Precision hydrolysis: tailored yeast processing enzymes for yeast-based products

Abstract

Abstract

Yeasts and yeast-based products are nutrient-rich bioresources with broad applications in technologies for the production of food, feed, medicine, and cosmetics. However, traditional processing often results in non-specific lysis and suboptimal product quality. Yeast extract can be used as a flavor enhancer, nutritional supplement, or fermentation substrate, and the other components of the yeast cell wall and nucleic acids can be processed into bioactive materials, including glucans and nucleotides. These materials offer both nutritional and therapeutic benefits. Precision hydrolysis, leveraging the high specificity of tailored enzymes, has emerged as a superior strategy for maximizing the yield and functional quality of high-value yeast-based products. It provides superior outcomes by improving the quality of yeast-based products. Tailored enzymatic strategies, leveraging mechanistically focused core enzymes, including proteases, β-glucanases, and coupled nucleases-deaminases, have demonstrated superior efficiency, nutritional enhancement, and sensory refinement. This review focuses on the mechanistic properties of yeast processing enzymes, emphasizing their functional classification and applications in precision hydrolysis. It details how such enzymes are optimized for the targeted release and modification of high-value components. Additionally, the review highlights recent strategies for tailored biosynthesis of yeast processing enzymes, including enzyme discovery, heterologous expression systems, and machine-learning-guided optimization. This review aims to support future innovations that will promote the development of sustainable, high-value, and diversified yeast-based bioproducts by optimizing the biosynthesis of processing enzymes, thus lowering the overall cost of precision hydrolysis.

Key points

• Precision hydrolysis enables the controlled release of yeast components in a specific pattern, yielding high-quality, specific yeast-based products.

• By leveraging the highly specific effects of enzymes, targeted product refinement and superior characteristics under mild processing conditions can be achieved.

• To avoid the high cost of precision hydrolysis, continuous advances in enzyme discovery, protein engineering, and metabolic engineering technologies are vital.

Introduction

Yeasts and yeast-based products, such as yeast extract and yeast cell walls, are increasingly recognized as versatile bioresources. They have applications in in food, animal feed, healthcare, cosmetics, and biotechnology (Fig. 1). Yeasts contain high-quality proteins and peptides, vitamins in the B-complex group (B₁, B_2, B₃, B₆, and folate), minerals such as selenium, zinc, and iron, microbial oils from oleaginous species, and cell wall polysaccharides like β-glucans and mannans. These components provide both nutritional and therapeutic benefits (Naik et al. 2024; Patterson et al. 2023). For example, the yeast components enhance nutrition (Martin and Chan 2024) and flavor (Demirgül et al. 2022) in food while reducing salt (Wu et al. 2024a). In animal feed, they promote growth and disease resistance (Bilal et al. 2023; Fan et al. 2025). Within healthcare and medicine, they provide trace elements (Sun et al. 2022b), and exhibit immune-regulating, antibacterial, antiviral, anti-carcinogenic, antioxidant, anti-inflammatory, and immunomodulatory properties (Abid et al. 2022). Yeast components are also applied in cosmetics for skin renewal and wound healing (Sousa et al. 2023), and are widely used in biotechnology as a cultivation medium (Tao et al. 2023).

Fig. 1.

Applications of yeast-based products in various sectors. This illustration highlights the diverse industrial uses of processed yeast products. In the food sector, yeast-based products serve as nutritional supplements, flavor enhancers, and salt-reducing agents, with representative food items depicted. In Feed, they promote growth, enhance disease resistance, and substitute protein sources, illustrated by feed-related materials. In medicine, yeast-based compounds exhibit immune-modulating, antibacterial, antiviral, anticancer, antioxidant, and anti-inflammatory activities, shown alongside pharmaceutical preparations. The cosmetics field employs yeast extracts for skin regeneration, antioxidation, and wound healing, demonstrated by skincare products. In biotechnology, yeast contributes to fermentation processes and environmental remediation, with relevant industrial reagents displayed

Yeast processing involves several steps: fermentation, cell disruption, centrifugation, evaporation, and concentration. Together, these steps yield yeast extract (Fig. 2). Yeast extract is the crude product obtained from the supernatant after yeast cell disruption, which is then concentrated into liquid, paste, or powder forms. After centrifugation, both the pellet and the supernatant are processed to recover bioactive and flavor-enhancing compounds (Marín-Sánchez et al. 2024; Moyo et al. 2025). Cell disruption and extraction rely on both mechanical and non-mechanical methods. However, enzymatic processing is often preferred because it enables precise lysis and targeted release of valuable components (Gautério et al. 2023). This selectivity reduces the challenges usually associated with downstream purification in traditional approaches.

Fig. 2.

Schematic diagram of the comprehensive yeast processing. Following fermentation, yeast cells undergo disruption via physical, chemical, or enzymatic methods (e.g., papain) and autolysis. Centrifugation separates the lysate into two fractions: the supernatant (cell extract) and the pellet (cell wall). The supernatant is processed into yeast extract or enzymatically treated with nucleases, deaminases, and flavor enzymes to yield guanosine monophosphate (GMP), inosine monophosphate (IMP), and flavor peptides, followed by concentration into liquid, paste, or powder forms. The pellet is treated with glucanases to isolate β-glucan

The primary enzymes employed in yeast processing include proteases, β-glucanases, nucleases, and deaminases. Proteases conduct the disruption of yeasts by weakening the cell wall by disrupting the mannoproteins and breaking cell membrane by hyrolyzing membrane proteins (Moyo et al. 2025). β-Glucanases degrade cell wall glucans, increasing the recovery of intracellular proteins, polysaccharides, and vitamins, while producing functional β-glucan fragments (Gautério et al. 2023). Nucleases break down nucleic acids into flavor-active 5′-nucleotides such as inosine monophosphate (IMP) and guanosine monophosphate (GMP), which interact synergistically with glutamate to enhance umami taste. Flavor enzymes are a specific complex of proteases that can effectively convert proteins or short peptides into substances with stronger umami taste, thereby improving the flavor acceptability of products (Merz et al. 2015). However, the widespread application of these enzymes still faces challenges: limited natural sources, rising production costs, and stringent industrial production standards.

This review presents detailed studies on the classifications, mechanisms, sources, applications, and biosynthetic engineering of enzymes critical for the production of yeast-based products. We aim to evaluate enzymatic applications and inform future advancements in high-performance enzyme production.

Precise hydrolysis and application of enzymes in yeast processing

Precision hydrolysis surpasses traditional autolysis by employing tailored enzymes to achieve specific, high-value outcomes. Proteases play a central role in yeast processing, whereas nucleases and deaminases are primarily refine β-glucans and generate flavor-active compounds (Fig. 2).

Proteases

Proteases catalyze the hydrolysis of peptide bonds in proteins, breaking them down into shorter peptides or, ultimately, individual amino acids. Proteases aid in yeast processing by serving two primary roles. First, proteases weakened the yeast cell wall by hydrolyzing the mannoproteins and broke the membrane protein (Moyo et al. 2025). Second, proteases performed intracellular precision hydrolysis, selectively degraded high-molecular-weight proteins into specific smaller fragments via site-specific cleavage to modulate flavor profiles (Tao et al. 2023). Specifically, flavor enzymes are a specialized complex of endo- and exopeptidases used in precision hydrolysis to selectively cleave peptides, reducing bitterness and enhancing the umami flavor of yeast products. Proteases are predominantly sourced from animals (such as calf stomach), plants (including pineapple and papaya), and microorganisms (such as Bacillus and Pseudomonas) (Table 1).

Table 1.

Proteases in yeast processing

Sources	Characteristics	Protease	Acquisition	Optimal pH	References
Plant-derived protease	Broad-spectrum hydrolytic activity	Papain	Carica papaya (fruit, root, and leaves)	5.5	(Babalola et al. 2023)
		Bromelain	Ananas comosus (stem and juices)	7	(Kumar et al. 2023)
		Ficin	Ficus carica	7	(Siar et al. 2024)
		Actinidin	Actinidia deliciosa (Kiwi fruit)	7.5	(Dhiman et al. 2021)
Animal-derived protease	High specificity	pancreatin	Animal pancreas	8.0	(Bayarjargal et al. 2014)
		Trypsin	Animal pancreas	8.5	(Takalloo et al. 2022)
Microbial-derived protease	Low-cost	Neutral protease	Submerged fermentation of Bacillus spp.	6.5–7.5	(Dumitrașcu et al. 2023; Zhu et al. 2025)
		Alkaline protease	Submerged fermentation of Bacillus spp. and Aspergillus spp.	7.5–9.0	(Pawar et al. 2023; Zhang et al. 2023)
		Acid protease	Aspergillus spp. and other fungal	2–6	(Song et al. 2023)

Papain

Due to its strong proteolytic activity, papain is widely used in various industrial processes, including meat tenderization, dairy processing, brewing, pharmaceuticals, cosmetics, and textiles (Choudhary et al. 2025). Papain’s application in yeast processing is often customized through two distinct approaches: assisting autolysis or acting as the primary lytic agent (Fig. 3). The choice between these methods depends on the desired balance of yield, flavor, processing speed, and cost (Moyo et al. 2025).

Fig. 3.

Comparative schematic of two papain-mediated yeast disruption methods. Papain, derived from Carica papaya, is shown as the catalytic agent in both processes. The left panel illustrates papain-assisted autolysis, conducted at 50–55 °C, pH 5.0–5.5, with 0.1–0.5% papain over 4–6 h. The right panel depicts enzymatic hydrolysis, performed at 50–60 °C, pH 5.5–6.5, using 0.5–2.0% papain for 1–2 h. Both protocols include enzyme inactivation and centrifugation steps

Papain (EC 3.4.22.2), a member of the cysteine protease family, is derived from Carica papaya. It is one of the most extensively studied and widely utilized proteases (Benito-Vázquez et al. 2024), known for its ability to release amino acids during protein hydrolysis and enhance the umami profile of yeast hydrolysates (Sirisena et al. 2024). The enzyme has a molecular mass of approximately 23.4 kDa and consists of a single polypeptide chain of 212 amino acids stabilized by four disulfide bridges (Choudhary et al. 2025). Its catalytic function relies on three key residues, Cys25, His159, and Asn175, and it preferentially cleaves peptide bonds adjacent to phenylalanine, lysine, and arginine. This selectivity is the cornerstone of precision hydrolysis, enabling the targeted release of umami-rich peptides and amino acids, thereby tailoring the yeast hydrolysate’s flavor profile.

Microbial synthesis offers a promising route to lower the production costs of papain. The papain gene has been successfully isolated from the papaya plant using reverse transcription (Makeri et al. 2024). Notably, codon-optimized recombinant papain has been effectively expressed in Pichia pastoris, achieving a purification titer of 463 mg/L (Werner et al. 2015). Besides, in one study, recombinant papain was produced at 350 mg/L in Escherichia coli (Hoque et al. 2025). These findings provide a solid foundation for the scaled-up, and commercial production of papain.

Animal- and microbial-derived proteases

Proteases from both animal and microbial sources have seen increasing use in yeast processing. Animal-derived proteases offer high specificity and efficiency, allowing precise protein hydrolysis and the generation of controlled peptide profiles for targeted flavor or functional properties. Microbial proteases, in contrast, provide cost-effective production, greater operational stability, and compatibility with genetic engineering, enabling customized applications and enhanced control over the hydrolysis process across diverse industrial contexts.

Studies have employed pancreatin to treat brewer’s yeast, using 2.5% pancreatin and 2.5% flavor enzymes to facilitate autolysis and enzymatic hydrolysis of spent yeast. The autolysis and hydrolysis processes were evaluated by measuring soluble solids, soluble protein concentration, and α-amino nitrogen content in the reaction mixture (Bayarjargal et al. 2014). Takalloo et al. (2020) conducted a series of studies to expand the application of microbial proteases in this field. In one investigation, the researchers treated baker’s yeast with Alcalase®, a serine alkaline endo-protease produced by Bacillus subtilis. Compared with autolysis and plasma disruption methods, enzymatic hydrolysis with Alcalase® yield higher solids and protein yields. In a subsequent study, trypsin digestion produced greater hydrolysate yields and higher α-amino nitrogen content than autolysis and Flavourzyme treatment (Takalloo et al. 2022).

Recombinant production of alkaline proteases has been successfully achieved using B. licheniformis as the expression host, yielding an enzyme activity of 15,435 U/mL (Zhang et al. 2023). Subsequent optimization of the fermentation medium resulted in a nearly threefold increase in enzymatic activity to 39,233.6 U/mL. In a separate study, the same bacterial strain was used to enhance aprE expression by identifying the most effective combination of plasmid and genomic integration site (Zhou et al. 2020). This genetic modification yielded a 62% increase in activity compared to the unaltered strain.

Various commercial protease preparations have been developed for yeast processing (Table 2). The role of these enzymes in yeast hydrolysis has been evaluated by comparing traditional methods, such as autolysis and mechanical disruption, with enzymatic hydrolysis using commercial proteases, including Brauzyn®, Alcalase™, Protamex™, and Flavourzyme™ (Marson et al. 2019). Hydrolysis with Brauzyn® increased crude protein content by 50%, protein recovery by 83%, and antioxidant activity by 63% relative to the control without enzyme. Sequential hydrolysis using Brauzyn® followed by Alcalase™ produced a hydrolysate with the highest solid content and antioxidant activity. Further studies demonstrated that protease blends, such as Alcalase™, Brauzyn®, and Protamex™, could generate spent brewer’s yeast (SBY) hydrolysates with enhanced physicochemical properties and improved antioxidant activity (Marson et al. 2020). On average, protein hydrolysis yielded 57% solids and 70% crude protein by weight from the raw material. The process achieved an intermediate hydrolysis degree by 15%, which promoted the release of hydrophobic residues 8% higher than in the control and conferred significant antioxidant properties.

Table 2.

Characteristics of main commercial proteases in yeast extract production

Commercial name	Biological source	Characteristic	Dose	Manufacturer	Working pH range	Working temperature range (○C)
Neutrase®	B. amyloliquefaciens	Broad-spectrum microbial endo-protease	0.5–1.5%	Novozyme	5.5–7.5	45–55
Alcalase®	B. licheniformis	Food-grade alkaline protease	0.5–2.0%	Novozyme	6.5–10	60–75
Protamex®	B. licheniformis	Broad-spectrum endo-protease	0.2–0.8 g/kg	Novozyme	6–9	30–65
Promod® 950L	Bacillus spp.	Microbial endopeptidase Preparation with broad substrate specificity	0.50–1.0%	Biocatalysis	5–7.5	50–60
Promod® 144MDP	Carica papaya	Botanical protease with at least five proteases of different specificity	0.4–0.7 g/kg	Biocatalysis	5.0–7.5	50–70
Promod® 184MDP	Ananas comosus	Bromelain product with a long history of use in protein modification	0.5–2.0%	Biocatalysis	5.0–7.0	45–55

Flavor enzymes

Protein hydrolysis often yields bitter peptides with molecular weights of 0.5–2.0 kDa, which can negatively affect product palatability (Liu et al. 2022). For example, the free amino acids leucine (Leu) and phenylalanine (Phe) only have a slight bitter taste (with a taste threshold of 15–20 mM), while the simple dipeptides formed by them (Leu-Leu, Leu-Phe) have a bitterness intensity 10 times that of their corresponding free amino acids (Maehashi and Huang 2009).

To improve flavor, precision hydrolysis utilizing flavor enzymes is applied to modify hydrophobic peptide segments containing amino acids such as leucine and isoleucine. This process reduces bitterness and enhances umami and mouthfeel, thereby optimizing the sensory quality of products like yeast extract. Compound proteases with both endopeptidase and exopeptidase activities are commonly used to achieve this effect by specifically targeting bitter-inducing hydrophobic amino acids, such as leucine and phenylalanine, from peptide termini and promoting the formation of umami amino acids like glutamic acid and taste-active peptides such as glutamyl dipeptides.

Flavor enzymes are a specific complex of proteolytic enzymes, consisting of a mixture of endopeptidases and exopeptidases. They are the core tool for achieving precision hydrolysis in flavor modulation. Commercial flavor enzyme complexes are specialized blends of enzymes designed to precisely cleave flavor-active peptides, enhancing taste, aroma, and mouthfeel in food products. These complexes typically combine multiple enzyme types to target proteins, nucleic acids, and other flavor precursors (Table 3).

Table 3.

Comparative overview of commercial flavor enzyme blends used in yeast extract processing

Supplier	Product name	Key enzymes	Optimum pH	Optimum temperature (○C)	Dose (‰)	Target functionality	Application notes
Novozymes	Flavorzyme	Endo- and exopeptidases from A. oryzae	5.0–7.0	50–55	0.2–3.0	Peptide breakdown, amino acid release, flavor depth	Common in savory hydrolysates, enhances glutamic acid and mouthfeel
Angel yeast	FF104AN	Exopeptidase from A. oryzae and endopeptidase from B. subtilis	6.5–8.5	50–55	1.0–3.0	Mitigating bitterness, enhancing flavor complexity, harmonizing sensory attributes, and elevating amino nitrogen levels	Widely used in yeast extract production, tailored for umami and clean taste profiles
AB enzymes	COROLASE®7089	Neutral endopeptidase from B. subtilis	7.0–8.0	55, up to 65	0.1–5.0	Hydrolyse a broad range of substrates, at neutral pH	Hydrolysis of proteins from several animal and plant sources
	COROLASE®8000	Fungal alkaline protease enzyme preparation	6.0–11.0	65, up to 80	0.5–5.0	Hydrolysing proteins under neutral to alkaline conditions and is used to achieve a high degree of hydrolysis	Hydrolysis of animal proteins

β-Glucanase

The yeast cell wall is primarily composed of β-glucans, including β−1,3-glucan (240 kDa) and β−1,6-glucan (24 kDa), which together account for 29–64% of the wall. Mannan-protein complexes (100–200 kDa) comprise approximately 31%, and chitin (25 kDa) represents 1–2%. Other components include proteins (13%) and lipids (9%) (Baek et al. 2024; Gautério et al. 2023) (Fig. 4A). Efficient extraction of valuable compounds from the yeast cell wall typically requires specific enzymes to degrade these structural components, facilitating the recovery of bioactive products.

Fig. 4.

Structure of yeast cell wall and chain structure of β-glucans. This figure illustrates the structure of the yeast cell wall and the backbone and branch chain structure of β-glucans. A The layered architecture of the yeast cell wall, highlighting the distribution of components including mannoproteins, β−1,6-glucans, β−1,3-glucans, chitin, the cell membrane, and integral proteins. B The chemical structure of β-glucans, with the red arrow indicating the branching site (where a side chain attaches to the main chain) and blue arrows marking the glycosidic linkages within the main chain of glucose residues

β−1,3-Glucanase and β−1,6-glucanase specifically hydrolyze β−1,3 and β−1,6 glycosidic linkages in yeast cell wall glucans (Fig. 4B). Acting sequentially, β−1,3-glucanase cleaves the backbone to disrupt the wall. β−1,6-glucanase removes side chains, enabling complete degradation and maximal release of intracellular components. This sequential action facilitates the selective isolation of high-purity components, including β-glucan, chitin and mannoproteins, from the complex cell wall matrix (Günal‐Köroğlu et al. 2025).

β−1,3-Glucanase

β−1,3-Glucanases are enzymes within the glycoside hydrolase family that break β−1,3-D-glycosidic bonds in glucans and, occasionally, β−1,4 linkages, acting on both linear and branched structures. They function either as endo-enzymes, which randomly cleave internal β−1,3 bonds to produce oligosaccharides. Alternatively, they function as exo-enzymes, which sequentially release sugar units from either the reducing or non-reducing ends (Gupta and Verma 2015). Endo-β−1,3-glucanases typically retain the anomeric configuration of the substrate through a double-displacement mechanism and may also catalyze transglycosylation reactions, forming new glycosidic bonds (Jiang et al. 2024). In contrast, exo-enzymes usually invert the anomeric configuration during hydrolysis, reflecting their distinct catalytic pathways.

β−1,3-Glucanase is primarily derived from fungi, bacteria, plants, insects, and mollusks. Table 4 summarizes representative β−1,3-glucanases from various sources that have demonstrated promising protential for application in yeast processing. Among bacteria, notable producers include Hymenobacter siberiensis (Kim et al. 2024), Alkalihalobacillus clausii (Zhang et al. 2024), and Streptomyces species (Elshami et al. 2025). Fungi serve as important sources of β−1,3-glucanase, with key examples including Aspergillus fumigatus, Trichoderma asperellum, and Phanerochaete chrysosporium (Jin et al. 2023).

Table 4.

Representative β−1,3-glucanases with yeast β-glucan as the substrate

Species	GH Families	Optimal temperatures (°C)/pHs	Substrates	Acting modes	References
Archangium sp. AcGluA	GH55	60/6.0	Yeast β-glucan	β−1,3-glucanase	(Wang et al. 2021a)
B. lehensis Blg32	GH16	70/8.0	Yeast β-glucan	Endo-β−1,3-glucanase	(Jaafar et al. 2020)
Flavobacterium sp. NAU1659	GH16	40/5.0	Yeast β-glucan	Endo-β−1,3-glucanase	(Wang et al. 2024b)
Alkalihalobacillus clausii	GH16	40/5.5	Yeast β-glucan	Endo-β−1,3-glucanase	(Zhang et al. 2024)
Paenibacillus sp. PsLam81A	GH81	60/6.5	Yeast β-glucan	Endo-β−1,3-glucanase	(Plakys et al. 2022)
Magnaporthe oryzae MoGluB	GH64	50/9.0	Yeast β-glucan	Endo-β−1,3-glucanase	(Wang et al. 2021b)
Flavobacterium sp. NAU1659	GH30	50/6.0	Yeast β-glucan	β−1,6-glucanase	(Xie et al. 2024)
Paenibacillus sp. GKG	GH30	50/5.5	Yeast cell wall	Endo-β−1,6-glucanase	(Plakys et al. 2024)

The extraction of β-glucan from the cell wall of yeast involves three steps: cell disruption, extraction of β-glucan from the insoluble cell wall, and chemical modification to convert it into a water-soluble form (Chioru et al. 2024). In contrast to the low purity typically associated with acid–base extraction, enzymatic approaches employ cell wall lytic enzymes or proteases to gently and efficiently disrupt yeast cell walls for β-glucan recovery (Wei et al. 2024). For instance, Yuan et al. (2022) extracted yeast β-glucan from Cyberlindnera jadinii by ultrasound-assisted β-glucanase catalysis, resulting in enhanced solubility and improved antioxidant activity of yeast β-glucan. However, excessive use of β−1,3-glucanase can severely disrupt cell wall integrity, reducing product yield by up to 50% (de Assis et al. 2022). Therefore, β−1,3-glucanase is critical for precision hydrolysis, as it controls the degree of cell wall permeabilization to maximize the yield of both intracellular and structural target products.

β−1,6-Glucanase

β−1,6-Glucanase (EC 3.2.1.75) selectively hydrolyze β−1,6-glycosidic linkages and is classified in glycoside hydrolase (GH) families 5 and 30 based on conserved catalytic residues. β−1,6-Glucanase was first isolated from Mucor hiemalis and identified in various fungi, including Tvir30 from Trichoderma virens (Klemanska et al. 2023), Pus30A from Lentinula edodes (Kobayashi et al. 2023), and β−1,6-glucanases from Aspergillus fumigatus and Epichloë festucae (Volkov et al. 2021). Although β−1,6-endoglucanases were initially considered rare in bacteria, they have been increasingly identified in recent years. The first bacterial β−1,6-endoglucanase was identified in Saccharophagus degradans 2-40T (Wang et al. 2017), followed by GluM from Corallococcus sp. strain EGB (Li et al. 2019), GH30A from Flavobacterium sp. NAU1659 (Xie et al. 2024), PsGly30A from Paenibacillus sp. (Plakys et al. 2024), and CpGlu30A from Chitinophaga pinensis (Lu et al. 2023c).

Functionally, these enzymes contribute to mycoparasitism, autolysis, and fungal cell wall remodeling. Based on β−1,6-glucanase, Qiao et al. (2022a) established a strategy producing mannoproteins MP112 from baker’s yeast with 11.1% yield and 90.7% purity. They also introduced a novel enzymatic method employing β−1,6-glucanase (GluM) to liberate immunologically active β-glucan from baker’s yeast, yielding a 17.8% output with 85.3% purity and a 75.4% recovery rate (Qiao et al. 2022b). The use of β−1,6-glucanase exemplifies precision hydrolysis, leveraging its high specificity to selectively remove β−1,6-linked side chains, thereby significantly enhancing the purity and functional integrity of target components like β-glucan or mannoproteins.

Nucleases and deaminases

Yeast species contain a substantial amount of RNA, typically ranging from 7 to 12% (w/w) on a dry weight basis (Jacob et al. 2019; Zhao and Fleet 2005). Such a high RNA content makes yeast species valuable raw materials for the production of 5’-nucleotides, particularly natural flavor enhancer compounds such as IMP and GMP. When combined with other flavor compounds (glutamic acid or salt), IMP and GMP enhance overall taste and reduce the need for salt in foods (Shan et al. 2025). To produce GMP and IMP from yeast extract, a two-step enzymatic process is employed (Fig. 5). First, the yeast cells are disrupted to release RNA through incubation at 45 °C for 2–3 h. Once hydrolysis is complete, the pH is raised to 7.5 using NaOH, and adenosine monophosphate (AMP) deaminase (AMPD) is introduced to convert 5’-AMP into 5’-IMP at 40 °C over the course of one hour. The final product is a nucleotide-rich solution containing both GMP and IMP, which are widely valued for their flavor-enhancing properties in food applications (Olmedo et al. 1994). This sequential and highly specific enzymatic conversion process represents a critical application of precision hydrolysis in flavor ingredient manufacturing, where the goal is the selective and complete transformation of RNA into specific 5'-nucleotides.

Fig. 5.

Schematic diagram of the enzymatic production of flavor nucleotides from yeast RNA. The process involves yeast cell disruption followed by a two-step enzymatic hydrolysis: first, RNase P1 hydrolysis to generate a mixture of mononucleotides, including cytidine monophosphate (CMP), uridine monophosphate (UMP), adenosine monophosphate (AMP), and guanosine monophosphate (GMP); and second, the specific conversion of adenosine monophosphate (AMP) into inosine monophosphate (IMP) using AMP deaminase. Reaction conditions are indicated for each step

Nuclease P1

Nuclease P1 (EC 3.1.30.1) is a well-characterized nuclease recognized for its specific 5’-phosphodiesterase activity. It selectively cleaves phosphodiester bonds at the 5’ position in single-stranded RNA and DNA, generating 5’-mononucleotides and shorter phosphooligonucleotides, while exhibiting no activity toward double-stranded DNA.

In food production, nuclease P1 acts on single-stranded nucleic acids in the substrate, breaking them down into 5’-nucleotides such as GMP, AMP, uridine monophosphate (UMP), and cytidine monophosphate (CMP). Traditional ingredients like dried bonito and shiitake mushrooms, which are naturally rich in IMP and GMP, respectively, are commonly used in Asian cuisine to enhance flavor. Yeast extracts that have undergone partial hydrolysis also provide an additional source of 5’-nucleotides for flavor enhancement (Magistà et al. 2017). The precise 5’-phosphodiesterase activity of nuclease P1 is the cornerstone of nucleotide precision hydrolysis, ensuring the quantitative conversion of high-molecular-weight RNA into usable flavor precursors (5’-mononucleotides) without the formation of undesirable 3’-nucleotides.

Nuclease P1 is sourced from specific, non-GMO fungi, primarily Penicillium citrinum (Zorn et al. 2024) and Leptographium procerum (Zubaer et al. 2025). Reflecting its established safety and utility in food processing, P. citrinum is officially recognized as a permitted source of ribonuclease in several countries, including China, France, and Japan (Okado et al. 2016).

The escalating global demand for nucleotides, particularly as flavor enhancers in the food industry, underscores the increasing importance of nuclease P1. Consequently, researchers have concentrated on strategies to boost production efficiency. For example, immobilizing nuclease P1 onto suitable carriers enhances its thermal and operational stability, enabling repeated use (Yin et al. 2025). Running P. citrinum in a semi-continuous fermentation setup has demonstrated improved average enzyme yields (Chen et al. 2025). Furthermore, aqueous two-phase extraction systems offer a promising route to simultaneously improve the recovery and purification efficiency of nuclease P1 from fermentation cultures (Chen et al. 2021). These technical advancements are essential for developing a more sustainable and cost-effective method for producing nucleotide-based flavor ingredients.

Adenylate deaminase

AMPD is the second enzyme in this pathway of precision hydrolysis, which selectively converts the non-flavor nucleotide (AMP) into the highly desirable flavor enhancer (IMP). AMP deaminase (AMPD, EC 3.5.4.6) is an essential enzyme involved in purine metabolism, catalyzing the irreversible deamination of AMP to IMP. AMP deaminase was first identified and characterized in Helix pomatia in 1983 and has since been found in other mollusks. Industrial production of AMPD predominantly relies on solid-state fermentation using species such as Aspergillus oryzae, Mucor circinelloides (Li et al. 2024), and Saccharomyces cerevisiae (Byun et al. 2024).

Recent research has increasingly focused on the heterologous expression of eukaryotic AMPD to support industrial applications. He et al. (2022) successfully overexpressed AMPD to improve the flavor profile of Neopyropia yezoensis. Other notable advancements include the expression of Streptomyces murinus AMPD in B. subtilis (Guo et al. 2015) and the expression of Aspergillus oryzae AMPD in E. coli (Li et al. 2016). Ke et al. (2020) further contributed by expressing the AMPD gene from Aspergillus oryzae GX-08 in E. coli, P. pastoris, and B. subtilis, with the latter demonstrating superior efficiency as a host system.

Tailored biosynthesis of yeast processing enzymes

With the yeast processing industry shifting toward high-value and diversified products, application scenarios for yeast processing enzymes are expanding, while biosynthetic technologies are advancing toward greater efficiency and precision. However, challenges remain, including insufficient enzyme activity, high production costs, and limited applicability, necessitating technological innovation.

Mining and heterologous production of novel enzymes

Extremophiles, microorganisms adapted to thrive in extreme environments, are promising candidates for industrial biotechnology. They produce enzymes, known as extremozymes, that exhibit exceptional resilience, including halotolerance, alkaline-stability, surfactant- and solvent-resistance, and robust activity at elevated temperatures (Ashaolu et al. 2024; Gallo and Aulitto 2024). Among the numerous enzymes discovered in extremophiles, proteases are one of the most common and industrially relevant classes. For instance, thermophilic bacteria yield various heat-stable proteases, which are crucial for initial yeast processing steps (Finore et al. 2023). Furthermore, they also produce lipases, xylanases, amylases, chitinases, pectinases, pullulanase, esterases, laccases, dehydrogenases, and isomerases, giving these extremozymes immense potential for a wide range of biotechnological and industrial applications. For example, a recent study on the thermophilic Anoxybacillus caldiproteolyticus 1A02591 strain revealed the secretion of various proteases (primarily metalloproteases and serine proteases) when cultured on casein. These enzymes exhibited an optimal activity at 70 °C and demonstrated the ability to degrade different proteins (Cheng et al. 2021), confirming the potential of thermophiles to provide highly efficient biocatalysts for high-temperature industrial processes.

Heterologous expression is a central strategy for the efficient biosynthesis of yeast processing enzymes, driven by the need for high-yield and functionally active biocatalysts. The three primary platforms for gene overexpression, E. coli, B. subtilis, and P. pastoris, have successfully expressed proteases from diverse sources (animals, plants, microbes, viruses) for industrial use (Song et al. 2023; Table 5). E. coli offers rapid growth and easy genetic manipulation, but limitations include the lack of post-translational modifications and the frequent formation of inclusion bodies (Pouresmaeil and Azizi-Dargahlou 2023). B. subtilis is the main prokaryotic host for proteases (Liu et al. 2024b), simplifying purification through protein secretion and being endotoxin-free. However, endogenous proteases and lower yields can be limiting. P. pastoris enables eukaryotic modifications and high-density fermentation, making it suitable for complex enzymes, despite non-mammalian glycosylation and methanol-related handling concerns (Hong et al. 2025). This platform is also showing promise in C1-based biomanufacturing, as demonstrated by the efficient production of lacto-proteins via CO₂ thermo-catalysis coupled with P. pastoris fermentation (Lv et al. 2023a, b).

Table 5.

Genomic modified microbial production of proteases

Host strains	Proteases or encoding genes	Strategies	MW/kDa	References
B. subtilis SCK6	Thermophilic serine protease gene P3862 from Ornithinibacillus caprae L9T	Cloned gene from O. caprae L9T; achieved heterologous expression in B. subtilis SCK6	29	(Li et al. 2022b)
B. subtilis WB600	BaApr1 from the Bacillus altitudinis W3	Sequenced B. altitudinis W3 genome (revealed nine aprE genes); heterologous expression in B. subtilis enabled high-efficiency protease identification	37.29	(Yang et al. 2020; Li et al. 2022a)
B. subtilis RIK1285	Alkaline protease (AprBcp) from Bacillus circulans R1	Optimized promoter/signal peptide; achieved efficient high-level production in B. subtilis via integrated strategies and scale-up cultivation	30	(Chen et al. 2022a)
B. subtilis	Alkaline serine protease gene (GsProS8) from Geobacillus stearothermophilus pWB980	Cloned and expressed in B. subtilis; yielded activity of 3807 U/mL via high-cell-density fermentation	27.2	(Chang et al. 2021)
B. subtilis WB800	Subtilisin-like alkaline serine protease (ASP) from B. halodurans C-125 pMA0911	PCR amplification and cloning into pMA0911 vector (PHpaII promoter); expressed extracellularly in protease-deficient B. subtilis WB800	28.3	(Tekin et al. 2021)
B. subtilis	Alanine aminopeptidase from B. licheniformis E7 pMA0911	Identified from B. licheniformis E7; heterologously expressed in B. subtilis	52	(Chen et al. 2022b)
E. coli	Thermostable serine alkaline protease from Geobacillus thermoglucosidasius SKF4	Cloned gene into linearized pEASY-Blunt E1 expression vector	28	(Allison et al. 2024)
E. coli BL21 (DE3)	Codon-optimized subtilisin gene from B. subtilis	Codon-optimized and cloned using pSUMO vector; Induced for high-level expression in E. coli BL21 (DE3)	40	(Shettar et al. 2023)
E. coli Shuffle®T7	Serine carboxypeptidases SCP3 from Nepenthes mirabilis	Bioinformatically characterized; SCP3 cloned, purified, and shown active after expression in E. coli Shuffle® T7	32	(Porfírio et al. 2022)
P. pastoris	Streptomyces griseus trypsin (SGT)	Engineered α signal peptide; identified key residues via evolutionary analysis; constructed variants with enhanced activity/specificity	23.1	(Shi et al. 2023)
P. pastoris	A novel aminopeptidase B from Aspergillus niger pPIC9K	Cloned apb-AN from A. niger into pPIC9K vector; expressed in P. pastoris	100	(Song and Feng 2021)
P. pastoris SMD1168 and X33	Serine alkaline protease from Melghiribacillus genus pPICZαC	Isolated, cloned, and sequenced sapN gene; expressed in E. coli and P. pastoris	39	(Mechri et al. 2021)
P. pastoris GS115	Recombinant protease MarP from Mycobacterium tuberculosis pPICZα	Constructed P. pastoris vectors (full/partial MarP); selected GS115 clones; assessed AOX1 recombination by PCR	28	(García-González et al. 2021)

Additionally, the rapid advancement of genomic editing technologies has greatly facilitated the bioproduction capability of proteases (Martínez-Medina et al. 2024; Humaira et al. 2024) and glucanases (Liu et al. 2024a) in microorganisms. Deaminases can be produced in both E. coli and B. subtilis, enabling the development of in vivo continuous directed evolution systems (Wang et al. 2025a) and dCas12a-based dual-function base editors (Wu et al. 2024b), which gives a promising future for heterologous production for AMPD.

Enzyme engineering strategies

In the past 10 years, site-directed mutagenesis and directed evolution have been extensively applied to enhance the catalytic efficiency and stability of various enzymes (Chai et al. 2025). In recent years, rational and semi-rational design approaches have gained prominence in enhancing the functionality of glucanases and proteases. For example, B. subtilis β-glucanase was improved through a two-stage engineering process that utilized the complementary computational tools Pythia and ESM-2 to simultaneously boost enzyme stability and activity (Zhang et al. 2025). Building on this computational momentum, Zhao et al. (2025) applied energy-based modeling to optimize the thermostability of a 1,3–1,4-β-glucanase variant, which resulted in increased thermal resistance, specific activity, catalytic efficiency, and protease tolerance.

By integrating semi-rational, rational, and in silico design strategies, recent studies have significantly improved protease performance. For the enhancement of endo-peptidases, Wang et al. (2024a) applied semi-rational design to carboxypeptidase BmeCPM32 from B. megaterium, achieving a 2.2-fold increase in specific activity, a 2.9-fold rise in catalytic efficiency, a 1.8-fold longer half-life at 60 °C, plus a 55% boost in umami intensity and an 83% reduction in bitterness of soy protein isolate hydrolysate. Yuan et al. (2024) employed targeted rational design, informed by homologous sequence comparison, to enhance the stability and performance of the serine protease BAPB92 from B. alcalophilus. The resulting optimal double mutant, BAPB92 (Q239R/V262I), achieved a 2.95-fold increase in specific activity and successfully elevated the optimal operating temperature from 50 to 60 °C. In silico approaches have achieved more in protease design. Duan et al. (2024) generated a calcium-independent proteinase K mutant that exhibits a 9.2-fold increase in activity. Similarly, Liu et al. (2025) optimized non-catalytic regions of elastase expressed in S. cerevisiae, targeting propeptide and signal peptide cleavage sites as well as N-glycosylation motifs, resulting in variants with up to 23% improvement.

Other strategies, such as immobilization, have improved enzyme stability and reusability while enabling continuous processing (Sun et al. 2022a; Wen et al. 2023). Continuous evolution methods have emerged as powerful tools for elucidating evolutionary mechanisms and enhancing cellular and enzymatic properties (Chu et al. 2024).

Machine learning in enzyme modification

Acting as a synergistic partner to traditional enzyme engineering, machine learning (ML) is now instrumental in pinpointing viable starting candidates (Cheng et al. 2023) and steering directed evolution toward application-specific performance enhancements (Yang et al. 2024). While the surge in enzyme-related data has fueled ML-driven design by exposing predictive patterns, creating models capable of broad generalization across varied enzyme families continues to pose a significant hurdle (Zhou and Huang 2024). Innovative approaches have been devised to navigate this complexity, such as isothermal compressibility-assisted dynamic squeezing index perturbation engineering (Wang et al. 2025b), which constructs hierarchical modular networks for enzymes with distinct structural and functional profiles. A case in point is the work of Lu et al. (2023a), who developed a structure-aware graph convolutional network that integrates protein–ligand interaction energetics from Rosetta. This model successfully predicted protease specificity for two noncanonical substrates, a finding subsequently leveraged to rationally redesign phosphatase BT4131 and modify its substrate preference (Lu et al. 2024).

Recent progress in ML-guided cell-free protein synthesis (CFPS) systems has further accelerated enzyme engineering workflows (Landwehr et al. 2025). Thornton et al. (2025) developed an integrated ML-driven platform combining cell-free DNA assembly, CFPS, and high-throughput functional assays to efficiently explore protein fitness landscapes and tailor enzymes for diverse catalytic activities. Genome-scale metabolic models enhance ML-guided CFPS by predicting precursor and cofactor requirements, informing extract and reaction design, and providing realistic training datasets (Bi et al. 2023; Lu et al. 2023b).

Together, these studies highlight the transformative role of ML in enzyme design, with particular promise for improving the performance of proteases, β-glucanases, deaminases, and other industrial enzymes, thereby enhancing the efficiency of yeast-based bioprocessing.

Conclusions and perspectives

As a foundation of yeast-base products’ development, the identification and enrichment of bioactive compounds, such as peptides, nucleotides, and antioxidants must be the priority to maximize the utility of yeasts (Ribeiro-Oliveira et al. 2021). By leveraging advanced processing methods like targeted enzymatic hydrolysis and fractionation, yeast extract can be transformed from a basic commodity into a premium bioproduct suitable for functional foods, therapeutics, and dermocosmetics (Mora and Toldrá 2023).

In parallel, the biosynthesis of yeast processing enzymes has seen substantial progress driven by interdisciplinary innovation. The search for novel biocatalysts has expanded into extreme environments, facilitated by metagenomics and bioinformatics (Ariaeenejad et al. 2024; Marzban and Tesei 2025). Production scalability has also improved through efficient heterologous expression in hosts like B. subtilis (Akinsemolu et al. 2024) and S. cerevisiae (Wang et al. 2025c). While traditional engineering strategies, such as directed evolution, have long been used to improve thermostability and specificity (Dinmukhamed et al. 2021), the recent integration of machine learning has revolutionized the field. These computational tools now allow for precise de novo enzyme design, accelerating optimization with speed and scalability that were previously unattainable (Orsi et al. 2024).

However, significant hurdles persist, particularly when considering proteases (Liu et al. 2024b). A primary bottleneck is the incomplete understanding of enzyme toxicity; their impact on host cell membranes and autodegradation regulation remains unclear, complicating viability maintenance. Furthermore, the pathways controlling enzyme secretion are not fully mapped, making it difficult to resolve productivity limits. Current eukaryotic systems also frequently lack the robustness needed for stable, high-yield production, and metabolic engineering efforts are often hindered by a limited grasp of the metabolic disturbances caused by heterologous expression. Finally, the field lacks predictive digital models that sufficiently integrate machine learning to forecast secretion dynamics. Addressing these gaps is crucial for meeting the growing industrial demand for diverse, high-quality, tailored yeast processing enzymes.

To overcome these hurdles, the future of yeast-based biomanufacturing lies in the convergence of these disciplines. Specifically, protein engineering can be used to screen and modify enzymes for high-performance yeast processing. Metabolic engineering provides tools to rebalance carbon flux and reduce host burden to increase the bio-productivity of enzymes. Machine learning enables data-driven prediction of enzyme secretion efficiency and stability. Integrating these strategies will be pivotal for transitioning from trial-and-error optimization to the precise and cost-effective production of high-performance yeast enzymes.

Author contribution

Jieying Deng prepared the original draft and contributed to the conception, with guidance from Long Liu. Long Liu and Xueqin Lv contributed to the conceptual design and secured funding for this work. Zhendong Li critically revised and edited the manuscript. All authors reviewed and approved the final version and agreed to be accountable for the integrity of the work.

Funding

This research was supported by the National Natural Science Foundation of China (Grant No. 32400054); Key R&D Program of Shandong Province, China (2023CXGC010714); Key Technological Project of Jiangxi Province (20244AFH82001); Jiangsu Basic Research Center for Synthetic Biology (Grant No. BK20233003).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethical approval

No work was undertaken with human participants or animals performed by any of the authors.

Competing interests

The authors declare no competing interests.