GcsR, a TyrR-Like Enhancer-Binding Protein, Regulates Expression of the Glycine Cleavage System in Pseudomonas aeruginosa PAO1

Glycine is required for various cellular functions, including cell wall synthesis, protein synthesis, and the biosynthesis of several important metabolites. Regulating levels of glycine metabolism allows P. aeruginosa to maintain the metabolic flux of glycine through several pathways, including the metabolism of glycine to produce other amino acids, entry into the trichloroacetic acid cycle, and the production of virulence factors such as hydrogen cyanide. In this study, we characterized GcsR, a transcriptional regulator that activates the expression of genes involved in P. aeruginosa PAO1 glycine metabolism. Our work reveals that GcsR is the founding member of a novel class of TyrR-like EBPs that likely regulate glycine metabolism in Pseudomonadales.

A mino acids are a preferred nutrient source for the Gram-negative opportunistic pathogen Pseudomonas aeruginosa (1,2). This bacterium can use almost any amino acid as either a sole carbon or a sole nitrogen source (2). The amino acid glycine is a major source of single carbon units for biochemical reactions in the cell and is necessary for the biosynthesis of several important metabolites such as purines, thymidine, methionine, threonine, and lipids (3)(4)(5). Aside from basic metabolism, glycine is also a precursor of hydrogen cyanide (HCN), which P. aeruginosa uses as a virulence factor (6).
The metabolism of glycine is dependent on a multienzyme complex known as the glycine cleavage system (GCS), which is composed of three proteins, GcvH, GcvP, and GcvT (7). GcvP is a dehydrogenase that catalyzes the decarboxylation of glycine and transfers the aminomethyl moiety to the lipoyl prosthetic group of GcvH. GcvT is an aminomethyltransferase that mediates the release of ammonia from the intermediate attached to GcvH and the synthesis of 5,10-methylene tetrahydrofolate (5,10methylene-THF) in the presence of tetrahydrofolate (THF). Overall, the GCS catalyzes the conversion of glycine into carbon dioxide, ammonia, and 5,10-methylene-THF. In addition to the GCS, there are two other proteins that play important roles in glycine metabolism. First, the serine hydroxymethyltransferase (GlyA) transfers a methylene group from 5,10-methylene-THF to glycine, thus forming serine and THF as products. The GlyA reaction is reversible and is necessary for the biosynthesis of glycine from serine (8,9). The other key protein for glycine metabolism is serine dehydratase (SdaA), which catalyzes the deamination of serine into pyruvate, an intermediate of central metabolism (10,11). The GCS-GlyA-SdaA pathway provides a route for the utilization of glycine as a carbon and nitrogen source by bacteria.
In the case of P. aeruginosa PAO1, two sets of genes coding for GCS proteins are found on the chromosome. There are also three genes coding for GlyA and two serine dehydratase genes in the P. aeruginosa PAO1 genome. Among these, one set of glycine metabolism genes, the gcvH2-gcvP2-glyA2-sdaA-gcvT2 genes, are located together in the chromosome and are collectively known as the gcs2 cluster. This gene cluster was previously found to be downregulated in the absence of the gene PA2449 (here gcsR [for glycine cleavage system regulator]) (12). Indeed, the gcsR gene was also essential for the utilization of glycine as a sole carbon source by P. aeruginosa PAO1 (12). The gcsR gene encodes a transcriptional regulator belonging to the enhancer-binding protein (EBP) family. EBPs are an interesting family of regulators because they possess domains that enable them to specifically interact with the alternative sigma factor 54 (RpoN) to initiate transcription from their target promoters (13). Upstream of the gcs2 cluster is a putative RpoN promoter, supporting a model in which GcsR regulates the transcription of gcs2 genes in response to glycine availability.
In addition to being an EBP, GcsR also shows Ͼ40% identity to the TyrR transcriptional regulator of Escherichia coli. This further classifies GcsR as a TyrR-like EBP. The members of the TyrR family of regulators are known for their roles in aromatic amino acid biosynthesis and metabolism (14)(15)(16)(17). The P. aeruginosa PAO1 genome has two genes encoding the TyrR-like EBPs PhhR and GcsR (18). PhhR was previously identified as a regulator of aromatic amino acid metabolism in P. aeruginosa PAO1 (16,19). GcsR shows 44% sequence homology to PhhR (12).
In this study, we show a novel mechanism for the regulation of the P. aeruginosa PAO1 gcs2 cluster by the TyrR-like EBP GcsR. We demonstrate that, unlike other TyrR regulators that respond to aromatic amino acids, GcsR activates the transcription of the gcs2 cluster in response to glycine. In addition, we show that the five gcs2 genes are transcribed as an operon and GcsR binds to an 18-bp tandem repeat sequence in the promoter region of the gcs2 operon to activate transcription. Although the P. aeruginosa PAO1 genome contains multiple homologs of each of the gcvH, gcvP, gcvT, glyA, and sdaA genes, our work indicates that only the gcs2 operon is essential for the metabolism of glycine as a sole carbon source. GcsR also appears to link glycine metabolism to virulence since the paralytic killing of the nematode Caenorhabditis elegans is significantly enhanced in the absence of gcsR.
Our work also reveals that the gcsR gene is conserved in nearly all of the sequenced genomes of the members of the order Pseudomonadales and is found adjacent to genes encoding a GCS. The results presented in this report indicate that GcsR is the prototype of a new family of TyrR-like EBPs that regulate glycine metabolism. The five glycine metabolism genes gcvH2, gcvP2, glyA2, sdaA, and gcvT2 were the only genes that were differentially expressed in both the gcsR transposon mutant and the ΔgcsR mutant. The transcription levels of these genes were also the most affected in ΔgcsR PAO1. Additionally, we found that, like the transposon mutant, the ΔgcsR mutant was unable to grow on glycine as a sole carbon source ( Fig. 1), providing evidence that the glycine metabolism genes are likely to be an important target of GcsR.
The gcs2 cluster genes are cotranscribed in P. aeruginosa PAO1. The glycine metabolism genes downregulated in the ΔgcsR PAO1 strain are arranged together in the gcs2 gene cluster, with the gcvH2 gene being the first gene, followed by the gcvP2, glyA2, sdaA, and gcvT2 genes ( Fig. 2A). The gcvH2, gcvP2, and gcvT2 gene products together make up the GCS (4), the glyA2 gene encodes a serine hydroxymethyltransferase that catalyzes the reversible conversion of glycine to serine (9), and sdaA encodes a serine dehydratase that catalyzes the deamination of serine to a pyruvate (11). Because of their organization in the genome, their function in glycine metabolism, and the fact that they were all downregulated in the ΔgcsR PAO1 strain, we proposed that these genes are cotranscribed and hence regulated by GcsR. In order to test this hypothesis, we determined the operon structure by reverse transcriptase (RT) PCR analysis of RNA isolated from P. aeruginosa PAO1 grown for 48 h to an optical density at 600 nm (OD 600 ) of 0.25 in minimal medium with glycine as the sole carbon source (Fig. 2B) from the isolated RNA and used as a template for PCR amplification with primers that amplified 500-to 600-bp fragments that spanned intergenic regions from the 3= end of the upstream gene to the 5= end of the adjacent downstream gene. PCR products were observed for all four primer sets, indicating that these five genes are transcribed together as a single operon (Fig. 2B).
Expression of a P gcvH2 -lacZ fusion in E. coli is dependent on GcsR and is regulated by glycine availability. GcsR was expected to regulate the expression of the gcs2 operon in response to glycine. To test this hypothesis, the 5= regulatory region (~500 bp) upstream of gcvH2 (P gcvH2 ) was fused to the ␤-galactosidase (lacZ) open reading frame (ORF) of E. coli. The P gcvH2 -lacZ fusion was then cotransformed with a plasmid harboring the gcsR gene into nonnative E. coli. As shown in Fig. 3A, ␤-galactosidase (LacZ) levels increased 3-fold with the addition of 10 mM glycine to our recombinant E. coli strain. The addition of serine or glutamate had no effect on LacZ levels, suggesting that induction of the P gcvH2 -lacZ fusion was specific to glycine.
Since wild-type E. coli can metabolize glycine, the internal glycine concentrations were too low to induce lacZ expression, as evidenced by the relatively minor differences in LacZ activity levels in the absence of glycine (20 Miller units [MU]) and in the presence of glycine (70 MU) (Fig. 3A). GcvA is an essential activator of the GCS in E. coli (22). Because an E. coli ΔgcvA mutant is unable to assimilate glycine, the LacZ activity of the P gcvH2 -lacZ fusion was 12-fold higher in this strain than in wild-type E. coli (Fig. 3A). In contrast, GcvR is a negative regulator of glycine metabolism in E. coli (23), so its absence was not expected to alter the regulation of the P gcvH2 -lacZ fusion by GcsR. Accordingly, in the absence of exogenous glycine, the LacZ level was 259 MU in the ΔgcvA mutant compared to 20 and 19 MU in the wild-type and ΔgcvR mutant strains, respectively (Fig. 3A). The addition of glycine increased the LacZ level by 2-fold in ΔgcvA mutant E. coli, which was similar to the fold change observed in wild-type and ΔgcvR mutant E. coli (Fig. 3A). Lastly, the presence of the gcsR gene was required for glycine induction of P gcvH2 -lacZ in E. coli (Fig. 3B). Replacement of the plasmid harboring the gcsR gene with either an empty plasmid equivalent or a plasmid having the gcsR gene in the opposite orientation did not generate significant levels of LacZ activity. Since GcsR is predicted to be an EBP and since there is a putative RpoN binding site in the P gcvH2 promoter, we wanted to check whether RpoN is required for transcription from this promoter. We found that an E. coli ΔrpoN mutant expressing GcsR and harboring the reporter plasmid showed relatively low LacZ levels and no significant difference in LacZ levels in the presence or absence of glycine (17.9 MU in the absence of substrate and 27.7 MU in the presence of glycine) (Fig. 3A). This indicates that RpoN is required for initiation of transcription from the P gcvH2 promoter.
In order to verify that GcsR regulates the P gcvH2 promoter in vivo, we measured the activity of the P gcvH2 -lacZ fusion in the wild-type PAO1, ΔgcsR PAO1, and ΔrpoN PAO1 strains (Fig. 4). We found that the LacZ levels in PAO1 increased 3-fold in the presence of exogenous glycine compared to that in cells grown with no substrate or in the presence of serine or glutamate. Additionally, there was no difference in LacZ levels in the presence or absence of glycine in the ΔgcsR PAO1 strain or in the ΔrpoN PAO1 strain, further indicating that both GcsR and RpoN are required for transcription from the P gcvH2 promoter.

GcsR binds to the gcvH2 promoter region with high specificity and affinity.
On the basis of homology, the gcsR gene encodes a TyrR family EBP (12). Since all of the characterized TyrR family EBPs regulate the metabolism of aromatic amino acids (14,15), this makes GcsR a unique member of this group. To understand how GcsR regulates glycine metabolism, we first identified its target promoters. The 54 Promoter Database (http://www.sigma54.ca/promoterdata/Web/data.aspx) predicts a putative strong RpoN binding site (score of 92) 79 bp upstream of the predicted gcvH2 ORF, the first gene of the gcs2 operon. Furthermore, the gcs2 operon is located adjacent to the gcsR gene in the PAO1 genome (18). This made the gcvH2 promoter the most likely target for GcsR. To verify this, we used electrophoretic mobility shift assays (EMSAs) to monitor the binding of His 6 -GcsR to a Cy5-labeled DNA probe containing the 200-bp region immediately upstream of the putative RpoN binding site of the gcvH2 promoter (P gcvH2 ). EMSAs showed that His 6 -GcsR was indeed able to bind to P gcvH2 (Fig. 5A). The affinity of GcsR for this piece of DNA is very high, judging from the shift produced by as little as 6.25 nM His 6 -GcsR, and the intensity of the shift increased with increasing concentrations of His 6 -GcsR (Fig. 5B). We also found that His 6 -GcsR bound to this region with high specificity. To test the specificity of His 6 -GcsR binding, we compared its binding of P gcvH2 with its binding to the nonspecific probe P PA5530 , which contains a 200-bp region upstream of the RpoN binding site on the PA5530 promoter (24). Figure 5A shows that while 200 nM His 6 -GcsR completely shifts the P gcvH2 probe (lane 4), there is no shift in the nonspecific P PA5530 probe (lane 2). Also, Fig. 5C shows that addition of the nonlabeled P gcvH2 probe led to depletion of the shift (lanes 3 to 5), but addition of the nonlabeled P PA5530 probe did not have any effect on the shift (lane 6). Taken together, these data suggest that the gcvH2 promoter region is indeed a target for GcsR. The EMSAs revealed that the binding of His 6 -GcsR with the P gcvH2 probe resulted in shifts to three distinct positions (Fig. 5). This suggests that there are at least three binding sites for GcsR in the gcvH2 promoter region. This is consistent with the binding patterns of known EBPs. EBPs bind as dimers to tandem repeat sequences known as enhancer elements upstream of the RpoN binding site in the target promoter region (13). Typically, three sets of EBP dimers bind to the promoter region. Upon activation, EBPs form a hexameric ring that goes on to interact with RpoN to initiate transcription with the help of ATP hydrolysis. Since we observed three distinct shifts, we expected there to be three GcsR binding sites in the gcvH2 promoter. Sequence analysis revealed three 18-bp tandem repeat sequences in the 200 bp P gcvH2 probe (Fig. 6A). In order to investigate the effects of these sequences on GcsR binding, we mutated the first tandem repeat sequence. We found that the binding pattern changed from three shifts with the wild-type sequence to two with the mutated sequence (Fig. 6B). We then inserted mutations into each of the other two tandem repeat sequences separately. We found that GcsR was able to bind each site on its own but the binding was abolished when the individual sites were mutated (Fig. 6C). Taken together, these data indicate that this 18-bp tandem repeat sequence is indeed the binding site for GcsR. Using these three binding sites, we were able to identify the following 18-bp consensus site for GcsR binding: TGTAACG-N 4 -CGTTCCG (Fig. 6D). The binding site is composed of two 7-bp palindromic arms separated by four nucleotides. Upon comparison with the TyrR binding site TGTAAA-N 6 -TTTACA, we find that although the GcsR and TyrR binding sites are very similar, the GcsR binding site has an invariable CG pair at positions 6 and 7 and a CG or GG pair at positions 12 and 13 that is not present at the TyrR binding site.
In addition to the glycine metabolism genes, the RNA-Seq analysis revealed that 15 other genes were differentially expressed in ΔgcsR PAO1 (Table 1). However, none of these had a putative RpoN binding site or a putative GcsR binding site and therefore they are probably not direct targets of GcsR.
GcsR shows TyrR-like phosphatase activity. The TyrR protein of E. coli exhibits phosphatase activity that is dependent on the presence of divalent cations (25). The role of this phosphatase activity in vivo has yet to be determined for TyrR. Nonetheless, we decided to determine if GcsR also possesses phosphatase activity in vitro. His 6 -GcsR was incubated with or without different divalent cations at 2 mM for 3 h at 37°C. Among the divalent cations tested, Zn 2ϩ and Co 2ϩ had the most significant effect on GcsR phosphatase activity (Fig. 7). The presence of glycine in the assay mixture did not have any effect on the phosphatase activity of GcsR.
GcsR is not essential for pyocyanin biosynthesis in P. aeruginosa PAO1. Our previous study indicated that GcsR was crucial for the production of pyocyanin in P. aeruginosa PAO1 (12). Specifically, a gcsR transposon mutant (PW5126) was observed to be deficient in pyocyanin production. Plasmid-derived expression of the gcsR gene did restore pyocyanin production in PW5126, suggesting that pyocyanin biosynthesis was dependent on gcsR. However, the gcsR mutant generated in this study exhibited wild-type levels of pyocyanin production (Fig. 8A), contradicting these previous findings.
A reasonable explanation that might account for the discrepancies in pyocyanin production between the two gcsR mutants is the deregulation of quorum sensingrelated genes in PW5126. Microarray analysis revealed that numerous quorum sensing genes were significantly downregulated in PW5126 (12). One of these genes was rhII, which encodes the N-butyryl homoserine lactone (C 4 -HSL) synthase (26,27). Because C 4 -HSL positively regulates pyocyanin biosynthesis (28), the downregulation of rhlI and, consequently, the absence of C 4 -HSL would explain why the PW5126 strain was unable to produce pyocyanin. Consistent with this hypothesis, plasmid-derived expression of rhlI did rescue pyocyanin production in PW5126 (Fig. 8A). In addition to downregulation of rhlI, the previous transcriptome analysis indicated that there was Ͼ100-fold upregulation of the mexEF and oprN genes in PW5126. The MexEF-OprN pump has been shown to be responsible for the efflux of biosynthetic precursors of 2-alkyl-4-quinolone, the Pseudomonas quinolone signal, which affects the expression of rhlI (29)(30)(31). Indeed, deletion of the mexF gene from PW5126 did restore pyocyanin production (Fig. 8A). This result suggests that upregulation of the mexEF and oprN genes from PW5126 was the likely cause of its pyocyanin deficiency. Unexpectedly, we found that GcsR did bind with low affinity to the promoter region of mexEF-oprN (Fig. 8B). This low-affinity binding might explain the previous observation that plasmid-derived expression of gcsR rescued pyocyanin biosynthesis in PW5126; i.e., low-affinity binding of GcsR to the P mexE promoter region antagonized the transcription of the mexEF and oprN genes.
HCN-induced killing of C. elegans by ⌬gcsR PAO1. One of the metabolic fates of glycine in P. aeruginosa is its conversion into HCN through the actions of an HCN synthase (HcnABC). HCN is a virulence factor and causes paralytic killing (HCN-mediated lethal paralysis) of C. elegans (32,33). Since GcsR was necessary for the assimilation of glycine, we wanted to determine if GcsR has any impact on the paralytic-killing capabilities of P. aeruginosa PAO1.
After 2 h of exposure to ΔgcsR PAO1,~60% of the worms were dead (Fig. 9A). Figure  9B and C show live, healthy, motile C. elegans on a lawn of PAO1 after 2 h of exposure. In contrast, Fig. 9D and E show worms on a lawn of ΔgcsR PAO1 after 2 h. The worms in Fig. 9 D and E are dead, as indicated by a lack of pharyngeal pumping. The worm have kinked or curved bodies, as dead C. elegans worms commonly do. In some worms, the nose was hypercontracted (data not shown), as previously described (32). C. elegans exposed to control strains P. aeruginosa lecA::luxΔlasR PAO1 and E. coli OP50 had motility and appearance similar to those of worms on wild-type P. aeruginosa PAO1 lawns (data not shown).
Despite the results of the paralytic-killing assays, RNA-Seq data (see Materials and Methods) showed no significant differences in the expression of the hcnABC genes in ΔgcsR PAO1 and EMSAs indicated that there was no binding of GcsR to the 5= upstream regulatory region of hcnA (data not shown). This suggests that GcsR does not directly regulate HCN production. Since glycine catabolism is abolished in the GcsR mutant, we hypothesize that more glycine flux into the HCN production pathway is occurring, leading to increased HCN production and lethality.

DISCUSSION
TyrR family EBPs regulate the expression of genes involved in aromatic amino acid metabolism (14,15,19,34). Many members of this family have been well characterized, including PhhR from P. aeruginosa PAO1 (16,19). Interestingly, in addition to PhhR, P. aeruginosa PAO1 has another TyrR homolog (18). The second one is GcsR, and we have found that, unlike other previously characterized TyrR homologs, it regulates the expression of genes required for glycine metabolism. The primary route of glycine catabolism is catalyzed by the GCS (4). P. aeruginosa PAO1 harbors two GCS gene clusters (18). We have found that the gcs2 gene cluster, comprising the gcvH2P2T2, glyA2, and sdaA genes, forms an operon and is essential for the metabolism of glycine as a sole carbon source. The best-characterized regulation of the expression of a bacterial gcs operon is that of E. coli, where its expression is regulated by the LysR-type transcriptional activator GcvA (22) and the repressor GcvR (23). Our work has revealed that in P. aeruginosa PAO1, a novel TyrR family EBP, GcsR, regulates the expression of this gcs2 operon and hence the metabolism of glycine. GcsR is the first characterized member of the TyrR family that regulates the metabolism of a nonaromatic amino acid.
GcsR has 41% sequence identity with E. coli TyrR. Sequence analysis shows that its domain architecture is identical to that of other TyrR homologs (15,35) RpoN interaction domain, and a C-terminal DNA-binding domain. Like other TyrR family EBPs, GcsR responds to an amino acid effector (glycine). We found that GcsR upregulates the expression of the gcs2 operon in the presence of glycine. Another piece of evidence indicating that GcsR is a TyrR homolog is that it has divalent-cationdependent phosphatase activity like E. coli TyrR (25). How GcsR binds to metals is unclear since its sequence does not have any recognizable metal-binding motifs such as stretches of multiple cysteine or histidine residues. Although the exact purpose of the phosphatase activity of TyrR or GcsR is not known, one possibility is that GcsR regulates expression from its target promoter by dephosphorylating other phosphorylated proteins. Further work is needed to better characterize the phosphatase activity of GcsR. P. aeruginosa PAO1 is not the only pseudomonad with a genome that harbors a gcsR gene. We found that there are GcsR orthologs in a number of other sequenced genomes ( Table 2). All of the genomes that harbor a GcsR ortholog belong to the order Pseudomonadales. Additionally, all of these GcsR orthologs are located adjacent to a glycine metabolism gene cluster (18) (Fig. 10A). This likely indicates a conserved mechanism for regulating glycine metabolism in members of the order Pseudomonadales.
The binding site of GcsR is an 18-bp tandem repeat sequence with two palindromic arms that is very similar to that of TyrR binding sites (Fig. 6D). We analyzed the genome sequences of P. putida KT2440, P. fluorescens Pf0-1, and P. syringae pv. tomato DC300 and identified binding sites for GcsR in the 5= regulatory regions of the gcs gene clusters in these genomes (Fig. 10B). The 54 Promoter Database (http://www.sigma54.ca/ promoterdata/Web/data.aspx) indicates that there are putative RpoN binding sites upstream of the gcs cluster in all three strains. Additionally, RpoN has been shown to be required for glycine metabolism in P. putida KT2440 (36). Taken together, these data suggest that the regulation of glycine metabolism by GcsR and RpoN is likely a common mechanism among members of the order Pseudomonadales. The P. aeruginosa genome harbors two homologs of the gcvHPT genes, three homologs of the glyA gene, and two homologs of the L-serine dehydratase genes, namely, sdaA and sdaB. However, this does not translate to redundancy in the function of the glycine metabolism genes. The glyA1 and sdaB genes have been shown to function in the glycine betaine catabolism pathway, and their expression is regulated by the AraC-like transcriptional activator GbdR (37,38). Our work indicated that only the gcs2 operon, gcvH2-gcvP2-glyA2-sdaA-gcvT2, is required for the metabolism of glycine as the sole carbon source. We also found that the growth defect displayed by ΔgcsR PAO1 with glycine as the sole nitrogen source was much less pronounced than the growth defect with glycine as the sole carbon source. This suggests that GcsR and, hence, the GCS2 proteins are not essential for the metabolism of glycine as a nitrogen source. Both the GCS and GlyA have been shown to be important for the metabolism of glycine as a nitrogen source in E. coli (39). However, since GcvH2P2T2 or GlyA2 does not appear to be essential for the metabolism of glycine as a nitrogen source by P. aeruginosa PAO1, it is likely that the other gcv genes and perhaps the glyA3 gene are involved in that process. This is an excellent example of the diverse metabolic capacity of P. aeruginosa PAO1.
Glycine also acts as the precursor for the production of HCN gas by an HCN synthase in P. aeruginosa (6). P. aeruginosa is one of the few bacteria that produce HCN (40). HCN has been detected in patients with P. aeruginosa infection and is thought to contribute to the pathogenesis of P. aeruginosa in cystic fibrosis patients (41). Although how HCN production by P. aeruginosa contributes to its virulence in cystic fibrosis lungs is not yet clear, it has been shown that P. aeruginosa uses HCN for paralytic killing of the nematode C. elegans (33). We found that the level of C. elegans paralytic killing by ΔgcsR PAO1 was much greater than that by the parent P. aeruginosa PAO1 strain. Since the GCS is not functioning in the ΔgcsR PAO1 mutant, more glycine is available for conversion to HCN, which is likely responsible for the increased lethality of the ΔgcsR PAO1 strain.
We had previously found that, in addition to regulating glycine metabolism, GcsR also plays a role in pyocyanin production (12). Pyocyanin is one of several redox-active phenazines produced by P. aeruginosa that are used as virulence factors (42)(43)(44)(45)(46)(47). Pyocyanin causes virulence by affecting gene expression, electron transport, cellular respiration, and innate immune mechanisms (42,46,47). We had previously found that pyocyanin production is reduced in the gcsR transposon mutant strain PW5126 (12). In this study, we found that this decrease in production of pyocyanin in PW5126 is due not to the direct regulation by GcsR of biosynthetic genes but to overexpression of the MexEF-OprN efflux pump. However, we found that this was not reproducible in the ΔgcsR PAO1 strain. When we tested the pyocyanin production of this strain, we found that it was not significantly reduced. Furthermore, in contrast to PW5126, RNA-Seq analysis of ΔgcsR PAO1 showed no significant change in mexEF-oprN expression. The parent strain of the gcsR transposon mutant strain PW5126 is the PAO1 strain from Colin Manoil's lab at the University of Washington (48). Our ΔgcsR PAO1 strain is a derivative of a different PAO1 strain from the Dieter Haas lab at the University of Lausanne in Switzerland. It has been shown that in many PAO1 strains originating from different laboratories, mexEF-oprN is quiescent and uninducible because of mutations in several regulatory genes (49,50). Recent studies have also shown differences in the genome sequences of different PAO1 strains originating from different laboratories (51). Additionally, unexpected mutations affecting the expression of mexEF-oprN have also been shown to alter the pyocyanin production of several PAO1 strains and a number of mutants from the PAO1 transposon library (20). Thus, one explanation for this discrepancy could be that the PW5126 strain had an inducible mexEF-oprN operon, whereas ΔgcsR PAO1 did not. The other possibility is that an adventitious mutation that allows the overexpression of mexEF-oprN could have occurred in PW5126 because of the transposon mutation. Further studies must be conducted with known mexEF-oprNinducible strains to fully understand the effect of GcsR on mexEF-oprN.
Our RNA-Seq analysis of the ΔgcsR PAO1 strain revealed that 20 genes were differentially regulated (Table 1). In addition to the five gcs2 operon genes, these genes include genes for several tRNAs, an operon encoding genes for formaldehyde metabolism, and another operon involved in nitrogen metabolism. However, none of the latter 15 genes are likely to be direct targets of GcsR. There are two lines of evidence to suggest this. First, we could not identify a putative GcsR binding site in the 5= regulatory region of any of these genes. Second, 14 out of these 15 genes do not have an RpoN binding site in their promoter region. The only gene with an RpoN binding site in its promoter region is dctA, which has been previously shown to be regulated by the DctD EBP (52). This, along with the fact that the dctA promoter region lacks a GcsR binding site, indicates that dctA is likely not a direct target of GcsR. All of these genes function in pathways that directly or indirectly depend on products from the glycine metabolism pathway. This could possibly explain why disruption of glycine metabolism likely causes changes in their expression patterns.
In conclusion, our data suggest that GcsR is the first characterized member of a novel family of TyrR-like EBPs that regulate genes required for glycine metabolism. Unlike E. coli, which uses a LysR-type regulator to regulate glycine metabolism, GcsR also presents a novel mechanism of regulation of glycine metabolism involving the alternative factor RpoN. Future work will determine how GcsR interacts with glycine to regulate the expression of its target genes.

MATERIALS AND METHODS
Bacterial strains, plasmids, and growth conditions. The bacterial strains, plasmids, and oligonucleotides used in this study are listed in Table 3. Bacteria were grown in Difco Lennox broth (LB), peptone broth (PB) (12), modified M9 minimal medium supplemented with 5 M FeSO 4 (12), 2ϫYT medium (53), or brain heart infusion (BHI) broth (RPI). Solid bacteriological medium was prepared with the addition of BD agar at 15 g liter Ϫ1 , unless otherwise specified. Liquid cultures were grown at 37°C with shaking at  (24,54) Briefly, the gcsR::Gm and mexF::Gm cassettes were cloned into pEX18ApGW to give plasmids pBRL528 and pBRL521, respectively. The gcsR gene was deleted from PAO1 with the pBRL528 plasmid, and the mexF gene was deleted from PW5126 with the pBRL521 plasmid. The Gm markers were removed with the pFLP2 plasmid. The ΔgcsR and ΔmexF mutations were verified by PCR.
Preparation of RNA. RNA of P. aeruginosa PAO1 and ΔgcsR PAO1 were isolated and purified as described previously (12). To isolate RNA for RNA-Seq, each strain was grown in quadruplicate in 50 ml of PB in 500-ml baffled shake flasks at 37°C and 200 rpm to an OD 600 of~0.5. To isolate RNA for RT analysis of operon organization, each strain was grown in quadruplicate in 50 ml of M9 minimal medium supplemented with 20 mM glycine as the sole carbon source in 500-ml baffled shake flasks at 37°C and 200 rpm to an OD 600 of~0.25. Cultures were immediately stabilized by adding 1 ml of RNAprotect Bacteria reagent (Qiagen) to 0.5 ml of culture. Cells were then lysed with lysozyme and proteinase K as described in the manufacturer's protocol. The total RNA was subsequently purified from the lysed cells with the RNeasy minikit (Qiagen) by using an on-column DNase digestion step. PCR and a Bioanalyzer were used to check the RNA for DNA contamination and quality.
RNA-Seq. The purified total RNA was sent to the Molecular Analysis Core at SUNY Upstate Medical University for mRNA isolation, cDNA library preparation, and RNA sequencing with the Illumina NextSeq 500 system. CLC Genomics workbench 8.5 (Qiagen) was used to map the sequencing reads to the P. aeruginosa PAO1 genome and to obtain differential gene expression data (a Ն2-fold change in the number of reads per kilobase of transcript per million mapped reads and a Bonferroni-corrected P value of Յ0.05) (55,56).
Analysis of operon organization of glycine metabolism genes. In order to map the operon organization of the gcs2 locus, RT analysis of isolated RNA was used. An RT reaction with 500 ng of each RNA sample was set up with the iScript cDNA synthesis kit (Bio-Rad) to obtain cDNA. Primers ZS448F/ ZS448R, ZS449F/ZS449R, ZS454F/ZS454R, and ZS451F/ZS451R were designed to amplify 500-to 600-bp regions spanning from the 3= end of gcvH2 to the 5= end of gcvP2, from the 3= end of gcvp2 to the 5= end of glyA2, from the 3= end of glyA2 to the 5= end of sdaA, and from the 3= end of sdaA to the 5= end of gcvT2, respectively. Primers ZS452F/ZS452R were designed to amplify the 50S ribosomal protein L21 gene rplU as a positive control (57). PCR analyses of the cDNA samples were subsequently performed with these primers. PCR analyses of the RNA samples and genomic DNA samples were performed as negative and positive controls, respectively. The PCR products obtained were analyzed by agarose gel electrophoresis.
LacZ reporter assays. Recombinant E. coli strains were grown in M9 minimal medium supplemented with glycine, glutamine, or serine at a 20 mM concentration. Each culture was grown in triplicate. Cells were grown to an OD 600 of 0.3. ␤-Galactosidase (LacZ) activity was measured with the Miller assay as described previously (12).
Heterologous expression and purification of GcsR. GcsR was expressed and purified as an N-terminally six-histidine-tagged fusion protein (His 6 -GcsR) from plasmid pBRL417 harboring the gcsR gene using E. coli BL21(DE3) as described previously (53). Cells were grown in 2ϫYT medium to an OD 600 of~0.6 at 37°C with shaking at 200 rpm. Protein expression was induced by the addition of 0.1 mM isopropyl-␤-D-thiogalactopyranoside, and cell cultures were incubated for 12 h at 16°C with shaking at 200 rpm. Cells were harvested, resuspended in lysis buffer (100 mM sodium phosphate [pH 8.0], 300 mM NaCl, 10% glycerol, 1 mg/ml lysozyme, 5 U/ml DNase I, 1 g/ml pepstatin, 1 g/ml leupeptin), and lysed by three 30-s sonication bursts. The lysate was centrifuged at 9,000 ϫ g to remove insoluble material, and the His 6 -GcsR was then purified with Ni-nitrilotriacetic acid Superflow resin (Qiagen). The protein was eluted off the resin by a step elution method with elution buffer (100 mM Tris [pH 8.0], 300 mM NaCl) containing 20, 100, and 250 mM imidazole. Purified protein was concentrated with Amicon Ultra centrifugal filter units (Millipore). Protein expression and purification were monitored visually by sodium EMSAs. DNA fragments containing the promoter region of the gcvH2 gene were amplified by PCR and cloned into the pJET1.2 vector. The mutated promoter regions were obtained with a quick-change site-directed mutagenesis kit (Qiagen). Cy5-labeled primers JRH05.f/JRH05.r were used to PCR amplify the probes from the resulting plasmids, yielding the Cy5-labeled probes used in subsequent EMSAs.
The binding of GcsR to P gcvH2 was investigated by EMSA (58). A 200 nM sample of His 6 -GcsR was incubated with 2 nM Cy5-labeled P gcvH2 probe (specific probe) or 2 nM Cy5-labeled P PA5530 probe (nonspecific probe) in EMSA buffer (25 mM Tris-acetate [pH 8.0], 8 mM magnesium acetate, 10 mM potassium chloride, 1 mM dithiothreitol) for 30 min at 30°C. For EMSA reactions to examine the binding affinity of His 6 -GcsR for P gcvH2 , 2.0 nM 5=-labeled Cy5 P gcvH2 probe was incubated with 0, 6.25, 12.5, 25, 50, 100, or 200 nM His 6 -GcsR. For EMSA reactions to determine the binding specificity of His 6 -GcsR for P gcvH2 , a 200 nM sample of His 6 -GcsR was incubated with 2.0 nM 5=-labeled Cy5 P gcvH2 probe in the presence of 0, 25, 50, or 100 nM unlabeled P gcvH2 probe (specific competitor probe) or 100 nM unlabeled P PA5530 probe (nonspecific competitor probe). For EMSA reactions to determine the binding site of GcsR on the gcvH2 promoter, a 200 nM sample of His 6 -GcsR was incubated with 2 nM each probe containing either a wild-type or a mutated binding site. For EMSA reactions to determine the binding of GcsR to the mexE promoter region, a 200 nM sample of His 6 -GcsR was incubated with 2 nM probe P mexE . The samples were then analyzed by PAGE under nondenaturing conditions and imaged with a Typhoon imager.
Phosphatase assay. Phosphatase activity of His 6 GcsR was measured as described previously for E. coli TyrR (25). Briefly, a 5 M sample of His 6 -GcsR was incubated at 37°C for 3 h in phosphatase buffer (100 mM p-nitrophenylphosphate, 100 mM HEPES [pH 6.5]) in the presence of 2 mM glycine, 2 mM MgCl 2 , 2 mM ZnCl 2 , and 2 mM CoCl 2 or in the absence of any divalent salts or glycine. Reactions were stopped by the addition of 1 ml of 1 M Na 2 CO 3 . The absorbance at 410 nm of the samples was measured with a Genesys 20 spectrophotometer to calculate the amount of substrate dephosphorylated.
Spectroscopic measurement of extracellular pyocyanin Experiments were done in triplicate. Cells were grown in 2 ml of PB at 37°C and 200 rpm for 24 h. Pyocyanin was measured as described previously (12). Briefly, cells were removed from the culture by centrifugation at 16,000 ϫ g for 5 min and passage through Acrodisc syringe filters with 0.2-m nylon membranes. The absorbance at 690 nm of the cell-free samples was measured with a Genesys 20 spectrophotometer and converted to the pyocyanin concentration by using a molar extinction coefficient of 4,130 M Ϫ1 cm Ϫ1 .
Paralytic-killing assay. The C. elegans Bristol (N2) strain was cultivated under standard conditions (59). From synchronized populations of worms, hermaphroditic gravid adults were transferred to fresh nematode growth medium agar on a lawn of E. coli for 2 to 4 h to lay eggs and then removed. Eggs were incubated at 20°C for 3 days to the young adult stage (60).
Bacteria were grown overnight in BHI broth at 37°C with shaking. Cultures were diluted 1:100 in BHI broth, and 170-l volumes were spread on BHI agar plates (1.7% Bacto agar, 60-mm petri plates) and then incubated at 37°C for 24 h (32). Synchronized worms were transferred to prepared BHI plates and incubated at room temperature for 6 h. Worms were examined for paralysis and scored for survival every 2 h with a stereomicroscope (ϫ4.5 magnification). Worm paralysis was defined as no movement after mechanical stimulation, and death was defined as cessation of pharyngeal pumping. Data were analyzed with GraphPad Prism (GraphPad Software Inc.). Differences were considered significant when the P value was Յ0.05.
Microarray data accession number. RNA-Seq data were posted to the Gene Expression Omnibus (GEO) under accession number GSE76522.