Histidine-Triad Hydrolases Provide Resistance to Peptide-Nucleotide Antibiotics.

The Escherichia coli microcin C (McC) and related compounds are potent Trojan horse peptide-nucleotide antibiotics. The peptide part facilitates transport into sensitive cells. Inside the cell, the peptide part is degraded by nonspecific peptidases releasing an aspartamide-adenylate containing a phosphoramide bond. This nonhydrolyzable compound inhibits aspartyl-tRNA synthetase. In addition to the efficient export of McC outside the producing cells, special mechanisms have evolved to avoid self-toxicity caused by the degradation of the peptide part inside the producers. Here, we report that histidine-triad (HIT) hydrolases encoded in biosynthetic clusters of some McC homologs or by standalone genes confer resistance to McC-like compounds by hydrolyzing the phosphoramide bond in toxic aspartamide-adenosine, rendering them inactive.IMPORTANCE Uncovering the mechanisms of resistance is a required step for countering the looming antibiotic resistance crisis. In this communication, we show how universally conserved histidine-triad hydrolases provide resistance to microcin C, a potent inhibitor of bacterial protein synthesis.

Although most McC is efficiently exported outside the producing cell by the MccC pump, intracellular processing by aminopeptidases should inevitably lead to the accumulation of toxic nonhydrolyzable aspartamide-adenylate and self-intoxication of the producer, since MccC does not export processed McC. Many mcc-like clusters acquired additional genes whose products help avoid self-intoxication. In the case of E. coli, the C-terminal domain of MccE, a Gcn5-related N-acetyltransferase (GNAT)-type acetyltransferase, acetylates the ␣-amino group of processed McC, making it unable to bind to AspRS (8). In addition, MccF peptidase cleaves the carboxamide bond between the C-terminal aspartamide and AMP of both intact and processed McC (9).
In this work, we report a novel pathway of McC inactivation by histidine-triad (HIT) superfamily hydrolases encoded in some mcc-like biosynthetic clusters or by standalone genes located elsewhere in bacterial genomes. Proteins of the HIT superfamily form two separate functional groups: the first group includes nucleotide hydrolases, represented by HinT (10)(11)(12)(13), Fhit (13), APTX (14), and Dcsp (15) enzymes, while the other group includes nucleotide transferases such as GalT (16). The most common members of the HIT superfamily, HinT proteins, were shown to possess phosphoramidase activity (17). We show that bacterial MccH, a product of a gene in an mcc-like cluster from Hyalangium minutum, as well as its homologs from Salmonella enterica, Nocardiopsis kunsanensis, and Pseudomonas fluorescens, are phosphoramidases that confer resistance to McC-like compounds by hydrolyzing the toxic aspartamideadenylate that is produced after intracellular processing of peptidyl-nucleotides.

RESULTS
Bioinformatic prediction and experimental validation of an unusual mcc operon of Hyalangium minutum. Bioinformatics analysis reveals a uniquely organized cluster in the genome of the Gram-negative bacterium Hyalangium minutum DSM 14724 that may determine the production of two putative McC-like compounds. The cluster contains two genes coding for putative precursor peptides, MccA 1 and MccA 2 , two mccB genes, encoding THIF-like adenylyl transferases, and three genes whose products likely constitute a complex ABC-type transporter with integrated HlyD-like translocator and C39-like peptidase (18) (Fig. 2A). An additional gene, mccH, is located downstream of the mccB 1 gene and encodes a protein belonging to a histidine-triad (HIT) superfamily (12).
To validate the predicted H. minutum mcc-like cluster, in vitro adenylation reactions of synthetic MccA 1 and MccA 2 peptides with recombinant MccB 1 and MccB 2 adenylyltransferases were performed and products analyzed by matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS). As can be seen from Fig. 2B  We next attempted to reconstruct the production of each of the H. minutum McC-like compounds in a heterologous E. coli host. Cognate mccA-mccB pairs were cloned on one expression plasmid, and the mccP 1 P 2 P 3 genes encoding the putative transporter were cloned on a compatible plasmid. Under conditions of induction of plasmid-borne genes, E. coli cells or cellular extracts harboring both plasmid pairs did not inhibit the growth of an McC-sensitive E. coli tester strain, and no mass ions corresponding to H. minutum McC-like compounds were detected in cultured medium (see Fig. S1A in the supplemental material). To test if peptidyl-adenylates are synthe- hinT Eco or hinT Hmi had no effect on the size of growth inhibition zones produced by McC 1120 (Fig. 3). All HIT proteins were produced in comparable amounts, as judged by SDS-PAGE (Fig. S2). We therefore conclude that the MccH Hmi but not HinT proteins tested can provide resistance to externally added toxic peptidyl-adenylate.  In the presence of HinT Eco or HinT Hmi , McC 462 remained largely intact, with only trace amounts of AMP formed in the course of the reaction. We therefore decided to assess whether HinT Hmi is an active phosphoramidase using AMP-N--(N-␣-acetyl-lysine methyl ester)-5=-phosphoramidate (K-AMP), a previously described HinT phosphoramidase model substrate (17). Incubation of K-AMP with HinT Hmi , HinT Eco , or MccH Hmi , followed by RP-HPLC and MALDI-TOF MS, revealed that both HinT Hmi and HinT Eco hydrolyzed it with the release of AMP, while MccH Hmi did not ( Fig. 5A to C). We therefore conclude that the MccH Hmi hydrolase cleaves the P-N bond in aspartamideadenylate but not in K-AMP. HinT Hmi and HinT Eco have different specificities, where they hydrolyze K-AMP well but are largely inactive toward processed McC 1120 (Fig. 4A and 5A). These results explain why both HinT enzymes failed to provide resistance to McC in the antibiotic susceptibility test under our conditions in vivo (Fig. 3).
MccH homologs are present in diverse bacteria. HIT domain-containing proteins are widespread among prokaryotes (see Materials and Methods for details on domain identification). A phylogenetic tree constructed using available HIT domain sequences revealed that MccH Hmi belongs to a distinct clade, highlighted in red in Fig. 6A (see also Fig. S4). This clade also contains proteins from putative mcc-like clusters from Nocardiopsis, Pseudomonas, and Thermobifida spp., as well as multiple proteins encoded by standalone genes. Genes encoding MccH homologs from the mcc-like cluster from Nocardiopsis kunsanensis DSM 44524, as well as standalone genes from Pseudomonas Mutational analysis of MccH active center. The mechanism of nucleotide phosphoramidate hydrolysis is best studied for the E. coli enzyme, HinT Eco (19), and its human homologue, hHint1 (20,21). The characteristic feature of the HIT superfamily proteins is a conserved histidine-triad motif, HxHxHxx, where H is a histidine, and x is a hydrophobic residue (12). The three essential catalytic histidines form a network of hydrogen bonds with the substrate that promotes proton transfer from the protonated C-terminal histidine of the triad (H103 in HinT Eco or H102 in HinT Hmi ) to phosphoramidate unbridged oxygens and amide nitrogen and facilitate nucleophilic attack of the central histidine (HinT Eco H101 or HinT Hmi H100) on the phosphorus atom, resulting in P-N bond hydrolysis (19,21,22). Another conserved His residue, H39 in HinT Eco (H38 in HinT Hmi ), located outside the triad motif closer to the N terminus of HIT enzymes, contributes to catalysis by stimulating the protonation of the third histidine of the triad by stabilizing its cationic state (21). Mutational analysis of human Hint1, a close homologue of Hint Eco , revealed that substitutions of conserved histidines equivalent to H39 and H103 in HinT Eco substantially reduce the catalytic activity (22).
Interestingly, HIT proteins of the MccH clade contain a modified motif where the third histidine is substituted for lysine (K103 in MccH Hmi ). Together with this substitu-  (Fig. 7A and S4). Structural modeling (see below) indicates that F44 makes hydrophobic contacts with the side chain of K103 in MccH Hmi (Fig. 8A). In addition, all members of MccH clade acquired glutamate (E93 in MccH Hmi ) five residues away from the triad motif. In the structural model of MccH Hmi , E93 makes a hydrogen bond (or a salt bridge) with K103, thus playing the same functional role as H39 in HinT Eco (H38 in HinT Hmi ) (Fig. 8). Since this triple substitution should still allow phosphoramide bond hydrolysis, we speculate that the positively charged lysine occupying position of H103 in MccH Hmi ultimately donates its stationary proton to the nitrogen of the phosphoramide bond (Fig. 8B). To test this conjecture and better understand the origin of MccH-like protein specificity, we prepared two MccH Hmi mutants harboring the single substitutions K103H and F44H. An MccH Hmi K103H-F44H double mutant was also engineered. As a key residue in the triad, H101 in MccH Hmi is supposed to directly participate in catalysis. Therefore, a protein with substitution H101N was prepared and tested for phosphoramidase activity. As expected, the H101N mutant, which served as a control, was catalytically inactive, i.e., no hydrolysis of processed McC 1120 was detected (Fig. 7B). The K103H substitution had also eliminated HIT-Like Proteins Confer Resistance to Microcin C ® the hydrolytic activity of MccH Hmi , confirming that a lysine characteristic of MccH-like proteins is essential for the catalytic function. The F44H mutant retained its activity, which is also an expected result. The K103H-F44H double mutant was insoluble, and thus, its enzymatic properties could not be assessed. To confirm the in vitro hydrolytic activity of the mutants, the E. coli cells harboring the corresponding plasmids were tested for their susceptibility to McC 1120 . As shown in Fig. 7C, the H101N and K103H substitutions completely abolished immunity to peptidyl-adenylates, while the phenotype of MccH Hmi F44H-expressing cells was indistinguishable from that of the cells producing wild-type MccH Hmi .

Structural model of MccH Hmi . The loss of functional activity by MccH Hmi mutant
K103H suggested that this residue, which is specific to MccH clade proteins, is involved in substrate binding and/or catalysis. We also hypothesized that some other active-site residues might spatially constrain the catalytic pocket environment favoring the flexible aliphatic side chain of lysine over a more rigid imidazole ring of histidine. To explore the possible spatial organization of the active center of MccH Hmi , we generated its threedimensional (3D) model using the SWISS-MODEL homology modeling program (23). The resulting model of MccH Hmi (Fig. 8A) is based on a top-ranked Mycobacterium paratuberculosis HIT-like protein structure (PDB 3P0T) (24) with a global model quality estimate (GMQE) quality score of 0.68, indicative of good reliability and accuracy. To reveal the potential interactions in the substrate-binding pocket, the model structure of MccH Hmi was superimposed with the crystal structure of the human histidine-triad nucleotide-binding protein 1 (hHINT1) in complex with AMP (PDB 3TW2) (25).
As expected, the overall structure of MccH Hmi and the spatial organization of its active site are very similar to those of other HinT and HIT-like proteins, with a notable exception of the C-terminal nonconserved 45 amino acids that model differently depending on the homology template used. In the model, MccH Hmi forms a symmetric homodimer with each protein monomer capable of binding and hydrolyzing the substrate (Fig. S5). The nucleoside-binding pocket is formed mostly by conserved hydrophobic residues F11, F12, L15, F34, P37, V46, F38, and I97. The hydroxyl group of T36 makes a hydrogen bond with ribose 2=-OH in AMP, thus contributing to nucleotide recognition. The N atoms of the side chains of catalytic H101 and K103 are positioned (2.5 to 2.7 Å) to make strong hydrogen bonds with the P and unbridged O atoms of the phosphate moiety of AMP, respectively (Fig. 8A). The interatomic distances (2.9 to 3.6 Å) between the peptide backbone amide and carbonyl groups of residues G96, I97, and H99 and their interacting partners (unbridged oxygen and the protonated N atom of H101, respectively) are within a range that is optimal for hydrogen bonding and consistent with the proposed catalytic mechanism (Fig. 8B). Unexpectedly, the carbonyl group of E93, a conserved residue among members of the MccH clade, is in close proximity (2.3 Å) to the protonated N atom of K103, suggesting a strong hydrogen bond or salt bridge that would stabilize the charged state of K103 and facilitate the catalysis. Furthermore, consistent with the results of our mutagenesis experiments (Fig. 7), the bulky hydrophobic side chains of F44 and F91 replacing conserved H39 and I86 in HinT Eco and HinT Hmi , about the side chain of K103, stabilize its conformation and direct it toward the phosphate. Thus, the positioning of F44, F91, and E93 in the active center explains the observed preference for a catalytic lysine in MccH instead of histidine in the HinT clade.

DISCUSSION
In this work, we uncover a novel mechanism of immunity to microcin C-like compounds by MccH Hmi , a HIT-like phosphoramidase encoded in the mcc cluster of H. minutum. The cluster produces two separate peptide-adenylates that are analogous to McC 1120 , a toxic maturation intermediate of E. coli McC that lacks the aminopropyl decoration. As of today, the mcc operon from H. minutum is the only validated operon that produces two McC-like compounds with different peptide parts. Since peptide parts determine the specificity of antibacterial action by allowing selective import into sensitive cells (26), H. minutum DSM 14724 may target distinct, nonoverlapping sets of its competitors by the McC-like compounds it produces.
Like other McC-producing organisms, H. minutum should experience the buildup of toxic processed products inside the cell, which could lead to the cessation of protein biosynthesis. The MccH Hmi enzyme, the product of the mcc operon, alleviates this problem by cleaving the bond between phosphorus and nitrogen in the toxic aspartamide-adenylate that is produced after proteolytic processing of either of the two McC-like compounds encoded by the operon, thus providing self-immunity to the producing cell. MccH Hmi makes cells resistant to the maturation intermediate of E. coli McC that lacks the aminopropyl decoration but not to fully mature McC 1177 . The HIT-Like Proteins Confer Resistance to Microcin C ® catalytic mechanism of phosphoramide hydrolysis requires a transient protonation of two unbridged oxygens (21,22). The presence of an additional aminopropyl group on the phosphate in unprocessed McC 1177 and processed McC 519 precludes the proton transfer reaction and renders the phosphorus center inaccessible to nucleophilic attack by the catalytic histidine (HinT Eco H101, HinT Hmi H100, or MccH Hmi H101). Thus, H. minutum and E. coli mcc operons use different strategies to overcome the self-intoxication of producers. The H. minutum mcc operon produces two peptidyl-adenylates without additional modifications, and the processing of both compounds leads to identical toxic aspartamide-adenylate. MccH Hmi hydrolyzes the phosphoramide bond in aspartamide-adenylate with the formation of AMP and aspartamide. The absence of additional genes in the H. minutum mcc operon that may be involved in self-immunity suggests that MccH Hmi is sufficient to counter the inhibitory effects caused by the buildup of the toxic product. In E. coli, the MccD/E enzyme complex installs the aminopropyl decoration at the phosphate of peptide-adenylate, which allows the potency of antibacterial action to be increased ϳ10-fold by increasing the affinity of the processed compound to its target, AspRS (4). The presence of activity-enhancing decoration renders the MccH Hmi enzyme inactive, necessitating another mechanism to overcome selfintoxication. MccE detoxifies both aminopropylated and nonaminopropylated aspartamide-adenylates by acetylating the amino group of the aspartate (8).
The structural model of MccH Hmi built based on a crystal structure of homologous HIT-like protein from M. paratuberculosis (PDB 3P0T) (24) provides a plausible view on a spatial organization of the active center and offers clues to understanding the enzyme's substrate specificity. Importantly, the model points to the functional role of the conserved hydrophobic (F44 and F91) and charged (E93) residues in the activation of catalytic K103 for the hydrolysis of aspartamide-adenylate that can be tested experimentally. It is also predicted that residues M95 and W115 of the C-terminal loop of one MccH monomer together with L110 from the adjacent C-terminal loop of the other MccH monomer form a tight aspartamide-binding site which would sterically occlude the binding of bulkier groups, such as the -lysine amide of K-AMP. This view is consistent with the fact that HinT Eco lacking the C-terminal extension present in MccH Hmi was active toward K-AMP but could not hydrolyze aspartamide-adenylate. It is also supported by previous observations that both deletion and swapping of the C-terminal loop between human HINT1 and HinT Eco strongly affect both the catalytic activity and substrate specificity (19,27,28). The proposed model will be validated in our future genetic, biochemical, and structural studies of bacterial MccH and HinT proteins.
Previous studies have shown that bacterial and human HinT proteins exhibit a broad substrate specificity; they can accommodate both purine and pyrimidine nucleotides with various substitutions in the aminoacyl moiety, including D-and L-stereoisomers of tryptophan and sterically hindering N--(N-␣-acetyl-lysine methyl ester)-adenosine phosphoramidates (11,27). Unlike HinT, the MccH Hmi and its homologues from four diverse bacterial species characterized in this work apparently have evolved much more specialized enzymes that show a clear preference for aspartamide-adenylate ( Fig. 4  presence of K-AMP by MALDI-TOF mass spectrometry, and fractions containing K-AMP were subjected to additional chromatographic purification on the same column in a linear gradient of acetonitrile (0 to 25%) in triethylammonium acetate (TEAA) buffer (pH 6.5).
In vitro phosphoramidase activity assay. To test the phosphoramidase activity of MccH Hmi , HinT Hmi , HinT Eco enzymes, and their respective variants, the processed forms of E. coli McC 1177 and McC 1120 (McC 519 and McC 462 , respectively) and K-AMP were used as the substrates. For production, purification, and processing of McC forms, refer to the study by Metlitskaya et al. (6). Fifty micromolar processed McC or K-AMP was mixed with 5 M the enzyme in the reaction buffer (20 mM HEPES [pH 7.2], 2.5 mM MgCl 2 , 2.5 mM MnCl 2 ). The reaction mixture was incubated at 25° for 30 min, terminated by the addition of 0.1% TFA, and analyzed for hydrolysis by HPLC and mass spectrometry.
Reverse-phase HPLC analysis of the products of the in vitro reactions. All biochemical reactions were analyzed on 1220 Infinity II LC system (Agilent), and the peak separation occurred on a Zorbax Eclipse Plus C 18 5-m (4.6 by 250 mm) column (Agilent) in a 0.1 M TEAA buffer system (pH 6.0) in the varying linear gradient of acetonitrile.
The products of McC 462 hydrolysis reactions were separated in a linear gradient of acetonitrile (0 to 20%) over a period of 15 min. After the incubation of McC 519 with HIT enzymes, the reaction products were separated in the acetonitrile gradient (0 to 22%) for 15 min. After hydrolysis of K-AMP by HIT enzymes, the reaction products were analyzed in the linear acetonitrile gradient (5 to 30%) lasting for 15 min. The chromatograms were processed with the use of the ChemStation software (Agilent), and elution profiles were exported in comma-separated values format.
Mass spectrometry analysis. One to two microliters of the sample aliquots was mixed with 0.5 l of matrix mix (Sigma-Aldrich) on a steel target. The mass spectra were recorded on an UltrafleXtreme MALDI-TOF/TOF mass spectrometer (Bruker Daltonics) equipped with a neodymium laser. The molecular MH ϩ ions were measured in reflector mode; the accuracy of the measured results was within 0.1 Da.
Sequence analysis. Proteins containing the histidine catalytic triad were identified in 4,621 completely sequenced genomes available in 2016. Profiles belonging to the NCBI CDD (30) superfamily cl00228 were used as PSI-BLAST (31) queries to search the protein sequences encoded in this set. The resulting set of 10,580 proteins was clustered using UCLUST (32) and aligned using MUSCLE (33) (Fig. S6); alignments were iteratively compared to each other using HHSEARCH and aligned using the HHALIGN program (34). The approximate ML tree was reconstructed using the FastTree program (35) with a WAG evolutionary model and gamma-distributed site rates.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.