Moonlighting Glutamate Formiminotransferases Can Functionally Replace 5-Formyltetrahydrofolate Cycloligase*

5-Formyltetrahydrofolate (5-CHO-THF) is formed by a side reaction of serine hydroxymethyltransferase. Unlike other folates, it is not a one-carbon donor but a potent inhibitor of folate enzymes and must therefore be metabolized. Only 5-CHO-THF cycloligase (5-FCL) is generally considered to do this. However, comparative genomic analysis indicated (i) that certain prokaryotes lack 5-FCL, implying that they have an alternative 5-CHO-THF-metabolizing enzyme, and (ii) that the histidine breakdown enzyme glutamate formiminotransferase (FT) might moonlight in this role. A functional complementation assay for 5-CHO-THF metabolism was developed in Escherichia coli, based on deleting the gene encoding 5-FCL (ygfA). The deletion mutant accumulated 5-CHO-THF and, with glycine as sole nitrogen source, showed a growth defect; both phenotypes were complemented by bacterial or archaeal genes encoding FT. Furthermore, utilization of supplied 5-CHO-THF by Streptococcus pyogenes was shown to require expression of the native FT. Recombinant bacterial and archaeal FTs catalyzed formyl transfer from 5-CHO-THF to glutamate, with kcat values of 0.1–1.2 min−1 and Km values for 5-CHO-THF and glutamate of 0.4–5 μm and 0.03–1 mm, respectively. Although the formyltransferase activities of these proteins were far lower than their formiminotransferase activities, the Km values for both substrates relative to their intracellular levels in prokaryotes are consistent with significant in vivo flux through the formyltransferase reaction. Collectively, these data indicate that FTs functionally replace 5-FCL in certain prokaryotes.

The main reaction mediated by SHMT, the reversible conversion of serine to glycine and 5,10-methylene-THF, is a key source of C 1 units and SHMT is an almost ubiquitous enzyme (11)(12)(13). The capacity to form 5-CHO-THF is thus also almost universal, and consequently so is the need for 5-CHO-THF removal. It might therefore be expected that all organisms would have 5-FCL, except those few that lack SHMT (14) or folates (15). However, in the course of a comparative genomic analysis of folate synthesis and metabolism involving hundreds of genomes (16) we were surprised to note that diverse prokaryotes had genes encoding SHMT (glyA) and other folate-dependent enzymes but no gene for 5-FCL (ygfA).
This lack of the ygfA gene appeared to signal a case of a missing enzyme, or pathway hole, i.e. where there is either a completely different enzyme (nonorthologous displacement) or an alternative pathway (16,17). Here, we report an investigation of these possibilities using comparative genomics analysis. This analysis, together with literature mining, predicted that the histidine degradation enzyme glutamate formiminotransferase (FT, EC 2.1.2.5) could functionally replace 5-FCL in organisms that lack this enzyme. This prediction was experimentally verified by functional complementation of an Escherichia coli ygfA deletant, by antifolate rescue experi-ments with Streptococcus pyogenes, and by biochemical characterization of the formyltransferase activity of recombinant FT proteins.

EXPERIMENTAL PROCEDURES
Bioinformatics-Genomes were analyzed using the SEED data base and its tools (18). Archaea lacking folates were excluded from the analysis. Genes with anticorrelated distributions to that of ygfA were sought using the Phylogenetic Profiler tool at JGI and the Signature Genes tool at NMPDR. Sequences were aligned using ClustalW or Multalin. Phylogenetic analyses were made using MEGA4 (19).

5-Formyltetrahydrofolate Metabolism
exposed to NH 3 vapor for 2 h. Fractions containing FIGLU were pooled and freeze-dried.
Mutant Construction-Primers are listed in supplemental Table S1. The ygfA::Kan deletion from Keio collection strain JW2879 (28) was transferred to E. coli strain MG1655 by P1 transduction. Clones harboring the deletion were selected on LB containing kanamycin, and verified by sequencing. The S. pyogenes FT knock-out was obtained by single crossover recombination. A 421-bp fragment of the S. pyogenes FT gene was cloned in pSK-Erm, a pBluescript vector modified to confer erythromycin resistance (29). S. pyogenes JRS4 competent cells were obtained as described (26). Briefly, cells were grown in 20 ml of TH-HS medium containing streptomycin until A 560 reached 0.3 and harvested by centrifugation at 4°C (4500 ϫ g, 5 min). The washed cell pellet was resuspended in 100 l of ice-cold 0.5 M sucrose. Electroporation was performed by adding 2 g of the Sp421bp-pSK-Erm construct to 40 l of fresh competent cells; a Bio-Rad gene pulser was used (peak voltage 2.5 kV; capacitance 25 F; pulse controller 200 ⍀). Cells were incubated for 2 h at 37°C in 1 ml TH-HS, followed by plating on TH-HS containing erythromycin. Colonies appeared after 2 days. Mutants were screened by PCR using gene-and vector-specific primers (SpKOver-F and M13rev-R; SpKOver-R and M13fwd-F). The queT gene was used as a positive control (primers QueT-F and QueT-R). FT knock-out strains were electroporated as above with the S. pyogenes FT gene under the control of the strong P23 promoter (30,31) in the shuttle vector pAT28 (32). pAT28 has the spc spectinomycin resistance gene and the pAM␤1 origin of replication, which confers high copy number in Gramϩ bacteria (33). The FT gene was first subcloned into BamHI-SalI digested pOri23 (30). The P23-FT fragment was then excised with EcoRI-SalI and cloned into EcoRI-SalI-digested pAT28.
Expression  Table S1) designed with restriction sites to subclone the amplified product into pBluescript II SK (Stratagene) for complementation assays, or pET28b or pET21a (Novagen) for overexpression of proteins carrying a C-terminal His tag. Constructs were verified by sequencing.
Functional Complementation-E. coli ⌬ygfA cells harboring pACYC-RP and pSC101-RIL (Stratagene) encoding rare tRNAs were transformed with pBS II SK alone (negative control) or harboring E. coli ygfA (positive control) or a FT gene, plated on LB containing 1 mM isopropyl-␤-D-thiogalactopyranoside (IPTG), kanamycin, chloramphenicol, streptomycin, and carbenicillin, and incubated at 37°C. The next day, independent clones were streaked on the M9 medium above or a modified version in which 50 mM glycine replaced 20 mM NH 4 Cl as nitrogen source. Plates were incubated for 4 days at 37°C.

A Candidate for 5-CHO-THF Removal Predicted by Com-
parative Genomics-A systematic analysis of 621 bacterial and 19 archaeal genomes using the SEED data base (18) revealed that 618 have glyA genes encoding SHMT. Of these 618, 555 have ygfA genes specifying 5-FCL, and 63 (a significant minority) do not. The ygfA-deficient group includes taxonomically diverse bacteria and archaea ( Fig. 2A). As some other gene is expected to replace ygfA when it is absent, we sought genes whose phylogenetic distribution is anticorrelated with that of ygfA. The gene encoding the histidine degradation enzyme FT emerged from this analysis as the strongest candidate. The FT gene is frequently present when ygfA is absent and almost always absent when ygfA is present ( Fig.  2A); globally, FT occurs in 46% of the genomes without ygfA, but in only 2% of the genomes with it.
Classical biochemical data on FT add to its plausibility as a substitute for 5-FCL. FT is a folate-dependent enzyme that mediates a late step in histidine degradation in mammals and certain bacteria; other bacteria use alternative, folate-independent routes (Fig. 2B) (18,36). Specifically, FT catalyzes the transfer of a formimino group from FIGLU to THF, yielding 5-formimino-THF (5-NHϭCH-THF), which is then converted to 5,10-CHϭTHF by the action of formiminotetrahydrofolate cyclodeaminase (CD) (Fig. 2C). In mammals and some bacteria, FT and CD are fused together in a single protein ( Fig. 2A). Besides formimino transfer, mammalian FT is known to mediate an analogous formyl transfer between Nformylglutamate and THF, and to do so reversibly (Fig. 2C) (36,40). The formyltransferase activity of mammalian FT is far lower than its formiminotransferase activity and has been viewed as physiologically insignificant (6,36). Nevertheless, formyltransferase activity in the 5-CHO-THF 3 THF direction (henceforth termed the forward direction) would in principle enable 5-CHO-THF removal.
Further comparative genomic evidence of an alternative function for FT comes from its distribution pattern relative to those of other histidine degradation enzymes. Most prokaryotic genomes having FT also have CD and the other genes (hutHUI) needed for a complete histidine degradation path-way ( Fig. 2A), and some (e.g. S. pyogenes) have these genes in an operonic arrangement (18). In these cases, the FT enzyme is predicted to retain a function in histidine breakdown. However, certain genomes such as E. minutum and S. aciditrophicus lack the hutHUI genes, and the latter lacks CD as well ( Fig. 2A). Such conservation of FT in the absence of other histidine utilization genes implies a second function.
Phenotype of the E. coli ygfA Deletant-In order to develop a functional complementation test for alternatives to 5-FCL, we first determined the effects on folate metabolism and growth of deleting the ygfA gene in E. coli (which has no FT gene). The deletion was made in the standard K12 strain MG1655 by P1-transduction from a Keio collection strain (28). The deletant grew normally on LB rich medium or M9 minimal medium, and folate analysis detected little or no intracellular accumulation of 5-CHO-THF (not shown). However, supplementing M9 medium with glycine caused a 6-fold accumulation of 5-CHO-THF (Fig. 3A), presumably because the SHMT side-reaction that generates 5-CHO-THF is promoted by bound glycine (2). Glycine supplementation did not affect growth of the deletant but completely replacing the NH 4 Cl nitrogen source in M9 medium with 50 mM glycine led to a severe growth defect (Fig. 3B). This defect is likely due to a buildup of 5-CHO-THF inhibiting SHMT (6) and glycine cleavage (41), both of which E. coli needs to use glycine as  DECEMBER 31, 2010 • VOLUME 285 • NUMBER 53 nitrogen source (42). The growth defect was reversed by expressing the E. coli ygfA gene from a plasmid (Fig. 3B). These observations establish the ygfA deletant as a suitable platform for complementation tests of possible alternatives to 5-FCL.

5-Formyltetrahydrofolate Metabolism
Complementation of the ⌬ygfA Strain by Prokaryotic Formiminotransferase Genes-The ⌬ygfA deletant was transformed with pBS II SK alone or containing FT genes from five bacteria and two archaea; two of the genes were FT-CD fusions ( Fig. 2A). E. coli ygfA served as a positive control. Transformants were plated on M9 medium containing NH 4 Cl or 50 mM glycine as nitrogen source. All strains grew well on NH 4 Cl, indicating that none of the FT genes was toxic. Six out of the seven FT genes tested permitted good growth on glycine (Fig. 4A), indicating that they can functionally replace 5-FCL. The exception, H. aurantiacus FT-CD, may be a pseudogene; uniquely, the histidine utilization (hut) operon in H. aurantiacus is entirely broken up, suggesting that this pathway no longer functions. Also, related bacteria (e.g. Roseiflexus spp.) entirely lack FT and most other histidine utilization genes. Folates were analyzed in ⌬ygfA cells harboring vector alone, or containing representative FT genes (from S. pyogenes and S. aciditrophicus) or E. coli ygfA as a control. The data confirmed that FTs can replace 5-FCL: both FT genes reversed the accumulation of 5-CHO-THF as effectively as E. coli ygfA (Fig. 4B); the effects of FT expression on other folates were minor.
Function of S. pyogenes FT in Its Native Host-To reinforce the above evidence from heterologous expression that FT functions in vivo in 5-CHO-THF metabolism, we investigated the function of the FT from S. pyogenes in its native host. S. pyogenes was selected because it is genetically manipulable and a model human pathogen (43). We first ablated the FT gene by single crossover integration (Fig. 5A); qRT-PCR tests confirmed lack of FT expression in the disruptant and detected weak expression in wild-type cells (not shown). The disruption had no impact on the growth or the folate profile of cells cultured on TH-HS or CDM medium (not shown). This may reflect the low level of FT expression in wild-type cells, itself due to the FT gene being in a histidine operon (Fig.  5A) that is repressed by the culture conditions (44,45). We therefore compared the FT disruptant strain with or without the native gene expressed from a plasmid. In this comparison, folate synthesis was blocked by the antifolate drug sulfathiazole, resulting in inhibition of growth, and 5-CHO-THF was supplied to attempt to reverse the inhibition. For reversal to occur, cells must take up 5-CHO-THF and convert it to other folates. Expression of S. pyogenes FT effectively reversed sulfathiazole inhibition (Fig. 5B), showing that this gene confers the capacity to metabolize 5-CHO-THF when expressed in its native host. This result also shows that S. pyogenes has a folate transport system, as do other Firmicutes (46).
In Vitro Activities of Formiminotransferase Proteins-To show directly that prokaryote FT and FT-CD proteins have

5-Formyltetrahydrofolate Metabolism
formyltransferase activity, His-tagged recombinant proteins were purified and assayed for formyltransferase activity in both directions. Formiminotransferase activity in the physiological direction (FIGLU ϩ THF 3 Glu ϩ 5-NHϭCH-THF) was measured for comparison. The proteins analyzed were the FTs from S. pyogenes, P. torridus, and T. acidophilum, the FT-CD from Acidobacteria bacterium Ellin 345, and (as a benchmark) porcine FT-CD. The prokaryote protein preparations were all Ն95% homogeneous as judged by SDS-PAGE (Fig. 6A).
All the prokaryote enzymes catalyzed glutamate-dependent THF formation from 5-CHO-THF (Fig. 6B); the fact that no THF was formed in the absence of glutamate demonstrates that the reaction is a formyl transfer, not simply a deformylation. The porcine FT-CD control mediated the same reaction, as expected (40). None of the prokaryote enzymes catalyzed detectable THF formation when glutamate was replaced by any of the other 19 protein amino acids, indicating high specificity for glutamate as formyl acceptor. Kinetic characteriza-tion of the formyltransferase activities (Table 1) showed K m values in the micromolar range for 5-CHO-THF and for its triglutamate form; as triglutamates have been found to be the predominant folates in growing cells of various bacteria (47)   DECEMBER 31, 2010 • VOLUME 285 • NUMBER 53 the latter result confirms the effectiveness of the physiological form of the cofactor. The K m values for glutamate were two orders of magnitude greater. The k cat values for the prokaryote enzymes were equal to or greater than that for the porcine enzyme, and in one case (P. torridus) the difference was 10-fold. Because P. torridus FT also had the lowest K m value for 5-CHO-THF (Table 1), this result suggests that the P. torridus enzyme may have become more efficient in formyl transfer relative to formimino transfer.

5-Formyltetrahydrofolate Metabolism
To investigate this possibility, we estimated k cat values for formiminotransferase activity (in the physiological direction) and compared them to the above values for formyltransferase in the forward direction (Fig. 7). We also measured formyltransferase k cat values for the reverse reaction, to juxtapose with the literature on the porcine FT-CD (Fig. 7); our values for this enzyme agreed with published ones (36). The forward formyltransferase/formiminotransferase ratio (in parentheses, Fig. 7, bottom panel) for P. torridus FT was 330-fold higher than that of the porcine enzyme, and 170 -660-fold higher than those of the other prokaryote enzymes. The high ratio for P. torridus FT reflects a high formyltransferase activity combined with an exceptionally low formiminotransferase activity (Fig. 7). The P. torridus enzyme is thus a markedly more efficient formyltransferase than the others in terms of (i) the k cat /K m ratio for this activity and (ii) the relative magnitudes of the formyltransferase and formiminotransferase k cat values. However, even for this enzyme, the formyltransferase k cat value is 30-fold less than the formiminotransferase k cat value; for the other prokaryote enzymes, and for porcine FT, the formyltransferase k cat values are ϳ5,000 -20,000-fold less than the formiminotransferase k cat values. Formyltransferase activity is thus not the major activity in any of the enzymes tested.

DISCUSSION
The comparative genomic, genetic, and biochemical data presented above support the conclusion that the formyltransferase activity of the histidine degradation enzyme FT replaces 5-FCL in diverse prokaryotes. Most of these enzymes presumably retain their principal function in histidine catabolism because they have high formiminotransferase activity in vitro and their genes co-occur with the other genes of histidine breakdown, sometimes in operonic arrangements. Such FTs can therefore be viewed as moonlighting enzymes, i.e. as metabolic enzymes with an additional functional activity (48), in this case 5-CHO-THF salvage. As previously noted, how-ever, some FTs occur in genomes that lack most or all other histidine degradation genes; in these cases the moonlighting activity may have become the only physiological one.
Although the in vitro formyltransferase activity of FTs is much lower than their formiminotransferase activity, three factors may enable the formyltransferase activity to function effectively in 5-CHO-THF removal. First, prokaryote FTs  a The cofactors used were the natural (S) form of 5-CHO-THF versus mixed natural and unnatural forms (R,S) of 5-CHO-THF-Glu 3 , so that K m values for 5-CHO-THF-Glu 3 should be halved for comparison with those for 5-CHO-THF. b Values for k cat were determined with 5-CHO-THF as cofactor; those determined with 5-CHO-THF-Glu 3 were similar.

5-Formyltetrahydrofolate Metabolism
have low K m values for 5-CHO-THF (0.4 to 5.4 M; Table 1). Second, the in vivo concentration of the glutamate substrate would be saturating, because K m values for glutamate are Յ1 mM (Table 1) and intracellular glutamate concentrations in prokaryotes are typically 30 -100 mM (49 -52). Third, no more than a low formyl transfer flux is needed to remove 5-CHO-THF, which is formed only by a minor side reaction (6) and which does not come to dominate the folate pool even if normal removal is blocked (Fig. 3A).
If FTs function simultaneously in histidine degradation and 5-CHO-THF salvage in the same cell, then they must carry flux in both directions because the formiminotransferase reaction consumes THF and produces glutamate and the formyltransferase salvage reaction does the opposite (Fig. 2C). Given that moonlighting functions can depend on differences in protein-protein interactions (48), it seems possible that running these opposing reactions concurrently is facilitated by FT joining/not joining a protein complex related to histidine degradation. In this connection, it is relevant that mammalian FT-CD is an octamer that can dissociate into two types of monofunctional dimers, and that while the FT and CD activities of the octamer function separately with monoglutamylated folates, polyglutamates are channeled from the FT to the CD active site (34,53). Analogous behavior of prokaryote FT oligomers, or perhaps FT-CD complexes, together with polyglutamate-mediated channeling between FT and CD reactions, could result in the functional partitioning, i.e. mutual non-interference, of formyl-and formiminotransferase activities.
One of the FTs examined (that from the archaeon P. torridus) was far more efficient as a formyltransferase than the other prokaryote enzymes, which were themselves comparable in efficiency to porcine FT-CD. These findings are consistent with natural selection of the P. torridus protein for increased formyltransferase activity. They also imply that mammalian FT-CD cannot be ruled out a priori as a contributor to 5-CHO-THF salvage, particularly because glutamate levels in liver(to which FT-CD is confined) far exceed the K m for glutamate (206 M; Table 1) (54). However, any contribution from FT-CD is probably small relative to that from 5-FCL because 5-FCL is highly expressed in liver (55) and its specific activity appears to be at least one order of magnitude greater than that of FT-CD formyltransferase (36,56).
In removing 5-CHO-THF, the formyltransferase reaction generates N-formylglutamate (Fig. 2C), which is a metabolic dead-end that can apparently be metabolized only via hydrolysis to glutamate and formate. Although a dedicated N-formylglutamate hydrolase (Fig. 2, NfoD) participates in histidine degradation in certain bacteria (18,57), this protein is absent from prokaryotes in which FT replaces 5-FCL (Fig.  2). However, deacylation of amino acids is a common reaction and nonspecific enzymes that hydrolyze N-formylglutamate have been reported from bacteria and other organisms (58,59).
Finally, because our comparative genomic analysis indicated that FT occurs in only 46% of the genomes with SHMT (glyA) and without 5-FCL (ygfA), it is likely that at least one other gene for 5-CHO removal remains to be discovered in prokaryotes. Such a hypothetical missing gene might also be present in plants. Ablating the single gene encoding 5-FCL in Arabidopsis results in only a 2-fold increase in leaf 5-CHO-THF content, suggesting that 5-FCL is not the sole route for 5-CHO-THF disposal (9).