Novosphingobium aromaticivorans uses a Nu-class glutathione S-transferase as a glutathione lyase in breaking the β-aryl ether bond of lignin

As a major component of plant cell walls, lignin is a potential renewable source of valuable chemicals. Several sphingomonad bacteria have been identified that can break the β-aryl ether bond connecting most phenylpropanoid units of the lignin heteropolymer. Here, we tested three sphingomonads predicted to be capable of breaking the β-aryl ether bond of the dimeric aromatic compound guaiacylglycerol-β-guaiacyl ether (GGE) and found that Novosphingobium aromaticivorans metabolizes GGE at one of the fastest rates thus far reported. After the ether bond of racemic GGE is broken by replacement with a thioether bond involving glutathione, the glutathione moiety must be removed from the resulting two stereoisomers of the phenylpropanoid conjugate β-glutathionyl-γ-hydroxypropiovanillone (GS-HPV). We found that the Nu-class glutathione S-transferase NaGSTNu is the only enzyme needed to remove glutathione from both (R)- and (S)-GS-HPV in N. aromaticivorans. We solved the crystal structure of NaGSTNu and used molecular modeling to propose a mechanism for the glutathione lyase (deglutathionylation) reaction in which an enzyme-stabilized glutathione thiolate attacks the thioether bond of GS-HPV, and the reaction proceeds through an enzyme-stabilized enolate intermediate. Three residues implicated in the proposed mechanism (Thr51, Tyr166, and Tyr224) were found to be critical for the lyase reaction. We also found that Nu-class GSTs from Sphingobium sp. SYK-6 (which can also break the β-aryl ether bond) and Escherichia coli (which cannot break the β-aryl ether bond) can also cleave (R)- and (S)-GS-HPV, suggesting that glutathione lyase activity may be common throughout this widespread but largely uncharacterized class of glutathione S-transferases.

The most distinguishing characteristic of Nu-class (Main.2 (16)) GSTs is their apparent ability to bind two molecules of GSH and/or a single molecule of GSSG (16 -20). Nu-class GSTs are found in many organisms (16,18), but their physiological roles are largely unknown. Several Nu-class GSTs have been tested for activity in vitro; most, including EcYfcG and EcYghU from Escherichia coli, showed disulfide bond reductase activity toward small molecules, such as 2-hydroxyethyl disulfide (16 -18, 21), which led to the proposal that this is their main physiological activity. A strain of E. coli in which the yfcG gene was deleted was impaired in its response to oxidative stress, leading to the proposal that EcYfcG may naturally function as a peroxidase, although this enzyme exhibited low in vitro peroxidase activity with model peroxides (22). At least one other Nu-class GST besides GST3 has been found to have glutathione lyase activity in vitro: PcUre2pB1 from the white-rot fungus Phanerochaete chrysosporium can cleave GS-phenacylacetophenone, which led to the suggestion that this enzyme may naturally function as a glutathione lyase (19). To investigate the prevalence of glutathione lyase activity throughout the Nu class, we assayed EcYfcG and EcYghU and found that each cleaves both ␤(R)and ␤(S)-GS-HPV in vitro, although less efficiently than NaGST Nu . Furthermore, EcYghU complements growth of an N. aromaticivorans ⌬NaGST Nu mutant, showing that it can perform these reactions in vivo. Our results thus suggest that glutathione lyase activity may be common throughout the Nu-class, and the lower catalytic efficiencies we find with EcYghU and EcYfcG toward GS-HPV may reflect the fact that this compound is not a natural substrate for E. coli.

GGE metabolism by sphingomonads
To test for the ␤-etherase pathway in the sphingomonad bacteria N. aromaticivorans DSM 12444, Novosphingobium sp. PP1Y, and S. xenophagum NBRC 107872, we fed them erythro-GGE ( Fig. 1), both alone and in the presence of another organic molecule, in case any of the strains can break the ␤-aryl ether bond of GGE but cannot use it to support growth, as is the case for Novosphingobium sp. MBES04 (6).
For these studies, we created a strain of N. aromaticivorans in which the Saro_1879 (putative sacB) gene has been deleted (12444⌬1879), so that we could make additional modifications to its genome using a sacB-containing cloning vector. When fed erythro-GGE alone, N. aromaticivorans 12444⌬1879 completely removed the compound from the medium (Fig. 2 (A and  B) and Fig. S1) and incorporated ϳ22% of its organic material (measured as chemical oxygen demand (COD)) into biomass (Table S1). The majority of the electrons from the erythro-GGE (ϳ53%) were probably combined with oxygen to support respiration during growth with the compound (Table S1). When fed erythro-GGE plus vanillate, 12444⌬1879 completely removed both substrates from the medium (Fig. 2, G and H) and incorporated ϳ41% of the COD from the vanillate (based on the results from a culture fed vanillate alone) and ϳ22% of the COD from the GGE into biomass (Table S1). The known ␤-etherase pathway intermediates threo-GGE, MPHPV, and HPV transiently appeared in the media of both erythro-GGE-fed 12444⌬1879 cultures (Fig. 2 (B and H) and Fig. S1), whereas the pathway intermediate guaiacol was only detected at a low level in the medium of the culture fed erythro-GGE plus vanillate (Fig. 2H). The predicted pathway intermediate GS-HPV was not detected in the medium of either culture. Wild-type N. aromaticivorans DSM 12444 behaved similarly under both of these growth conditions, justifying our use of 12444⌬1879 as a parent strain in this study.
When fed only erythro-GGE, Novosphingobium sp. PP1Y slowly converted most of it into ␤-etherase pathway intermediates but did not completely remove these compounds from the medium and only achieved a low cell density (Fig. S2, A and B). When fed erythro-GGE plus vanillate, Novosphingobium sp. PP1Y rapidly consumed the vanillate, but the maximum cell density and amount of COD incorporated into biomass in this culture were the same as when the strain was fed vanillate only ( Fig. S2E and Table S1), suggesting that the strain did not convert a significant amount of GGE into biomass in the presence of vanillate. S. xenophagum converted some of the erythro-GGE it was supplied into ␤-etherase pathway intermediates (Fig. S2, D and H) but did not assimilate organic material from it into cellular biomass in any culture tested: when fed erythro-GGE alone, no cell density was detected, and when fed erythro-GGE plus glucose (we found this strain to be unable to metabolize vanillate), the maximum cell density and amount of COD incorporated into biomass were the same as when the strain was fed glucose only ( Fig. S2 (C and G) and Table S1). These results suggest that N. aromaticivorans is the best of the three organisms we analyzed for studying various aspects of the ␤-etherase pathway.

Deglutathionylation of a GS-phenylpropanoid by Nu-class GSTs Transcripts of predicted N. aromaticivorans ␤-etherase pathway genes increase in abundance in the presence of GGE
To test the transcriptional response of N. aromaticivorans to GGE, we investigated expression levels of several of its genes predicted to code for enzymes of the ␤-etherase pathway, including Saro_2595, which encodes a Nu-class glutathione S-transferase (named here NaGST Nu ). With the exception of ligL, transcript levels from these genes were higher (by Ͼ6-fold) in cells grown in the presence of GGE versus its absence (Table 1).

NaGST Nu is necessary for GGE metabolism by N. aromaticivorans
To test for an in vivo role for NaGST Nu , we generated an N. aromaticivorans strain (12444⌬2595) lacking Saro_2595, the gene that encodes NaGST Nu . When 12444⌬2595 was pro-vided only erythro-GGE, a small amount of MPHPV and threo-GGE appeared in the medium (Fig. 2D), but no cell density was detected (Fig. 2C), and no detectable COD was converted into biomass (Table S1). When 12444⌬2595 was fed both vanillate and erythro-GGE, all of the vanillate was consumed, and almost all of the GGE was converted into MPHPV (Fig. 2J). The maximum cell density and amount of COD incorporated into biomass were the same for this culture as for a culture fed vanillate only ( Fig. 2I and Table S1), suggesting that 12444⌬2595 cannot convert GGE into cell material. A small amount of guaiacol appeared in the medium of the 12444⌬2595 culture fed vanillate and erythro-GGE (Fig. 2J), showing that this strain cleaved some MPHPV. However, unlike for its parent strain (12444⌬1879), no extracellular HPV was detected in any erythro-GGE-fed 12444⌬2595 culture (Fig. 2, D and J). These results show that NaGST Nu is necessary for complete GGE metabolism by N. aromaticivorans.

NaGST Nu is sufficient and necessary for conversion of GS-HPV into HPV in N. aromaticivorans
To determine which step in the ␤-etherase pathway requires NaGST Nu , we incubated cell extracts of 12444⌬2595 and its parent strain (12444⌬1879) with racemic MPHPV and GSH. With the 12444⌬1879 extract, MPHPV was completely converted to roughly equimolar amounts of guaiacol and HPV, along with a small amount of GS-HPV (Fig. 3A). In contrast, the 12444⌬2595 extract incompletely cleaved the MPHPV, producing roughly equimolar amounts of guaiacol and GS-HPV (along with a low level of HPV, ϳ2% of the level of GS-HPV

Deglutathionylation of a GS-phenylpropanoid by Nu-class GSTs
formed; Fig. 3B). Thus, the 12444⌬2595 extract was defective in converting GS-HPV into HPV. When recombinant NaGST Nu was added to the 12444⌬2595 extract after 1 day of incubation with MPHPV (Fig. 3B), the initially accumulated GS-HPV rapidly disappeared, with a concomitant increase in HPV, showing that the defect in GS-HPV cleavage by the 12444⌬2595 extract was caused by the lack of NaGST Nu .

Structural characterization of NaGST Nu
To gain insight into the mechanism of GS-HPV cleavage by NaGST Nu , we solved two structures of the enzyme with resolutions of 1.25 (PDB code 5UUO) and 1.45 (PDB code 5UUN) Å ( Table 3). The structures align with each other with an r.m.s. distance of 0.108 Å over 7381 atoms. NaGST Nu is a homodimer; each subunit contains a characteristic N-terminal GST1 (thi-  (Table S4). a Gene product from Sphingobium sp. SYK-6 (SLG) or Novosphingobium sp. MBES04 (MBENS4) that has been shown to catalyze a reaction in the ␤-etherase pathway and that was used in a BLASTp search to identify the indicated N. aromaticivorans (Saro) gene. Bit score and percentage identity from the BLASTp searches are shown in parentheses. b -Fold change is the ratio of the normalized transcript level in cells grown in the presence of GGE to that in cells grown in the absence of GGE for the N. aromaticivorans genes. c Only Saro_2405 showed high homology (Bit score Ͼ100) to both SLG_08660 (LigE) and SLG_32600 (LigP).

Table 2
Kinetic parameters for the enzymatic conversion of GS-HPV into HPV Kinetic parameters are from non-linear least squares best fits to plots of initial rate ([HPV] formed/s) versus GS-HPV concentration using the Michaelis-Menten equation (Fig. S5).

GS-HPV
0.017 Ϯ 0.004 130 Ϯ 40 0.14 Ϯ 0.06 a Stereoisomer of GS-HPV used in reaction (see Fig. 1). Structures of Nu-class GSTs solved in the presence of GSH typically contain either two GSH molecules (e.g. PDB codes 3C8E and 4IKH) or a single GSSG molecule (e.g. PDB codes 4MZW and 3GX0) in each active site. Consistent with this, each NaGST Nu subunit contained electron density data that were best modeled as a mixed population of GSH1 and GSH2 thiols with an HS-SH distance of 2.4 Å (ϳ60% occupancy) and a GSSG disulfide with S-S distance of 2.0 Å (ϳ40% occupancy) ( Fig. 4B and Fig. S6), suggesting a heterogeneous population of protein molecules in the crystals in which each subunit contained either two GSHs or a single GSSG (see legend to Fig. S6).

Deglutathionylation of a GS-phenylpropanoid by Nu-class GSTs
In NaGST Nu , seven residues make close contacts with GSH1 (Thr 51 , Asn 53 , Gln 86 , Lys 99 , Ile 100 , Glu 116 , and Ser 117 ), and three residues make close contacts with GSH2 (Asn 25 , Asn 53 , and Arg 177 from the opposite chain in the dimer) (Fig. 4B). These residues and contacts with the GSHs are conserved throughout much of the Nu class ( Fig. S3) (16,19,20), although members of this family with truncated N termini lack an analogue of Asn 25 ( Fig. S3; see supplemental materials of Ref. 19).
In NaGST Nu , a short channel connects the active site to the solvent ( Fig. 5 and Fig. S7). This channel is also present in Nuclass GSTs from E. coli (EcYghU; PDB code 3C8E (18)) and Streptococcus sanguinis SK36 (SsYghU; PDB code 4MZW). In most other structurally characterized Nu-class GSTs, the active site is more solvent-exposed, because these proteins lack N-terminal residues that contribute to one of the channel walls ( Fig.  S3; see supplemental materials of Ref. 19).
In each NaGST Nu homodimer we solved (PDB codes 5UUN and 5UUO), the individual monomers differ in the positioning of residues Gln 282 -Phe 288 (Fig. 5, A-C). The C-terminal Phe 288 resides near the active-site channel in one subunit (NaGST Nu closed; Fig. 5B) and ϳ18 Å away from the channel entrance in the other (NaGST Nu open; Fig. 5C), resulting in a difference between the monomers in the opening to the active-site pocket. The closed NaGST Nu configuration is stabilized by hydrogen bonds between Lys 286 and both Lys 262 and Arg 220 . The open configuration lacks these interactions, and its Arg 220 side chain has two rotamer positions. In other structurally characterized Nu-class GST homodimers, the individual subunits are symmetric, with C termini that generally extend away from the active site (alignments for EcYghU and SsYghU subunits are shown in Fig. 5 (A, D, and E)).

Modeling of substrate binding and proposed reaction mechanism
We separately modeled the GS-moiety of ␤(R)and ␤(S)-GS-HPV into the GSH2 active-site position (Fig. 4B) of the closed C-terminal configuration subunit of NaGST Nu and found that the HPV moieties extend into the active-site channel in different orientations without generating unfavorable steric clashes  (Fig. 6, A and B). For ␤(R)-GS-HPV, the ␥-hydroxyl is also predicted to hydrogen-bond with the ␣-ketone, which in turn is predicted to hydrogen-bond with the hydroxyl of Tyr 166 (Fig. 6A). For ␤(S)-GS-HPV, the hydroxyl of Tyr 166 is predicted to form a hydrogen bond with the HPV aromatic ring (24) and a long-range interaction with the HPV ␣-ketone (Fig. 6B).
We propose that the thiol of a GSH molecule in the GSH1 active-site position of NaGST Nu is activated by hydrogenbonding with the side-chain hydroxyl of Thr 51 (3.0 Å) and the side-chain amide of Asn 53 (3.2 Å) (Figs. 4B and 7 (A and B)). These residues are part of a Thr-Pro-Asn motif that is highly conserved throughout the Nu class, the Asn residue of which was found to be critical for the glutathione lyase activity of PcUre2pB1 (19). Interactions with active-site residues have been shown to lower the pK a of a GSH thiol and stabilize the reactive thiolate anion in members of other GST classes (e.g. the GSH thiol pK a is lowered to ϳ6.6 by hydrogen bonds with Tyr and Arg side chains in human GSTA3-3 (25)). In addition, the Thr 51 hydroxyl closely contacts the backbone amides of Asn 53 and Gly 54 , and the side-chain amide of Asn 53 closely contacts the side-chain amine of Lys 56 and the Gly carboxylate of the GSmoiety of GS-HPV (Fig. S8). Extended hydrogen-bonding networks ("second-sphere interactions") such as this are proposed to contribute to GSH thiolate stabilization in other GST classes (26).
In our proposed reaction mechanism for NaGST Nu , the GS1 thiolate attacks the thioether of GS-HPV to form a disulfide GS-SG. Because the C-S bond lyase reaction proceeds in the absence of a redox cofactor, such as flavin or pyridine nucleotide, we propose that a transient enolate (Fig. 7, C and D) stores the 2e Ϫ reducing equivalents released by disulfide bond formation. Formation of this enolate is proposed to be facilitated by polarization of the GS-HPV ␣-ketone, which is stabilized by the  This proposed mechanism for NaGST Nu , in which the activated thiol of a glutathione molecule is used to break the thioether bond of GS-HPV, is different from that proposed for the Omega-class GST LigG, in which the activated thiol of a Cys residue in the enzyme breaks the thioether bond (27). In contrast to NaGST Nu , LigG is not known or predicted to be able to bind two glutathione molecules in a single active site (27,28). NaGST Nu contains a single Cys residue (Fig. S3), whose thiol is Ͼ20 Å away from either GSH thiol, and so is unlikely to be involved in the lyase reaction.

Single-amino acid variants of NaGST Nu are catalytically impaired
To test the plausibility of our proposed reaction mechanism, three residues in NaGST Nu predicted to be important for catalysis were separately mutated: Thr 51 into Ala (T51A), Tyr 166 into Phe (Y166F), and Tyr 224 into Phe (Y224F). All three mutated enzymes had decreased reactivity with GS-HPV compared with wildtype NaGST Nu (Fig. S5 (B-D) and Table 2). The T51A variant had K m values similar to wildtype NaGST Nu but k cat values that were ϳ1000-fold lower, consistent with our proposal that Thr 51 is primarily involved in stabilizing the reactive GSH1 thiolate and not in binding GS-HPV. Compared with wildtype NaGST Nu , the Y166F variant had k cat values that were ϳ1000and ϳ400-fold lower and K m values that were ϳ2.5and ϳ15-fold higher for the ␤(R)and ␤(S)-stereoisomers of GS-HPV, respectively. These results suggest that the Tyr 166 hydroxyl is critical for catalysis with both GS-HPV stereoisomers and is important for binding of ␤(S)-GS-HPV. Compared with wildtype NaGST Nu , the Y224F variant had k cat values that were ϳ40and ϳ60-fold lower and K m values that were ϳ7.5and ϳ6-fold higher for the ␤(R)and ␤(S)-stereoisomers of GS-HPV, respectively. These results implicate the Tyr 224 hydroxyl in both binding and catalysis, although it is less important to catalysis than that of Tyr 166 .

Modeling GS-conjugated syringyl phenylpropanoids into the NaGST Nu active site
Besides GS-HPV, depolymerization of natural lignin oligomers via the ␤-etherase pathway is expected to produce syringyl phenylpropanoids (3). Modeling of the ␤(R)and ␤(S)-isomers of a GS-conjugated syringyl phenylpropanoid into the NaG-ST Nu active site predicts that they should bind in essentially the same orientations as the corresponding GS-HPV isomers, because the active-site channel can accommodate each syringyl conjugate's additional methoxy group (Fig. S7, C and D). Thus, Nu-class GSTs may be the only enzymes necessary for the glu- Deglutathionylation of a GS-phenylpropanoid by Nu-class GSTs tathione lyase step in deconstructing natural lignin oligomers via the ␤-etherase pathway.

Nu-class GSTs from E. coli can cleave ␤(R)and ␤(S)-GS-HPV
To investigate the prevalence of glutathione lyase activity throughout the GST Nu-class, we tested two Nu-class GSTs from E. coli (EcYghU and EcYfcG), an organism not known to metabolize lignin-derived glutathione conjugates. We found that recombinant EcYghU cleaves ␤(R)and ␤(S)-GS-HPV in vitro (Fig. S5F); whereas its K m values were comparable with those of NaGST Nu , its k cat values were much lower, resulting in k cat /K m values for EcYghU ϳ100-fold lower than those of NaG-ST Nu ( Table 2). To investigate the cause of these differences in catalytic efficiency, we modeled ␤(R)and ␤(S)-GS-HPV into the active site of EcYghU (PDB code 3C8E). Whereas EcYghU is similar to NaGST Nu in sequence (61% identical, including analogues of catalytic residues Thr 51 , Tyr 166 , and Tyr 224 ; Fig. S3) and structure (r.m.s. deviation of 0.49 Å over 3116 atoms; Fig.  5), there are steric differences in the interior of the active-site channel that lead to different predicted orientations of the bound substrates between the enzymes (Fig. 6 and Fig. S7). The hydroxyls of Tyr 167 and Tyr 225 in EcYghU (analogues of Tyr 166 and Tyr 224 in NaGST Nu ) are predicted to interact with the HPV moiety of both ␤(R)and ␤(S)-GS-HPV (Fig. 6, C and D). For ␤(R)-GS-HPV, Tyr 225 is predicted to form hydrogen bonds with both the ␥-hydroxyl and ␣-ketone, and Tyr 167 is predicted to form hydrogen bonds with the ␣-ketone and the aromatic ring (Fig. 6C). For ␤(S)-GS-HPV, Tyr 225 is predicted to form a hydrogen bond with the ␥-hydroxyl, and Tyr 167 is predicted to form a hydrogen bond with the aromatic ring and a long-range interaction with the ␥-hydroxyl (Fig. 6D). These interactions could help stabilize an enolate intermediate in the reactions between EcYghU and GS-HPV, so EcYghU could follow a reaction mechanism similar to that proposed for NaGST Nu (Fig. 7). However, the reactive atoms (S-C ␤ -(C ␣ ϭO)-aryl) of both ␤(R)and ␤(S)-GS-HPV bound to EcYghU are predicted to be ϳ45°o ut of alignment ( Fig. 6 (C and D) and Fig. S7 (E and F)), which could hinder formation of the enolate and lead to the lower k cat values for EcYghU than for NaGST Nu (for which these reactive atoms were in a roughly planar configuration (Fig. 6 (A and B) and Fig. S7 (A and B)).
Despite these lower in vitro k cat values, we found that EcYghU was able to substitute for NaGST Nu in the ␤-etherase pathway in vivo: a strain of N. aromaticivorans (12444EcyghU) in which Saro_2595 was replaced in the genome by the E. coli yghU gene completely removed the GGE from the media of both of its cultures and assimilated the GGE into biomass (Fig.  2 (E, F, K, and L) and Table S1), whereas 12444⌬2595 (which lacks NaGST Nu ) could not (Fig. 2, C, D, I, and J). 12444EcyghU metabolized GGE slower than 12444⌬1879 (which contains NaGST Nu ) (Fig. 2), which we propose reflects the slower conversion of GS-HPV into HPV in 12444EcyghU than in 12444⌬1879, due to the lower catalytic efficiencies of EcYghU than NaGST Nu . , contacts between active-site residues and the GSH1 and GSH2 dithiol (60% occupancy; orange and cyan carbon atoms, respectively) and the GS-SG disulfide (40%; gray carbon atoms). NaGST Nu residues are colored according to domain origin in A. Interactions involving protein residues are shown in black; those between GSH1 and GSH2 are in silver. Selected distances between interacting atoms are shown.

Deglutathionylation of a GS-phenylpropanoid by Nu-class GSTs
The other E. coli Nu-class GST we tested, EcYfcG, is less similar to NaGST Nu in sequence (42% identical; Figs. S3 and S9) and structure (PDB code 3GX0; r.m.s. deviation of 0.775 Å over 2016 atoms and lacking an enclosed active-site channel (17)). Whereas we found that recombinant EcYfcG cleaved both ␤(R)and ␤(S)-GS-HPV in vitro (Fig. S5G), K m values were higher and k cat values were much lower than those of NaGST Nu , leading to k cat /K m values with GS-HPV ϳ10,000-fold lower for EcYfcG than for NaGST Nu ( Table 2). The lower catalytic efficiencies of EcYfcG are probably due, at least in part, to the absence of analogues of NaGST Nu residues Tyr 166 and Tyr 224 (Fig. S3), which were found to be important for catalysis and substrate binding in NaGST Nu . Given the dramatically lower catalytic efficiencies of EcYfcG, it is unlikely that EcYfcG could substitute for NaGST Nu in N. aromaticivorans.

Discussion
In developing bio-based systems to generate products from renewable materials such as lignin, optimized cellular and enzyme catalysts are needed. Toward this end, we tested sphingomonads for the ability to break the ␤-aryl ether bond commonly found in lignin (Fig. 1) and identified and characterized a Nu-class glutathione S-transferase from N. aromaticivorans that acts as a glutathione lyase in the process (NaGST Nu ).

Differences in ␤-aryl ether bond breaking by sphingomonads
We found that N. aromaticivorans was the most effective species studied here at breaking the ␤-aryl ether bond of the dimeric aromatic compound GGE and assimilating GGE into cellular biomass. The rate of GGE ␤-aryl ether bond breaking by N. aromaticivorans (ϳ165 M in ϳ25 h; Fig. S1) is comparable with rates for Erythrobacter sp. SG61-1L (ϳ180 M in ϳ74 h (5)) and Novosphingobium sp. MBES04 (ϳ700 M in ϳ40 h (6)). Sphingobium sp. SYK-6 (ϳ180 M in ϳ160 h (5)), Novosphingobium sp. PP1Y (Fig. S2), and S. xenophagum (Fig.  S2) are slower and/or less efficient at breaking the ␤-aryl ether bond of GGE under laboratory conditions, although they each contain enzymes implicated in the ␤-etherase pathway. Understanding the bases for these differences can aid future efforts to develop microbial systems for converting lignocellulosic biomass into commodities.

GGE metabolism by N. aromaticivorans
The transient appearance of extracellular MPHPV, threo-GGE, and HPV in N. aromaticivorans cultures fed erythro-GGE suggests that the bacterium excreted these metabolites and then subsequently transported them back into the cell (Fig. 2). The production of threo-GGE in cultures fed erythro-GGE suggests that MPHPV reduction occurs in vivo, as was previously found for Pseudomonas acidovorans D3 (29), because the two forms of GGE are not directly interconvertible. The low level of extracellular guaiacol and the absence of extracellular GS-HPV suggest that MPHPV cleavage occurred intracellularly, as was proposed for Novosphingobium sp. MBES04 (6), and that the products of this cleavage were largely retained within the cells. Because we did not observe any additional aromatic compounds in our culture media, any aromatic metabolites downstream of HPV and guaiacol were also probably largely retained within the cells.
Because NaGST Nu catalyzes the conversion of GS-HPV into HPV, the accumulation of extracellular MPHPV by N. aromati-

Deglutathionylation of a GS-phenylpropanoid by Nu-class GSTs
civorans 12444⌬2595 fed erythro-GGE and vanillate (Fig. 2J) was unexpected. Cleavage of MPHPV into guaiacol and GS-HPV is catalyzed by LigF and LigE (Fig. 1), enzymes that are probably expressed in 12444⌬2595, because crude extract from this strain can cleave MPHPV (Fig. 3B). We hypothesize that without NaGST Nu , GS-HPV accumulates in 12444⌬2595, and cells become limited for the free GSH that is needed to cleave MPHPV.
It is unclear whether the trace amount of HPV formed in assays using 12444⌬2595 extract (Fig. 3B) resulted from activity of an unknown enzyme or spontaneous cleavage of GS-HPV. Even if there were another enzyme in N. aromaticivorans with low glutathione lyase activity, our results with the 12444⌬2595 strain and its extract show that NaGST Nu is sufficient for cleavage of GS-HPV and necessary for complete metabolism of GGE by N. aromaticivorans.

The role of Nu-class GSTs in the ␤-etherase pathway
This work demonstrates that a Nu-class GST is probably the only enzyme necessary for metabolizing the ␤(R)and ␤(S)isomers of lignin-derived GS-phenylpropanoids (both guaiacyl (HPV) and syringyl) in nature. This raises the question of why some sphingomonads (such as Sphingobium sp. SYK-6) contain both a Nu-class GST and LigG, an Omega-class GST that is reported to be specific for the ␤(R)-isomer (Fig. 1) (13). Indeed, cell extract from a Sphingobium sp. SYK-6 ⌬LigG mutant was found to completely cleave racemic GS-HPV, presumably because the lysate contained active SYK6GST Nu (13). Notably, we found that the k cat /K m value for SYK6GST Nu with ␤(R)-GS-HPV (ϳ240 mM Ϫ1 s Ϫ1 ; Table 2) is ϳ7-fold lower than that reported for LigG with ␤(R)-GS-HPV (ϳ1700 mM Ϫ1 s Ϫ1 (28)). Thus, whereas SYK6GST Nu probably cleaves ␤(S)-GS-HPV in Sphingobium sp. SYK-6, LigG may play an important role in cleaving ␤(R)-GS-HPV in that organism.
Novosphingobium sp. MBES04 contains not only a LigG homologue (GST6, which preferentially reacts with ␤(R)-GS-HPV (6)), but also two Nu-class GSTs (GST3 and a protein encoded by MBENS4_4395). Whereas GST3 cleaves ␤(R)and ␤(S)-GS-HPV in vitro (6), its catalytic efficiencies and physiological role have not been reported. No investigation of MBENS4_4395 has been reported, but it is 66% identical to NaGST Nu in amino acid sequence and contains analogues of catalytic residues Tyr 166 and Tyr 224 (Fig. S3) and so can probably react with ␤(R)and ␤(S)-GS-HPV. Whereas the relative roles of these enzymes in Novosphingobium sp. MBES04 are unknown, given the effect of deleting NaGST Nu on N. aromaticivorans (its only known glutathione lyase) (Fig. 2, C, D, I, and J), perhaps they represent redundant enzymes or are specialized for different GS-conjugates.

The potential roles of Nu-class GSTs as glutathione lyases in organisms that do not contain the ␤-etherase pathway
All five of the Nu-class GSTs that have been tested for GS-HPV cleavage (NaGST Nu , SYK6GST Nu , EcYghU, and EcYfcG (tested here), and GST3 (6)) show in vitro glutathione lyase (deglutathionylation) activity with the ␤(R)and ␤(S)-stereoisomers of this substrate, although with a wide range of catalytic efficiencies. Phylogenetic analysis of Nu-class GSTs shows that these five enzymes lie in separate subclades (Fig. S9), although the residues predicted to be involved in the binding of two GSH molecules (or one GSH molecule and the glutathione moiety of a GS-conjugate) (Fig. 4B) and in activating the thiol of the GSH1 molecule (Thr 51 and Asn 53 ) are conserved in many members of the class (Fig. S3) (16,19,20). Another Nu-class GST, PcUre2pB1 from P. chrysosporium, has also been shown to exhibit in vitro glutathione lyase activity (19). Glutathione lyase activity may thus be widespread throughout this large class of GSTs, most of which are found in organisms (like E. coli) not known or predicted to break the ␤-aryl ether bond of lignin. Whereas EcYghU and EcYfcG cleave ␤(R)and ␤(S)-GS-HPV in vitro with lower catalytic efficiencies than NaGST Nu and SYK6GST Nu (Table 2), this may reflect the fact that GS-HPV is not a natural substrate for the E. coli enzymes. Although the overall structures of Nu-class GSTs are similar, differences between them, particularly in the region surrounding the active site where the GS-conjugate would be located, could make individual enzymes optimally suited for binding and cleaving GSconjugates that their respective organisms encounter more frequently. For example, we propose that steric differences in the  Fig. S7 shows space-filling models of the enzyme-substrate complexes.

Deglutathionylation of a GS-phenylpropanoid by Nu-class GSTs
active-site channels between NaGST Nu and EcYghU contribute to the observed differences in catalytic efficiencies between these enzymes with GS-HPV as substrate ( Fig. 6 and Fig. S7).

Conclusions
The rate of GGE metabolism by N. aromaticivorans makes it an attractive organism for studying various aspects of the ␤-etherase pathway and may allow it to be developed into a biological system for converting lignin oligomers into useful compounds. Our finding that NaGST Nu is the only enzyme needed for the glutathione lyase step of the ␤-etherase pathway is notable, because the other pathway steps require multiple stereospecific enzymes (Fig. 1). In addition, this is the first demonstration of a Nu-class GST having a defined physiological role as a glutathione lyase in an organism. Our analyses of wildtype NaGST Nu and its variants define roles for residues conserved throughout the Nu class in GS-HPV cleavage. Our identification of SYK6GST Nu probably solves the enigma of how both ␤(R)and ␤(S)-GS-HPV are cleaved in Sphingobium sp. SYK-6: based on kinetic analyses, we propose that SYK6GST Nu cleaves ␤(S)-GS-HPV, whereas LigG and SYK6GST Nu both contribute to cleaving ␤(R)-GS-HPV in that organism.
Finally, the ability of EcYghU and EcYfcG to cleave GS-HPV shows that Nu-class GSTs from organisms lacking the ␤-etherase pathway can also act as glutathione lyases, offering new possibilities for the physiological roles of members of this large enzyme class.

Bacterial strains and growth media
Strains used are listed in Table S2. We deleted Saro_1879 (putative sacB; SARO_RS09410 in the recently reannotated genome in NCBI) from the N. aromaticivorans DSM 12444 genome to create a strain (12444⌬1879) amenable to markerless genomic modifications using a variant of pK18mobsacB (30), which contains sacB and a kanamycin resistance gene. We used 12444⌬1879 as parent strain to generate strains in which Saro_2595 (SARO_RS13080 in the recently reannotated genome in NCBI) was deleted from the genome (12444⌬2595) and in which Saro_2595 was replaced in the genome by the E. coli DH5␣ yghU gene (12444EcyghU). Detailed methods for modifying the genome of N. aromaticivorans are described in the supporting information. All plasmids and primers used for

Deglutathionylation of a GS-phenylpropanoid by Nu-class GSTs
genomic modifications are contained in Tables S2 and S3. Fig.  S10 shows genotypes for 12444⌬2595 and 12444EcyghU.
Unless otherwise noted, E. coli cultures were grown in lysogeny broth and shaken at ϳ200 rpm at 37°C. For routine storage and manipulation, sphingomonad cultures were grown in lysogeny broth or GluSis at 30°C. GluSis is a modification of Sistrom's minimal medium (31) in which the succinate has been replaced by 22.6 mM glucose. Standard mineral base (SMB) (32) used for growth experiments contains 20 mM Na 2 HPO 4 , 20 mM KH 2 PO 4 , 1 g/liter (NH 4 ) 2 SO 4 , and 20 ml of Hutner's vitaminfree concentrated base (adapted from Ref. 33 but lacking nicotinic acid, thiamin, and biotin; see supporting information) per liter, final pH 6.8. SMB was supplemented with carbon sources as described below. Where needed to select for plasmids, media were supplemented with 50 g/ml kanamycin and/or 20 g/ml chloramphenicol.

Sphingomonad growth experiments
Cell densities were measured using a Klett-Summerson photoelectric colorimeter with a red filter. For N. aromaticivorans, 1 Klett unit (KU) is equal to ϳ8 ϫ 10 6 cfu/ml (Table S4). Experimental cultures of N. aromaticivorans and Novosphingobium sp. PP1Y were grown in SMB containing either vanillate or GGE alone (4 and 3 mM, respectively) or a combination of vanillate and GGE (4 and 1.5 mM, respectively). For S. xenophagum cultures, vanillate was replaced by glucose, because we found this strain to be unable to metabolize vanillate. N. aromaticivorans was also grown in SMB containing 165 M GGE. Starter cultures were grown in SMB with 4 mM vanillate or glucose, and cells were pelleted and washed with PBS (10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , 137 mM NaCl, 2.7 mM KCl, pH 7.4). Pellets were resuspended into culture medium and used to inoculate experimental cultures to initial cell densities of Ͻ5 KU.
Cultures were grown aerobically at 30°C, in 125-ml conical flasks containing 20 -40 ml of medium and shaken at ϳ200 rpm. Aliquots (400 -600 l) were removed at specified time points and filtered through 0.22-m syringe tip filters (e.g. Whatman Puradisc filters, GE Healthcare) before HPLC analysis of extracellular aromatics. Every culture was grown at least three times; data shown are from representative cultures.

Enzyme purifications
Saro_2595 from N. aromaticivorans, yghU and yfcG from E. coli DH5␣, and SLG_04120 (SLG_RS02030 in the recently reannotated genome in NCBI) from Sphingobium sp. SYK-6 (codon-optimized for expression in E. coli) were individually cloned into plasmid pVP302K (34) so that transcripts from the plasmids would be translated into proteins containing a His 8 tag connected to the N terminus via a tobacco etch virus (TEV) protease recognition site (construction of plasmids described in the supporting information). The plasmid containing Saro_ 2595 was modified via PCR to generate plasmids for expressing variants of NaGST Nu containing the single-residue mutations T51A, Y166F, and Y224F. Recombinant proteins were expressed in E. coli B834 (35,36) containing plasmid pRARE2 (Novagen, Madison, WI) grown for ϳ25 h at 25°C in ZYM-5052 autoinduction medium (37) containing kanamycin and chloramphenicol. Recombinant proteins were purified as described previously (15) (see supporting information for modifications to the procedure). After removal of His 8 tags using TEV protease, recombinant proteins retained a Ser-Ala-Ile-Ala-Gly-peptide on their N termini, derived from the linker between the protein and the TEV protease recognition site. Recombinant LigE and LigF1 from N. aromaticivorans were purified as described previously (34).

Kinetics of converting GS-HPV into HPV
The reaction buffer (RB) consisted of 25 mM Tris-HCl (pH 8.5) and 25 mM NaCl. The ␤(R)and ␤(S)-stereoisomers of GS-HPV were separately generated by incubating racemic ␤(S)and ␤(R)-MPHPV (0.46 mM) in RB with 5 mM GSH and either 38 g/ml LigF1 or 36 g/ml LigE for several hours (Fig. S4). This sample, containing a single GS-HPV stereoisomer, guaiacol, and the unreacted MPHPV stereoisomer (as well as LigE or F1), was diluted with RB to achieve the desired concentration of GS-HPV for the time course reaction (0.005, 0.01, 0.02, 0.1, or 0.2 mM). An additional 5 mM GSH (dissolved in RB) was added before initiation of each time course. At time 0, 100 l of the indicated enzyme (resuspended in RB) was combined with 1800 l of the diluted GS-HPV reaction mixture to achieve final concentrations of 8 nM NaGST Nu , 20 nM NaGST Nu (T51A), 100 nM NaGST Nu (Y166F), 10 nM NaGST Nu (Y224F), 195 nM EcYghU, 195 nM EcYfcG, or 47 or 18 nM SYK6GST Nu (for the ␤(R)and ␤(S)-GS-HPV reactions, respectively). Reactions were performed at 25°C. At specified time points, 300 l of the reaction was removed and combined with 100 l of 1 M HCl (Acros Organics, Geel, Belgium) to stop the reaction before HPLC analysis to quantify HPV formed.

N. aromaticivorans cell extract assays
N. aromaticivorans cells were grown in 500-ml conical flasks containing 267 ml of SMB medium with 4 mM vanillate and 1 mM GGE. When cell densities reached ϳ1 ϫ 10 9 cells/ml, cells were lysed by the sonication procedure used to generate E. coli lysates for protein purification (see supporting information). Samples were centrifuged at 7000 ϫ g for 15 min, and the supernatants were used as crude cell extracts.
Assays containing 50 mM Tris-HCl (pH 8.0), 50 mM NaCl, 10 mM GSH, 0.407 mM racemic ␤(R)and ␤(S)-MPHPV and cellular extract from 12444⌬1879 or 12444⌬2595 (final concentrations of 269 and 186 g of protein/ml, respectively) were performed at 30°C. At defined time points, 300-l aliquots were combined with 100 l of 1 M HCl to stop the reaction before HPLC analysis. At the indicated time, recombinant NaGST Nu (30 g of protein/ml) was added to the 12444⌬2595 cell extract reaction, along with an additional 10 mM GSH.

HPLC analysis
After extracellular aromatics were identified using LC-MS with a PFP column (see supporting information), routine analysis and quantification of aromatics were performed using an Ultra AQ C18 5-m column (Restek) attached to a System Gold HPLC system (Beckman Coulter) with running buffers described in Fig. S11A. The eluent was analyzed for light absorbance between 191 and 600 nm, and absorbances

Deglutathionylation of a GS-phenylpropanoid by Nu-class GSTs
at 280 nm were used for quantification of aromatic metabolites by comparing peak areas with those of standards (retention times of measured metabolites are shown in Figs. S4 and S11B).

Production of cDNA libraries from N. aromaticivorans cultures and real-time quantitative PCR
N. aromaticivorans cultures were grown in 120 ml of SMB containing either 4 mM vanillate or 4 mM vanillate and 1 mM GGE. Cells were harvested when the vanillate concentration of a culture's medium was ϳ20% of its initial value (cell densities of ϳ8 ϫ 10 8 cells/ml); at this point, ϳ65% of the GGE initially present in the GGE-fed culture was converted into downstream intermediates. Because both cultures were actively metabolizing vanillate, and only one culture was actively metabolizing GGE, the differences in transcript levels between the cultures should largely be related to GGE metabolism. 40 ml of harvested culture was combined with 5.71 ml of ice-cold stop solution (95% ethanol, 5% acid phenol/chloroform (5:1 solution, pH 4.5)). These mixtures were centrifuged at 4°C for 12 min at 6000 ϫ g. Cell pellets were resuspended into 2 ml of lysis solution (2% SDS, 16 mM EDTA in RNase-free water) and then incubated at 65°C for 5 min. RNA purification and cDNA synthesis were performed as described previously (38), using SuperScript III reverse transcriptase (Thermo Fisher Scientific) to construct the cDNA library. Genes selected for transcript analysis were the top BLASTp hits in the N. aromaticivorans genome of gene products from Sphingobium sp. SYK-6 or Novosphingobium sp. MBES04 previously shown to catalyze reactions of the ␤-etherase pathway (Table 1).
Real-time qPCR was performed on a 7500 real-time PCR system (Applied Biosystems, Forest City, CA) using SYBR Green JumpStart Taq ReadyMix (Sigma-Aldrich). Primers used to detect transcripts are contained in Table S5. Transcript levels were normalized to those of Saro_0141 (rpoZ, coding for the RNA polymerase Omega subunit).

Determination of COD
Initial COD values for cultures were obtained either from uninoculated medium or from inoculated medium that was immediately passed through a 0.22-m filter. Final COD samples were collected when cultures reached their maximum cell densities, both from unfiltered culture (cells and medium) and filtered culture (medium). The difference in COD between the unfiltered and filtered final samples is defined as the COD of cellular biomass. Samples were diluted as needed and combined with High Range COD Digestion Solution (Hach, Loveland, CO). The mixtures were heated to 150°C for 120 min to oxidize the materials before absorbances were measured at 600 nm. Standards with known COD values were analyzed in parallel.
A racemic mixture of MPHPV was synthesized by dissolving erythro-GGE into ethyl acetate (Fisher) and then slowly adding 1.25 molar eq of DDQ and stirring for 30 min. The reaction was washed three times with saturated NaHCO 3 to remove DDQH 2 formed during the reaction. The MPHPV was purified via flash chromatography using hexane/ethyl acetate (0.33:0.67, v/v), as described previously (15), and then crystallized from the eluent via solvent evaporation. HPV was synthesized as described previously for synthesis of ␤-deoxy-␣-veratrylglycerone, except using 4-O-benzyl-acetovanillone as starting material, rather than acetoveratrone (15). Synthesis of HPV required an additional debenzylation step that was unnecessary in the synthesis of ␤-deoxy-␣veratrylglycerone.

Structure determination
NaGST Nu was screened for crystal formation against several commercial screens at 277 and 293 K using a TTP Labtech Mosquito crystallization robot. The best-diffracting ammonium acetate-precipitated crystal was obtained at 293 K, by mixing 0.2 l of protein solution (277 M protein preincubated for 50 min with 10 mM GSH (neutralized with NaOH)) with 0.2 l of reservoir solution (4 M ammonium acetate buffered with 100 mM sodium acetate, pH 4.6). This crystal was mounted directly from the growth solution by drawing it through a layer of fomblin oil, thinning the surrounding liquid with a paper wick, and plunging into liquid nitrogen. The best-diffracting ammonium sulfate-precipitated crystal was obtained at 293 K using 0.13 l of protein solution and 85 nl of reservoir solution (1.35 M ammonium sulfate, 0.1 M lithium sulfate, and 0.1 M Bistris propane, pH 7.5). This crystal was cryopreserved by adding 0.5 l of a solution composed of 2 parts reservoir solution and 1 part undiluted glycerol to the droplet containing the crystal and equilibrating for 11 min before looping and plunging into liquid nitrogen.
Diffraction data were obtained at the GM/CA beamline at Argonne National Laboratory with an Eiger 16M detector (39). Data were collected on the ammonium acetate (PDB code 5UUN) and ammonium sulfate (PDB code 5UUO) crystal forms using 1.033 Å (for 1.45 Å resolution) or 0.7749 Å (for 1.25 Å resolution) X-rays, respectively. Diffraction data were reduced using XDS (40). Both crystals belonged to space group P2 1 2 1 2 1 with a predicted solvent content of 60%. The structure was solved by molecular replacement with Phaser (41) in the Phenix suite (42), using a search model based on PDB entry 3C8E:A (EcYghU (18)) modified with phenix.sculptor (43), based on primary sequence alignment. Phenix.refine (44) and COOT (45) were alternatively used to refine the structure and fit the model to electron density maps. Structure solution revealed two copies of the protein per asymmetric unit, with strong electron density present for the paired active-site glutathione molecules. The composite simulated annealing omit map for the active-site glutathiones was calculated in Phenix using default parameters and contoured at 2. The active sites were modeled containing either two GSH molecules or a single GSSG molecule, and difference density maps were calculated by setting the occupancy of either conformation to one or zero and then running one round of standard refinement, contoured at 4. The omit map data were best fit to a model in which each active site contained electron density from both a pair of GSHs

Deglutathionylation of a GS-phenylpropanoid by Nu-class GSTs
(locked to be of equal occupancy, Q(GSHA) ϭ Q(GSHB)) and a single GSSG (occupancy 1-Q(GSHA,B)), which was interpreted as a heterogeneous population of protein molecules containing either two GSHs or one GSSG throughout the crystal (Fig. S6).

Molecular modeling of substrate binding
For modeling ␤(R)and ␤(S)-GS-HPV and their syringyl phenylpropanoid analogues into the active sites of NaGST Nu and EcYghU, the PyMOL Builder function (46) was used to create molecules of GS-HPV or GS-syringyl by adding atoms onto the GSH2 molecule bound in each active site. Atoms were added so as to visually minimize steric clash with the proteins. The potential energies of the protein-GS-phenylpropanoid complexes were minimized using the Minimize Structure function of UCSF Chimera (23). For the energy minimization, all of the atoms of the protein-GS-phenylpropanoid complex were held rigid except for those of the phenylpropanoid moiety and the Cys side chain of GSH2. 100 steepest descent steps were run, followed by 20 conjugate gradient steps, and all step sizes were 0.05 Å.