Isolation of enzyme-producing microbial strains and their effects on the color and quality of tobacco stems Authors: Honglin Zhang, Jinghua Zhang, Tingju Liu, Jinhui Jiang, Chaojun Shi, Ze Liu, Lương Hùng Tiến, Hoàng Trung Tín, Jindan Wang, Neng Yang, Yinbiao Zhou, Ruilin Fu, Jiahao Jia, Yumei Dong, Jun Li, Tao Liu Published: 2025 ABSTRACT China is one of the world’s major tobacco producers, yielding more than 2.2 million tons of tobacco leaves annually, among which approximately 25–30% constitute by-products such as tobacco stems. Dark-colored stems are favored for their high contents of reducing and total sugars, porous structure, thermal stability, and rich aroma precursors; however, their limited availability constrains large-scale industrial utilization. In contrast, light-colored stems are frequently discarded because of inadequate natural aging, resulting in significant resource waste. To enhance stem utilization and improve cigarette quality, enzyme-producing strains were isolated from dark-colored stems (cellulase: C8, Z16; polyphenol oxidase: A59, Z1), and a synthetic microbial community (C8–Z1) was established to ferment light-colored stems. Amplicon-based high-throughput sequencing and UPLC-QTOF-MS non-targeted metabolomics were utilized to comprehensively characterize microbial succession, the dynamics of volatile aroma compounds, and their correlations with dominant genera during fermentation. The results demonstrated that microorganisms facilitated stem darkening and aroma enhancement through macromolecule degradation, enzymatic browning, and modulation of pigment metabolism. Key genera, including Bacillus, Aspergillus, and Filobasidium were identified as major contributors to aroma formation and color regulation. This study elucidates the microbial enzymatic mechanisms underlying the improvement of stem quality and offers theoretical as well as technical support for reducing the aging period and enhancing the industrial value of tobacco stems.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12870-025-07729-7. Background China is one of the world’s leading tobacco producers, yielding over 2.2 million tons of tobacco leaves annually [1], among which approximately 25–30% exist as by-products in the form of tobacco stems [2]. Tobacco stems are widely utilized in producing expanded stems, reconstituted tobacco, blended stems, and cigarette filters. As important fillers, tobacco stems contribute to reducing production costs, improving smoke characteristics, and enhancing cigarette quality [3–6]. After two years of natural aging, stems are categorized into light-, medium-, and dark-colored types, with color being closely associated with their chemical composition and sensory properties [7, 8]. Dark-colored stems are characterized by higher contents of reducing and total sugars, denser porous structures, greater thermal stability, and more abundant aroma compounds, which collectively enhance smoking quality when incorporated into cigarettes [9]. However, the production of dark-colored stems remains limited, and their utilization in cigarette blending is relatively low; meanwhile, most light-colored stems are discarded owing to inadequate aging, leading to substantial resource waste [10]. Therefore, elucidation of the mechanisms underlying stem darkening and quality enhancement is essential for improving stem utilization and overall cigarette quality, warranting systematic investigation. Color is a key indicator of the appearance, sensory attributes, and overall quality of tobacco leaves and stems, directly influencing aroma, strength, and aftertaste [11]. Aging, as a critical process for enhancing tobacco quality, depends on the synergistic interactions among enzymatic reactions, microbial metabolism, and chemical transformations, which degrade macro-molecules and increase small-molecule aroma compounds, thereby intensifying color and improving smoking properties [12]. Previous studies have demonstrated that microorganisms colonizing the leaf surface secrete diverse degradative enzymes, including cellulases, pectinases, and amylases, thereby increasing the levels of reducing sugars, amino acids, and polyphenols, which subsequently accelerate enzymatic browning reactions [13, 14]. In addition, the degradation of chlorophyll and the accumulation of carotenoids are major contributors to color darkening, accompanied by the formation of aroma precursors such as neophytadiene and ionone derivatives [15]. The degradation of membrane lipids further generates aldehydes and ketones, which participate in the evolution of color. Color progression is governed by multiple factors, including leaf variety, plastid pigments, polyphenol content, lipid composition, and environmental parameters such as temperature, light intensity, and oxygen availability [16, 17]. Although the microbial contribution to aging has been widely acknowledged, studies focusing on enzyme-mediated regulation of stem color and quality remain scarce [18]. Targeted screening of efficient enzyme-producing strains is therefore essential for elucidating the mechanisms underlying stem aging and for providing a theoretical foundation and technical guidance for improving stem quality. Cellulose constitutes the primary structural component of the cell walls in tobacco leaves and stems. Excessive cellulose content leads to coarse texture, inferior color, and increased smoking harshness. During the aging process, cellulose is efficiently degraded by microbial cellulases, thus enhancing leaf texture and color [12, 19]. Cellulases, comprising exo- and endo-β-glucanases as well as β-glucosidases, catalyze the hydrolysis of cellulose into monosaccharides. This degradation is frequently accompanied by the breakdown of hemicellulose and lignin, resulting in darker coloration and reduced brightness. The sugars, polyphenols, and other small molecules released from cell wall degradation subsequently participate in enzymatic browning, thereby contributing to uniform color development [20]. Microorganisms including Bacillus, Aspergillus terreus, and Trichoderma viride exhibit strong cellulase-producing abilities, serving as valuable microbial resources for enhancing tobacco quality [20–23]. Polyphenol oxidase, a key copper-containing oxidoreductase in tobacco leaves, catalyzes the oxidation of phenolic substrates—including chlorogenic acid and caffeic acid—into quinones. These quinones polymerize with sugars and amino acids to generate brown pigments, thereby promoting enzymatic browning, intensifying leaf coloration, reducing astringency, and enhancing overall quality. Polyphenol oxidase activity is typically elevated during the early stages of aging and stabilizes thereafter, influencing the color characteristics of tobacco throughout the aging process [24]. Several Polyphenol oxidase-producing microorganisms have also been isolated, offering a theoretical foundation for selecting functional strains aimed at improving tobacco quality [25]. Unaged tobacco leaves and stems are rich in macro-molecular substances—including cellulose, starch, proteins, and nicotine—which deteriorate combustibility and aroma quality while intensifying smoke harshness and elevating potential health risks [26–28]. With increasing concern over smoking-related health issues, an urgent need has arisen to enhance tobacco quality and minimize harmful constituents. Microbial fermentation, recognized as a green and sustainable aging strategy, plays a pivotal role in this process through the secretion of degradative enzymes such as cellulase and polyphenol oxidase. These enzymes catalyze the hydrolysis of macro-molecules into reducing sugars, amino acids, and fatty acids, thereby facilitating the formation of aroma-active metabolites (e.g., alcohols, ketones, and esters) and initiating Maillard reactions that enhance aroma, soften the taste, and alleviate smoke irritation [20, 29]. Previous studies have shown that microorganisms exhibiting strong degradative and aroma-producing activities—such as Bacillus, Bacillus amyloliquefaciens, Aspergillus, and various yeasts—can significantly enhance tobacco quality [12, 30, 31]. Therefore, the integration of functional microbial strains with omics-based approaches to explore microbial succession and aroma metabolism during tobacco stem fermentation is expected to yield valuable theoretical insights and technical support for improving stem quality and promoting industrial utilization. To address the issues of uneven coloration and inconsistent quality during the natural aging of tobacco stems, enzyme-producing strains (cellulase- and Polyphenol oxidase-producing) isolated from dark-colored stems were employed to artificially ferment light-colored stems. Amplicon sequencing and Ultra-Performance Liquid Chromatography–Quadrupole Time-of-Flight Mass Spectrometry (UPLC–QTOF–MS)-based untargeted metabolomics were utilized to systematically characterize microbial community dynamics, profile aroma-related volatile compounds, and determine their associations with dominant genera during fermentation. This study elucidates the intrinsic mechanisms underlying stem darkening and quality enhancement, providing a scientific basis for shortening the aging cycle and improving the utilization efficiency of tobacco stems. Materials and methods Experimental materials Tobacco stem materials of cultivar K326, which had undergone two years of natural aging, were supplied by China Tobacco Yunnan Industrial Co., Ltd. (Fig. S1). The samples were utilized for the isolation of enzyme-producing strains and for subsequent artificial fermentation experiments. All tobacco stem samples were stored at 4 °C prior to subsequent analyses. Isolation, screening, and identification of enzyme-producing strains Isolation of enzyme-producing strains Enzyme-producing strains were isolated through the standard dilution–plating technique. Crushed dark-colored tobacco stems were vortexed thoroughly and serially diluted, after which aliquots were spread onto Luria–Bertani (LB) and Potato Dextrose Agar (PDA) plates and incubated at 30 °C for 24–48 h. Distinct colonies were selected, purified through repeated streaking, and assigned unique identification numbers. Bacterial isolates were cultured on LB agar plates, whereas fungal isolates were inoculated on PDA plates. After three successive rounds of purification, the isolates were preserved in 50% (v/v) glycerol at –80 °C. All procedures were performed under strictly aseptic conditions. Screening of cellulase-producing strains Isolates were inoculated onto primary screening plates to assess cellulase activity and incubated at 30 °C for 4 days. Plates were subsequently stained with 1 mg/mL Congo red for 30 min and destained using a NaCl solution to visualize hydrolysis zones. The colony diameter (d) and hydrolysis zone diameter (D) were measured, and the enzyme activity was quantified by calculating the hydrolysis capacity index (Hc = D/d). Strains exhibiting higher Hc values were selected for secondary screening. For this purpose, isolates were cultivated in 100 mL of seed medium (LB for bacteria or PDB for fungi) at 28 °C and 200 rpm for 2–3 days, then transferred (20%, v/v) into secondary screening medium and incubated for 5 days. The culture supernatant was collected by centrifugation and served as the crude enzyme extract. Enzyme activity was measured using commercial assay kits (Beijing Boxbio Science & Technology Co., Ltd., China), Specifically, enzyme activities were measured following the manufacturer’s protocols with slight modifications to ensure reproducibility. The assay conditions were as follows: Crude enzyme extracts were prepared by centrifuging microbial cultures to collect cells. Samples were processed at a ratio of cell number (10,000 cells) to extraction buffer volume (mL) of 500–1000:1. Cells were disrupted by ultrasonic treatment in an ice bath (200 W, pulse 3 s, interval 10 s, repeated 30 times). The lysates were centrifuged at 8,000 g for 10 min at 4℃, and the supernatant was collected and kept on ice for enzyme activity assays. Enzyme activity was determined using a microplate reader (preheated for 30 min) set to 540 nm, following the manufacturer’s protocol (Beijing Boxbio Science & Technology Co., Ltd., China). Distilled water was used to set the zero baseline. For each assay, 200μL of reaction mixture was loaded into a 96-well plate or micro cuvette. Absorbance values were recorded as A1, A2, A3 and A4, Calculate:A5 = A1-A2, A6 = A3-A4, Note: Each measurement tube must include a control tube. Blank tubes need to be measured only 1–2 times. Standard Curve Preparation: Standard solutions at concentrations of 3.0, 2.0, 1.5, 1.0, 0.5, and 0.25 mg/mL were prepared. The concentration (x, mg/mL) was plotted on the x-axis and the corresponding A6 on the y-axis to construct a standard curve. The resulting linear regression equation was: y = kx + b, by substituting A5 into the equation, x (mg/mL) was obtained. This value was then used to calculate enzyme activity according to the appropriate formula provided by the assay kit(cellulase activity(U/104) = (233*x)/cell number). All experiments were performed in triplicate. Screening of polyphenol oxidase -producing strains Isolates were inoculated onto PDA plates supplemented with guaiacol and were incubated at 28 °C for 6–8 days. Strains that produced purple discoloration zones were selected. Colony diameter (d) and discoloration zone diameter (D) were measured, and the Hc values (D/d) were subsequently calculated to evaluate enzyme activity. Strains exhibiting larger Hc values and more intense discoloration were subjected to secondary screening. The selected isolates were cultured in 100 mL of liquid seed medium at 28 °C with shaking at 180 rpm for 4–5 days, and were then inoculated (10%, v/v) into 150 mL of secondary screening medium, which was incubated for 6–8 days. After centrifugation, the supernatant was collected as a crude enzyme extract, and the Polyphenol oxidase activity was determined using commercial assay kits. Specifically, enzyme activities were measured following the manufacturer’s protocols with slight modifications to ensure reproducibility. The assay conditions were as follows: Preparation of Crude and Inactivated Enzyme Extracts and Absorbance Measurement: Crude enzyme extracts were prepared by centrifuging microbial cultures to collect cells:Samples were processed at a ratio of cell number (10,000 cells) to extraction buffer volume (mL) of 500–1000:1. The collected cells were disrupted in an ice bath by ultrasonic treatment (power at 20% or 200 W, pulse 3 s, interval 10 s, repeated 30 times). The lysates were centrifuged at 8,000 g for 10 min at 4 °C, and the supernatant was collected and kept on ice until further analysis.The microplate reader was preheated for 30 min, and the wavelength was set to 410 nm. According to the manufacturer’s instructions, reagents were sequentially added for the assay.Preparation of Inactivated Enzyme Extracts: An appropriate amount of crude enzyme extract was heated at 95 °C for 5 min in a sealed container to prevent moisture loss. The sample was then cooled to room temperature and thoroughly mixed to obtain the inactivated enzyme extract. Absorbance Measurement: 200 μL of supernatant was transferred to a well of a 96-well plate. Absorbance was measured at 410 nm and recorded as A1and A2, The net absorbance was calculated as A3 = A1-A2, Note: Each sample was measured with a corresponding control tube to ensure accuracy. This value was then used to calculate enzyme activity according to the appropriate formula provided by the assay kit(Polyphenol oxidase(U/104 cell) = (120*A3)/cell number). All experiments were performed with three biological replicates. Identification of enzyme-producing strains Strains were inoculated onto LB or PDA plates and incubated at 30 °C for 2–3 days. Colony morphology, including size, shape, color, transparency, and texture, was recorded. Bacterial isolates were identified by Gram staining and observed under a LEICA DM2500 microscope: purple cells were recorded as Gram-positive, and red cells as Gram-negative. Fungal isolates were stained with lactophenol cotton blue (prepared from phenol, lactic acid, glycerol, distilled water, and cotton blue), and hyphal and spore structures were examined microscopically and photographed. Physiological and biochemical identification Physiological and biochemical properties of bacterial isolates were assessed using carbohydrate fermentation, methyl red (MR), Voges–Proskauer (V-P), starch hydrolysis, and catalase tests. Carbohydrate fermentation was evaluated in fermentation media with different substrates, and acid production was determined by indicator color change. MR and V-P tests were used to detect mixed acid fermentation and neutral metabolites, respectively. Starch hydrolysis was confirmed by iodine staining, and catalase activity was determined by the release of bubbles after exposure to hydrogen peroxide. All tests were conducted with three biological replicates. Molecular identification Genomic DNA was extracted from purified isolates. The ITS region of fungi was amplified using universal primers ITS1 (5’-TCCGTAGGTGAACCTGCGG-3’) and ITS4 (5’-TCCTCCGCTTATTGATATGC-3’). Bacterial 16S rDNA was amplified using primers 27 F (5’-AGAGTTTGATCCTGGCTCAG-3’) and 1492R (5’-TACGGCTACCTTGTTACGACTT-3’). PCR products were verified by 1% agarose gel electrophoresis and sequenced bidirectionally (Tsingke Biotechnology Co.,Ltd, Beijing, China). Sequences were compared with those in the NCBI database using BLAST, and phylogenetic trees were constructed with MEGA (v11.0) software to determine taxonomic status. Artificial fermentation of tobacco stems Antagonism test of enzyme-producing strains Strains with the highest enzyme activity were tested for compatibility using plate confrontation assays. Bacterial and fungal strains with different enzymatic functions were co-inoculated on PDA plates and incubated at 30 °C for 5–7 days. The presence of inhibition zones indicated antagonism, whereas their absence indicated compatibility. Strain combinations without antagonistic interactions were selected for artificial fermentation of tobacco stems. Preparation of microbial suspensions Bacterial suspensions: Target bacterial strains were activated 2–3 times on LB agar plates at 30 °C for 1–2 days. Single colonies were inoculated into 2 mL LB broth and cultured at 28 °C, 180 rpm for 24 h to obtain seed cultures. These were transferred (1% inoculum) into 500 mL LB broth and cultured at 28 °C, 200 rpm for 48 h. Cells were harvested by centrifugation (5000 rpm, 20 min, 4 °C), washed, and resuspended in sterile water to a final concentration of 4% (w/v). The suspension was adjusted to an Optical Density600(OD600) of 1–2 for fermentation experiments. Fungal suspensions: Fungal strains were cultured on PDA plates at 30 °C for 5–7 days and activated 2–3 times before inoculation into 500 mL PDB broth. Cultures were incubated at 28 °C, 200 rpm for 4 days. Mycelia and spores were collected by centrifugation (5000 rpm, 20 min, 4 °C), disrupted by ultrasonication, and resuspended in sterile water to a final concentration of 4% (w/v), yielding spore suspensions for subsequent fermentation. Artificial fermentation of tobacco stems Microbial suspensions were mixed according to the preset ratios (Table 1) and uniformly applied to light-colored tobacco stems using a 10 mL sterile sprayer at an inoculation rate of 30% (v/w). The treated samples were incubated in a constant temperature and humidity chamber at 30 °C and 85% relative humidity for 40 days. During fermentation, the samples were manually turned every 4 h, and changes in stem color and odor were recorded. At the end of fermentation, the samples were heated at 85 °C for 5 min to inactivate the microbial strains. The control group (CK) was treated with sterile water under the same fermentation conditions to serve as a negative control. All treatments were conducted in triplicate.Table 1Different ratios of bacterial solutions for different treatmentsTreatB: FCKSterile waterT11:0T20:1T31:1 Table 1Different ratios of bacterial solutions for different treatmentsTreatB: FCKSterile waterT11:0T20:1T31:1 Different ratios of bacterial solutions for different treatments Determination of conventional chemical components in fermented tobacco stems After artificial fermentation, tobacco stem samples were collected and dried in an oven at 60 °C for 2 h until the moisture content reached 10–15%. The dried samples were ground into powder, passed through a 60-mesh sieve (0.25 mm), and labeled. The contents of total sugar, reducing sugar, potassium (K), chloride (Cl), total nitrogen, total alkaloids, and pectin were then determined. UPLC-QTOF-MS non-targeted metabolomics analysis of fermented tobacco stems Headspace solid-phase extraction combined with UPLC-QTOF-MS non-targeted metabolomics was used to detect the main volatile compounds affecting the quality and color changes of tobacco stems. Differential volatile compounds between CK and treated samples were analyzed to identify key compounds and metabolic pathways associated with color, aroma, and flavor changes. Sample preparation and instrumental analysis were performed by Wekemo TechGroup Co., Ltd. Shenzhen China.The specific chromatographic and mass spectrometric conditions used in this study are described as follows: Analyses were carried out using a LECO Pegasus® 4D instrument (LECO, St. Joseph, MI, USA) consisting of an Agilent 8890 A GC (Agilent Technologies, Palo Alto, CA, USA) system equipped with a split/splitless injector, and dual stage cryogenic modulator (LECO) coupled with TOFMS detector (LECO). The separation system was a DB-Heavy Wax (30 m × 250 μm × 0.5 μm) (Agilent, USA) column. The carrier gas was high purity helium (> 99.999%) at a constant flow rate of 1.0mL/min. The temperature program of the oven was as followed: the oven temperature was held at 50℃ for 2 min at first, and increased to 240 °C at a rate of 5°C/min, held for 5 min. Flavor substance was performed on LECO Pegasus BT 4D.The transfer line and TOF MS ion source temperature were set at 250℃ and 250℃, respectively. The acquisition frequency was 10 spectra/s. The mass spectrometer was operated in the EI mode at 70 eV using a range of m/z 35–550 and the detector voltage of 2035 V. Microbial community analysis of fermented tobacco stems The 16S rRNA and ITS amplicons of microorganisms in fermented tobacco stems were sequenced. DNA extraction, PCR amplification, product purification, library preparation and quality control, and sequencing were all carried out by Wekemo TechGroup Co., Ltd. Shenzhen China. Correlation analysis between dominant microbial genera and volatile compounds Based on an Orthogonal Partial Least Squares-Discriminant Analysis (OPLS-DA) model, the top 20 most abundant bacterial and fungal genera were selected. Spearman correlation analysis was used to construct clustered heatmaps showing the relationships between dominant genera and volatile compounds. Data analysis Experimental data were processed and analyzed using Excel 2019 and Word 2019. Statistical significance was assessed using SPSS Statistics 27(IBM, USA), data were expressed as mean ± standard deviation (SD) of at least three independent biological replicates. Significance was determined at P < 0.05 using analysis of variance (ANOVA) followed by least significant difference(LSD)and Duncan’s multiple range test. Phylogenetic trees were constructed with MEGA 11.0. Figures were generated using Origin 2022. Principal component analysis (PCA) was performed with SIMCA 2022. Spearman correlation coefficients and P-values were calculated using R software (v3.6.2). Results Isolation of enzyme-producing strains Using the plate dilution method, a total of 27 bacterial strains representing seven genera were isolated from dark-colored tobacco stems (Table S1), with Bacillus identified as the dominant genus. In addition, 9 fungal strains representing four genera were obtained (Table S2), among which Penicillium species were found to be the most prevalent. Screening of enzyme-producing strains Screening for cellulase-producing strains The isolated strains were inoculated onto primary screening media to assess cellulase activity. Among them, 12 bacterial strains and 8 fungal strains produced visible clear zones on the plates. The Hc values (D/d) and corresponding enzyme activities are presented in Tables S3. During bacterial screening, strain D12 exhibited the highest Hc value (3.66), which was significantly higher than those of the other strains; strains M1 and D9 also displayed relatively high Hc values. Strains C8, A55, M33, B53, and B47 did not differ significantly. Among the fungal strains, Z16 exhibited the highest Hc value (2.22), which was significantly higher than those of the remaining strains, followed by Z10. Strains Z39, Z4, Z31, Z59, Z3, and Z45 displayed lower Hc values or differences that were not statistically significant. In shake-flask rescreening, strain C8 exhibited the highest cellulase activity (147.55 U/mL), which significantly exceeded that of the other bacterial strains. Strain D12 also demonstrated high activity, ranking second only to C8. There were no significant differences in the activities of M1, M33, and B53, whereas H3, D9, A55, and B47 exhibited lower activities that were not statistically significant. Among the fungal strains, Z16 exhibited the highest activity (193.27 U/mL), which did not differ significantly from Z59 but was significantly higher than that of the other strains. Strains Z10 and Z39 exhibited similar activities, both of which were higher than those of Z39, Z31, Z45and Z1, whereas Z45 and Z31 did not differ significantly. Overall, strains C8 and Z16 exhibited the highest cellulase production, suggesting substantial application potential. Screening for polyphenol oxidase producing strains For Polyphenol oxidase screening, the isolated strains were inoculated onto primary screening media supplemented with guaiacol. Five bacterial strains and two fungal strains formed visible purple halos, with the coloration intensity and enzyme activity presented in Table S4. Strain D9 exhibited the most intense coloration, followed by A48 and B53, whereas R6 displayed the weakest coloration. In shake-flask rescreening, strain A59 exhibited the highest enzyme activity (50.8 U/mL), followed by D9 (28.4 U/mL) and A48 (21.2 U/mL), whereas B53 displayed the lowest activity (16.0 U/mL). Among the fungal strains, both produced halos, with Z1 exhibiting a higher Hc value and more intense coloration than Z10 (Table S4-1). Rescreening confirmed that Z1 exhibited the highest Polyphenol oxidase activity (111.2 U/mL), which was significantly higher than that of Z10. Therefore, strains A59 and Z1 were selected as the most efficient Polyphenol oxidase-producing bacterial and fungal strains, respectively, indicating substantial potential for application. Compatibility testing of enzyme-producing strains Four high-activity strains were selected, including the cellulase-producing C8 (bacterium) and Z16 (fungus), as well as the Polyphenol oxidase-producing A59 (bacterium) and Z1 (fungus). Based on their functional characteristics, three bacterial–fungal combinations were evaluated for antagonistic interactions (Table S5). The C8–Z31 combination exhibited significant antagonism, demonstrating the strongest inhibitory effect. The A59–Z31 combination also displayed antagonistic activity, albeit weaker. No antagonistic effect was observed in the C8–Z1 combination. Among the four highly active strains (C8, Z16, A59, and Z1), strains C8 and Z1 were selected for further fermentation experiments due to their lack of antagonistic effects and comparatively higher enzyme activities.(Table S3-5). Therefore, the C8–Z1 combination was selected for artificial fermentation of tobacco stems, followed by subsequent functional strain screening and identification. Identification of enzyme-producing strains Morphological identification To observe morphological characteristics and staining results, strain C8 was streaked on LB agar plates three times and incubated at 30 °C for 48 h (Fig. 1). C8 grew well, forming pale yellow, opaque, and glossy colonies with irregular edges, a moist surface, non-sticky texture, and a distinctive odor; the colonies were easily picked and exhibited rapid growth (Fig. 1a). Gram staining revealed purple, rod-shaped cells, indicating that C8 is a Gram-positive bacterium (Fig. 1b).Strain Z1 was inoculated onto PDA plates and purified three times, followed by incubation at 30 °C for 5–7 days. During the early growth stage, colonies were orange-yellow with a pale-yellow substrate; hyphae extended inward, the central region was sparse, and the edges were dense. At later stages, colonies turned dark green to brown, accompanied by pigment production (Fig. 1c, d). Microscopic observation revealed lemon-shaped or oval, brown ascospores with a stalk, containing eight spores, clustered with terminal protrusions, and easily disintegrated (Fig. 1e). The ascus wall was spherical or oval, densely covered with unbranched, wavy setae, with.Fig. 1Morphological characteristics of strains C8 (a–b) and Z1 (c–f). a Colony morphology of C8 on agar plate; b Gram staining of C8 cells; c Front view of Z1 colony on agar plate; d Reverse view of Z1 colony; e: Ascospores of Z1; f: Ascomata (perithecia) of Z1 Morphological characteristics of strains C8 (a–b) and Z1 (c–f). a Colony morphology of C8 on agar plate; b Gram staining of C8 cells; c Front view of Z1 colony on agar plate; d Reverse view of Z1 colony; e: Ascospores of Z1; f: Ascomata (perithecia) of Z1 ascospores accumulated around it (Fig. 1f). Physiological and biochemical identification Purified strain C8 was subjected to multiple physiological and biochemical tests, including sugar and alcohol fermentation (glucose, sucrose, maltose, galactose, mannitol, and sorbitol), methyl red (MR), Voges–Proskauer (V–P), starch hydrolysis, and a 3% catalase assay (Table S6). Strain C8 exhibited positive fermentation for all six carbon sources, indicating a strong carbon utilization capacity and good adaptability. Both MR and V–P tests were positive, suggesting the production of acidic metabolites, including formic acid and lactic acid. Starch hydrolysis was positive, demonstrating the strain’s capacity to degrade macro-molecular starch. The catalase test was negative, indicating the absence of catalase production and a propensity for growth under anaerobic conditions. Overall, the physiological and biochemical characteristics of strain C8 were consistent with the typical features of the genus Bacillus. Molecular identification The 16S rRNA gene of strain C8 was amplified, sequenced, and analyzed using the NCBI BLAST database (Fig. S2a; the original full-length gel image is provided in Fig. S3). The results indicated the highest sequence similarity (99.79%) to B. amyloliquefaciens (Accession No. NR_116022.1). Phylogenetic analysis conducted using MEGA 11.0 indicated that C8 clustered closely with B. amyloliquefaciens strain BCRC 11601 (NR_116022.1), confirming a close genetic relationship. Combined with morphological and physiological characteristics, strain C8 was therefore identified as B. amyloliquefaciens. For strain Z1, the ITS region was amplified and sequenced, and BLAST analysis in the NCBI database indicated the highest similarity (99.28%) to C. nepalense (Accession No. NR_159782.1) (Fig. S2b; the original full-length gel image is provided in Fig. S4). Phylogenetic tree analysis indicated that Z1 clustered with C. nepalense CBS 288.83, exhibiting the closest phylogenetic relationship and 96% sequence homology. Based on morphological features, strain. Z1 was identified as C. nepalense. Analysis of conventional chemical components in post-fermentation tobacco stems To evaluate the impact of fermentation on tobacco stem quality, the contents of total sugar, reducing sugar, potassium (K), chloride (Cl), total nitrogen (TN), total alkaloids, and pectin were determined under different treatments (Fig. 2). As shown in Fig. 2, inoculation with functional strains significantly altered the basic chemical composition of tobacco stems. Compared with the ck, the total sugar and reducing sugar contents were markedly increased in all treatment groups, indicating that microbial enzymatic activities accelerated polysaccharide degradation and saccharification, thereby enhancing soluble sugar accumulation. The potassium content was also higher than that of the CK, possibly due to microbial secretion of organic acids and extracellular enzymes that promoted ion exchange and the release of cell wall–bound potassium. In contrast, the pectin content significantly decreased, suggesting that microbial pectinase activity contributed to cell wall depolymerization, which could facilitate the formation and release of aroma precursors. However, no significant differences were observed in total nitrogen, chlorine, or total alkaloid contents, implying limited microbial involvement in nitrogen and chlorine containing compound metabolism. Overall, the C8–Z1 fermentation system improved the chemical composition of tobacco stems mainly by enhancing saccharification, promoting mineral element release, and degrading pectin, thereby providing a biological basis for quality improvement.Fig. 2Conventional chemical content of tobacco stem in different treatments(p < 0.05, n = 3) Conventional chemical content of tobacco stem in different treatments(p < 0.05, n = 3) Analysis of volatile compounds in post-fermentation tobacco stems Principal component analysis (PCA) of volatile compounds in fermented tobacco stems revealed clear differences between the CK and treatment groups (T1, T2, T3), with distinct separation among samples (Fig. S5). After quality assessment and control of the raw metabolomic data, samples exhibited good clustering, demonstrating high reproducibility and stability (Fig. S5a). The first two principal components explained 48.65% of the total variance, indicating that the different treatment groups could be effectively distinguished within the 95% confidence interval (Fig. S5b). To evaluate the effects of microbial fermentation on the composition of volatile metabolites in tobacco stems, a non-targeted metabolomic analysis was performed using UPLC-QTOF-MS. The relative abundances of the major metabolites in each treatment group were visualized using bar plots (Fig. 3a). In the CK, ethanol and acetic acid were identified as the predominant volatile components. Following microbial fermentation, the metabolite profiles in T1–T3 treatments became increasingly complex, with the appearance of multiple characteristic compounds. Notably, nitrosomethane, 2,6-dimethylnonane, and propylene glycol exhibited specific increases in the treated groups, with nitrosomethane content markedly elevated in T3, suggesting that higher inoculum or prolonged fermentation may activate specific metabolic pathways. Additionally, compared with CK, the relative proportions of ethanol and acetic acid decreased in the treated groups, suggesting that microbial treatment promoted the transformation of volatile compounds, which may reduce irritation and improve sensory quality. A total of 72 significantly differential volatile compounds were identified (Fig. 3b), which are likely key contributors to alterations in color and aroma. These compounds were classified into alkanes (24), esters (10), ketones (7), alcohols (5), aldehydes (6), acids (3), alkenes (5), phenols (3), ethers (3), heterocycles (4), and terpenes (2). Among these compounds, two were associated with terpene metabolism, one with phenylalanine degradation, and seven with Maillard reaction products.Fig. 3Relative abundance distribution of volatile metabolites (a) and distribution of PLS-DA differential volatile compounds (b) in tobacco stem of different treatment groups Relative abundance distribution of volatile metabolites (a) and distribution of PLS-DA differential volatile compounds (b) in tobacco stem of different treatment groups To further investigate the key differential metabolites among treatments, the top 30 metabolites with the largest differences were subjected to hierarchical clustering and visualized as a heatmap (Fig. 4). Figure 4 showed distinct differences in volatile compound profiles between treatments, particularly in T2, which exhibited significantly higher levels of Maillard reaction products and terpenoid metabolites compared to CK. These changes likely result from the metabolic activity of the inoculated strain during fermentation, which can produce enzymes and metabolites that alter substrate composition and promote the formation of flavor-related compounds. For example, the substantial increase in 2-pentylfuran (up to 243.4% compared to CK) and ethyl acetate (up to 166.5% compared to CK suggests enhanced lipid oxidation and esterification processes, consistent with microbial fermentation pathways reported in previous studies [32–34]. The higher content of nerolidol in T2 further supports this, as terpenoids are known for their aromatic properties and are often produced or enhanced by microbial metabolism. Nerolidol has a sweet aroma and may act as a natural flavor enhancer and color regulator in tobacco stems, indicating a potential role of the inoculated strain in improving product quality. Additionally, the elevated levels of 6-ethyl-5,6-dihydro-2H-pyran-2-one and 4-(hydroxymethyl)-dihydrofuran-2(3H)-one in T2 suggest that the strain could influence the synthesis of key ketone aroma compounds, which are important for sensory quality. The highest nicotine content observed in T2 further implies that the microbial fermentation process may affect alkaloid metabolism, possibly by altering precursor availability or activating specific metabolic pathways. Overall, these results suggest that inoculation with the tested strain not only changes the chemical composition of tobacco stems but also enhances aromatic quality, supporting its potential application in tobacco fermentation and flavor improvement.Fig. 4Cluster analysis of key differential volatile compounds across treatments Cluster analysis of key differential volatile compounds across treatments Volatile compounds constitute the primary chemical basis determining the quality of tobacco stems and play a crucial role in their visual color. Using non-targeted metabolomics via UPLC-QTOF-MS, a total of 360 metabolites were detected across the various treatment groups. Based on the selection criteria (Similarity ≥ 800, SN ≥ 50, Quant_SN ≥ 50, Area_percent ≥ 0.001), 136 differential volatile compounds were ultimately identified (Fig. S6). These compounds were classified into ten chemical classes: alkanes (24), benzenes and their derivatives (16), carboxylic acids (8), alcohols (2), aromatics (12), esters (1), phenols (4), lipids (16), pyridines (7), and others (43). Among the aroma-active compounds, lipids accounted for the largest proportion (12%), followed by aromatics (9%), whereas esters contributed the least (0.8%). The top 20 high-abundance volatile compounds in each sample are presented in Fig. 5. Notably, neo-plant dienes exhibited the highest content in T1, being 7.23-, 2.03-, and 4.19-fold higher than in CK, T2, and T3, respectively, although the difference with T2 was not statistically significant. The Maillard reaction product 2-acetylpyrrole was significantly enriched in T2, with levels 49.8-, 4.96-, and 5.29-fold higher than in CK, T1, and T3, respectively. Stimulatory compounds, including acetic acid and ethanol, were most abundant in CK. In contrast, microbial fermentation markedly increased aroma-active compounds, including isopropanol, 1-methyl-2-(3-pyridyl)pyrrolidine, γ-butyrolactone, and o-methoxyphenol, indicating that fermentation treatment effectively enhanced both the aromatic quality and sensory properties of tobacco stems.Fig. 5Top 20 volatile compounds in abundance for different treatments Top 20 volatile compounds in abundance for different treatments To further elucidate the biosynthetic mechanisms of key volatile metabolites during tobacco stem fermentation, differential volatile compounds were subjected to Kyoto Encyclopedia of Genes and Genomes(KEGG) pathway enrichment analysis (Fig. 6). Figure 6 shows the enrichment overview of metabolic pathways associated with differential volatile metabolites during tobacco stem fermentation. Seven key pathways were identified, with degradation of aromatic compounds, microbial metabolism in diverse environments, and sulfur metabolism showing high statistical significance (P < 0.01), though their fold enrichment was relatively low. These pathways suggest that the inoculated microbial strains enhanced the degradation of complex aromatic structures and improved sulfur transformation, which can help reduce undesirable smoky or harsh odors. Moderately enriched pathways included toluene degradation and monoterpenoid biosynthesis (P < 0.05), both of which are closely related to the generation of terpenoid-derived aroma compounds that contribute to pleasant floral and fruity notes. Moreover, aminobenzoate degradation and furfural degradation showed the highest fold enrichment, particularly the latter, indicating an intensive microbial transformation of aldehydes and furans. This process likely decreases irritating volatiles while promoting the formation of sweet and caramel-like aromas.Fig. 6Significant metabolic pathway analysis for Over-Representation Analysis(ORA) enrichment Significant metabolic pathway analysis for Over-Representation Analysis(ORA) enrichment Collectively, these results demonstrate that the inoculated strains actively regulate multiple metabolic pathways related to aromatic compound degradation and terpenoid biosynthesis, ultimately enhancing the aroma profile and color stability of fermented tobacco stems. Microbial community analysis of fermented tobacco stems To assess the impact of fermentation on microbial communities in tobacco stems, bacterial and fungal Operational Taxonomic Unit(OTU) distributions were analyzed, and Venn diagrams were constructed (Fig. S7). In total, 1,525 bacterial and 431 fungal OTUs were detected. Only 39 bacterial and 9 fungal OTUs were shared among CK, T1, T2, and T3, indicating limited overlap. CK contained the most unique bacterial OTUs (937), far exceeding T1 (76), T2 (186), and T3 (142), suggesting higher bacterial diversity in unfermented samples. Fermentation reduced unique taxa and led to convergence of bacterial communities. Fungal communities showed greater differentiation: CK (136) and T1 (127) had far more unique OTUs than T2 (33) and T3 (92). Only 9 fungal OTUs were common to all groups, implying strong selective effects of fermentation on fungal composition. Alpha-diversity analysis of bacterial communities (Table 2) revealed Coverage indices exceeding 99% in all samples, indicating that the sequencing depth was sufficient to capture the majority of microbial taxa. Chao1, ACE, Shannon, and Simpson indices exhibited an initial increase followed by a decline across treatments, reflecting dynamic alterations in species richness and diversity. CK consistently exhibited the highest diversity and richness, representing the most complex bacterial community structure. T1 exhibited the lowest diversity indices, suggesting that early-stage fermentation exerted an inhibitory effect on bacterial populations. Diversity and richness slightly recovered in T2, implying potential microbial restoration or replacement, whereas T3 displayed a further decline, indicating a shift in community composition toward specific taxa. Statistical analysis revealed that, compared with CK, all fermentation treatments significantly reduced bacterial diversity and richness (P < 0.05), suggesting that fermentation mediated by enzyme-producing strains markedly altered the bacterial community, likely through nutrient competition, metabolic inhibition, or environmental modulation, thereby impacting the microbial ecosystem of tobacco stems.Table 2Analysis of bacterial alpha diversityTreatmentShannon indexChao1 indexSimpson indexAce indexCoverage indexCK5.99 ± 0.45a533.38 ± 41.42a0.94 ± 0.03a533.50 ± 41.42a99.33%T10.30 ± 0.13c69.67 ± 30.90c0.05 ± 0.02c69.67 ± 30.90c100%T21.18 ± 0.31b140.33 ± 27.43b0.24 ± 0.07b140.40 ± 27.50b99.67%T30.36 ± 0.19c89.00 ± 28.16bc0.06 ± 0.04c89.00 ± 28.16bc100% Table 2Analysis of bacterial alpha diversityTreatmentShannon indexChao1 indexSimpson indexAce indexCoverage indexCK5.99 ± 0.45a533.38 ± 41.42a0.94 ± 0.03a533.50 ± 41.42a99.33%T10.30 ± 0.13c69.67 ± 30.90c0.05 ± 0.02c69.67 ± 30.90c100%T21.18 ± 0.31b140.33 ± 27.43b0.24 ± 0.07b140.40 ± 27.50b99.67%T30.36 ± 0.19c89.00 ± 28.16bc0.06 ± 0.04c89.00 ± 28.16bc100% Analysis of bacterial alpha diversity Alpha-diversity analysis of fungal communities is summarized in Table 3. Coverage indices for all samples reached 100%, indicating that the sequencing depth was sufficient to comprehensively capture fungal diversity, ensuring high data reliability. Chao1, ACE, Shannon, and Simpson indices exhibited a decreasing trend followed by an increase across treatments, suggesting that fermentation significantly influenced fungal richness and diversity. CK displayed the highest species richness and evenness, representing the most complex fungal community structure. After fermentation, T1 exhibited a decline in fungal diversity and richness, yet these remained at relatively high levels. T2 exhibited the lowest diversity and richness, suggesting that certain fungal populations were likely suppressed or competitively excluded during this stage. T3 demonstrated partial recovery of diversity and richness, implying potential restoration or restructuring of fungal communities during the later fermentation phase. Overall, compared with CK, all fermentation treatments significantly reduced fungal diversity and richness (P < 0.05), indicating that enzyme-producing strains substantially altered the fungal community structure of tobacco stems, potentially affecting their subsequent metabolic potential and functional stability.Table 3Analysis of fungal alpha diversityTreatmentShannon indexChao1 indexSimpson indexAce indexCoverage indexCK4.76 ± 0.17a70.33 ± 7.57a0.94 ± 0.02a70.33 ± 7.57a100%T13.25 ± 0.31b64.00 ± 22.52b0.74 ± 0.02b64.00 ± 22.52a100%T21.72 ± 0.21c19.00 ± 3.00c0.58 ± 0.01c19.00 ± 3.00b100%T31.95 ± 0.14c48.00 ± 24.00c0.57 ± 0.06c48.00 ± 24.00ab100% Table 3Analysis of fungal alpha diversityTreatmentShannon indexChao1 indexSimpson indexAce indexCoverage indexCK4.76 ± 0.17a70.33 ± 7.57a0.94 ± 0.02a70.33 ± 7.57a100%T13.25 ± 0.31b64.00 ± 22.52b0.74 ± 0.02b64.00 ± 22.52a100%T21.72 ± 0.21c19.00 ± 3.00c0.58 ± 0.01c19.00 ± 3.00b100%T31.95 ± 0.14c48.00 ± 24.00c0.57 ± 0.06c48.00 ± 24.00ab100% Analysis of fungal alpha diversity Principal coordinate analysis (PCoA) was performed to further investigate changes in microbial community structure across treatments (Fig. S8). For bacterial communities, the first two principal components accounted for 97.3% of the total variance. CK samples were clearly separated from fermentation-treated groups, reflecting significant differences, whereas treated samples clustered closely, indicating similar community structures (Fig. S8a). For fungal communities, the first two principal components explained 63.2% of the variance. CK remained distinctly separated from treated groups, whereas T1 and T3 samples clustered closer together, suggesting more similar fungal community compositions (Fig. S8b). Collectively, these results indicate that microbial fermentation markedly reshapes both bacterial and fungal community structures in tobacco stems. To further elucidate compositional differences in microbial communities of tobacco stems under different treatments, OTU annotation was conducted at 99% sequence similarity. Analysis of bacterial community structure revealed that, in CK, the dominant phyla were Proteobacteria and Bacteroidota, whose relative abundances were significantly higher than those observed in bacterial inoculation treatments. Representative genera within Proteobacteria, including Pseudomonas and Enterobacter, are ubiquitous environmental bacteria capable of organic matter degradation. However, their metabolic products may include ammonia, phenols, and other compounds potentially detrimental to tobacco aroma quality. Bacteroidota, although capable of polysaccharide degradation, possess limited aromatic metabolic pathways and therefore contribute minimally to flavor formation. In contrast, bacterial inoculation treatments resulted in a significant increase in Firmicutes, which became the dominant phylum. This phylum predominantly comprises functional genera, including Bacillus and Lactobacillus, which efficiently secrete cellulases, proteases, and other enzymes, enabling rapid decomposition of macro-molecular organic matter in tobacco stems. Their metabolic activities also produce lactic acid, acetic acid, and various aromatic esters, thereby contributing to improved aroma and flavor profiles (Fig. 7a–b). These results indicate that bacterial inoculation markedly reshapes the tobacco stem microbiota, with the predominance of Firmicutes facilitating flavor enhancement.Fig. 7Microbial community composition of tobacco stem samples at the phylum and genus levels. a Bacterial community composition at the phylum level; (b) bacterial community composition at the genus level; (c) fungal community composition at the phylum level; (d) fungal community composition at the genus level Microbial community composition of tobacco stem samples at the phylum and genus levels. a Bacterial community composition at the phylum level; (b) bacterial community composition at the genus level; (c) fungal community composition at the phylum level; (d) fungal community composition at the genus level Fungal community analysis revealed that CK was overwhelmingly dominated by Ascomycota, primarily represented by Aspergillus and Mycosphaerella, resulting in a relatively simple community structure. In inoculated treatments (T1–T3), fungal community composition was significantly altered, with the relative abundance of Ascomycota increasing, whereas Basidiomycota decreased, particularly in T3. Aspergillus and Mycosphaerella nearly occupied all ecological niches in treated samples, indicating that inoculation promoted proliferation of dominant fungi. At the genus level, functional fungi, including Meyerozyma, Aspergillus, and Chaetomium were enriched. Meyerozyma, a yeast, possesses strong carbohydrate fermentation and ester synthesis capabilities, thereby contributing to the formation of fruity compounds, including ethyl acetate and ethyl lactate. Aspergillus secretes cellulases and proteases, thereby facilitating macro-molecule breakdown and the release of aroma precursors. Chaetomium excels in lignocellulose degradation and promotes the accumulation of aromatic alcohols and phenolic compounds. The combined metabolic activities of these genera enhance metabolite diversity and flavor complexity, providing a crucial microbial foundation for improving the sensory quality of fermented tobacco stems (Fig. 7c–d). Thus, the succession of fungal communities plays an essential role in generating a more harmonious and complex aroma profile, thereby supporting overall flavor enhancement during fermentation. The metabolic functions of bacterial and fungal communities in tobacco stems subjected to different treatments were predicted using Phylogenetic Investigation of Communities by Reconstruction of Unobserved States 2(PICRUSt2) based on the MetaCyc database (Fig. 8). This approach infers microbial functional profiles from 16S rRNA and ITS sequences and subsequently maps them to metabolic pathways for functional annotation. KEGG pathway analysis revealed that bacterial communities were predominantly enriched in core metabolic processes, including amino acid metabolism, carbohydrate metabolism, fatty acid biosynthesis, secondary metabolite biosynthesis, and energy metabolism. Compared with CK, bacterial inoculation significantly increased the abundance of these functional pathways (Fig. 8a). Fungal metabolic functions were primarily associated with aerobic respiration, nucleic acid metabolism, fatty acid β-oxidation, and amino acid metabolism, processes closely linked to improvements in tobacco aroma and color. Among the treatments, the T2 group exhibited the highest functional abundance. These results indicate that bacterial inoculation promotes the activation of key microbial metabolic pathways, with fungal communities potentially playing a particularly important role in the biosynthesis of flavor compounds (Fig. 8b).Fig. 8Functional modules of bacterial communities in MetaCyc tertiary pathways, (a) Bacterial functional prediction module; (b) Fungal functional prediction module Functional modules of bacterial communities in MetaCyc tertiary pathways, (a) Bacterial functional prediction module; (b) Fungal functional prediction module Correlation analysis between dominant microbial taxa and volatile compounds in fermented tobacco stems To elucidate potential associations between microbial communities and volatile compounds in fermented tobacco stems, OPLS-DA modeling was performed to integrate microbiome and non-targeted metabolome data (Fig. S9). The first two principal components accounted for 41.8% of the total variance, and all samples fell within the 95% confidence interval. The CK was clearly separated from all treated groups (T1, T2, and T3), indicating that bacterial inoculation significantly altered both microbial composition and volatile metabolic profiles of tobacco stems. Differentially abundant dominant taxa and key volatile compounds were subsequently identified, suggesting their potential regulatory roles in flavor formation during fermentation. Based on the OPLS-DA model, 12 dominant bacterial genera and 16 dominant fungal genera were selected, and Spearman correlation analysis, integrated with Cytoscape, was used to construct microbial–metabolite correlation networks (Fig. S9). In the bacterial community (Fig. 9a), Bacillus exhibited significant positive correlations with multiple key flavor compounds, particularly Maillard reaction products, including neophytadiene and isopropanol (P < 0.05), as well as various fatty acids and esters (e.g., dodecane, γ-butyrolactone, dimethyl phthalate). Conversely, Bacillus showed negative correlations with pungent compounds, including acetic acid, ethanol, and nicotine. Other bacterial genera, including Acinetobacter, Microbacterium, and Methanobacterium, were positively correlated with 2-acetylpyrrole, acetic acid, and guaiacol, but negatively correlated with most other compounds, highlighting Bacillus as the dominant genus in flavor enhancement. In the fungal community (Fig. 9b), Chaetomium was significantly associated with 15 volatile compounds, showing positive correlations with neophytadiene, isopropanol, and dodecane (P < 0.05), and negative correlations with acetic acid and guaiacol. Other fungal genera, including Penicillium, Naganishia, Epicoccum, and Kurtzmaniella, exhibited positive correlations with 2-acetylpyrrole and acetic acid. Overall, Bacillus and Chaetomium played dominant roles in regulating Maillard reaction product formation, thereby driving aroma and flavor enhancement in fermented tobacco stems, whereas genera such as Acinetobacter and Penicillium may contribute to 2-acetylpyrrole accumulation and color development, providing metabolic support for quality improvement.Fig. 9Cluster analysis of the correlation between dominant bacterial genera and volatile compounds. a bacteria; b Fungi; Note: The correlation clustering heatmap displays the Spearman correlation analysis results of 12 dominant bacterial genera, 16 dominant fungal genera (columns), and 20 volatile compounds (rows). The color of the ruler in the figure represents the significance of key differences in volatile compounds. Among them, "1" represents complete positive correlation (dark red), and "−1" represents complete negative correlation (dark blue) Cluster analysis of the correlation between dominant bacterial genera and volatile compounds. a bacteria; b Fungi; Note: The correlation clustering heatmap displays the Spearman correlation analysis results of 12 dominant bacterial genera, 16 dominant fungal genera (columns), and 20 volatile compounds (rows). The color of the ruler in the figure represents the significance of key differences in volatile compounds. Among them, "1" represents complete positive correlation (dark red), and "−1" represents complete negative correlation (dark blue) Discussion The enzyme-producing microbial isolates identified in this study demonstrated significant potential to improve the color and sensory quality of tobacco stems through enzymatic bio-transformation. The conventional chemical composition of tobacco stems is closely associated with their visual appearance and aroma characteristics, and previous studies have shown that sugars, nicotine, and nitrogenous compounds undergo degradation and transformation during fermentation, influencing color and flavor development [8, 26, 35–38]. In this study, microbial fermentation notably increased total and reducing sugars and modulated nitrogen content, indicating that microbial enzymes facilitated the conversion of macro-molecules into smaller, aroma-active compounds. The pronounced darkening of T2-treated stems was likely attributable to enzymatic browning reactions between sugars and amino acids catalyzed by microbial oxidases, producing brown pigments that contributed to improved color. Although total alkaloids were highest in CK, the moderate decrease in treated groups suggested a balancing effect on alkaloid content, which could reduce bitterness and harshness while enhancing smoothness and aroma. Mechanistically, cellulase and polyphenol oxidase were identified as key enzymes affecting color and aroma formation. Cellulase degrades cellulose into soluble sugars, loosening plant cell walls and releasing intracellular pigments and aroma precursors, whereas polyphenol oxidase catalyzes the oxidation of phenolic compounds into quinones that polymerize to form pigments and volatile aromatic intermediates [18, 39]. The synergistic action of these enzymes constitutes the core biochemical mechanism underlying color and aroma improvement in fermented tobacco stems, which aligns with the observed compositional and sensory changes in this study. Comprehensive volatile analysis revealed that microbial fermentation significantly increased the abundance of Maillard reaction products and terpenoids, particularly in the T2 treatment. Key compounds such as hexanal, ionone, 6-methyl-5-hepten-2-one, ethyl acetate, 2-pyrrolidone, 2-pentylfuran, and nonanal were markedly enriched, contributing sweet, fruity, and floral notes characteristic of high-quality tobacco [40–43]. The generation of 2-pentylfuran and ionone suggests microbial participation in amino acid and carotenoid metabolism [44, 45], while increased neophytadiene levels indicate active degradation of carotenoids, imparting green and honey-like nuances [43, 46]. Enrichment of pathways related to aromatic compound degradation, monoterpenoid biosynthesis, and furfural degradation further supports the role of enzyme-producing microbes in regulating key metabolic networks associated with aroma formation [47–49]. High-throughput sequencing revealed pronounced shifts in microbial community composition during fermentation. The dominance of Bacillus, Meyerozyma, Aspergillus, and Chaetomium after inoculation indicated that enzymatic strains reshaped the microbial ecosystem, promoting functional specialization. Bacillus abundance increased dramatically, consistent with its reported ability to degrade macro-molecules and generate flavor compounds [12, 20, 21, 50–52]. Fungal genera such as Aspergillus and Chaetomium contributed to the synthesis of secondary metabolites involved in aroma enhancement [27]. Functional prediction indicated that bacterial pathways were enriched in amino acid and carbohydrate metabolism, while fungal pathways were associated with fatty acid degradation and energy metabolism. These processes provided essential precursors for Maillard reactions and volatile compound formation, consistent with previous reports [53–55]. Collectively, microbial restructuring enhanced enzyme activity and metabolic flux, accelerating the conversion of primary metabolites into volatile aromatic compounds and improving both color and flavor quality. Correlation analysis further revealed that Bacillus was positively associated with Maillard-derived volatiles, such as neophytadiene and isopropanol, and with fatty acid-derived esters, while negatively correlated with nicotine and other irritant compounds. Chaetomium and Aspergillus showed strong associations with terpenoid volatiles, indicating cooperative interactions that enhanced aroma intensity [13, 56–59]. The simultaneous increase in Bacillus and Aspergillus abundance in T2 samples corresponded to the highest aroma compound levels and darkest color, suggesting synergistic metabolism between bacterial and fungal enzymes. Moreover, reduced nicotine and ethanol levels indicated the mitigation of off-flavors and improved overall sensory balance. From a microbiological perspective, the construction of a simplified synthetic community comprising strains C8 and Z1 was based on their compatibility and cooperative fermentation behavior, consistent with prior reports that bacterial–fungal consortia can efficiently degrade macro-molecular substrates [60]. The stable coexistence of these strains enhanced enzymatic diversity, fermentation efficiency, and metabolite accumulation, demonstrating the importance of strain compatibility in designing functional microbial consortia. Nonetheless, this study was limited to a two-strain system under controlled laboratory conditions. Future work should expand to multi-strain co-cultures, optimize fermentation environments, and scale up validation under industrial conditions to ensure reproducibility and practical applicability. In summary, this study provides new insights into the enzymatic and microbial mechanisms driving the bio-transformation of tobacco stems. Enzyme-producing strains such as B. amyloliquefaciens and C. nepalense not only restructure the microbial community but also catalyze key metabolic pathways that improve tobacco color, aroma, and sensory quality. These findings highlight a sustainable biotechnological strategy for upgrading low-value tobacco by-products into high-quality materials through controlled microbial fermentation. Conclusions This study identified two high-activity enzyme-producing strains, B. amyloliquefaciens C8 and C. nepalense Z1, which can be co-applied for artificial tobacco stem fermentation. Microbial inoculation reshaped the metabolic and microbial profiles of tobacco stems, enhancing sugars, minerals, and key aroma-related metabolites such as terpenoids and Maillard reaction products. These changes collectively improved stem color and aroma quality. Correlation analysis highlighted the pivotal roles of Bacillus and Chaetomium in driving the biosynthesis of characteristic volatiles. Overall, this work provides novel insights into the microbial mechanisms underlying tobacco stem fermentation and offers a foundation for developing bio-based strategies to enhance tobacco processing and quality improvement. Supplementary Information Supplementary Material 1. Supplementary Material 2. Supplementary Material 1. Supplementary Material 2.