The tandem ureohydrolase genes encode a metformin hydrolase
We heterologously expressed a part of the operon associated with metformin degradation containing both ureohydrolase family genes and the nickel chaperones hypA and hypB but without the nickel transporter hupE of P. mendocina MET, A. niigataensis MD1 and A. niigataensis DSM7050 in E. coli (Fig. 1A). For purification, the first of the ureohydrolase family proteins contained an N-terminal 6x His- or Strep-tag (for the different expression constructs used for the enzyme production see Supplementary Fig. 1). Strikingly, two bands were observed on denaturing polyacrylamide gels after affinity purification, matching the predicted sizes of the first and second ureohydrolase family proteins (Fig. 1C; Supplementary Fig. 2). Peptide mass fingerprint analysis of the proteins confirmed that the untagged second ureohydrolase family proteins were co-purified together with the tagged proteins (Extended Data Table 1). To avoid co-purification of naturally His-rich HypB and interference between the Ni-NTA matrix used for purification and the putatively nickel-dependent ureohydrolases, we mostly used the Strep-tagged versions of the first ureohydrolase for the enzymatic characterization. Key experiments were also performed with the 6x His-tagged versions and yielded very similar results.
Incubation of metformin with the purified ureohydrolases from P. mendocina MET and A. niigataensis MD1 yielded guanylurea and dimethylamine (Fig. 1B), which were detected by LC-MS. We tested the specificity of the enzymes by incubation with several other guanidine compounds as potential substrates (Extended Data Table 2). Of all tested compounds only metformin and dimethylguanidine were hydrolyzed. For all other compounds, including highly similar methylguanidine, neither consumption of the substrate nor product formation were detected. The hydrolysis of dimethylguanidine produced dimethylamine and urea. Since urea is not ionized during LC-MS analysis, it was detected by a colorimetric assay 26 that was also used in the quantitative analyses of enzyme activity. We developed a quantitative method to detect guanylurea by LC-MS (Supplementary Fig. 3, see Methods section for details) and determined the specific metformin and dimethylguanidine hydrolysis activities of the three different enzymes (Fig.1D). The enzymes of P. mendocina MET and A. niigataensis MD1 catalyzed the hydrolysis of both metformin and dimethylguanidine. However, the specific activity for dimethylguanidine hydrolysis was approximately 20-fold lower (Fig. 1D). It has been noted before that the type strain A. niigataensis DSM 7050 was not able to grow on metformin as the sole carbon source 22 despite possessing a highly similar operon (~ 93% identical amino acids for both the first and second ureohydrolase family proteins (Supplementary Fig. 4). Indeed, the enzyme of A. niigataensis DSM 7050 only hydrolyzed dimethylguanidine but not metformin (Fig. 1D). In the following, we refer to the gene products from the metformin-degrading strains as metformin hydrolase α and β (MefHα and MefHβ) and the proteins from A. niigataensis DSM 7050 as dimethylguanidine hydrolase α and β (DmgHα and DmgHβ). When PmMefHα or PmMefHβ were expressed as individual proteins in combination with the nickel chaperones, the purified proteins did not exhibit metformin hydrolase activity. Therefore, we concluded that only heteromeric complexes of both subunits are catalytically active. MefH and DmgH are used throughout the manuscript to refer to the enzymatically active heteromeric complexes.
Enzyme characterization
The presence of the nickel chaperones hypA and hypB in all three operons strongly suggested that the enzymes are dependent on nickel. PmMefH was expressed in the presence of Ni2+, Mn2+, Zn2+ and Co2+ that have been reported as metal co-factors for ureohydrolases. Subsequently, the enzyme was purified (Supplementary Fig. 5) and the specific activity with 2 mM metformin was determined. The enzyme with the highest activity was obtained when expression occurred in the presence of Ni2+ with only residual activity for the other metal ions. Consistent results were obtained for DmgH and dimethylguanidine as substrate (Supplementary Fig. 6). Therefore, we conclude that both MefH and DmgH are indeed nickel-dependent enzymes. Active preparations of DmgH expressed in the presence of Ni2+ were subjected to ICP-OES analysis to determine the metal content. The nickel content was 12 µmol (mg protein)-1 and additionally 2 µmol (mg protein)-1 manganese were detected, but no other transition metals (Supplementary Fig. 7).
After establishing that the enzymes are dependent on nickel, the kinetic parameters were determined. Enzymes were expressed in the presence of Ni2+ and purified by Strep-Tactin affinity chromatography. Purified enzymes were incubated with 0.16-100 mM of their preferred substrate (Fig. 1E). KM values for the enzymes were very similar with 61.9 ± 4.4 mM, 57.8 ± 4.3 mM and 57.1 ± 4.3 mM for PmMefH, AnMefH and DmgH, respectively. Based on the observation of two catalytic sites per hexameric holoenzyme (see below), we calculated the turnover numbers as kcat = 0.9 s-1 for DmgH and 0.8 s-1 for PmMefH but only 0.4 s-1 for AnMefH. Therefore, hydrolysis by DmgH and PmMefH was more efficient than by AnMefH. All enzymes were highly resistant to thermal inactivation. DmgH exhibited a slightly higher apparent temperature optimum than AnMefH and PmMefH around 56°C compared to 50°C, respectively (Fig. 1F)
Crystal Structure of MefH
To determine the stoichiometry and architecture of the MefH heteromer, we resolved its three-dimensional atomic structure using X-ray crystallography. For crystallization, we used AnMefH as it shares a higher sequence identity with DmgH than PmMefH, so that residues responsible for the change in substrate specificity are easier to identify. In the obtained structure, both MefHa and MefHb adopt an almost identical three-layer alpha-beta-alpha fold (Fig. 2A,B) typical of members of the ureohydrolase family 27, but diverge notably at the N-terminal loop regions that are unique to each subunit type (Fig 2C and Supplementary Fig. 4). The global structure of MefH is distinct amongst ureohydrolase family proteins since it is a heterohexamer consisting of a dimer of heterotrimers, each with αβ2 stoichiometry (Fig. 2). The trimers interact with each other generating a two-fold symmetry axis along the enzyme. The MefHα subunits are located in apical position and make a close contact between both trimers. In contrast, the MefHβ subunits interact weakly with their counterparts in the opposite trimer. They contain flexible and partially disordered loops at the interface that result in the presence of a cleft between the trimers (Fig. 2B). Each MefHα subunit coordinates two Ni2+ ions in its active site through four Asp (D183, D187, D276 and D278) and two His (H158 and H185) residues. In the crystal structure, the Ni2+ ions coordinate a molecule of urea (Supplementary Fig. 8). In contrast, the MefHβ subunits do not contain any metal ions and four of the six residues that bind Ni2+ in the MefHα subunits are replaced by amino acids that are not prone to coordinate cations (Fig. 2D). Another specific feature of the MefHα subunits is the binding of one molecule per subunit in the N-terminal loop segment that was interpreted as biuret (Supplementary Fig. 9). The electron density could also satisfactorily be interpreted as guanylurea. However, biuret is a known contaminant of urea that was used for soaking. The biuret binding site is distant from the di-metal center marking the active site and it is unknown whether it might hold functional significance.
Investigation of substrate specificity determinants
To investigate the similarity between the different enzyme systems and subunits, we constructed a multiple sequence alignment of all six subunits of both MefH and DmgH enzymes (Supplementary Fig. 7). AnMefHα and DmgHα share 93.5% sequence identity, whereas AnMefHα and AnMefHβ share 33% sequence identity. MefHβ and DmgHβ exhibit 93% sequence identity. Thus, the protein subunits within one heteromeric enzyme are more dissimilar than the corresponding subunits of the two distinct enzymes. AnMefHα and PmMefHα share a sequence identity of 97.5% and their β-subunits are 100% conserved. As observed in the crystal structure, the metal coordination site is degenerated in MefHβ and DmgHβ. A phylogenetic analysis including further ureohydrolase sequences with >24% sequence identity to MefHα confirmed that DmgH is likely the direct evolutionary precursor of MefH (Supplementary Fig. 10). The α and β subunits of homologous proteins form separate branches in the tree and are clearly separated from other ureohydrolase family proteins that are encoded by single genes. A second tandem arrangement of ureohydrolase genes in putative Limnocylindrales bacteria seems to have evolved independently.
In order to investigate the contribution of the subunits to the substrate specificities of MefH and DmgH, we combined MefHα with DmgHβ and DmgHα with MefHβ. Both combinations yielded stable heteromers and DmgHα/MefHβ exclusively hydrolyzed dimethylguanidine, whereas MefHα/DmgHβ preferentially hydrolyzed metformin (Fig. 3B and Supplementary Fig. 11). However, the specific activity of DmgHα/MefHβ was approximately 5 times lower than for native DmgH, whereas MefHα/DmgHβ was less affected in comparison to MefH. These results clearly show that the catalytic activity and the substrate selectivity is mainly mediated by the MefHα and DmgHα subunits. The degenerated subunits MefHβ and DmgHβ were crucial for activity, but were exchangeable and had only a minor effect on substrate specificity.
As the proteins are very similar with only 23 amino acid residues distinguishing MefHα from DmgHα, we decided to investigate variants of DmgHα for changed substrate specificity. Especially, a stretch of three amino acids in close vicinity to the two Ni2+ ions in the active site aroused our interest. Instead of Ser288-Thr289-Ser290 in DmgHα, these positions are exchanged for Asn288-Ser289-Ala290 in AnMefHα and PmMefHα (Fig. 3A and Supplementary Fig. 4). We mutated all three single positions and the three amino acid stretch in the DmgH expression construct and determined the specific activities of the resulting protein variants (Fig. 3B and Supplementary Fig. 11). Remarkably, the Thr289Ser exchange resulted in higher dimethylguanidine hydrolase activity. All other variants exhibited a loss of DmgH activity without gaining metformin hydrolase activity. For variant Thr289Ser we determined a lower KM of 28 ± 2 mM and a slightly higher kcat of 1 s-1 compared to the wild-type DmgH (Supplementary Fig. 12), reflecting the improved catalytic performance. However, no metformin hydrolase activity was observed for these variants. Thus, mutation Ser184Cys that is close to the active site was introduced in the DmgHα variant with the whole stretch mutated (Fig. 3A and Supplementary Fig. 11). The dimethylguanidine hydrolase activity remained low, but comparable metformin hydrolase activity was observed (Fig. 3B). Thus, these four mutations could be the tipping point for the evolution of MefH from DmgH.
Growth analysis of A. niigataensis DSM7050
In contrast to A. niigataensis MD1, A. niigataensis DSM7050 was not able to grow on metformin as the sole nitrogen, carbon, and energy source (Fig. 4)22. Our analyses of the recombinant proteins established that the tandem ureohydrolase genes of A. niigataensis DSM7050 encode DmgH and thus allow the hydrolysis of dimethylguanidine rather than metformin. To investigate if A. niigataensis DSM7050 could assimilate dimethylguanidine, the bacteria were incubated in minimal medium containing either 10 mM dimethylguanidine or 10 mM metformin as the only nitrogen, carbon and energy source. As control, the minimal medium was supplemented with 10 mM ammonium chloride and 0.5% w/v glycerol. Cultures were incubated at 30°C and OD600 was recorded to assess growth. A. niigataensis DSM7050 grew on dimethylguanidine with a doubling time (tD) of 12.9 h, corresponding to approximately half the growth rate of the control culture with ammonium and glycerol (tD= 6.4 h).