to a Eukaryote and Bacterial common ancestor that possessed this gene in its chromosome. Furthermore, it is most likely that ACC deaminase originated as a consequence of specific mutations in an already existing PLP dependent enzyme showing high similarity to tryptophan synthase beta subunit. This is consistent with the results obtained by Todorovic and Glick [58] showing that small amino acid changes in related enzymes can be responsible for the ability to use a specific substrate.

Through time, it is probable that the acdS gene evolved by continuous vertical transmission, in which different constraints like habitat adaptation led to acdS divergence and sometimes gene loss. Intragenomic transfers of acdS genes from primary chromosomes to plasmids may have been selected for as a consequence of the advantage of ACC deaminase production, and this probably led to HGT events and increased divergence of acdS genes. These intragenomic transfer events and the presence of acdS on plasmids may have also lead to gene loss in many organisms. This is consistent with the results obtained by Prigent-Combaret et al. [40] showing that Azospirillum lipoferum 4B loses the plasmid containing an acdS gene during phenotypic variation events.

The role of ACC deaminase production in microorganism's ecology and fitness

From the available information, it is observed that many of the acdS+ organisms here described were isolated from heavily contaminated soils or otherwise stressed environments (Tables S1–S5), suggesting that ACC deaminase-producing microorganisms are more prevalent and better able to live in such conditions.

Organisms that produce ACC deaminase normally bind to plant tissues, and take up ACC to convert into ammonia and α-ketobutyrate [2]. The products of ACC cleavage are potential nitrogen and carbon sources [2], [33] that can play a role in the microorganism's fitness under stressful situations. Under stress conditions plants produce higher levels of the phytohormone ethylene, which means that the plants also produce higher levels of ACC [3]. Microorganisms that bind to plant tissues typically utilize plant exudates as a nutrient source. Under stress conditions, not only is the amount of ACC produced by the plant increased, the vast majority of rhizosphere microorganisms produce the phytohormone indoleacetic acid (IAA) which acts to loosen plant cell walls thereby facilitating root exudation. Bacterial IAA production has also been shown to increase ACC synthase expression in plants [93]. Thus, microorganisms that can both produce IAA and utilize ACC may have a competitive advantage over other soil microorganisms [2], [94].

Importantly, a recent study by Timmusk et al. [51], showed that ACC deaminase-producing organisms were more much abundant in the rhizosphere of wild barley (Hordeum spontaneum) growing in a stressed environment than they were in a similar (nearby) less stressed environment. This result was obtained despite the fact that both environments had similar soil, rock and topology characteristics. In addition, ACC deaminase-producing bacteria were abundant in plant rhizosphere samples and almost nonexistent in bulk soil samples. This suggests that organisms that produce ACC deaminase more readily survive in stressed environments by the mutualistic interaction with a plant host.

By degrading ACC, microorganisms decrease plant ethylene levels that under stress conditions are responsible for plant senescence and ultimately plant death [3]. Therefore, these organisms facilitate plant health under stress conditions. In turn, healthier plants provide their associated microorganisms with more nutrients thereby increasing the proliferation of these microorganisms.

Chen et al. [95] demonstrated that ACC deaminase-producing bacteria are also present in the casing soil of the ethylene-producing fungi Agaricus bisporus. The authors proposed a new model for the interaction between fungi and ACC deaminase producing bacteria. Bacteria possessing an acdS gene were able to increase fungal primordium initiation and proliferation by reducing endogenous ACC levels and consequently the inhibitory ethylene levels known to affect fungal development. These results show that ACC deaminase-producing bacteria might not only associate with plants but also with fungi, bringing significant advantages to fungal colonization in soil. On the other hand, bacteria producing ACC deaminase gain significant advantages by associating with extreme soil and plant colonizers like fungi. Being that these organisms constantly produce ACC, bacteria able to degrade ACC may gain extra nutrient sources as previously suggested.

ACC deaminase in Fungi: Relationship with plants or regulation of endogenous ACC levels?

The production of ACC deaminase by Trichoderma asperellum T230 has been shown to be an important mechanism for the plant growth promotion abilities of this fungal strain [16]. When ACC deaminase production is impaired, the fungal ability to promote canola root elongation is decreased, therefore, suggesting that ACC deaminase may act in a similar way as previously described by Glick et al. [2] for plant growth-promoting bacteria.

Nonetheless, it has been shown by Jia et al. [65] that ACC deaminase in Penicillium citrinum is produced independently of a relationship with a plant host. This happens because Pennicillium citrinum is capable of producing and accumulating ACC in its tissues. That is, Penicillium citrinum possesses not only an acdS gene but also an ACC synthase gene. Jia et al. [96] found that the ACC deaminase was induced by the presence of accumulated ACC in the intracellular spaces of Penicilium citrinum, indicating that ACC deaminase may participate in the regulation of ACC levels in this strain. As a consequence, ethylene production by P. citrinum can also be regulated by ACC deaminase. In fact, our search of the database revealed the presence of ACC synthase homologs in most fungal strains that possess an ACC deaminase (data not shown). Together these results suggest that ACC deaminase production by Fungi can account for the regulation of endogenous ACC concentrations, and therefore regulation of ethylene levels which can inhibit primordium initiation and formation.

What is the role of ACC deaminase in pathogenic microorganisms?

Surprisingly, acdS and acdR genes are found in a wide range of plant and human pathogenic microorganisms (Tables S1-S5), suggesting that ACC deaminase may play a role in these microrganisms' ecology. For example, the production of ACC deaminase has been reported in the human pathogenic Burkholderia cenocepacia J2315 [64]. However, this bacterial strain, like other pathogenic Burkholderia strains, is predominant in soils where it normally associates with plants [97]–[99]. The acdS gene is also found in pathogenic fungi like Aspergillus spp. and Myceliophthora thermophila. Despite causing severe diseases in immunocompromised humans, these strains are mainly found in soil [100], [101]. This data suggests that the presence of acdS genes in human pathogenic organisms may not be related to their human pathogenesis mechanisms but rather to their possible ecological role in soil. Also, it is possible that the presence of acdS gene in these strains and in plant pathogenic bacteria is related to the continuous acdS vertical transmission and not to any beneficial effects of ACC deaminase production.

Nevertheless, ACC deaminase production by pathogenic microorganisms may ultimately play a role in: (a) obtaining extra nutrients sources from ACC or ACC like substrate degradation, (b) the plant or fungi growth promoting abilities of these organisms when they are not acting as human or plant pathogens (“opportunistic” pathogens), (c) augmenting the ability to overcome ethylene or ACC mediated plant response systems, (d) regulation of endogenous ACC levels, or a combination of these factors.