Prebiotics are defined as non-digestible dietary components that reach the colon intact and selectively stimulate the cell growth and metabolism of beneficial probiotics, thereby exerting health-promoting effects on the host [1]. The ingestion of prebiotics can induce significant shifts in the composition and functionality of the colonic microbiota. Among prebiotics, non-digestible oligosaccharides with a degree of polymerization (DP) ranging from 2 to 10 represent a predominant category. These oligosaccharides exhibit stability in the gastrointestinal tract, demonstrating resistance to enzymatic hydrolysis, acidic pH, and bile salts, while specifically enhancing the growth and biological activity of health-associated intestinal microorganisms [1, 2]. Emerging evidence indicates that polyphenols also exhibit prebiotic-like properties owing to their antioxidant and anti-inflammatory effects, facilitated by their gut microbial metabolism [3]. Despite these findings, carbohydrate-based prebiotics remain a focal area of research and industry due to their broad-spectrum bioactivity, low caloric content, and versatile applicability across pharmaceuticals, functional foods, and nutraceutical formulations [4].

Prebiotic oligosaccharides can be categorized based on their chemical composition into sucrose-based fructo-oligosaccharides (FOS), lactose-based galacto-oligosaccharides (GOS), starch-based isomalto-oligosaccharides (IMOS), soy-derived oligosaccharides, chitosan-derived oligosaccharides, hemicellulose-based xylo-/arabinoxylo-/arabino-oligosaccharides (XOS, AXOS, and AOS), and so on [4]. Of these, pentose-based oligosaccharides from hemicellulosic biomass, such as XOS, AXOS, and AOS, have gained substantial interest due to their synbiotic characteristics with health-benefit and selective bifidogenic activities [5]. Given the ubiquity and structural diversity of heteroxylans in plant cell walls, extensive research has focused on the production and functional characterization of XOS and AXOS, and their prebiotic potential [5-8].

Arabinan, a homopolysaccharide predominantly extracted from sugar beet, composed of α-1,5-linked L-arabinofuranosyl residues, often substituted with α-1,2- and/or α-1,3-linked L-arabinofuranosyl branches, serves as a valuable precursor for AOS production. The enzymatic breakdown of arabinan into AOS and L-arabinose relies on the synergistic activity of exo-acting α-L-arabinofuranosidase (ABF; EC 3.2.1.55) and endo-1,5-α-L-arabinanase (ABN; EC 3.2.1.99). Exo-type ABFs are widespread across microbial taxa and are capable of selectively cleaving α-1,2-, α-1,3-, and/or α-1,5-L-arabinofuranosidic bonds from arabinan and arabinoxylan substrates [9]. On the other hand, ABNs are comparatively rare in nature and exhibit endo-acting specificity, hydrolyzing internal α-1,5-L-arabinofuranosidic linkages to generate various AOS intermediates. To date, extensive research has focused on the biochemical properties and biotechnological applications of ABFs [9-11]. However, recent genomic studies have revealed arabinan utilization gene clusters containing ABN-encoding genes in diverse microbial species [12-15]. These discoveries highlight the emerging potential of AOS as next-generation pentose-based prebiotic candidates with functional benefits for gut health modulation [16-20]. This review systematically categorizes microbial α-1,5-specific endo- and exo-arabinanases, examining the structure-function relationships governing their enzymatic mechanisms and outlining strategies for the enzymatic production of diverse prebiotic AOS.

AOS are derived from the enzymatic breakdown of arabinan, a hemicellulosic polysaccharide abundantly present in plant cell walls. Arabinan exhibits significant resistance to enzymatic hydrolysis in the gastrointestinal tract, remaining largely unaffected by intestinal epithelial enzymatic activity. This inherent resistance highlights the prebiotic potential of both arabinan and its derivative AOS, enabling them to selectively support the growth of beneficial gut microbiota [21, 22]. Prebiotic effects of AOS on the growth of probiotic microbial cells are summarized (Table 1).

Table 1 . Prebiotic effects of arabino-oligosaccharides (AOS) on microbial cell growth..

Prebiotic AOS candidates	Growth-stimulated target microorganisms	Ref.
Linear AOS (DP 2~10)	Bifidobacterium longum ATCC 15707, B. breve ATCC 15700, B. adolescentis ATCC 15703, Bacteroides vulgatus ATCC 8482	[23]
Arabinan-hydrolyzed AOS fractions	Bifidobacterium, Bacteroides spp.	[19]
Acid-hydrolyzed AOS (DP 1~11)	Bifidobacterium, Bacteroides/Prevotella spp.	[24]
Linear and branched AOS (DP 2~10)	Bifidobacterium, Lactobacillus spp. (from patients with ulcerative colitis)	[20]
Feruloylated and non-feruloylated AOS	Bifidobacterium spp.	[18]
High (>1 kDa) and low M.W. AOS	Bifidobacterium spp.	[17]
Linear AOS (Mainly DP 2~4)	Bifidobacterium longum ATCC 15707, Lactobacillus brevis ATCC 14869, Bacteroides fragilis ATCC 25285	[16]
Branched AOS, debranched AOS	Bacteroides cellulosilyticus, B. dorei, B. eggerthii, B. finegoldii, B. fragilis, B. intestinalis, B. massiliensis, B. oleiciplenus, B. ovatus, B. stercoris, B. thetaiotaomicron, B. vulgatus, B. xylanisolvens	[13]

In vitro fermentation of sugar beet arabinan (SA) and AOS by the human gut microbiota indicated that AOS with various DP determine the fermentation characteristics by human gut bacteria. Different catalytic reactions were observed in bifidobacteria depending on the DP of AOS [19]. Through confirmation of growth in 24 strains of gut bacteria under given conditions, the digestive ability in the human gastrointestinal tract and prebiotic effects in the colon were established for linear arabino-oligosaccharides (LAOS) with mainly DP 2~4. LAOS has been shown to promote the growth of Lactobacillus brevis ATCC 14869, Bifidobacterium longum ATCC 15707, and Bacteroides fragilis ATCC 25285, increasing the population of bifidobacteria and influencing the production of short-chain fatty acids (SCFAs) [16]. Feruloylated and non-feruloylated AOS derived from sugar beet pectin were confirmed to selectively stimulate the growth of Bifidobacterium spp. through in vitro fermentation [18]. Candidate substances, such as phenylalanine, xanthine, linoleic acid, or its derivatives, confirmed to influence microbial communities as metabolites, were found to be generated more abundantly during AOS fermentation compared to FOS [17]. Microbial strains whose growth is stimulated by AOS are predominantly identified as bifidobacteria and bacteroides, with limited prebiotic effects observed in certain lactobacilli. The limited research on the prebiotic effects and health benefits of AOS, compared to L-arabinose monosaccharides and arabinan polymer, can be attributed to the absence of established mass-production technology for AOS. To address this limitation, it is crucial to discover novel ABNs with diverse hydrolytic properties and develop robust industrial-scale AOS production technologies.

Recent advancements in biological big data, including genomic and proteomic information, have facilitated the comprehensive analysis of microbial genomes, accelerating the identification and functional characterization of novel genes encoding ABNs with promising industrial applications. The L-arabinose/arabinan utilization gene cluster consists of genes involved in the enzymatic degradation of arabinan and AOS, as well as those related to AOS transport and L-arabinose metabolism. Typically, carbohydrate utilization gene clusters encode a variety of exo-/endo-type and intracellular/extracellular hydrolases that function synergistically in substrate breakdown. While a-1,2-/a-1,3-specific ABF genes can be found in both arabinan and arabinoxylan utilization gene clusters, a-1,5-specific ABN genes are exclusively localized within arabinan utilization clusters. Consequently, ABN-encoding genes serve as specific genetic markers for identifying and characterizing arabinan utilization gene clusters, which provides a powerful tool for exploring microbial pathways related to arabinan metabolism and facilitating biotechnological applications [15].

Previous studies investigating the structure and function of genes involved in arabinan utilization have been conducted across a diverse range of microorganisms such as Bacillus subtilis [25, 26], Bacteroides spp. [13, 27], Hypocrea jecorina [28], Bifidobacterium longum spp. [15, 29], Geobacillus stearothermophilus [14], Lactobacillus crispatus [30], and Aspergillus niger [31]. These studies have provided valuable insights into the genetic organization, regulatory mechanisms, and enzymatic properties of arabinan-degrading enzymes, contributing to our understanding of microbial arabinan utilization and its potential for biotechnological applications.

As mentioned above, arabinan is a pectic polysaccharide with an α-1,5-L-arabinofuranosyl backbone and α-1,2 and/or α-1,3-L-arabinofuranosyl branches. Complete degradation of arabinan into L-arabinose units is achieved by the synergistic action of two types of hydrolases, exo-acting ABF (EC 3.2.1.55) and endo-ABN (EC 3.2.1.99). ABF mainly debranches α-1,2 and/or α-1,3-linked side chains, whereas ABN hydrolyzes α-1,5-linked backbone of arabinan to produce AOS. According to the CAZy (Carbohydrate-active enzymes database; http://www.cazy.org/), enzymes involved in substrate degradation are classified into distinct glycoside hydrolase (GH) families based on their catalytic mechanisms and sequence/structural homology [32]. Among these, most ABFs are distributed across GH families such as GH51, 43, and 61, while ABNs are predominantly found in GH43, characterized by a five-bladed β-propeller-shaped catalytic domain structure [33]. Almost all ABNs, except for a limited number of exo-1,5-α-L-arabinanases GH93 (exo-ABNs GH93), are known to share the structural fold of GH43. Microbial ABNs GH43 are divided into two enzyme groups, endo- and exo-ABNs, which are distinguished based on the overall structure of the catalytic cleft and the structure of the glycon site, determining their endo- or exo-action on arabinan or AOS [33-36]. The protein information and enzymatic properties of various microbial endo-ABNs GH43 known to date are compared and summarized (Table 2).

Table 2 . Characteristics of microbial endo-1,5-α-L-arabinanases Glycoside Hydrolase (GH) 43 family..

No.	ABNs	Microbial origins	GenBank ID	AAa	Substratesb	Productsc	Ref.
Prokaryotic endo-ABNs GH43
1	Endo-1,5-α-L-arabinanase	Bacillus licheniformis	BAI47539.1	291*	DA	AOS	[37]
2	BLABNase	Bacillus licheniformis	WP_011197792.1	291*	DA≥LA	AOS (2-4)	[38]
3	AbnA	Bacillus subtilis	WP_003229496.1	291*	LA>SA	AOS	[26]
4	Abn2	Bacillus subtilis	WP_003244268.1	443*	LA≥SA>>P	AOS	[25]
5	ABN-TS	Bacillus thermodenitrificans	BAB64339.1	313	DA≥SA	AOS (2)	[39]
6	AbnA	Bacteroides sp. (Termite gut)	CCO20984.1	618*	SA>DA	AOS (4-5)	[27]
7	BT0360	Bacteroides thetaiotaomicron	AAO75467.1	622*	SA>>DA	AOS	[13]
8	BT0367	Bacteroides thetaiotaomicron	AAO75474.1	488*	DA>>SA	AOS	[13]
9	BflsABN43A	Bifidobacterium longum subsp. suis	WP_032684165.1	505*	DA>>SA	AOS (2-4)	[15]
10	BflsABN43B	Bifidobacterium longum subsp. suis	WP_032684164.1	581	DA	AOS (2)	[15]
11	Endo-1,5-α-L-arabinanase	Caldicellulosituptor Saccharolyticus	WP_011917087.1	488	DA>>SA	AOS	[40]
12	AbnB	Geobacillus stearothermophilus	ACE73676.1	315	DA	AOS (2-3)	[34]
13	AbnA	Lactobacillus crispatus	WP_100732642.1	773*	LA>SA	AOS	[30]
14	AbnB	Lactobacillus crispatus	WP_181577067.1	518*	LA	AOS	[30]
15	AbnZ1	Paenibacillus polymyxa	AHA50062.1	288*	LA>DA>SA	AOS (2-3)	[41]
16	AbnA	Thermotoga petrophila	ABQ46657.1	452*	SA>DA	AOS (2-3)	[42]
17	Tth Abn	Thermotoga thermarum	AEH51204.1	447*	LA≥DA>SA	AOS (2-3)	[43]
18	MC72GH43-2	Xylanivirga thermophila (Biogas reactor metagenome)	UKG18730.1	859	DA>SA	AOS (2-3)	[44]
19	MC68GH43-1	Xylanivirga thermophila	UKG18727.1	524	SA>DA	AOS (2-3)	[44]
Eukaryotic endo-ABNs GH43
20	ARA1	Aspergillus aculeatus	AAG27441.1	302*	DA>SA	AOS (2-3)	[45]
21	BcAra1	Botrytis cinerea	XP_001551065.1	290*	DA≥LA	AOS	[46]
22	AbnS1	Penicillium Chrysogenum	BAJ49868.1	293*	DA	AOS (2-3)	[47]
23	ABN1	Penicillium purpurogenum	ATY46024.1	309*	DA>>SA	AOS	[48]
24	PcARA	Phanerochaete chrysosporium	AFN42888.1	293*	DA	AOS (2-3)	[49]
25	RmArase	Rhizomucor miehei	AHF50152.1	288*	DA>SA	AOS (2)	[50]

^aAmino acid residues (AA) constituting each enzyme is presented with an asterisk if the predicted signal peptide sequence was excluded..

^bAbbreviations: DA, debranched arabinan; LA, linear arabinan; SA, sugar beet arabinan; P, pectin; AOS, arabino-oligosaccharides..

^cEndo-ABNs can produce linear and/or branched AOS from DA and SA, respectively, and the degrees of polymerization (DP) of resulting major linear AOS products were shown in parentheses..

Endo-ABNs GH43 can randomly hydrolyze internal α-1,5-L-arabinofuranosidic backbone of debranched (linear) and/or branched substrates. These hydrolases demonstrate higher catalytic activity on debranched arabinan (DA) and LAOS compared to branched substrates such as sugar beet arabinan (SA) [26, 34, 38]. While the enzymatic hydrolysis efficiency on SA is relatively low, endo-ABNs GH43 can generate branched AOS (BAOS) with diverse DP and branching patterns from SA [13, 27, 42, 44]. In contrast, short-chain LAOS, predominantly with DP ranging from 2 to 4, can be produced through the endo-hydrolysis of DA or long-chain LAOS. This selective enzymatic behavior highlights the substrate specificity and product distribution of endo-ABNs, which are critical factors for optimizing the industrial production of targeted AOS with desired structural and functional properties.

Exo-1,5-α-L-arabinanases (exo-ABNs) are categorized into two distinct glycoside hydrolase families, GH43 and GH93 (Table 3). Exo-ABNs belonging to GH43 exhibit a five-bladed β-propeller architecture, sharing structural similarities with endo-ABNs GH43, particularly in the substrate-binding site conformation.

Table 3 . Characteristics of exo-1,5-α-L-arabinanases Glycoside Hydrolase (GH) 43 and 93 families..

No.	ABNs	Microbial origins	GenBank ID	AAa	Substratesb	Products	Ref.
exo-ABNs GH43
1	Arb43A	Cellvibrio japonicus (Pseudomonas cellulosa)	ACE84667.1	316*	CMA≥LA, AOS	A3	[51]
2	ARN3	Bovine ruminal metagenome	ADB43999.1	327*	CMLA≥DA≥LA, AOS	A1	[52]
exo-ABNs GH93
3	ReAbn93	Rasamsonia emersonii	XP_013327992.1	362*	A4≥A3≥LA	A2	[53]
4	Arb93A	Fusarium graminearum	CAX17703.1	365*	DA>>SA, AOS	A2	[54]
5	Abnx	Penicillium chrysogenum	BAC76689.1	354*	DA>>SAG≥SA, AOS	A2	[55]
6	Arap2	Talaromyces purpureogenus (Penicillium purpurogenum)	AKH40271.1	363*	DA>>BX≥RA≥SA, AOS	A2	[56]
7	Abn93T	Thermothielavioides terrestris	XP_003653830.1	364*	DA, AOS	A2	[57]
8	Abn2	Chrysosporium lucknowense	XP_003660737.1	361*	LA>SA, AOS	A2	[58]

^aAmino acid residues (AA) constituting each enzyme is presented with an asterisk if the predicted signal peptide sequence was excluded..

^bAbbreviations: CMA, carboxymethyl arabinan; LA, linear arabinan; CMLA, carboxymethyl linear arabinan; DA, debranched arabinan; SA, sugar beet arabinan; BX, birchwood xylan; RA, rye arabinoxylan; SAG, soybean arabinogalactan; AOS, arabinooligosaccharides; A1, L-arabinose; A2, arabinobiose; A3, arabinotriose; A4, arabinotetraose..

Despite these structural parallels, an ABN GH43 from Cellvibrio japonicus (CjArb43A) displays a unique exo-type hydrolysis pattern, selectively producing only arabinotriose from DA substrate [33, 51]. In contrast, ARN3, isolated from the bovine ruminal metagenome, demonstrates an entirely different exo-type catalytic behavior, producing only L-arabinose as the final hydrolysis product [52]. Unlike the conventional ABNs GH43, the other group of exo-ABNs GH93 possesses a six-bladed b-propeller fold structure, which primarily releases arabinobiose from DA, but hardly hydrolyzes SA [53-58]. Their modes of action are distinctly different from endo-ABNs, which typically generate a series of short-chain AOS through random internal cleavage of α-1,5-L-arabinofuranosidic bonds. These findings underscore the functional diversity among exo-ABNs, despite their structural similarity to endo-ABNs, and highlight the substrate specificity and unique product profiles dictated by subtle differences in the active site architecture and substrate interaction mechanisms.

The three-dimensional structures of endo- and exo-ABNs GH43 predominantly feature a catalytic domain with a five-bladed β-propeller fold. This structural motif facilitates the enzymatic breakdown of arabinan and AOS with high specificity and efficiency. However, a subset of exo-ABNs GH93 possesses a six-bladed β-propeller fold in their catalytic domain, distinguishing them structurally and functionally from ABNs GH43. To elucidate the structural differences and similarities between these ABNs, comparative analyses of their primary and tertiary structures were performed and comprehensively illustrated (Fig. 1).

Figure 1. (A) Phylogenetic relationship of endo- and exo-arabinanases GH43 and GH93 and (B) Structures of multi-domain endo-arabinanases GH43. The phylogenetic tree was drawn using MEGA11 software and iTOL web server. Multi-domain structures were analyzed using InterPro and SignalP web servers. The numbering assigned to each endoand exo-ABNs corresponds to the identifiers presented in Tables 2 and 3. The black triangle indicates a signal peptide for protein secretion.

Phylogenetic analysis of various ABNs based on amino acid sequence identity revealed a clear evolutionary distinction between exo-ABNs GH93 and endo-ABNs GH43 (Fig. 1A). This result highlights significant differences in amino acid sequences and catalytic actions between exo-ABNs GH93 and ABNs GH43, reflecting their functional and structural divergence. In contrast, two exo-ABNs GH43 exhibit higher sequence similarity with single-domain endo-ABNs GH43 than with exo-ABNs GH93. This closer evolutionary relationship suggests a shared structural and functional ancestry between exo- and endo-ABNs GH43, despite their distinct catalytic modes of action. Furthermore, some endo-ABNs GH43 possess multi-domain architectures, comprising an additional C-terminal domain alongside the catalytic domain for GH43 hydrolase. Beyond these core domains, certain ABNs include optional domains such as fibronectin type-III (FN3) domain, surface layer protein A (SlpA) domain, atrophied bacterial Ig (aBig) domain, or laminin-G domain, respectively (Fig. 1B).

Comparative structural analysis provides critical insights into the molecular mechanisms underlying substrate recognition and catalytic specificity of ABNs GH43 and GH93. These contribute significantly to their potential biotechnological applications, particularly in the targeted production of AOS. The first resolved three-dimensional structure of an ABN GH43 with a five-bladed β-propeller fold was reported for CjArb43A from Cellvibrio japonicus [33]. This enzyme contains three essential catalytic acidic residues within its active site and hydrolyzes arabinan and AOS into arabinotriose via an inverting catalytic mechanism. Three-dimensional structures of endo-ABNs GH43 in complex with AOS substrates have been elucidated from Geobacillus thermodenitrificans (PDB: 6A8I with arabinohexaose) [59], G. stearothermophilus (3D61 with arabinobiose, 3D5Z with arabinotriose) [34], and Bacillus subtilis (2X8S with arabinotriose) [35]. Additionally, the structures of a thermostable endo-ABN GH43 from Thermotoga petrophila RKU-1 (4KC8) and ARN2 GH43 discovered from the bovine rumen metagenome (4KCA) were also reported [60].

The crystal structures of an intracellular, monomeric, and single-domain endo-ABN GH43 (GsAbnB; 315 amino acids) from Geobacillus stearothermophilus T-6, in complex with various AOS, revealed the mechanism of substrate-binding and catalytic action of typical endo-ABNs GH43.

The first acidic residue (E201) of GsAbnB acts as a general acid, protonating the leaving aglycon and making it a better leaving group. The second acidic residue (D27) acts as a general base, activating a water molecule for a single-displacement attack on the anomeric carbon, resulting in the inversion of the anomeric configuration. The third residue (D147) probably maintains the correct alignment of the general acid residue with the substrate, modulates the pK_a of the general acid residue, and stabilizes transition states such as the oxocarbenium ion through interactions with the 2-OH of the glycon [33, 34, 61-63]. The structures of GsAbnB have been comparatively analyzed in complex with arabinobiose (PDB: 3D61), arabinotriose (3D5Z), and arabinopentaose (6F1G). Isothermal titration calorimetry (ITC) analysis of catalytic mutants of GsAbnB interacting with various AOS revealed that the active site comprises at least five subsites (-2, -1, +1, +2, and +3) capable of accommodating arabinopentaose [34]. The catalytic triad residues (D27, D147, and E201) are precisely positioned with appropriate spatial distances and orientations to hydrolyze an α-L-arabinofuranosidic linkage between L-arabinose units at subsites -1 and +1. Arabinobiose is challenging to hydrolysis because it primarily binds to the +1 and +2 subsites of the enzyme, preventing proper positioning within the catalytic cleft. As a result, GsAbnB exhibits extremely low hydrolytic activity on arabinobiose, preventing its decomposition into L-arabinose monomers. In the case of arabinotriose, it binds weakly to the subsites -1, +1, and +2 and is slowly decomposed into L-arabinose and arabinobiose by GsAbnB. Arabinopentaose interacts with all five subsites (-2, -1, +1, +2, and +3) within the active site. This extended binding interaction enhances substrate affinity and supports efficient catalytic activity, making arabinopentaose and longer substrates such as DA the preferred substrates for GsAbnB.

Exo-ABNs are enzymes known for their exo-acting hydrolytic activity, and they exhibit different modes of action compared to endo-ABNs and exo-ABFs. For instance, an exo-ABN GH43 from Cellvibrio japonicus (CjArb43A, PDB: 1GYD) releases only arabinotriose from the non-reducing end of arabinan [33, 51] (Fig. 2B). Although CjArb43A shares a typical five-bladed β-propeller structure with common endo-ABNs GH43, it differs in the loop sequence connecting the second and the third blades. This loop sequence consists of 13 amino acids, in contrast to 4-6 amino acid residues found in the short loop of typical endo-ABNs GH43. However, the loop structure in CjArb43A has been confirmed not to influence its exo-type modes of action. Mutagenesis studies revealed that the amino acid residues responsible for steric restrictions at subsite -3 play a crucial role in its hydrolytic modes of action, leading to a transition from exo-acting to endo-acting activity [64]. CjArb43A is structurally very similar to the typical endo-ABNs GH43 but has a high substrate affinity at subsite -3, so that the long-chain substrates can bind in a specific orientation and position, exhibiting the exo-acting hydrolysis characteristics. In contrast, ARN3 (PDB: 4KCB), an exo-ABN GH43 discovered from the rumen metagenome, was known to have a loop structure (28 amino acids) affecting its exo-acting activity [60]. ARN3 catalyzes the exo-acting degradation of linear arabinan from the reducing end to produce L-arabinose units [52]. When the loop sequence of ARN3 was replaced with that of endo-ABN (SRGEEP, 5 amino acids), the mutant ARN3 showed increased catalytic efficiency, enhanced substrate affinity, and a shift to an endo-acting mode of action, respectively [60]. While most exo-acting ABNs GH43 are known to act on the non-reducing end, only a few exo-hydrolases were confirmed to act on the reducing end.

Figure 2. Three-dimensional structures of microbial exo- and endo-1,5-α-L-arabinanases (ABNs) GH43 and GH93 in complex with substrates. Surface, cartoon, and closed-up structures are represented for (A) endo-ABN GH43 from Geobacillus stearothermophilus (PDB: 3CU9, catalytic residues of D27, D147, and E201 in blue) in complex with arabinobiose (3D61 in cyan), arabinotriose (3D5Z in magenta), and arabinopentaose (6F1G in green), (B) exo-ABN GH43 from Cellvibrio japonicus with arabinohexaose (PDB: 1GYD and 1GYE, catalytic residues of D38, D158, and E221 in blue), and (C) exo-ABN GH93 from Penicillium chrysogenum with arabinobiose (PDB: 3A72, catalytic residues of E170, E242, and Y337 in blue). The substrate-binding site for each enzyme is illustrated with designated subsite numbers. The substrate cleavage sites and non-reducing ends of substrates are marked with arrowheads and asterisks, respectively. Each structure is visualized using PyMOL molecular graphics tool.

Some exo-ABNs have a six-bladed β-propeller fold structure, classifying them under the GH93 family. The Abnx from Penicillium chrysogenum 31B is an exo-ABN belonging to the GH93 family and hydrolyzes DA from the non-reducing end to release only arabinobiose as a product [55]. The structure of Abnx (PDB ID: 3A72) was shown in complex with arabinobiose (Fig. 2C). It is difficult to understand the exact substrate binding and hydrolysis mechanism of exo-ABNs GH93 because a few complex structures with substrates have been elucidated to date. Based on the structure in complex with arabinobiose, the location of subsites -1 and -2 can be predicted, which explains the structural reason why Abnx can exclusively generate arabinobiose as a product, not L-arabinose. The structure of Arb93A from Fusarium graminearum (PDB ID: 2W5N) is another example of exo-ABN GH93, and its modes of action involve the non-reducing end of linear arabinan [54].

Since arabinan is a homopolymer composed exclusively of L-arabinose, early research primarily focused on achieving high-yield production of L-arabinose through the synergistic action of endo-ABN and exo-ABF [65, 66]. However, the complete decomposition of arabinan into L-arabinose is possible not only by enzymes but also by acid treatment. Accordingly, the research focus has shifted towards the production of AOS using endo-ABNs, as AOS exhibit prebiotic properties and functional benefits beyond L-arabinose. This transition reflects an emerging industrial interest in value-added bioproducts, where ABNs play a pivotal role in controlling the DP and structural profiles of AOS. As described above, microbial ABNs found in nature have been reported to exhibit diverse structures, substrate specificities, and hydrolysis mechanisms. In this study, it is suggested that the appropriate utilization of exo- and endo-type ABNs can enable the production of AOS with a wide range of DP and diverse chemical structures (Fig. 3).

Figure 3. Scheme for enzymatic production of prebiotic arabino-oligosaccharides using exo- and endo-1,5-α-Larabinanases (ABNs) GH43 and GH93. CjArb43A, exo-ABN GH43 from Cellvibrio japonicus; exo-ABF, α-Larabinofuranosidase with debranching activity.

For BAOS production, specific endo-ABNs GH43 with hydrolytic activity toward SA should be employed. While most endo-ABNs GH43 exhibit higher activity on DA, some enzymes have been reported to demonstrate significant hydrolytic activity toward SA as well. When SA is treated with these specialized endo-ABNs GH43, BAOS with varying degrees of branches and diverse branched structures can be produced through the hydrolysis of α-1,5-L-arabinofuranosidic linkages within SA. For example, BAOS were produced through the enzymatic treatment of three arabinohydrolases, Abn1 (endo-ABN), Abn2 (exo-ABN), and Abn4 (exo-ABF), derived from Chrysosporium lucknowense [58]. The resulting BAOS were classified into two distinct types, α-1,3-single-substituted and α-1,2-/α-1,3-double-substituted BAOS. The chemical structures of eight different BAOS were elucidated through detailed structural analysis [67]. LAOS can be produced using endo- or exo-ABNs with various hydrolytic properties, using DA as the substrate. However, the production of DA as a starting substrate is a prerequisite, which necessitates the use of ABFs capable of efficiently debranching the α-1,2- and α-1,3-linked side chains from SA. These ABFs selectively hydrolyze side-chain substitutions, enabling the conversion of branched arabinan (SA) into a linear form (DA) suitable for subsequent enzymatic hydrolysis by ABNs [68]. Once DA is generated from SA through ABF treatment, it can serve as a precursor to produce various AOS. When DA is treated with general endo-ABNs GH43, a mixture of short-chain LAOS with a DP ranging from 2 to 4 can be produced. The composition and distribution of LAOS in the mixture may vary depending on the specific hydrolytic characteristics of the ABN employed. The desired AOS fractions can be isolated and purified using column chromatography, filtration, and/or crystallization techniques. The purified AOS fractions are then utilized for further studies and functional analyses to explore their prebiotic potential and biological activities [17, 19, 20].

When exo-ABNs GH43 are employed, AOS with a specific DP can be produced, depending on the hydrolytic characteristics of the enzyme used. This selective production of single AOS species offers a significant advantage in achieving economic feasibility during the industrial downstream processes, as it simplifies successive purification and fractionation steps. For example, when exo-type CjArb43A GH43 is applied to DA, the main product is exclusively arabinotriose, depending on its unique hydrolytic modes of action. In contrast, most exo-type ABNs GH93 can predominantly produce arabinobiose, offering a targeted and efficient AOS production process. The enzyme-specific hydrolysis patterns enable precise control over AOS production, facilitating their application in functional foods, pharmaceuticals, and nutraceutical industries where consistency and structural specificity are critical.

The demand for next-generation prebiotic carbohydrate materials as alternatives to widely used existing prebiotics is steadily increasing. Pentose-based oligosaccharides derived from hemicellulose, such as AOS, are gaining significant attention due to their potential health benefits and selective prebiotic properties. To produce structurally and functionally diverse AOS, the development of novel enzymes with unique hydrolytic properties is essential. In this review, we have presented the prebiotic effects of AOS, the classification of various ABNs available for AOS production, enzymatic properties, their structure-function relationships, and the foundation for utilizing enzymes in the efficient production of AOS. Recently, global interest in the relationship between dietary intake and gut microbiome diversity has surged, leading to an increase in research on functional synbiotics, which involves the strategic pairing of prebiotics and probiotics [69-75]. This research field aims to explore how specific combinations of prebiotic carbohydrates and probiotic strains can synergistically enhance gut health, modulate microbial composition, and improve host well-being. This requires an integrated approach, including (1) discovery and characterization of enzyme resources with novel catalytic properties, (2) development of industrial-scale mass production and purification techniques for enzymes, (3) enhancement of enzyme properties through protein engineering, (4) establishment of an efficient mass production process for prebiotic materials via enzyme treatment, (5) elucidation of the functional effects of synbiotics using advanced multi-omics (e.g., genomics, transcriptomics, proteomics, metabolomics, and so on) analyses. Integrating these processes enables a comprehensive understanding of enzyme-substrate interactions, microbial metabolic pathways, and prebiotic utilization mechanisms, for the efficient design and production of next-generation synbiotics with enhanced functional properties and commercial availability. Microbial exo- and endo-ABNs are indispensable biocatalysts for the enzymatic production of prebiotic AOS, offering significant potential for applications across the health, food, pharmaceutical, and agricultural sectors. Ongoing advancements in enzyme technology will be essential for unlocking the full commercial and biotechnological potential of prebiotic AOS, paving the way for innovative applications in gut health and functional nutrition.

This research was supported by Korea Institute of Planning and Evaluation for Technology in Food, Agriculture and Forestry (IPET) through Agricultural Microbiome R&D Program for Advancing Innovative Technology, funded by Ministry of Agriculture, Food and Rural Affairs (MAFRA) (RS-2024-00401736) and by "Regional Innovation Strategy (RIS)" through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (MOE) (2021RIS-001).

The authors have no financial conflicts of interest to declare.