Redundancy in Aromatic O-Demethylation and Ring-Opening Reactions in Novosphingobium aromaticivorans and Their Impact in the Metabolism of Plant-Derived Phenolics

For lignocellulosic biorefineries to effectively contribute to reduction of fossil fuel use, they need to become efficient at producing chemicals from all major components of plant biomass. Making products from lignin will require engineering microorganisms to funnel multiple phenolic compounds to the chemicals of interest, and N. aromaticivorans is a promising chassis for this technology.

this pathway, allowing the engineering of an improved strain for production of PDC from lignin. KEYWORDS Novosphingobium aromaticivorans, O-demethylases, PDC, aromatic metabolism, biological funneling, dioxygenases, lignocellulose, syringic acid L ignocellulosic plant biomass, composed of cellulose, hemicellulose, and lignin, is the most abundant organic material on the planet with potential to support a sustainable economy based on renewable feedstocks (1). Numerous studies predict that the economic and environmental viability of lignocellulosic biomass utilization for fuel and chemical production will be increased by the utilization of as much of these polymers as possible, including the use of lignin for production of chemicals (2)(3)(4). We are interested in deciphering the bacterial metabolism of phenolic compounds to engineer bacterial hosts to convert biomass-derived lignin into chemicals.
Lignin is an amorphous heteropolymer containing mainly syringyl (S; two methoxy groups), guaiacyl (G; one methoxy group), and p-hydroxyphenyl (H; no methoxy groups) phenolic structures that differ in the number of methoxy groups attached to the aromatic ring (5). One approach to valorizing lignin that has gained significant attention is to first use chemical techniques to deconstruct plant biomass and generate mixtures containing a large fraction of lower-molecular-weight phenolic compounds that could then be transformed by engineered microbes to a single valuable product (6). This funneling of phenolic mixtures to single compounds has been demonstrated with engineered strains of Pseudomonas putida (6,7), Rhodococcus jostii (8), and Novosphingobium aromaticivorans (9). In addition, other microbes, such as the yeast Rhodosporidium toruloides (10) and the photoheterotrophic bacterium Rhodopseudomonas palustris (11)(12)(13), have been extensively studied for their ability to transform the plant-derived phenolic compounds often present in deconstructed plant biomass.
Among the desirable features for a microbial strain to be used as a chassis for the development of microbial lignin valorization strategies are an ability to metabolize the majority of the biomass-derived phenolic compounds and to funnel them into native convergent metabolic pathways (14). We are studying the sphingomonad bacterium N. aromaticivorans as a platform microbe for lignin valorization because it efficiently and simultaneously utilizes a large variety of S, G, and H phenolics (9,15) and because it and other sphingomonads have enzymes to cleave different intersubunit bonds in the lignin polymer (16)(17)(18)(19)(20). These features, when combined with the genetic and genomic information on sphingomonads like N. aromaticivorans, could support strategies to maximize the number and type of lignin depolymerization products (e.g., phenolic monomers, oligomers) that can be microbially transformed into a range of valuable chemicals.
Metabolic pathways for the degradation of phenolics in sphingomonads have been proposed for Sphingobium sp. strain SYK-6, based on experiments with mutant strains and purified enzymes (21)(22)(23), and for N. aromaticivorans, based on analysis of a genome-scale transposon library and a set of targeted deletion mutants (9,15). These studies have revealed several commonalities in the phenolic metabolism pathways of both organisms (Fig. 1), but there remain significant knowledge gaps that limit the engineering of strains with increased transformation of phenolics to desired products.
To illustrate some of these knowledge gaps, syringic acid, which is an abundant biomass component that has been analyzed as a model S phenolic in both organisms, is demethylated to 3-methoxygallic acid (3-MGA), after which multiple pathways have been proposed for 3-MGA transformation by Sphingobium sp. SYK-6 (21,22). In one proposed pathway, the demethylation of 3-MGA produces gallic acid (GA), whose aromatic ring is cleaved by a dioxygenase to produce 4-oxalomesaconate (OMA) (24). Alternatively, it is proposed that the aromatic ring of 3-MGA is cleaved by a dioxygenase to produce 4-carboxy-2-hydroxy-6-methoxy-6-oxohexa-2,4-dienoate (CHMOD), which could be converted to OMA by an unknown enzyme (21,22). Furthermore, in vitro experiments with a purified dioxygenase from Sphingobium sp. SYK-6 (LigAB) and 3-MGA as the substrate showed rapid production of 2-pyrone-4,6-dicarboxylic acid (PDC) in addition to the ring cleavage product CHMOD, a result that led to the hypothesis that LigAB catalyzes the transformation of 3-MGA to both CHMOD and PDC (22). In N. aromaticivorans, the only pathway thus far proposed for syringic acid metabolism (15) is via demethylation to 3-MGA, ring cleavage to CHMOD, and conversion to OMA. However, an N. aromaticivorans deletion mutant lacking the proposed enzymes for CHMOD conversion to OMA (DesC and DesD) resulted in accumulation of PDC, suggesting the possibility for cyclization of CHMOD to PDC that is independent of these enzymes (9).

Functional Pathway Redundancy in N. aromaticivorans
Applied and Environmental Microbiology Sphingobium sp. SYK-6, an O-demethylase (LigM) has been proposed to be active on both 3-MGA and vanillic acid (25).
To address knowledge gaps in the reactions and enzymes involved in S, G, and H phenolic metabolism by N. aromaticivorans, we analyzed putative O-demethylases and aromatic ring opening dioxygenases that are predicted to be involved in the metabolism of these plant-derived aromatics. We present results of experiments with purified enzymes and with deletion mutants that provide evidence for functionally redundant pathways and for enzymes with activity on more than one substrate in this organism. Our studies led us to identify an aromatic ring-opening dioxygenase (LigAB2) and an O-demethylase (DmtS) that have not been previously shown to have these activities in sphingomonads. In addition, the newly acquired knowledge on enzyme redundancy and substrate specificity allowed us to engineer a second-generation N. aromaticivorans DSM 12444 strain with improved yields of PDC from plantderived phenolics.
(i) Effect of deleting putative O-demethylase genes on N. aromaticivorans growth. To begin evaluating the involvement of ligM, desA, and dmtS in the degradation of S and G phenolics in N. aromaticivorans, we generated mutants ( Table 1) containing combinations of deletions of these three genes in a parent strain (12444 D1879) and in a strain (12444PDC) in which deletions in PDC and CHMOD degradation genes led to accumulation of PDC from S, G, and H aromatics (9). Figure 2 shows growth curves for the parent strain and for its corresponding set of mutant strains. When cultured in the presence of only syringic acid, the parent strain and the mutant strains 12444 DligM, 12444 DdmtS, and 12444 DligM DdmtS had similar growth patterns, whereas all mutant strains lacking desA were unable to grow ( Fig. 2A).
a Construction of the parent and PDC-producing strains was described previously by Perez et al. (9). The parent strain has a deletion of Saro_1879 (sacB) to render it susceptible to genetic manipulation using sacB as a counterselection marker (9).
This suggests that desA is essential for N. aromaticivorans growth on syringic acid, whereas ligM and dmtS are not. When cultured in the presence of vanillic acid as the sole carbon source, three distinct growth patterns were observed for this set of mutants (Fig. 2B). First, all mutant strains with an intact ligM (12444 DdesA, 12444 DdmtS, and 12444 DdesA DdmtS) exhibited a similar growth pattern as the parent strain, suggesting that the DesA and DmtS enzymes are not essential for N. aromaticivorans growth on vanillic acid. Second, strains 12444 DligM and 12444 DligM DdmtS showed lower growth rates and a lower final cell density than the parent strain, and third, strains simultaneously lacking desA and ligM, 12444 DdesA DligM and 12444 DdesA DligM DdmtS, were unable to grow in the presence of vanillic acid.
Because the growth defects caused by gene deletions in the parent strain do not necessarily reveal which specific intracellular reactions are being affected, and can obscure the roles of genes with redundant functions, we also used the PDC-producing strain as a background to construct mutants lacking the same combinations of putative O-demethylase genes ( Table 1). With this second set of mutants, which require a nonaromatic carbon source for growth, we used PDC production as a proxy to elucidate the metabolic pathways affected by the gene deletions.
In the presence of glucose and syringic acid as an aromatic carbon source, the original PDC-producing strain (12444PDC) completely removed the syringic acid from the medium and produced PDC with a yield of ;85% ( Fig. 3A; Table 2). All strains with deletion of desA (12444PDC DdesA, 12444PDC DdesA DligM, 12444PDC DdesA DdmtS, and 12444PDC DdesA DligM DdmtS) were not able to degrade syringic acid (Fig. 3B, E, F and H). These results support an essential role for DesA in syringic acid metabolism by N. aromaticivorans, which is proposed to be its demethylation to 3-MGA (Fig. 1).
Furthermore, all the mutant strains with intact desA that were constructed in the PDC-producing background (12444PDC DligM, 12444PDC DdmtS, and 12444PDC DligM DdmtS) consumed syringic acid and accumulated PDC. Notably, the PDC yield from syringic acid by strain 12444PDC DligM DdmtS was stoichiometric and in all cases statistically higher (P , 0.05) (see Table S1 in the supplemental material) than the yield of this product obtained in the other strains ( Table 2). This suggests that the simultaneous  Table 1 for the definition of strain names. Values correspond to the average results for three biological replicates. deletion of both ligM and dmtS blocks another previously unknown pathway for syringic acid metabolism that normally detracts from PDC production in the 12444PDC strain.
We predict that this alternative pathway for syringic acid metabolism involves Odemethylation (by LigM and/or DmtS) of 3-MGA to GA, since that transformation has been shown to occur in Sphingobium sp. SYK-6 ( Fig. 1) (25). An observed small and transient accumulation of 3-MGA in the culture medium when 12444PDC DligM and 12444PDC DligM DdmtS ( Fig. 3C and G) are grown in the presence of syringic acid supports this hypothesis. To further test this hypothesis, we grew the 12444PDC strain and its corresponding set of putative O-demethylase deletion mutants in the presence of glucose and 3-MGA (Fig. 4). We posited that using 3-MGA as the aromatic substrate would provide a growth substrate that requires only one demethylation reaction, compared to growth with syringic acid that is predicted to require the removal of two methyl groups. In this set of experiments, strain 12444PDC was able to completely remove 3-MGA from the medium, converting ;94% of it into PDC. Moreover, in contrast with the experiments using syringic acid (Fig. 3), each of the putative O-demethylase mutant derivatives of 12444PDC were able to completely remove 3-MGA from the medium (Fig. 4). This result indicates that none of these putative O-demethylases is essential for conversion of 3-MGA to PDC. In addition, all strains with deletions of ligM (12444PDC DligM, 12444PDC DdesA DligM, 12444PDC DligM DdmtS, and 12444PDC  Table 1 for the definition of strain names. Values correspond to the average results for three biological replicates. Error bars are included but in most cases not visible because they are smaller than the symbols in the plots. Abbreviations: 3-MGA, 3-methylgallate; PDC, 2-pyrone-4,6-dicarboxylic acid. DdesA DligM DdmtS) had statistically higher PDC production (P , 0.05) (Table S1) than strain 12444PDC (Table 2), a result that suggests the involvement of LigM in the proposed alternative pathway of syringic acid metabolism. Notably, stoichiometric conversion of 3-MGA to PDC was achieved in the mutants with simultaneous deletion of ligM and dmtS (12444PDC DligM DdmtS and 12444PDC DdesA DligM DdmtS), which, in agreement with the experiments using syringic acid, suggests that the loss of LigM and DmtS blocks the alternative pathway for syringic acid metabolism in this bacterium.
In an additional experiment with the PDC-producing strain and its derivatives, we analyzed the role of the putative O-demethylases in the metabolism of vanillic acid. In the presence of glucose and vanillic acid, strain 12444PDC completely removed the vanillic acid from the medium and converted it into PDC with ;96% yield ( Table 2). Transient extracellular accumulation of a small amount of PCA during the course of the experiment (Fig. 5A) supports the predicted role of demethylation in the degradation of vanillic acid (Fig. 1). Analysis of the mutants lacking single putative O-demethylases ( Fig. 5B to D) revealed a decrease in the vanillic acid consumption rate when ligM was deleted (Fig. 5C), compared to the other two strains, suggesting a role for LigM in vanillic acid metabolism. In the set of double O-demethylase mutants ( Fig. 5E to G), no effect was seen when desA and dmtS were deleted (Fig. 5F), but a decrease in vanillic acid consumption rates was evident with all double mutants that lacked the ligM gene. This observation is also consistent with the reduced rate of vanillic acid degradation in the single ligM deletion mutant (Fig. 5C). Notably, the rate of vanillic acid consumption decreased the most when ligM and desA were both deleted (Fig. 5E), suggesting that DesA can partially substitute for LigM in the degradation of vanillic acid. Finally, minimal vanillic acid degradation was observed in a mutant that lacked all three of the putative O-demethylase genes (Fig. 5H), consistent with the lack of LigM and DesA activity in this strain. Deleted gene(s) a All experiments used 3.2 mM glucose and a 3 mM concentration of the corresponding aromatic substrate as organic carbon sources. b Strains are defined in Table 1. c Yield reported as average 6 standard deviation of results for three biological replicates. d N/C, PDC yield was not calculated for conditions under which the aromatic substrate was not consumed.
(ii) Activity of LigM and DesA with aromatic substrates. In vitro experiments were performed with purified recombinant LigM and DesA proteins to test their activity with the methoxylated aromatic substrates. These assays were performed only with LigM and DesA, since we have thus far been unsuccessful in purifying an active recombinant DmtS protein. The LigM and DesA homologues from Sphingobium sp. SYK-6 have been shown to be tetrahydrofolate (H 4 folate)-dependent O-demethylases (25, 26), a prediction that was experimentally confirmed (Fig. S1).
The recombinant LigM protein of N. aromaticivorans was able to convert 3-MGA into GA and vanillic acid into PCA at comparable rates under our assay conditions ( Fig.  6C and E). However, under identical conditions, the recombinant LigM protein was unable to convert a detectable amount of syringic acid into 3-MGA (Fig. 6A). These results are consistent with the observations in growth experiments with mutant strains, which predicted LigM's involvement in vanillic acid and syringic acid metabolism.
We also found that the recombinant DesA of N. aromaticivorans demethylated both syringic acid and, somewhat more slowly, vanillic acid ( Fig. 6B and F) but was not active in demethylating 3-MGA (Fig. 6D). These results support the critical role of DesA in the demethylation of syringic acid that was predicted by analyzing growth of mutant strains (Fig. 2) and its potential role in vanillic acid transformation when ligM is deleted (Fig. 5) and offer new evidence supporting the hypothesis that this enzyme does not participate in 3-MGA demethylation.  Table 1 for the definition of strain names. Values correspond to the average results for three biological replicates. Error bars are included but in most cases not visible because they are smaller than the symbols in the plots. Abbreviations: 3-MGA, 3-methylgallate; PDC, 2-pyrone-4,6-dicarboxylic acid.
Identification of putative aromatic ring-opening dioxygenases in N. aromaticivorans. The opening of the aromatic ring is an essential step in the assimilation of plantderived phenolic compounds into intermediary metabolism (Fig. 1). In N. aromaticivorans, the only aromatic ring-opening dioxygenase that has been previously identified is a LigAB homologue, encoded by Saro_2813 (ligA) and Saro_2812 (ligB) (15), whose subunits have ;67% and ;70% amino acid sequence identity with LigA and LigB of Sphingobium sp. SYK-6, respectively. A search of the N. aromaticivorans genome reveals genes that could encode another aromatic ring-opening dioxygenase encoded by Saro_1233 and Saro_1234 (hereafter referred to as ligA2 and ligB2, respectively), whose subunits have amino acid sequences that are ;33% and ;42% identical to the LigA and LigB of N. aromaticivorans (41% and 43% amino acid identity to the LigA and LigB of Sphingobium sp. SYK-6, respectively).
(i) Effect of deleting genes encoding putative aromatic ring-opening dioxygenases on growth of N. aromaticivorans. To evaluate the roles of LigAB and LigAB2 in the degradation of S and G phenolics by N. aromaticivorans, we tested growth and aromatic metabolism by mutants containing combinations of deletions in ligAB and ligAB2 in the parent strain (12444 D1879) and the PDC-producing strain (12444PDC) ( Table 3). When cultured in the presence of either syringic acid or vanillic acid, the parent strain and strain 12444 DligAB2 both grew well, whereas strains with deletion of ligAB (12444 DligAB and 12444 DligAB DligAB2) were unable to grow (Fig. 7). This indicates that  Table 1 for the definition of strain names. Values correspond to the average results for three biological replicates. Error bars are included but in most cases not visible because they are smaller than the symbols in the plots. Abbreviations: PCA, protocatechuic acid; PDC, 2-pyrone-4,6-dicarboxylic acid.
LigAB is necessary for N. aromaticivorans growth on both syringic and vanillic acids, whereas LigAB2 is not.
Investigating the metabolism of syringic acid by mutants containing deletions in ligAB and ligAB2 in the PDC-producing strain (12444PDC) (Fig. 8), we observed that all strains consumed syringic acid, but only the mutant with intact LigAB (Fig. 8B) produced PDC with a high yield (;84%) ( Table 4), supporting the hypothesis that LigAB plays an important role in PDC production. The proposed pathway for PDC production from syringic acid includes ring cleavage of 3-MGA to CHMOD, which is then converted to PDC (Fig. 1). The high PDC yield observed in cells in which LigAB is present (Table 4) indicates that this enzyme plays a major role in the transformation of 3-MGA. It also suggests that in N. aromaticivorans, 3-MGA ring opening is the primary route for 3-MGA metabolism, while demethylation of 3-MGA to GA is a secondary route responsible for only a small fraction of the metabolized 3-MGA. In support of this, the tests with the two mutants that were missing LigAB ( Fig. 8A and C) showed lower rates of syringic acid consumption and accumulation of 3-MGA and GA, the predicted metabolic intermediates in this secondary metabolic route. Strain 12444PDC DligAB also converted only ;2% of the syringic acid to PDC ( Table 4), suggesting that LigAB2 may have some role in 3-MGA metabolism, whereas strain12444PDC DligAB DligAB2 did not produce any detectable PDC (Table 4), suggesting a complete interruption of 3-MGA aromatic ring opening in the mutant that lacked both of the putative dioxygenases, LigAB and LigAB2.
Notably, in the two experiments in which we observed GA accumulation ( Fig. 8A  and C), the medium darkened in color, which could be due to a rapid nonenzymatic transformation of GA under the conditions used for these experiments. Since the sum of the molar concentration of unreacted syringic acid plus the accumulated metabolites does not add up to 100% in either experiment, these results also suggest that the GA produced was partially degraded either abiotically or biologically.
The role of the putative ring-opening dioxygenases in vanillic acid degradation was also tested using the deletion mutants in the PDC-producing strain background ( Fig.  9). Both the LigAB and LigAB2 mutant derivatives of the PDC-producing strain showed degradation of vanillic acid, but only the mutant with intact ligAB genes produced high yields of extracellular PDC ( Fig. 9B; Table 4). The strains with deleted ligAB genes, 12444PDC DligAB (Fig. 9A) and 12444PDC DligAB DligAB2 (Fig. 9C), had a lower consumption of vanillic acid, removing only ;80% of it from the medium and converting ;48% of it into extracellular PCA over the course of the experiments. These observations suggest that LigAB has a significant role in PCA ring opening in N. aromaticivorans. Notably, strain 12444PDC DligAB also converted ;12% of the vanillic acid into PDC, whereas 12444PDC DligAB DligAB2 did not accumulate any detectable PDC in the medium, suggesting that in the absence of LigAB, LigAB2 may also function in PCA ring opening, while the absence of both LigAB and LigAB2 completely eliminates the PCA ring-opening activity in N. aromaticivorans.
(ii) Activity of LigAB and LigAB2 with aromatic substrates. To investigate the predicted activities of LigAB and LigAB2 in aromatic ring opening, we purified recombinant forms of the proteins and tested them for activity in vitro (Fig. 10). With 3-MGA as the substrate, LigAB completely converted the 3-MGA from the reaction mixture within 2 h (Fig. 10A), whereas LigAB2 converted only ;20% of the 3-MGA over the course of 24 h (Fig. 10B). With each purified enzyme, two products that transiently accumulated in the assay were identified as isomers of CHMOD (see supplemental material), and a third product was identified as PDC. These results show that LigAB catalyzes the trans- a Construction of the parent strain (12444 D1879) and the PDC-producing strain (12444PDC) was described previously by Perez et al. (9). See also footnote a in Table 1.  Table 3 for the definition of strain names. Values correspond to the average results for three biological replicates. Error bars are included but in most cases not visible because they are smaller than the symbols in the plots.
formation of 3-MGA to CHMOD and that CHMOD is subsequently transformed to PDC, as proposed previously (9). They also show that LigAB2 can also convert 3-MGA to CHMOD, albeit at a lower rate than with purified LigAB under the assay conditions used, which is consistent with results from the in vivo analyses of defined mutants in each of the corresponding genes ( Fig. 7 and 8).
Both LigAB and LigAB2 also showed activity when tested with GA as the substrate (Fig. 10C and D). One compound that accumulated in these assays has been identified as OMA, the expected product of this ring-opening reaction (see supplemental material). Under identical assay conditions, LigAB2 degraded ;31% of the GA after 2 h of reaction (Fig. 10D), whereas LigAB degraded only ;11% (Fig. 10C). These findings suggest that of these two ring-opening dioxygenases, N. aromaticivorans LigAB2 has higher catalytic activity with GA than purified LigAB.
When tested with PCA as the substrate under identical assay conditions, LigAB converted ;81% of the PCA from the reaction mixture after 2 h (Fig. 10E), while LigAB2 converted only ;14% of the PCA over the same time span (Fig. 10F). The reaction product from both assays was identified as CHMS (see supplemental material), consistent with the expected ring-opening product of these enzymes with PCA as a substrate. The relative rates of PCA disappearance in assays using LigAB and LigAB2 suggest that under the assay conditions, LigAB is more catalytically active with PCA, which is consistent with the in vivo experiments with the individual mutant strains (Fig. 9).

DISCUSSION
Strategies to successfully engineer efficient microbial catalysts that produce valuable compounds from chemically depolymerized lignocellulosic biomass have several  Table 3. c Yield reported as average 6 standard deviation of results for three biological replicates. d Results for strain 12444PDC are the same as reported in Table 2 and are included here to facilitate comparison.  Table 3 for the definition of strain names. Values correspond to the average results for three biological replicates. Error bars are included but in most cases not visible because they are smaller than the symbols in the plots. Abbreviations: 3-MGA, 3-methylgallate; PDC, 2-pyrone-4,6-dicarboxylic acid.
requirements, including the need for the microorganisms to funnel a heterogenous mixture of plant-derived phenolic compounds through central pathways and the ability to genetically engineer the microorganisms to direct the flow of carbon from central aromatic metabolic pathways to the production of valuable compounds. To develop these strategies, it is necessary to acquire a detailed understanding of the native aromatic metabolic pathways in the microorganisms to be used as chassis for lignin valorization. This study focuses on advancing the knowledge of native aromatic metabolism in N. aromaticivorans, a sphingomonad of interest because it efficiently degrades the major S, G, and H phenolic substituents of plant biomass (9, 15) and because it is equipped with metabolic pathways for breaking down interunit linkages in lignin (17)(18)(19)(20). The ability of N. aromaticivorans to efficiently degrade many aromatic compounds may be linked to having functionally redundant aromatic degradation pathways and enzymes with broad substrate specificity. While having redundant pathways and enzyme promiscuity may confer this microorganism with ecological advantages in nature, these features can create challenges or opportunities when engineering such microorganisms to produce high yields of a desired compound. For example, we reported earlier that the 12444PDC strain of N. aromaticivorans is able to funnel multiple lignin-derived aromatic compounds into PDC, but the PDC yields from S phenolics were lower than those from G and H phenolics (9). A possible explanation for this observation is pathway redundancy (16), which would allow N. aromaticivorans to channel a fraction of the S phenolics through one or more uncharacterized pathways that were not blocked in the 12444PDC strain. Based on the published analysis of aromatic metabolism of Sphingobium sp. SYK-6 ( Fig. 1), we posited that one uncharacterized step might be the O-demethylation of 3-MGA to form GA with subsequent aromatic ring opening to produce OMA (Fig. 1). Because O-demethylation and aromatic ring opening are also involved in other branches of the aromatic metabolism pathways, in this work we systematically evaluated these two functions in N. aromaticivorans. Below we discuss the new information derived from the integration of in vivo experiments with mutant strains and in vitro experiments with purified enzymes (summarized in Fig. 11). The knowledge gained from these experiments helps us define roles for enzymes not previously described in N. aromaticivorans or other sphingomonads, identify functional pathway redundancy in the metabolism of S phenolics by this organism, and refine predictions of the substrate specificity of key N. aromaticivorans enzymes.
O-Demethylation reactions. The two H 4 folate-dependent O-demethylases, DesA and LigM, and the newly identified DmtS, a potential new O-demethylase, were each shown to have a role in the metabolism of S and G phenolics in N. aromaticivorans (Fig. 11). While we obtained genetic evidence for the role of DmtS as an O-  Table 3 for the definition of strain names. Values correspond to the average results for three biological replicates. Error bars are included but in most cases not visible because they are smaller than the symbols in the plots. Abbreviations: PCA, protocatechuic acid; PDC, 2-pyrone-4,6-dicarboxylic acid. demethylase, we were not able to purify recombinant DmtS, so we lack direct information about its substrate specificity. Of known O-demethylases, DmtS is most similar in its predicted amino acid sequence (;27% identical) to the monooxygenase component (subunit A) of the vanillic acid O-demethylase VanAB, whose function has been demonstrated in other bacteria (29,30) and that plays a key role in aromatic metabolism in P. putida (27,28). In P. putida and other bacteria, the vanA gene is found in an operon with vanB, which encodes the putative reductase, VanB. However, dmtS is not contained in a gene cluster in N. aromaticivorans with a gene that encodes a protein with amino acid sequence identity to a known VanB reductase. Nevertheless, we were able to use mutants lacking dmtS to confirm its role in aromatic metabolism in N. aromaticivorans.  other aromatic compounds showed no detectable activity with this substrate (Fig. 6). Second, in vivo experiments with the set of putative O-demethylase mutants showed that DesA was essential for N. aromaticivorans growth on syringic acid (Fig. 2). Third, the experiments with the set of putative O-demethylase mutants constructed in the previously reported PDC-producing strain (12444PDC) showed that all of the mutants lacking desA did not degrade syringic acid (Fig. 3).
To our knowledge, no enzymes other than DesA have been reported to demethylate syringic acid and produce 3-MGA in any sphingomonad. Indeed, in Sphingobium sp. SYK-6, DesA has been proposed to be the only enzyme responsible for syringic acid O-demethylation, and, similarly to N. aromaticivorans, the deletion of the desA gene prevents the mutant strain from growing on syringic acid as the sole substrate (25). In addition, deletion of the desA gene in strain 12444PDC completely blocks syringic acid degradation and PDC production in all the mutant strains even when cell growth is promoted by the addition of glucose in the growth medium.
(ii) O-Demethylation of 3-MGA. Although O-demethylation of 3-MGA to GA has been described in Sphingobium sp. SYK-6 (26), evidence for enzymes involved in this reaction in N. aromaticivorans was lacking before our studies. Here, we provide several lines of evidence that 3-MGA is converted to GA in N. aromaticivorans, but it is not the major route of 3-MGA metabolism. Instead, our results support the hypothesis that aromatic ring opening of 3-MGA to CHMOD is the major route for the metabolism of syringic acid. First, the nonstoichiometric conversion of syringic acid to PDC in the 12444PDC strain (Fig. 3A) is evidence for the existence of a secondary pathway that supports the degradation of a small fraction (;15%) of the syringic acid in the PDC- producing strain. Second, stoichiometric conversion of syringic acid to PDC is achieved in the mutant that has an intact desA (necessary for the conversion of syringic acid to 3-MGA) and lacks both ligM and dmtS (Fig. 3G), leading to the hypothesis that both LigM and DmtS are active in this secondary pathway. This hypothesis was further confirmed in experiments using 3-MGA as the aromatic substrate (Fig. 4), where stoichiometric conversion of 3-MGA to PDC is achieved only by simultaneous deletion of ligM and dmtS (Table 2). Third, the in vitro assays with purified recombinant LigM (Fig. 6) showed that GA was stoichiometrically produced from 3-MGA. Finally, when comparing PDC yields from the 12444PDC strain and the strains with deletion of either dmtS or ligM (Table 2), a small but reproducible increase in PDC yield occurs when ligM is deleted, suggesting that in vivo, LigM is more active in 3-MGA demethylation than DmtS.
Based on a comparison of our results to that from other labs, the role of 3-MGA Odemethylation in the metabolism of syringic acid appears to be different in N. aromaticivorans and Sphingobium sp. SYK-6. For instance, inactivation of ligM in N. aromaticivorans did not cause a detectable effect in growth on syringic acid ( Fig. 2A), whereas inactivation of this gene in Sphingobium sp. SYK-6 has a detrimental effect on both growth rate and final cell density when cells use syringic acid (25). This suggests that between the two functionally redundant pathways for converting 3-MGA to OMA in N. aromaticivorans, ring opening to CHMOD carries more flow of carbon than demethylation to GA, which is the opposite of what was reported for Sphingobium sp. SYK-6. In addition, our results also suggest redundancy in the enzymes that demethylate 3-MGA to GA in N. aromaticivorans, with both LigM and DmtS having this activity. Sphingobium sp. SYK-6 appears to have less enzyme redundancy in this reaction, as deletion of ligM eliminated 3-MGA conversion to GA in cell extract experiments (25).
(iii) O-Demethylation of vanillic acid. Based on our data, we propose that LigM plays a major role in demethylation of vanillic acid but that DesA can perform this reaction, although with reduced efficiency (Fig. 11). First, the in vitro assays showed that both of these enzymes could convert vanillic acid to PCA, with LigM having a higher degradation rate (Fig. 6). Second, in vivo experiments with the set of O-demethylase mutants in the wild-type background showed that all mutants with an intact ligM gene grew as well as the parent strain, strains lacking ligM but with intact desA showed a detectable growth defect, and strains lacking both ligM and desA could not grow in the presence of vanillic acid (Fig. 2). Consistent with these findings, our in vivo experiments with the set of mutants constructed in the 12444PDC background confirmed that deleting ligM lowered the rates of vanillic acid degradation, deleting desA alone did not have an effect, but that deleting both ligM and desA prevented vanillic acid degradation (Fig. 5). Finally, a role for DmtS in vanillic acid O-demethylation was not evident in the results of any of the experiments. These observations are consistent with the predicted role of LigM and DesA in the metabolism of vanillic acid by Sphingobium sp. SYK-6 (25).
Aromatic ring-opening reactions. Aromatic ring opening of phenolic compounds in sphingomonads is predicted to be catalyzed by extradiol dioxygenases (cleaving at the 4,5 position), which often exhibit broad substrate specificity (16,31,32). In Sphingobium sp. SYK-6, at least three dioxygenases are reported to be involved in the metabolism of S and G compounds: LigAB, with highest activity on PCA (33); DesZ, with highest activity on 3-MGA and GA (21); and DesB, specific for GA (24). In contrast, previous analysis of N. aromaticivorans has implicated only one extradiol dioxygenase, LigAB, in aromatic ring opening of phenolic compounds (15). In this study, we showed that two aromatic ring-opening enzymes in N. aromaticivorans, LigAB and LigAB2, catalyzed extradiol cleavage of 3-MGA, GA, and PCA, the three intermediates in the pathways for metabolism of S, G, and H aromatics whose rings are capable of being enzymatically opened by a 4,5-dioxygense (Fig. 11). In addition, our results show that LigAB has higher activity with 3-MGA and PCA than LigAB2, whereas LigAB2 has higher activity with GA than LigAB.
(i) Aromatic ring opening of 3-MGA. Our results support the hypothesis that LigAB is the primary enzyme for 3-MGA ring opening in N. aromaticivorans (15). They also provide new evidence that LigAB2 has activity with 3-MGA and might partially substitute for the role of LigAB.
The in vitro assays using recombinant LigAB and LigAB2 proteins with 3-MGA as a substrate showed accumulation and disappearance of the known stereoisomers of CHMOD and accumulation of PDC (Fig. 10A and B). In vitro accumulation of PDC from 3-MGA has also been observed in enzyme assays with the LigAB of Sphingomonas sp. SYK-6 (22). Rapid PDC accumulation in those experiments led to the proposal that PDC is a product of LigAB activity in Sphingomonas sp. SYK-6. Several results from our reactions of LigAB and LigAB2 from N. aromaticivorans with 3-MGA led us to the different conclusion that CHMOD is the direct product of the LigAB and LigAB2 reactions with 3-MGA and that CHMOD is nonenzymatically converted to PDC under the conditions of the assays. First, CHMOD is unstable in aqueous solution, and nonenzymatic cyclization of CHMOD to PDC at neutral pH has been reported (34). Unfortunately, the CHMOD appears to cyclize under our reaction conditions too fast for us to be able to purify this compound for determination of its molar concentration in our assays. However, the sum of the 3-MGA and PDC molar concentrations over the course of the enzymatic reactions inversely correlated with the accumulation of the CHMOD isomers (see Fig.  S7 in the supplemental material), suggesting that CHMOD is an intermediate in the conversion of 3-MGA to PDC. Third, at the end of the LigAB reaction, when 3-MGA and CHMOD are not detectable (Fig. 10A), PDC accumulation reaches the initial concentration of 3-MGA in the assay, indicating the absence of any other potential products of CHMOD degradation. Finally, the results of in vitro assays of LigAB and LigAB2 with other substrates (Fig. 10) produced products consistent with the proposed function of these enzymes as extradiol aromatic ring-opening dioxygenases, and CHMOD is the expected product of such reaction when 3-MGA is the substrate (Fig. 11).
(ii) GA ring opening. The in vitro evidence obtained in this study indicates that LigAB2 reacts more rapidly with GA than with 3-MGA or PCA and confirms OMA as the product of GA ring opening (Fig. 10). The experiments with the set of deletion mutants in the 12444PDC background (Fig. 8) also provided evidence that either LigAB or LigAB2 might be responsible for metabolizing the fraction of syringic acid that is normally channeled through GA in the 12444PDC strain. However, under the conditions of our studies, the GA that accumulated in the medium apparently converted abiotically to an unknown product. Thus, further research is needed to specifically ascertain the role of LigAB and LigAB2 on GA transformation in vivo.
(iii) PCA ring opening. The results of in vitro (Fig. 10) and in vivo (Fig. 9) experiments with vanillic acid provided evidence that the ring opening of PCA is catalyzed primarily by LigAB but that LigAB2 could partially substitute in vivo when ligAB was deleted. Furthermore, deleting both sets of genes eliminated growth (Fig. 7) and PDC production (Fig. 9), indicating that no other enzyme in N. aromaticivorans could catalyze the ring opening of PCA under our growth and medium conditions. Implications for PDC production from lignin-derived aromatics by N. aromaticivorans. We have previously tested PDC production from plant-derived phenolics by N. aromaticivorans (9), because this native pathway intermediate is a potential building block for bio-based plastic and epoxy adhesives (35). We showed that strain 12444PDC, which contains mutations that block the conversion of PDC and CHMOD to OMA, could transform S, G, and H phenolics to PDC but exhibited lower PDC yields from S phenolics than from G or H phenolics (9). Based on this observation, we proposed that strain 12444PDC contained additional, previously unidentified, gene products that could metabolize S phenolics (9). In this study, we identified these previously unknown enzymes in the proposed alternative pathway for metabolism of syringic acid in N. aromaticivorans that O-demethylates 3-MGA to GA (Fig. 11). We further showed that these same enzymes catalyze ring opening of GA to OMA (Fig. 11). The activity of the enzymes in this previously unknown alternative pathway helps explain why strain 12444PDC had lower PDC yields from S phenolics, since our data suggest that ;15% of the 3-MGA follows the GA-OMA route and does not contribute to PDC production in strain 12444PDC. Although deleting the genes in this alternative pathway might increase the yield of PDC, the promiscuity of the other N. aromaticivorans O-demethylases described in this study implies that single gene deletions may not be sufficient to maximize PDC yields. As predicted, stoichiometric production of PDC from either syringic acid (Fig. 3G) or 3-MGA (Fig. 4G) was possible only with the deletion of both ligM and dmtS from the 12444PDC background. However, since O-demethylases are also needed in the demethylation of G phenolics, a potential side effect of ligM deletion (the main enzyme we found to be responsible for vanillic acid demethylation) could be the reduced conversion of G phenolics into PDC. Indeed, we observed the predicted negative effect in the rate of vanillic acid conversion when ligM was deleted (Fig. 5C), although the PDC yield was not affected when this gene was inactivated in the 12444PDC strain ( Table 2). Since DesA was shown to have activity with vanillic acid, one potential strategy to compensate for the absence of LigM might be to engineer PDC-producing strains that overexpress desA, as has been done with overexpression of vanAB in P. putida to accelerate vanillic acid degradation (36).
The results of our studies also make a new prediction that deleting enzymes responsible for the aromatic ring-opening step does not appear to be a productive strategy for reducing the flow of substrates through this alternative N. aromaticivorans pathway due to the promiscuity of these dioxygenases. For example, while in vitro assays with the newly discovered LigAB2 showed its preference for GA over 3-MGA and PCA, its inactivation did not produce measurable effects on PDC yield (Table 4). Furthermore, inactivation of both LigAB and LigAB2 would also prevent aromatic ring opening with other biomass-derived phenolic substrates. From our data, we propose that the main function of LigAB2 in vivo is in ring-opening reactions of aromatic compounds other than S, G, and H phenolics. While Sphingobium sp. SYK-6 has a predicted homologue of N. aromaticivorans LigAB2 (encoded by SLG_37520 and SLG_37530), we are not aware of any studies that investigate its role in S, G, or H aromatic metabolism.
In conclusion, we identified a previously unknown alternative route for syringic acid catabolism, described new enzymes that are involved in this pathway (DmtS and LigAB2), and performed a systematic genetic and enzymatic analysis of the O-demethylases and aromatic ring-opening dioxygenases that function in N. aromaticivorans. The studies revealed that these O-demethylases and dioxygenases have activity with multiple substrates, suggesting that they are able to participate in different reactions within individual pathways (Fig. 11). The new knowledge on the metabolism of aromatic compounds by N. aromaticivorans obtained in this work has allowed us to design a strain (12444PDC DligM DdmtS) with increased yield of PDC from syringic acid. It also will enable better predictions of metabolic routes that will facilitate engineering strains for improved yields of other desirable products from biomass-derived aromatics. While the interruption of 3-MGA O-demethylation (by deleting ligM and dmtS) in strain 12444PDC resulted in stoichiometric PDC production from syringic acid and vanillic acid, this new strain exhibited a reduced rate of vanillic acid degradation. This observation illustrates additional steps in aromatic acid degradation that are targets for future strain improvement.
All genetic modifications were carried out using a variant of the plasmid pK18mobsacB (37), which contains sacB and a kanamycin resistance gene. The gene deletions were performed as previously described (9), with the details of the processes used here described in the supplemental material. All primers, plasmids, and Escherichia coli strains used for cloning and protein expression are listed in Tables 5, 6, and 7, respectively.
E. coli cultures were grown in LB medium containing 50 mg ml 21 kanamycin at 37°C. N. aromaticivorans cultures were grown in SMB medium (see reference 17 for recipe) supplemented with the indicated carbon source at 30°C. For routine culture and storage, SMB medium was supplemented with 1 g liter 21 glucose. For constructing mutants, SMB medium was supplemented with 1 g liter 21 glucose and either N. aromaticivorans growth experiments. N. aromaticivorans cultures were grown overnight in SMB medium supplemented with 1 g liter 21 glucose. Cultures were diluted 1:1 with fresh medium containing 1 g liter 21 glucose and incubated for 1 h to resume cell growth. Cells from 2 ml of the growing culture were pelleted (5 min at 2,300 Â g), washed with 1 ml SMB medium without carbon, and then resuspended into 600 ml SMB medium with no added carbon. Growth experiments were initiated by adding 80 ml of the cell suspension into 8 ml of fresh SMB medium supplemented with either 3 mM syringic acid or 3 mM vanillic acid. Cultures were grown aerobically in 20-ml test tubes, with shaking at 200 rpm at 30°C. Each experiment was repeated 3 times.
N. aromaticivorans extracellular metabolite analysis. Bacterial cell cultures were grown overnight in 20 ml of SMB medium supplemented with 1 g liter 21 glucose and then reactivated by adding 2 ml of fresh medium containing 1 g liter 21 glucose and incubated for 1 h. Experiments were initiated by inoculating 5 ml of active culture into 15 ml of fresh SMB medium supplemented with 5 mM glucose and either 4 mM syringic acid, 4 mM 3-MGA, or 4 mM vanillic acid. Cultures were grown aerobically in 125-ml conical growth flasks, with shaking at 200 rpm at 30°C. Samples were collected periodically, filtered (through 0.22-mm pores) to remove cells, and immediately analyzed by high-performance liquid chromatography-mass spectrometry (HPLC-MS) to monitor the extracellular aromatic compounds. Each experiment was repeated 3 times. PDC yields were calculated as the amount of PDC produced divided by the amount of aromatic substrate consumed. Comparison of PDC yields was done with an unpaired, two-tailed t test, with a P of #0.05 taken as a statistically significant difference in the observed yields.
Recombinant enzyme expression and purification. Genes Saro_2812/3 (ligAB), Saro_1233/4 (ligAB2), Saro_2861 (ligM), and Saro_2402 (desA) from N. aromaticivorans were independently cloned into plasmid pVP302K (19), which incorporates a His 8 tag to the N terminus of the translated transcripts connected via a tobacco etch virus (TEV) protease recognition site (see the supplemental material for plasmid construction details). The expression plasmids were transformed into E. coli B834 (38, 39) containing plasmid pRARE2 (Novagen, Madison, WI), and transformants were grown in ZYM-5052 autoinduction medium (40) containing 50 mg ml 21 kanamycin and 20 mg ml 21 chloramphenicol. Recombinant proteins were purified by passing crude E. coli lysates through a nickel-nitrilotriacetic acid (Ni 21 -NTA) column as described previously (17). His 8 tags were cleaved from recombinant proteins using TEV protease, and proteins were passed again though a Ni-NTA column to remove the cleaved His 8 tag and the TEV protease (which contains its own His tag).
In vitro aromatic ring-opening dioxygenase assays. Preliminary experiments with recombinant LigAB (Saro_2812/3) and LigAB2 (Saro_1233/4) purified in the presence of air suggested that the enzymes were catalytically inactive, as reported for other homologues of these proteins (41). We thus separately mixed LigAB and LigAB2 with reactivation buffer (buffer A, containing 20 mM Na 2 HPO 4 ,  SO 4 , and 1 mM ascorbic acid and prepared anaerobically) to a concentration of 2 mM enzyme active sites under anaerobic conditions and incubated them anaerobically at 30°C for 21 h to reactivate the enzymes. Three solutions containing buffer B (20 mM Na 2 HPO 4 , 20 mM KH 2 PO 4 , and 1 mM ascorbic acid) and either 2 mM 3-MGA, 2 mM PCA, or 2 mM GA were prepared in the presence of air. Enzyme assays were performed in 2-ml polypropylene vials inside an anaerobic chamber at 30°C by mixing 750 ml of reactivated enzyme in buffer A with 750 ml of aromatic substrate in buffer B. As a control, we mixed 750 ml of buffer A without enzyme with 750 ml of each aromatic substrate in buffer B. After 30 min, the vials were exposed to the ambient atmosphere outside the anaerobic chamber for 10 min to expose the reactions to the O 2 predicted to be a substrate for the ring-opening reaction and then transferred back into the anaerobic chamber. Exposure to O 2 was repeated at 2 h and 4 h after the assay was initiated. Samples (250 ml) were collected at time zero, at 30 min, and at 1, 2, 4, and 24 h, immediately mixed with 50 ml of 1 N HCl to terminate the reaction, and then analyzed by HPLC-MS to quantify substrate disappearance and formation of any products. Identification of OMA, CHMOD, and PDC was performed by gas chromatography (GC)-MS of samples extracted with ethyl acetate immediately after collection. CHMS was converted into 2,4-pyridinedicarboxylic acid (PDCA) by the addition of (NH 4 ) 2 SO 4 (33), which was then identified by HPLC/UV. Assays were performed in triplicate.
Aromatic O-demethylase enzyme assays. Enzyme assays were performed in triplicate under anaerobic conditions at 30°C (because of the expected O 2 sensitivity of H 4 folate). Recombinant LigM (Saro_2861) and DesA (Saro_2404) were separately mixed with buffer C (20 mM Na 2 HPO 4 , 20mM KH 2 PO 4 , and 1 mM or 2 mM H 4 folate; prepared anaerobically) to concentrations of 2 mM enzyme active sites under anaerobic conditions. Three independent solutions containing buffer D (20 mM Na 2 HPO 4 and 20 mM KH 2 PO 4 ) and either 1 mM 3-MGA, 1 mM PCA, or 1 mM GA were prepared aerobically. Enzyme reactions were initiated in 2-ml polypropylene vials inside an anaerobic chamber at 30°C by mixing 750 ml of an enzyme in buffer C with 750 ml an aromatic substrate in buffer D. A control was run by mixing 750 ml of buffer C without enzyme with 750 ml each aromatic substrate in buffer D. Samples (250 ml) were collected at time zero, at 30 min, and at 1, 2, 4, and 24 h, immediately mixed with 50 ml of 1 N HCl to stop the reaction, and then analyzed by HPLC-MS to quantify substrate disappearance and product formation.
Analysis of extracellular metabolites and enzyme reaction products. All culture supernatants and in vitro enzyme assay samples were filtered (0.2 mm) prior to chemical analysis. Quantitative analyses of aromatic compounds were performed on a Shimadzu triple-quadrupole liquid chromatography mass spectrometer (LC-MS) (Nexera XR HPLC-8045 MS/MS). Reverse-phase HPLC was performed using a binary gradient mobile phase consisting of solvent A (0.2% formic acid in water) and methanol (gradient profile shown in Fig. S8 in the supplemental material) and a Kinetex F5 column (2.6-mm pore size, 2.1-mm inside diameter, 150-mm length; product no. 00F-4723-AN). All compounds were detected by multiplereaction monitoring (MRM) and quantified using the strongest MRM transition (Table S2).
Identification of OMA and PDC was performed by GC-MS. Sample aliquots (150 ml) were acidified with HCl to pH ,2 and extracted with ethyl acetate (3 times, 500 ml each time). The three extraction samples were combined, dried under a stream of N 2 at 40°C, derivatized by the addition of 150 ml of pyridine and 150 ml of N,O-bis(trimethylsilyl)trifluoroacetamide with trimethylchlorosilane (99:1 [wt/wt]; Sigma), and incubated at 70°C for 45 min. The derivatized samples were analyzed on an Agilent GC-MS (GC model 7890A; MS model 5975C) equipped with a (5% phenyl)-methylpolysiloxane capillary column (Agilent model HP-5MS). The injection port temperature was held at 280°C, and the oven temperature program was held at 80°C for 1 min and then ramped at 10°C min 21 to 220°C, held for 2 min, ramped at 20°C min 21 to 310°C, and held for 6 min. The MS used an electron impact (EI) ion source (70 eV) and a single-quadrupole mass selection scanning at 2.5 Hz, from 50 to 650 m/z. The data were analyzed with the Agilent MassHunter software suite.
The product of PCA aromatic ring opening, predicted to be CHMS, was analyzed by its conversion into 2,4-pyridinedicarboxylic acid (PDCA). One hundred microliters of sample was pH neutralized by the addition of 10 ml of a 1.67 N solution of NaOH. In addition, 5 ml of a 10% solution of (NH 4 ) 2 SO 4 was added and then incubated at room temperature for 24 h. Samples were analyzed by HPLC-UV using the same HPLC conditions described above. Eluent was analyzed for light absorbance between 190 and 400 nm with a Shimadzu SPD-M20A spectrophotometer.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 1.1 MB.