A novel decarboxylating amidohydrolase involved in avoiding metabolic dead ends during cyanuric acid catabolism in Pseudomonas sp. strain ADP

Cyanuric acid is a common environmental contaminant and a metabolic intermediate in the catabolism of s-triazine compounds, including atrazine and other herbicides. Cyanuric acid is catabolized via a number of bacterial pathways, including one first identified in Pseudomonas sp. strain ADP, which is encoded by a single, five-gene operon (atzDGEHF) found on a self-transmissible plasmid. The discovery of two of the five genes (atzG and atzH) was reported in 2018 and although the function of atzG was determined, the role of atzH was unclear. Here, we present the first in vitro reconstruction of the complete, five-protein cyanuric acid catabolism pathway, which indicates that AtzH may be an amidase responsible for converting 1,3-dicarboxyurea (the AtzE product) to allophanate (the AtzF substrate). We have solved the AtzH structure (a DUF3225 protein from the NTF2 superfamily) and used it to predict the substrate-binding pocket. Site-directed mutagenesis experiments suggest that two residues (Tyr22 and Arg46) are needed for catalysis. We also show that atzH homologs are commonly found in Proteobacteria associated with homologs of the atzG and atzE genes. The genetic context of these atzG-atzE-atzH clusters imply that they have a role in the catabolism of nitrogenous compounds. Moreover, their presence in many genomes in the absence of homologs of atzD and atzF suggests that the atzG-atzE-atzH cluster may pre-date the evolution of the cyanuric acid catabolism operon.


Introduction
The symmetrical triazine cyanuric acid (1,3,5-triazine-2,4,6-triol) is a common anthropogenic compound that is used in the synthesis of a variety of disinfectants (e.g., trichlorocyanuric acid) and herbicides (e.g., atrazine). It is catabolized and used as a nitrogen source by a number PLOS ONE | https://doi.org/10.1371/journal.pone.0206949 November 6, 2018 1 / 17 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 by 1-carboxybiuret amidohydrolase (AtzE) to form 1,3-dicarboxyurea [8]. 1,3-Dicarboxyurea must then undergo a single decarboxylation step to deliver allophanate to AtzF, which is again deaminated to ultimately produce ammonia and carbon dioxide [10][11][12][13]. Several of the metabolic intermediates in the 1-carboxybiuret-dependent pathway are unstable under physiological conditions: 1-carboxybiuret forms biuret and CO 2 , 1,3-dicarboxyurea and allophanate both form urea and CO 2 , and dicarboxyammonia (from allophanate deamination) forms ammonia and CO 2 . Pseudomonas sp. strain ADP is unable to degrade biuret or urea, and these solvent-mediated reactions effectively form metabolic 'dead-ends' [2,10]. Presumably, the enzyme-catalyzed reactions have rates that are sufficient to avoid this outcome, effectively using the enzyme-mediated rate enhancement to control which hydrolyses occur. In this context, it is perhaps counter-intuitive that the conversion of 1,3-dicarboxyurea to allophanate would be uncontrolled, which would potentially lead to urea accumulation via the solvent-mediated process; however, to date no enzyme has been identified that catalyzes this step [8].
Previously, proteomic analysis had demonstrated that AtzH is expressed from the cyanuric acid catabolism operon. However, its function had not been explored in previous studies [8]. Herein, we present the structure of AtzH and explore its potential function.

Cloning and mutagenesis
The AtzH gene, coding for the protein WP_006390399.1 (KSW21442.1), was cloned directly from Pseudomonas sp. strain ADP genomic DNA with appropriate primers (5' to 3': CGACGA CATATGCTCGAGATGCAAATTAATCTACC and TGCTGCGAGCTCCCTAGGTCAGGAAACG GGC), before being subcloned into the NdeI and BamHI sites of the pETcc2 [9], which enabled production of AtzH with an N-terminal hexa-his-tag with thrombin cleavage site (S1 Fig). Cloning of the other genes used in this work has been described elsewhere [9,12,13].

Heterologous protein expression and purification
The expression vectors were used to transform Escherichia coli BL21 (λDE3) cells (Invitrogen). Bacteria were grown on Luria-Bertani (LB) medium containing ampicillin 100 μg/mL for the pETcc2 constructs or chloramphenicol 34 μg/mL for the pACYCDuet-1 construct (used to coexpress AtzE and AtzG). Cells were grown with shaking at 200 rpm at 28˚C. Protein expression was induced at an OD 600 of 0.8 by addition of isopropyl-β-D-1-thiogalactopyranoside (IPTG; 1 mM final concentration).
Cells were harvested 24 hours after induction by centrifugation at 5000 x g for 15 minutes using an Aventi J-E centrifuge (Beckman Coulter, Indianapolis, USA), resuspended in lysis buffer (25 mM potassium phosphate, 5 mM imidazole, pH 7.5) and lysed by passage through a Microfluidics homogenizer M-110P (Massachusetts, USA) five times at 15000 PSI. The lysis was followed by centrifugation at 18000 x g for 45 minutes to pellet the cellular debris, and the soluble fraction was used for further purification.
The soluble fraction was syringe filtered through a 0.22 μm filter (Merck Millipore, USA). The His 6 -tagged proteins were isolated from the filtrate using a 5 mL Ni-NTA HisTrap HP (Healthcare Life Sciences) with an imidazole gradient from 5 mM to 500 mM over 10 column volumes (CV). SDS-PAGE gel analysis was performed to assess the purity of the fractions.
Fractions containing AtzH were pooled and concentrated to 8 mg/mL for crystallography using an Amicon Ultra-15 centrifugal filter unit. Finally, size exclusion chromatography was performed using a 130 mL column packed with Superdex 76 preparation grade resin (GE Healthcare Life Sciences), equilibrated with 25 mM HEPES and 200 mM NaCl, pH 7.5, over 1.5 CV.
All chromatography steps were performed using an Ä KTA purifier UPC 10 (GE Healthcare Life Sciences).

Differential Scanning Fluorimetry (DSF)
Initially, buffer screening by DSF was performed on the wild-type protein-thirteen different combinations of salt/buffer at different pHs were tested in triplicate (Buffer screen 9 protocol) [14]. Protein was tested at 8 mg/mL in gel filtration buffer; 300 nL protein was diluted into 19.4 μL of each buffer, and 300 nL of a 1:20 dilution of sypro orange (Sigma S5692) was added. The temperature was increased from 20 to 100˚C in increments of 0.5˚/ 5 sec. This gave a T m of 55.8 +/-0.1˚C in the gel filtration HEPES buffer. A slight improvement was seen when the protein was in 50 mM Tris pH 8, 200 mM NaCl (T m of 57.1 +/-0.1˚C). The fourteen mutants were also tested for stability using DSF. Each sample was diluted to 1 mg/mL in 50 mM HEPES pH 7.5, 200 mM NaCl, and 1 μL of protein was added to 19 μL of mix of 4 mL 50 mM HEPES pH 7.5, 200 mM NaCl, 2.5 μL Sypro Dye. Each variant was run in 6-fold replicates, and compared to the wild-type protein which was included as a control (under these conditions the wild type protein had a T m of 55.0 +/-0.1˚C). A complete table of the results is shown in S2 Table. Crystallography Purified AtzH at 8 mg/mL protein was used to set up a PCT test (Pre-Crystallization Test [15]), and immediately showed microcrystal formation in the PEG based condition of the PCT. The protein was diluted to 4 mg/mL and then set up in 3 initial PEG-rich screens (shotgun, PACT and C3_3; conditions available from c6.csiro.au [16]), in SD2 sitting drop plates (Molecular Dimensions, UK) with an equal volume of crystallant (150 nL protein plus 150 nL reservoir) at 20˚C. Crystals appeared overnight in many conditions. Further PEG based screens (C3_1->C3_4, Morpheus_C3) were also set up, with the protein at 4 mg/mL with and without added thrombin. Crystals continued to appear in many PEG-based conditions for up to 3 weeks after setting up the plates. The AtzH wild-type protein and the variants were also set up in co-crystallization trials with the PPDi inhibitor. 1 μL of a 40 mM solution of the inhibitor in protein buffer (50 mM HEPES, 200 mM NaCl pH 7.5) was added to 40 μL of the 4 mg/mL protein. The variants were each set up in two screens (Shotgun, C3_3) at 4 mg/mL (if accessible) with and without added inhibitor and incubated at 20˚C. The mutants in general crystallized poorly, and only a few variants gave crystals suitable for harvesting. Crystals were cryoprotected with addition of glycerol to a final concentration of 20% and cryo-cooled in liquid nitrogen prior to data collection. Crystals used in data collection were harvested directly from the screening plates.
The native AtzH P1 crystal form crystal was harvested from a 300 nL drop consisting of a 1:1 ratio of protein and reservoir, where the reservoir consisted of 200 mM ammonium acetate, 30% PEG 4000 and 100 mM sodium acetate buffer pH 5.0. The native C222 1 crystal form was harvested from a 300 nL drop consisting of a 1:1 ratio of protein and reservoir, where the reservoir contained 20% PEG 6000, 2.5% v/v tert-butanol with 100 mM sodium citrate buffer at pH 5.5.
The R73K mutant (which crystallized in space group P2 1 ) was harvested from a 300 nL drop consisting of a 1:1 ratio of protein/inhibitor and reservoir, where the reservoir contained 0.2 M diammonium tartrate, 20% w/v PEG 3350. Data sets were obtained at the Australian Synchrotron beamline MX2 using a Dectris Eiger 16M detector and a full 360 degrees of data were taken. The data were indexed using DIALS [17][18][19][20][21], scaled using Aimless [22] and the original structure solution was determined by molecular replacement using Phaser [20,21] with PDB code 2RCD as the model, after sequence alignment with Clustal-Omega and pruning using Chainsaw [23,24] to give an appropriate starting model. The model was rebuilt manually using Coot [25] and refined using Refmac [24,26].
In silico substrate docking was done using Accelrys Discovery Studio (v4.1). The Define and Edit Binding Site program was used to identify cavities in AtzH. Models of 1,3-dicarboxyurea were prepared using the same client in doubly protonated, deprotonated and singly protonated states and docked into the cavity identified previously using the CDOCKER program with default parameters.

Enzyme assays
Reaction rates were monitored by detecting the production of ammonium in μmoles per second using the glutamate dehydrogenase GDH-coupled assay, described previously [3]. GDH catalyzes the NADH-dependent amination of alpha-ketoglutarate. Ammonia production was followed using the decrease of absorbance by UV spectrophotometry at 340 nm, due to the oxidation of NADH by GDH. 1.25 U of GDH was used in a 250 μL reaction volume, the final concentrations of alpha-ketoglutarate and NADH were 3.5 mM and 0.2 mM, respectively. The buffer was 25 mM potassium phosphate, pH 8.5. Cyanuric acid hydrolysis by AtzD was followed by UV spectrophotometry at 214 nm [13]. In all the reaction mixes, 12 μM of AtzD, 0.17 μM of AtzE, 0.17 μM AtzF and 3.3 μM of AtzH or its variants were used in presence of 1 mM of cyanuric acid.

Structural characterization of the Pseudomonas sp. strain ADP AtzH
We obtained the X-ray crystal structure of the AtzH. His 6 -tagged AtzH was produced in E. coli and purified, yielding 5 mg of AtzH per 1 L of culture. Size exclusion chromatography performed on AtzH after removal of the his-tag, suggested that the native protein had a molecular mass of~29 kDa, consistent with a dimer (2 x 14,655 kDa) (S2 Fig). Differential Scanning Fluorimetry (DSF) was used to estimate the stability of the protein and showed that the native protein had a T m of 57.2˚C. Well diffracting crystals of AtzH grew readily, and two different space groups were observed: P1 and C222 1 . X-ray diffraction data were collected for both.
The structure of AtzH was solved by molecular replacement, using the structure of a functionally uncharacterized protein from Pectobacterium atrosepticum strain SCRI 1043 (PDB accession number: 2RCD) [27] as the model. 2RCD is 56% identical to AtzH and there is a 1.0 Å rmsd between Cα atoms when superposing the structures (x-ray data are in Table 1). The structures exhibit minor differences at the N-and C-termini and in some loop regions (including the loop with weak density in AtzH, residues 66 to 73). Additionally, 2RCD has a dimer in the asymmetric unit, which mirrors the dimer seen in AtzH where two, six-stranded β-sheets sit face-on at an angle to form the interface. One long helix and two short helices sit on the outside of this interface and the overall dimer structure is very compact (Fig 2). Analysis with PISA [28] gives an interfacial surface area of almost 3,200 Å 2 (or about 29% of the total surface area of the protein).

A physiological role for AtzH
We investigated the effect of AtzH on the cyanuric acid catabolic pathway by reconstituting the complete cyanuric acid mineralization pathway in vitro using purified enzymes; i.e., AtzD, AtzEG and AtzF with and without AtzH. Inclusion of AtzH led to an increased rate of ammonia production from cyanuric acid by the pathway (Fig 3C). This is consistent with an enzymatic role for AtzH, potentially fulfilling the role of a decarboxylating 1,3-dicarboxyurea amidohydrolase. Unfortunately, both the substrate and product for the putative AtzH-mediated reaction are unstable under the conditions required for enzyme function (i.e., aqueous buffer at physiological pH) and so the reaction could not be followed directly. However, it was possible to probe the role of AtzH by varying the composition of the in vitro pathway. The AtzD substrate (cyanuric acid) is stable and its hydrolysis easily followed by measuring the reduction in absorbance at 214 nm caused by AtzD-mediated ring-opening [9]. It was therefore possible to directly test the influence of AtzH on this step of the pathway by incubating AtzD and its substrate in the presence or absence of AtzH, and AtzH did not influence the rate of this step (Fig 3A). The influence of AtzH on the AtzEG-catalyzed reaction was also investigated (Fig 3B). The substrate for AtzEG (1-carboxybiuret) is unstable and was therefore produced in situ via AtzD-mediated ring opening of cyanuric acid. One product of the AtzEG reaction is ammonia, which was used to determine the rate of the reaction through a GDHcoupled assay. In vitro cascades containing AtzD and AtzEG evolved ammonia from cyanuric acid at the same rate whether or not AtzH was included in the incubation (Fig 3B).
These observations suggest that the AtzH-dependent increase in the rate of ammonia production from the pathway was also AtzF-dependent; i.e., AtzH likely increases the rate at which AtzF produces ammonia, possibly by increasing the rate at which allophanate is fed to AtzF. To further investigate this, an in vitro pathway comprised of AtzD, AtzEG and a variant of AtzF that lacked the active-site nucleophile (AtzF Ser189Ala) was also tested. In this reaction, the presence of AtzH also failed to improve the rate of ammonia production (Fig 3D), confirming that the AtzH-mediated effect observed in the fully reconstituted pathway is dependent on the presence of active AtzF. These observations imply that AtzH is either increasing the rate of allophanate production from 1,3-dicarboxyurea, or enhancing the activity of AtzF itself. To test whether AtzH increases the catalytic activity of AtzF, an alternative in vitro pathway was tested in which biuret was incubated with the biuret amidohydrolase BiuH [2] and AtzF. BiuH produces allophanate directly, and so an increase in the rate of ammonia production in this pathway would indicate that AtzH is an allosteric activator of AtzF. However, the presence of AtzH made no measurable difference to the rate of ammonia production (Fig 3E), suggesting that AtzH does not 'activate' AtzF. The simplest explanation for these observations is that AtzH mediates the decarboxylation of 1,3-dicarboxyurea, the product from AtzEG, to form allophanate, the substrate for AtzF.

Defining the active site of AtzH
The AtzH active site was sought by examining the X-ray structure. As 1,3-dicarboxyurea is unstable, it was not possible to produce crystals of AtzH in its presence. However, an in silico docking simulation between the X-ray crystal structure of the AtzH dimer and 1,3-dicarboxyurea elucidated a potential active site (Fig 4A and 4B). The putative substrate bound with good complementarity to an arginine-rich pocket comprised of Tyr22, Arg46, Arg63, Arg66, Arg73, Phe94, Arg96, Gln106 and Gln108 (Fig 4A and 4B). The arginine residues form an extensive hydrogen bonding network with the carbonyl oxygens of 1,3-dicarboxyurea. Tyr22 is positioned within 3 Å of the carbonyl carbon of one of the two terminal carboxylic acids, and Arg46 is within hydrogen-bonding distance of the hydroxyl group of Tyr22. Tyr22 and Arg46 are therefore well positioned to form a catalytic dyad, albeit one of atypical composition.
Guided by the results of the docking experiment, fourteen variants of AtzH were produced: Tyr22Ala, Tyr22Phe, Arg46Ala, Arg46Lys, Arg63Ala, Arg63Lys, Arg66Ala, Arg66Lys, Arg73Ala, Arg73Lys, Arg96Ala, Arg96Lys, Gln106Ala, Gln108Ala (S1 Table). All of the variants were correctly folded, as assessed by DSF (S2 Table). Furthermore, X-ray structures of the R73K variant (PDB: 6D63) was also obtained and closely resembled the wild-type enzyme (PDB: 6BJT, 6BJU) except for the intended amino acid substitution. The AtzH variants were added to in vitro pathways comprised of AtzD, AtzEG and AtzF and incubated with cyanuric acid. Five of the variants (Tyr22Ala, Tyr22Phe, Arg46A, Arg46Lys and Arg63Lys) were found to enhance the rate of ammonia production, albeit to a significantly lower extent than the wild-type AtzH (Fig 5). The loss of function of Arg63Lys, but not of Arg63Ala, may suggest that Arg63 does not have a catalytic role. However, AtzH Tyr22Ala/Phe was indistinguishable from the AtzH-free cascade (Fig 5). The Arg46Ala and Arg46Lys variants both increased the rate of ammonia production by the cascade, but to a lesser degree than that of the wild-type AtzH (Fig 5). This may suggest that Tyr22 is the catalytic amino acid residue in AtzH, possibly acting as a proton donor (consistent with the catalytic role of tyrosine in other enzymes) [41], with Arg46 fulfilling a non-critical function As the substrate analog binds between Arg46 and Tyr22, it seems unlikely that Arg46 has a role in activating Tyr22 (noting that the substrate analog may not adopt exactly the same conformation as the bound substrate). However, this positioning of the active site residues relative to the bound analog is similar to that of the catalytic residues in inverting β-glycosidases that use an acid-base catalytic mechanism [42], and it is plausible that that a similar mechanism may be used here. Further work is needed to test this hypothesis.
The wild-type and variants were co-crystallized with 1,3-dicarboxyacetone, a stable structural analogue of 1,3-dicarboxyurea. In the AtzH Arg73Lys structure (PDB 6D63) 1,3-dicarboxyacetone was found to occupy the predicted active site of AtzH in a subset of the protomers (25%), indicating that this pocket is accessible by 1,3-dicarboxyurea-like small molecules (Fig 4C and 4D). The hydrogen bonding network is largely as predicted for 1,3-dicarboxyurea, and Tyr22 is well positioned as the catalytic residue. Amino acids residues involved in binding 1,3-dicarboxyurea are shown as green sticks; C-D) X-ray structure of AtzH Arg73Lys co-crystalized with the substrate analog 1,3 dicarboxyacetone (pink). The distance between Tyr22 and the carbon of the carboxyl group is <3 Å; E) Polder map (3 sigma density contour) of 1,3 dicarboxyacetone bound to the AtzH Arg73Lys variant.

AtzH homologs in other bacterial genomes
AtzH is the first DUF3225 family protein that has been shown to possess a catabolic function, likely decarboxylating the product of AtzEG-mediated deamination of 1-carboxybiuret as part of the cyanuric acid mineralization pathway. Homologs of the other members of the pathway are found in different metabolic contexts: the AtzD homolog barbiturase participates in pyrimidine catabolism [43], the AtzEG homologs GatCA are involved with acyl-tRNA metabolism [44] and the AtzF homolog allophanate hydrolase is used in urea catabolism [45]. We were therefore interested to investigate whether AtzH homologs may also participate in alternative metabolic contexts.
A thorough search of microbial genomes in the GenBank database revealed more than thirty examples of genomes containing homologs of atzG, atzE, and atzH organized as a single gene cluster with the same organization as that of the Pseudomonas sp. strain ADP cyanuric acid catabolism operon (Fig 6). Perhaps surprisingly, the atzG-atzE-atzH clusters were rarely associated with atzD or atzF homologs.
The most common arrangement, found in species of Enterobacteriaceae (Pectobacterium, Serratia, Enterobacter, Klebsiella, Kosakonia, Pantoea, Yersinia and Erwinia), has the atzG-atzE-atzH cluster immediately downstream of a gene predicted to encode γ-glutamyl transferase (ggt, E.C. 2.3.2.2). In these genomes, the ggt-atzG-atzE-atzH cluster is found downstream and on the opposite strand from a gene cluster encoding a LysR-family regulator, an ABC transporter, a transaminase and an allantoate amidohydrolase. The overall genetic organization of these clusters is similar to that of the cyanuric acid catabolism operon of Pseudomonas sp. strain ADP, albeit lacking atzF and with atzD replaced by ggt. (Fig 6B, S3 Fig) γ-Glutamyl transferase, which also acts as glutathione hydrolase (E.C. 3.4.19.13), is responsible for the amide hydrolysis of glutathione to produce L-glutamate and L-cysteinylglycine (Fig 7). Further, the substrate for allantoate hydrolase, an enzyme found encoded in close proximity to the Enterobacteriaceae atzG-atzE-atzH clusters described here, contains two   Structure and function of AtzH ureido groups (Fig 7) [46]. It is possible that atzG-atzE-atzH clusters form components of catabolic pathways in which the substrates possess amide groups.
There are also gene clusters found in β-proteobacter from the orders Comamonadacaea (Rhodoferax and Variovorax) and Burkholderiales (Rhizobacter, Leptothrix, Methylibium and Rhizobiales) that contain an allophanate hydrolase encoding gene. In all cases, the allophanate hydrolase gene is upstream of the atzG-atzE-atzH cluster and encoded on the same strand; in about half of the genomes there are genes encoding an ABC transporter separating the allophanate gene from the atzG-atzE-atzH cluster. Downstream, and encoded on the same strand as these atzG-atzE-atzH clusters, are genes encoding ABC transporters and amidase enzymes. This may indicate that these clusters are for the uptake and catabolism of nitrogenous compounds. (Fig 6C, S3 Fig) Other Burkholderiales (Bordetella and Achromobacter) possess atzG-atzE and atzH clusters, but where the cluster also includes genes of unknown function that are annotated by homology as a transcriptional regulator, a monooxygenase and an FMN-reductase. In these clusters atzH is located between the genes encoding the monooxygenase and reductase, rather than immediately downstream of atzG and atzE. Finally, a number of diverse α-, β-, and γ-proteobacteria (Pseudorhodoplanes, Methylobacterium, Thiomonas, Hydrogenophaga, Comamomonas, Pseudomonas, and Halothiobacillus) have atzG-atzE-atzH clusters in unique genetic contexts, including the cyanuric acid catabolic operon of Pseudomonas sp. strain ADP. (Fig 6D, S3 Fig).
The occurrence of the atzG-atzE-atzH cluster appears to be far more wide-spread than the cyanuric acid catabolic pathway, and it is frequently associated with genes that may be involved in the catabolism of diverse nitrogen-containing compounds. However, the genetic contexts in which atzH-like genes are found provides little further evidence to suggest the specific function of the AtzH protein.

Conclusion
The data presented here imply that AtzH is a decarboxylating, 1,3-dicarboxyurea amidohydrolase. As yet its catalytic mechanism is not known, but our results suggest it involves the action of Tyr22 and Arg46, and that these amino acid residues are positioned relative to the substrate in a manner consistent with an acid-base mechanism. AtzH also facilitates the only previously uncontrolled step in the catabolic pathway, and as such, we believe that we have now identified the last remaining enzyme in the Pseudomonas sp. strain ADP cyanuric acid catabolism pathway ( Fig 1B). The pathway appears to be dependent on enzymatic steps that outcompete solvent mediated hydrolysis that would otherwise lead to the accumulation of dead-end metabolites (biuret and urea). With no endogenous biuret or urea hydrolases, formation of these metabolites effectively prevent access to two of the three nitrogen atoms of the triazine ring.
Unexpectedly, atzG-atzE-atzH -like clusters appear to be relatively common within the Proteobacteria. Given the amidase function of AtzE and proposed amidase activity of AtzH, it is perhaps unsurprising that the atzG-atzE-atzH-like clusters tend to be associated with genes predicted to be involved in nitrogen compound metabolism. Previously, we had demonstrated that atzG and atzE were likely to have co-evolved from components of the bacterial GatCAB (transamidosome) complex. Here, we report the discovery of atzG-atzE-atzH-like clusters associated with diverse genetic contexts and metabolic functions, implying that the formation of the atzG-atzE-atzH cluster may predate the evolution of the cyanuric acid catabolism operon. Bacteria containing an allantoate hydrolase and glutamyl-transferase enzymes upstream of the atzG-atzE-atzH segment; B. Bacteria found to be containing an allophanate hydrolase homologue of AtzF found upstream of the atzG-atzE-atzH segment; C. Bacteria containing a nonadjacent AtzH homologue, located between a monooxygenase and reductase enzymes, downstream of the atzG-atzE segment; D. Bacteria presenting the atzG-atzE-atzH cluster within unique gene organizations. A legend is shown at the top of the figure. Green arrows represents amidase type enzymes, black and grey tones represent ABC transporters, discontinued dark blue arrows represent the presence of a LysR type regulator, genes encoding for AtzG, AtzE and AtzH homologues are shown by orange, yellow and pink arrows and light blue arrows represent other types of enzymes, respectively. This figure is not to scale. (DOCX) S1