Database Mining for Novel Bacterial β-Etherases, Glutathione-Dependent Lignin-Degrading Enzymes

The use of biomass as a renewable source and replacement for crude oil for the provision of chemicals and fuels is of major importance for current and future societies. Lignin, the most abundant aromatic polymer in nature, holds promise as a renewable starting material for the generation of required aromatic structures. However, a controlled and selective lignin depolymerization to yield desired aromatic structures is a very challenging task. In this regard, bacterial β-etherases are especially interesting, as they are able to cleave the most abundant bond type in lignin with high selectivity. With this study, we significantly expanded the toolbox of available β-etherases for application in lignin depolymerization and discovered more active as well as diverse enzymes than previously known. Moreover, the identification of further β-etherases by sequence database mining in the future will be facilitated considerably through our deduced etherase-specific sequence motifs.

ABSTRACT

Lignin is the most abundant aromatic polymer in nature and a promising renewable source for the provision of aromatic platform chemicals and biofuels. β-Etherases are enzymes with a promising potential for application in lignin depolymerization due to their selectivity in the cleavage of β-O-4 aryl ether bonds. However, only a very limited number of these enzymes have been described and characterized so far. Using peptide pattern recognition (PPR) as well as phylogenetic analyses, 96 putatively novel β-etherases have been identified, some even originating from bacteria outside the order Sphingomonadales. A set of 13 diverse enzymes was selected for biochemical characterization, and β-etherase activity was confirmed for all of them. Some enzymes displayed up to 3-fold higher activity than previously known β-etherases. Moreover, conserved sequence motifs specific for either LigE- or LigF-type enzymes were deduced from multiple-sequence alignments and the PPR-derived peptides. In combination with structural information available for the β-etherases LigE and LigF, insight into the potential structural and/or functional role of conserved residues within these sequence motifs is provided. Phylogenetic analyses further suggest the presence of additional bacterial enzymes with potential β-etherase activity outside the classical LigE- and LigF-type enzymes as well as the recently described heterodimeric β-etherases.

IMPORTANCE The use of biomass as a renewable source and replacement for crude oil for the provision of chemicals and fuels is of major importance for current and future societies. Lignin, the most abundant aromatic polymer in nature, holds promise as a renewable starting material for the generation of required aromatic structures. However, a controlled and selective lignin depolymerization to yield desired aromatic structures is a very challenging task. In this regard, bacterial β-etherases are especially interesting, as they are able to cleave the most abundant bond type in lignin with high selectivity. With this study, we significantly expanded the toolbox of available β-etherases for application in lignin depolymerization and discovered more active as well as diverse enzymes than previously known. Moreover, the identification of further β-etherases by sequence database mining in the future will be facilitated considerably through our deduced etherase-specific sequence motifs.

INTRODUCTION

Lignin, one of the three main constituents of plant cell walls (lignocellulose), represents the most abundant aromatic polymer on earth and a promising renewable source for the production of biofuels and aromatic chemicals (1). The latter, however, requires a selective depolymerization of lignin, which is a major obstacle due to the highly complex and heterogeneous structure of this aromatic polymer (2, 3). Lignin consists of guaiacyl (G), syringyl (S), and p-hydroxyphenyl (H) phenylpropanoid units, which are randomly connected by various C-O and C-C bonds (2, 3). The β-O-4 aryl ether bond is the most abundant one, accounting for 45 to 60% of all linkages present in lignin (3). Various chemical and thermic lignin depolymerization processes have been described (for recent reviews see Xu et al. [4] and Rinaldi et al. [5]), which usually rely on harsh reaction conditions and high energy consumption and are mostly unselective. These are some of the reasons why lignin has not been valorized on the industrial scale until now but rather used for energy and heat production (6). On the other hand, enzymes such as peroxidases, laccases, and β-etherases are described to depolymerize lignin in nature (7). Among these, bacterial β-etherases have been shown in previous studies to exhibit high activities and absolute selectivity for cleavage of β-O-4-aryl ether bonds present in various lignin model substrates as well as lignin polymers (8–12). Hence, these enzymes are highly interesting for application in lignin valorization.

β-Etherases belong to the superfamily of glutathione S-transferases (GST) and are part of the glutathione (GSH)-dependent β-O-4 aryl ether-degrading pathway, which was first discovered and investigated in Sphingobium sp. strain SYK-6 (13–17). This pathway starts with the selective oxidation of a hydroxyl group in α-position of the β-O-4 aryl ether bond catalyzed by four NAD⁺-dependent, stereospecific alcohol dehydrogenases (LigD, LigL, LigN, and LigO) (16, 18). Only after formation of this benzylic keto group are β-etherases able to cleave the adjacent β-O-4 aryl ether bond in a highly stereoselective manner by following an S_N2-type mechanism (Fig. 1) (17, 19). Hence, inversion of the stereoconfiguration at the β carbon atom is observed. While β-etherases LigE and LigP of Sphingobium sp. strain SYK-6 cleave ether bonds in substrates with (R)-configured β-carbon, resulting in the corresponding (S)-configured glutathione adducts, LigF converts the corresponding (S)-substrate enantiomers (19). Afterwards, the chiral glutathione adducts are further converted by GSH-dependent glutathione lyases LigG and SYKGST_Nu of Sphingobium sp. strain SYK-6, catalyzing thioether cleavage and resulting in the release of oxidized glutathione (GSSG) (17). While LigG displays high stereopreference for cleavage of (R)-configured thioethers, SYKGST_Nu was found to be slightly S-selective. Similar pathways were also described for some other bacteria of the order Sphingomonadales (20). Even though Cα dehydrogenases, β-etherases, and glutathione lyases are intracellular enzymes and, hence, likely act on soluble lignin degradation products in nature (12), their potential for lignin polymer degradation has been demonstrated in several studies (8, 10, 11). When applying this enzymatic pathway for lignin depolymerization, aromatic monomers such as guaiacyl hydroxypropanone (GHP) and syringyl hydroxypropanone (SHP) could be released rather selectively (10, 11).

FIG 1.

β-Etherase-catalyzed cleavage of the β-O-4 aryl ether bond present in lignin model compounds [2,6-MP-VG, β-(2,6-dimethoxyphenoxy)-α-veratrylglycerone; VN-VG, β-vanillyl-α-veratrylglycerone; MU-VG, β-(4-methylumbelliferyl)-α-veratrylglycerone].

Unfortunately, the number of known and characterized β-etherases is still rather limited. Only 14 β-etherases have been studied so far (five LigE-type and six LigF-type enzymes as well as three heterodimeric β-etherases), all originating from Sphingobium or Novosphingobium species and sharing high sequence homology (9, 14, 15, 20, 21). In LigE-type β-etherases, sequence identities vary between 56% and 85%. LigF-type enzymes seem to be separated in two distinct subgroups, LigF homologs (including LigF, LigF-NA, LigF-NS, GST4, and NaLigF-1) and NaLigF-2 (22). Sequence identities among LigF homologs vary between 56% and 96%, whereas NaLigF-2 is only 36% to 42% identical to the other LigF enzymes. Whereas classic β-etherases are homodimeric enzymes, Kontur et al. recently described a novel heterodimeric type of β-etherases displaying R-selectivity (22). Here, sequence identities among the monomers of described heterodimeric β-etherases vary between 52% and 74%.

One major obstacle in identifying putatively novel β-etherases is our limited knowledge of their catalytic mechanism and important amino acids involved in catalysis. Previous BLASTP searches performed by Picart et al. and Gall et al. provided only a few new β-etherase enzymes, all originating from Sphingobium and Novosphingobium species (9, 21). In contrast, enzymes displaying somewhat lower sequence identities to known β-etherases, such as LigP-SC from Sorangium cellulosum and RpHypGST from Rhodopseudomonas palustris, did not show the expected etherase activity (9, 21). Clearly, the unambiguous identification of novel β-etherase sequences in public databases is especially challenging due to the high abundance of related glutathione S-transferases (GST). In fact, up to 1% of all enzymes of an organism belong to the GST superfamily (EC 2.51.18) (23). Therefore, BLAST analyses will usually reveal staggering numbers of sequences related to β-etherases, requiring an additional method to distinguish true β-etherases from other GST enzymes.

Here, we decided to use peptide pattern recognition (PPR), established by Busk and Lange in 2013, to distinguish sequences encoding new β-etherase members from the majority of other GST sequences obtained in a conventional BLAST search (24). While BLASTP starts with short local alignments, which are expanded to maximum length, leading to the identification of homologs (25), PPR is a bioinformatic approach that classifies and clusters proteins on the basis of small peptide patterns instead. Thus, if enzymes share small peptide patterns, e.g., around the active site, then they might also exhibit similar activities despite low global homology. For more information on the theoretical background of PPR as well as details regarding the PPR algorithm, the reader is referred to the original publication by Busk and Lange (24). This approach has already been quite successful in predicting the functions of members of the glycoside hydrolase family with high accuracy (24). In the latter study, 118 enzymes of the GH5 class were predicted with 97% accuracy and 540 enzymes of the GH13 class with 82% accuracy. Additionally, a β-glucosidase was successfully identified from the genome of Mucor circinelloides with the help of PPR (26). Furthermore, the PPR algorithm was applied to group polysaccharide monooxygenases and glucuronoyl esterases successfully (27, 28).

After successfully applying the PPR algorithm to the results of conventional BLASTP searches with β-etherase queries, 96 putatively new β-etherase sequences belonging to the LigE and LigF subtype could be discovered. Of these, a representative set comprising 13 enzymes, six LigE-type and seven LigF-type β-etherases, has been studied regarding their activity and selectivity. Moreover, the presence of conserved residues as well as their potential functional roles have been analyzed.

RESULTS

Identification of new putative β-etherases by PPR analysis.

To identify new putative β-etherases, the PPR algorithm was applied on two separate databases which contained either LigE- or LigF-type candidate proteins from BLASTP searches. In the case of the LigE-type database, BLASTP resulted in 1,171 unique proteins with 27,967 hexamer peptides in total. PPR analysis clustered these into 11 groups, of which one group comprised all known LigE-type enzymes together with 45 new sequences (see Table S4 in the supplemental material). For LigF-type enzymes, the database consisted of 1,956 individual enzymes with 64,584 hexamer peptides in total. Here, the PPR algorithm created 54 groups, of which one group contained all known LigF-type enzymes, except for NaLigF-2, together with 51 other enzyme sequences (Table S4).

Most of the new putative enzymes have their origin in alphaproteobacteria of the order Sphingomonadales. In addition to Sphingobium or Novosphingobium species, now also Erythrobacter and Altererythrobacter species were found as host organisms. Interestingly, putative β-etherase-encoding sequences WP_104830666 (LigE-type) and WP_104831261 (LigF-type) are derived from the genome sequence of Marinicaulis flavus, a marine alphaproteobacterium of the family Parvularculaceae. Additionally, the host organism of LigF215 (GenBank accession number OGT78215) is classified as a gammaproteobacterium.

To analyze the diversity of the extended enzyme family, all sequences of the LigE- or LigF-containing PPR groups were aligned separately using webPRANK as a basis for maximum likelihood tree building by the IQ-TREE webserver (Fig. 2). Of these, six putative LigE sequences (LigE179, LigE283, LigE491, LigE760, LigE889, and LigE915) and seven putative LigF sequences (LigF008, LigF215, LigF729, LigF755, LigF921, LigF935, and LigF965) were chosen for detailed characterization of a diverse candidate subset (Table 1) and for their comparison to LigE from Sphingobium sp. strain SYK-6 and LigF-NA from Novosphingobium aromaticivorans as reference enzymes. Those newly identified sequences were selected based on their placement in the phylogenetic tree (Fig. 2) and their sequence identities to known β-etherases to ensure sufficient coverage of the observed phylogenetic diversity among LigE- and LigF-type β-etherases.

FIG 2.

Maximum likelihood trees of putatively novel as well as known β-etherases using glutathione lyase LigG-TD from Thiobacillus denitrificans as outgroup. The tree on the left includes all LigE-type enzymes, whereas the tree on the right covers all LigF-type enzymes. Previously known β-etherases are highlighted in blue, while those enzymes that were selected for further characterization within this study are marked in red. Respective GenBank accession numbers of the sequences included in both phylogenetic trees are listed in Table S4.

TABLE 1.

List of putatively novel β-etherases investigated in this study

Name	NCBI accession no.	Organism
LigE-type enzymes
LigE179	WP_046903179	Altererythrobacter atlanticus
LigE283	OJU60283	Altererythrobacter sp. strain 66-12
LigE491	WP_044331491	Sphingomonas hengshuiensis
LigE76	ODU84760	Novosphingobium sp. strain SCN 63-17
LigE889	WP_055920889	Altererythrobacter sp. strain Root672
LigE915	WP_062781915	Novosphingobium capsulatum
LigF-type enzymes
LigF008	WP_055919008	Altererythrobacter sp. strain Root672
LigF215	OGT78215	Gammaproteobacteria
LigF729	ODU83729	Novosphingobium sp. strain SCN 63-17
LigF755	WP_066854755	Sphingobium sp. strain TCM1
LigF921	WP_054529921	Erythrobacter sp. strain SG61-1L
LigF935	OJU59935	Altererythrobacter sp. strain 66-12
LigF965	WP_068075965	Novosphingobium lentum

In addition to the increased number of putative β-etherases, the sequence diversity also is increased. In our test set, the lowest sequence identities among LigE- and LigF-type enzymes are 53% and 54%, respectively. Sequence identities for the whole set of putative novel β-etherases (Table S4) go as low as 49% in both cases.

Purification and characterization of putative β-etherases.

Recombinant expression and subsequent purification by immobilized metal ion affinity chromatography based on the N-terminal His₆ tag resulted in enzyme preparations with purities of >90% (determined by SDS-PAGE; data not shown) and yielded 72 to 497 mg protein per liter of medium, except for LigF965 (31 mg liter⁻¹) (Table S2). In the case of LigF215, the general purification protocol had to be modified due to its low stability and the observed enzyme precipitation under purification conditions. To stabilize LigF215, GSH and glycerol were added to all buffers during purification (see also supporting information in Fig. S4).

For biochemical characterization, activities of the purified putative β-etherases were investigated with the lignin model compound β-(2,6-dimethoxyphenoxy)-α-veratrylglycerone (2,6-MP-VG) in high-performance liquid chromatography (HPLC) assays. All 13 selected enzymes displayed the expected β-etherase activity and cleaved the β-O-4-aryl ether bond of the lignin model compound (Table 2). LigE283, LigE889, and LigF008 even displayed two to three times higher activities than their respective reference enzymes LigE and LigF-NA, while most of the remaining enzymes (LigE179, LigE491, LigE915, LigF729, LigF755, LigF921, LigF935, and LigF965) were similarly as active as the reference enzymes. In contrast, LigE760 and LigF215 displayed only 37% and 6% activity, respectively, compared to that of LigE and LigF-NA.

TABLE 2.

Specific activity and enantioselectivity of the investigated enzymes in the cleavage of 2,6-MP-VG

Enzymea	Sp act (U mg−1)	Selectivity	E value
LigE	0.65	(R)-selective	>200
LigE179	0.94	(R)-selective	>200
LigE283	1.86	(R)-selective	>200
LigE491	0.98	(R)-selective	>200
LigE760	0.24	(R)-selective	>200
LigE889	1.67	(R)-selective	>200
LigE915	0.66	(R)-selective	>200
LigF-NA	2.06	(S)-selective	>200
LigF008	4.69	(S)-selective	>200
LigF215	0.12	(S)-selective	>200
LigF729	1.77	(S)-selective	>200
LigF755	2.20	(S)-selective	>200
LigF921	3.48	(S)-selective	>200
LigF935	2.92	(S)-selective	>200
LigF965	2.07	(S)-selective	>200

^aCorresponding GenBank accession numbers and host organisms are listed in Table S3.

The enzymes’ enantioselectivity in the kinetic resolution of 2,6-MP-VG was analyzed using chiral HPLC, and all enzymes displayed absolute enantioselectivity with E values higher than 200 (Table 2). In agreement with the selectivity of the reference enzymes, all LigE-type β-etherases were (R)-selective, whereas LigF-type enzymes converted the (S)-enantiomers of the lignin model substrate.

All enzymes were also characterized with respect to their thermal stabilities as well as temperature and pH optima. In activity assays at different pH values, all enzymes displayed activity in the range between pH 6 and pH 10 with an optimum clearly in the alkaline region at pH 9 (Table 3 and Fig. S2). Optima in the alkaline pH range have been reported for all previously known β-etherases as well, and they are very likely caused by the pK_a of the glutathione thiol (reported pK_a of 9.65 for free GSH) (9, 12, 20). Hence, a high pH facilitates deprotonation of the GSH thiol, which is a prerequisite for nucleophilic attack of the ether substrate. The optimal reaction temperature for activity was observed in a range between 20 and 40°C (Table 3 and Fig. S3). Nevertheless, many enzymes still show relevant residual activity above 40°C. As an example, for LigF935 60% relative activity still was found at 50°C.

TABLE 3.

Temperature and pH optima as well as apparent melting temperatures of investigated β-etherases

Enzymea	pH optimum	Temp optimum (°C)	Apparent melting temp (°C)
LigE	9.0	40	52.5
LigE179	9.0	30	53.8
LigE283	9.0	40	55.8
LigE491	9.0	30	51.3
LigE760	9.0	40	51.3
LigE889	9.0	30	50.3
LigE915	9.0	40	69.0
LigF-NA	9.0	30	62.8
LigF008	9.0	35	57.5
LigF215	9.0	35	42.8
LigF729	9.0	25	55.0
LigF755	9.0	20	48.5
LigF921	9.0	20	65.0
LigF935	9.0	40	67.2
LigF965	9.0	35	51.5

^aCorresponding GenBank accession numbers and host organisms are listed in Table S3.

Thermostabilities of all enzymes were investigated using the thermofluor assay to determine individual apparent melting temperatures (T_m). As a result, obtained T_m values ranged between 43°C and 69°C (Table 3). The T_m of LigF215 is 43°C, significantly lower than the T_m values of most other tested β-etherases, which possess melting points above 50°C. This lower thermal stability of LigF215 is in direct agreement with the observed low stability and precipitation of the enzyme encountered during purification.

Sequence and structure-function analysis.

As the PPR algorithm dissects enzyme sequences in small hexamer peptides and analyzes those as an indicator of functional similarity, recognized peptide patterns can aid the functional analysis of β-etherases. Overall, β-etherases consist of two domains: the N-terminal thioredoxin and the C-terminal helical domain (19, 23). In the case of LigE-type enzymes, the majority of the conserved peptides identified by the PPR algorithm are found in the N-terminal thioredoxin domain (residues 1 to 82 in LigE), which also harbors the GSH-binding site. This high level of conserved amino acids in the thioredoxin domain is also observed in the corresponding webPRANK alignment.

The conserved hexamer peptides of the LigE group can be assembled into larger motifs with putative importance and function in LigE-type enzymes. In the thioredoxin domain, three motifs were identified: GxTxSPxVWxxxxAxxHKG, RxPxIxDxG, and LDSWxIxExLD (Fig. 3A). The amino acids S21, P60, D71, and S72 (LigE numbering; highlighted in boldface above) are strictly conserved and, in the LigE crystal structure (PDB entry 4YAN), seem to interact with cosubstrate GSH (Fig. 3B). In contrast, no long conserved motifs could be identified in the C-terminal helical domain (residues 93 to 255 in LigE), but still, several amino acids, such as W105, W107, Y122, F142, and W197, are highly conserved. Unfortunately, except for GSH, the enzymes’ exact substrate binding mode is not known, as no substrate-bound crystal structures of LigE and LigF could be obtained (19). On the other hand, it is highly remarkable that many hydrophobic and especially aromatic amino acids are highly conserved in the C-terminal domain of LigE-type enzymes. With lignin being an aromatic and highly hydrophobic substrate, such conserved hydrophobic and aromatic amino acids will likely be important for substrate binding.

FIG 3.

Visualization of conserved amino acids and sequence motifs in LigE-type β-etherases. (A) Sequence logos of the conserved amino acid motifs present in LigE-type enzymes. (B) The active site of LigE (PDB entry 4YAN) with cofactor GSH (turquoise) and the docked substrate 2,6-MP-(R)-VG (magenta). Conserved amino acids S21, Y23, P60, D71, S72, W107, Y122, F142, and W197, surrounding the GSH or the substrate, are visualized in orange. Residue numbering is according to the amino acid sequence of LigE.

To better understand which amino acids are involved in these interactions, substrate 2,6-MP-VG was docked into the crystal structure of LigE. Even though none of the resulting possible binding modes seemed to resemble the productive substrate conformation, as the distance between the GSH thiolate and the substrate’s β-carbon, as well as the substrate geometry, were in each case not in agreement with an S_N2 mechanism, the conformation exhibiting the shortest S-to-C_β distance (4.9 Å) was used for further analysis. In the docked structure, amino acids Y23 (thioredoxin domain), W107, Y122, F142, and W197 show direct interactions with the substrate (Fig. 3B). W107 and Y122 are strictly conserved, whereas only hydrophobic residues are found at positions 23 (F and Y), 142 (F, L, and W), and 197 (F and W). In addition to these residues with clear substrate interaction, residue W105 is conserved in every LigE-type enzyme but only flanks the substrate-binding region while being part of the dimer interface in LigE.

In LigF-type β-etherases, GSH binding seems to be highly conserved as well. The crystal structure of LigF (PDB entry 4XT0) indicates that amino acids N13, S14, and K16 (LigF numbering; highlighted in boldface in the following motif) of the thioredoxin domain (residues 1 to 76 in LigF) motif LYSFGPxANSxKP, which is found in nearly all LigF-type enzymes, interact with the cocrystallized cofactor GSH (Fig. 4). Also, S67 interacts with GSH and is part of motif TESTVICEYLEDxxP, which is present in all LigF-type enzymes.

FIG 4.

Visualization of conserved amino acids and sequence motifs in LigF-type β-etherases. (A) Sequence logos of the conserved amino acid motifs present in LigF-type enzymes. (B) Active site of LigF (PDB entry 4XT0) with cofactor GSH and the docked substrate 2,6-MP-(S)-VG shown in turquoise and magenta, respectively. (C) Dimer interface of LigF with chain A, shown in gray, and chain B, shown in blue. Conserved amino acids interacting with either GSH (F8, N13, S14, K16, Q53, and S67) or the substrate (W109, V111, S112, W116, I120, W149, and I200), as well as conserved residues in the dimer interface (A94, R97, and K101), are visualized in orange. Residue numbering is according to the amino acid sequence of LigF.

In contrast to the LigE-type etherases, long stretches in the helical domain (residues 93 to 242 in LigF) are highly conserved in LigF-type enzymes. In particular, the long motif AxMRxWTKWVDEYFCWCVSTxGW is striking merely due to its size. A part of this motif (residues A94, R97, and K101, highlighted in boldface in the previous motif) is located in the dimer interface and is very likely involved in interactions between both monomers (Fig. 4).

As for LigE-type enzymes, the exact substrate-binding mode of LigF-type enzymes is still unknown. Docking substrate 2,6-MP-VG into the crystal structure of LigF yielded a conformation that likely resembles the productive binding mode based on the S-to-C_β distance and substrate geometry (i.e., the leaving group is on the opposite site of the β-carbon relative to the approaching thiolate). Inspection of this substrate-bound structure revealed the strictly conserved amino acids F8 (thioredoxin domain), W109, V111, S112, W116, and W149, as well as positions 120 (I and V) and 200 (I and V), as interaction partners of the substrate. Residues W109, V111, S112, and W116 (highlighted in boldface in the following motif) belong to the second half of the aforementioned motif, AxMRxWTKWVDEYFCWCVSTxGW, underlining its importance for substrate binding as well as dimerization of LigF-type enzymes. Additionally, several hydrophobic amino acids between positions 179 and 209, such as L179, L185, and L192, are conserved but do not form a closed motif.

This sequence and structure analysis with our increased set of β-etherases will be a good starting point for further functional analysis of these enzymes to better understand their catalytic mechanism and the functional role of conserved amino acids.

Phylogenetic analysis.

To analyze the phylogenetic relationship of β-etherases and homologous glutathione S-transferases, a new BLASTP search was performed using all previously known β-etherases and the 13 β-etherases confirmed here as queries and limiting the search to 1,000 hits per query sequence. The resulting 5,026 sequences (1,118 found by BLASTP based on LigE-type query sequences and 3,908 found by BLASTP based on LigF-type query sequences after removal of double entries) were used to construct a phylogenetic tree based on average linkage (unweighted pair group method using average linkages [UPGMA]) using the MAFFT server (Fig. 5).

FIG 5.

Phylogenetic analysis of β-etherases and homologous sequences. The phylogenetic tree was generated using MAFFT (44) based on 5,026 protein sequences obtained after standard BLASTP searches with all previously known β-etherases and 13 β-etherases confirmed here as queries (limited to 1,000 sequence hits per query; all sequences were combined and double entries were removed). LigE-type sequences are colored in green, whereas LigF-type sequences are colored in red. Sequences grouped together with NaLigF-2 by the PPR algorithm are colored in blue. The phylogenetic branches containing both sequences encoding the recently described heterodimeric β-etherase BaeAB (BaeA and BaeB) from Novosphingobium aromaticivorans are colored in magenta.

This phylogenetic analysis confirms that LigE- as well as LigF-type sequences are clearly separated from each other. Also, LigF-type enzymes and NaLigF-2 are placed on separate branches of the phylogenetic tree. The latter observation is in agreement with the significantly lower sequence homology of NaLigF-2 to other known LigF-type enzymes and the placement of NaLigF-2 in a different PPR group. Hence, LigF-type and NaLigF-2-type enzymes can be distinguished among bacterial (S)-selective β-etherases.

The phylogenetic branches covering LigE- and LigF-type enzymes contain the same sequences, which were also clustered together by the PPR algorithm. Moreover, NaLigF-2 was clustered together with 52 other homologs by the PPR algorithm, which are found on the same phylogenetic branch as NaLigF-2 (GenBank accession numbers of sequences grouped together with NaLigF-2 by the PPR algorithm are listed in Table S4). Hence, the PPR result is congruent with the phylogenetic placing of β-etherases. The same phylogenetic arm carrying LigF- and NaLigF-2-type enzymes also contains the sequences coding for the recently described (R)-selective heterodimeric β-etherases BaeAB (22). Hence, all three β-etherase types seem to originate from a common ancestor.

Extending the set of phylogenetically investigated sequences to a tree generated from 5,000 BLASTP hits per query, the same monophyletic origin of LigE- and LigF-type enzymes is found (not shown). Moreover, among all sequences used for the generation of the phylogenetic tree shown in Fig. 5, only three sequences of a characterized enzyme with reported glutathione transferase activity could be found in addition to the previously known and presently characterized β-etherases (29–31). Those three sequences refer to GSTF-3 from Zea mays (maize), one of the leading enzymes in maize herbicides detoxification. Additionally, for one sequence from Marinobacter aquaeolei VT8, the crystal structure of the corresponding protein (PDB entry 4N0V) has been solved, but without further characterization of the putative glutathione transferase.

DISCUSSION

In this study, 96 new putative LigE- and LigF-type β-etherases have been identified using the peptide pattern recognition (PPR) algorithm based on sequence data sets that had been generated by homology searches. Until now, PPR has mainly been used to group known sets of enzymes, all active on (poly)saccharides, into subgroups, as well as for the identification of a novel β-glucosidase-encoding sequence in a fungal genome (24, 26). In contrast, in this study the PPR algorithm has been applied in combination with homology searches for the identification of a large number of so-far unknown members of an enzyme group in public databases. Alternatively, when performing BLASTP homology searches in combination with subsequent phylogenetic analysis of the resulting sequence set, the same putative β-etherase-encoding sequences could be identified based on their clustering together with known β-etherase sequences in the phylogenetic tree. Hence, in our specific example, PPR and phylogenetic analysis of the same set of homologous sequences yielded the same result; therefore, the two approaches can be regarded as being equivalent. This is actually not surprising, considering the fact that two sequences sharing high sequence homology, an indication for a close phylogenetic relationship, will also share a higher number of identical peptide sequences than two distantly related sequences with lower sequence similarity. On the other hand, Busk and Lange hypothesized that their PPR algorithm could also identify new enzymes that are phylogenetically unrelated to the previously known ones while sharing the same activity (24). Such an example, however, has yet to be demonstrated. In our case, only sequences sharing highest sequence homology to known LigE- and LigF-type β-etherases were grouped together with respective known β-etherase sequences during the PPR run.

Comparing PPR and phylogenetic analysis, both offer individual benefits. Applying PPR on a large sequence data set, including thousands of sequences, is an easier and quicker approach than phylogenetic analysis, including multiple-sequence alignment and tree building of the same data set, especially for inexperienced users. Moreover, it helped in our case to distinguish between true LigE-/LigF-homologous β-etherases and other sequence-related glutathione transferases. On the other hand, phylogenetic analysis of a given data set of homologous sequences allows for grouping sequences into different subgroups according to branches of the phylogenetic tree and visualizes the phylogenetic relationship of the different subgroups (22, 32). Thus, it was discovered that LigF-, NaLigF-2-, and heterodimeric BaeAB-type β-etherases are placed on the same phylogenetic arm, also including additional branches with so-far unexplored sequences that could encode putative β-etherases (Fig. 5). A similar phylogenetic relationship between LigF-, NaLigF-2-, and BaeAB-type β-etherases was recently reported by Kontur et al. using a significantly smaller sequence set for phylogenetic analysis (22). Exact branching points for the separation of the three β-etherase types, however, differ between the two phylogenetic trees. This is not surprising, considering the low bootstrap values at the major branching points for separation of LigF-, NaLigF-2-, BaeA-, and BaeB-type sequences in the previously published phylogenetic tree (22).

The PPR algorithm further enables the automatic identification of conserved sequence patterns in sequences that were grouped together. The same patterns, in principle, can also be inferred manually from a multiple-sequence alignment of the same sequence set. Hence, different sequence motifs including strictly conserved amino acids could be obtained for LigE- and LigF-type β-etherases. These include residues S21 in LigE- as well as N13 and S14 in LigF-type enzymes (residue numbering according to LigE and LigF, respectively), which have been suggested to affect β-etherase activity by improved deprotonation of the GSH thiol, a prerequisite for nucleophilic attack at the β-carbon of the β-O-4 aryl ether bond (19, 22). Accordingly, mutagenesis of S21 in LigE and S14 in LigF, as well as mutagenesis of N14 in the heterodimeric β-etherase BaeAB from N. aromaticivorans (corresponding to N13 in LigF), resulted in significantly reduced to almost complete loss of enzyme activity (19, 22). Further investigation is required to elucidate the structural and/or functional roles of other conserved residues of LigE- and LigF-type β-etherases highlighted in the present study.

The sequence motifs of LigE- and LigF-type β-etherases identified here are also useful tools for the future identification of LigE and LigF homologs in public databases. Using the three combined motifs GxTxSPxVWxxxxAxxHKG, RxPxIxDxG, and LDSWxIxExLD with query LigE or LYSFGPxANSxKP, TESTVICEYLEDxxP, and AxMRxWTKWVDEYFCWCVSTxGW with query LigF in PHI-BLAST searches (33) on the nr database of GenBank (release 232), the same (putative) β-etherase sequences were obtained that had previously been found in the respective PPR groups of LigE- and LigF-type β-etherases. Moreover, 17 additional LigE- and 17 additional LigF-type sequences (listed in Table S5 in the supplemental material) were retrieved that constitute new putative members of both etherase types that had been included in the updated GenBank release.

Our biochemical characterization of 13 new β-etherases (6 LigE-type and 7 LigF-type enzymes) especially revealed enzymes from Altererythrobacter and Erythrobacter spp. (LigE283, LigE889, LigF008, LigF921, and LigF935) to be highly active on the tested lignin model substrates. Previously, LigE from Sphingobium sp. strain SYK-6 and LigF-NA from N. aromaticivorans were reported to display highest specific activities among the known β-etherases (9). These new enzymes are two to three times more active than LigE and LigF-NA. Indications for the presence of β-O-4 aryl ether-cleaving enzymes in Erythrobacteraceae have been reported previously. Palamuru et al. demonstrated that a bacterial isolate belonging to the family Erythrobacteraceae could grow on the lignin model compound guaiacylglycerol-β-guaiacyl-ether as a sole carbon source (34). Additionally, Cα dehydrogenases, catalyzing the first step in the GSH-dependent β-O-4 aryl ether degradation pathway by oxidation of the Cα-hydroxyl group, have been identified in this bacterium (34). Hence, beside β-etherases, as demonstrated in this study, it is expected that sequences encoding glutathione lyases of the Omega or Nu class of glutathione S-transferases will be present in Erythrobacter and Altererythrobacter spp. as well to complete the GSH-dependent pathway for degradation of lignin-derived β-O-4 aryl ethers (35–37). In fact, genome analyses revealed that in all ten bacterial strains from which a β-etherase had been characterized in this study, both Cα dehydrogenase and Nu-class glutathione lyase homologs were present, whereas only six genomes also contained a homolog of an Omega-class glutathione lyase.

Beside sequences originating from bacteria of the families Sphingomonadaceae and Erythrobacteraceae (both order Sphingomonadales), we identified LigE- and LigF-type β-etherases in the marine alphaproteobacterium Marinicaulis flavus, belonging to the family Parvularculaceae (order Parvularculales) (38). Hence, in contrast to previous assumptions, this is an indication that enzymes with β-etherase activity also can be found in bacteria outside the order Sphingomonadales (21). Homology searches in the genome of M. flavus further revealed the presence of sequences encoding homologs of the Nu-class glutathione lyases and Cα dehydrogenases. Therefore, we hypothesize that M. flavus harbors a functional GSH-dependent β-O-4 aryl ether degradation pathway. In addition, LigF215 also seems to originate from a nonsphingomonad bacterium, as the gene was found in a metagenome-assembled genome (MAG) from environmental DNA, which has been classified as a gammaproteobacterium. This classification, however, should be taken with great care, as it has been shown that such MAGs might contain substantial contamination, leading to less reliable phylogenetic binning (39). Moreover, more than 3% of the respective contig carrying the LigF215 gene contains multiple gaps between 26 and 189 nucleotides, indicating a rather low assembly quality.

To conclude, using database mining, the number of LigE- and LigF-type β-etherases was significantly expanded in this study. Although only 13 of the 96 putative enzymes have been investigated here, confirming their expected β-etherase activity, we can safely assume that the other 83 putative enzymes will exhibit β-etherase activity too, considering their placement together with known LigE- and LigF-type enzymes in the phylogenetic tree and their grouping by PPR. As some of the enzymes even display significantly higher activities than previously known β-etherases, it will be interesting to study their activity together with other enzymes of the GSH-dependent β-O-4 aryl ether degradation pathway (Cα dehydrogenases and glutathione lyases) for lignin depolymerization in the future. β-Etherases are especially useful for application in lignin valorization due to their selectivity for cleavage of the β-O-4 aryl ether bond and their non-radical-based mechanism (8, 10, 11).

Moreover, our phylogenetic data are also a great starting point to investigate putative enzymes with potential activity for β-O-4 aryl ether cleavage beside the classic LigE- and LigF-type enzymes. The NaLigF-2 homologs, also grouped together by the PPR algorithm, as well as enzymes from LigE- and LigF-neighboring branches of the phylogenetic tree could reveal interesting activities and enzymatic properties. In particular, the subtree with LigF-, NaLigF-2-, and BaeAB-type enzymes seems to be very interesting, as (R)- and (S)-selective β-etherases are located here.

MATERIALS AND METHODS

Chemicals.

All chemicals were of analytical grade or higher quality and purchased from Sigma-Aldrich (Steinheim, Germany), Carl Roth (Karlsruhe, Germany), Acros Organics (Geel, Belgium), VWR Chemicals (Darmstadt, Germany), or Alfa Aesar (Karlsruhe, Germany). The synthesis of lignin model substrates for β-etherase reactions was performed according to published protocols, with minor modifications (9). Detailed information is given in the supplemental material.

Identification of putative β-etherases.

Genes coding for new putative β-etherases were identified by using the peptide pattern recognition (PPR) tool provided by Busk and Lange (24). To generate an input database for the PPR tool, two separate databases for LigE-type and LigF-type enzyme candidates were created using BLASTP, with LigE, LigE-NA, LigF-NS, LigE-NS, and GST5 as well as LigF, LigF-NA, and GST4 (Table S3), respectively, as queries for homology searches in the nr database of GenBank (GenBank release 226). One thousand sequences from each BLASTP search were combined and yielded the respective LigE-type and LigF-type database after removal of duplicate entries. Both databases were used as input for PPR runs with default parameters.

Strains and plasmids.

Synthetic genes were ordered from Eurofins Genomics (Ebersberg, Germany) after codon optimization for heterologous expression in Escherichia coli and cloned into expression vector pET-28a according to standard protocols using restriction enzymes NdeI, HindIII, and T4 DNA ligase (all from NEB, Frankfurt, Germany). Respective gene and protein sequences of all newly characterized β-etherases are listed in the supplemental material.

The strains E. coli XL-1 Blue and E. coli BL21(DE3) Gold (Agilent Technologies, Waldbronn, Germany) were used for cloning and expression, respectively.

Enzyme expression and purification.

For expression, 2 ml of an E. coli BL21(DE3) Gold overnight culture carrying one of the resulting β-etherase genes in pET-28a was used to inoculate 0.25 liters of terrific broth (TB) medium with kanamycin (50 mg liter⁻¹) and 0.1 mM isopropyl-β-d-thiogalactopyranoside (IPTG). Expression cultures were incubated for 24 h at 20°C with shaking (200 rpm). Cells were harvested by centrifugation for 10 min at 6,000 × g. The resulting pellet was frozen at –20°C until further use.

For purification of β-etherases, cell pellets from expression cultures were resuspended in 20 ml binding buffer (20 mM potassium phosphate buffer, pH 7.4, 500 mM NaCl, 20 mM imidazole) and lysed by sonification (4-min active pulsing, 2-s pulse, 5-s pause, 60% amplitude). Afterwards, cell debris was removed by centrifugation (18,000 × g, 30 min) and subsequent filtration (0.45-μm pore size). The resulting cell extract was applied to a 5-ml HisTrap-FF column (GE-Healthcare, Berlin, Germany) using an ÄKTA start system (GE-Healthcare, Pittsburgh, PA) with a flowrate of 2 ml min⁻¹. Weakly bound proteins were eluted from the column by washing with 10 column volumes (CV) of binding buffer and then 5 CV of 2.5% elution buffer (20 mM potassium phosphate buffer, pH 7.4, 500 mM NaCl, 500 mM imidazole) in binding buffer at a flow rate of 5 ml min⁻¹. Target protein was eluted with 70% elution buffer in binding buffer at a flow rate of 1 ml min⁻¹ (fraction size, 1 ml). The collected protein fractions were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis with Coomassie brilliant blue G-250 staining. Fractions containing the target protein were combined, concentrated (molecular weight cutoff, 10 kDa; Vivaspin Turbo 15; Sartorius, Göttingen, Germany), and desalted via PD-10 columns (GE-Healthcare, Berlin, Germany) with 20 mM Tris buffer, pH 7.5, containing 20% glycerol. Protein concentration was determined spectrophotometrically (NP80 nanophotomer; Implen, Munich, Germany) by absorption at 280 nm. The purified proteins were stored at –80°C.

In the case of LigF215, the purification protocol was modified to a three-buffer system. For cell lysis, sample loading, and initial washing, standard binding buffer (buffer A) was used. The buffer on the column then was exchanged by buffer B (20 mM Tris, pH 8, 100 mM NaCl, 2 mM GSH, 20% glycerol) using 5 CV and a flow rate of 5 ml min⁻¹. Subsequently, weakly bound proteins were eluted from the column using 5 CV of 2.5% buffer C (20 mM Tris, pH 8, 100 mM NaCl, 2 mM GSH, 20% glycerol, 500 mM imidazole) in buffer B at a 5-ml min⁻¹ flow rate. Finally, the His-tagged protein was eluted in 1-ml fractions using 25% buffer C in buffer B at a flow rate of 1 ml min⁻¹. Fractions were collected in 2-ml reaction tubes that already contained 1 ml of buffer B for direct protein dilution.

Enzyme characterization.

To determine β-etherase activity by HPLC, 1-ml reactions were performed containing 0.4 mM racemic 2,6-MP-VG (dissolved in dimethyl sulfoxide [DMSO]; end concentration, 10%) and 5 μg enzyme in 100 mM glycine buffer, pH 9, and 1 mM GSH at 25°C and 800 rpm. Samples for HPLC measurements were taken after 1, 3, 5, 7, 9, and 11 min. Reactions were stopped by adding sulfuric acid (5% end concentration), and precipitated protein was removed by centrifugation (13,000 × g, 10 min). Enzyme activity was calculated based on the production of 2,6-dimethoxyphenol detected at 280 nm using a calibration curve of commercial 2,6-dimethoxyphenol.

For enantioselectivity determination, similar reactions were performed, but substrate was dissolved in isopropanol (5% end concentration) and reactions were stopped after 4 h. Enantioselectivity was calculated based on the conversion of 2,6-MP-VG and substrate enantiomeric excess according to Chen et al. (40).

To determine the pH optimum of each β-etherase, a fluorescence assay based on substrate β-(4-methylumbelliferyl)-α-veratrylglycerone (MU-VG) was used. Reactions were performed in 96-well flat-bottom microtiter plates (Sarstedt, Nümbrecht, Germany) and were measured in a CLARIOstar 96-well microplate reader (BMG Labtech, Ortenberg, Germany). The reaction volume was 200 μl, containing 0.1 mM racemic MU-VG (dissolved in DMSO; end concentration, 10%) and 1 mM GSH in 100 mM buffer at 25°C. For pH 5 an acetate buffer, for pH 5 and 6 a phosphate buffer, for pH 8 a Tris buffer, for pH 9, 9.5, and 10 a glycine buffer, and for pH 11 a carbonate buffer was used. Additionally, a mixture of Tris and glycine buffer (100 mM Tris plus 100 mM glycine) was tested for pH 8.5, 9, and 9.5. For LigF-NA, LigF008, LigF729, LigF755, LigF921, LigF935, and LigF965, the enzyme concentration was 25 μg ml⁻¹; for LigE, LigE179, LigE283, LigE491, LigE889, and LigE915, the enzyme concentration was 300 μg ml⁻¹; for LigE760 and LigF215, an enzyme concentration of 400 μg ml⁻¹ was used. The increase in fluorescence due to formation of 4-methylumbeliferone was measured using an excitation wavelength of 360 nm (±10 nm) and an emission wavelength of 450 nm (±15 nm). For each reaction, the slope of the first 5 min was determined, and relative activities were calculated by setting the highest slope of each enzyme to 100%.

Temperature optima of all β-etherases were determined using an absorption assay based on the substrate β-vanillyl-α-veratrylglycerone (VN-VG). Reactions were performed in microcuvettes (Brand, Wertheim, Germany). The reaction volume was 0.4 ml, containing 0.5 mM racemic VN-VG (dissolved in DMSO; end concentration, 10%) and 1 mM GSH in 100 mM glycine buffer, pH 9. The amount of enzyme per reaction mixture was 1 μg ml⁻¹ for LigF008 and LigF935; 4 μg ml⁻¹ for LigE, LigE283, LigE491, LigE889, LigE915, LigF-NA, LigF729, LigF755 LigF921, and LigF965; 8 μg ml⁻¹ for LigE179; 10 μg ml⁻¹ for LigF215; and 50 μg ml⁻¹ for LigE760. Reaction temperature varied from 10 to 50°C using steps of 5°C. Before starting a reaction, all reaction components were incubated separately for 5 min at the desired reaction temperature. Product formation was measured for 7.5 min in a Cary 60 UV-visible spectrophotometer (Agilent Technologies, Santa Clara, CA, USA) by following the absorbance at 360 nm. The slope of the first 2 min was used to determine a relative activity value compared to the highest slope obtained with the same enzyme, which was set to 100%.

Apparent melting temperatures (T_m) of β-etherases were determined by thermal unfolding using the thermofluor assay (also called the thermoshift assay) (41). Each 50-μl sample was composed of 40 μl 20 mM Tris buffer, pH 7.5, containing 20% glycerol, 5 μl 50× SYPRO Orange (Thermo Fisher Scientific, Waldbronn, Germany), and 5 μl purified protein solution of 2 mg ml⁻¹ concentration in an iQ 96-well real-time PCR plate (Bio-Rad Laboratories, Munich, Germany). Dilution of proteins to the desired concentration of 2 mg ml⁻¹ was performed using the respective storage buffer, except for LigF215, for which deionized water was used. The plate was sealed and measured in a reverse transcription-PCR machine (CFX96 real-time PCR detection systems; Bio-Rad Laboratories, Munich, Germany) over a linear gradient from 10 to 90°C in 0.5°C steps using an excitation wavelength of 490 nm and an emission wavelength of 575 nm. The final T_m was derived from the local minimum of the negative first derivative of the measured relative fluorescence plotted versus the temperature.

To determine optimal buffer conditions for LigF215 for purification and storage, increasing concentrations of glycerol and GSH in 20 mM Tris buffer, pH 7.5, as well as increasing concentrations of elution buffer (20 mM potassium phosphate buffer, pH 7.4, 500 mM NaCl, 500 mM imidazole) in binding buffer (20 mM potassium phosphate buffer, pH 7.4, 500 mM NaCl, 20 mM imidazole) were tested (see Fig. S4 in the supplemental material).

Phylogenetic and structure-function analysis.

Putative β-etherase sequences, grouped by the PPR analysis, were aligned using webPRANK (42), and phylogenetic trees were generated using the IQ-TREE web server (43).

A large sequence data set for phylogenetic analysis was generated based on BLASTP searches in the nr database of GenBank (GenBank release 226) using all known and presently confirmed LigE-, LigF-, and NaLigF-2-type β-etherases as queries (24 sequences in total; Table S3). After removal of duplicate entries, the resulting data set was aligned and the phylogenetic tree was generated based on average linkage (UPGMA) using the MAFFT server (FFT-NS-2) (44). Logos of conserved amino acid motifs were generated using WebLogo (45).

Docking of 2,6-MP-VG in the crystal structures of LigE (PDB entry 4YAN) and LigF (PDB entry 4XT0) was performed using AutoDock Vina (46) implemented in YASARA structure (http://www.yasara.org/) with 999 docking runs.

HPLC analyses.

Samples for enzyme activity measurements were analyzed on a Nexera XR20 HPLC system (Shimadzu, Duisburg, Germany) equipped with an achiral Nucleosil C₁₈ column (length, 250 mm; inner diameter, 5 mm; particle size, 5 μm; Macherey-Nagel, Düren, Germany). As a mobile phase, an isocratic mixture of water, acetonitrile, and TFA (49.95/50/0.05, vol/vol/vol) was used with a flow rate of 1 ml min⁻¹.

Chiral HPLC analysis was also performed on a Nexera XR20 HPLC system using a Chiralcel OD-RH column (length, 150 mm; inner diameter, 4.6 mm; particle size, 5 μm; Daicel, Illkirch, France). As a mobile phase, an isocratic mixture of water and acetonitrile (70/30, vol/vol) was used with a flow rate of 1 ml min⁻¹.