Metabolomics aided by machine learning decodes adaptive remodeling of Bacillus biofilms in response to pasteurization stress

Abstract

The popularity of low-temperature dairy products is challenged by Bacillus species, whose heat-resistant spores and biofilms often survive pasteurization. Moreover, heat treatment can paradoxically enhance biofilm formation in some Bacillus spp., a phenomenon whose metabolic basis is not fully understood. Combining untargeted metabolomics with random forest analysis, we decoded the metabolic adaptations behind this heat-induced biofilm enhancement in raw milk Bacillus isolates. Our results demonstrate strain-specific mechanisms: in BC01, heat stress activated glutaminase, depleting L-glutamine and free histidine to relieve metabolic inhibition and activate biofilm genes, while reduced xanthosine promoted the biofilm-state transition. In BS01, metabolic network restructuring led to decreased synthesis of arginine, D-amino acid, dopamine, and arachidonic acid, thereby mitigating their known inhibitory effects on biofilm formation. This study clarifies the metabolic drivers of biofilm adaptation under heat stress, highlighting novel targets for metabolic intervention in dairy safety.

Introduction

Milk is an essential dietary source, providing abundant calcium and diverse nutrients that support bone health, regulate skin metabolism, and improve sleep quality¹. With the advancement of cold chain logistics, low-temperature dairy products, such as pasteurized milk, have gained substantial market share due to their ability to preserve nutritional value and provide product diversity. Importantly, these products are not completely sterile. Bacillus spp. present in raw milk can produce heat-resistant spores that survive pasteurization and subsequently colonize production environments as biofilms, posing a persistent contamination risk in dairy production^2,3. In the dairy industry, Bacillus biofilms are particularly problematic because they reduce heat transfer efficiency and increase potential food safety risks⁴.

Biofilms, protective structures formed by bacteria under stressful conditions, significantly enhance bacterial resistance and persistence in production environments, thereby posing a greater risk of food contamination⁵. Conventional pasteurization, though effective against Bacillus spore-forming cells, also generates sublethal stress that may trigger adaptive responses. Previous studies have shown that environmental stress can enhance bacterial biofilm formation. For example, subinhibitory concentrations of tetracycline promote biofilm synthesis in Salmonella typhimurium M37⁶; and oxygen stress enhances biofilm formation in Campylobacter⁷. Similarly, the effect of heat treatment on Bacillus biofilm formation is not solely inhibitory. This study found that the biofilm formation ability of some strains actually increased after pasteurization. These bacteria not only survive the sterilization process, but their progressive biofilm development may also lead to higher energy consumption during sterilization and increased spoilage risks in the dairy industry, underscoring the critical need for biofilm management.

Metabolic intervention to inhibit bacterial biofilm formation is an effective emerging strategy for controlling biofilm contamination. Previous studies have demonstrated that exogenous inhibitors can suppress biofilm synthesis by regulating the extracellular polysaccharide metabolic pathway. For instance, the antimicrobial peptide Citrocin inhibits Pseudomonas aeruginosa biofilm formation by degrading extracellular polysaccharides⁸. Similarly, the combination of ultrasound and rhamnolipids reduces the levels of extracellular polysaccharides and proteins in Bacillus cereus, thereby disrupting its biofilm formation⁹. However, despite transcriptomics and proteomics studies revealing the regulatory network of Bacillus biofilm formation^10,11, the adaptive flexibility of metabolic pathways under pasteurization stress remains poorly understood, particularly concerning the metabolic response mechanisms of high-risk strains. To address this gap, this study integrates non-targeted metabolomics with machine learning algorithms, employing random forest analysis to identify key metabolites involved in Bacillus biofilm formation and elucidate the potential molecular mechanisms underlying the metabolic remodeling of Bacillus biofilms induced by pasteurization. These findings provide new insights into the mechanisms of Bacillus biofilm formation and highlight potential targets for metabolic intervention to prevent biofilm contamination through metabolic perturbation.

Results

Changes in biofilm formation capacity of Bacillus spp. before and after pasteurization

Pasteurization treatment altered the biofilm-forming capacities of 14 Bacillus strains isolated from raw milk (Fig. 1). Under non-pasteurized conditions, all Bacillus spp. strains except BC07, BL01, and BP02 were capable of forming biofilms on polypropylene vinyl surfaces and exhibited adhesive capabilities. Bacillus subtilis and Bacillus licheniformis displayed weak biofilm formation, Bacillus B. cereus showed moderate adhesive capabilities, and the adhesion abilities of different B. subtilis strains varied significantly. Specifically, BS01 and BS02 demonstrated weak adhesion, whereas BS03 and BS04 exhibited strong adhesion. After pasteurization, strains exhibiting moderate adhesion lost this capability. Both B. subtilis and B. licheniformis completely lost their adhesion and became non-adherent. The biofilm-forming ability of most B. cereus and B. subtilis strains generally decreased; specifically, strains BC02, BC03, BC05, BC06, BC07, and BS03 showed a reduction in adhesion from strong or moderate to weak or non-adherent. Interestingly, certain strains, including BC01, BC04, BS01, BS03, and BS04, exhibited enhanced biofilm formation, increasing their adhesion from moderate or weak to strong. This finding suggests that pasteurization heat treatment may not inhibit and may even promote the biofilm formation of these bacteria. Considering that the above bacteria were isolated from different raw milk samples from the same region (with potentially similar or identical genotypes), we aimed to investigate the mechanism behind the pasteurization treatment’s promoting effect on biofilm formation in B. cereus and B. subtilis. We selected these strains for pulsed-field gel electrophoresis (PFGE) typing to identify and screen representative strains for subsequent studies.

Fig. 1. Biofilm formation capacity of 14 Bacillus strains before (left panel) and after (right panel) pasteurisation.

Biofilm formation was quantified via optical density measurement at 595 nm (OD₅₉₅), and the strains were categorized into four biofilm-forming levels (strong, medium, weak, and non-adherent) based on the OD₅₉₅ values (corresponding to red, yellow, blue, and green dots, respectively). Error bars represent the standard deviation of three independent experimental replicates.

Bacillus spp. isolated from raw milk were subjected to PFGE typing and cluster analysis. The Dice coefficient was used to the similarity between PFGE band types. The typing results indicated that the Bacillus group exhibited a high overall similarity, exceeding 90% in PFGE banding. Further analyses revealed that B. cereus and B. subtilis each formed two distinct PFGE band types. Notably, Bacillus strains with similar biofilm-forming abilities showed a perfect 100% genetic similarity in the clustering analysis (Supplementary Fig. 1). Specifically, PFGE profiles of strains BC02, BC03, BC05, BC06, and BC07 was identical, while BC01 and BC04 shared the same PFGE typing. Similarly, PFGE typing of BS01, BS03, and BS04 strains were identical. These results indicate that Bacillus strains with similar biofilm-forming abilities possess highly consistent genetic fingerprints. From each group of strains with identical PFGE banding, one bacterial strain was selected as a representative (BC01, BC02, BS01, BS02) for subsequent studies.

Changes in biofilm-forming capacity of Bacillus spp. in dairy matrices processed by heat treatment

A 304 stainless steel sheet, commonly used as a contact material in dairy processing equipment, served as the adhesion medium. Aseptic UHT-sterilized milk was used as the culture medium. The biofilm formation ability of four Bacillus species strains was assessed before and after pasteurization. The results indicated that under simulated conditions, the biofilm formation trends of Bacillus after pasteurization were consistent with those observed in BHI medium (Fig. 2). Specifically, the biofilm formation ability of strains BC01 and BS01 increased, transforming them from weakly adherent to strongly adherent strains. Conversely, the biofilm formation ability of strains BC02 and BS02 decreased, changing them from strongly/medium adherent to weakly adherent strains. The magnitude of changes in biofilm formation was more pronounced in milk than in BHI medium with polypropylene vinyl as the adhesion medium. After pasteurization, the biofilm formation trends of the same bacterial strains were consistent across different media and adhesion media, suggesting that heat treatment has a specific mechanism for regulating biofilm formation. To explore the potential reasons for the enhanced biofilm formation following heat treatment, controlled experiments with strains BC01, BC02, BS01, and BS02 were conducted in this study.

Fig. 2. Biofilm formation capacity of representative Bacillus strains (BC01, BC02, BS01, BS02) before (left panel) and after (right panel) simulated pasteurisation processing.

Biofilm formation was quantified by measuring the optical density at 595 nm (OD₅₉₅), and the strains wereclassified into four biofilm-forming levels (strong, medium, weak, and non-adherent) corresponding to red, yellow,blue, and the baseline (blank) dots, respectively. Error bars represent the standard deviation of three independentexperimental replicates.

Microstructure of Bacillus-milk component complexes

In order to visually demonstrate the microstructure of Bacillus biofilm on stainless steel plates after different heat treatments, we observed it using SEM. Overall, the simulated pasteurization heat treatment induced unique structural changes in the biofilm of B. subtilis (Fig. 3). Specifically, the unvaccinated blank control surface only showed sparse, amorphous milk protein and fat particles (white arrows), without biofilm structure, providing a clean background for subsequent observation (Fig. 3A). In the untreated control group, a relatively classical biofilm structure was formed. Rod shaped bacterial cells are embedded in a mesh matrix mainly composed of EPS (yellow arrow), scattered with dispersed milk protein particles (white arrow). After pasteurization heat treatment, the biofilm morphology of BC01 and BS01 underwent fundamental changes, and their biofilm formation ability was significantly enhanced. The most prominent feature is that the EPS matrix no longer exhibits a fibrous network structure, but rather fuses into dense, blocky, and irregularly shaped large aggregates (yellow arrows). These large matrix aggregates encapsulate more bacterial cells and milk protein components (white arrows), covering most of the field of view. This intuitively explains why CV staining showed an increase in the production of biofilm complexes between these two bacterial strains. This structure suggests that heat treatment may be an environmental stressor for these bacteria, triggering them to secrete a large amount of EPS and alter their structure, forming a more protective physical barrier. On the contrary, the EPS matrix of BC02 and BS02 no longer exhibited a fibrous network structure, but formed sparse, smaller block like complexes containing small amounts of Bacillus and milk components (compared to the control group). Heat treatment may weaken the ability of these bacteria to secrete EPS, thereby reducing the production of biofilm milk protein complexes.

Fig. 3. Representative scanning electron micrographs of the biofilm-milk fouling complexes formed by different Bacillus spp. before and after pasteurization.

A Sterile milk control; B B. cereus BC01; C B. cereus BC02; D B. subtilis BS01; E B. subtilis BS02. For each strain, the left panel shows the untreated control, and the right panel shows the condition after pasteurization. Yellow arrows indicate the mesh matrix composed of extracellular polymeric substances (EPS); white arrows indicate dispersed milk protein particles.

Changes in hydrophobicity of Bacillus spores under pasteurization treatments

The hydrophobicity of Bacillus spores is a key determinant of their adhesion capability and is typically quantified by their affinity for n-hexadecane. We compared the hydrophobicity of bacterial spores in their free state with those in biofilms before and after pasteurization (Table 1). Conventional wisdom suggests that increased spore hydrophobicity enhances biofilm formation. However, the results of this study contradict previous findings and reveal an unexpected pattern of association. Specifically, we observed that the hydrophobicity of spores from bacterial strains with enhanced biofilm formation ability (BC01 and BS01) decreased after pasteurization, whereas the hydrophobicity of spores from strains with reduced biofilm formation ability (BC02 and BS02) increased. Furthermore, this negative correlation was evident when comparing the hydrophobicity of spores in the free state with those in biofilms: strains with enhanced biofilm formation exhibited lower spore hydrophobicity within biofilms compared to free state spores, while strains with reduced biofilm formation showed the opposite trend. These findings indicate that changes in spore hydrophobicity are not directly responsible for the enhanced biofilm formation ability of the BC01 and BS01 strains post-pasteurization. Instead, we speculate that pasteurization-induced metabolic remodeling of Bacillus biofilms, significantly altering their metabolic pathways or metabolites, thereby affecting both biofilm formation ability and spore hydrophobicity.

Table 1.

Spore hydrophobicity of Bacillus after different pasteurization treatments

Name of the strain	Treatment	Cell state	A0	A1	RHS (%)
BC01	Control	free	0.87 ± 0.04	0.77 ± 0.02	11.49
BC01	Pasteurization	free	1.72 ± 0.05	1.51 ± 0.02	12.21
BC01	Pasteurization	biofilm	1.61 ± 0.04	1.46 ± 0.02	9.31
BC02	Control	free	0.73 ± 0.09	0.69 ± 0.02	5.47
BC02	Pasteurization	free	1.42 ± 0.05	1.18 ± 0.02	16.90
BC02	Pasteurization	biofilm	0.27 ± 0.04	0.21 ± 0.00	22.22
BS01	Control	free	1.77 ± 0.03	0.77 ± 0.01	56.49
BS01	Pasteurization	free	0.41 ± 0.02	0.23 ± 0.01	43.90
BS01	Pasteurization	biofilm	1.56 ± 0.06	1.35 ± 0.01	13.46
BS02	Control	free	0.81 ± 0.07	0.78 ± 0.03	3.70
BS02	Pasteurization	free	0.32 ± 0.03	0.17 ± 0.02	46.88
BS02	Pasteurization	biofilm	0.27 ± 0.19	0.07 ± 0.00	74.07

Metabolomics analysis: significant difference metabolite analysis

We analyzed the metabolomic changes in Bacillus species in response to pasteurization treatment and their effects on biofilm formation using LC-MS/MS-based untargeted metabolomics. PCA analysis and pairwise OPLS-DA comparisons were performed on samples of the same bacteria before and after pasteurization treatment. The results indicated that pasteurization treatments led to significant differences in bacterial metabolism. Differentially expressed metabolites (DEMs) for each bacterial strain between treatment groups were identified using an OPLS-DA VIP value > 1 and a p < 0.05 as screening criteria (Fig. 4). Specifically, for BC01, a total of 312 significantly DEMs were identified in the pasteurization-treated group compared to the untreated control group. These mainly included up-regulated lipids and lipid-like molecules, organic acids and derivatives, organoheterocyclic compounds, and benzenoid-related metabolites, while some homogeneous non-metal compounds were down-regulated. For B. cereus BC02, 247 significantly DEMs were identified following pasteurization treatment. The up-regulation trend was similar to that of BC01, but the down-regulated metabolites mainly involved homogeneous non-metal compounds. For B. subtilis BS01, 119 significantly DEMs were identified in the pasteurized treatment group compared to the untreated control. The main up-regulated metabolites were organoheterocyclic compounds, while the down-regulated metabolites included lipids and lipid-like molecules, organic acids, and derivatives. For BS02, 286 significant DEMs were identified in the pasteurization group, with pasteurization treatment significantly up regulating the expression of all the aforementioned categories of metabolites. Despite the enhanced biofilm formation ability of both BC01 and BS01 after pasteurization, their metabolic changes were significantly different, necessitating separate and in-depth analysis to understand the underlying mechanisms.

Fig. 4. KEGG pathway enrichment analysis of four Bacillus spp. strains (BC01, BC02, BS01, BS02) between their control groups (C) and pasteurization treatment groups (P).

Each subpanel corresponds to a strain-specific comparison: A BC01P vs BC01C, B BC02P vs BC02C, C BS01P vs BS01C, D BS02P vs BS02C. Enriched KEGG pathways are listed on the left of each subpanel; the horizontal axis denotes compound number, and bar length reflects the association extent of the pathway with the pasteurization treatment. The color gradient (red to yellow) indicates pathway enrichment p-values (darker red = smaller p-value, i.e., more significant enrichment), with individual p-values labeled for each pathway.

KEGG pathway enrichment analysis

To investigate the specific effects of pasteurization on the metabolic pathways of Bacillus spp., KEGG pathway analysis was employed (Fig. 5). In Bacillus, pasteurization induced significant changes in ABC transporter pathways, potentially representing a key mechanism enabling energy-efficient survival under heat stress. Besides the notable alterations in the ABC transporter protein metabolic pathway, the changes varied among different Bacillus species. In B. cereus, pasteurization induced significant alterations in 16 metabolic pathways in BC01, primarily involving protein digestion and absorption as well as aminoacyl-tRNA biosynthesis. In BC02, 9 metabolic pathways were significantly affected, primarily involving protein digestion and absorption and mineral absorption. For B. subtilis, pasteurization treatment caused significant changes in 11 metabolic pathways in BS01, with the most significant being the mTOR signaling pathway and the synaptic vesicle cycle. Conversely, pasteurization treatment led to significant changes in 20 metabolic pathways in BS02, with the most significant being the biosynthesis of amino acids, protein digestion and absorption, and the biosynthesis of valine, leucine, and isoleucine. Cross-species comparisons revealed that, although alterations in ABC transporter metabolism represent a common heat stress response in Bacillus, B. cereus relies more on protein metabolic reconfiguration, whereas B. subtilis adapts to heat stress primarily via enhanced amino acid metabolism and optimization of neurotransmitter-related pathways. This differentiation in metabolic strategies reflects the distinct survival strategies developed by different Bacillus species during evolution. In combination with the changes in biofilm formation in different Bacillus species post-pasteurization treatment, strains with enhanced biofilm formation due to heat treatment exhibited more significant optimization of signaling molecule-related metabolism. Metabolic remodeling plays a crucial role in the changes in bacterial biofilm formation capacity, and the potential regulatory mechanisms require further analysis based on DEMs in metabolic pathways.

Fig. 5. Random forest analysis of key metabolites in four Bacillus strains (BC01, BC02, BS01, BS02) between their pre-pasteurization control groups (C) and post-pasteurization treatment groups (P).

Each subpanel corresponds to a strain-specific comparison: A BC01C vs BC01P, B BC02C vs BC02P, C BS01C vs BS01P, D BS02C vs BS02P. MeanDecreaseAccuracy = metabolite importance for group differentiation (higher = more impactful). Color: red = metabolite upregulation; blue = metabolite downregulation (intensity = change magnitude). Each dot = individual metabolite (listed left).

Random forest analysis for key metabolite identification

A random forest analysis was employed to identify key metabolites in metabolic pathways that were significantly altered in Bacillus following pasteurization treatment. The top 15 features were selected as key metabolites to explore potential reasons for the enhanced biofilm formation capacity of Bacillus post-heat treatment (Fig. 6). Metabolite profiles of B. cereus (BC01 and BC02) and B. subtilis (BS01 and BS02) exhibited significant changes following pasteurization, which were closely associated with post-heat stress regulation and biofilm formation.

Fig. 6. Relative expression levels of biofilm-associated genes (Spo0A, SinR, TasA, EpsA) in Bacillus spp., quantified by the △△Ct method (housekeeping gene: 16S rRNA).

Blue bars = pre-pasteurization group; cyan bars = post-pasteurization group. Y-axis denotes relative expression level; error bars represent the standard deviation of experimental replicates.

In B. cereus, heat treatment resulted in the down-regulation of DL-proline levels across all strains, indicating a correlation with the bacterium’s heat stress regulatory mechanisms. In BC01, biofilm formation-related metabolites, including L-glutamine, histidine, xanthosine, betaine, L-carnitine, and salicylic acid, were significantly down-regulated following pasteurization. These changes suggest that the increased biofilm production by BC01 post-thermal excitation may be related to the reduced inhibitory effect of secondary metabolites on biofilm production, the regulation of bacterial community sensing, and enzyme activity modulation. Conversely, BC02 showed a distinct metabolomic response to heat treatment, characterized by significant down-regulation of key metabolites such as D-glucosamine, DL-glutamate, and butyric acid, which are linked to biofilm formation. These changes imply that the decreased biofilm production ability of BC02 post-thermal excitation might be associated with its diminished initial colonization ability, the down-regulation of key adhesion factors, and a reduction in biofilm substrates.

In B. subtilis, heat treatment induced a general down-regulation of raffinose levels, indicating a typical heat stress response in this bacterium. For BS01, metabolites such as arginine, dopamine, arachidonic acid, and leucine, which are associated with biofilm formation, were significantly down-regulated post-heat treatment. These changes suggest that the enhanced biofilm formation ability of BS01 after heat treatment may result from modulation of amino acid metabolism and a reduction in the inhibitory effects of secondary metabolites on biofilm production. Conversely, BS02 showed a distinct metabolomic response after heat treatment, characterized by significant up-regulation of several biofilm formation-related metabolites, including succinate, DL-tryptophan, α-ketoglutarate, orotate, and D-arabinose. These changes suggest that the altered biofilm production capacity of strain BS02 post-thermal excitation may be closely related to the energy deficit caused by TCA cycle disruption and the enhanced inhibitory effect of secondary metabolites on biofilm production.

In summary, Bacillus adjusts its biofilm-forming capacity after heat treatment through modulation of key mechanisms, including amino acid metabolism, sugar metabolism, enzyme activity, and energy metabolism. The identification of these key metabolites provides crucial insights into the biofilm formation mechanism of Bacillus and potential targets for preventing and controlling Bacillus biofilm hazards, meriting further in-depth research and discussion.

Analysis of biofilm-related gene expression changes

The relative expression of the Spo0A, sinR, TasA, and EpsA genes associated with biofilm production in Bacillus before and after pasteurization was examined using RT-qPCR (Fig. 6). Overall, the expression of Spo0A, TasA, and EpsA was up-regulated, while sinR was down-regulated in Bacillus strains BC01 and BS01, which showed increased biofilm production post-pasteurization. Conversely, in Bacillus strains BC02 and BS02, which exhibited decreased biofilm production post-pasteurization, SpoA, TasA, and EpsA were down-regulated, while sinR was up-regulated. These gene regulation patterns are closely related to the mechanisms of biofilm gene regulation in Bacillus. SinR is a transcriptional regulator whose DNA-binding activity inhibits the downstream epsA-O and tapA-sipW-tasA operons in the CSP quorum sensing system. The expression of sinR is negatively regulated by biofilm formation¹². In contrast, Spo0A, TasA, and EpsA, which are key transcription factors and genes involved in spore formation, biofilm matrix amyloid production, and extracellular polymer generation respectively, were positively regulated in relation to biofilm formation. The observed changes in gene expression following Bacillus pasteurization provide targets for further gene expression analysis to elucidate its metabolic regulatory mechanisms.

Discussion

Pasteurization kills most heat-intolerant microorganisms in milk while retaining as much of the nutrients and natural flavor as possible¹³. However, Bacillus species are difficult to completely eradicate during pasteurization due to their capacity to form heat-resistant spores⁵. The formation of biofilms by these bacteria on dairy production lines poses an ongoing threat to the quality and safety of dairy products, becoming a critical issue for the dairy industry⁶. Given that Bacillus species exhibit varying biofilm-forming potentials due to genomic and metabolite expression differences, this study focused on screening strains that displayed significant changes in biofilm formation after pasteurization. The aim was to investigate potential mechanisms underlying enhanced biofilm formation following heat treatment. Regarding factors influencing Bacillus biofilm formation, Zhou et al. found that both culture medium composition and adherence medium affect biofilm formation¹⁰. In this study, comparing experimental group A (BHI medium, polystyrene adhesive medium) with experimental group B (UHT milk medium, 306 adhesive medium), it was found that the trend in biofilm formation ability of the same Bacillus strain after heat treatment was consistent across different media and adhesive surfaces. The inherent physiological characteristics of the bacteria and metabolic changes post-heat treatment are the most critical factors influencing biofilm formation. Bacillus biofilm formation proceeds through four stages: initial adhesion, cell proliferation and multilayer structure formation, establishment of stable biofilm structures, and eventual biofilm detachment¹⁴. Spore hydrophobicity is widely recognized to facilitate the initial adhesion step¹⁵, and a positive correlation with biofilm formation is often reported. However, our study reveals a noteworthy divergence: heat treatment induced a negative correlation between spore hydrophobicity and biofilm formation capacity. This intriguing observation suggests that under heat stress, the influence of traditional physical properties like hydrophobicity may be overridden by more dominant biological processes. We propose two non-exclusive mechanisms to explain this phenomenon. First, heat stress may trigger such a profound metabolic reprogramming that its driving effect on biofilm development surpasses the contribution of hydrophobicity. This is strongly supported by our metabolomics data and the significant upregulation of key biofilm-related genes (spo0A, tasA, epsA) in strains BC01 and BS01. The resultant enhancement in EPS secretion could effectively compensate for the reduced adhesion potential of less hydrophobic spores, by serving as a copious hydrophilic ‘bio-glue’ that facilitates attachment and matrix formation. Second, these strains may employ an alternative adhesion strategy that is less dependent on inherent spore hydrophobicity, instead relying more on the prolific secretion of hydrophilic EPS for initial surface colonization. These hypotheses, centered on heat-induced metabolic remodeling as the primary driver of biofilm adaptation, were subsequently investigated and validated through our non-targeted metabolomics analysis and the construction of a predictive regulatory model, as detailed in the following sections.

Analysis of bacterial metabolic changes revealed that the KEGG pathways down-regulated by heat treatment in the two B. cereus strains included: ABC transporter proteins, protein digestion and absorption, aminoacyl-tRNA biosynthesis, mineral absorption, and phytohormone signal transduction. In both B. subtilis strains, the down-regulated KEGG metabolic pathways were ABC transporter proteins, protein digestion and uptake, and aminoacyl-tRNA biosynthesis. These findings significantly diverge from previous studies. ABC transporter proteins, responsible for ATP-driven translocation of substrates across membranes for transporting essential nutrients, have long been recognized as positive regulators of biofilm formation¹⁶. The upregulation of the ABC transporter protein pathway is considered more favorable for biofilm formation¹⁷. The aminoacyl-tRNA biosynthesis pathway, which transports amino acids for protein synthesis and ensures accurate transcription and translation of gene sequences to maintain intracellular homeostasis, is also deemed crucial for regulating bacterial biofilms¹⁸. Notably, pasteurization resulted in down-regulation of aminoacyl-tRNA biosynthesis and ABC transporter proteins in all Bacillus species, suggesting that heat treatment generally suppresses energy and amino acid metabolism in Bacillus. Comparing the energy and amino acid metabolism of bacteria in biofilm and planktonic states, Sadiq et al. found that these metabolic pathways are down-regulated in biofilm bacteria due to the relative scarcity of nutrients in the biofilm¹⁹. This phenomenon is attributed to the lower metabolic activity of biofilm bacteria compared to their planktonic counterparts²⁰. Importantly, this study focused on metabolic differences in the same biofilm bacteria before and after heat treatment, rather than comparisons between biofilm and planktonic states. Compared to biofilm cells in normal physiological conditions, heat-treated biofilm cells are more nutrient-deprived, which reduces metabolic activity. Pasteurization heat treatment reduced the biofilm metabolic activity of all Bacillus species, but the amount of biofilm produced by heat-treated BC01 and BS01 increased instead. The study found no correlation between the reduction in biofilm metabolic activity of some Bacillus spp. and the quantity of biofilm they produced.

Based on the key metabolites identified in the KEGG-enriched metabolic pathway and changes in gene expression related to biofilm formation, we elucidated the potential metabolic mechanisms underlying the alterations in biofilm formation in Bacillus following pasteurization heat treatment. Notable strain-level differences were observed in the metabolic regulatory mechanisms among B. cereus strains. Biofilm production was markedly increased in the BC01 strain after thermal excitation, a phenotypic change resulting from the synergistic effect of multiple metabolic and signaling regulatory mechanisms. Specifically, heat treatment modulated the biofilm-forming capacity of BC01 via three principal mechanisms. The first mechanism involved remodeling of the bacterial metabolite-enzyme system. The significant down-regulation of L-glutamine and histidine levels after heat excitation reflected the adaptive adjustment of bacterial metabolic strategies. L-glutamine, a key nutrient component of milk, possesses antibacterial and anti-inflammatory activities and inhibits Bacillus biofilm formation through its D-isomer^21,22. Glutaminase converts L-glutamine to L-glutamate, an important component of the biofilm substrate polyglutamic acid (PGA)²³. While glutaminase is typically produced by B. subtilis or B. licheniformis, Singh et al. demonstrated that certain B. cereus strains also produce this enzyme²⁴. In our study, the ability of BC01 to produce glutaminase was confirmed by the determination of the pga gene. Therefore, we hypothesized that the reduction of L-glutamine levels in the BC01 strain after pasteurization heat treatment may be related to heat-stimulated activation of glutaminase expression. BC01 may accelerate the conversion of L-glutamine to L-glutamate by heat-stimulated up-regulation of glutaminase synthesis, providing raw material for biofilm formation and reducing its metabolic inhibition of biofilm formation. Meanwhile, histidine kinase KinD plays a central regulatory role in Bacillus biofilm formation through a two-component signaling system²⁵. The signal transduction mechanism of this system relies on the phosphorylation modification of histidine residues, with the imidazole heterocyclic structure of histidine serving as a key molecular target for KinD’s catalytic activity²⁶. Heat-excited treatment may accelerate the phosphorylation modification of non-phosphorylated histidine by enhancing the phosphotransferase activity of KinD, leading to down-regulation of intracellular free histidine levels. The generated phosphorylation signals promoted the expression of biofilm-related genes eps and tasA through a cascade reaction (Fig. 7), ultimately enhancing the biofilm-generating capacity of the BC01 strain. Second, heat stress modulates the bacterial community sensing system. Cyclic di-guanosine monophosphate (c-di-GMP) acts as a bacterial second messenger, controlling transitions between planktonic and biofilm states²⁷. Environmental purines (e.g., xanthosine) reduce c-di-GMP in bacteria, inducing a transition from the biofilm state to the planktonic state²⁸. The down-regulation of xanthosine levels in the BC01 strain after heat excitation may have lifted its inhibition of c-di-GMP synthesis, increasing intracellular c-di-GMP levels and thereby activating the expression of biofilm-related genes (eps, tasA), enhancing the bacteria’s biofilm-forming ability. Finally, heat stress also attenuated the inhibitory effect on biofilm formation by down-regulating the synthesis of secondary metabolites. Secondary metabolites such as betaine, L-carnitine, and salicylic acid not only possess inhibitory activities but also inhibit the synthesis and assembly of the biofilm matrix by interfering with QS signals or altering cell membrane properties^29–31. The heat-excitation-induced reconfiguration of secondary metabolite profiles provided additional metabolic support for the increase in BC01 biofilm production.

Fig. 7. Metabolic regulation models of biofilm formation in four Bacillus strains (BC01, BC02, BS01, BS02) under pasteurization-induced heat stress.

Each subpanel corresponds to one strain; ↑ = upregulation, ↓ = downregulation. The models illustrate heat stress-mediated changes in metabolites, signaling pathways (e.g., TCS system) and biofilm-related components (e.g., EPS, γ-PGA) that regulate biofilm formation.

In contrast to the significant increase in biofilm production observed in the BC01 strain following heat excitation, the BC02 strain exhibited a marked decrease in biofilm production capacity under the same conditions. This difference is attributed to the multiple inhibitory effects of heat excitation on the substrates for biofilm synthesis and key adhesion factors. Firstly, heat stress caused a severe shortage of raw materials necessary for biofilm substrate synthesis. The synthesis of extracellular polysaccharides (EPS), a major structural component of bacterial biofilms, requires specific glycosaminoglycans and amino acids as precursors³². D-glucosamine and DL-glutamic acid have been identified by several studies as the core substrates for EPS biosynthesis in various bacteria³³. The levels of these two key metabolites were significantly down-regulated in the BC02 strain after thermal excitation treatment. Additionally, the expression of the eps gene was significantly down-regulated. This down-regulation of metabolites and gene expression indicates a limitation in EPS synthesis, a reduction in biofilm substrate production, and an inability to form effective spatial network structures to support bacterial colonization on solid surfaces. Secondly, heat stress reduced the induction of biofilm formation by milk substrates. Lipolysis of milk lipids releases free fatty acid butyric acid (BA), a molecule shown to be a key trigger for the formation of the tubular structure of Bacillus biofilms³⁴. Butyric acid promotes adhesive aggregation between bacteria and the construction of three-dimensional biofilm structures by regulating cell membrane fluidity and signal transduction pathways³⁴. However, heat excitation treatment led to a significant decrease in butyric acid concentration in the BC02 strain’s culture system, weakening its chemotactic effect on biofilm formation. The absence of this environmental signal, combined with the insufficiency of raw materials for EPS synthesis, further exacerbated the decline in biofilm generation capacity.

Similarly, significant strain-level differences were observed in the metabolic regulatory mechanisms of various B. subtilis strains. The substantial increase in biofilm production in the BS01 strain following thermal excitation was attributed to a combination of amino acid metabolism modulation and the attenuation of secondary metabolite inhibitory effects. Specifically, thermal excitation reshaped the biofilm-forming capacity of BS01 through a dual mechanism: first, it triggered an adaptive remodeling of the amino acid metabolic network, significantly downregulating arginine synthesis. The arginine kinase McsB and the ATPase ClpC form a complex that acts as a key negative regulator of biofilm formation in Bacillus spp. McsB modulates the activity of Arg protein substrates through phosphorylation, while ClpC, in synergy with the protease ClpP, participates in the degradation of unfolded proteins, thereby maintaining intracellular protein homeostasis³⁵. Recent studies have shown that the McsB-ClpC complex inhibits biofilm formation by repressing the expression of the transcriptional regulators Spo0A and SinI, which subsequently downregulates the transcription of the extracellular polysaccharide synthesis gene eps³⁶. This is consistent with the significant upregulation of spo0A and eps expression in BS01 observed after heat treatment in this study. Heat stress may reduce the negative regulatory effect of the McsB-ClpC complex by modulating amino acid metabolic pathways and decreasing arginine production. This alleviates the suppression of Spo0A and SinI expression, thereby activating eps gene transcription, promoting biofilm matrix synthesis and secretion, and ultimately enhancing BS01’s biofilm production capability. Secondly, heat stress also mitigated the inhibitory effect of D-amino acids on biofilm formation by regulating their metabolism. Amyloid fibrils in B. subtilis biofilms are crucial for maintaining the biofilm’s three-dimensional structural stability. However, a mixture of D-amino acids, such as D-leucine, D-methionine, D-tyrosine, and D-tryptophan, can trigger the depolymerization of amyloid fibrils, leading to biofilm structure disruption³⁷. Heat stress may provide additional metabolic support for increased biofilm production in strain BS01 by modulating amino acid metabolic pathways, reducing the production of D-amino acids like D-leucine, and diminishing their disruptive effects on biofilm structure. Finally, heat stress also reduced the inhibitory effect on biofilm formation by downregulating the synthesis of secondary metabolites. The downregulation of specific secondary metabolites, including dopamine and arachidonic acid, directly contributes to the enhanced biofilm formation in BS01 by attenuating their inherent anti-biofilm activities^38,39. Dopamine, a structural analog of quorum-sensing autoinducers, can infiltrate and disrupt established biofilm architectures by interfering with bacterial cell-to-cell communication³⁸. The observed downregulation of dopamine thus likely stabilizes cell-cell signaling and biofilm architecture by removing this inhibitory signal. Arachidonic acid has been demonstrated to suppress biofilm formation in pathogens like Streptococcus mutans by downregulating the transcription of key biofilm-associated genes³⁹. Its downregulation in BS01 presumably alleviates this transcriptional repression, thereby facilitating the expression of genes necessary for biofilm matrix production and assembly. The heat-excitation-induced reconfiguration of secondary metabolite profiles provided metabolic support for the increased biofilm production in BS01.

Under identical heat stress conditions, BS01 and BS02 strains exhibited markedly different biofilm formation responses: the former showed a significant increase in biofilm production, while the latter showed a significant decrease. The strain-specific difference may be attributed to two inhibitory effects: first, the impact of heat stress on the energy metabolism of strain BS02, and second, the alterations in secondary metabolite profiles that affect biofilm formation. Firstly, heat stress severely impaired the tricarboxylic acid (TCA) cycle function in the BS02 strain. Pasteurization treatment may cause the abnormal accumulation of metabolic intermediates α-ketoglutarate and succinate by disrupting the key enzyme activities in the TCA cycle, such as α-ketoglutarate dehydrogenase and succinate dehydrogenase. This metabolic blockage not only results in a crisis in the synthesis of energetic substances like ATP and NADPH, forcing bacteria to prioritize the allocation of limited metabolic resources to survival mechanisms such as heat shock proteins rather than to biofilm synthesis processes. Concurrently, the abnormal accumulation of metabolites also inhibits the QS system signaling pathway, as evidenced by the significant down-regulation of spo0A and up-regulation of sinR in the CSP QS system. Both of these factors synergistically down-regulate the expression of the extracellular polysaccharide synthesis gene eps and the amyloid synthesis gene tasA, ultimately weakening the biofilm construction ability of the bacteria. Secondly, heat stress-induced remodeling of secondary metabolite profiles further exacerbated the inhibitory effect on biofilm formation in BS02. Metabolites such as dL-tryptophan and D-arabinose, known for their anti-biofilm activities, were synthesized at significantly higher levels following heat stress treatment. dL-tryptophan inhibited initial biofilm adhesion by interfering with the expression of bacterial surface adhesins, while D-arabinose possibly inhibited initial biofilm adhesion by altering cell membrane charge properties or inhibiting the polymerization of extracellular polysaccharides, thus preventing stable deposition of biofilm substrates. The synergistic effect of these two metabolites, along with the energy metabolism crisis, constitutes the metabolic basis for the decreased biofilm formation capacity of strain BS02

In this study, we comprehensively investigated the effects of heat treatment on Bacillus spp. exhibiting enhanced biofilm formation capabilities. Using non-targeted metabolomics analysis and random forest analysis methods, combined with previous related studies, we identified the potential mechanisms underlying changes in biofilm formation capacity of Bacillus following heat treatment. Heat-treated biofilm cells exhibited greater nutrient deficiencies and consequently lower metabolic activity compared to biofilm cells under normal physiological conditions. In both BC01 and BS01, the increased biofilm formation was primarily attributed to the down-regulation of secondary metabolites after pasteurization, which reduced their inhibitory effect on biofilm production. Additionally, BC01 showed an increase in biofilm formation due to the regulation of enzyme activity, while BS01 showed an increase due to the remodeling of amino acid metabolism. This study elucidates the potential reasons for increased biofilm production in Bacillus following heat treatment and provides valuable insights into the prevention and control of biofilm-associated hazards based on metabolic perturbations.

Finally, this study was designed to investigate the potential mechanisms by which pasteurization heat treatment alters the biofilm-forming capacity of different Bacillus species on stainless steel surfaces in a milk matrix, potentially through metabolic perturbations. We explicitly state that the experimental focus was placed on revealing the metabolic impact of pasteurization on biofilm formation by Bacillus strains under controlled laboratory conditions, rather than fully simulating the dynamic and variable environments of actual production lines. Consequently, the experiments did not account for complex factors such as hydrodynamic conditions or transient temperature fluctuations in industrial settings, nor did they involve continuous monitoring of the dynamic progression of biofilms throughout their entire lifecycle. It is precisely the findings of this work that establish a solid foundation and provide clear direction for subsequent, more in-depth investigations that better approximate real-world conditions. Future research will build upon this basis by constructing dynamic reactor systems to simulate actual processing environments. Through systematic sampling at different time points, these studies will precisely characterize the complete dynamic process of Bacillus biofilm development, from initial adhesion and maturation to dispersal, thereby informing the development of more timely intervention strategies for enhanced hygienic control in the food industry.

Methods

Isolation and identification of Bacillus in raw milk

To thoroughly investigate the Bacillus spp. capable of generating biofilm and affecting the quality of dairy products during dairy processing, this study collected raw milk produced on the same day from 40 large-scale breeding farms in Shandong Province, a major raw milk-producing region in China. Mannitol yolk polymyxin agar (MYB) medium was used to screen for Bacillus in the raw milk. The isolated bacteria were identified through a combination of 16S rDNA sequencing and physiological and biochemical tests to determine specific species. Fourteen strains of Bacillus spp. were isolated from the 40 raw milk samples, including 7 strains of B. cereus, 4 strains of B. subtilis, 2 strains of Bacillus pumilus, and 1 strain of B. licheniformis. The Bacillus spp. were stored at −80 °C in TSB medium containing 20% glycerol for subsequent experiments.

Changes in biofilm formation capacity of Bacillus sp. before and after pasteurization heat treatment

Pasteurization heat treatment increases the survival pressure of Bacillus species. Bacillus generally enhances its environmental resistance by forming biofilms⁴⁰. Heat treatment may alter Bacillus biofilm formation, including accelerated development, structural optimization, and modulation of gene expression. To investigate the effect of pasteurization heat treatment on Bacillus biofilm formation capacity, we explored changes in 14 strains of Bacillus isolated from raw milk before and after pasteurization heat treatment. For untreated bacteria, the biofilm formation capacity was examined by diluting the Bacillus broth using BHI medium until the broth reached 0.5 McFarland turbidity to observe the biofilm formation capacity of the 14 strains under normal physiological conditions. For bacteria heat-treated by pasteurization, the diluted bacterial broth (0.5 McFarland turbidity) was placed in a 75 °C water bath, and the center temperature of the broth was monitored. The center temperature was maintained at 75 °C for 15 s to simulate the pasteurization heat treatment process. The biofilm formation capacity of the bacterial solutions subjected to pasteurization heat treatment was then examined to observe changes in biofilm formation capacity of the 14 Bacillus strains. For the biofilm assay, 200 μL of each bacterial suspension was added to a 96-well plate in six replicates. BHI medium served as a blank control. Plates were incubated at 30 °C for 24 h. After incubation, the BHI medium or bacterial solution was discarded, the plate was washed with sterile phosphate-buffered saline (PBS), and dried at room temperature for 30 min. Biofilms were fixed with 200 μL of methanol per well for 30 min, then stained with 1% crystal violet for 30 min. Wells were washed with PBS, dried, and the dye was solubilized with 95% ethanol. Optical density (OD) at 580 nm was measured using an enzyme-linked immunosorbent assay (ELISA) reader. The mean value of the optical density (OD) plus three standard deviations (3 SD) of the blank wells was designated as ODc. The Bacillus were classified into four categories: non-adhesive (OD ≤ ODc), weakly adherent (ODc < OD ≤ 2ODc), moderately adherent (2ODc < OD ≤ 4ODc), and strongly adherent (OD > 4ODc). In this study, Bacillus species were isolated from different raw milk samples from the same region, suggesting they may have similar or identical genotypes. To explore mechanisms underlying pasteurization-enhanced biofilm formation, PFGE typing was performed on strains showing significant biofilm changes pre- and post-treatment. The goal was to identify representative strains for subsequent experiments.

Changes in the biofilm formation ability of Bacillus spp. on stainless steel surface after simulated dairy pasteurization treatment

Replicating the industrial processing conditions of pasteurized milk under laboratory settings remains challenging. To bridge the gap between laboratory models and industrial reality, we designed a simulated processing system that incorporates the key environmental factors influencing biofilm formation in the dairy industry. Previous studies have confirmed that culture medium and adhesion medium are important factors affecting the ability of bacteria to form biofilms¹⁰. In dairy production, Bacillus proliferates in raw milk, which is pre-cooled and transported via cold-chain logistics to processing plants for pasteurization⁴¹. Bacillus typically adheres to the stainless steel surfaces of processing equipment, forming biofilms⁴². We used sterile UHT milk as the culture medium to simulate the nutrient matrix and 304 stainless steel coupons as the adhesion substrate, reflecting the most common material in dairy processing equipment. The experimental workflow simulated the entire post-harvest chain: inoculated milk was pre-cooled at 4 °C for 24 h, subjected to pasteurization (75 °C for 15 s in a water bath), and rapidly cooled to mimic industrial conditions. This approach provides a more industrially relevant context for studying Bacillus biofilm formation than standard laboratory media.

The detection of changes in Bacillus biofilm formation on 304 stainless steel followed the previously described procedure, with minor modifications. The primary difference was the adhesion medium. A sterile 304 stainless steel sheet, degreased with anhydrous ethanol, was added to each well of a 6-well plate. Each well was then filled with 3 mL of milk bacterial solution, with sterile UHT milk serving as a blank control. Plates were incubated at 30 °C for 24 h, after which the stainless steel sheets were rinsed three times with sterile PBS and air-dried at room temperature for 30 min. Biofilms on the stainless steel sheets were fixed by adding 3 mL of methanol per well for 30 min. The biofilms were then stained with 1% crystal violet for 30 min, washed with PBS, and air-dried. To dissolve the dye, 95% ethanol was added, and the OD was measured at 580 nm using an ELISA reader. It is important to note the specific interpretation of the CV signal in this complex matrix. While CV traditionally quantifies total biomass (cells and EPS) in laboratory media, it also undergoes non-specific binding to negatively charged milk components, such as casein and fat globule membranes, in our system. However, this “limitation” accurately reflects the industrial challenge, where biofilms do not exist in isolation but as composite “biofilm-milk residue complexes” (or milk fouling) on equipment surfaces. Therefore, the CV absorbance in this study represents the total adherent composite—the true source of microbial persistence and fouling in dairy processing.

Scanning electron microscopy (SEM)

In this study, the target of CV staining was the “biofilm milk component complex”, reflecting the total adhesive biomass produced by bacterial isolates on the surface of stainless steel. In order to visually confirm the presence of structured biofilms in the composite, we conducted supplementary scanning electron microscopy (SEM) studies under the same conditions. Using SEM, we verified that Bacillus strains established biofilms within milk matrices across different treatment groups. Moreover, we visualized the fine architecture of the Bacillus-induced “biofilm–milk component complex,” thereby bridging CV staining data to the practical hygiene implications of biofilm contamination. For SEM preparation, 304 stainless steel sheet from were washed gently three times with PBS to eliminate planktonic cells and residual medium. Fixation was performed overnight at 4 °C with 2.5% glutaraldehyde in PBS. After fixation, samples were rinsed three times in PBS (15 min per wash) to clear any fixative traces. Dehydration was then carried out in a graded ethanol series (50%, 70%, 80%, 90%, 95%, and two changes of 100%), allowing 15 min per concentration. Critical point drying was employed to dry the specimens thoroughly. Dried samples were mounted on stubs with conductive tape and coated with a roughly 10 nm thick gold layer via ion sputtering before being imaged under the scanning electron microscope.

Changes in spore hydrophobicity of Bacillus spp. in dairy products after pasteurization treatment

The hydrophobicity of Bacillus spores is strongly associated with their ability to adhere to surfaces. Highly hydrophobic spores are more likely to interact with hydrophobic surfaces (e.g., stainless steel, plastics), thereby initiating and promoting biofilm formation⁴³. Spore hydrophobicity is considered a key factor in assessing their adhesion potential⁴⁴. By comparing the changes in the hydrophobicity of Bacillus spores under different conditions, we can gain insight into the regulatory pathways of biofilm formation and bacterial adhesion. Therefore, in this study, the hydrophobicity of representative Bacillus spores was examined. Spore broth of the same bacterial strain in different states (free and biofilm) was prepared before and after pasteurization heat treatment. Following the method of Jindal et al., spore hydrophobicity was tested using their affinity for n-hexadecane⁴⁴. Specifically, the Bacillus spp. broth cultured in a 6-well plate was used as the free-state bacteria. The Bacillus biofilm formed on the surface of 304 stainless steel was scraped with a sterile cell spatula and suspended in PBS solution as the biofilm bacteria. The bacterial solution was incubated at 90 °C for 30 min to kill the vegetative cells. Subsequently, the treated broth was centrifuged at 12,000 g for 10 min at 4 °C, and the supernatant was the spore solution. 3.00 mL of spore solution was mixed with 0.60 mL of n-hexadecane, vortexed for 20 s, and left to stand for 10 min. The OD600 values before and after mixing were recorded as A0 (initial spore solution) and A1 (spore-hexadecane mixture), respectively. The hydrophobic spore (RHS) ratio was calculated as (A0-A1)/A0 × 100%.

Non-targeted metabolomics analysis

Non-targeted metabolomics analysis was conducted on biofilms formed by four selected Bacillus strains on stainless steel sheets. Each strain was divided into two treatment groups: one before and one after pasteurization treatment. The biofilm was scraped off the 304 stainless steel sheet using a sterile cell spatula, washed three times with sterile PBS buffer, and centrifuged at 12,000 rpm for 14 min at 4 °C. The cellular precipitate, representing the Bacillus biofilm metabolites, was frozen in liquid nitrogen for 14 min and then stored at −80 °C for subsequent processing. Six biological replicates were prepared for each treatment group for each bacterial strain. For sample preparation, the samples were thawed slowly at 4 °C. An appropriate amount of the thawed sample was then added to a pre-cooled methanol/acetonitrile/water solution (2:2:1, v/v). The mixture was vortexed, sonicated at low temperature for 30 minutes, left to stand at −20 °C for 10 min, centrifuged at 14,000 g for 20 min at 4 °C, and the supernatant was dried under vacuum. For mass spectrometry analysis, 100 μL of an acetonitrile/water solution (1:1, v/v) was added to re-dissolve the dried supernatant, which was then vortexed and centrifuged at 14,000 g for 14 min at 4 °C. The resulting supernatant was used for LC-MS/MS analysis.

The samples were analyzed using an Agilent 1290 Infinity LC ultra-performance liquid chromatography system coupled with an AB Triple TOF 6600 mass spectrometer. The metabolome samples were separated on a HILIC column at 25 °C with a flow rate of 0.5 mL/min and an injection volume of 2 μL. The mobile phase A was water with 25 mM ammonium acetate and 25 mM ammonia, and mobile phase B was acetonitrile. The elution gradient was as follows: 0–0.5 min, 95% B; 0.5–7 min, 95–65% B; 7–8 min, 65–40% B; 8–9 min, 40% B; 9–9.1 min, 40–95% B; 9.1–12 min, 95% B.

UPLC-MS/MS raw data were converted to .mzXML format using ProteoWizard, and peak alignment, retention time correction, and peak area extraction were performed using XCMS software. The extracted data were used to identify metabolites of Level 2 or higher using an in-house database from Shanghai Applied Protein Technology. DEMs were identified based on criteria of OPLS-DA VIP > 1 and an FDR-adjusted p-value < 0.05. DEMs identified from both positive and negative ion patterns were subjected to KEGG pathway annotation, mapping the DEMs to metabolic pathways. The random forest analysis was employed to identify key metabolites in the metabolic pathways that undergo significant changes in Bacillus after pasteurization treatment. The random forest model was built and evaluated using 10-fold cross-validation. The importance of metabolites was ranked based on the mean decrease in accuracy from the random forest model.

RT-qPCR validation of biofilm-related gene expression changes

The spo0A, sinR, epsA, and tasA genes are critical regulators of biofilm formation in Bacillus. These genes function within the spo0A–sinI–sinR regulatory circuit and participate in the synthesis of the TasA amyloid biofilm matrix^45,46. In this study, the relative expression levels of the spo0A, sinR, epsA, and tasA genes in four Bacillus strains were assessed using 16S rRNA as the internal reference gene, before and after pasteurization heat treatment. The primers for these genes were obtained from Yi et al.⁴⁷ mRNA was extracted from Bacillus biofilms on stainless steel sheets using the RNeasy® Plus Mini Kit (QIAGEN, Germany). RNA reverse transcription and RT-qPCR were conducted using the SYBR® PrimeScript™ PLUS RT-PCR Kit. Each reaction had a total volume of 20 μL, consisting of 10 μL 2× One Step SYBR RT-PCR Buffer, 0.4 μL PrimeScript PLUS RTase Mix, 1.2 μL TakaRa Ex Taq HS Mix, 5.4 μL RNase-free dH₂O, 0.4 μL ROX Reference Dye II, 1 μL RNA template, and 0.8 μL each of the forward and reverse primers. The annealing and extension temperature was set at 60 °C, with 40 cycles performed to measure relative gene expression changes.

Ethical approval

This article does not contain any studies with human participants or animals performed by any of the authors.

Author contributions

L.L. wrote the main manuscript text; L.L. and P.W. designed the research. L.L., P.W., X.Z., Z.W., and B.X. performed the research. L.L. and Q.J. analyzed the data. D.W. and Y.C. conducted project administration and funding acquisition. All authors reviewed the manuscript.

Data availability

All data generated or analysed during this study are included in this published article and its supplementary information files.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41538-026-00712-y.