High-Throughput Generation of Product Profiles for Arabinoxylan-Active Enzymes from Metagenomes

Arabinoxylan is mainly found in the hemicellulosic fractions of rice straw, corn cobs, and rice husk. Converting arabinoxylan into (arabino)xylo-oligosaccharides as added-value products that can be applied in food, feed, and cosmetics presents a sustainable and economic alternative for the biorefinery industries. Efficient and profitable AX degradation requires a set of enzymes with particular characteristics. Therefore, enzyme discovery and the study of substrate preferences are of utmost importance. Beavers, as consumers of woody biomass, are a promising source of a repertoire of enzymes able to deconstruct hemicelluloses into soluble oligosaccharides. High-throughput analysis of the oligosaccharide profiles produced by these enzymes will assist in the selection of the most appropriate enzymes for the biorefinery.

ABSTRACT

Metagenomics is an exciting alternative to seek carbohydrate-active enzymes from a range of sources. Typically, metagenomics reveals dozens of putative catalysts that require functional characterization for further application in industrial processes. High-throughput screening methods compatible with adequate natural substrates are crucial for an accurate functional elucidation of substrate preferences. Based on DNA sequencer-aided fluorophore-assisted carbohydrate electrophoresis (DSA-FACE) analysis of enzymatic-reaction products, we generated product profiles to consequently infer substrate cleavage positions, resulting in the generation of enzymatic-degradation maps. Product profiles were produced in high throughput for arabinoxylan (AX)-active enzymes belonging to the glycoside hydrolase families GH43 (subfamilies 2 [MG43₂], 7 [MG43₇], and 28 [MG43₂₈]) and GH8 (MG8) starting from 12 (arabino)xylo-oligosaccharides. These enzymes were discovered through functional metagenomic studies of feces from the North American beaver (Castor canadensis). This work shows how enzyme loading alters the product profiles of all enzymes studied and gives insight into AX degradation patterns, revealing sequential substrate preferences of AX-active enzymes.

IMPORTANCE Arabinoxylan is mainly found in the hemicellulosic fractions of rice straw, corn cobs, and rice husk. Converting arabinoxylan into (arabino)xylo-oligosaccharides as added-value products that can be applied in food, feed, and cosmetics presents a sustainable and economic alternative for the biorefinery industries. Efficient and profitable AX degradation requires a set of enzymes with particular characteristics. Therefore, enzyme discovery and the study of substrate preferences are of utmost importance. Beavers, as consumers of woody biomass, are a promising source of a repertoire of enzymes able to deconstruct hemicelluloses into soluble oligosaccharides. High-throughput analysis of the oligosaccharide profiles produced by these enzymes will assist in the selection of the most appropriate enzymes for the biorefinery.

INTRODUCTION

Metagenomic studies of the microbial communities associated with plant cell wall degraders reveal a large number of gene sequences coding for potential carbohydrate-active enzymes (CAZymes) (1). Accurate functional analysis of new CAZymes must accompany this continuous discovery at the genomic level. Such newly functionally validated enzymes can be applied in the biorefinery industry and/or for further protein engineering (2–5). Glycoside hydrolases (GHs), which cleave polysaccharide main chains and/or substituents, can have complex substrate preferences, often showing multisubstrate specificities (6–9). These substrate preferences are often also dependent on main chain substituents. Consequently, the degradation of complex carbohydrate polymers often requires the synergistic or combinatorial action of multiple enzymes. Tedious techniques, long analysis times, demanding hands-on assays, specialized equipment, and lack of appropriate representative substrates contribute to the existing gap between enzyme discovery and functional characterization (10). Accordingly, high-throughput (HT) techniques that can deal with a large number of metagenome-derived putative enzymes and that can give insight into the substrate specificities of unannotated enzymes in a relatively short time are required (11). DNA sequencer-aided fluorophore-assisted carbohydrate electrophoresis (DSA-FACE) offers an interesting approach to study the substrate specificities of CAZymes, primarily due to the possibility of using substrates that represent the natural carbohydrates instead of artificial aryl glycoside substrates like p-nitrophenyl- and 4-methylumbelliferyl derivatives (12). In fact, the latter substrate derivatives may mask the real enzymatic substrate specificity due to, for example, steric differences in comparison to the natural oligosaccharides. DSA-FACE has shown outstanding oligosaccharide resolution and sensitivity (with detection limits ranging from 38 to 55 pM for the substrates studied) and short hands-on time and analysis time (13). In addition, DSA-FACE allows analysis of the substrate specificities in HT when using multiple parallel capillaries as in standard capillary sequencing devices.

In this work, we focus on arabinoxylan (AX), which is a hemicellulosic polysaccharide that may contain a range of substitutions, including α-l-arabinofuranosyl, α-d-glucuronic acid, 4-O-methyl-α-d-glucuronic acid, α-d-galactopyranose, and ferulic acid residues, depending on the source (14). Deconstruction of AX by GHs leads to useful sugars for the production of bioethanol, food, and added-value nutraceutical products like xylitol (15–18) and prebiotics (19–23). GHs with diverse substrate specificities for the degradation of complex AX structures are being discovered continuously and annotated in the various protein databases (24). The Carbohydrate-Active enZYmes Database (CAZy) classifies GHs into families according to amino acid sequence similarity (25). Although structural similarity often correlates with enzyme substrate specificity, the CAZy family division cannot always be used to predict enzyme substrate specificity because of the different specificities assigned per GH family. Additionally, at this time, only approximately 1% of the annotated GHs in CAZy have been experimentally characterized (25). Due to the abundance of glycoside hydrolase family 43 (GH43) members and the large variety of substrate specificities found within this family, the GH43 family was further divided into subfamilies on the basis of sequence analyses, suggesting that the correlation between functional annotation of the enzyme and subfamily assignment is more accurate (26). The GH43 family subdivision and the analysis of its members rely on computational and experimental data, which are mainly based on synthetic substrates like p-nitrophenyl (pNP) monosaccharides. To obtain a more accurate understanding of the substrate specificity of AX-active enzymes, we have recently used DSA-FACE to analyze the hydrolysates of AX-active enzymes with natural representative (arabino)xylo-oligosaccharides [(A)XOS] (13). In this work, DSA-FACE is used to elucidate the substrate specificities of enzymes derived from metagenomic studies on the North American beaver (Castor canadensis) fecal microbiome (27). Three genes belonging to the uncharacterized or poorly characterized subfamilies 2, 7, and 28 of GH43 (MG43₂, MG43₇, and MG43₂₈ where MG stands for metagenomic, followed by the glycoside hydrolase family and subfamily numbers [in subscript]) and one from GH8 (MG8) that were identified in active fosmids were shown to be active on AX by preliminary functional screening tests with aryl glycosides and/or high-performance anion exchange chromatography with pulsed amperometric detection (HPAEC-PAD) (Fig. 1) (27). To gain further insight into patterns of AX degradation by the aforementioned enzymes, we introduce DSA-FACE product profiles and associated sequential degradation maps for a convenient representation of the activity(ies) of each enzyme on 12 (A)XOS substrates. By implementing DSA-FACE product profiles, we reveal the preferred sites of substrate cleavage by a GH43₂₈ member, the modular specificity of a GH43₂-GH8 enzyme, and the dependence of the enzymatic activity of a GH43₇ member on the activity of the aforementioned GH43₂-GH8 enzyme.

FIG 1.

Preceding functional screening of putative enzymes derived from metagenomics on beaver fecal samples. (A) Upon environmental sample collection and genomic DNA (gDNA) extraction, a metagenomic DNA library of 4,500 clones suitable for heterologous expression was constructed. These clones were expressed and checked for active hits by high-throughput preliminary functional screening methods. Fifty-one active hits were sequenced, and 135 putative glycoside hydrolases (GHs) from 28 GH families were identified by in silico analysis. (B) Three GH43 genes, one of which is modular with an additional GH8 domain, and two mutants thereof were characterized by enzymatic activity tests with aryl glycosides and by HPAEC-PAD using representative arabinoxylan oligosaccharides (27).

RESULTS

DSA-FACE allows rapid evaluation of product profiles of metagenome-derived AX-active enzymes.

In this study, we used DSA-FACE to set up product profiles for three newly discovered enzymes of the GH43 family active on AX. These enzymes were identified through a preceding metagenomic analysis of the North American beaver (Castor canadensis) fecal microbiome (27). The first selected enzyme (MG43₂₈) contains a GH43 subfamily 28 domain. The second enzyme (MG43₂-8) is a modular enzyme composed of a GH43 subfamily 2 and a GH8 domain. To differentiate the specificities of each domain, we also set up product profiles of the two mutated variants, in which the respective domains are inactivated by mutagenesis of the catalytic acid residue (27), resulting in MG43₂ and MG8, respectively (Table 1). The third enzyme, MG43₇, comprises a single domain assigned to GH43 subfamily 7. These enzymes were selected because of the limited characterization of these subfamilies. An initial analysis of these enzymes with chromogenic substrates and HPAEC-PAD confirmed that they had AX-acting activities, as summarized in Table 1. The initial low expression yields of MG43₂-8, MG43₂, and MG43₇ were optimized by variation of the expression strains, growth temperature, and induction and purification protocols (Table 2). Eluted fractions containing the desired proteins obtained from the different expression/purification procedures were pooled for further analysis.

TABLE 1.

Metagenomic AX-active enzymes analyzed in this study^a

^aThe CAZy family (and subfamily in subscript) is given for the modules that constitute each enzyme. Domains inactivated by mutagenesis are indicated with a red slash mark. Activity tests on aryl glycosides p-nitrophenyl β-d-xylopyranoside (pNP-X), 4-methylumbelliferyl β-d-xylopyranoside (MU-X), 6-chloro-4-methylumbelliferyl β-d-xylopyranoside (CMU-X), p-nitrophenyl α-l-arabinofuranoside (pNP-Ara), and 4-methylumbelliferyl α-l-arabinofuranoside (MU-Ara) and HPAEC-PAD analysis using A³X, A²XX, XA³XX, and XA²XX as substrates were performed previously (27). The proteins studied in this work are named MG (metagenomic) followed by the number of the glycoside hydrolase family (and subfamily) number. The second column refers to the code names used in a preceding study (27). CBM, carbohydrate-binding module; Rex, reducing-end xylose-releasing exo-oligoxylanase.

TABLE 2.

Expression conditions of the metagenome-derived enzymes studied in this work

Enzyme	Expression host	Host growth and induction	Purification method
MG4328	E. coli BL21(DE3)	24 h at 37°C, 250 rpm on LBE-5052 autoinduction medium	HisGraviTrap
MG432-8	E. coli BL21 CodonPlus(DE3)	37°C, 250 rpm; 1 mM IPTG at 16°C for 18 to 20 h, 250 rpm	HisPur Ni-NTA Superflow agarose (250/100 μl resin)
	E. coli ArcticExpress	30°C, 250 rpm; 1 mM IPTG at 16°C for 24 h, 250 rpm	HisPur Ni-NTA Superflow agarose (100 μl resin)
MG432	E. coli BL21 CodonPlus(DE3)	30°C/37°C, 250 rpm; 1 mM IPTG at 16°C for 18 to 20 h, 250 rpm	HisPur Ni-NTA Superflow agarose (500/100 μl resin)
MG437	E. coli BL21 CodonPlus(DE3)	18 h at 30°C, 250 rpm on LBE-5052 autoinduction medium	HisGraviTrap
		30°C, 250 rpm; 1 mM IPTG at 16°C for 18 h, 250 rpm	HisPur Ni-NTA Superflow agarose (500 μl resin)
MG8	E. coli BL21(DE3)	37°C, 250 rpm; 1 mM IPTG at 16°C for 18 h, 250 rpm	HisPur Ni-NTA Superflow agarose (500 μl resin)

DSA-FACE product profiles are a qualitative representation of the carbohydrates present after enzymatic reactions with different (A)XOS (Fig. 2A). Twelve oligosaccharides (5 XOS and 7 AXOS) were used in these enzymatic reactions. Reaction mixture hydrolysates were then analyzed with DSA-FACE. The output of DSA-FACE is electropherograms that need to be interpreted by referencing to (A)XOS standards to reveal the identity of the resulting products. Peak areas are quantified and normalized to calculate the relative conversion of each substrate and relative proportions of the products. This is exemplified in Fig. 2B for a previously characterized α-l-arabinofuranosidase from Bifidobacterium adolescentis (BaAXH-d3) (13). α-l-Arabinofuranosidases are classified into arabinoxylan arabinofuranohydrolases (AXH) that hydrolyze the O-2 and/or O-3 arabinofuranosyl monomers from the doubly substituted xyloses (AXH-d2, AXH-d3, and AXH-d2,3) or from the monosubstituted xyloses (AXH-m2, AXH-m3, and AXH-m2,3). To simplify the representation of enzymatic substrate preferences, we introduce here product profiles with a fixed color code instead of electropherograms as the final DSA-FACE outcome. Substrate conversions are easily observed in the product profiles by a color change. Accordingly, there is only a color change in the cases of A²⁺³XX and XA²⁺³XX for BaAXH-d3. Next to the product profiles, degradation maps highlighting cleavage positions for the different (A)XOS tested are elaborated to assist in evaluating enzyme substrate preferences.

FIG 2.

Product profiles and degradation maps. (A) General approach to establish DSA-FACE product profiles and a corresponding degradation map in four steps. (B) Product profiles for BaAxhd3 from Bifidobacterium adolescentis after reaction with A²XX, A²⁺³XX, XA²XX, XA³XX, and XA²⁺³XX. (1) Electropherograms a, c, e, and g show the substrate blanks, whereas electropherograms b, d, f, and h show the corresponding hydrolysates upon enzymatic reaction with BaAxhd3. The peaks are compared to standards for carbohydrate peak identification. Electropherogram i corresponds to the enzyme blank reaction mixture. (2) A qualitative interpretation of the electropherograms is then displayed on a product profile (bars are labeled with the letters of corresponding electropherograms). Substrate conversions are easily observed by a color change. The first bar corresponds to the substrate blank and is followed by bar(s) showing colors corresponding to the (A)XOS found upon enzymatic reaction. (3) A degradation map is obtained from the different product profiles for BaAxhd3.

Different product profiles for increasing MG43₂₈ concentrations.

Purified MG43₂₈ was tested against the panel of five XOS and seven AXOS using four different enzyme concentrations (0.3, 1, 6, and 32 μM). The product profiles for MG43₂₈ showed diverse hydrolytic products that changed depending on the enzyme concentration (Fig. 3). At the lowest concentration tested (0.3 μM), MG43₂₈ completely removed an O-2 arabinose from a non-reducing-end singly substituted xylose, i.e., A²XX to X₃ (AXH-m2 activity). In addition, at concentrations at or above 0.3 μM MG43₂₈, both O-2 and O-3 arabinoses were partially removed from the non-reducing-end doubly substituted xylose of A²⁺³XX, resulting in X₃ (AXH-d2,3 activity), but higher concentrations (6 μM) were needed for a full conversion. The internal O-2 arabinose from XA²XX was only removed at concentrations at or above 1 μM, and even at the highest concentration studied (32 μM), MG43₂₈ was not able to remove O-2 and O-3 arabinoses from XA²⁺³XX, indicating that the internal single/double arabinose substitutions were less accessible for hydrolysis. Removal of the O-3 arabinoses from A³XX and XA³XX (AXH-m3 activity), resulting in X₃ and X₄, respectively, was observed at or above 1 μM MG43₂₈. MG43₂₈ thus exerted diverse arabinofuranosidase specificities, with the best conversion of A²XX. Xylanolytic activity was visible at concentrations at or above 1 μM as well, with a preference for the longest xylo-oligosaccharide tested (X₆). Also, AXOS with removed arabinoses were further partially degraded (e.g., XA³XX was converted to X₂ and X₄).

FIG 3.

Product profiles starting from 12 (A)XOS and degradation map for MG43₂₈. (A) The product profiles show the hydrolysis products obtained after 22 h of enzymatic reactions with 0.3, 1, 6, and 32 μM MG43₂₈ (1, 2, 3, and 4, respectively). The (A)XOS used as substrates for the enzymatic reactions are identified as “s.” (B) A degradation map is presented, with a schematic representation of the (A)XOS structures used as substrates and the ones obtained as hydrolysis products, using corresponding colors. Based on hydrolysis products obtained, cleavage positions for MG43₂₈ are indicated (see key at bottom left).

The product profiles for MG43₂-8 are the sum of the product profiles of its respective domains.

Enzymatic reactions with purified MG43₂, MG8, and MG43₂-8 were performed. The product profiles shown in Fig. 4 demonstrate that MG43₂ is a β-xylosidase and MG8 is a reducing-end xylose-releasing exo-oligoxylanase (Rex). At a concentration of 3 μM, MG43₂ partially converted XOS into X₂. It seemed there were more structural hindrances or product inhibition than with MG8, based on the lower observed degradation rate. MG43₂ did hydrolyze the nonreducing xylose monomers from XA²XX, XA³XX, and XA²⁺³XX (Fig. 4A), but not the nonreducing arabinose-substituted xylose monomers (A²XX, A³XX, and A²⁺³XX). At a concentration of 3 μM, MG8 fully converted X₃, X₄, and X₅ to X₂, but we did not observe visible conversion of X₂ to X. However, X₆ seemed to be converted to X₂ more slowly, due to possible hindrances in accommodating long XOS into MG8 catalytic subsites or to product inhibition, indicating a preference for smaller XOS. As is typical for Rex enzymes, MG8 required two nonsubstituted xyloses from the reducing end to hydrolyze the reducing-end xylose (28, 29). This was observed when the reducing-end xylose monomer was hydrolyzed from A²XX, A³XX, A²⁺³XX, XA²XX, XA³XX, and XA²⁺³XX, but not when A³X was used as a substrate. We also evaluated different concentrations of MG8 (0.2, 0.8, 3, and 17 μM) (Fig. S1 in the supplemental material). Notably, at the minimum MG8 concentration tested (0.2 μM), only X₃ to X₆ were partially hydrolyzed into a smaller degree of polymerization (dp) XOS, showing the preference for XOS over AXOS by MG8. At concentrations from 0.8 to 17 μM, MG8 showed the same product profiles as were obtained in the reactions performed with 3 μM enzyme (Fig. 4).

FIG 4.

Product profiles of MG43₂, MG8, and MG43₂-8. (A) The product profiles show the hydrolysis products obtained after 22 h of enzymatic reactions with 3 μM MG43₂ (1), MG8 (2), and MG43₂-8 (3). The (A)XOS used as substrates for the enzymatic reactions are identified as “s.” (B) A degradation map is presented, with a schematic representation of the (A)XOS structures used as substrates and the ones obtained as hydrolysis products, using corresponding colors. Based on hydrolysis products obtained, cleavage positions for MG43₂ and MG8 are indicated (see key at bottom left).

At a concentration of 3 μM, MG43₂-8 displayed the sum of both β-xylosidase and Rex activities (Fig. 4). When we evaluated MG43₂-8 at a higher concentration (17 μM), MG43₂-8 also exhibited α-l-arabinofuranosidase activity when hydrolyzing the mixture of XA²XX and XA³XX into A²X, A³X, and X₂ (Fig. S2). This was likely performed by the MG43₂ domain, since 17 μM MG8 did not show α-l-arabinofuranosidase activity on any of the substrates tested and 3 μM MG43₂ showed a small amount of α-l-arabinofuranosidase activity on A³X but not on A³XX or XA³XX (Fig. 4).

Product profiles of MG43₇ change with increasing MG43₇ concentration and in the presence of MG43₂-8/MG43₂/MG8.

MG43₇ activity was only detected at the highest concentration tested (38 μM). MG43₇ showed xylanase activity on X₄, X₅, and X₆ (Fig. S3). And yet, when enzymatic reactions were performed with 8 μM MG43₇ in the presence of 3 μM MG43₂-8, A³X was converted to X₂, showing an additional AXH-m3 activity, which did not happen when enzymatic reactions were performed with the same concentrations of MG43₂-8 or MG43₇ alone (Fig. 5). X₂ and X₃ could also be observed after enzymatic reactions with 8 μM MG43₇ in the presence of 3 μM MG43₂-8 and A³XX or XA³XX, again showing an additional AXH-m3 activity. Notably, O-2 arabinofuranosyl substitutions were not a substrate, since A²XX was not further hydrolyzed to X₂. When enzymatic reactions were performed with elevated concentrations (38 μM MG43₇ in the presence of 17 μM MG43₂-8) and A³X, the end product was again X₂, showing no further hydrolysis, but when the same concentrations were tested against A²XX, A²X and X₂ appeared as reaction products (Fig. S2).

FIG 5.

Product profiles of MG43_7, MG43₂-8, and MG43₇ in the presence of MG43₂-8. (A) The product profiles show the hydrolysis products obtained after 22 h of enzymatic reactions with 8 μM MG43₇ (1), 3 μM MG43₂-8 (2), and 8 μM MG43₇ in the presence of 3 μM MG43₂-8 (3). The (A)XOS used as substrates for the enzymatic reactions are identified as “s.” The dashed-line box means there was no reaction performed to test the hydrolysis of A²⁺³XX by MG43₇. (B) A degradation map is presented, with a schematic representation of the (A)XOS structures used as substrates and the ones obtained as hydrolysis products. Based on hydrolysis products obtained, cleavage positions for MG43₇ and MG43₂-8 are indicated (see key at bottom left).

To discriminate whether the additional AXH-m3 or AXH-m2 activity that appeared when combining MG43₇ and MG43₂-8 came from either MG43₇ or MG43₂-8, we again investigated the activity of MG43₇ in the presence of the derivatives of MG43₂-8 in which one domain was inactivated by mutation (MG43₂ and MG8). When A³X reacted with 8 μM MG43₇ and 3 μM MG43₂ (Fig. S4), approximately 50% A³X was converted to X₂. The same reaction mixture but in combination with 3 μM MG8 also resulted in a minor fraction of X₂ (Fig. S1). These data indicated that the AXH-m3 activity did indeed result from MG43₇. This found further support as AXH-m3 activity was also detected in both cases when a mixture of A²XX/A³XX reacted with MG43₇ in the presence of either MG43₂ (Fig. S4) or MG8 (Fig. S1). Likely MG43₇ had a preference for nonreducing O-3-substituted xyloses, since 8 μM MG43₇ and 3 μM MG8 did not hydrolyze XA³XX further into X₂. These findings are consistent with the previously identified arabinofuranosidase activity of MG43₇ on A³X detected by HPAEC-PAD (27). In sum, the scans indicated that MG43₇ showed AXH-m3 activity on small O-3 arabinose-substituted AXOS (A³X and A³XX) in the presence of MG43₂-8 and showed xylanase activity at elevated concentrations.

DISCUSSION

Functional metagenomic studies of the beaver fecal microbiome revealed enzymes from subfamilies 2, 7, and 28 of the GH43 CAZy family. Whereas subfamily 2 has two characterized α-l-arabinofuranosidases from Chitinophaga pinensis strain DSM2588 and Mucilaginibacter mallensis strain MP1X4 (3, 30), subfamilies 7 and 28 have no characterized enzymes to date. Activity on CMU-X (6-chloro-4-methylumbelliferyl β-d-xylopyranoside) indicated that MG43₂₈ is a β-xylosidase (27). However, the MG43₂₈ product profiles reveal that MG43₂₈ is able to hydrolyze all (A)XOS substrates except XA²⁺³XX, showing xylanase, AXH-m2,3, and AXH-d2,3 activities. We cannot confirm whether the observed xylanase activity is due to endo- or sequential exo-xylanase activity. If MG43₂₈ acts as a β-xylosidase, it would be expected that X₂ is also degraded to X, which does not seem to happen even at the highest concentration tested. However, monomeric xylose cannot be detected by DSA-FACE. Similarly, GH43 PcAxy43A from Paenibacillus curdlanolyticus (GH43 subfamily 35) is also unable to hydrolyze XA²⁺³XX and shows endoxylanase, β-xylosidase, AXH-d2,3, and AXH-m2,3 activities in a single catalytic domain (31), but the presence of both exo and endo activity in a single enzyme can be considered unusual. Ara 1 isolated from barley malt also has both AXH-m2,3 and AXH-d2,3 activities, and a 4-times-higher enzyme concentration is needed for conversion of XA²⁺³XX into X₄ than for conversion of A²⁺³XX (32). Besides an N-terminal GH43₂₈ domain, a C-terminal discoidin domain has been identified in MG43₂₈ using the Conserved Domain database. This discoidin domain has putative lectin-like properties, binding carbohydrates (33). To unravel how MG43₂₈ deals with such a variety of substrates, the influence of this C-terminal domain on the observed multiple activities may be investigated. A blastp analysis of MG43₂₈ against all GH43 CAZy family characterized sequences (182 characterized sequences out of 16,250) revealed an enzyme from Belliella baltica strain DSM 15883 (accession number AFL85801.1) as the most similar sequence (E value 5 × 10⁻¹⁰, with 45% query cover and 25% percentage identity). This enzyme is annotated in GH43 subfamily 31 of the CAZy database and was identified as a β-d-galactofuranosidase in the study presented in reference 3. The endo-1,4-β-xylanase from uncultured bacterium URE4 (accession number ACM91046.1) shows the maximum query cover of 85% (E value 1 × 10⁻⁴, with 23% percentage identity). This enzyme is annotated in subfamily 29 of the CAZy database. These relatively low similarities spread over different subfamilies emphasize the need for detailed analyses of the substrate specificities, as is done here for MG43₂₈.

The substrate preferences of MG43₂-8, MG43₂, MG8, and MG43₇ in the presence of MG43₂-8 were previously analyzed by HPAEC-PAD upon enzymatic reactions with 0.5 μM purified enzyme and 4 mM A³X, A²XX, and a mixture of XA²XX and XA³XX (27). Under the conditions tested, MG43₂, MG8, and MG43₇ showed β-xylosidase, Rex, and AXH-m3 activities, respectively. Due to the limited HT capacity of HPAEC-PAD, a restricted number of substrates were tested, omitting doubly substituted XOS, for example. The minor α-l-arabinofuranosidase activity of the MG43₂ β-xylosidase against A³X indicates that MG43₂ shows both β-xylosidase and AXH-m3 activities at concentrations higher than 3 μM. Bifunctional β-xylosidase and α-l-arabinofuranosidase activities have already been reported before in the GH43 CAZy family but not yet in subfamily 2. It seems these enzymes can accommodate both xylose and arabinose units in their active sites not only due to obvious structural similarities between arabinose and xylose sugars but also to rotations on the α-arabinose linkage to xylose that can resemble a β-xylose linkage in the main chain (34, 35). Accordingly, it can be questioned whether MG43₂ is actually bifunctional or misrecognizes the substrate, which can be observed at an elevated enzyme concentration.

Xylanolytic activity of MG43₂-8 and MG43₂ against X₂ remains ambiguous. DSA-FACE is not able to detect xylose and, thus, cannot detect possible X₂ degradation. Previous HPAEC-PAD analyses show that X₂ was not hydrolyzed by MG43₂-8 and MG43₂ (27). In accordance with these results, MG43₂-8 and MG43₂ were also not active against chromogenic pNP-X. However, MG43₂-8 and MG43₂ showed activity with fluorogenic CMU-X, suggesting X₂ xylanolytic activity. This discrepancy may be explained by the higher sensitivity when using fluorogenic substrates. In sum, if MG43₂-8 and MG43₂ can hydrolyze X₂, it will be with a low activity. This contrasts with many GH43 β-xylosidases that digest X₂ (36). And yet, β-xylosidases like XylB from Bifidobacterium adolescentis that prefer longer dp XOS over X₂ have also been reported (37).

DSA-FACE demonstrated a strict Rex substrate specificity for MG8 and showed complete substrate conversion at the maximum concentration tested. At this time, there are only four GH8 Rex enzymes characterized in the CAZy database, including enzymes from Bacillus halodurans (38), Bifidobacterium adolescentis (39), Bacteroides intestinalis (40), and Paenibacillus barcinonensis (29). MG8 shows a typical Rex activity (like the characterized Rex enzymes listed above): MG8 does not hydrolyze pNP-X, is active on XOS with dp 3 to 6, and has a preference for short-dp XOS (41). Similar to Rex8A from Paenibacillus barcinonensis, which was the first one tested against branched oligosaccharides (MeGlcA-decorated xylooligomers), MG8 is able to hydrolyze the reducing-end xylose of branched-AX oligosaccharides. Notably, the rex8A gene from Bacteroides intestinalis is located downstream from a xyl3A gene, which encodes a β-xylosidase. The X₂ generated by Rex8A is therefore hydrolyzed by the Xyl3A β-xylosidase. This is not the case for MG43₂-8, since neither MG43₂ nor MG8 can efficiently hydrolyze X₂, as shown here and previously with HPAEC-PAD. β-Xylosidases like MG43₂ and Rex enzymes like MG8 that have low or no X₂ hydrolytic activity are interesting for the incomplete degradation of AX into X₂. X₂ has been demonstrated to be the most efficient prebiotic among the xylose polymers in terms of promoting higher growth of Bifidobacterium and Lactobacillus strains and presenting increased sweetness power in comparison to that of sucrose (42).

MG43₇ shows a unique substrate specificity pattern, requiring the presence of MG43₂-8/MG43₂/MG8 for activity. In fact, the MG43₂-8 and MG43₇ coding sequences were identified in the same operon, already suggesting a natural synergy between these two enzymes. Notably, MG43₇ only handles O-3 arabinofuranosyl substitutions of rather small AXOS. Previously, a GH43₁₈ metagenome-derived enzyme also showed a single preference for A³X from the (A)XOS studied (13). At higher enzyme concentrations, MG43₇ shows xylanase activity (either endo- or exo-xylanase activity, as discussed for MG43₂₈) on higher-dp XOS and AXH-d2 activity on internal-arabinose-substituted xyloses. Further investigation should be made to understand such particular substrate recognition by MG43₇ and GH43₁₈.

It is worth noting that our study provides detailed insights into substrate preferences but not into kinetics. Overnight reactions were performed, but often, incomplete conversions were observed. Similar observations of incomplete conversions were observed previously with AX-acting enzymes (31, 43–45). This may indicate low rates, enzyme death, and/or product inhibition. The latter is less likely to be an issue in natural systems where other enzymes further convert the product of the first reaction. And yet, product inhibition is highly relevant in industrial applications, where high substrate concentrations are used. Consequently, enzymes are either selected based on low product inhibition levels (46, 47) or their crystal structure is determined to unravel the structural basis of product inhibition, giving rise to opportunities for protein engineering to release or reduce product inhibition (48).

In conclusion, DSA-FACE enables HT analysis of the enzymatic substrate preferences of AXOS-acting enzymes in a relatively short experimental and analysis time. Thanks to the HT nature of the approach, by performing enzymatic reactions at different enzyme concentrations, different (A)XOS structures can be ranked as preferred substrates and sequential enzymatic cleavages can be determined. This approach allowed us to create degradation maps for five metagenome-derived enzymes for 12 different (A)XOS substrates. The knowledge of the exact substrate preferences is undoubtedly essential to either achieve desired hydrolysis products (e.g., prebiotics) or come to a full hydrolysis. Finally, given the variety, promiscuity, and flexible substrate preferences of the majority of carbohydrate-active enzymes, DSA-FACE may be explored for analysis of activities other than (arabino)xylanolytic activities.

MATERIALS AND METHODS

Expression and purification of metagenome-derived enzymes.

pET28 plasmids containing the enzyme DNA sequences were obtained as described previously (27). Chemically competent Escherichia coli TOP10 and E. coli BL21(DE3), E. coli BL21 CodonPlus (DE3), or E. coli ArcticExpress strains prepared according to the rubidium chloride method were transformed with these plasmids.

Table 2 gives an overview of the enzyme expression conditions obtained after the preceding optimization steps. For optimal aeration, Erlenmeyer flasks exceeding at least four times the expression culture volume were used. The different expression hosts were grown at indicated temperatures (Table 2) in lysogeny broth (LB) with appropriate antibiotics until reaching an optical density at 600 nm (OD₆₀₀) of approximately 0.6, followed by IPTG (isopropyl β-d-1-thiogalactopyranoside) induction. LBE-5052 autoinduction medium consisted of 1% tryptone, 0.5% yeast extract, 40 mM K₂HPO₄, 10 mM KH₂PO₄, 50 mM NH₄Cl, 5 mM Na₂SO₄, 2 mM Mg₂SO₄, 0.5% glycerol, 0.05% glucose, 0.2% lactose, 50 μg/ml kanamycin, and a trace metal mixture (50 μM FeCl₃, 20 μM CaCl₂, 10 μM MnCl₂, 10 μM ZnSO₄, 2 μM CoCl₂, 2 μM CuCl₂, and 2 μM NiCl₂).

Cells were harvested by centrifugation at 3,100 × g for 30 min at 4°C. Pellets were then suspended in 1/25 of the original volume in equilibration buffer for metal affinity chromatography (see below) and 1 mg/ml lysozyme and incubated on ice for 30 min. After three freeze-thaw cycles, sonication was performed on ice (3 times for 30 s with 30-s intervals at 40% amplitude). Cell debris was removed by centrifugation at 20,000 × g for 30 min at 4°C, and the resulting supernatants were clarified by filtration with a 0.45-μm filter. Purifications by metal affinity chromatography were performed either with His GraviTrap columns (GE Healthcare) or HisPur Ni-nitrilotriacetic acid (NTA) superflow agarose (Thermo Fisher Scientific). The manufacturer’s protocols were followed in both cases, with the exception for the latter that the sample-resin incubation time was extended to 1 h and the buffers used were modified (20 mM sodium phosphate, 500 mM NaCl, 20 mM imidazole, pH 7.4, as the equilibration buffer; 20 mM sodium phosphate, 500 mM NaCl, 50 mM imidazole, pH 7.4, as the wash buffer; and 20 mM sodium phosphate, 500 mM NaCl, 500 mM imidazole, pH 7.4, as the elution buffer).

Eluted samples were diluted with reducing sample buffer, boiled for 5 min, and analyzed by 12% SDS-PAGE (ROTI-mark standard from Carl Roth was used). Fractions containing the protein of interest were then dialyzed against 20 mM HEPES-NaOH buffer, pH 7.0, and 300 mM NaCl, pooled, and concentrated with Vivaspin concentrators when necessary. Dialysis was done with Slide-A-Lyzer mini-dialysis devices (molecular-weight cutoff [MWCO] of 3,500) (Thermo Fisher Scientific) or with Servapor dialysis tubing (MWCO of 12,000 to 14,000, regenerated cellulose [RC], 16-mm diameter). Protein concentrations were measured with the Abs280nm app of the DeNovix DS-11 series spectrophotometer. Extinction coefficients were calculated with the ProtParam tool (ExPASy).

Enzymatic reactions of metagenome-derived enzymes with (A)XOS.

Metagenome-derived enzymes were tested against (A)XOS (Fig. 6) supplied by Megazyme (Megazyme International Ireland, Bray, Ireland), which have a minimum purity of 95%, except for the mixture of A²XX and A³XX, which has a minimum purity of 90%, and XA²⁺³XX, which has a minimum purity of 85%. Enzymatic reaction mixtures with a total volume of 100 μl (or 50 μl, to achieve the desired enzyme concentration when there was only a limited enzyme volume available) in a 96-well plate contained 0.2 to 38 μM enzyme, 10 μM (A)XOS, 50 mM HEPES-NaOH, 50 mM NaCl, pH 7.0. Mineral oil (30 to 50 μl) was used to avoid evaporation from the 96-well plates during enzymatic reactions (Fig. 7). Substrate and enzyme blanks, where enzyme and substrate (respectively) were replaced by the corresponding buffer, were added. Some repetitions of reactions were performed in a 1.5-ml Eppendorf tube for reasons of simplicity. Enzymatic reaction mixtures were incubated at 37°C and 750 rpm in a Thermomixer comfort (Eppendorf). The number of replicates done per enzyme/substrate combination is given in Table S1 in the supplemental material. After 22 h, reactions were stopped by incubation at 80°C for 30 min.

FIG 6.

Twelve different (arabino)xylo-oligosaccharides [(A)XOS] used as substrates in the enzymatic reactions and as standards for the DSA-FACE analysis. AXOS are named according to the nomenclature proposed in reference 50.

FIG 7.

Protocol for high-throughput study of substrate specificities of arabinoxylan-active enzymes by DSA-FACE. (A) Putative AX-active enzymes were incubated with (A)XOS for 22 h. Six enzymes were tested against 12 (A)XOS, including 12 substrate blanks and 6 enzyme blanks (90 samples in total). (B) Reaction mixture hydrolysates were then diluted with ultrapure water and lyophilized. (C) Afterwards, reductive amination reactions were performed to derivatize the carbohydrates at their reducing end with the negatively charged and fluorescent reagent APTS (8-aminopyrene-1,3,6-trisulfonic acid trisodium salt). (D) Ten microliters of derivatized reaction mixture hydrolysate were analyzed by DSA-FACE. All steps were done in a 96-well plate, and 90 samples were analyzed in approximately 14 h.

Analysis of enzymatic-reaction-mixture hydrolysates by DSA-FACE.

Reaction mixture hydrolysates were diluted 10-fold with ultrapure water, and 10-μl amounts were lyophilized. Carbohydrates present in the lyophilized fraction were then derivatized with APTS (8-aminopyrene-1,3,6-trisulfonic acid trisodium salt) by reductive amination as described in reference 13. Afterwards, samples were quenched by diluting the reaction mixtures 200-fold with ultrapure water. Ten microliters of derivatized hydrolysate was analyzed using the Applied Biosystems 3130 Genetic Analyzer with 36-cm capillaries filled with Applied Biosystems POP-7 polymer as described in reference 13 (Fig. 7).

Through DSA-FACE electropherograms, the carbohydrates before and after enzymatic reactions are identified by comparison to standards. Xylose and arabinose monomers are not detected by DSA-FACE, as they fall into the DSA-FACE noise region due to their high electrophoretic mobility. Since A²X, A²⁺³X, XA²X, XA³X, and XA²⁺³X standards are not commercially available, they were identified by comparison between the electrophoretic mobilities of the hydrolysates and the electrophoretic mobilities of the available standards and based on spiking experiments (Fig. S5 and S6). Previously, it was seen that AXOS with dp z present electrophoretic mobilities between XOS with dp z − 1 and z, showing increased electrophoretic mobilities in comparison with those of XOS with the same dp. For example, A²XX and A³XX are therefore expected to have electrophoretic mobilities between those of X₃ and X₄ (13).

DSA-FACE product profiles.

DSA-FACE product profiles were made with the Excel graph function. Peak areas were collected with GeneMapper software, version 4.0. DSA-FACE peak area reproducibility is dependent on the amount of labeled carbohydrate injected in each run, which may vary due to the electrokinetic injection mechanism of the 3130 Genetic Analyzer. Intrinsic carbohydrate electrophoretic mobilities affect the amount of sample injected by the electrokinetic mechanism (49). Therefore, peak areas are corrected by dividing the hydrolysate peak areas by the peak area of the blank with the same (A)XOS structure. When this AXOS was not one of the standard AXOS, the peak area of an AXOS with the same dp is taken. The average of the corrected peak areas is then taken for the DSA-FACE product profiles. To normalize all peak areas obtained for the same enzyme but different enzyme concentrations and substrates, the largest peak area (or the largest sum of the carbohydrate peak areas when more peaks are present in a hydrolysate) is taken as the maximum amount of carbohydrate possibly found in a hydrolysate. All product profiles revealed by DSA-FACE are summarized in Table S2.

Data availability.

The GenBank accession numbers for the enzyme DNA sequences 12_H03-13 (MG43₂-8), 12_H03-12 (MG43₇), and 12_J03-18 (MG43₂₈) (Table 1) are MT603581, MT603582, and MT603583, respectively.