The uridylyltransferase GlnD and tRNA modification GTPase MnmE allosterically control Escherichia coli folylpoly-γ-glutamate synthase FolC

Folate derivatives are important cofactors for enzymes in several metabolic processes. Folate-related inhibition and resistance mechanisms in bacteria are potential targets for antimicrobial therapies and therefore a significant focus of current research. Here, we report that the activity of Escherichia coli poly-γ-glutamyl tetrahydrofolate/dihydrofolate synthase (FolC) is regulated by glutamate/glutamine-sensing uridylyltransferase (GlnD), THF-dependent tRNA modification enzyme (MnmE), and UDP-glucose dehydrogenase (Ugd) as shown by direct in vitro protein–protein interactions. Using kinetics analyses, we observed that GlnD, Ugd, and MnmE activate FolC many-fold by decreasing the Khalf of FolC for its substrate l-glutamate. Moreover, FolC inhibited the GTPase activity of MnmE at low GTP concentrations. The growth phenotypes associated with these proteins are discussed. These results, obtained using direct in vitro enzyme assays, reveal unanticipated networks of allosteric regulatory interactions in the folate pathway in E. coli and indicate regulation of polyglutamylated tetrahydrofolate biosynthesis by the availability of nitrogen sources, signaled by the glutamine-sensing GlnD protein.

Protein-protein interactions regulate many processes in Escherichia coli and probably in all living cells. Carbon source availability is a signal for the recently identified global regulation of glycolysis and energy (ATP) balance by the phosphocarrier protein of the bacterial phosphotransferase system (PTS) HPr (1). Coregulation of carbon and nitrogen metabolism resulting, for example, in activation of glucosamine 6-P deaminase, NagB, by HPr, synergistically with uridylylated PII under limiting nitrogen conditions, is important for nutrient homeostasis (2). We have collaboratively published evidence for the existence of a network of protein-protein interactions (the interactome) in E. coli (3). This work provided the incentive and a guide for the research reported in this and previous works describing allosteric regulatory phenomena (1,2).
Bifunctional folylpoly-␥-glutamate synthase, or dihydrofolate (DHF) 3 synthase (FolC), catalyzes polyglutamylation of DHF; tetrahydrofolate (THF); 5,10-methylene-THF; and 10-formyl-THF using two additional substrates, ATP and L-glutamate. The enzyme also catalyzes glutamylation of 7,8dihydropteroate, the biosynthetic precursor of folic acid. E. coli is able to synthesize DHF de novo, but eukaryotes must take up extracellular folate using FolT from the reduced folate carrier family (Transporter Classification Database (TC) no. 2.A.48). Bacteria use vitamin uptake transporters of the ECF family (TC no. 2.A.88) (4) or a member of the folate-biopterin transporter (FBT) family present in protists, cyanobacteria, and plants (TC no. 2.A.71) (5)(6)(7)(8)(9). Once inside the cell, folate is reduced via DHF reductase (FolA) to DHF and then to THF (Fig. 1), the precursor to many folate derivative cofactors (5). Upon reduction to THF, FolC catalyzes the addition of glutamyl residues, resulting in the formation first of THF-monoglutamate and then of THF-polyglutamate derivatives; the latter exhibit full cofactor activity. This process is tightly regulated in mammals (10,11). Folate derivatives are important cofactors for enzymes that are involved in several metabolic processes including serineglycine interconversion, methionine recycling, purine biosynthesis, the thymidylate cycle, and one-carbon metabolism (11,12). As a result of their involvements in these key processes, folate derivatives are universally required for cell growth (13). However, humans lack the ability to synthesize these compounds de novo and must rely on their diets to acquire adequate levels of folate (10).
GlnD senses the cytoplasmic L-glutamine and L-glutamate concentrations, and signal transduction systems deliver output signals, even in complex media (14). Variations in the concen-trations of proteins and nutrients stimulate signal transmission for optimal enzymatic activity and growth. GlnD catalyzes uridylylation of the PII protein but has not been known as a regulator of folate biosynthesis. We show that it interacts with and activates FolC in the presence of low concentrations of glutamate (Fig. 1). These observations may explain the fact that an experimental approach showed that GlnD, but not GlnA, GlnB, or GlnE, is an essential protein in E. coli (15,16), suggesting that it could be important for the design and characterization of inhibitors acting on the FolC/GlnD complex.
Ugd produces UDP-glucuronate, a precursor of colanic acid (17), an enterobacterial exopolysaccharide produced in response to stress conditions. With low cytoplasmic Mg 2ϩ concentrations, E. coli Wzc, a bifunctional two domain tyrosine kinase/colanate exporter responsible for the export of colanic acid (18 -23), phosphorylates Ugd. Under these conditions, charge repulsion between lipopolysaccharide molecules increases, preventing aggregation of cells (18,23). The participation of several combinations of regulatory systems, such as PhoP (in response to low Mg 2ϩ concentration) (21,24), RcsB, and PmrA, control ugd gene expression in response to a variety of signals including changes in temperature, osmolarity, and overexpression of membrane proteins.
MnmE is a THF-dependent tRNA modification GTPase found in complex with MnmG, another tRNA modification enzyme (25,26). Its abnormally high GTPase activity depends on the cytoplasmic potassium concentration; however, classical small GTP-binding proteins are usually regulated by activating proteins and guanine nucleotide-exchange factors (25,27). Typically, GTP-binding proteins show low intrinsic GTPase activity, but MnmE is the exception, being active in the homodimeric state in the presence only of potassium ions.
All four nucleotides in tRNAs can be posttranslationally modified, and modification of the anticodon is particularly important (26). Specifically, position 34, U34, known as the wobble position, is most often modified. The 5-carboxymethyl-aminomethyl-2-thio-type modification ensures fidelity of regions ending in A or G. Unmodified U34 will interact with all four nucleotides and lead to misincorporation of amino acids into the growing peptide, leading, for example, to poor growth of Salmonella enterica serovar Typhimurium (28). The pathway of U34 modification is complex and involves many proteins. Depending on the MnmEG substrate, an unmodified U34 is converted to aminomethyluridine using ammonium or 5-carboxy-methyl-aminomethyluridine using glycine. MnmE has also been found to play a potential role in virulence interactions of the hosts with S. enterica serovar Typhimurium and Pseudomonas syringae (29).
Interestingly, GadE is a key regulator of the major acid resistance system in E. coli which is composed of the glutamate decarboxylase isoenzymes GadA and GadB and a dedicated glutamic acid/GABA antiporter GadC. During the bacterial response to acid stresses, GadA and GadB catalyze the decarboxylation of glutamic acid, yielding GABA, which is subsequently exported by GadC in exchange for another molecule of glutamic acid. The anaerobic transcriptional activation of the gadE-mdtEF operon in E. coli is largely dependent on the global regulators ArcA and MnmE, as deletion of MnmE causes a significant decrease of the transcription of gadE-mdtEF (30,31).
Cytoplasmic GTP availability is very important for bacteria, and as shown here, MnmE GTPase activity, associated with tRNA modification, is inhibited by FolC at 75 M aminopterin (a THF analogue) when the GTP concentration is low (Fig. 1).
Folate pathway-related inhibition and resistance mechanisms have been actively studied in bacteria (32)(33)(34)(35). The previously mentioned metabolic processes involving THF operate more slowly with monoglutamylated folyl coenzymes than with polyglutamylated cofactors because folate-dependent enzymes exhibit low affinity for the monoglutamylated cofactors (36). As such, folylpoly-␥-glutamate synthase (FolC) is required for normal cell growth (37). Taken together, this makes FolC an ideal target for antimicrobial therapy, although resistance to FolC inhibitors in Mycobacteria has been documented (38).
We here demonstrate that activation of FolC by Ugd, MnmE, and GlnD occurs via novel regulatory mechanisms, impacting the activation of one-carbon metabolism in response to stress

GlnD, MnmE, and Ugd activate FolC in E. coli
conditions. FolC has the same binding site for THF and dihydropteroate (39). The modeling of the Ugd-GlnD-FolC interactions revealed the probable residues involved in this interaction. We also show that the GTPase activity of the tRNA modification enzyme, MnmE, is inhibited by FolC, thus revealing a novel mode of functional protein-protein interactions in E. coli. The importance of these interactions was confirmed by in vivo growth experiments. For the first time, regulation of polyglutamylated THF biosynthesis by the availability of a nitrogen source, sensed by the glutamine-sensing GlnD protein, is demonstrated for E. coli.

THF synthase (FolC) regulation
FolC polyglutamylates (derivatizes with glutamate) THF, yielding the active cofactor. To study FolC regulation, FolC, MnmE, GlnD, Ugd, TdcB (serine/threonine dehydratase), HybD (hydrogenase 2 maturation protease), and the PII protein were all purified (see "Experimental Procedures"). TdcB and PII were refolded as described in Ref. 1. Interactome data (3) had suggested that FolC interacts with proteins: Ugd, GlnD, MnmE, TdcB, and HybD. They were examined for their effects on FolC activity as shown in Figs. 2 and 3. GlnD activated at least 4-fold, whereas TcdB, HybD, and PII did not activate or inhibit appreciably under the conditions used for the experiment, using 350 M aminopterin and 2 mM L-glutamate. To investigate the effect of L-glutamate on the activation by GlnD, we measured the kinetics under constant conditions but with varying concentrations of L-glutamate with THF (Fig. 2B). Additionally, experiments were conducted with aminopterin, and the concentration of substrate that produces half-maximal enzyme velocity (K half ) was measured at 0.2 mM aminopterin (Fig. S1). GlnD activated FolC, reducing the K half for glutamate around 10-fold (Table 1). FolC activation by GlnD at low concentrations of L-glutamate and 0.15 mM tetrahydrofolate was possible only in the presence of 7 mM L-glutamine (Fig. 2B). Titration of
The presence of 1.5 mM ␣-ketoglutarate abolished activation by GlnD in the presence of 7 mM L-glutamine. Titration with GlnD at 1 mM L-glutamate and aminopterin, using the Pgi protein as a negative control, is shown in Fig. 3A. Twelve-fold activation by GlnD (but not by Pgi) using a 0.5 M protein concentration was found. FolC activation by Ugd at 4 mM L-glutamate and 0.4 mM aminopterin was most efficient at 5 mM ATP (Fig. 3B). We tested separately the effect of L-glutamine (1-30 mM) on activation of FolC by GlnD (0.3 M) with 0.5 mM L-glutamate, but no inhibitory or activation effect was found using aminopterin as substrate (data not shown). A small inhibitory effect on FolC activity was noticed for ␣-ketoglutarate at a concentration of 3 mM when using 1 mM L-glutamate and 0.12 mM aminopterin in the presence of GlnD (Fig.  3A). No synergistic or additive effect was observed for Ugd and GlnD (data not shown) suggesting either that both proteins bind to the same FolC allosteric-binding site, or that their binding sites overlap. The kinetics with varying concentrations of aminopterin are shown at Fig. S1.

Activation of FolC by MnmE and reciprocal inhibition of MnmE by FolC
To investigate the effect of L-glutamate on the activation by MnmE, we measured the kinetics under constant conditions but with varying concentrations of L-glutamate with THF or aminopterin (Fig. 2B). Activation by MnmE using a 0.8 M protein concentration was not observed with aminopterin as substrate, but activation using a 0.5 M protein concentration with THF was substantial. The kinetics followed the Michaelis-Menten model, and GraphPad Prism 7 was used for calculations to fit the equation. The Michaelis constant K m measured for L-glutamate without a protein-effector was 9.8 mM. MnmE substantially decreased the K m to 3.8 mM in the presence of 10 mM L-glutamine.
MnmE GTPase activity was measured using the assay described in "Experimental Procedures" in the presence and absence of aminopterin or L-glutamate, and also with and without FolC. These were added in different combinations. An increase in the GTPase activity of the methyl-THF-binding tRNA-modification protein MnmE was observed (2-fold) when the THF analogue aminopterin was added (data not shown). The presence of FolC in the assay mixture reduced the activation by aminopterin. Titration with GTP (Fig. 4) confirmed FolC-dependent inhibition of MnmE activity in the presence of the THF analogue aminopterin.

FolC/MnmE/Ugd docking model
The model for FolC binding showed both Ugd and MnmE binding to the C-terminal (smaller) domain of FolC. Interestingly, although the docking was performed separately for Ugd and MnmE, they showed considerable shape complementarity and possible interfaces with each other (Fig. 5). Also, their binding did not appear to sterically hinder access to the FolC active site in the N-terminal/C-terminal domain cleft. Residues involved in these interactions are listed in Table S1.

folC, glnD, and ugd genetically interact with each other as well as with genes involved in metabolic and folate biosynthetic pathways
To identify E. coli genes that genetically interact with folC, glnD, and ugd, we used our previously published genetic interaction dataset using the E. coli Synthetic Genetic Array (eSGA) approach (40). Using this approach, double mutants were derived by conjugating the donor HfrC strain (marked with ⌬::Cm R ) with single gene deletions or hypomorphic (i.e. partial loss of gene function) mutant strains of an essential gene with all other recipient nonessential or essential E. coli genes (marked with ⌬::Kan R ) under both prototrophic (rich medium) and auxotrophic (minimal medium) culture conditions. The colony growth and relative fitness of the double mutants that survived the double antibiotic selection were digital imaged and quantified to generate genetic interaction scores. This resulted in cases of synthetic sick or lethal (double mutants grew more

GlnD, MnmE, and Ugd activate FolC in E. coli
slowly than expected) or alleviating (double mutants grew more rapidly than expected) growth phenotypes. In accordance with our protein-protein interaction data and enzyme kinetics (Fig.  2), folC or ugd displayed strong synthetic sick or lethal with the carbon metabolism genes (accBC, frdBCD) and alleviating interactions with the genes involved in folate biosynthesis (folAM, pabC) and glutamate metabolism (purB, glmS) (Table  S2). Additionally, we confirmed the recipient glnD and ugd with a modest synthetic sick growth defect when combined with folC or glnD donor mutant strains (Fig. 6). This observation appears to be bona fide as no detectable growth defects were observed when folC or glnD donor mutants were combined with a functionally unrelated gene (40). Taken together, our results suggest that folC, glnD, and ugd function redundantly with each other and with the carbon metabolic genes, and they cooperate with genes involved in glutamate and folate biosynthesis.

Discussion
Glutamine metabolism is regulated by the PII protein (GlnB), uridylylated by GlnD (41)(42)(43). The cytoplasmic L-glutamate/Lglutamine concentrations determine the rates of GlnD uridylyltransferase activity (14). We measured activation of FolC by GlnD, working as a potential sensor of the glutamate/glutamine concentrations in the cell. The binding of GlnD to an allosteric site in FolC reduced the K half of FolC for glutamate maximally about 10-fold (Fig. 2, Table 1).
PII and GlnE (bifunctional glutamine synthetase adenylyltransferase/adenylyl-removing enzyme) are not essential proteins in E. coli, but GlnD is essential (15,16). We suggest that the mechanism of GlnD essentiality may involve activation of FolC. FolC activity decreased substantially without GlnD, and FolC activity is essential for E. coli growth, thus possibly rendering GlnD essential. Other possible mechanisms explaining GlnD essentiality may involve GlnD interactions with other essential enzymes, such as RnhB (RNase HII), which degrades DNA-RNA hybrids, and/or XerC (tyrosine recombinase) (3). Such potential interactions could also render GlnD proteinprotein interactions essential, but these possibilities have not been examined.
The colanic acid transporter Wzc regulates the activity of Ugd, activating it by phosphorylating a tyrosyl residue in the protein (19). Stress conditions, such as low magnesium, have been shown to activate expression of the ugd gene (21), and now we find that similar conditions affect one-carbon metabolism via protein-protein interactions involving Ugd and FolC, activating FolC.
Among the proteins interacting with FolC (3), MnmE activated FolC (Fig. 5), but PII and several other proteins tested had no obvious effect (data not shown). These results clearly substantiate some of the published interactome data (3) and suggest that they are physiologically important. The consequences of the other interactions reported have not yet yielded positive results. Future studies will be required to determine whether these proteins do regulate FolC or if FolC regulates the activities of some of these other proteins.
As noted above, regulatory effects of FolC on MnmE GTPase activity were detected (Fig. 5). This is the first reported proteindependent regulatory effect so far observed for MnmE. Typically, GTP-binding proteins have very low GTPase activity compared with MnmE (44). They require GTPase-activating proteins and GTP/GDP exchange factors to maximize their GTP hydrolase activities. In contrast, the G-domain of E. coli MnmE has high activity without added factors and has low affinity for GDP. The FolC inhibiting effect on the MnmE GTPase activity may allow growth by preventing GTP depletion. Thus, tetrahydrofolate-dependent activation of MnmE and compensatory inhibition of MnmE by FolC may be essen- Figure 6. Genetic screening among folC, glnD, and ugd interaction pairs. The Hfr Cavalli donor mutants, folC and glnD, were crossed with the indicated F Ϫ recipient mutant strains (*, essential hypomorph), and the resulting double mutants were selected on rich medium plates containing chloramphenicol (Cm R ) and kanamycin (Kan R ). The same recipient plate was used for screening of folC and glnD donors.

GlnD, MnmE, and Ugd activate FolC in E. coli
tial for the proper functioning of the MnmEG-protein complex. It appears that FolC activation by GlnD, but not by Ugd, is essential for the growth of E. coli strains (Fig. 6).

Protein purification
The clones encoding FolC, GlnD, Ugd, MnmE, TdcB, HybD, and GlnB (PII) were used from the ASKA collection (45) following verification by sequencing. For protein overproduction, the cells were grown at 37°C, induced with isopropyl ␤-D-1-thiogalactopyranoside for 4 h, and harvested. Then cells were resuspended in 20 mM HEPES buffer, pH 7, containing 100 mM NaCl, 2 mM ␤-mercaptoethanol, and 0.3% Brij 35 with 2 mM phenylmethylsulfonyl fluoride present. Cells were lysed by incubation with lysozyme (1 mg/ml) for 30 min, followed by a freeze-thaw cycle and sonication as described (1). After centrifugation, the supernatant was loaded onto a nickel-nitrilotriacetic acid agarose minicolumn (0.3 ml) from Qiagen (Valencia, CA). After bound proteins were washed with 2 ml of At-buffer containing 50 mM Tris-HCl buffer (pH 8), 0.5 M NaCl, 5 mM imidazole, and 0.3% Brij 35, they were eluted with 0.4 ml of the same buffer supplemented with 300 mM imidazole. Protein size, expression level, and purity were monitored by SDS-PAGE. All proteins were obtained in sufficient yield (ϳ0.3-0.5 mg) and purity (80 to 90% pure). Protein concentrations were measured using the Bradford assay kit (Bio-Rad).

FolC enzymatic activity measurements using an improved FolC assay
A novel assay was developed to test the effects of four different proteins and the negative control protein, phosphoglucoisomerase (Pgi), separately and simultaneously on FolC activity. This assay depends on the detection of ADP. At pH 8.7, where FolC activity is low, the interactions with Ugd and GlnD activate FolC substantially (Ͼ10-fold) (see "Results"). FolC catalyzes the ATP-dependent addition of L-glutamate to the ␥-carboxyl moiety of a glutamyl residue present in THF. FolC converts ATP to ADP while glutamylating THF, and ADP can be used by pyruvate kinase and lactate dehydrogenase to convert phosphoenolpyruvate to pyruvate and further to lactate, following the decrease in absorbance at 340 nm, resulting from the oxidation of NADH in a coupled assay. We added FolC (120 -200 nM) to 100 l of a reaction mixture containing 50 mM KCl, 50 mM Tris-HCl buffer, pH 8.7, 10 mM MgSO 4 , 0 -150 M aminopterin (the 4-amino derivative of folic acid) or THF, 0 -25 mM glutamate, 3 mM DTT, 1.2 mM ATP, 1.2 mM phosphoenolpyruvate, 0.3 mM NADH, 1.2 units of pyruvate kinase and lactate dehydrogenase. Reaction rates were compared with controls in which glutamate or aminopterin was absent. The kinase activity detection kit was used as described (46). When tetrahydrofolate instead of aminopterin was the substrate, DMSO was added to 10%, and PBS, diluted 10-fold, was used in addition to the other conditions described above for the FolC assay.

MnmE activity measurements
For the MnmE GTPase assay, the formation of phosphate was detected using a malachite green assay as described in Ref. 47. This sensitive and reproducible assay was used for the detection of P i in 96-well format. MnmE activity was measured as a function of GTP concentration, and FolC inhibition was quantitated. The assay mixture contained 80 mM Tris-HCl, pH 8.0, 2 mM DTT, 1 mM EDTA, 50 mM KCl, 10% glycerol, 0 -75 M aminopterin or THF, and 10 mM MgCl 2 , assayed at 37°C and incubated for 1 h. The malachite green reagent was added to stop the reaction, and the absorbance at A 595 nm was measured. Low FolC GTPase activity was observed relative to MnmE activity, and this was used as the background value.

FolC docking
FolC was docked to Ugd and MnmE. For FolC, PDB ID 1W78 chain A was used, whereas for Ugd a homology model using PDB ID 3PID chain A (83% sequence identity) was used, and for MnmE a homology model using PDB ID 3GEH chain A (38% sequence identity) was used. Both homology models were full length. For GlnD, a homology model could be generated for the central HD domain, but this could not be docked because of nonspecific predicted binding residues.
Binding residues were predicted for FolC, Ugd, and MnmE structures using the CPORT web server (50) with the highest sensitivity setting. Docking was performed employing the HADDOCK web server (51) using the aforementioned structures as well as CPORT predictions as input.
Selection of representative complexes from docking output was performed slightly differently for the two pairs, FolC-Ugd and FolC-MnmE. For FolC-Ugd, the top cluster was the best-scoring across nearly all criteria and was a clear choice. For FolC-MnmE, the top four clusters were all significant, but three of the four were very close spatially (low root mean square deviation) and the fourth was an outlier. The top-scoring cluster was therefore also used for FolC-MnmE. The representative structures for FolC-Ugd and FolC-MnmE were aligned in Pymol by the FolC chains. This aligned pair of complexes is referred to as the "model" for FolC binding.

Genetic crosses
Mini-array screens were performed following the E. coli Synthetic Genetic Array strategy (48) in rich medium conditions using folC and glnD as HfrC donor mutant strains. The folC hypomorphic allele was constructed with the C-terminal SPA (Sequential Peptide Affinity) tag extension engineered by homologous recombination essentially as described previously (40,49).