Dual activity of PNGM-1 pinpoints the evolutionary origin of subclass B3 metallo-β-lactamases: a molecular and evolutionary study

ABSTRACT Resistance to β-lactams is one of the most serious problems associated with Gram-negative infections. β-Lactamases are able to hydrolyze β-lactams such as cephalosporins and/or carbapenems. Evolutionary origin of metallo-β-lactamases (MBLs), conferring critical antibiotic resistance threats, remains unknown. We discovered PNGM-1, the novel subclass B3 MBL, in deep-sea sediments that predate the antibiotic era. Here, our phylogenetic analysis suggests that PNGM-1 yields insights into the evolutionary origin of subclass B3 MBLs. We reveal the structural similarities between tRNase Zs and PNGM-1, and demonstrate that PNGM-1 has both MBL and tRNase Z activities, suggesting that PNGM-1 is thought to have evolved from a tRNase Z. We also show kinetic and structural comparisons between PNGM-1 and other proteins including subclass B3 MBLs and tRNase Zs. These comparisons revealed that the B3 MBL activity of PNGM-1 is a promiscuous activity and subclass B3 MBLs are thought to have evolved through PNGM-1 activity.


Introduction
Antibiotic resistance is a steadily increasing global problem which could lead to a fundamental upheaval in clinical care with the potential to return us to the pre-antibiotic era [1][2][3]. The production of β-lactamases, a group of enzymes that confer β-lactam resistance in Gram-negative bacteria, is now one of the major barriers in treating Gram-negative infections [4][5][6]. β-Lactamases are classified according to their catalytic mechanisms into serine β-lactamases (classes A, C, D) and metallo-β-lactamases (MBLs, class B enzymes further divided into three subclasses B1-B3) [7,8]. There are functional and structural similarities between serine β-lactamases and penicillin-binding proteins (PBPs), and so serine β-lactamases are thought to have evolved from a PBP [8,9]. Due to the functional and structural differences between MBLs and serine β-lactamases, the MBLs, hydrolyzing almost all β-lactam antibiotics and thereby representing a critical antibiotic resistant threat, are thought to have evolved from a protein other than a PBP [9][10][11][12].
A member of the MBL superfamily, tRNase Z, is a tRNA processing enzyme which removes the 3 ′ trailer from pre-tRNA [21]. Most tRNase Zs cleave pre-tRNA immediately downstream of a discriminator nucleotide (nt), onto which the CCA residues are added to produce mature tRNA. tRNase Z can cleave unstructured single-strand RNAs that are unrelated to pre-tRNA in vitro [22]. The tRNase Z catalytic centre is formed by the major five well conserved residues [His-48, His-50, Asp-52, His-53, and His-222 in the case of T. maritima tRNase Z (prokaryotic tRNase Z S )] together with two zinc ions [23]. Residues Asp-52 and His-222 are thought to directly contribute as donors during the catalytic proton transfer [23].
To date, there is a study [15] or hypothesis [24] on the evolution of MBLs. The structure-based phylogeny [15] showed that MBLs and rubridoxin oxidoreductase (an MBL fold protein without β-lactamase activity, a member of MBL superfamily) descended from a common ancestor. However, this ancestor and the origin of MBLs remains unknown. Due to the structural similarity between the two halves (αβ and βα domains) of MBLs, there was a proposal that MBLs might be arisen by a gene-duplication event [24]. However, there is no evidence showing the event. PNGM-1 (a subclass B3 MBL) identified in deep-sea sediments that predate the antibiotic era [13] showed unique characteristics as follows. First, PNGM-1 contains 386 amino acids (AAs), making it longer by about 40-100 AAs in length than the length (280 to 342 AAs) of amino acid sequences of subclass B3 MBLs. Second, while checking on the NCBI database using the BLASTP programme (https://blast.ncbi.nlm.nih.gov/Blast.cgi?PAGE= Proteins), we found that PNGM-1 showed high sequence identities (74 and 71%, respectively) to the MBL fold metallo-hydrolases (members of MBL superfamily; GenBank accession numbers: WP_146397269 and TDI33721, respectively) of Planctomycetes bacterium CA13 and Acidobacteria bacterium identified from the marine sediment metagenome. The annotated sequences of, and limited information about, the two MBL fold proteins were released at the GenBank database from March to August 2019. Therefore, knowledge of the enzymatic functions of these MBL fold metallo-hydrolases remains incomplete. However, PNGM-1 showed a low sequence identity (≤18%) to MBLs with β-lactamase activity. These results suggest that PNGM-1 is able to possess an enzymatic activity (other than β-lactamase) which is one of the various biological functions of the MBL superfamily, including DNA or RNA processing [16].
In this study, we found that in addition to β-lactamase activity, PNGM-1 possesses endoribonuclease (tRNase Z) activity on both pre-tRNA substrates and on small unstructured single-strand RNA substrates. Our functional, phylogenetic, and structural analyses of PNGM-1 and their comparison with the properties of other MBLs (proteins containing the αββα-fold with β-lactamase activity) and structurally representative MBL fold proteins (proteins having αββα-fold without β-lactamase activity) of the MBL superfamily, reveal the evolution of tRNase Z to subclass B3 MBLs through the dual activity of PNGM-1.

Strains and plasmids
Escherichia coli BL21 (DE3) and plasmid-containing E. coli strains were used for all cloning and expression studies. The pET-28a(+)/His 6 -PNGM-1 plasmid has been described previously [13]. The strains and plasmids used in this study are listed in Table S1.

Steady-state kinetic analysis
Kinetic assays were conducted at 30°C with a Shimadzu UV-1650PC spectrophotometer (Shimadzu Corp.). β- Steady-state kinetic constants were determined by fitting the initial rates (in triplicate) directly to the Michaelis-Menten equation using nonlinear regression with the programme DynaFit [26]. To determine whether PNGM-1 was halophilic or halotolerant to NaCl (from 0 to 2 M), enzyme activity was measured using benzylpenicillin and cephalothin as the substrate, as previously described [27].

Structure determination of PNGM-1
Crystallization and X-ray diffraction data collection of PNGM-1 was carried out as previously published [29]. All X-ray diffraction data were integrated and scaled using the DENZO and SCALEPACK crystallographic data-reduction routines [30].
The PNGM-1 structure was solved by the SAD method using SeMet PNGM-1. The interpretable electron density was obtained at 2.3 Å resolution for the SeMet PNGM-1 data set using the single wavelength SAD protocol of AUTO-RICKSHAW, an automated crystal structure determination platform [31] in the P2 1 space group. The native PNGM-1 structure was determined by molecular replacement with MOLREP [32] using the SeMet PNGM-1 structure (Table S4).

In vitro RNA cleavage assay
Various RNA substrates and a negative control DNA substrate were mixed with PNGM-1 in 50 μl of a 0.2-1 mM MES (pH 6.8) buffer or with tRNase Z in 50 μl of 4 mM Tris-HCl (pH 8.0) buffer in the presence of 5-50 mM MgCl 2 or MnCl 2 at 25-80°C for 30-90 min. The reaction products for the total RNA of the human cell line KMM-1 were analyzed by microfluidics-based automated electrophoresis with the Agilent 2100 Bioanalyzer (Agilent Technologies, CA, USA) according to the manufacturer's protocol. The reaction products for the other substrates were separated on a 20% polyacrylamide 8 M urea gel or in the case of the reaction products for the DNA substrate with MnCl 2 , a 20% polyacrylamide native gel. After resolution of the reaction products, the gel was analyzed with a Gel-Doc imager (Bio-Rad Laboratories, CA, USA). Total RNA isolated from the human cell line KMM-1 was incubated with 15 μM PNGM-1 in the presence of 10 mM Mg 2+ or Mn 2+ at 37°C for 60 min, and analyzed by microfluidics-based automated electrophoresis. Three substrates [usRNA1 (24 nt), usRNA9 (22 nt), and usDNA1 (24 nt)], which were 5 ′ -labeled with 6-carboxyfluorescein, were incubated with 15 μM PNGM-1 in the presence of 10 mM Mg 2+ at 50°C for 30-90 min, and the products were analyzed by denaturing polyacrylamide gel electrophoresis.

Data availability
Coordinates for the PNGM-1 atomic model have been deposited in the Protein Data Bank under the accession codes 6J4N (native) and 6JKW (SeMet).

β-Lactamase activity of PNGM-1 and PNGM-1 mutants
The subclass B3 MBL, PNGM-1, has recently been discovered from the metagenomic library of deep-sea sediments from the Edison seamount that predates the antibiotic era [13]. While PNGM-1 has low amino acid sequence identity with other B3 MBLs, it does possess the unique metal binding motif (H 116-XH 118 XD 120 H 121 ; numbering according to the BBL scheme [8]) that is only present in subclass B3 MBLs. To test the β-lactamase (MBL) activity of PNGM-1, four active-site mutants, two at the first zinc-binding site [H91A (H116A, numbering according to the BBL scheme [8]) and H93A (H118A)] and two at the second zinc-binding site [D95A (D120A) and H96A (H121A)], were generated by site-directed mutagenesis (Tables S1 and S2).
Most of the enzymes from deep sea are halophilic or halotolerant [34]. PNGM-1 exhibited the highest activity in the presence of 50 mM NaCl (Table S5). Moreover, the activity was maintained with up to 2 M NaCl (Table S5). Therefore, PNGM-1 is halotolerant (salt-resistant) to NaCl.

The insight into the evolutionary origin of subclass B3 MBLs
The evolutionary origin of subclass B3 MBLs has not yet been identified. However, MBLs belong to an ancient MBL superfamily [15,16] that evolved billions years ago [35] and to date, there are more than 80 different types of MBLs in the Beta-Lactamase Data-Base (http://bldb.eu/). A phylogenetic analysis was carried out with all MBL types, PNGM-1, 14 structurally representative MBL superfamily members [MBL fold proteins having no β-lactamase activity (Table S3) such as Bs-tRNase Z, Tm-tRNase Z, and Ec-tRNase Z [18]], and two MBL fold metallo-hydrolases (Pl and Ac with no functional analysis). Our phylogenetic analysis revealed that (i) PNGM-1 was not grouped with subclass B3 MBLs but grouped with MBL fold proteins (Figure 1(A)); (ii) PNGM-1 was closest to the tRNase Zs of the MBL fold proteins (Figure 1 (A)); and (iii) the HXHXDH motif was completely conserved in PNGM-1, the subclass B3 MBLs (AIM-1, GOB-18, and FEZ-1), and MBL fold proteins, but not in subclasses B1/B2 MBLs (Figure 1(B)). These results suggest that PNGM-1 gives us an insight into the evolutionary origin of subclass B3 MBLs.

Structural similarity between PNGM-1 and tRNase Z
To examine the evolutionary relationship between PNGM-1 and tRNase Z, we determined the PNGM-1 structure and compared the structure with those of structurally representative tRNase Zs from B. subtilis, E. coli, and T. maritima (Figure 2). All the structures have the characteristic αββα-fold of the MBL superfamily with the unique metal binding motif. The core of the PNGM-1 monomer is well superimposed on the tRNase Z monomers of Bs-tRNase Z, Ec-tRNase Z, and Tm-tRNase Z using Coot programme (http:// bernhardcl.github.io/coot/), with the root-meansquare-deviation of 2.61 Å in the Cα atom of 178 aligned residues, 2.46 Å in 175 residues, and 2.05 Å in 202 residues, respectively. In addition, comparison of the PNGM-1 structure with all structures in the Protein Data Bank, using DALI server (http:// ekhidna2.biocenter.helsinki.fi/dali/), revealed that the highest structural similarity is for Bs-tRNase Z (Dali Z-score: 29.  (Table S6).
Structures of all tRNase Z enzymes showing dimerization [19] is essential for tRNA binding and tRNase Z activity; tRNA is mostly bound to subunit A with the 3' end bound to the active site of subunit B ( Figure  2). PNGM-1 structure also maintains the dimerization interactions and the dimer interface of PNGM-1 is strictly conserved to that of tRNase Z ( Figure S1). Specifically, the dimer interface consists of three helices and a loop: three helices of residues 43-49, (4.7 ± 0.2) × 10 2 ND ND ND ND ND a Kinetic data for PNGM-1 were identical to data from reference [13]. b His-257 of PNGM-1 is located at the position corresponding to His-222 of T. maritima tRNase Z. c ND, not detectable. Data are mean ±s.d. of three assays. The closed triangle shows the difference between subclass B2 (CphA and SFH-1) and others, with the differing residues highlighted in blue, while the closed square shows the difference between B1 and B2 MBLs (NDM-1, VIM-1, IMP-1, CphA, and SFH-1) and others, with the differing residues highlighted in green. The closed circles highlight residues which are conserved in all MBLs and MBL fold proteins. Accession numbers of all enzymes used for phylogenetic analysis are listed in Table S3.
70-82, and 95-106; and a loop of residues 9-18 in Bs-tRNase Z and three helices of residues 71-77, 98-110, and 132-143; and a loop of residues 39-48 in PNGM-1 ( Figure S1). When the tRNA-bound Bs-tRNase Z dimer was superimposed on the PNGM-1 dimer, tRNA can be bound to PNGM-1 subunit A without major hindrance and the 3' end (to be processed) of tRNA is well positioned in the active site of subunit B (Figure 2). The structurally representative subclass B3 MBLs, AIM-1, GOB-18, and FEZ-1, exist as monomers due to the presence of bulge-like structures at the proposed dimer interface which prevents dimerization due to steric hindrance ( Figure 2). tRNA is a bulky substrate similar to tRNase Z enzyme monomer. However, the dimerization of subclass B3 MBLs is not necessary for the binding of much smaller β-lactam substrates. Therefore, subclass B3 MBLs are unable to interact with tRNA except the 3' end of tRNA to be processed ( Figure 2). tRNase Z has a conserved tRNA-binding arm, which is lost in PNGM-1 and subclass B3 MBLs (Figure 2). This analysis of the superimposed structures of these six enzymes with PNGM-1 showed the gradual change from tRNase Z to subclass B3 MBL (Figure 2). These structural similarities between tRNase Zs and PNGM-1 prompts us to investigate the possibility of PNGM-1 possessing tRNase Z activity.

PNGM-1 possesses RNase activity
To test whether PNGM-1 possesses RNase activity, we examined whether PNGM-1 can degrade total RNA isolated from the human cell line KMM-1. Total RNA was incubated with 15 μM PNGM-1 in the presence of 10 mM Mg 2+ or Mn 2+ at 37°C for 60 min, and analyzed by microfluidics-based automated electrophoresis. Both 18S and 28S rRNAs were degraded to smaller RNAs, and RNA species (∼100 nt) accumulated under both conditions (10 mM Mg 2+ and Mn 2+ ) (Figure 3). The RNA integrity numbers in the assays with Mg 2+ and Mn 2+ decreased by 25% and 18%, respectively. These results suggest that PNGM-1 possesses a ribonuclease activity.

PNGM-1 can cleave unstructured RNAs endoribonucleolytically
Next, we examined PNGM-1 for RNase activity on two small RNA substrates. The unstructured RNAs, usRNA1 (24 nt) and usRNA9 (22 nt), were used as they are cleaved by tRNase Z. An unstructured 24-nt DNA (usDNA1), corresponding to usRNA1 was used as a negative control substrate. These three substrates, which were 5 ′ -labeled with 6-carboxyfluorescein, were incubated with 15 μM PNGM-1 in the presence of 10 mM Mg 2+ at 50°C for 30-90 min, and the products were analyzed by denaturing polyacrylamide gel electrophoresis. Fragments of various sizes were generated from usRNA1 and usRNA9 but not from usDNA1 by PNGM-1 (Figure 4(A)). The pattern of generated RNA fragments suggests that PNGM-1 possesses an endoribonuclease activity, not an exoribonuclease activity, and has no DNase activity. The amounts of usRNA1 and usRNA9 cleavage products increased in a dose-dependent manner, with the optimal temperature for RNase activity estimated to be around 50°C  ( Figure 4(B) and (C)). RNase activity was also assessed by varying the concentration (5-50 mM) of MgCl 2 or MnCl 2 , and the optimal concentrations of Mg 2+ and Mn 2+ were estimated to be around 10 mM and 5-10 mM, respectively (Figure 4(D)). It should be noted that PNGM-1 shows no RNase activity without Mg 2+ or Mn 2+ ions. This metal ion requirement for the RNase activity would be related to stabilizing the RNA/PNGM-1 interaction, the chemical step of cleavage, or both of these.
Asp-95 and His-257 are essential for the RNase activity of PNGM-1 To rule out the possibility that the observed RNase activity of PNGM-1 originates from unidentified contaminant RNases, we examined the five PNGM-1 mutants. These mutants contained single amino-acid substitutions of residues that were likely essential for RNase activity on usRNA1 and usRNA9. The PNGM-1 mutants (H91A, H93A, D95A, H96A, and H257A: five mutants with alanine replacements of His-48, His- 50, Asp-52, His-53, and His-222 at structurally equivalent positions in Tm-tRNase Z, respectively) had a single substitution of alanine for histidine or aspartic acid. The corresponding amino-acid substitutions in Tm-tRNase Z are known to abolish its activity without Mn 2+ ions [23]. All five PNGM-1 mutants showed little or no RNase activity in the presence of Mg 2+ (Figure 5(A)). These results indicate that the observed RNase activity of PNGM-1 is genuine.
RNase activity was recovered, albeit inefficiently, in the presence of Mn 2+ for mutants H91A, H93A, and H96A but not D95A and H257A (Figure 5(B)). This Mn 2+ -rescue phenomenon, which was first observed for the pre-tRNA cleavage reaction by Tm-tRNase Z [23], suggests that Asp-95 (Asp-52 in Tm-tRNase Z) and His-257 (His-222 in Tm-tRNase Z) are essential and directly contribute to proton transfer for the RNase activity of PNGM-1, further corroborating the RNase activity of PNGM-1. It has been suggested that the Mn 2+ -rescue is primarily due to the property that Mn 2+ ions have higher affinity for either the active site or a nearby site, or both, than Mg 2+ ions and that Mn 2+ ions are able to be coordinated even in either the catalytic site or the nearby site, or both, even when they are distorted due to the lack of one of the important amino acids [23]. Interestingly, the turnover number (k cat ) of H257A for all tested β-lactams decreased in comparison to that of wild-type PNGM-1 (Table 1), which suggests that H257 (as well as H91A, H93A, D95A, and H96A) of PNGM-1 plays an important role in both its RNase and β-lactamase activities.
Cleavage site preference of the endoribonucleolytic activity of PNGM-1 We examined the cleavage site preference of the endoribonucleolytic activity of PNGM-1 on usRNA1 and usRNA9. The cleavage reaction products for usRNA1 and usRNA9 were analyzed at nt resolutions, with corresponding substrate ladders, on a 20% polyacrylamide denaturing gel. The cleavage patterns of usRNA1 and usRNA9 by PNGM-1 were unique and different from those for Tm-tRNase Z (Figure 6(A) and (B)). PNGM-1 appears to have a tendency to cleave the RNA substrates between pyrimidine nucleotides.

PNGM-1 can process a pre-tRNA substrate
To test whether PNGM-1 can process pre-tRNA similar to tRNase Z, the 84-nt human pre-tRNA Arg was used as a pre-tRNA substrate ( Figure 6(C)). The pre-tRNA Arg was incubated with PNGM-1, Tm-tRNase Z or human Δ30 tRNase Z L , with cleavage products analyzed by denaturing polyacrylamide gel electrophoresis. PNGM-1 cleaved the pre-tRNA Arg similar to Tm-tRNase Z and human Δ30 tRNase Z L (Figure 6 (D)). Major cleavage by PNGM-1 occurred after the 76th nt (3 nt downstream of the discriminator), whereas major cleavage by Tm-tRNase Z and human Δ30 tRNase Z L occurred after the 75th nt and the discriminator (73rd nt), respectively, as expected [21,23]. Therefore, PNGM-1 is the first unique enzyme that has dual activity i.e. β-lactamase (MBL hydrolyzing all tested and clinically-used β-lactam antibiotics; Table 1) and tRNase Z activity.

Discussion
MBLs hydrolyze almost all β-lactam antibiotics including carbapenems, thereby representing a critical antibiotic resistant threat [11,12]. The evolutionary origin of subclass B3 MBLs is currently unknown. Recently, we discovered a novel B3 MBL named PNGM-1, which was the first subclass B3 MBL enzyme obtained from a functional and bacterial metagenomic library of deep-sea sediments from the Edison seamount, which existed prior to the antibiotic era. Our phylogenetic analysis revealed that PNGM-1 was grouped with MBL fold proteins and was closest to the tRNase Zs of the MBL fold proteins with known enzymatic functions, which suggest that PNGM-1 gives us an insight into the evolutionary origin of subclass B3 MBLs. Although enzymatic functions of two MBL fold metallo-hydrolases (Pl and Ac) remain unknown, they showed higher sequence identity with PNGM-1 than tRNase Zs. These results suggest that Pl (or Ac) seems to have the characteristics of PNGM-1.
Our functional analyses of PNGM-1 showed that PNGM-1 has β-lactamase (MBL) and tRNase Z activity (Table 1 and Figures 3-7). We demonstrate the dual activity of PNGM-1, which strongly suggests that PNGM-1 has evolved from a tRNase Z [MBL fold protein with a single native activity and without β-lactamase activity (Table S7)]. This is in agreement with earlier reports suggesting that novel enzymes could evolve from an existing enzyme with single native activity through the recruitment of multiple functions (e.g. dual activity) [18,36]. It is also consistent with previous reports showing that (i) MBLs descended from a common ancestor (MBL fold protein) of the MBL superfamily [15] and (ii) the last common ancestor of the MBL superfamily is a phosphodiesterase, such as Ec-tRNase Z (PDB code: 2CBN), involved in tRNA processing [18,37].
All PNGM-1 molecules, eight monomers in the asymmetric unit, in the native structure have two zinc ions in the active site. Both zinc ions are tightly coordinated with the distance between 2.0 and 2.4 Å ( Figure S2). When PNGM-1 structure is superimposed to Bs-tRNase Z, the position of the two zinc ions in both structures is conserved ( Figure  S3). We compared the PNGM-1 structure with carbapenem-bound subclass B3 MBL structures such as biapenem-bound CphA and doripenem-bound SMB-1 ( Figure S4A and B). The superimposed biapenem and doripenem are almost in the same position as the 3' end of tRNA close to the zinc ions in the active site (Figures 2 and 7). The substrate-binding pocket between PNGM-1 and subclass B3 MBLs is well conserved (Figure 7(B-D)). When we soaked doripenem in the crystals of PNGM-1, the doripenem with the hydrolyzed β-lactam ring was determined in the doripenem-bound PNGM-1 mutant (H257A) structure ( Figure S5). However, the efficiency (k cat / K m for carbapenems [imipenem, meropenem, and ertapenem]) of PNGM-1 MBL activity is significantly lower than that of the single native MBL activity of AIM-1, GOB-18 or FEZ-1 by 3 or 4 orders (Tables 1 and S7) and lower than that of Tm-tRNase Z by 3 orders [21]. β-Lactam-containing antibiotics including carbapenems (substrates for MBLs) existed in natural environments [9] and natural antibiotics appear to exist in the biosphere billions years ago [38,39]. It means that the B3 MBL activity of PNGM-1 is a promiscuous activity and subclass B3 MBLs is thought to have evolved from PNGM-1. It is consistent with the Jensen conceptualized enzyme evolution showing that newly specialized enzymes are generated from promiscuous activities [18,40].
Based on our results, we suggest that subclass B3 MBLs arose through an evolutionary trajectory (tRNase Z → PNGM-1→ subclass B3 MBLs) and the origin of subclass B3 MBLs is a tRNase Z. The evolutionary trajectory would be strongly supported if two MBL fold metallo-hydrolases (Pl and Ac) might have the dual activity of PNGM-1. If a B1/B2 MBL with dual activity, like PNGM-1, could be identified in the future, we would be able to understand the evolutionary origin of subclass B1/B2 MBLs, as MBLs have evolved independently twice (B1/B2 and B3 subclasses) billions years ago [15,18,41]. These evolutionary processes help us to understand where MBL genes came from and predict the future evolution of MBL genes, as previously described [39]. In addition, the molecular evolutionary process (dimer interfacial deformation to prevent dimerization of an ancestral protein such as a tRNase Z, dimer) provides a new strategy to develop novel enzymes (e.g. class B MBLs, monomer). To date novel enzymes have rationally been designed by two main approaches: (i) grafting of a reactive catalytic motif into a protein scaffold; and (ii) amino acid substitutions to improve catalytic activity or to create new catalytic activities [42].