A Marine Alkaline Protease from Bacillus safensis DL12: Heterologous Expression, Purification and Preliminary Application in Animal Feed

Abstract

Cottonseed meal (CSM) is a cost-effective protein source, but its application is limited by the toxicity of free gossypol. Traditional physical and chemical detoxification methods are costly, energy-intensive, and cause nutrient loss, while microbial fermentation-based biological detoxification is considered more sustainable than chemical or physical approaches. This study reports an alkaline protease from the marine strain Bacillus safensis DL12 isolated from Yellow Sea sediments. Following cloning of its encoding gene and heterologous expression, enzymatic characterization of the purified enzyme revealed optimal activity at pH 8.0 and 50 °C, with Fe²⁺, Cu²⁺, Ni²⁺, and dithiothreitol (DTT) significantly enhancing its activity. Substrate hydrolysis analysis using the purified enzyme on soybean meal, peanut meal, rapeseed meal, and cottonseed meal demonstrated that, compared to the control group, cottonseed meal hydrolysates exhibited a 55.6% relative increase in peptide content and a 41.5% relative improvement in the degree of hydrolysis (DH), indicating higher hydrolysis efficiency among the four substrates. Notably, when hydrolyzing cottonseed meal with purified enzyme versus crude enzyme preparation at equivalent activity, the purified enzyme effectively reduced free gossypol content by 70% compared to the control, achieving more efficient detoxification than the crude enzyme preparation and most reported microbial treatments. These results highlight the potential of B. safensis DL12 protease as a marine-derived enzyme, offering promising prospects for enhancing protein digestibility and addressing the long-standing challenge of gossypol toxicity in cottonseed meal utilization.

1. Introduction

The global livestock industry is experiencing growing demand for high-quality protein feed, concurrent with rising prices of premium soybean meal. Cottonseed meal (CSM) presents a cost-efficient alternative [1]; however, its application is constrained by the presence of gossypol and other anti-nutritional components that negatively affect animal performance and raise concerns regarding food safety [2,3,4,5]. Therefore, effective detoxification methods are essential to enhance the utilization of CSM and reduce dependency on more expensive protein sources.

Physical and chemical detoxification methods are costly, energy-intensive, and may lead to nutrient loss [6]. Biological methods primarily rely on microbial fermentation, utilizing microorganisms such as Meyerozyma guilliermondii, Bacillus, and Lactobacillus mucosae to remove antinutritional factors including gossypol and other similar compounds. Proteases play a role in this process [7,8,9]. The majority of commercial proteases applied in animal feed are alkaline proteases derived from terrestrial Bacillus species, including B. subtilis, B. licheniformis, B. pumilus, and B. amyloliquefaciens. These enzymes are highly valued for their robust catalytic activity, shortened microbial fermentation cycles, and extracellular secretion [10,11,12,13]. Since the first identification of microbial alkaline protease from B. licheniformis in 1945 [14], research has expanded to include enzymes from alkaliphilic microorganisms, heterologous expression systems in Escherichia coli, and proteases derived from fungal and yeast sources [15,16]. Such investigations have substantially enhanced our understanding of the structural diversity, catalytic mechanisms, and prospective applications of microbial alkaline proteases. However, most industrial enzymes still exhibit limited thermal stability, poor pH adaptability, and insufficient resistance to inhibitory metal ions, which compromises their performance under the high-temperature, high-salt, and alkaline conditions commonly encountered in feed processing. Moreover, their efficiency in dealing with small-molecule anti-nutritional factors, such as gossypol, remains suboptimal.

Marine microorganisms are an untapped source of novel enzymes with superior industrial properties. Adapted to extreme conditions of high salinity, alkalinity, pressure, and fluctuating temperatures, they produce enzymes with exceptional stability and substrate adaptability [17,18]. Marine-derived alkaline proteases often show enhanced salt and alkali tolerance, heat stability, and broad substrate specificity, making them promising for improving the utilization of unconventional protein feed resources [19]. Yet, studies on marine proteases from the Yellow Sea, particularly the Dalian coastal region, are limited.

In this study, an alkaline protease-producing strain was isolated from Yellow Sea marine sludge and identified as B. safensis. The protease gene was cloned, heterologously expressed, and the recombinant enzyme purified. Its biochemical properties, including optimal temperature and pH, thermostability, metal ion effects, substrate specificity, and kinetic parameters, were systematically characterized. The enzyme’s ability to improve protein digestibility and reduce gossypol content in CSM was further evaluated under simulated feed processing conditions. These results provide a foundation for developing marine-derived alkaline proteases as cost-effective feed additives to enhance CSM value and help address protein feed shortages in the livestock industry.

2. Materials and Methods

2.1. Strain Isolation and Culture

Based on tidal forecasts and referencing the experimental method of Wang et al. [20] with minor modifications, sampling was conducted multiple times during ebb tide. At low tide, surface sediment samples (5–20 cm depth) (Supplementary Figure S1) were collected from the intertidal zone of Dalian Bay, China (39°00′04″ N, 121°38′26″ E) (Figure 1) using a sterile sampler. Samples were placed in low-density polyethylene plastic bottles. Seawater was collected from a depth greater than 30 cm below the surface (temperature: 27.4 °C) and transported to the laboratory within 1–2 h. The seawater was vacuum-filtered through a 0.45 µm membrane filter, sealed, and stored refrigerated for subsequent microbial medium preparation. Approximately 1 g of sediment was suspended in 9 mL of sterile saline solution, vortexed, and allowed to settle. One milliliter of the supernatant was inoculated into enrichment medium (beef extract 5 g, peptone 1 g, yeast extract 5 g, seawater 1000 mL, pH 8.0). To enhance microbial acclimatization to laboratory cultivation temperatures and screen for more temperature-adapted strains, initial incubation was conducted at 28 °C with 140 rpm shaking. After 12 h, the temperature was elevated to 33 °C, followed by a further increase to 37 °C at the 24 h mark, with continued shaking until 48 h. After cultivation, serial dilution was performed, and diluents were spread onto selective casein agar plates (casein 10 g, yeast extract 1 g, agar 15 g, seawater 1000 mL, pH 8.0), followed by incubation at 37 °C for 48 h. Colonies exhibiting clear hydrolysis zones were picked and repeatedly streaked for purification on 2216E medium (peptone 5 g, yeast extract 1 g, ferric phosphate 0.1 g, agar 15 g, seawater 1000 mL, pH 8.0). Purified strains were stored on 2216E slant medium for subsequent experiments.

For screening, bacterial strains were inoculated onto casein agar plates and incubated at 37 °C for 24 h. The hydrolysis zone diameter (D) and colony diameter (d) were measured and recorded. The D/d ratio was calculated as an indicator of proteolytic activity. To enhance the visibility of hydrolysis zones, 10% (w/v) trichloroacetic acid (TCA) solution was uniformly added over the plate surface. This solution precipitates undigested casein, thereby improving contrast.

2.2. Protease, Amylase and Cellulase Activity Assay

Protease activity, amylase activity and cellulase activity assays were performed according to the method of Chen et al. [21] with minor modifications. Protease activity was determined using the Folin–Ciocalteu method. Under specific assay conditions, one unit of enzyme activity (U) was defined as the amount of enzyme required to liberate 1 µg of tyrosine per minute from casein hydrolysis. A standard curve was prepared using tyrosine (Aladdin, Shanghai, China) as the standard. The crude enzyme solution was obtained by inoculating strain DL12 into enrichment medium or fermentation medium (tryptone 10 g/L, yeast extract 3 g/L, sodium chloride 10g/L, artificial seawater 1000 mL), incubating at 37 °C for 24 h, followed by centrifugation to collect the supernatant. The reaction mixture for the test group consisted of 1 mL of crude enzyme solution and 1 mL of 2% (w/v) casein solution, incubated at different temperatures and pH for 10 min. The reaction was terminated by adding 2 mL of 0.4 M TCA (with blanks prepared by adding TCA first, incubating for 10 min, and then adding the casein solution). After standing for 15 min, the mixture was filtered. One milliliter of the filtrate was mixed with 5 mL of 0.4 M Na₂CO₃ and 1 mL of Folin–Ciocalteu reagent (Solarbio, Beijing, China). The mixture was thoroughly mixed and incubated at 40 °C for 20 min. The absorbance was measured at 680 nm, and the enzyme activity was calculated based on the standard curve. Amylase and cellulase activities were determined using the 3,5-dinitrosalicylic acid (DNS) method. Under specified assay conditions, one unit of enzyme activity (U) was defined as the amount of enzyme required to liberate 1 µg of glucose per minute from starch or cellulose hydrolysis. A standard curve was prepared using glucose (Sangon, Shanghai, China) as the standard. The test group reaction mixture contained 1 mL of centrifuged crude enzyme solution and 1.5 mL of starch solution (0.1 g/mL) or cellulose solution (0.01 g/mL) (the control group was added with an equivalent volume of sterilized fermentation medium), incubated under respective conditions in a water bath for 30 min. Then, 1.5 mL of DNS reagent was added (potassium sodium tartrate 18.2 g, 3,5-dinitrosalicylic acid 0.63 g, NaOH 2.096 g, phenol 0.5 g, sodium sulfite 0.5 g, deionized water 100 mL), followed by boiling in a water bath for 5 min. After removal and cooling, the mixture was diluted to 25 mL with deionized water. After thorough mixing, absorbance was measured at 540 nm wavelength using the sterilized fermentation medium group as the blank control.

2.3. Phylogenetic Tree Construction

Referring to the experimental method of Wang et al. [20] with minor modifications, 1 mL of logarithmic-phase culture was collected and delivered to Sangon Biotech Co., Ltd. (Shanghai, China) for DNA sequencing service; the 16S rDNA gene of bacterial isolates was amplified using universal primers 27F (5′-AGAGTTTGATCCTGGCTCAG-3′) and 1492R (5′-GGTTACCTTGTTACGACTT-3′), yielding an amplified fragment of approximately 1500 bp. The obtained sequences were analyzed by BLAST (v2.15.0) against the NCBI database (v2.10.1) to identify the most closely related taxa. Phylogenetic analysis was performed using the neighbor-joining method in MEGA software (v11.0.13), and its taxonomic position was further confirmed by multi-locus sequence typing (MLST).

2.4. Genomic DNA Extraction and Sequencing

Genomic DNA extraction was performed according to the method of Wang et al. [22], as in the following text. Strain DL12 was cultured to the logarithmic growth phase, and bacterial pellets were harvested by centrifugation at 13,000 rpm for 30 min. Genomic DNA was isolated using the Bacterial DNA Isolation Kit (FOREGENE, Chengdu, China) based on the sodium dodecyl sulfate (SDS) method. DNA quality was assessed by agarose gel electrophoresis, while purity and concentration were determined by measuring absorbance ratios at 260/280 nm (A₂₆₀/A₂₈₀) and quantified using a Qubit^® 2.0 Fluorometer (Thermo Scientific, Waltham, MA, USA). Whole-genome sequencing was conducted by Novogene Bioinformatics Technology Co., Ltd. (Beijing, China)using single-molecule real-time (SMRT) technology on the PacBio platform. Raw reads were quality-filtered using SMRT Link v5.0.1. De novo assembly was performed with the same software to generate preliminary contigs. Filtered reads were mapped back to these contigs to evaluate sequencing depth distribution. Contigs were classified as chromosomal or plasmid sequences based on length and mapping characteristics, with circularization verified for plasmid contigs (Supplementary Tables S1 and S2).

Genomic component prediction included the prediction of coding genes, non-coding RNAs, genomic islands, prophages, and Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR). We used the GeneMarkS program to retrieve relevant coding genes. Transfer RNA (tRNA) genes were predicted by tRNAscan-SE. Ribosomal RNA (rRNA) genes were analyzed by rRNAmmer. Genomic islands were predicted using the IslandPath-DIOMB program. Prophage prediction was performed using PHAST (https://phaster.ca/ accessed on 31 January 2024.), and Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) identification was performed using CRISPRFinder (v4.3.2).

2.5. Genome Annotation

The predicted coding sequences (CDSs) were annotated through alignments using the following methods: Clusters of Orthologous Groups (COG), aligned against the COG database using DIAMOND software (v2.0.7); Kyoto Encyclopedia of Genes and Genomes (KEGG), aligned against the KEGG database using DIAMOND software; Non-Redundant protein database (NR), aligned against the NR database using DIAMOND software; Gene Ontology (GO), aligned against the GO database using InterProScan software (v5.72-103.0); Transporter Classification Database (TCDB), aligned against the TCDB database using DIAMOND software; Comprehensive Antibiotic Resistance Database (CARD), aligned against the CARD database using RGI (v5.1.0) software; Protein families database (Pfam), aligned against the Pfam-A database using InterProScan software; Carbohydrate-Active enZYmes Database (CAZy), aligned against the CAZy database using DIAMOND software; and Swiss-Prot Protein Knowledgebase (Swiss-Prot), aligned against the Swiss-Prot database using ExPASy Server (https://www.expasy.org/ accessed on 13 February 2024). The alignment results were subsequently integrated with the target species’ gene information and corresponding functional annotations to generate database-specific annotation outcomes.

2.6. Heterologous Expression, Purification, and Basic Characterization of Recombinant Alkaline Protease

Based on genomic annotation, genes exhibiting high sequence similarity to known alkaline proteases were selected from strain DL12 for heterologous expression. Primary structure (amino acid sequence), theoretical molecular weight, and isoelectric point were calculated using the ExPASy ProtParam tool (https://web.expasy.org/protparam/ accessed on 12 March 2024.). Phylogenetic analysis of the protease sequence was performed with MEGA software (v11.0.13) to confirm its taxonomic classification. The target gene was amplified via Polymerase Chain Reaction (PCR), cloned into E. coli DH5α for plasmid propagation, and subsequently transformed into various expression hosts (Tuner(DE3), Rosetta(DE3), C43(DE3), BL21(DE3), and BL21(DE3)-pLysS) through heat shock [23]. All competent cells were purchased from Sangon Biotech Co., Ltd. (Shanghai, China). Protein expression was induced with varying concentrations of isopropyl β-D-1-thiogalactopyranoside (IPTG; 0.2, 0.4, 0.6, 0.8 mM) to determine the optimal inducer concentration. Expression conditions were further optimized by testing different induction temperatures and durations. Cells were lysed by ultrasonication and centrifuged at high speed to obtain soluble protease-containing supernatant. The supernatant was filtered through a 0.22 μm aqueous membrane and incubated with HisSep Ni-NTA resin. After incubation, the flow-through was discarded. The resin was washed with low-imidazole buffer (20 mM) to remove non-specific proteins, followed by an intermediate wash with 50 mM imidazole buffer. The target protein was then eluted stepwise with increasing imidazole concentrations (100, 200, and 300 mM). The eluted protease was concentrated using ultrafiltration tubes and further purified by gel filtration chromatography on an NGC Scout 10 Plus system. High-purity protease fractions were collected based on 280 nm absorbance peaks, with final purity verified by SDS-PAGE analysis.

The measurement method for alkaline protease activity is as described in Section 2.2 above. Specific activity was expressed as U/mg. The eluted protease was concentrated using ultrafiltration tubes and further purified by gel filtration chromatography on an NGC Scout 10 Plus system. High-purity protease fractions were collected based on 280 nm absorbance peaks, with final purity verified by SDS-PAGE analysis.

2.7. Effects of Metal Ions and Chemical Reagents

The enzyme reaction system was consistent with the aforementioned method, with metal ions including Ca²⁺ (calcium chloride), Mg²⁺ (magnesium chloride), Mn²⁺ (manganese chloride), Zn²⁺ (zinc sulfate), Fe²⁺ (ferrous sulfate heptahydrate), Cu²⁺ (copper sulfate pentahydrate), Ni²⁺ (nickel chloride hexahydrate), Co²⁺ (cobalt chloride hexahydrate), and K⁺ (potassium chloride) added to the enzyme solution at a final concentration of 1 mM (all metal salts used were of analytical grade). After adding 1% casein, the mixture was reacted at the optimal temperature for 10 min. Enzyme activity changes were measured, with enzyme solution without additives serving as the control. Relative activity was calculated. For chemical agent tests, enzyme solutions were mixed with 1 mM concentrations of the following reagents: Triton X-100, 1% SDS solution, β-mercaptoethanol (β-ME), Tween-20, Tween-80, phenylmethylsulfonyl fluoride (PMSF), ethylenediaminetetraacetic acid (EDTA), and dithiothreitol (DTT). After incubation at 25 °C for 1 h, the solutions were mixed with 1% casein and reacted at the optimal temperature for 10 min. Enzyme activity changes were subsequently measured.

2.8. Preparation of Enzyme-Hydrolyzed Feed

Referring to the method of Yu et al. [24] with minor modifications, the experimental groups involved grinding 25 g each of cottonseed meal (Yihai Grain & Oil Co., Ltd., Changji, China), rapeseed meal (Lankun Feed, Heze, China), soybean meal (Jiahui Feed, Datong, China), and peanut meal (Tongsheng Feed, Jining, China) into fine powder. The powder was mixed with an enzyme solution containing 5 mg of protein. The mixture was brought to a final volume of 50 mL with sterile deionized water. The blank control group (CK) received only 50 mL of deionized water. The DL12 group received crude enzyme solution from the DL12 strain with equivalent protease activity to that used in the experimental groups. After thorough homogenization, semi-solid-state enzymatic hydrolysis was carried out at 50 °C for 48 h. Upon completion of enzymatic hydrolysis, the samples were dried at 35 °C to constant weight. The basic component contents of the respective feed materials are shown in Supplementary Table S3.

2.9. Determination of Peptide Content

Referring to the method of Yu et al. [24] with minor modifications, the dried feed sample (0.500 g) was suspended in 50 mL of ultrapure water and shaken for 6 h. After centrifugation, the supernatant was mixed with an equal volume of 0.4 M TCA and incubated for 30 min, followed by another centrifugation. One milliliter of the resulting supernatant was reacted with four volumes of biuret reagent (copper sulfate 0.15 g, sodium potassium tartrate 0.6 g, sodium hydroxide 3 g, potassium iodide 0.1 g, and deionized water 100 mL) and incubated at room temperature for 30 min. The absorbance was measured at 540 nm. A standard curve was prepared using bovine serum albumin (BSA) (Solarbio, Beijing, China), and the peptide content was calculated.

2.10. Determination of Hydrolysis Degree

Referring to the method of An et al. [25] with minor modifications, the feed sample (0.500 g) was divided into two groups: one group was hydrolyzed with 10 mL of 6 M HCl at 110 °C for 24 h; the other parallel sample was treated with 10 mL of ultrapure water under the same conditions. After centrifugation, 400 μL of the supernatant was mixed with 3 mL of o-phthalaldehyde (OPA) reagent (sodium tetraborate decahydrate 11.43 g, sodium dodecyl sulfate 300 mg, o-phthalaldehyde 240 mg, DTT 264 mg, and deionized water 300 mL). The absorbances of samples were collected at 340 nm using a UV–visible spectrophotometer (UV2450, Shimadzu, Kyoto, Japan). A standard curve was prepared using serine (Solarbio, Beijing, China), and the degree of hydrolysis (DH) was calculated. The result from the hydrochloric acid treatment was designated as NH_{2 total}, and the result from the ultrapure water treatment was designated as NH_{2 free}. The calculation formula is as follows:

DH (%) = (NH_{2 free}/NH_{2 total}) × 100

2.11. Antioxidant Activity Assays

The method of Bhoopathy et al. [26] was used with minor modifications. For the detection of purified enzyme-hydrolyzed feed, 1 g of dried feed was extracted with 1 mL of ethanol at 4 °C for 1 h, vortexed for 2 min, and centrifuged. The supernatant was filtered and used for antioxidant activity assays. ABTS assay: 0.1 mL of the extract was mixed with 0.9 mL of ABTS⁺ solution (ABTS radical solution prepared by reacting 7 mM ABTS stock solution with 2.45 mM potassium persulfate, stored at room temperature and protected from light overnight, and then diluted with ethanol to an absorbance of 0.700 ± 0.05 at 734 nm). The mixture was incubated in the dark for 15 min, and the absorbance was measured at 734 nm. Ethanol was used as the control group under identical conditions, and ascorbic acid (Aladdin, Shanghai, China) was used as the positive control. DPPH assay: 1 mL of the extract was mixed with 0.1 mM DPPH solution (0.0034 g DPPH dissolved in 100 mL ethanol). The mixture was incubated in the dark for 15 min, and the absorbance was measured at 517 nm. Ethanol was used as the control group under identical conditions, and ascorbic acid (Aladdin, Shanghai, China) was used as the positive control. The scavenging rate of the positive control was set as 100%, and relative values were calculated. The calculation formula is as follows:

Radical scavenging activity (%) = [(A_c − A_s)/(A_c − A_a)] × 100%

where A_c is the absorbance of the control group, A_s is the absorbance of the sample, and A_a is the absorbance of the positive control (ascorbic acid).

2.12. Free Gossypol Determination

Free gossypol content was determined according to the method of Karishma et al. [27] and the National Standard “GB 13086-91 Method for Determination of Free Gossypol in Feeds”. A mixed solvent of isopropanol:n-hexane in a volume ratio of 6:4 was prepared. One gram of the feed sample was weighed into a conical flask, and 50 mL of extraction reagent (1 mL 3-amino-1-propanol, 4 mL glacial acetic acid, and 25 mL water, brought to 500 mL with the isopropanol mixed solvent) was added. The mixture was shaken for extraction for 1 h. After extraction, the mixture was filtered. Twenty milliliters of the filtrate were divided into two portions of 10 mL each, transferred into two brown volumetric flasks labeled a and b as the sample assay solutions. Twenty milliliters of the extraction reagent were divided into two portions of 10 mL each, transferred into two brown volumetric flasks labeled c and d as the blank assay solutions. Flask a and flask c were each filled to 25 mL with the isopropanol-n-hexane mixed solvent and mixed well. Two milliliters of aniline were added to flask b and flask d, and the mixtures were heated in a boiling water bath for 30 min for color development. After cooling to room temperature, the mixtures were each filled to 25 mL with the isopropanol-n-hexane mixed solvent, mixed well, and allowed to stand for 1 h. The absorbance was measured at a wavelength of 440 nm: the absorbance of d was measured using c as the blank reference solution, and the absorbance of b was measured using a as the sample reference solution. The corrected absorbance (A) was obtained by subtracting the absorbance of the blank assay solution from the absorbance of the sample assay solution. The free gossypol content was calculated as follows:

X(mg/kg) = [(A × 1.25)/ (α × m × V)] × 10⁶

where A is the corrected absorbance, α is the absorptivity coefficient of free gossypol (62.5 cm²/g), m is the sample mass (g), and V is the volume of the filtrate (mL).

2.13. Statistical Analysis

All data are presented as mean ± standard deviation (SD). Statistical analyses were performed using GraphPad Prism 10.0. One-way analysis of variance (ANOVA) was used to assess differences among groups, followed by Tukey’s post hoc test for multiple comparisons. Differences were considered statistically significant at p < 0.05, highly significant at p < 0.01, and extremely significant at p < 0.001. Graphs were generated using GraphPad Prism 10.0 and Origin 2018.

3. Results

3.1. Isolation and Identification of a High-Protease-Producing Strain from Marine Sediments

From marine sediment samples, microbial enrichment, preliminary screening, and isolation were conducted. Colonies with diverse morphologies were observed, ranging in color from creamy white to pale yellow and orange, with diameters of 1–8 mm and variable margins and surface textures (Figure 2A). On casein agar, 55 isolates formed distinct hydrolysis zones and were therefore considered potential alkaline protease producers. Following purification, the hydrolysis-to-colony diameter ratio (D/d) was determined, and 12 strains with higher D/d values were selected. Among these, strain DL12 exhibited the largest ratio (D/d = 3.23) (Figure 2B), suggesting strong proteolytic capacity. Cultured using enriched medium, protease activity in liquid crude enzyme solution was further evaluated using Folin’s phenol method. Strains DL07, DL12, and DL31 displayed the highest enzyme activities of 129.67, 156.48, and 179.39 U/mL, respectively (

Table 1. Protease activity of bacterial strains isolated by dilution plating.

Strain Number	Enzyme Activity (U/mL)
7	129.67
11	53.57
12	156.48
13	49.82
14	95.59
15	50.60
18	55.96
20	50.79
24	83.36
26	55.81
29	93.25
31	179.39

). Under subsequent experimental conditions, strains DL07, DL12, and DL31 were compared, with DL12 exhibiting faster growth rate and more stable enzyme production. During serial subculture experiments, the enzyme-producing capacity of strain DL31 gradually decreased and occasionally disappeared entirely. In contrast, DL12 demonstrated consistent and stable enzyme production unaffected by successive subcultures. Therefore, strain DL12 was selected for further experimentation.

Microscopic and molecular characterization revealed that DL12 was a Gram-positive bacillus (Figure 2C). Sequencing of the 16S rDNA gene demonstrated 99% similarity to B. safensis in GenBank, and phylogenetic analysis confirmed that DL12 clustered closely with B. safensis (Figure 2D). Accordingly, the isolate was designated B. safensis DL12. The 16S rDNA gene sequence of the bacterial strain has been deposited in GenBank under accession number PX884774. Enzyme activity assays revealed that strain DL12 exhibits only good protease activity (Supplementary Figure S2).

Cultured using fermentation medium, preliminary enzymatic characterization indicated that the crude protease retained detectable activity across 10–65 °C, with maximum activity at 37 °C but a sharp decline above 40 °C (Figure 2E). The enzyme maintained full activity after 60 min at 35 °C, although thermal stability decreased progressively at higher temperatures. Activity assays across different pH values revealed robust performance within pH 8.0–13.0, with maximum activity under alkaline conditions (Figure 2F), confirming its alkaline-adapted nature. The enzyme activity exhibited a certain difference from that in Table 1, possibly due to the use of different media in the two trials. Preliminary enzymatic characterization of B. safensis DL12 crude enzyme enables initial screening and rapid identification of fundamental enzymatic characteristics of the protease produced by this strain, thereby providing critical directional guidance for subsequent heterologous expression, purification, and precise characterization of the recombinant enzyme. Moreover, combined with the stable enzyme-producing capacity of DL12, the preliminary enzymatic properties confirm its potential industrial application value, allowing exclusion of strains with unstable enzymatic properties during early screening. This approach establishes a natural reference for the enzymatic properties of the purified recombinant enzyme, facilitating elucidation of the impacts of expression systems, purification processes, and absence of endogenous cofactors on protease enzymatic properties, and further clarifying the structural and functional features of the target protease itself.

3.2. Genomic Insights into Protease Potential of Strain DL12

The draft genome of B. safensis DL12 is 3.79 Mb in size with a GC content of 41.55%, containing 3990 predicted coding sequences (

Table 2. Genomic characteristics of strain DL12.

Feature	Value
Genome size (bp)	3,787,914
Chromosome	1
Topology	Circular
G + C content (%)	41.55
Protein-coding genes (CDS)	3990
Gene total length (bp)	3,366,324
Gene len/genome (%)	88.87
Intergenetic region length (bp)	421,590
Intergenetic length/genome (%)	11.13
rRNA (5 s rRNA, 16s rRNA, 23 s rRNA)	24
tRNA	81
Genomic islands	8
Prophage	11
CRISPR	11

). The annotation results for the target genes showed an 82% consistency rate across the NR, Pfam, and SwissProt databases, with core functional domains (e.g., alkaline protease catalytic domains) being fully consistent. For discrepancies between databases, manual curation has been performed to confirm the accuracy of core functional annotations. By aligning the genomic sequences against eight commonly used databases, we matched 3886, 2769, 3776, 3307, 502, 142, 2769, and 2914 genes to sequences in the NR, GO, KEGG, COG, TCDB, CAZy, Pfam, and Swiss-Prot databases, respectively (Supplementary Table S4). Functional annotation revealed a substantial proportion of genes related to amino acid transport and metabolism, as well as carbohydrate utilization, reflecting the strain’s strong capacity for protein and polysaccharide degradation. Of particular interest, 50 genes were directly associated with protease activity, while KEGG pathway analysis highlighted extensive enrichment in amino acid and protein metabolism, forming the genetic basis for efficient protease production (Supplementary Figures S3–S5).

142 carbohydrate-active enzyme (CAZy) genes were detected, including glycoside hydrolases and carbohydrate-binding modules, indicating that strain DL12 possesses the ability to degrade polysaccharides such as amylose, cellulose, and chitin (

Table 3. Gene categories and copy numbers encoding various types of carbohydrate-active enzymes and proteases in the genome of strain DL12.

Carbohydrate Enzymes and Proteases	Number of Genes
Glycoside hydrolases	56
Glycosyl transferases	28
Carbohydrate esterases	19
Polysaccharide lyases	2
Auxiliary activities	1
Carbohydrate-binding modules	36
Proteases	73

). In addition, genome mining further identified 73 protease-related genes, including 22 alkaline proteases, 21 metalloproteases, and seven cysteine proteases (Table 4). The abundance of alkaline proteases is especially relevant for feed applications, given their stability and functionality under gastrointestinal conditions. However, enzymatic activity assays revealed that the amylose- and cellulose-degrading capabilities of strain DL12 were not prominent (Supplementary Figure S2). This may be attributed to the protease-focused screening process and the use of culture media containing beef extract and peptone, which reduced the strain’s requirement for polysaccharide degradation and thereby suppressed the expression of related genes.

Together, these genomic features indicate that DL12 possesses not only a broad reservoir of protease-encoding genes but also auxiliary carbohydrate-degrading enzymes, underscoring its potential as a versatile microbial source for feed protein hydrolysis and nutritional enhancement.

3.3. Cloning and Optimized Heterologous Expression of DL12 Alkaline Protease

The complete CDS sequence of the target protease was obtained through genomic sequencing and annotation, and deposited in GenBank under accession number PRJNA1404957. Genomic annotation revealed 89% homology with known feed-grade alkaline proteases, with a calculated molecular weight of 13.3 kDa and isoelectric point (pI) of 5.7. Phylogenetic analysis confirmed its closest relationship to alkaline protease (Supplementary Figure S6). Based on genome annotation, a gene with high sequence similarity to known alkaline proteases was selected from strain DL12 for heterologous expression. The target gene was cloned into the pET-28a(+) vector and expressed in Escherichia coli. To enhance soluble protein yield, expression parameters were systematically optimized, including host strain, induction temperature, IPTG concentration, and induction duration.

Among the tested hosts, E. coli Rosetta (DE3) produced the most distinct recombinant protein bands with minimal background expression (Figure 3A), and was therefore chosen for subsequent optimization. Induction at 24 °C yielded the highest level of soluble protein (Figure 3B), indicating that reduced temperature favored proper folding. IPTG concentrations between 0.1 and 1.0 mM showed comparable expression levels; however, 0.2 mM was selected as the optimal concentration owing to its balance of efficiency and cost-effectiveness (Figure 3C). Time-course analysis revealed that maximum soluble protein accumulation occurred after 20 h of induction (Figure 3D).

For purification, Ni²⁺ affinity chromatography was applied, and the target protein was predominantly eluted at 200 mM imidazole, producing a clear 13.3 kDa band (Figure 4A). Subsequent gel filtration chromatography further improved purity, with a major peak at 41 mL corresponding to the expected protease; the protein purification achieved a 5.7-fold purification with a 56.7% recovery yield (Figure 4B). SDS-PAGE analysis confirmed that the molecular weight matched the predicted enzyme. SDS-PAGE analysis confirmed that the target protease band aligned with the theoretical molecular weight trend and was close to the predicted 13.3 kDa, but a slightly higher apparent molecular weight (approximately 14.4 kDa, Figure 4C) was observed during gel electrophoresis. This electrophoretic retardation phenomenon (apparent molecular weight exceeding theoretical value) is primarily attributable to three structural characteristics of the protease: the abundance of acidic amino acid residues on the protein surface, the presence of polyglycine/polyserine flexible regions in the primary structure, and the distribution of charged residue clusters. These factors collectively alter the migration rate of the protein in the polyacrylamide gel matrix under denaturing conditions without affecting the catalytic activity of the protease. Crucially, the specific target band at this position was further verified by nickel affinity chromatography purification, exhibiting a single elution peak at 200 mM imidazole concentration (Figure 4A). The final recombinant protein concentration reached approximately 9.07 mg/mL (Figure 4C).

Through cloning, host selection, and expression optimization, high-yield soluble production of the DL12 alkaline protease was successfully achieved. This provides a robust foundation for subsequent enzymatic characterization and evaluation of its potential as a feed additive.

3.4. Biochemical Characteristics Relevant to Feed Applications

To assess the suitability of the DL12-derived alkaline protease for feed and industrial applications, its catalytic properties, stability, and sensitivity to inhibitors and metal ions were systematically evaluated.

The enzyme exhibited optimal activity at pH 8.0, consistent with its alkaline nature (Figure 5A). More than 65% of activity was retained between pH 6.0 and 8.0, with >80% residual activity at both pH 7.0 and 10.0. The enzyme also displayed strong stability across pH 6.5–8.0 after prolonged incubation (Figure 5B), supporting its adaptability to neutral and mildly alkaline conditions relevant to feed processing.

Maximum activity was observed at 50 °C (Figure 5C). The enzyme maintained >50% activity after 2 h incubation at 4–40 °C, indicating moderate thermal stability. However, exposure to 60–70 °C for 30 min reduced activity to <40% (Figure 5D), suggesting sensitivity to high-temperature denaturation. Although less thermostable than some Bacillus proteases active at ≥60 °C, its temperature tolerance remains compatible with conventional feed pelleting and enzymatic hydrolysis conditions.

Table 4. Prediction of protease-encoding gene types and their copy numbers in the genome of strain DL12 using NR, KEGG, COG, TCDB, Pfam, and SwissProt databases.

Type of Proteases	Number of Predicted Proteases
Alkaline proteases	22
Metalloproteases	21
Aspartic proteases	0
Cysteine proteases	7
Threonine proteases	0
Others	23
Total	73

Table 4. Prediction of protease-encoding gene types and their copy numbers in the genome of strain DL12 using NR, KEGG, COG, TCDB, Pfam, and SwissProt databases.

Type of Proteases	Number of Predicted Proteases
Alkaline proteases	22
Metalloproteases	21
Aspartic proteases	0
Cysteine proteases	7
Threonine proteases	0
Others	23
Total	73

The protease was strongly inhibited by PMSF, reducing activity below 50%, confirming its classification as a serine protease (Figure 5E). Activity was also markedly suppressed by EDTA (~25% residual), indicating partial dependence on metal ions for catalysis. Interestingly, the reducing agent DTT enhanced activity to ~200%, suggesting a structural requirement for disulfide bond reduction in maintaining the active conformation.

At 1 mM, Fe²⁺, Cu²⁺, and Ni²⁺ significantly enhanced activity (170%, 160%, and 122% of control, respectively), implying cofactor-mediated activation (Figure 5F). In contrast, Mn²⁺, Co²⁺, K⁺, Mg²⁺, and Zn²⁺ strongly inhibited activity. Non-ionic surfactants such as Tween-20, Tween-80, and Triton X-100 also caused measurable inhibition, reflecting sensitivity to detergent environments.

Collectively, the recombinant DL12 protease demonstrates robust activity under alkaline pH, moderate thermal stability, and distinct responsiveness to specific ions and inhibitors. These properties highlight its potential for feed applications, particularly in protein hydrolysis and nutrient release under conditions simulating the animal digestive tract or feed-processing systems.

3.5. In Vitro Evaluation of DL12 Alkaline Protease in Plant-Based Feed Substrates

To further evaluate the practical applicability of DL12 alkaline protease, its hydrolytic efficacy was tested on four plant-based feed substrates: cottonseed meal, rapeseed meal, soybean meal, and peanut meal. The peptide content, DH, and antioxidant activities, ABTS and DPPH radical scavenging capacity, of the enzymatic hydrolysates were determined.

Compared with the control group, among the four feed ingredients, peptide content analysis (Figure 6A) showed that cottonseed meal hydrolysates exhibited the highest relative increase at 55.6%, followed by rapeseed meal at 25.5%. In contrast, DH results (Figure 6B) revealed that cottonseed meal achieved the highest hydrolysis efficiency, reaching 41.5%, suggesting strong substrate–enzyme compatibility, with rapeseed meal ranking second at 28.5%. Regarding antioxidant properties, ABTS radical scavenging capacity was highest in peanut meal hydrolysates, with rapeseed and cottonseed meals showing moderate activity and soybean meal the lowest. However, no significant differences were observed among the four hydrolysates in terms of DPPH radical scavenging capacity (Figure 6C). Collectively, these results indicate that DL12 alkaline protease is particularly effective for hydrolyzing cottonseed, and rapeseed meals, thereby improving digestibility and enhancing functional small peptide production.

Given the limited use of cottonseed meal in livestock feeds due to the antinutritional factor free gossypol, further in-depth evaluation was performed on this substrate. Comparative analysis of purified DL12 protease and DL12 crude enzyme solution with equivalent enzymatic activity revealed distinct differences (Figure 6D). Compared with the CK group, treatment with purified protease significantly reduced free gossypol levels in cottonseed meal, exhibiting a reduction of 70%, whereas the crude enzyme solution broth unexpectedly increased free gossypol content, likely due to interfering metabolites or unpurified components. These findings highlight that the purified DL12 alkaline protease, rather than crude enzyme solution, holds substantial potential for both effective protein hydrolysis and antinutritional factor removal in cottonseed meal.

4. Discussion

In this study, a high-yield alkaline protease-producing strain, B. safensis DL12, was isolated from Bohai Sea sediments. As a spore-forming and non-pathogenic species, B. safensis is widely recognized as biosafe [28] and has been increasingly applied in biotechnological processes [29,30,31,32]. Genomic analysis of DL12 further supported its safety profile (

Table 5. Annotation categories and counts of antimicrobial resistance genes for strain DL12 in the CARD database.

Antibiotic Resistance Gene Classes	Number of Genes
Glycopeptide antibiotic	3
Phenicol antibiotic	1
Disinfecting agents and antiseptics	2
Total	6

) and revealed a diverse repertoire of genes related to amino acid metabolism and protease biosynthesis (Supplementary Figures S3–S5), suggesting strong inherent potential for protein degradation and nutrient conversion.

Compared with previously reported protease-producing strains, DL12 possesses a higher abundance of genes associated with amino acid metabolism and protease functions, offering a genomic advantage for efficient enzyme production [33,34,35]. Heterologous expression in E. coli Rosetta (DE3) successfully addressed the limitations of natural producers by improving protein solubility and yield, thereby establishing a feasible approach for scalable enzyme preparation. Biochemical characterization identified the DL12 protease as an alkaline serine protease with optimal activity at pH 8.0 and 50 °C, moderate thermal stability, and strong adaptability under neutral to alkaline conditions [36]. The crude protease and purified protease from Bacillus safensis DL12 exhibited markedly different optimal reaction conditions, with the crude enzyme showing maximum activity at 37 °C and pH 10–11, whereas the purified enzyme reached its peak activity at 50 °C and pH 8.0. These discrepancies likely arise from the distinct biochemical environments of the two enzyme preparations. In the crude fermentation broth, the protease coexists with various endogenous components, such as metal ions, short peptides, and other extracellular proteins, secreted by B. safensis DL12. These molecules may form transient complexes with the enzyme, subtly modifying its conformation and surface charge distribution, thereby shifting its apparent optimal pH toward a more alkaline range (pH 10–11). In contrast, the purified protease, isolated as a single protein species, exhibits its intrinsic catalytic characteristics, showing maximal activity at a mildly alkaline pH (~8.0) consistent with the Ser–His–Asp catalytic triad of typical serine proteases. The elevation of the optimal temperature from 37 °C (crude) to 50 °C (purified) may reflect differences in protein folding and microenvironmental stabilization. During heterologous expression in E. coli Rosetta (DE3) and subsequent purification involving mild denaturation/renaturation, ultrasonic disruption, and Ni-NTA affinity chromatography, partial refolding or conformational tightening may have enhanced the enzyme’s intrinsic thermal stability. In contrast, the crude enzyme operates within a complex extracellular milieu that restricts thermal adaptability, reflecting the moderate temperature niche (20–30 °C) of the Yellow Sea sediments from which the strain originated. Taken together, the observed variations in optimal temperature and pH likely result from the combined effects of cofactors, microenvironmental components, and structural stabilization during expression and purification. Further analyses, such as metal ion chelation, reconstitution experiments, or spectroscopic assessment of protein conformation, will be required to elucidate the precise molecular basis of these differences. These features align well with feed-processing requirements. Its inhibition by PMSF and EDTA, together with activation by Fe²⁺, Cu²⁺, and Ni²⁺, suggests partial metal ion dependence and a catalytic mechanism distinct from some previously reported Bacillus proteases. PMSF inhibition assays confirmed that DL12 protease is a serine protease. Regarding the mechanism of Ca²⁺-induced inhibition, we propose that it occurs through the protease’s dependence on the spatial conformation of the Ser-His-Asp catalytic triad. Specifically, Ca²⁺ likely binds to acidic amino acid residues (e.g., Asp/Glu) on the enzyme surface, inducing local conformational folding. This reduces the binding affinity of the active site for substrates (such as casein and cottonseed meal protein), ultimately diminishing catalytic efficiency. While sensitivity to surfactants may limit its direct application in detergent formulations, this feature is less critical for feed use.

Importantly, functional assays demonstrated the enzyme’s dual role in feed protein hydrolysis and detoxification. In this study, when compared to the control group (CK), the addition of the purified DL12 enzyme resulted in the highest relative increase in peptide content in cottonseed meal and the highest relative improvement in the degree of hydrolysis in cottonseed meal among the four substrates. Therefore, DL12 protease can effectively hydrolyze cottonseed meal, leading to increased peptide release and enhanced digestibility [37]. The hydrolysates also demonstrated antioxidant potential. Notably, peanut meal hydrolysate exhibited the strongest ABTS radical scavenging activity, highlighting the value of DL12 protease in producing functional feed peptides. In the antioxidant capacity assays, a discrepancy was observed between the ABTS and DPPH results. This phenomenon likely arises because the ABTS radical scavenging reaction occurs in an aqueous environment. Under these conditions, some hydrophobic peptides may aggregate, impairing their electron transfer capability and consequently reducing their ABTS radical scavenging capacity. Notably, compared to the enzyme-free control group (CK), the purified DL12 protease reduced the free gossypol content in cottonseed meal. This reduction is likely attributable to the formation of stable complexes between the protease and free gossypol, which in turn decreased the detectable levels of free gossypol in the system [2]. Specifically, the protease and its hydrolytic products are rich in active functional groups such as amino, sulfhydryl, and imidazole groups. These may form covalent or non-covalent bonds with the aldehyde groups and aromatic rings of gossypol, resulting in stable protein–gossypol complexes. Regarding the current methods for free gossypol degradation reported in the literature, Wu et al. utilized solid-state fermentation with Paenibacillus sp. and Cohnella xylanilytica and achieved a 23.33% reduction [38]. In contrast, Hu et al. employed fermentation with Paenibacillus sp. and B. amylolyticus and observed only a slight decrease [39]. The purified DL12 protease achieved a 70.18% degradation rate of free gossypol in cottonseed meal compared to the CK control. This performance significantly outperforms crude enzyme solution and the aforementioned reported microbial treatments and is comparable to the efficiency achieved by P. shawrang using 20 kGy gamma irradiation [40]. This clearly demonstrates its unique advantage in overcoming the long-standing limitations of cottonseed meal utilization.

Taken together, the results demonstrate that B. safensis DL12 and its alkaline protease represent a promising resource for feed biotechnology, with particular application in enhancing protein quality and mitigating anti-nutritional factors. Beyond its direct feed application, DL12 protease also enriches the microbial resource pool of alkaline proteases and provides a foundation for future protein engineering aimed at improving stability, catalytic efficiency, and broader industrial applicability.