2026 年 73 巻 2 号 論文ID: 7302103
Gum arabic (GA) is a highly branched arabinogalactan-protein complex widely used in food and pharmaceutical industries. Its complex side chains contain L-arabinofuranose residues linked via α-(1→3) and α-(1→4) bonds, which represent the final structural barrier to complete enzymatic degradation of GA. Here, we isolated and characterized two complementary α-L-arabinofuranosidases, FoAF2 and FoAF3, from Fusarium oxysporum 12S. FoAF2, a glycoside hydrolase (GH) family 54 enzyme, preferentially cleaves α-(1→3)-arabinosyl residues while exhibiting weak activity toward α-(1→4) linkages at high enzyme concentrations. In contrast, the GH43_34 enzyme FoAF3 displays strict specificity for α-(1→4)-arabinosyl residues but requires prior removal of neighboring α-(1→3) substituents for efficient catalysis. Structural modeling using AlphaFold 3 revealed that the constrained catalytic pocket of FoAF3 is highly sensitive to steric hindrance from adjacent branches, explaining its dependence on sequential FoAF2 action for complete debranching. This two-enzyme system functions similarly to the bifunctional Bifidobacterium BIAraE, but achieves the same effect using separate enzymes rather than fused domains. Combined with previously characterized F. oxysporum enzymes, FoAF2 and FoAF3 complete a comprehensive toolkit enabling systematic GA degradation from complex side chains to monosaccharides. These findings provide molecular insights into the mechanisms underlying α-L-arabinofuranosidase specificity and establish a foundation for the enzymatic modification of GA for industrial applications.
AFase, α-L-arabinofuranosidase; Ara, L-arabinose; Araf, L-arabinofuranose; CAZy, Carbohydrate-Active enZymes; ESI, electrospray ionization; CBM, carbohydrate-binding module; cDNA, complementary DNA; FA, formic acid; GA, gum arabic; Gal, D-galactose; GH, glycoside hydrolase; GlcA, D-glucuronic acid; GSOs, GA side chain oligosaccharides; HPAEC-PAD, high-performance anion-exchange chromatography with pulsed amperometric detection; LC/IT/TOF MS, liquid chromatography/ion-trap/time-of flight mass spectrometer; NMR, nuclear magnetic resonance; PCR, polymerase chain reaction; PDB, Protein Data Bank; pLDDT, predicted local distance difference test; pNP, p-nitrophenyl; rAoAFQ1, recombinant AoAFQ1; rFoAF2, recombinant FoAF2; rFoAF3, recombinant FoAF3; Rha, L-rhamnose; SDS-PAGE, sodium dodecyl sulfate-polyacrylamide gel electrophoresis; TFA, trifluoroacetic acid.
Gum arabic (GA) is a water-soluble polysaccharide obtained from the sap of Acacia senegal and related Acacia species. It is widely utilized in the food, pharmaceutical, and cosmetic industries as an emulsifier, stabilizer, and coating agent [1, 2]. The major component of GA is an arabinogalactan protein complex with a highly branched structure composed of a β-(1→3)-galactosyl backbone decorated with numerous side chains [3, 4, 5, 6]. These side chains are primarily composed of D-galactose (Gal), L-arabinofuranose (Araf), D-glucuronic acid (GlcA), and L-rhamnose (Rha) residues. This complex branching architecture is considered to be responsible for its functional properties, particularly its high hydrophilicity and strong emulsifying capacity.
The structure of GA has been investigated using chemical approaches such as methylation analysis and nuclear magnetic resonance (NMR) spectroscopy [7, 8]. However, due to the complexity of its branched side chains, the detailed structural features remain incompletely understood. Enzymes that specifically cleave glycoside linkages within polysaccharide chains are strong tools for elucidating such complex structures and for modulating their physicochemical properties.
We identified and characterized various GA-degrading enzymes derived from the plant-pathogenic filamentous fungus Fusarium oxysporum [9, 10, 11, 12]. Nevertheless, research on GA-degrading enzymes remains incomplete. In particular, enzymes capable of specifically cleaving the highly branched α-(1→3)/α-(1→4)-linked Araf residues in GA side chains, such as α-L-arabinofuranosidase (AFase, EC 3.2.1.55) have not yet been reported. As a result, the removal of Araf residues constitutes the final key step required for the complete enzymatic degradation of GA side chains. Therefore, elucidating the substrate specificity of AFases from F. oxysporum toward GA side chains is essential for understanding the overall degradation mechanism.
AFases are enzymes that release Araf residues from the non-reducing ends of polysaccharides or oligosaccharides. It acts cooperatively with other enzymes in the degradation of hemicelluloses such as arabinoxylan and pectic substances such as branched arabinan [13]. AFases are classified into several glycoside hydrolase (GH) families based on their amino acid sequences in the Carbohydrate-Active enZymes (CAZy) database, including GH2, GH3, GH43, GH51, GH54, and GH62. Members of each family exhibit distinct specificities for the position and substitution pattern of Araf residues [14]. For example, GH43 enzymes typically act on α-1,5-arabinooligosaccharides and on single α-(1→2)- or α-(1→3)-substituted residues present in branched arabinan [15, 16], whereas GH54 enzymes are frequently reported to remove both mono- and di-substituted Araf residues [17]. Although AFases acting on GA from Bifidobacterium have been reported [18], no previous study has achieved complete sequential degradation of GA using enzyme combinations.
In this study, we isolated two AFases, designated FoAF2 and FoAF3, from the culture supernatant of F. oxysporum. These enzymes specifically target the α-(1→3)/α-(1→4) linkages within GA side chains. The corresponding genes were cloned and heterologously expressed in Escherichia coli and Pichia pastoris, and the enzymatic properties of the recombinant proteins were characterized. Based on their biochemical features, we comprehensively elucidated the GA degradation mechanism in Fusarium and established a complete enzyme toolkit for structural analysis of GA polysaccharides. The enzymes described herein are expected to be valuable tools not only for fundamental studies on carbohydrate structure elucidation but also for the modification of physicochemical properties for industrial applications.
Reagents, columns, equipments, and primers
Mono Q HR 5/5, Resource PHE (6 mL), and HisTrap HP (5 mL) columns were obtained from GE HealthCare UK, Ltd. (Little Chalfont, UK). Toyopearl DEAE-650M was purchased from Tosoh Corp. (Tokyo, Japan). Amylose resin was from New England BioLabs Inc. (Ipswich, MA, USA). p-Nitrophenyl (pNP) glycosides and GA were acquired from Sigma-Aldrich Co. (St. Louis, MO, USA). Protease inhibitor cocktail was from Nacalai Tesque, Inc. (Kyoto, Japan). L-Arabinan and debranched arabinan were acquired from Megazyme International Ireland, Ltd. (Wicklow, Ireland). A Dionex ICS-5000 system and CarboPac PA1 (4 × 250 mm) and PA20 (3 × 150 mm) columns were obtained from Thermo Fisher Scientific Inc. (Waltham, MA, USA). A PSQ-1 protein sequencer was from Shimadzu Corp. (Kyoto, Japan). A Minimate TFF Capsule 650 Da Omega ultrafiltration membrane and Nanosep 10K Omega centrifugal device were purchased from Pall Co. (Port Washington, NY, USA). All other reagents, unless specified, were of reagent grade and purchased from FUJIFILM Wako Pure Chemical Co. (Osaka, Japan). The primers used in this study were synthesized by Fasmac Co., Ltd. (Kanagawa, Japan) and their sequences are listed in Table S1 (see J. Appl. Glycosci. Web site).
AFase assays
AFase activity was measured using a standard assay with pNP-α-L-arabinofuranoside (pNP-Araf) as the substrate. The reaction mixture consisted of 190 µL of 0.02 % pNP-Araf in 20 mM Na-acetate buffer (pH 5.0) and 10 µL of enzyme solution, which was incubated at 37 °C for 10 min. The reaction was stopped by adding 1.8 mL of 0.2 M Na2CO3, and the amount of released p-nitrophenol was determined by measuring the absorbance at 420 nm. One unit of enzyme activity was defined as the amount of enzyme required to release 1 µmol of p-nitrophenol per min at 37 °C.
Substrate specificity towards various synthetic pNP-glycosides was tested using 2 mU of the enzymes in the standard AFase assay, except that the incubation time was 16 h.
Enzyme activity in the inhibition experiment with L-arabinose (Ara) was measured using a standard AFase assay, except that the incubation time was 30 min. The enzyme was used at 4 mU, and Ara solutions at different concentrations were added to the reaction mixture.
Partial purification of FoAF2 and FoAF3 from the culture filtrate of F. oxysporum 12S
To purify FoAF2 and FoAF3, F. oxysporum 12S was cultured as described by Sakamoto et al. [19]. The culture supernatant (1 L) was concentrated, dialyzed against 20 mM Na-phosphate buffer (pH 6.0), and loaded onto a Toyopearl DEAE-650M column. The bound proteins were eluted by a linear NaCl gradient. Enzyme activities were detected in the non-adsorbed and adsorbed fractions, designated FoAF2 and FoAF3, respectively, and each enzyme was further purified while monitoring activity against pNP-Araf after each step.
The enzyme solution containing FoAF2 was adjusted to 50 % ammonium sulfate saturation and put on a Toyopearl Butyl column equilibrated with 20 mM Na-acetate buffer (pH 5.0) containing 50 % ammonium sulfate. Bound proteins were eluted by a linear gradient (50-0 % of saturation). The same procedure was repeated using a Resource PHE column.
FoAF3-containing fractions from the Toyopearl DEAE column were dialyzed against 20 mM Na-acetate buffer (pH 5.0) and put on a Mono Q column. Bound proteins were eluted by a linear NaCl gradient. Active fractions were pooled, adjusted to 50 % ammonium sulfate saturation, and loaded onto a Resource PHE column.
Cloning and expression of Foaf2 in E. coli, and purification of the recombinant enzyme
The N-terminal amino acid sequence of FoAF2 was determined to be GPXDIYKXGGTP (X: unidentified) using a PSQ-1 protein sequencer, identical to AFase (ABFB) from F. oxysporum f. sp. dianthi (GenBank AJ310126). Based on the abfb gene sequence, upstream (abfb-5/1, -5/2) and downstream (abfb-3/1, -3/2) primers were designed for nested polymerase chain reaction (PCR), and cassette ligation-mediated PCR was performed using genomic DNA of F. oxysporum 12S to determine sequences flanking the Foaf2 gene.
To amplify the mature Foaf2, PCR was performed using two primers, af2/mature/Neco and af2/mature/Cpst, with single-stranded complementary DNA (cDNA) as the template. EcoRI and PstI sites were introduced to the primers, respectively. The PCR product was cloned into pMAL-c2X (New England BioLabs Inc.) to yield pMAL-Foaf2, which was used to transform E. coli BL21. The Foaf2 cDNA sequence was deposited in GenBank AB468068.
For production of recombinant FoAF2 (rFoAF2), 1 % of an overnight culture of the E. coli transformant was inoculated into 500 mL of LB medium containing 50 μg/mL ampicillin and incubated at 37 °C for 1 h. Expression was induced with 0.1 mM IPTG at 15 °C for 24 h. Cells were harvested, lysed by sonication in amylose column buffer (20 mM Tris-HCl, pH 7.0, 200 mM NaCl, and 1 mM EDTA, 1 mM PMSF, protease inhibitor cocktail), and the supernatant was applied to an amylose resin column. Bound proteins were eluted by a linear 0-2 mM maltose gradient. The rFoAF2-containing fractions were pooled, dialyzed against 20 mM K-phosphate buffer (pH 8.0), and loaded onto a Mono Q column, and eluted by a 0-0.5 M NaCl gradient. Active fractions were collected based on AFase activity.
Cloning and expression of Foaf3 in P. pastoris, and purification of the recombinant enzyme
Internal amino acid sequences of FoAF3 were determined by liquid chromatography/ion-trap/time-of flight mass spectrometer (LC/IT/TOF MS) after trypsin digestion [12], revealing high similarity to the unnamed protein product (GenBank CEF77714.1) belonging to GH43_34 from F. graminearum PH-1. Based on GH43_34 sequences from Fusarium spp., degenerate primers, AF3-Fw and AF3-Rv, were designed to amplify the Foaf3 gene (excluding the signal sequence) with genomic DNA of F. oxysporum 12S as the template. Signal peptides were predicted using SignalP 6.0.
To construct a plasmid encoding mature FoAF3, two primers (pPICZαA/AF3EcoFw and pPICZαA/AF3XbaRv) were designed based on the Foaf3 sequence. PCR was performed using single-stranded cDNA from F. oxysporum 12S as the template. The forward primer contained EcoRI site, and the reverse primer contained an XbaI site and lacked a stop codon. The PCR product was inserted into the pPICZαA (Thermo Fisher Scientific Inc.) to generate pPICZ-Foaf3, which was transformed into E. coli DH5α and cultured in low-salt LB medium with 25 μg/mL Zeocin. The Foaf3 cDNA sequence was deposited in the GenBank LC681953.
Pichia pastoris X-33 transformed with BglII-linearized pPICZ-Foaf3 was cultured in BMMY medium for 6 days to produce recombinant FoAF3 (rFoAF3), following a previously reported method [20] with minor modifications. The culture supernatant was concentrated using a 10-kDa ultrafiltration membrane, dialyzed against binding buffer (10 mM K-phosphate buffer, pH 8.0, containing 150 mM NaCl and 20 mM imidazole), and applied to a HisTrap HP column. Bound proteins were eluted with 500 mM imidazole. Active fractions, identified by pNP-Araf assay, were collected and dialyzed against 10 mM K-phosphate buffer (pH 8.0).
Preparation of recombinant AoAFQ1
The GH51 AFase gene Aoafq1 from Aspergillus oryzae (GenBank Q2U790.2) was cloned as a control, following a previously reported method [21], using primers AoAFQ1-Fw and AoAFQ1-Rv. Recombinant AoAFQ1 (rAoAFQ1) was expressed as described previously and purified using Toyopearl DEAE and Superdex 75 columns.
Preparation of oligosaccharides from GA
To examine the substrate specificity of rFoAF2 and rFoAF3, four GA-derived oligosaccharides (GSO3A, 3B, 4A, and 5) were prepared. Sphingomonas sp. 24T exo-β-(1→3)-galactanase (Exo-1,3-Gal) cleaves β-(1→3)-galactan while bypassing branching points [22], yielding GA side chain oligosaccharides (GSOs). GSO7, a major hepta-oligosaccharide product, was purified and characterized as previously described [11]. The structures of the GSOs used in this study are summarized in Fig. 1.
GSO3A, GSO4A, and GSO5 were prepared from GSO7 by enzymatic treatment. GSO7 in 20 mM Na-acetate buffer (pH 5.0) was incubated with GH79 4-O-α-L-rhamnosyl-β-D-glucuronidase (FoBGlcA) [11] and GH39 3-O-α-D-galactosyl-α-L-arabinofuranosidase (GAfase) [23] for GSO3A, FoBGlcA and GH27 α-(1→3)-galactopyranosidase (FoGP1) [10] for GSO4A, or with GAfase alone for GSO5, at 37 °C until completion (Fig. 2A). GAfase was kindly provided by Prof. Kiyotaka Fujita (Kagoshima University, Kagoshima, Japan). Reaction mixtures were boiled to stop the reaction and GSOs were purified by high-performance anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD). Fractions were desalted on activated carbon columns and used for experiments. HPAEC was performed according to program No. 1 (Table S2; see J. Appl. Glycosci. Web site).
Red arrows indicate the enzymatic cleavage sites. Experimental details are described in the “Materials and Methods” section.
Exo-1,3-Gal also releases GSO6A from GA. GSO3B was prepared from GSO6A by treatment with FoBGlcA and FoGP1, following the same procedure as for GSO4A (Fig. 2A). The structural determination of GSO6A is described in the Results section.
HPAEC conditions
HPAEC-PAD for carbohydrate analysis was performed in a Dionex ICS-5000 system. One of two elution programs was selected depending on the sample (Table S2; see J. Appl. Glycosci. Web site).
Mass spectrometry
GSO6A purified by HPAEC-PAD was desalted using a porous graphitic carbon cartridge (25 mg, 30-40 μm; Thermo Fisher Scientific Inc.) [24, 25]. The sample was dried, redissolved in 0.1 % trifluoroacetic acid (TFA), loaded onto the prewashed cartridge, washed with water, and eluted with acetonitrile:water:TFA (40:60:0.05). After drying, it was redissolved in 0.1 % formic acid (FA), diluted with acetonitrile containing 0.1 % FA, and 5 μL was injected into an InertSustain Peptides C18 column (0.1 × 150 mm, GL Sciences Inc., Tokyo, Japan). Elution was performed with acetonitrile:water:FA (50:50:0.1) at 0.5 μL/min and 20 °C. The eluate was ionized in positive ion mode by electrospray ionization (ESI) (4.0 kV) and analyzed using an LCQ Fleet ion-trap mass spectrometer (Thermo Fisher Scientific Inc.). Mass spectra (m/z 150-2,000) were acquired, and major ions were fragmented by data-dependent CID using Xcalibur 4.2 SP1 (Thermo Fisher Scientific Inc.).
Preparation of GA with enzymatically modified structure
FoGP1 (2.2 U) was added to 30 mL of 20 mM Na-acetate buffer (pH 5.0) containing 1 % GA and incubated at 37 °C for 2 days. After boiling to stop the reaction, the mixture was dialyzed and designated as GA-GP1 (Fig. 2B). FoAF2 (0.7 U) was then added to GA-GP1 (6 mL), followed by incubation at 37 °C for 8 h. The enzyme was inactivated by boiling, and the mixture was dialyzed to obtain GA-GP1/AF2 (Fig. 2B).
Substrate binding prediction
The 3D structures of FoAF2 and FoAF3 were predicted using local AlphaFold 3 [26] and Protenix server [27], an updated deep-learning based protein structure prediction model. Amino acid sequences (FoAF2: BAG80559.1; FoAF3: BDF83204.1) were retrieved from GenBank, and the highest-confidence model (mean predicted local distance difference test (pLDDT) > 90) was selected from five predictions.
Ligand binding was analyzed using the “Ligand Binding” module in AlphaFold 3 and Protenix server. GSO3A, GSO3B, and GSO4A ligand structures were generated using the CHARMM-GUI server [28]. For each ligand, five independent predictions were performed, and the most probable binding pose was determined based on confidence score and binding free energy estimation. The predicted binding sites were mapped onto the 3D structure and compared with known catalytic residues. Structural superimposition and visualization were conducted using PyMOL (v3.1.0, Schrödinger, LLC, New York, NY, USA).
Purification of FoAF2 and FoAF3 from F. oxysporum 12S
FoAF2 and FoAF3 were purified by anion-exchange and hydrophobic chromatographies using culture supernatant (1 L) of F. oxysporum 12S. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis of the partially purified FoAF2 and FoAF3 showed protein bands with molecular masses of 60 and 35 kDa, respectively (data not shown). The characteristics of the enzymes were determined using recombinant enzymes because the quantities and purities of proteins obtained were low. These results are described below.
Cloning and expression of Foaf2
The N-terminal amino acid sequence of FoAF2 matched that of the ABFB from F. oxysporum f. sp. dianthi. Based on the nucleotide sequence of the abfb gene, cassette ligation-mediated PCR using the designed primer sets successfully amplified the regions flanking the Foaf2 gene from the genomic DNA of F. oxysporum 12S, allowing identification of the start and stop codons of the Foaf2 gene.
The mature Foaf2 cDNA fragment was successfully amplified by PCR using the designed primers and inserted into the pMAL-c2X vector. The resulting plasmid, pMAL-Foaf2, was confirmed by sequencing and used to transform E. coli BL21. No introns are present in the Foaf2 gene sequence. The mature Foaf2 gene encodes a 479-amino-acid protein, with a calculated molecular weight of 49,614 Da. This theoretical molecular weight differed from the 60 kDa observed for the enzyme produced by the parental fungus. We consider this discrepancy to be due to glycosylation occurring in the fungus. FoAF2 was classified into the GH54 family in the CAZy database. BLAST analysis of the predicted amino acid sequence of FoAF2 (Table S3; see J. Appl. Glycosci. Web site) revealed 100 % identity to the predicted AFase B (GenBank KAG7406539.1) from F. oxysporum f. sp. rapae.
The rFoAF2 was efficiently expressed in E. coli BL21 as an MBP-fusion protein and purified to apparent homogeneity by amylose resin and Mono Q column chromatography. SDS-PAGE analysis of the purified enzyme showed a single band with a molecular mass of approximately 100 kDa (Fig. S1; see J. Appl. Glycosci. Web site), consistent with the calculated mass of rFoAF2 comprising an MBP tag (~42 kDa) fused to its N-terminus.
Cloning and expression of Foaf3
Internal amino acid sequence analysis of FoAF3 using LC/IT/TOF MS identified several tryptic peptides showing high similarity to the unnamed protein product (GenBank CEF77714.1) belonging to the GH43_34 subfamily from F. graminearum PH-1. Based on this information, degenerate primers targeting the mature region of GH43_34 homologs from Fusarium species were designed, and PCR with these primers amplified a partial Foaf3 gene fragment corresponding to the mature region from the genomic DNA of F. oxysporum 12S. The cDNA corresponding to the mature region (excluding the signal sequence) was subsequently obtained and inserted into the pPICZαA vector, yielding the construct pPICZ-Foaf3. The Foaf3 gene sequence does not contain any introns. The mature Foaf3 gene encodes a protein of 312 amino acids, with a calculated molecular weight of 34,638 Da. Based on the amino acid sequence, FoAF3 was classified into the GH43_34 family in the CAZy database. The results of BLAST searches using the predicted amino acid sequence of FoAF3 are summarized in Table S4 (see J. Appl. Glycosci. Web site). FoAF3 showed the highest sequence identity (99 %) to a predicted xylosidase/arabinosidase (GenBank KAG7406630.1) from F. oxysporum f. sp. rapae.
As rFoAF3 expressed in E. coli was entirely insoluble, we used P. pastoris X-33, where rFoAF3 was successfully produced as a secreted protein. The enzyme was purified by HisTrap column chromatography to apparent homogeneity. SDS-PAGE analysis of the purified enzyme revealed a single band with a molecular mass of approximately 40 kDa (Fig. S1; see J. Appl. Glycosci. Web site). The difference in migration between the recombinant and native FoAF3 proteins on SDS-PAGE suggests that rFoAF3 undergoes N-linked glycosylation. Consistent with this, FoAF3 contains two predicted N-glycosylation sites at Asn22 and Asn99.
Enzymatic activity of rFoAF2 and rFoAF3 on pNP-glycosides
The activities of rFoAF2 and rFoAF3 were tested against various pNP-glycosides, including Araf, α-L-arabinopyranoside, β-L-arabinopyranoside, α-D-galactopyranoside, β-D-galactopyranoside, α-D-glucopyranoside, β-D-glucopyranoside, α-L-rhamnopyranoside, and β-D-xylopyranoside. The rFoAF2 showed the highest activity toward pNP-Araf and weak but detectable activity toward pNP-β-D-xylopyranoside, whereas rFoAF3 was active only on pNP-Araf. The weak β-xylosidase activity of rFoAF2 likely reflects structural similarities between Araf and xylopyranose moieties-both pentoses with comparable conformations that can be accommodated by flexible substrate-binding pockets. Similar low-level activity toward xylosyl substrates has also been reported for other GH43 AFases.
Effect of pH and temperature
The optimal reaction conditions were pH 3-4 and 60 °C for rFoAF2, and around pH 6 and 50 °C for rFoAF3 (Fig. S2; see J. Appl. Glycosci. Web site). These properties are typical of extracellular fungal enzymes. In terms of thermal stability, rFoAF2 retained more than 80 % of its activity after a 1 h incubation at temperatures up to 50 °C, whereas rFoAF3 maintained similar stability up to 40 °C. Regarding pH stability, rFoAF2 retained over 80 % of its activity between pH 4 and 8 after a 16 h incubation at 4 °C, while rFoAF3 remained stable across an exceptionally broad pH range (pH 3.2-13.0).
Effect of arabinose on enzyme activity
The inhibitory effect of the reaction products (L-Ara) on AFase activities of rFoAF2 and rFoAF3 was examined. The activity of rFoAF2 decreased to approximately 50 % in the presence of 1 mg/mL Ara and to 15 % at 10 mg/mL (Table 1). In contrast, rFoAF3 activity was only slightly affected by Ara (Table 1). Product inhibition has also been reported for other GH54 enzymes [29, 30]. In structural studies of GH54 AFase, co-crystallization of the GH54 enzyme A. kawachii AkAbfB with Ara revealed one Araf-binding site in the catalytic domain and two additional binding sites in the carbohydrate-binding module (CBM) 42 domain [31]. It is therefore speculated that Araf binding to either the β or γ subdomain of CBM42 induces a structural change that is transmitted to the catalytic domain, leading to the product inhibition. Further experiments using a CBM42-deficient FoAF2 mutant are expected to clarify this mechanism.
Table 1. Effect of arabinose concentration on AFase activity.
| Arabinose (mg/mL) | AFase activity (%) | |
| rFoAF2 | rFoAF3 | |
| 0 | 100 | 100 |
| 0.5 | 67 ± 4.7 | 93 ± 1.6 |
| 1 | 52 ± 4.1 | 92 ± 2.9 |
| 10 | 15 ± 1.4 | 70 ± 1.9 |
Enzyme activity was measured using a standard AFase assay with a 30 min incubation. The enzyme concentration was 4 mU, and arabinose solutions at various concentrations were added to the reaction mixture. The released arabinose was quantified by HPAEC-PAD (program No. 2). All experiments were performed in triplicate, and data are presented as mean ± SD.
Specificity of rFoAF2 and rFoAF3 on GSOs
The substrate specificity of rFoAF2 was examined using GSO4A as the substrate. When the enzyme reaction was carried out at different enzyme concentrations, it was found that at low concentrations, the α-(1→3)-arabinosyl side chain was cleaved from GSO4A, producing GSO3A. At high concentrations, both the α-(1→3)- and α-(1→4)-Araf side chains were degraded, resulting in the formation of β-(1→6)-Gal2 (Fig. 3). These results reveal that rFoAF2 exhibits high activity towards α-(1→3)-Araf side chains and weak activity towards α-(1→4)-Araf side chains. The peak at around 12.5 min increased with the amount of enzyme solution, although its identity remains unknown. This observation indicates that it likely arises from a component in the enzyme preparation.
GSO4A dissolved in 20 mM Na-acetate buffer (pH 5.0) was incubated with rFoAF2 at three enzyme concentrations (40, 200, and 400 mU) at 37 °C overnight. Reaction products were analyzed by program No. 1 of HPAEC-PAD.
The substrate specificity of rFoAF3 was then examined using GSO3A, GSO3B and GSO4A as substrates. The rFoAF3 degraded only GSO3A, producing β-(1→6)-Gal2 and Ara (Fig. 4). This indicates that rFoAF3 does not act on α-(1→3)-Araf side chains, but is specific for α-(1→4)-Araf side chains. The rFoAF3 showed no activity against GSO4A, indicating that the presence of α-(1→3)-Araf side chains in the vicinity of α-(1→4)-Araf side chains interferes with the enzyme's activity. These findings demonstrated that the two AFases possess distinct substrate specificities and that their combined action enables efficient degradation of GSO4A. Collectively, these results demonstrate that rFoAF2 and rFoAF3 possess distinct but complementary substrate specificities. Their sequential action enables efficient debranching of complex GSOs, highlighting the synergistic potential of these two enzymes.
GSO3A, GSO3B, and GSO4A dissolved in 20 mM Na-acetate buffer (pH 5.0) were incubated with 200 mU of rFoAF3 at 37 °C overnight. Reaction products were analyzed by program No. 1 of HPAEC-PAD.
Recently, Sasaki et al. [32] reported a unique GH43 AFase, BIAraE, from Bifidobacterium species that represents a novel architectural solution to arabinogalactan degradation. BIAraE is a bifunctional, two-domain enzyme in which a single open reading frame encodes two distinct catalytic domains: GH43_22 and GH43_34. Detailed substrate specificity analysis revealed that the GH43_22 domain cleaves α-(1→3)-Araf residues, while the GH43_34 domain is specific for α-(1→4)-Araf linkages in GA side chains. This tandem arrangement of complementary catalytic functions within a single polypeptide represents an elegant evolutionary strategy for achieving complete debranching capability.
The Fusarium system described in this study achieves a functionally analogous outcome through a fundamentally different organizational strategy. Rather than fusing two catalytic domains into a single enzyme, Fusarium employs two enzymes, FoAF2 and FoAF3, that work sequentially and synergistically. While FoAF2 shares functional similarity with the GH43_22 domain of BIAraE in targeting α-(1→3)-Araf linkages, FoAF3 parallels the GH43_34 domain in its specificity for α-(1→4)-Araf residues.
An intriguing distinction lies in the dual specificity of FoAF2, which exhibits primary activity toward α-(1→3)-linkages but also possesses secondary activity toward α-(1→4)-linkages at higher enzyme concentrations. This promiscuous substrate recognition contrasts with the strict division of labor observed between the two catalytic domains of BIAraE and may reflect differences in active site architecture or substrate-binding groove flexibility. Such dual activity potentially provides Fusarium with additional versatility in polysaccharide degradation, enabling partial cleavage of α-(1→4)-linkages even before FoAF3 engagement.
The observation that FoAF3 activity is inhibited by α-(1→3)-Araf side chains positioned adjacent to its α-(1→4)-Araf target represents a clear example of steric hindrance affecting enzyme-substrate interactions. Similar phenomena have been reported in other GH systems, where neighboring substituents can hinder access to the active site. The requirement for prior debranching by FoAF2 to enable FoAF3 activity highlights the crucial role of enzyme synergy in the complete degradation of complex heteropolysaccharides.
Structural determination of GSO6A
GSO6A, which was used as a substrate for the preparation of GSO3B, was purified and its structure was determined. GSO6A was obtained by treating GA with exo-β-(1→3)-galactanase (Exo-1,3-Gal). The GSOs released from GA by enzymatic treatment were partially purified using a Bio-Gel P2 column according to a previous report [11], and GSO6A was further purified by fractionation using program No. 1 described in Table S2 (see J. Appl. Glycosci. Web site).
LC-MS analysis of GSO6A revealed a precursor ion peak at m/z 981.37 corresponding to [Gal3ArafGlcARha+Na]＋ (theoretical molecular weight: 980.80) (Fig. S3; see J. Appl. Glycosci. Web site). MS/MS analysis of GSO6A detected product ion peaks at m/z 835.12 [Gal3ArafGlcA+Na]＋ (theoretical: 834.66) derived from Rha cleavage and m/z 659.20 [Gal3Araf+Na]＋ (theoretical: 659.54) derived from Rha-GlcA cleavage (Fig. S3; see J. Appl. Glycosci. Web site).
The structure of GSO6A was analyzed by enzymatic degradation. First, GSO6A was hydrolyzed by GAfase to yield GSO4B (designated as GSO4 in a previous report [11]) and Gal-Ara. The resulting Gal-Araf oligosaccharide was further hydrolyzed by FoGP1 to yield Gal and Ara (Fig. S4; see J. Appl. Glycosci. Web site). Based on these results, the sugar chain structure of GSO6A was determined as shown in Fig. S3 (see J. Appl. Glycosci. Web site).
Specificity of rFoAF2 and rFoAF3 on native and structure-modified GA
We investigated whether the substrate specificity observed for both enzymes towards oligosaccharides is similarly manifested with polysaccharides. Focusing GSO7, one of the major side chains of GA, we prepared two modified substrates-designated GA-GP1 and GA-GP1/AF2-by enzymatically removing only the terminal Gal residue or both the terminal Gal and α-(1→3)-Araf residues, respectively, from GA (Fig. 2B).
The enzymatic activities of rFoAF2 and rFoAF3 against GA, GA-GP1, and GA-GP1/AF2 were measured (Table 2). The rFoAF2 showed activity against all substrates, with more than twice the activity against GA-GP1 compared to GA. GA-GP1 contains a greater number of terminal α-(1→3)-Araf residues than GA, indicating that FoAF2 efficiently targets α-(1→3)-Araf side chains in polysaccharides, consistent with its specificity towards oligosaccharides. The activity of rFoAF2 against GA suggests that the native polysaccharide contains side chains terminating in α-(1→3)-Araf. In contrast, its markedly reduced activity against GA-GP1/AF2, from which α-(1→3)-Araf residues had been removed, confirms its limited ability to act on α-(1→4)-Araf linkages.
Table 2. Enzymatic activity of AFases on native and enzyme-treated GA.
| Substrate | Amount of reducing sugar (µg/mL) | ||
| rFoAF2 | rFoAF3 | rAoAFQ1 | |
| GA | 43.7 ± 3.5 | 3.3 ± 1.3 | ND |
| GA-GP1 | 97.1 ± 5.6 | 5.0 ± 2.3 | ND |
| GA-GP1/AF2 | 5.1 ± 1.3 | 43.8 ± 1.3 | ND |
Reaction mixtures containing 1 % GA, GA-GP1, or GA-GP1/AF2 in 20 mM Na-acetate buffer (pH 5.0) and 2 mU of rFoAF2, rFoAF3, or rAoAFQ1 were incubated at 37 °C for 16 h. Enzyme activity was determined by measuring the amount of reducing sugars produced using the Somogyi-Nelson method [35, 36]. All experiments were performed in triplicate, and data are presented as mean ± SD. ND, not detectable.
On the other hand, rFoAF3 showed only weak activity against GA and GA-GP1 but exhibited significantly enhanced activity against GA-GP1/AF2. This demonstrates that rFoAF3 is specific for α-(1→4)-Araf side chains and that its activity is hindered by adjacent α-(1→3)-Araf residues, consistent with results obtained using oligosaccharides as substrates. The weak activity of rFoAF3 toward GA and GA-GP1 also suggests the presence of structures such as GSO3A, in which α-(1→4)-Araf residues occur alone as side chains.
These results indicate that the synergistic action of rFoAF2 and rFoAF3 enables efficient degradation of Araf residues in the side chains of GA. As a control, rAoAFQ1 (GH51) showed no detectable activity toward any of the substrates under the tested conditions. Since the specificities of rFoAF2 and rFoAF3 were clearly identified, their activities on Araf-containing polysaccharides can be used to estimate the abundance of terminal α-(1→3)- or α-(1→4)-linked Araf residues in these polymers.
Activity of rFoAF2 and rFoAF3 on arabinan and arabino-oligosaccharides
Since the specificity of rFoAF2 and rFoAF3 was elucidated, we further validated these findings by measuring their activity towards L-arabinan, which contains abundant α-(1→3)-linked Araf side chains on an α-(1→5)-arabinan backbone, and for comparison, debranched arabinan, which has only minor Araf side chains (Table 3). The rFoAF2 showed high activity against L-arabinan but only weak activity against debranched arabinan, indicating that rFoAF2 functions as a debranching enzyme with strong preference for α-(1→3)-Araf side chains. On the other hand, rFoAF3 did not degrade either substrates even after prolonged reaction, indicating that these lack α-(1→4)-Araf side chains. In addition, when activity against α-(1→5)-arabinobiose was assessed, rAoAFQ1 showed the highest activity, rFoAF2 showed moderate activity, and rFoAF3 showed no activity (Table 4).
Table 3. Enzymatic activity of AFases on arabinans.
| Substrate | Amount of reducing sugar (µg/mL) | |||||
| rFoAF2 | rFoAF3 | rAoAFQ1 | ||||
| 2 h | 16 h | 2 h | 16 h | 2 h | 16 h | |
| L-Arabinan | 32.7 ± 1.3 | 212.6 ± 25.6 | ND | ND | 5.9 ± 1.4 | 42.6 ± 5.3 |
| Debranched arabinan | 3.8 ± 0.8 | 8.0 ± 2.8 | ND | ND | ND | 7.2 ± 4.3 |
Enzyme activity was measured using 0.1 % L-arabinan or debranched arabinan as substrates in 20 mM Na-acetate buffer (pH 5.0). Reactions were performed with 2 mU of enzyme at 37 °C for 2 h or 16 h. The amount of reducing sugars produced was quantified using Somogyi-Nelson method [35, 36]. All experiments were performed in triplicate, and data are presented as mean ± SD. ND, not detectable.
Table 4. Enzymatic activity of AFases on α-(1→5)-arabinobiose.
| Substrate | Amount of released arabinose (µg/mL) | |||||
| rFoAF2 | rFoAF3 | rAoAFQ1 | ||||
| 15 min | 60 min | 15 min | 60 min | 15 min | 60 min | |
| α-(1→5)-Arabinobiose | 4.3 ± 1.7 | 9.2 ± 0.5 | ND | ND | 17.4 ± 4.5 | 38.4 ± 2.6 |
Enzyme activity was evaluated using 0.1 % α-(1→5)-arabinobiose as the substrate in 20 mM Na-acetate buffer (pH 5.0). Reactions were carried out with 2 mU of enzymes at 37 °C for 15 or 60 min. After boiling of the mixture for 5 min, the released Ara was quantified by program No. 2 of HPAEC-PAD. All experiments were performed in triplicate, and data are presented as mean ± SD. ND, not detectable.
The substrate specificity profiles of rFoAF2 and rFoAF3 are consistent with the characteristic features of their respective CAZy families. Our previous studies have demonstrated that GH51 AFase derived from Penicillium (the homolog of AoAFQ1 in this study) exhibits high activity toward low-degree-of-polymerization α-(1→5)-arabinooligosaccharides [33]. In contrast, members of the GH54 family, to which rFoAF2 belongs, predominantly consist of AFases that show high activity toward α-(1→2)/α-(1→3)-linked branched residues in L-arabinan and type I arabinogalactan side chains, while retaining moderate activity against α-(1→5)-arabinooligosaccharides [17]. This broad substrate recognition mechanism of GH54 AFases is consistent with the relaxed substrate specificity observed for rFoAF2, suggesting its involvement not only in cleavage of α-(1→3)-Araf residues from GA but also potentially in L-arabinan utilization. By contrast, rFoAF3 showed no enzymatic activity toward L-arabinan or α-(1→5)-arabinooligosaccharides, demonstrating strict specificity for α-(1→4)-Araf residues in gum arabic. Since α-(1→4)-Araf linkages are uniquely found in GA among natural polysaccharides, rFoAF3 appears to be a GA-specific enzyme, representing a highly specialized member of GH43 AFases.
Putative degradation pathway of GA by F. oxysporum 12S
This study revealed that the F. oxysporum 12S possesses a well-organized enzymatic system for the complete degradation of GA. Figure 5 illustrates the proposed pathway for the breakdown of the GSO7 side chain and backbone structure of GA by F. oxysporum. First, GSO7 is degraded to GSO5 through the sequential action of GH27 α-galactosidase (FoGP1; 10) and GH54 AFase (FoAF2: this study). GSO5 is further degraded to GSO4B by GH43 AFase (FoAF3: this study). Next, GH79 4-O-α-L-rhamnosyl-β-D-glucuronidase (FoBGlcA; 11) cleaves the capping Rha-GlcA disaccharide from GSO4B. The order of action between FoAF3 and FoBGlcA can be inferred from their enzyme activities on GSO3A and GSO5 (Fig. S5A; see J. Appl. Glycosci. Web site). The activity of rFoAF3 toward GSO5 was reduced to approximately 60 % of that toward GSO3A (Fig. S5B; see J. Appl. Glycosci. Web site). However, FoBGlcA activity has been reported to be significantly inhibited by the presence of α-(1→4)-Araf residue [11]. This suggests that FoBGlcA releases Rha-GlcA only after the α-(1→4)-Araf residue is cleaved by FoAF3. Rha-GlcA is cleaved by FoRham1 into Rha and unsaturated GlcA [12], and the remaining backbone is subsequently cleaved by GH43_24 exo-β-(1→3)-galactanase [9] into Gal and β-(1→6)-Gal2. β-Galactosidase capable of degrading β-(1→6)-Gal2 was not detected in the culture supernatant of F. oxysporum 12S; however, a GH2 β-galactosidase with this activity was identified intracellularly and is currently being characterized (unpublished data). It is inferred that extracellular β-(1→6)-Gal2 is transported into the cell and cleaved into two Gal molecules. The absence of this enzyme in the extracellular space suggests that efficient uptake and intracellular metabolism of GA degradation products are crucial to the utilization strategy of F. oxysporum. This two-step degradation strategy appears to optimize both energy efficiency and nutrient utilization.
The red arrows indicate the enzyme cleavage sites.
The degradation pathway elucidated in this study deepens our understanding of fungal mechanisms for utilizing complex polysaccharides and highlights its potential for industrial applications in GA modification.
Structural prediction and ligand binding analysis
Structure prediction using AlphaFold 3 and Protenix generated high-confidence models for both mature FoAF2 and FoAF3 (mean pLDDT > 90) (Fig. S6; see J. Appl. Glycosci. Web site). FoAF2 revealed a two-domain architecture consisting of a β-sandwich catalytic domain typical of the GH54 and a C-terminal CBM42 module, whereas FoAF3 displayed a GH43 β-propeller catalytic fold. Although the overall confidence for FoAF3 was high, its N-terminal region showed low pLDDT scores, indicating structural flexibility.
Ligand docking of GSO3A, GSO3B, and GSO4A revealed clear differences in substrate recognition. All docked GSOs bound to the putative −1 subsite of FoAF2 with consistent orientation, reflecting its broad substrate specificity. In contrast, only GSO3A docked properly in FoAF3, whereas GSO3B and GSO4A failed to adopt the correct orientation (Fig. 6), mirroring FoAF3's strict specificity for α-(1→4)-Araf residues. Since AFases require proper non-reducing end binding at the −1 subsite, the structural model indicates that α-(1→3)-Araf branches can cause steric hindrance, thereby necessitating prior action by FoAF2.
The upper panels show the most stable binding model for each ligand. The second panels display the superposition of the five model structures. The third panels display the pLDDT-colored predictions models obtained by AlphaFold 3 and Protenix server. The schematic diagram in the bottom panels indicate the substrate binding sites in the model structure and the actual enzyme activity.
For FoAF2-ligand complexes, both + and − subsites were confidently predicted, whereas only the − subsite was resolved for FoAF3 (Fig. 6 and Fig. S6; see J. Appl. Glycosci. Web site). Superposition with known crystal structures (Protein Data Bank (PDB) ID: 1WD4 [31] for FoAF2 and PDB ID: 3AKH [34] for FoAF3) revealed high structural similarity (Fig. S7; see J. Appl. Glycosci. Web site), supporting the accuracy of the docking models.
FoAF2's preference for α-(1→3)-linked Araf substituents over α-(1→4)-linkages was rationalized by hydrogen bonding differences at the +1 subsite (Fig. S8; see J. Appl. Glycosci. Web site). In GSO3B, both Asn224 and Asp191 formed hydrogen bonds with the O-4 and O-5 positions of the Gal residue, whereas in GSO3A only Asn224 was involved. The docking models suggest that these additional interactions in GSO3B may enhance binding affinity, although site-directed mutagenesis is required to verify this hypothesis. Docking also clarified FoAF3 specificity: GSO3A fit precisely into its deep catalytic pocket, with the α-(1→4)-linked Araf residue positioned at the −1 subsite. This binding mode closely matched the GH43 crystal structure (PDB ID: 3AKH; 34), supporting the validity of the predicted model. In contrast, the α-(1→3) branch of GSO4A induced steric clashes, blocking correct positioning of the α-(1→4)-Araf residue (Fig. S9; see J. Appl. Glycosci. Web site). This suggests that FoAF3 is, at the molecular level, highly sensitive to neighboring substitutions, which may explain its strict substrate specificity and the complementary requirement for sequential FoAF2 activity.
Collectively, these findings provide the structural basis for the relatively relaxed substrate recognition of GH54 enzymes and the strict substrate specificity of GH43 enzymes, elucidating how their complementary functions enable efficient degradation of the complex side chains of GA. Future site-directed mutagenesis studies based on these predicted structures are expected to validate the roles of key amino acid residues and further clarify the catalytic mechanism.
The authors declare no conflict of interests.
We sincerely thank Prof. Kiyotaka Fujita (Kagoshima University) for generously providing GAfase, which was essential for this study. We would like to thank Prof. Shinya Fushinobu and Assistant Prof. Toma Kashima of the University of Tokyo for their support with AlphaFold 3. This work was funded by institutional support from Osaka Metropolitan University.
