The 14 samples came from different tobacco growing areas in Fujian and were divided into three grades, namely high, middle, and low. Samples were provided by the Fujian China Tobacco Industry Co., Ltd. The sample information is listed in Table 1.

Test strain: Paenibacillus amylolyticus A17 # was screened from the surface of flue-cured tobacco samples by the Institute of Food Science and Technology of Fuzhou University, and the molecular biological identification was completed by Bioengineering (Shanghai) Co., Ltd. using the 16 S rDNA identification method with the registration number of SUB13351325 Paenibacillus OQ991262.

In order to prepare the tobacco sample for analysis, 10 grams of aged flue-cured tobacco were added to 60 milliliters of pH 7.4 phosphate buffer in the conical flask. The mixture was shaken and washed at 200 rpm for 40 min before the washing solution was collected. A total volume of 40 mL of phosphate buffer was added to the flue-cured tobacco sample several times. The two washing solutions, which were combined after full oscillation mixing, were filtered with a 0.22-μm filter membrane to enrich the microorganisms. Ten milliliter of sterile water was utilized to wash the filter membrane. Subsequently, the bacterial solution was collected, centrifuged at 7,000 × g for 20 min, and stored at −80°C for microbial community detection. The detection and analysis of the microbial community entrusted Shanghai Meiji Bio-pharmaceutical Technology Co., Ltd. to use the Illumina MiSeq sequencing platform for high-throughput sequencing and the Meiji Bio-cloud platform for data analysis.

PICRUSt can predict the function of metagenomes through marker gene data and a reference genome database (Langille et al., 2013). Based on 16S rRNA sequencing data, it can also be utilized to find direct evidence of the functional ability of microbial communities (Yin and Wang, 2021). The PICRUSt software calculates the abundance of different functional classes for each sample by making functional predictions based on the high-throughput sequencing results and the gene database.¹

The volatile substances were extracted by headspace solid-phase extraction. The samples were crushed into powder and passed through a 425-μm sieve. After being screened, 0.4 g of the sample was put into a 20 mL sample extraction bottle, and 3 μL of phenylethyl acetate, whose concentration was 3.432 mg·L⁻¹, was added as the internal standard. The headspace bottle was put into the CTC automatic sampling tray. The extraction parameter was set at 100°C for 10 min, and the volatiles were adsorbed by a 50/30 μm divinylbenzene/carboxen/poly-dimethylsiloxane (DVB/CAR/PDMS) fiber (Supelco, Bellefonte, PA, United States) for 40 min. After 4 min of analysis at 250°C at the sample inlet of the gas chromatography-mass spectrometer (model 7890B-5977A, Agilent, Palo Alto, CA, United States), the program was started.

Gas chromatography–mass spectroscopy (GC–MS) conditions: a Hp-5 ms chromatographic column (60 m × 0.25 mm × 0.25 μm) was used. The carrier gas is 99.999% pure helium at a flow rate of 1 mL· min⁻¹. The temperature program was as follows: the initial temperature was held at 50°C for 2 min and programmed to increase at 8°C·min⁻¹ to 280°C, and then was held for 25 min. The post-run temperature was 280°C, and the post-run time was 5 min. The Splitless injection was used. The solvent delay was 4 min.

Mass spectrum conditions: the electron energy was 70 eV, and the ion source temperature was 230°C. The interface temperature was 280°C, and the transfer line temperature was 150°C. The scanning mode was full scanning and the mass scanning range was 35–350 m/z.

Qualitative and quantitative analysis of volatile substances: The total ion chromatogram of volatile substances were obtained by GC–MS analysis and were retrieved by the NIST 14 spectrum library. Qualitative analysis was carried out together with reference documents, and semi-quantitative calculation was carried out for each component based on the concentration of the internal standard. Content of volatile components (μg·L⁻¹) = concentration of internal standard × three × Peak area of component/peak area of internal standard.

In the early stage of the experiment, a batch of strains producing starch hydrolysis circles in the amylase screening medium plate were isolated from the surface of Fujian flue-cured tobacco. The strains with a large hydrolysis circle were selected for further comparison of amylase activity in the fermentation broth; incidentally, the fermentation broth was evenly sprayed on the cut tobacco for sensory evaluation. Finally, Paenibacillus amylolyticus A17 #, with strong amylase activity that has significantly improved the sensory of flue-cured tobacco, was selected for follow-up tests.

The activated A17 # strain was incubated in TSB medium at 200 r·min⁻¹ and 37°C for 12 h at a constant temperature, and 5 mL (per 100 mL fermentation liquor) of bacterial liquid was taken into the fermentation medium. The fermentation conditions were 37°C and 200 r·min⁻¹ for 24 h. Based on the fermentation medium, the fermentation broth was taken at 3 h, 6 h, 12 h, 18 h, 24 h, and 36 h to determine the growth of the strain (calculated by OD_600nm). The fermentation broth was centrifuged at 9,000 × g at 4°C for 10 min, and its supernate (crude enzyme solution; Chen et al., 2017) was concentrated and filtered through ultrafiltration membranes of 100 kDa and 30 kDa to produce the liquid amylase preparation for enzymatic activity determination.

The determination method of amylase activity is adjusted with reference to Yoo’s modified method (Yoo et al., 1987). The amylase activity is defined under the conditions of pH 5.8 and temperature 37°C, and the amount of enzyme required to hydrolyze 1 mg of starch in 5 min is one enzyme activity unit (U), expressed in U·mL⁻¹.

The determination method of pectinase activity refers to the determination method of pectinase activity of Tang et al. (2019). The pectinase activity is defined under the conditions of pH 9.4 and temperature 50°C, and the amount of enzyme required to catalyze pectin hydrolysis to form 1 μg of galacturonic acid is one enzyme activity unit (U), expressed in U·mL⁻¹.

The determination method of cellulase activity refers to the determination method of cellulase activity of Bedade et al. (2017). The cellulase activity is defined under the conditions of pH 5 and temperature 50°C, and the amount of enzyme required for the hydrolysis of CMC-Na to form 1 μg of glucose in 1 min is one enzyme activity unit (U), expressed in U·mL⁻¹.

The determination method of protease activity is determined according to the national standard GB23527-2009 Folin method. The protease activity is defined under the conditions of pH 7.5 and a temperature of 40°C, and the amount of enzyme required for the hydrolysis of casein to form 1 μg of tyrosine is one enzyme activity unit (U), expressed in U·mL⁻¹.

The contents of pectin and cellulose were determined with reference to YC/T 346–2010 (Tobacco and tobacco products—Determination of pectin, 2010) and YC/T 347–2010 (Tobacco and tobacco products—Determination of neutral detergent fiber, acid detergent fiber and acid detergent lignin, 2010) respectively. Starch and protein were determined by using an automatic continuous flow analyzer (France Aliance, model Futura) according to standards YC/T 216–2013 (Tobacco and tobacco products—Determination of starch, 2013) and YC/T 249–2008 (Tobacco and tobacco products—Determination of protein, 2008) respectively.

Total alkaloids, water-soluble total sugar, and reducing sugar were determined by the continuous flow method, using the same macromolecular content determination equipment as above, and referring to YC/T 160–2002 (Tobacco and tobacco products—Determination of total alkaloids, 2002), YC/T 217–2007 (Tobacco and tobacco products—Determination of Potassium, 2007) and YC/T 159–2019 (Tobacco and tobacco products─Determination of water soluble sugars, 2019) respectively.

The redried tobacco leaves that had been made into cut tobacco from the lower grade SLC4F-2020 in Nanping, Fujian Province, were selected as the samples for the sensory test. Simulate the workshop operation process: 50 g of tobacco leaves were taken and laid on the sample tray, and the amylase solution was evenly sprayed on the tobacco leaves by a throat sprayer at an application rate of 1,000 U·kg⁻¹ of tobacco at a 5% application ratio. Subsequently, the treated tobacco leaves were put in a bag at room temperature for 6 h for enzymatic hydrolysis. Later, the enzyme was inactivated at 135°C for 1 min. The treated samples were bagged in a constant temperature and humidity incubator at 22°C and (60 ± 5) % relative humidity to balance the moisture for 48 h. In the meantime, the same amount of water and inactivated enzyme solution were taken as the control. Finally, the samples obtained were made into cigarettes, randomly numbered, and then distributed to 10 professional reviewers for sensory evaluation. Refer to GB 5606.4–2005 (Cigarettes, 2005). The six items of each cigarette, including aroma, delicacy, impurity, irritation, etc., were evaluated and scored, and the results of the sensory evaluation were presented as the average scores of 10 experts.

All samples were conducted at least in triplicate and the results were expressed as means ± standard deviation. Statistical analysis of the data was performed using Statistical Package for the Social Sciences (SPSS) 23.0 (IBM, Armonk, New York). Student’s t-test was used for two-group comparisons (*p < 0.05, **p < 0.01),

A total of 80 strains were isolated and purified from the surface of 14 Fujian flue-cured tobacco samples, of which 7 strains had strong starch-degrading abilities in a starch-solid screening medium. Based on the results of the smoking evaluation after spraying the crude enzyme solution of the strain on the flue-cured tobacco, strain A17 #, identified as Paenibacillus amylolyticus, was screened, which could significantly improve the quality of tobacco leaves.

According to the determination of its enzyme-producing growth curve, which was shown in Figure 5. Paenibacillus amyloliquefaciens A17 # grew logarithmically in 0–12 h under the basic enzyme-producing medium and was in a stable phase after 12 h. Amylase was produced at 12 h and rapidly accumulated at 18–24 h. To sum up, the strain reached the peak of enzyme production at the stable growth stage, and the highest amylase activity was 53.28 U·mL⁻¹ at 24 h. With the prolongation of fermentation time, amylase activity showed a downward trend in 24–36 h due to the reduction of nutrients, the accumulation of waste, the inhibition of cell death, and the decomposition of metabolites (Mao et al., 2020).

Meanwhile, the cellulase, pectinase, and protease activities of the liquid enzyme preparation of Paenibacillus amyloliquefaciens A17 # strain were measured. The results show that the A17 # strain could produce pectinase and cellulase in addition to producing more amylase under the starch-induced enzyme-producing medium, and the pectinase activity can reach 280 U·mL⁻¹ (Figure 6), which indicated that the preparation of Paenibacillus amyloliquefaciens A17 # was a complex enzyme system.

The content of macromolecular substances in tobacco leaves before and after the treatment of enzyme preparation was determined on the basis of taking the inactivated enzyme-treated group as the control group. As shown in Figure 7A, the starch content in the flue-cured tobacco was significantly reduced after the enzyme treatment, with a decrease of 15.05% relative to the control, indicating that the enzyme treatment achieved the degradation of starch in the flue-cured tobacco. Moreover, after measuring the content of other macromolecules, it was found that the content of pectin (Figure 7B) and cellulose (Figure 7C) were significantly reduced, decreasing by 10.71 and 31.89%, respectively, compared with the control group. While the content of protein (Figure 7D) did not change significantly after the enzyme treatment. These results further revealed the microbial enzyme preparation was a compound enzyme that can degrade the cell wall substances of tobacco leaves. It was found that pectinase could cleave the intercellular layer and primary wall of tobacco leaves α- 1,4 glycosidic bond, which loosens the binding between the concentric layers of fibers, and further plays the role of cellulase to break the cellulose structure of the cell wall (Wang et al., 2020). Combined with the results of enzymatic activity characterization of enzyme preparation in the early stage, it is speculated that the synergism of gelatinase and cellulase causes a large decrease in cellulose content. Based on the change in macromolecule content and the subsequent sensory evaluation results, it was confirmed that reducing the macromolecule content in tobacco leaves could significantly improve the original defects of tobacco leaves, such as scorching and choking, and improve the quality of tobacco leaves.

The content of water-soluble total sugar and reducing sugar in tobacco leaves are important parameters in the routine chemical analysis of flue-cured tobacco. These parameters are critical in determining the mellowness of flue-cured tobacco smoke and serve as key indicators of its overall quality (Fan, 2017). In our study, the content of total water-soluble sugar (Figure 8A) in the flue-cured tobacco was significantly higher in the enzyme-treated group, with an increase of 3.13%, and the content of reducing sugar (Figure 8B) was significantly increased by 2.40% compared to the control group. The enzyme-treated group had been found to significantly increase the content of water-soluble sugar and reducing sugar, which was an important factor in the enhanced sweetness of the flue-cured tobacco and a more delicate and balanced smoke.

The level of sugar in tobacco products is an important factor in determining the preferences of both tobacco producers and consumers. Typically, tobacco products with sugar content ranging from 8 to 30% are preferred (Talhout et al., 2006). Alkaloids, which mainly consist of more than 90% nicotine (Zhou et al., 2012), are responsible for the strength of smoke. However, when the content of alkaloids is too high, it can cause discomfort, such as choking and irritation, and even pose risks to human health. In this study, it was found that the content of alkaloids (Figure 8C) in flue-cured tobacco after the enzyme treatment showed a downward trend, with a decline of 0.32%, which could further enhance the safety of tobacco.

The inactivated enzyme-treated group was used as the control group. The volatile components in flue-cured tobacco were analyzed using HS-SPME/GC–MS technology. A total of 68 volatile substances were identified, including 15 ketones, 6 aldehydes, 3 alcohols, 6 acids, 19 esters, 10 heterocyclic compounds, and 9 other compounds.

The value of variable importance in the projection (VIP) can be used to quantify the contribution of each variable to the classification (Liu et al., 2018). In the study, the VIP values of volatile components in flue-cured tobacco before and after the enzyme treatment were calculated using SIMCA software (v 14.1, MKS Umetrics AB). A total of 29 volatile components were found to have VIP values greater than 1 (Figure 9A), indicating their significant contribution to the classification. The load diagram showed that there were more volatile aroma components near the position of the enzyme-treated group (AT), indicating that the aroma components of flue-cured tobacco after the enzyme-treated were richer and the aroma characteristics were more obvious (Figure 9B).

The significance analysis of the important volatile components (VIP > 1) screened out is shown in Figure 9C. Compared with the control group, 2,3-dihydro-3,5-dihydroxy-6-methyl-4H-Pyran-4-one (C2), phenylacetic acid (C26), ethyl 13-methyl tetra decanote (C37), ethyl palmitate (C41), ethyl linoleate (C47), and linolenic acid ethyl ester (C48) in the enzyme-treated group were significantly increased. Phenylacetic acid (C26), with a sweet and floral fragrance, and ethyl palmitate (C41) which causes the sweet sense were the products of the Maillard reaction (Li et al., 2017). Previous studies have shown that the significant increase in the content of Maillard reaction products not only has a positive and direct contribution to the improvement of the aroma quality, aroma quantity, aftertaste, and total smoking score of tobacco leaves but also has an indirect effect on the reduction of impurities and irritation (Shi et al., 2013; Wang et al., 2019). After the enzyme treatment, sweet esters such as ethyl linoleate (C47) and ethyl linolenic acid (C48) increased significantly. The ester components have a certain impact on the aroma and taste of tobacco leaves, which could increase the mellow feeling of flue gas (Zuo et al., 2020).

The enzyme preparation was evenly sprayed on the tobacco leaves by a throat sprayer at an application rate of 1,000 U·kg⁻¹ of tobacco at a 5% application ratio, while two control groups were sprayed with the same amount of water and inactivated enzyme solution, respectively. The ten professionals evaluated treated tobacco leaves based on four aspects: style characteristics (sweetness), aroma characteristics (aroma quality, aroma quantity, impurity), smoke characteristics (fineness, concentration, strength), and taste characteristics (irritation, aftertaste). The results showed that the enzyme treatment improved the style characteristics of the tobacco leaves, making them sweeter. The aroma quality and quantity were also enhanced, while impurities were reduced. The smoke characteristics were improved, with increased fineness, concentration, and moderate strength. The taste characteristics were also improved, with reduced irritation and a better aftertaste (Figure 10). In conclusion, enzymatic hydrolysis could effectively improve the characteristics of flue-cured tobacco, making it sweeter, with better aroma, smoke, and taste characteristics.