Strategies for GH functional annotation: toward harmonized and predictive CAZymology

Abstract

Background

The functional annotation and biochemical characterization of glycoside hydrolases (GHs) for polysaccharides and oligosaccharides deconstruction are a major challenge. Here, we traced the transition from traditional “activity first” approaches, based on biochemical assays, to modern “sequence first” discovery methods enabled by modern genomics.

Results

We reported 169 glycosidase activities, the substrates, and the techniques used for the characterization of GHs. We highlight the challenges posed by the diversity of GH families including overlapping substrate specificities, heterogeneous assay conditions, and inconsistent reporting standards. To address these issues, we compiled a list of substrates for each GH family’s glycosidase activities as a practical guide to serve as an initial framework for characterizing newly identified GHs.

Conclusions

The adoption of standardized substrates and protocols, uniform reporting practices, and community-driven data sharing initiatives provides a practical roadmap toward harmonized GH functional annotation and predictive enzymology while acknowledging current limitations in data standardization and reporting heterogeneity.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13068-026-02781-3.

Background

Glycoside hydrolases (GHs), also called glycosidases, are carbohydrate-active enzymes (CAZymes) that, along with polysaccharide lyases (PLs) and "auxiliary activities" (AAs), are the primary enzymes supporting the deconstruction of polysaccharides and oligosaccharides across environments [1]. The varying chemistry of single sugars, how these sugars assemble in oligosaccharides and polysaccharides, and their association with proteins (i.e., glycoproteins) and lipids (i.e., glycolipids) produce a vast repertoire of substrates for GHs [2, 3]. Like other CAZymes, the classification of GHs in protein families mirrors their overall structure and sequence similarities [4]. GHs accommodating and cleaving a multitude of compounds have found applications in various industries (e.g., biofuels, drug development, food processing, bioremediation or detergents). In fact, the development of modern enzymology itself is closely linked to the enzymatic characterization of GH activities. The first "bioactive compound", later known as an amylase enzyme, was described by Payen and Perzos in 1833 when they identified the "diastase" reagent, meaning "separation" in Greek, used to separate starch from barley grain [5]. Although no protein was purified then, this diastase was the first biological catalyst to be scientifically described and used in industry (e.g., brewing). Interestingly, the "-ase" suffix in enzyme names (e.g., amylase, cellulase) is derived from the name "diastase" coined by Payen and Perzos to describe their "separation" process. Similarly, the "cytase" activity, for cellulose deconstruction, coined by Green in 1888, was instrumental for the development of enzymology and for understanding the breakdown of polysaccharides [6].

Like the diastase and the cytase activities, many glycosidase activities were identified before their corresponding genes. In these "activity first" discoveries, the active component of diastase and cytase was not precisely identified nor purified. However, thanks to the development of molecular biology and protein biochemistry (e.g., recombinant DNA technology, protein purification), suspensions of active compounds could be fractioned; and, by the 1950 s, glycosidase proteins were routinely purified and characterized (e.g., [7]). In parallel, methods for substrate characterization and preparation improved, leading to pure natural and synthetic substrates replacing mixed substrate suspensions (e.g., grain, plant cell wall).

Activity-first approaches are still used in modern microbiology and biochemistry as they allow for the identification of new enzymes or activities with no a priori knowledge (e.g., [8, 8–10]). Often, when performing activity-driven screening to identify new enzymes, isolates and clones are plated on media with limited sources of carbon or supplemented with modified substrates (e.g., AZCL-xylan). Then colonies that form or degrade the substrate are collected and analyzed (e.g., sub-cloning and sequencing). Together these "activity first" approaches contributed to the growing number of GH families identified over the years.

However, with DNA sequencing’s increase in accessibility, “sequence first” approaches have progressively become a cornerstone in the enzyme discovery process and, currently, most methods rely on sequence comparison to identify new variants of known GHs (e.g., [11–13]). Alternatively, some sequence-driven techniques rely on comparative genomic techniques to identify new candidate genes and enzymes using conserved genomic features such as the synteny and the clustering of genes (e.g., [14, 15]).

Finally, whether new enzymes are selected through activity- or sequence-driven approaches, the biochemical characterization of the glycosidase activity generally depends on: (i) a priori knowledge about the GH families; (ii) the availability of adequate substrates, and (iii) the technical capacity to perform the sometimes complex biochemical assays.

As most GH families contain multiple activities, identifying the GH family alone is not enough to infer the function of newly identified enzymes [16–18]. However, knowing the GH family can reduce the number of activities and substrates to consider. Recent efforts such as the CAZac framework also aim to improve CAZyme functional annotation by incorporating mechanistic and structural descriptors beyond EC classifications [19]. In vitro characterization of GHs for biomass deconstruction generally requires soluble substrates and the ability to identify and quantify the hydrolysis products. This is typically achieved by titrating the reducing ends generated during hydrolysis or using chromogenic substrates. In addition, since glycosidases can release multiple products, particularly when acting on natural substrates, chromatographic techniques can identify individual intermediate hydrolysis products enabling estimation of kinetic parameters and even revealing the precise mode of action of the enzymes, among other benefits.

When considering the diversity of GH families, glycosidase activities, existing substrates, and described biochemical assays, characterizing new GH proteins can be a daunting task. Among others, the complexity of the CAZyme classification system, the wide range of supported activities, the variety of substrates and detection techniques available, combined with the lack of standardized approaches for characterizing and reporting the activity of new GHs lead to inconsistent and difficult to interpret results. In addition, the limited availability of substrates, due to technical or commercial constraints, and the restriction of complex biochemical assays (e.g., HPLC-PAD) to large, well-funded labs hinder the integration and interpretation of existing literature on GH characterization.

Here, we investigated 169 reported glycosidase activities (i.e., 3.2.1.-), their substrates, and the techniques used for their characterization. Additionally, we compiled the reported activity for each GH family and proposed a streamlined list of substrates to test when characterizing new GHs. Our recommendations are based on published results, commercial availability of substrates, and ease of detection and quantification. These recommendations are intended as a starting point for the characterization of newly identified GHs that could be complemented with more advanced techniques. While focusing on GHs for polysaccharide deconstruction, we recognize that activities targeting glycoproteins, lipopolysaccharides, and short glycosylated secondary metabolites and drugs generally require advanced or specialized approaches, not discussed here.

Data collection and processing

First, data for biochemically characterized GH proteins were retrieved from the CAZy-db web interface, including their family associations and reported enzymatic activity (i.e., EC numbers), as described previously [18]. Next, for the listed enzymatic activities, we retrieved the complete list of substrates from the BRENDA database [20]. These two datasets were merged, and data processing, statistical analysis, and visualization were performed using R (version 4.4.3) and the following packages: gplots (3.2.0), ggplot2 (3.5.1), and dplyr (1.1.4).

Finally, the GH families, corresponding activities, and substrate lists were compiled into a comprehensive table (Supplementary data 1) and into a user-friendly Excel spreadsheet (Supplementary data 2) including formulas allowing the user to enter either GH families or EC numbers and see the distribution of EC numbers or substrates, respectively.

Distribution of activities across the GH classification

Linking new GHs to a specific activity is not trivial. Since the early-90s, enzymes active on carbohydrates have been categorized in the Carbohydrate Active enZYme database (CAZy-db, [4]) based on sequence and structural similarities. As of January 2025, 189 distinct GH families have been described, and this number is still increasing. By November 2025, the classification included 194 GH families. However, over the years, several historical GH families (GH21, GH40, GH41, GH60, GH61, GH69, GH145, and GH155) have been deleted or reclassified and are therefore no longer considered active GH families within the current CAZy framework. Briefly, the classification of GHs into families mirrors their conserved structure and sequence, not their activity. In fact, many GH families, and eventually subfamilies, are associated with several distinct enzymatic activities revealing diverging evolutionary processes among closely related enzymes [18, 21]. Furthermore, many described activities have been identified in multiple unrelated GH families, highlighting some converging evolutionary processes across CAZymes [16]. Finally, many individual GHs can accommodate and process multiple substrates, albeit with varying activity [22, 23] adding to the complexity of linking newly identified GH-sequences to specific activities.

As of 2025, most known enzymatic activities are categorized using the Enzyme Commission number (EC) first introduced in 1961 (for review see [24]) and then updated by the International Union of Biochemistry and Molecular Biology (IUBMB, https://iubmb.org). Per the IUBMB, glycosylases (i.e., EC.3.2) are divided into glycoside hydrolases (GHs), also known as glycosidases (i.e., EC.3.2.1), that cleave O- and S-glycosyl compounds, and other enzymes which hydrolyze N-glycosyl-compounds (i.e., EC.3.2.2) such as purine nucleosidases (EC.3.2.2.1) and uridine nucleosidases (EC.3.2.2.2). In its initial release, the EC classification listed just 33 glycosidase activities, some of which still exist as of 2025 such as EC.3.2.1.1 (α-amylase) and EC.3.2.1.2 (β-amylase). However, other entries were revised, such as EC.3.2.1.4 (cellulase), last revised in 2001, and some, such as EC.3.2.1.16 (alginase), were deleted. Overall, the approved list of glycosidase activities expanded as more substrates with varying chemistry were tested. As of January 2025, over 200 described glycosidase activities (EC.3.2.1.-) are unevenly distributed across more than 180 GH families referenced in the CAZy-db. However, the total number of characterized GH enzymes (~10,000) pales in comparison to uncharacterized GHs currently listed on CAZy-db (~2 million).

Since CAZy-db lists the reported enzymatic activities of biochemically characterized GHs, we retrieved the EC numbers associated with each family (Fig. 1). Several GH families were associated with a single activity. Many of these monospecific GH families have been established recently or are associated with few characterized enzymes.

Fig. 1.

The number of reported enzymatic activities (i.e., EC#) mirrors the overall number of characterized proteins across GH families (r_pearson = 0.84, p < 0.01). For clarity, only GH families with more than 2 reported activities and more than 5 characterized proteins are shown

On the other hand, many populated GH families were associated with multiple activities [4]. For example, more than 25 activities are associated with the GH5 family, a marker of diverging evolution [21]. Alternatively, the apparent polyspecificity of some GH families could mirror biases in enzyme characterization, challenges in data reporting from a growing body of literature, or the pressure to find novelty even in well-characterized GH families.

In this context, knowing the exact distribution of reported enzymatic activities in each GH family (Fig. 2) (i) reveals what enzymatic activity can be expected from the characterization of newly identified proteins from known families; (ii) help select the substrate to be tested for the characterization of new proteins (see below); and (iii) eventually predict the activity of identified sequences.

Fig. 2.

Distribution (%) of reported enzymatic activity (EC#) within GH families, as of January 2025. For each GH family, the number of characterized enzymes (per the CAZy-db) is displayed. Monospecific GH families and activities found in only one GH family were excluded for clarity. (The full dataset, with no exclusions, is available in supplementary data 1.)

Polysaccharides and oligosaccharides for GH characterization

In nature, biomass polysaccharides can provide structure (e.g., cellulose, chitin, peptidoglycan) or store energy (e.g., amylose in starch). Pure polysaccharides (e.g., cellulose, amylose) used for the enzymatic characterization of GHs are extracted from raw materials (e.g., wood, grain, tuber) using chemical extraction and purification. Most structural polysaccharides (e.g., cellulose, chitin) have a linear structure, are sometimes decorated (e.g., xylan, pectin), have a high degree of polymerization (DP), and are generally insoluble in water. Conversely starch and glycogen, used to store energy, are composed of mostly insoluble branched amylopectin (i.e., mixed α−1,4/6-linked glucopyranoses) and mostly soluble, linear amylose (i.e., α−1,4 linked glucopyranoses). Most polysaccharides are abundant across environments and often associate with each other and with other compounds such as lignin in plants [25].

Since the precise characterization of enzymes requires preparation of pure and soluble compounds, techniques have been developed to (i) solubilize and purify polysaccharides, generally under harsh conditions (e.g., phosphoric acid-swollen cellulose, PASC [26]); (ii) chemically modify substrates to improve their solubility (e.g., carboxymethyl-cellulose, CMC [27]); (iii) remove chemical substitutions from decorated polysaccharides (e.g., xylan [28]); and (iv) distinguish between linear and branched polysaccharides (e.g., amylose vs. amylopectin [29]). Although many extraction–purification procedures involve harsh chemical treatments, some enzymatic processing techniques exist [30].

Many polysaccharides are routinely incorporated in culture media to serve as substrates in "activity first" screening projects. The degradation of polysaccharides within media post-incubation is revealed by staining (e.g., Congo-Red staining) [31] or through the addition of low doses of inert stains (e.g., Trypan blue) [32, 33]. In addition, polysaccharides are sometimes modified to facilitate these screening procedures. Notably, azurin-cross-linked polysaccharides (e.g., AZCL-xylan) from MegaZyme/Neogen, are frequently used. These substrates are insoluble polysaccharides associated with the azurin dye that can be incorporated directly in agar media. Strains or clones degrading the substrate release the azurin dye producing blue hydrolysis halos around the colonies.

Polysaccharides are also used to determine the kinetic parameters of individual enzymes. Soluble substrates are generally required as Michaelis–Menten kinetics procedure measures the initial velocity of a reaction with increasing substrate concentrations [34]. However, most preparations of polysaccharides contain substrates with varying degrees of polymerization (DP) making it impossible to know the exact substrate’s molar concentration. Instead, when testing polysaccharides, the substrate concentration is often expressed as a percentage (w/v) or gram per liter and apparent kinetic parameters (i.e., Km', kcat', and Vmax') deduced (e.g., [32]).

Importantly, the complete enzymatic deconstruction of most natural polysaccharides is not the result of a single catalytic activity but rather of synergistic and sequential enzyme systems [35–37]. For example, efficient cellulose hydrolysis typically requires the concerted action of endo-cellulases (EC.3.2.1.4) that introduce internal chain breaks, exo-cellulases or cellobiohydrolases (EC.3.2.1.91) that release cellobiose units from chain ends, and β-glucosidases (EC.3.2.1.21) that convert soluble cellodextrins and cellobiose into glucose. Similarly, complete xylan degradation often depends on endo-xylanases (EC.3.2.1.8) acting in synergy with debranching enzymes such as α-L-arabinofuranosidases (EC.3.2.1.55), acetyl xylan esterases (EC.3.1.1.72), and β-xylosidases (EC.3.2.1.37) to remove substitutions and generate fermentable sugars. Starch degradation also illustrates this principle, where α-amylases (EC.3.2.1.1) randomly cleave α−1,4 linkages, β-amylases (EC.3.2.1.2) release maltose from non-reducing ends, pullulanases (EC.3.2.1.41) hydrolyze α−1,6 branch points, and glucoamylases (EC.3.2.1.3) and α-glucosidases (EC.3.2.1.20) complete saccharification. Such interdependent activities can profoundly impact kinetic measurements as the accumulation of intermediate products can inhibit upstream enzymes (i.e., inhibition by the product), relieve product inhibition upon downstream processing, or artificially inflate reducing-end assays when multiple enzymes are present. Consequently, kinetic parameters obtained with isolated enzymes on simple substrates may not reflect their behavior within multi-enzyme systems acting on natural polymers. Accounting for enzyme synergy and sequential cooperativity is therefore essential when interpreting apparent kinetic constants, comparing activities across studies, or attempting to extrapolate in vitro measurements to biological or industrial processes [38–40].

Oligosaccharides, containing few sugars, are often derived from longer polysaccharides although some are synthesized de novo such as sucrose [41] and lactose [42]. They serve as energy storage molecules or act as antidesiccants by regulating osmotic pressure [43, 44], among other functions.

Although some oligosaccharides such as fructooligosaccharides (FOS) and galactooligosaccharides (GOS) are extracted directly from plants and milk/whey, respectively [45], most are intermediates released during catabolic processes. The production and purification of oligosaccharides remain challenging as they are typically obtained by partial degradation of longer polysaccharides or through synthesis, followed by purification. Specifically, solid-phase synthesis [46, 47] or enzymatic methods [48, 49] have been used. However, because these methods become more intricate as the number of sugar units increases, the larger the oligosaccharide, the more expensive it is. In addition, the solubility of oligosaccharides generally decreases with the DP and complexity. Thus, the overall production and purification of large and complex oligosaccharides can be expensive and sometimes prohibitive. However, the use of oligosaccharides in enzymology allows for more accurate quantification of the substrate and the product(s) and thus can produce exact kinetic parameters.

To facilitate navigation across the broad range of analytical approaches currently used for GH characterization, Table 1 summarizes the principal experimental techniques, the associated substrate types, and the type of functional information that can be obtained from each method. Detailed substrate–activity associations are available in Supplementary Data 1 and 2.

Table 1.

Overview of common analytical approaches used for GH characterization, associated substrate types, and principal information obtained

Technique/assay	Typical substrate types	Example substrates	Primary information obtained	Major advantages	Main limitations
Growth-based or plate screening	Insoluble natural polysaccharides; dyed polysaccharides	Cellulose, xylan, AZCL-xylan, CMC agar	Presence/absence of hydrolytic activity	Simple, inexpensive, high-throughput screening	Mostly qualitative; limited mechanistic information
Reducing-end assays (DNS, Somogyi–Nelson)	Soluble oligo- and polysaccharides	CMC, xylan, starch, cellooligosaccharides	Total reducing sugar release; apparent kinetic parameters	Fast, inexpensive, broadly applicable	No product identification; cannot distinguish endo/exo activity
Viscometry	Soluble polysaccharides	CMC, starch, soluble glucans	Mode of action (endo vs exo activity); polymer depolymerization	Useful for monitoring polysaccharide breakdown	Mostly qualitative; limited structural information
Thin-layer chromatography (TLC)	Soluble oligosaccharides and hydrolysis products	Cellobiose, xylooligosaccharides, arabinooligosaccharides	Product profiles; approximate mode of action	Low-cost and accessible	Semi-quantitative; co-migration of sugars possible
HPLC-RI	Mono- and oligosaccharides	Glucose, cellobiose, xylooligosaccharides	Product identification and quantification	Routine quantitative analysis; non-destructive	Lower sensitivity; limited structural resolution
HPLC/HPAEC-PAD	Neutral and charged oligosaccharides	Cellooligosaccharides, acidic sugars, substituted glycans	Product identification; kinetic parameters; product ratios	Highly sensitive and quantitative	Requires specialized instrumentation
LC/MS	Native or derivatized oligosaccharides	Xyloglucan fragments, substituted glycans	Molecular mass; oligomer composition; partial structural information	High sensitivity; compatible with complex mixtures	Regioisomer distinction can be difficult
GC/MS (methylation analysis)	Derivatized monosaccharides and oligosaccharides	Permethylated glycans	Glycosidic linkage positions; structural characterization	High structural resolution	Labor-intensive derivatization
NMR spectroscopy	Purified oligosaccharides and glycans	Defined glycan products	Linkage configuration; stereochemistry; complete structural characterization	Precise structural information	Expensive instrumentation; requires expertise
Chromogenic assays (pNP substrates)	Synthetic glycosides	pNP-β-D-glucopyranoside, pNP-cellobioside	Specific glycosidase activity; kinetic parameters	Sensitive, simple, quantitative	Synthetic substrates may not reflect natural substrate behavior
Fluorogenic assays (MUF substrates)	Synthetic fluorogenic glycosides	MUF-β-D-glucopyranoside	Low-level activity detection; high-throughput screening	Very sensitive; real-time monitoring possible	Expensive substrates; synthetic substrate bias

Measuring the GH activity on natural substrates

Most simple carbohydrates such as glucose and galactose have a free aldehyde or ketone group that can be oxidized in a redox reaction. Several chromogenic reagents (oxidants) can be reduced by these aldehyde or ketone groups providing an effective way to measure the quantity of free reducing ends in solution. For the characterization of glycosidase activities, as the reaction proceeds, the substrate is cleaved and generates a new reducing end. Tracking the concentration of these reducing ends over time allows for the estimation of kinetic parameters. However, these techniques destroy the sample necessitating large quantities of enzymes and substrates.

The 3,5-dinitrosalicylic (DNS) acid assay is a popular technique used to quantify the reducing ends generated during hydrolysis of polysaccharides and oligosaccharides. It was first introduced in 1921 to detect sugars in urine [50]. Under alkaline conditions, the sugar’s free aldehyde and ketone groups reduce the 3,5-dinitrosalicylic acid (yellow) and produce 3-amino-5-nitrosalicylic acid (red). Performing the DNS assay requires a standard curve prepared with known concentrations of a reducing sugar (e.g., cellobiose or glucose) to calculate the sugar concentration in unknown samples. For the characterization of GH activities, reaction mixtures are incubated in various conditions (e.g., pH, temperature) in the presence of known substrate concentrations for different amounts of time, and the enzymatic reaction is stopped by adding the DNS reagent. The alkaline pH of the reagent will generally stop the reaction as most GHs are active in acidic or neutral conditions. Next, the mixture is heated (typically to 90–100 °C for 5–10 min) to facilitate the chemical reaction and then cooled on ice to halt it. Finally, the intensity of the color is measured spectrophotometrically at 540 nm. The color intensity correlates with the concentration of reducing ends which can be deduced from the standard curve. To ensure consistency, all assays (standard curve and unknown samples) must be processed using identical conditions (e.g., reagent mixing, heating and cooling temperature, and time). The DNS assay is easy to perform with minimal equipment making it suitable for routine lab use, it works with many substrates, is inexpensive, and the DNS reagent is readily available from many vendors (e.g., SigmaAldrich #D0550).

The Somogyi–Nelson test also quantifies reducing sugars. This method involves a two-step process. First, the sample containing reducing sugars reacts with an alkaline copper reagent (typically copper sulfate), reducing Cu²⁺ to Cu⁺, which forms a colorful precipitate of cuprous oxide (Cu₂O). Next, the cuprous oxide is dissolved in an acidic solution, and the remaining copper ions are measured spectrophotometrically at 560 nm. The color intensity is directly proportional to the concentration of reducing sugars [51].

Both the DNS and the Somogyi–Nelson assays are quite sensitive and can detect low levels of reducing sugars; however, the DNS assay is more commonly used in laboratories because of its (i) ease of implementation, (ii) speed and versatility given a single reagent, and (iii) resistance to interference by substances that affect Cu reduction. Although both DNS and the Somogyi–Nelson assays can be used to investigate the enzymatic deconstruction of polysaccharides and oligosaccharides, they can only quantify the release of reducing ends and fail to identify the actual products or the mode of action of the enzymes (e.g., endo- vs. exo-cellulase).

Generally, the distinction between endo- and exo-acting enzymes is achieved by chromatographic techniques (see below) or viscometry. As soluble polysaccharides (e.g., starch or CMC) increase the viscosity of solutions, their enzymatic breakdown lowers the molecular weights of the oligosaccharides thus reducing the viscosity. As endo-acting enzymes (e.g., endo-cellulases—EC.3.2.1.4) attack random points along the polysaccharide chains, they release intermediates of variable DP which leads to a rapid drop in viscosity. On the contrary, exo-acting enzymes (e.g., cellobiohydrolases—EC.3.2.1.91) release short oligosaccharides from the substrate’s ends, with no intermediate deconstruction products, and display a slower reduction in viscosity. Although several methods have been developed to link the drop of viscosity to enzyme kinetics [e.g., [52–54]), this is essentially used as a qualitative method to investigate the mode of action of enzyme preparations [32]. Unlike viscometry, the precise identification of intermediates and final reaction products through chromatographic techniques can also provide key information about enzyme functioning.

Identification of deconstruction products

Identification of deconstruction products can be achieved by analytical techniques where soluble sugar products (i.e., mono- and oligosaccharides) are released through enzymatic or acid hydrolysis, or other chemical treatments for characterization. The precise identification of these products is instrumental to understand enzymes at the molecular level. For example, when considering GHs targeting specific linkages (e.g., GH16 β‑1,3/1,4‑glucanase) in complex substituted and branched substrates, the precise identification of the produced oligomers reveals the position where the enzyme cuts in the polysaccharide (e.g., [55]). Next, when applied to samples collected at various time intervals, these techniques can even be used to infer the enzyme kinetic parameters (e.g., [56, 57]). Also, as GHs “encode” their selectivity in their subsite layout [58]; the produced oligosaccharides are the observable readout of that selectivity and thus provide information about the nature of the subsites along the catalytic cleft [59, 60]. Finally, when combined with well-characterized enzymes, identifying the deconstruction products can reveal the substrate structure (i.e., branching, substitution, motif patterns) (e.g., [61, 62]).

Thin layer chromatography (TLC) [63, 64], HPLC-PAD (high-performance liquid chromatography with pulse amperometric detection) [65], and HPLC-RI (refractive index) are routinely used to identify GH hydrolysis products. TLC is a widely used, cost-effective, method for analyzing carbohydrates and their enzymatic breakdown. It is particularly useful for characterizing the composition of monosaccharides, oligosaccharides, and polysaccharides, in addition to monitoring enzymatic deconstruction of complex carbohydrates. TLC separates sugars based on their differential affinity for the stationary phase (i.e., the TLC plate—often silica gel on a glass or plastic sheet) and the mobile phase (the solvent mixture). As the silica gel has a polar surface, sugars carried by the mobile phase are separated based on their polarity and size [63]. After the sugars separate, the TLC plates are dried, and the sugars are visualized using aniline hydrogen phthalate [66], diphenylamine (DPA) [67], silver nitrate [68] or orcinol/sulfuric acid [69, 70]. When investigating enzymatic deconstruction of polysaccharides, the appearance of new spots or changes in the intensity of spots can be indicative of sugar hydrolysis or cleavage. Distinct spots can be identified using individual known standards. Applied to sugar characterization, TLC is a qualitative approach that can help identify the mode of action of enzymes. Indeed, endo-acting enzymes (e.g., endo-cellulase, EC.3.2.1.4) generate a mixture of intermediate products of various sizes whereas exo-acting enzymes (e.g., cellobiohydrolase, EC.3.2.1.91) release only short oligosaccharides (e.g., cellobiose) with no intermediates.

As TLC requires no major equipment, it is accessible to most laboratories. However, at best, TLC is semi-quantitative as there is no easy way to convert the spot’s intensity or size to actual sugar concentrations. In addition, some products tend to co-migrate (e.g., hexose monosaccharides) [71, 72] and thus require repeated runs on TLC or, potentially tedious, tuning of mobile and fixed phases. Finally, TLC requires a high concentration, limiting its use in some fields (e.g., medical glycobiology) where working concentrations are in the nano- or micro-molar range.

In HPLC-PAD analysis, after large polysaccharides and debris are removed (clarification process), short sugars are separated based on their physical and chemical properties, such as size, charge, or hydrophobicity [73]. Reversed-phase columns (e.g., C18) [74] are typically used for neutral sugars whereas anion-exchange columns for HPAEC (e.g., Carbopac PA1 or Carbopac PA-20) [75] are often used for acidic sugars, sugars with charged groups (including oligosaccharides with hydroxyl and amino groups), or negatively charged sugar derivatives, such as carboxylic acids in glucuronic or galacturonic acids. Sugars are then detected and quantified in the elution using a pulsed-amperometric detector (PAD). The unique retention/elution times for different sugars allow for their identification through comparison with known standards (e.g., glucose, cellobiose, cellotriose). The combination of the retention times and the characteristic peak shapes and area helps distinguish similar sugars, such as glucose versus fructose [76, 77]. Unlike TLC, HPLC-PAD for sugar characterization and investigation of enzymatic reactions requires specialized equipment and trained personnel to perform and interpret the results. Although the setup and maintenance costs associated with HPLC/HPAEC-PAD can limit accessibility for some laboratories, these approaches provide highly sensitive and quantitative characterization of enzymatic hydrolysis products. In addition, HPLC/HPAEC-PAD can quantify products with high sensitivity, detecting even nanomolar concentrations [78].

High-performance liquid chromatography (HPLC) using refractive index (RI) detection coupled with classic ion-moderated partition chromatography is also a well-established method for sugar analysis in GH product deconstruction and characterization. Similarly to HPLC-PAD, in this approach, carbohydrates are separated on a sulfonated polystyrene–divinylbenzene resin in the presence of a dilute acidic mobile phase (commonly water with a small concentration of sulfuric acid), where separation is governed by a combination of size exclusion, ligand exchange, and ion-moderated partition mechanisms. Neutral mono- and oligosaccharides produced during GH enzymatic hydrolysis are resolved based on differences in their interactions with the resin and their degree of polymerization. RI detection, while less sensitive than MS-based methods, measures changes in the refractive index of the mobile phase as analytes elute from the column. Since sugars alter the refractive index relative to the pure mobile phase, they can be detected without prior derivatization. This makes RI detection particularly suitable for routine analysis of mono-, di-, and oligosaccharides in food, fermentation broths, and enzymatic hydrolysis samples for polysaccharide deconstruction [79].

Although RI detection is considered less sensitive than mass spectrometric or evaporative light scattering detectors, it offers several advantages: it is non-destructive, simple to operate, and provides a nearly universal response for compounds present at relatively high concentrations. However, it requires isocratic elution (because it is sensitive to changes in mobile phase composition) and careful temperature control to maintain baseline stability. Despite these limitations, RI detection remains a standard and cost-effective choice for quantitative sugar analysis in many laboratory and industrial settings [80].

Finally, other analytical techniques have been developed to understand the molecular mass and linkage composition of the products generated during the deconstruction of polysaccharides. Specifically, liquid chromatography coupled with mass spectroscopy (LC/MS) has been used to analyze unidentified oligosaccharides where no known standard is available (e.g., [69]). Briefly, MS techniques like MALDI-TOF and ESI–MS can determine the molecular weight and composition of oligosaccharide products and infer the monosaccharide constituents and sequence through a fragmentation pattern. However, one limitation of MS-based approaches is that structurally related oligosaccharides and regioisomers can display identical or highly similar masses, complicating their discrimination using mass information alone. Nevertheless, the combination of chromatographic separation, tandem MS fragmentation, derivatization strategies, and complementary analytical approaches can substantially improve regioisomer discrimination and structural characterization of complex carbohydrates (e.g., flavonoids and galactoside isomers) [81–83]. In GC/MS-based methods, polysaccharides are typically subjected to permethylation, followed by hydrolysis, reduction, and acetylation to generate partially methylated alditol acetates. These derivatives are sufficiently volatile and thermally stable for separation by gas chromatography. The resulting mass spectra provide characteristic fragmentation patterns that reveal the positions of glycosidic linkages, since hydroxyl groups involved in linkages remain unmethylated prior to hydrolysis and are therefore distinguishable after derivatization [84]. GC/MS offers high chromatographic resolution and well-established spectral libraries, making it particularly powerful for detailed structural elucidation of neutral sugars, although the derivatization process can be labor-intensive and less suitable for labile or highly charged carbohydrates.

LC/MS, by contrast, often requires less extensive derivatization and is especially advantageous for analyzing polar, non-volatile, or thermally unstable sugars. Using techniques such as hydrophilic interaction liquid chromatography (HILIC) or reverse-phase chromatography with appropriate labeling reagents, LC/MS can separate native or derivatized oligosaccharides prior to mass analysis. Coupled with soft ionization methods like electrospray ionization (ESI), LC/MS enables the detection of intact molecular ions and supports tandem MS (MS/MS) experiments for detailed structural characterization, including linkage and branching information. While LC/MS may not always provide the same level of linkage-specific fragmentation as classical GC/MS methylation analysis without additional strategies, it excels in sensitivity, speed, and compatibility with complex biological samples, making it an increasingly important tool in glycomics and carbohydrate structural analysis [85].

Ultimately, NMR can provide detailed structural information such as identification of isomers and the precise connections between sugars in oligosaccharides. Although the acquisition and analysis of complicated NMR spectra require major equipment and trained scientists, these techniques allow for the precise structural identification of the product, eventually discriminating between α- and β-linkages in the glycosidic bonds [86, 87].

Measuring the GH activity with synthetic substrates

The availability of synthetic substrates to characterize enzymes has enabled detailed investigations of substrate specificity and identification of various modes of action. Two types of synthetic substrates are routinely used for the characterization of GHs. On the one hand, para-nitrophenyl (4-NP/p-NP) derivatives are chromogenic substrates used in alkaline conditions, while para-methylumbelliferone (4-MUF/p-MUF or simply MUF) derivatives are fluorogenic compounds and can be adjusted to work across a broad pH-range, albeit with variable sensitivity.

The first p-NP substrate, p-nitrophenyl acetate, was synthesized by Hugo Erdmann to study ester hydrolysis [88]. Cleavage of the p-NP group releases p-nitrophenol (pKa ~ 7.15) which is modified to yellow p-nitrophenolate (peak absorbance at 405 nm) under alkaline conditions. This property made p-NP-linked substrates popular in enzymology starting in the mid-twentieth century after p-NP-β-D-glucopyranoside was first synthesized for the characterization of β-glucosidase activity [89].

In a typical "p-NP assay", the substrate is incubated with an enzyme under various conditions (e.g., temperature, pH) and the reaction is stopped by adding an alkaline solution (e.g., sodium carbonate), which reveals the yellow color of the released p-NP. Then, the absorbance of the solution is measured at 405 nm to determine the concentration of released p-NP. To improve reproducibility and accurate quantification, standard curves and absorbance measurements should be generated using replicate measurements and under the same experimental conditions (e.g., pH, temperature, buffer composition, and ionic strength) as the enzymatic assays. For enzymes active in alkaline conditions, the A_405nm can be monitored continuously, allowing the initial rate of the reaction to be measured directly, substantially reducing the amount of enzyme required to obtain kinetic parameters.

Overall, the assay is straightforward and highly sensitive, allowing the detection of low activity or analysis using low substrate/enzyme concentrations. A variety of p-NP-linked glycosides have been synthesized to study glycosidase activity and are commercially available (Table 2). Being synthetic, the substrate can be tailored for specific activity based on the sugar moiety. In addition, working with pure synthetic substrates ensures consistency across experiments and minimizes cross-reactivity. However, as p-NP substrates are not natural, results from the hydrolysis of these substrates cannot simply be transposed to the hydrolysis of natural substrates like cellulose or hemicellulose. Finally, synthetic p-NP substrates, like p-NP-oligosaccharides can be expensive, especially for longer/harder to obtain glycosides. Finally, when working with long p-NP-linked substrates, one should consider the breakdown of the oligosaccharide without release of the p-NP group.

Table 2.

Some of the commercially available p-NP and p-MUF substrates used for characterization of GH activities

Substrate compound (CAS#)	Enzyme activity (EC#)	Citations
Arabinose
p-NP-α-L-arabinopyranoside (1223-07-0)	3.2.1.55	[100]
p-MUF-α-L-arabinopyranoside (69414-26-2)	3.2.1.55	[101]
Cellobiose
p-NP-β-D-cellobioside (3482-57-3)	3.2.1.91	[102]
p-MUF-β-D-cellobioside (72626-61-0)	3.2.1.91	[103, 104]
Galactose
p-NP-α-D-galactopyranoside (7493-95-0)	3.2.1.22	[105]
p-MUF-α-D-galactopyranoside (3150-24-1)	3.2.1.22	[106]
p-NP-β-D-galactopyranoside (38597-12-5)	3.2.1.23	[107–109]
p-MUF-β-D-galactopyranoside (6160-78-7)	3.2.1.23	[110–112]
Glucose
p-NP-α-D-glucopyranoside (3767-28-0)	3.2.1.20	[113]
p-MUF-α-D-glucopyranoside (2492-87-7)	3.2.1.20	[114]
p-NP-β-D-glucopyranoside (17833-43-1)	3.2.1.21	[115, 116]
p-MUF-β-D-glucopyranoside (18997-57-4)	3.2.1.21	[117–119]
Mannose
p-NP-α-D-mannopyranoside (10357-27-4)	3.2.1.24	[120]
p-MUF-α-D-mannopyranoside (28541-83-5)	3.2.1.24	[121]
Xylose
p-NP-β-D-xylopyranoside (2001-96-9)	3.2.1.37 or 3.2.1.72	[122]
p-MUF-β-D-xylopyranoside (6734-33-4)	3.2.1.37 or 3.2.1.72	[123]

p-MUF-linked compounds are fluorogenic substrates widely used to characterize glycoside hydrolases, esterases, and proteases among other enzymes. The cleavage of these p-MUF-substrates releases fluorescent methylumbelliferone (MUF) which, upon excitation around 360 nm, emits a photon around 450 nm requiring a fluorometer for detection [90–92] Although MUF emits maximally under alkaline conditions (~ pH 10), it can be used under a broad range of pHs after standardization. Its emission intensity at pH 6 is only ~10% of the emission recorded at pH 10, and the emission is further reduced in more acidic conditions [93, 94]. Standard curves of free MUF in various pHs are used for precise quantification, especially under acidic conditions (e.g., [95, 96]). The use of p-MUF derivatives allows for the detection of very low activities or diluted enzyme preparations. After standardization, these substrates allow continuous, real-time monitoring of enzyme activity without the need to stop the reaction, reducing enzyme/substrate preparation. p-MUF substrates are well-suited for high-throughput enzyme screening in microplate formats and similar to the p-NP assay, many substrates have been synthesized and are commercially available (Table 2). However, like the p-NP-substrates, results from the hydrolysis of p-MUF substrates cannot simply be transposed to the hydrolysis of natural substrates, and the cost of p-MUF compounds also increases with the DP. Finally, p-MUF substrates are generally more expensive than the corresponding p-NP-linked substrates.

Conclusion and future directions

Although CAZy-db reports the activity (i.e., EC number) of biochemically characterized proteins in each family, more information could and should be reported. Knowing which enzymatic activities have been reported within each GH family, together with the substrates historically used to test these activities, as presented here, provides a structured framework for guiding the characterization of newly identified enzymes. The compiled information (Supplementary data 1 and 2) narrows down the range of plausible activities to investigate and helps researchers select the most relevant and informative substrates for initial screening. By aligning substrate choice with the known activity spectrum of a given GH family, researchers can design assays that are both efficient and targeted, reducing unnecessary experimentation while maximizing the likelihood of detecting true enzymatic function. Ultimately, this knowledge-driven approach could facilitate the functional annotation of new GHs, especially for scientists new to the field, and has the potential to improve the consistency and comparability across studies.

While this work proposes a structured framework for GH characterization, several limitations should be considered. First, the compiled dataset relies on publicly available entries from CAZy and BRENDA, which are subject to reporting biases and uneven characterization across GH families. As a result, the distribution of activities and substrates likely reflects historical research focus rather than the full functional diversity of GHs. Next, our analysis compiles reported activities and substrates without re-evaluating primary kinetic data or weighting studies by experimental quality, which may introduce variability in the observed trends. Third, the proposed substrate panels are derived from reported usage and practical considerations (e.g., availability, assay compatibility) and should therefore be interpreted as practical starting points rather than exhaustive solutions. Finally, this framework has not been experimentally validated and does not resolve functional ambiguity within highly diverse GH families or subfamilies.

Despite these limitations, as the repertoire of GH families and enzymes continues to expand, so does the number of reactions making the linking of sequence to function more complex. The current landscape of enzyme characterization is fragmented: different GH families are assayed on heterogeneous substrates, often under variable conditions, using assays that differ in sensitivity, specificity, and reporting formats. While these approaches have been sufficient to establish broad activity profiles as displayed on CAZy-db [4], they fall short of enabling consistent, cross-comparable datasets. Eventually, subfamily classification is an established approach that increases functional resolution within several large CAZy families (e.g., GH5) divided into subfamilies displaying improved functional coherence [4, 21, 97]. However, CAZy does not currently provide publicly released computational definitions (e.g., HMM-profiles or explicit sequence boundaries) for systematic subfamily assignment. As a result, reproducible subfamily-level annotation requires independent reconstruction of clustering criteria and may not be uniformly applied across studies. Furthermore, although some subfamilies are functionally coherent, others remain sparsely characterized or still encompass multiple reported EC activities, meaning that experimental ambiguity persists even at the subfamily level. This persistent heterogeneity obscures our ability to detect subtle trends in enzyme specificity and hinders the predictive power of computational models. Our work is therefore complementary: by quantifying EC-activity distributions across families and compiling practical substrate panels and detection strategies, we provide an experimentally actionable framework that supports enzyme characterization regardless of whether subfamily information is available or unambiguously interpretable.

Moving forward, the field of enzymology would benefit from a global effort to establish standardized protocols for GH characterization. Such standards should define (i) recommended substrate panels for each GH family, balancing natural, synthetic, and chromogenic compounds if possible; (ii) clear and uniform units for substrate concentration and kinetic parameter reporting; and (iii) minimum reporting requirements—including negative results—for assay conditions, enzyme preparation, and detection methods as implemented in sister disciplines [98]. The adoption of such a framework would facilitate reproducibility and ensure that data generated across laboratories are interoperable.

Compiling functional data from the literature remains a daunting task. Extracting kinetic parameters and substrate specificities from publications often requires specialized training to navigate diverse assay formats, access to paywalled articles, and significant effort to “homogenize” the information into a comparable format. Variability in how results are reported (i.e., incomplete metadata, inconsistent units) further complicates this process and limits large-scale integration. In this context, the STRENDA initiative provides relevant guidelines for reporting and accessing enzymatic activity [99]. Recognizing these challenges, the CAZy team has also initiated efforts to streamline data collection. In particular, they have developed a dedicated platform for reporting functional information, including kinetic parameters, of newly characterized GHs and other CAZymes (https://www.cazy.org/Functional_data.html). In addition, the CAZac initiative represents another advance in CAZy functional annotation and provides a multicriterion activity descriptor that captures mechanistic features (e.g., catalytic mechanism, glycosidic bond orientation, subsites) for individual activities [19]. CAZac is primarily designed to codify mechanistic and structural descriptors that improve how individual activities are represented and compared within CAZy. By contrast, the present study provides a quantitative overview of how reported EC activities are distributed across GH families (i.e., the fraction and types of EC numbers observed per family) and compiles a practical substrate panel and assay recommendations intended to guide initial experimental screening. While CAZac and the CAZy functional portal provide mechanistic and curated functional entries; our dataset provides family-level distributional values and an experimentally oriented substrate resource to support empirical enzyme characterization. These initiatives represent critical steps toward building community-driven resources that will enhance data accessibility, promote consistency in reporting, and ultimately accelerate the use of functional data for predictive and computational modeling, assuming the data are made publicly accessible.

Importantly, functional data should not be reduced to the EC number alone. While the CAZy functional data page currently reports EC assignments, the true value of biochemical characterization lies in the experimental context: the substrates tested (both positive and negative reporting), the measured kinetic parameters, and the assay conditions such as optimum temperature, pH, and cofactors. Sharing this richer layer of information, in a manner similar to the comprehensive entries available in the BRENDA database (https://www.brenda-enzymes.org) [20] and STRENDA [99], would transform scattered GH reports into interoperable datasets, greatly enhance reproducibility, enable meaningful cross-comparisons, and provide the high-resolution functional insights required to develop robust predictive models of enzyme activity.

To fully realize the potential of standardized datasets, it is essential that the entire research community actively contributes to these efforts. Beyond reporting functional data from newly characterized enzymes, there is immense value in curating and submitting information from past publications, which often contain rich but underutilized datasets scattered across decades of research. By contributing both historical and contemporary data to platforms such as the CAZy functional data portal, BRENDA, and STRENDA, researchers can help build a more complete, accessible, and harmonized record of GH activities. This collective effort should not only reduce redundancy in future characterizations but also empower the development of robust AI and machine learning models, enabling predictive insights that extend beyond the limits of individual studies. Standardized, community-driven data submission must therefore become a shared priority, ensuring that the field of enzymology moves from fragmented reports to a comprehensive, predictive science [99].

For biotechnology, harmonized and FAIR-compliant functional datasets would substantially accelerate enzyme discovery and engineering pipelines. Standardized reporting of substrates, kinetic parameters, assay conditions, as well as negative results would enable direct cross-comparison of candidate enzymes for targeted applications such as lignocellulose saccharification and food processing. Interoperable datasets would facilitate meta-analyses, automated data mining, and machine learning approaches aimed at identifying sequence–function relationships or engineering new variants. In particular, structured and accessible functional outputs would allow researchers to efficiently identify enzymes with desired properties (e.g., thermostability, substrate range, catalytic efficiency on defined substrates) without re-characterizing previously studied systems. By aligning experimental reporting practices with FAIR data principles, the field can transition from fragmented enzyme descriptions to reusable functional knowledge, thereby enhancing translational impact in industrial biotechnology and bio-based production systems.

While this work does not impose a universal standard, it delineates a structured and reproducible pathway toward harmonized GH characterization. The datasets and substrate recommendations presented here provide an operational starting point upon which community consensus, FAIR-compliant reporting, and interoperable databases can be built. Ultimately, the harmonization of enzyme characterization represents more than a methodological refinement. It is a paradigm shift. By bridging the gap between biochemical experimentation and computational modeling, standardized enzyme characterization may help support ongoing efforts toward more predictive enzymology.

Author contributions

Conceptualization (D.S.E., R.B.), data collection and processing (D.S.E., N.G.G., A.E.H., and R.B.), manuscript writing (D.S.E., N.G.G., A.E.H., J.M.-M., and R.B.).

Funding

The material generated and analyzed in this study is based on work supported by the National Science Foundation under Award Number #2138880 (RB).

Data availability

No datasets were generated or analyzed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Consent to publish has been given by all authors (all displays are the original work of the authors).

Competing interests

The authors declare no competing interests.