Bioaugmentation of Bacillus amyloliquefaciens-Bacillus kochii co-cultivation to improve sensory qualities of ue-cured tobacco

Flue-cured tobacco (FCT) with irritating and undesirable avor must be treated with aging process. However, the spontaneous aging usually takes a very long time for the low eciency. Bioaugmentation with functional strains is a promising method to improve FCT quality with high aging eciency. To ease the adverse effect of excessive starch or protein content on the FCT quality, we screened B. amyloliquefaciens LB with high alpha-amylase and B. kochii SC with high neutral protease from the FCT microora with the ow cytometry. The mono, co-culture of strains was subjected to solid-state fermentation with FCT. B. amyloliquefaciens monoculture at 2 days and B. kochii monoculture at 2.5 days achieved the optimum quality. B. amyloliquefaciens-B. kochii co-culture at a ratio of 3:1 fermenting for 2 days demonstrated a more comprehensive quality enhancement and higher functional enzyme activity than mono-cultivation. With the analysis of OPLS-DA model, there were 38 differential compounds between bioaugmentation samples. In co-cultivation, most of Maillard reaction products and terpenoid metabolites were at a higher level than other samples, which promoted an increase in aroma, softness and a decrease in irritation. This result was in line with our design about the quality enhancement of FCT with co-cultivation. In our study, we presented a promising bioaugmentation technique for changing the quality attributes of FCT in a short aging time. which was implied that there was a good synergy on quality improvement of FCT. 38 differential compounds were analysis through OPLS-DA model. In co-cultivation, most of Maillard reaction products and terpenoid metabolites were at a higher level than others, which conrmed the hypothesis of screening protocol. However, the interaction mechanism between two strains needed further study. In this study, it provided a new reference for the function expansion that B. kochii was screened and application in FCT. In this study, we presented an effective bioaugmentation technique for changing the sensory attributes of FCT, which will be signicance for improving raw material utilization and reducing costs in tobacco industry.

Tobacco (Nicotiana tabacum L.) is a special industrial crop planted in over 120 countries and regions (Ye et al., 2017). Most of fresh tobacco is made into ue-cured tobacco (FCT). Flue-cured but unaged tobacco is usually irritating and undesirable smoke (Su et al., 2011). Therefore, the aging is an essential process to improve the FCT quality. The aging process usually takes 1 to 2 years, which is long along with low e ciency. It is necessary to seek approaches to enhance sensory the aging e ciency (Ye et al. 2017;. Previous studies has reported that some speci c functional microorganisms were high effective in improving the quality (Chortyk O. T. et al. 1973;He 2013;Li et al. 2019) or degrading nicotine of FCT (He et al. 2019) in aging process.
The starch is an important carbohydrate in FCT. However, More than 3% starch content can have a negative impact on the sensory quality of FCT. Meanwhile, starch degrades into the soluble polysaccharides or reducing sugars which can improve the sweetness and aftertaste of FCT (Feizi et al. 2020). The protein content of tobacco should drop to around 5% before the aging process, and excess protein content in FCT may be more irritation and off-avor (Han et al. 2010). In addition, reducing sugars and amino acids rich in FCT can promote the formation of Maillard reaction products to enhance the softness and aftertaste after burning. Therefore, we hypothesized that co-cultures of strains that degrade starch and protein could promote the enhancement of quality and reduce the off-avors and irritation.
To prove our hypothesis, we sorted functional microorganisms with high degradability of starch or protein by ow cytometry. Strains were inoculated on FCT by mono-or co-culture of solid-state fermentation. We dynamically studied the changes of FCT sensory qualities, enzyme activity along with volatile components in bioaugmentation. we enhanced the effect via the co-cultivation of two strains. The biocultivation in FCT will be an effective technique for changing the sensory attributes. It will be signi cance for improving raw material utilization and reducing costs of aging process in industry.

Single-cell sorting by ow cytometry
To get more strains from the microbial community, single-cell was sorted by ow cytometry. First, we prepared the mixture of ora on FCT by the following method. The faint-avor type of FCT sampled from China Tobacco Si Chuan industrial Co, Ltd (Chengdu County, Sichuan Province, China). A 10 ± 0.1 g of sample added into 200 mL sterilized 0.1 mol/L phosphate buffer solution (pH 7.2) and shocked at 220 rpm, 30 o C for 2 h. The mixture was sonicated for 5 min and ltered through the sterile two-layer of absorbent gauze. The ltrate was centrifuged at 500 × g for 5 min to separate the non-cellular particles, then centrifuged at 7,000 × g for 10 min to collect the microorganisms (Su et al., 2011), repeatedly 2 or 3 times. The precipitation was stained with 7-AAD solution (BD Pharmingen, New Jersey, USA) for the optimized 20 min, and removed the residual dye with the sterile re-distilled water, then diluted to a biomass with the absorbance of 0.3 at 600 nm (OD 600 ). Suspension was ltered through a 40 µm lter and stored in ice for use.
Second, 1 mL suspension was input into the sample bin of ow cytometry (FACSAria III Cell Sorter, BD Biosciences, USA). Single-cell allowed to pass one by one through the nozzle and sorted into the 96 -well plate. Reference to Zhang' s method (2018), the main parameters were following: Dulbecco' s phosphate buffer (DPBS) used as a sheath solution to protect the cells and clean the nozzle; nozzle size was selected as100 µm; the pressure of sheath uid was 30 psi (1 psi = 6.895 kPa); oscillation frequency was 4000 Hz and amplitude was 30-50 V; sorting mode set as "Single-0.5 Droplet"; the pressure difference between the sample and the sheath uid was 0.2 to 0.3 psi to ensure 3,000 to 5,000 cells per second through the nozzle. The 96-well plates were tted with the100 µL basic medium. Bacteria cultured in Nutrient Broth (NB) medium at 37 o C for 24-72 hours. Fungi cultured in Bengal Red (BR) medium at 30 o C for 72-120 hours.

Screening of functional strains
The colony-forming strains in 96-well plates were inoculated into the selective medium. The selective medium for alpha amylase-producing strains used soluble starch 10 g/L as the only carbon source in NB medium. The selective medium for protease-producing strains used soluble casein 2 g/L replacing peptones of BR medium. After cultivation, the alpha-amylase activity of strains was investigated based on DNS (3,5-dinitrosalicylic acid) assay (Suraiya et al. 2018). One unit (U) of alpha-amylase activity de ned as the amount of enzyme required to produce 1 mg of reducing sugar per minute at 60 o C, pH 6.0. Protease activity detected according to the Folin color method (Anson, 1938). One unit of neutral protease activity (U) de ned as the amount of enzyme that produces 1 µg tyrosine per minute under 40 o C, pH 7.0. The strains with the highest alpha-amylase or protease would be screened out.

Identi cation of strains
Strains identi ed via 16S rDNA gene sequencing, which was determined according to the method described by Thakur et al. (2015). Total DNA was extracted using the MiniBEST Bacteria Genomic DNA Extraction Kit (Takara, Japan) following the manufacturer's instructions. The polymerase chain reactions (PCR) were ampli ed in 50 µL reaction volume with the forward primer 27F (5′-GAGAGTTTGATCC-TGGCTCAG-3′) and 1492R (5′-ACGGGCGGTGTGTRC-3′) at 95 o C for 5 min, 30 cycles at 94 o C for 90 s, 55 o C for 30 s, nally an extension at 72 o C for 2 min. The PCR product sequenced in Sangon Biotech (Shanghai) Co., Ltd. (Shanghai, China). The sequencing data uploaded to NCBI database (National Center for Biotechnology Information, https://blast.ncbi.nlm.nih.gov/) and compared with the identi ed species using BLAST (Basic local alignment search tool). The phylogenetic tree was built with MEGA7 software (version 7.0.14, Mega Limited, Auckland, New Zealand).

Preparation of the bioaugmentation inoculation
The bioaugmentation inoculation strategy included three steps which were 1) growth of mono-culture, 2) preparation of pure starter, and 3) application of bioaugmentation.
First, strains were cultured in NB medium at 37 o C, 220 rpm for 24 h. The growth of mono-culture reached the 7-8 log CFU/mL. Then, the pure starters were inoculated at 5% (v/v) of mono-culture. Strain with high alpha-amylase activity was cultivated in the optimized medium (soybean meal powder 24 g/L, ammonium sulfate 6 g/L maize our 1 g/L, calcium chloride 0.3 g/L). Strain with high protease activity was grown in the optimized medium (glucose 3 g/L, wheat our 7 g/L, corn extract 25 g/L, yeast extract 7.5 g/L, sodium chloride 0.1 g/L). All pure starters were incubated at 37 o C, 220 rpm for 24 h. At last, the suspension of starts with cells and enzymes was sprayed evenly onto the surface of FCT at a 20% (v/w) inoculum. Then, FCT with bioaugmentation was cultured in a 500-mL wide-mouth bottle sealed with eight layers of gauze and stirred every 4 hours. After culturing for 1, 1.5, 2, 2.5 and 3 days, FCT samples were frozen immediately in liquid nitrogen, collected into sterile polyethylene bags and stored at -20 o C for further analysis. In addition, the control group was treated with sterile distilled water under the same conditions and steps as above.

Evaluation of sensory quality
The assessment panel comprised two female and ve male quali ed professional assessors, 25-40 years old. According to the Chinese tobacco industry recommended standards (YC/T138-1998, YC/T496-2014), the sensory quality of FCT samples was evaluated using eight evaluation indexes. In which, aromatic quality, aromatic intensities and pleasant odor indicate the fragrance attributes, while smoke strength and smoke intensities indicate the avor attributes; Softness, aftertaste and sweetness were used to evaluate the mouthfeel of smoking. The score of sensory evaluation was the sum of all indicator scores.

Measurement of chemical indicators in FCT
The bioaugmentation samples were ground into powder at 60 Hz for 90 s using a grinder (TL-48R, Jingxin, ShangHai, China). The chemical indicators were measured with Bruker FT-NIR spectrometer (Flyer MATRIX-F, Bruker, Germany), referring to Wang's method (Wang C. et al., 2018).

Analysis of volatile compositions in FCT
Volatiles were investigated with untargeted metabolomics techniques based on the combination of the Agilent 7890-Pegasus HT GC/MS system and Agilent DB-5MS column (30 m×250 µm×0.25 µm) (Agilent, USA; LECO, USA).
A 2.00 ± 0.01 g of the homogenized FCT powder into the 20 mL headspace vessel was extracted for 30 min at 60 o C with the SPME ber (DVB/CAR/PDCS, divinylbenzene / carboxen / polydimethylsiloxane, 50/30 µm) (Supelco, Inc., Bellefonte, PA, USA). The chromatographic conditions were that: helium as the carrier gas owed at 1 mL/min rate; an injection port temperature was 250 o C; the heating held at 40 o C for 2 min and increased at a rate of 10 o C pre minute to 250 o C for 6 min; the ion source was an adopted electron bombardment model with electron energy of 70 eV; the ion source and transmission line temperature were 210 o C and 280 o C, respectively. The mass spectrometry data was collected in full-scan mode with the range of 33-400 atomic mass units at a rate of 10 specs/s after a solvent delay of 3 min.
At last, volatiles were identi ed comparing with the Wiley library (NY, version 9.0) and the MS library of National Institute for Standards and Technology (NIST, Search Version 1.6) based on the matching of the mass spectrometry (MS) information. The compositions with a matching score of over 700 were reserved for further analysis.

Statistical analysis
All samples conducted at least in triplicate and data presented as mean ± standard error of the mean (SEM). The difference of volatiles in samples was assessed with MetaboAnalyst 5.0 website (https:// www. metaboanalyst.ca/MetaboAnalyst/home. Xhtml) or SIMCA software (Sweden, Umetrics, v 14.1). Analysis of variances was carried out with GraphPad Prism 6 (GraphPad Software, San Diego, CA, USA) and p < 0.05 by paired t-test was considered statistically signi cant.

Acquisition of functional microorganisms
In order to ease the adverse effects of high starch and protein in FCT on sensory quality, we screened the functional strains from microbial consortia of FCT with ow cytometry. Following the single-cell sorting scheme (Fig. 1A), there was a high yield and activity of single cells after staining of dead and live cell and division of screening area according to scatter plot (FSC, SSC). The number of holes with single-cell clones was about 50-80 pre 96-well plate. Strains were transferred into deep-well plates with selection media for multiplication and then analyzed their functions. At last, the strain with the highest activity of alpha-amylase was selected out and name as LB. The strain with the highest activity of neutral protease was named as SC. BLAST analysis of the 16S rDNA gene sequence, strain LB had the highest sequence identity (99.86%) with Bacillus amyloliquefaciens strain 4 (GenBank accession number MG822731.1). Strain SC possessed the highest sequence identity with Bacillus kochii strain WCC 4581 (GenBank accession number FR845720.1). The phylogenetic tree was shown in Fig. 1B. The nucleotide sequence determined of strain LB and SC submitted to GenBank and the accession number is respectively MZ197997 and MZ198211 (https://submit.ncbi.nlm.nih.gov/subs/search=SUB9650406; https://submit.ncbi.nlm.ni h.gov/subs/search = SUB9650898).
Flow cytometry sorting is a powerful tool for monitoring, screening and separating single cells (Vitelli et al. 2021). Compared with traditional cultural methods, ow cytometry can isolate the individual cells with high precision, high throughput and low cost . According to Wang's studies (2018), the bacterial diversity of FCT in aging process showed Bacillus sp. was the dominant genera. It is implied that Bacillus sp. played an unignored role in the improvement of FCT quality. Bacillus sp. have been applied in many elds for rapid reproduction and secreting a variety of enzymes (Seiler et al. 2012), such as alpha-amylase ( Ullah et al. 2021), protease (Yogesh and Halami 2015), lipase (Adetunji and Olaniran 2021), cellulase (Araújo et al. 2021) and so on. B. amyloliquefaciens was rst isolated from the soil and named after its noteworthy character of producing the alpha-amylase (Karuppiah et al. 2019a). B. kochii has been isolated from some food sources, manufacturing sites (Seiler et al. 2012) and soil (Liu et al. 2017). Feizi et al. (2020) rst reported that B. kochii isolated from compost had the ability of phenanthrene biodegradation, however, there are not more studies. In our study, B. kochii was rst found that had a high neutral protease activity and co-cultured with B. amyloliquefaciens to promote the FCT sensory quality.

Sensory quality responses to mono, co-cultivation
The FCT sensory quality of mono, co-cultivation at different time had been dynamically investigated as shown in Fig. 2. There was the most obvious improving quality with B. amyloliquefaciens LB monoculture at two days ( Fig. 2-a), which the indicators of sweetness, aftertaste, pleasant odor, aroma quality and intensities had signi cant changes, but the indicator of softness related to irritation was only a slight change. In B. kochii SC monoculture, the indicators of softness, aroma quantity and pleasant odor had the most pronounced improvement at 2.5 days (Fig. 2-b), however, sweetness and aftertaste increased slightly. In order to achieve a further quality enhancement, the co-culture with B. amyloliquefaciens LB and B. kochii SC fermented at different ratios for 2 days (Fig. 2-c). Interestingly, B. amyloliquefaciens LB-B. kochii SC co-culture at a ratio of 3:1 had the higher sensory evaluation than one of monoculture, and there was a very signi cant difference between co-cultivation and control group (two-tailed, p < 0.0001). As expected, B. amyloliquefaciens LB-B. kochii SC co-cultivation has the obvious quality improvement at 2 days ( Fig. 2-d). FCT quality demonstrated a more comprehensive enhancement in terms of different indicators, which implied a synergistic effect of two strains on the quality. The accumulation of amino acids and reducing sugars is the basis for the formation of the products of the Maillard reaction that enhances softness and aromatic intensities of FCT by offering a nutty, sweet, or popcorn avor (Hinneh et al.2018).
The quality differences in the effects of bio-inoculation on FCT were closely related to the secreted enzymes. The functional enzymes of bio-inoculation were investigated dynamically in Fig. 3. In the control group, there were the low alpha-amylase activity and almost no neutral protease activity. The alpha-amylase activity of B. amyloliquefaciens LB monoculture was higher than that of B. kochii SC. The alpha-amylase achieved the maximum at 3 days of fermentation, while the neutral protease activity was almost undetectable except at 2.5 days. B. kochii SC can also secrete a certain amount of alpha-amylase with the maximum at two days. Both alpha-amylase and protease activities of co-cultivation were much higher than those of the monoculture. The highest enzyme activity was obtained at 2 days of fermentation, which was consistent with the optimal time point of quality evaluation. It suggested that the co-cultivation of B. amyloliquefaciens LB and B. kochii SC synergistically promote the functional enzyme activity. Previous studies have reported that B. amyloliquefaciens has an eco-friendly effect with other strains. Valliappan K. et al. (2021) thought that the co-cultivation with B. amyloliquefaciens induced the enzymes related in metabolic pathways by offering differentiated cellular environs. B. amyloliquefaciens used in the co-cultivation with rhizobacteria (Chowdhury S. P. et al. 2015) and T. asperellum (Karuppiah V. et al., 2019 b) by secreting multiple hydrolytic enzymes. In our work, the cocultivation of B. amyloliquefaciens and B. kochii has exhibited the higher enzyme activity and avorrelated compounds content than mono-cultivation.

Changes of avor-related compositions in FCT with mono, co-cultivation
Through bio-inoculation, sensory attributes of FCT were enhanced. It is closely related to the impact of strains on FCT compositions. Basic chemical indicators were changed by bio-inoculation. The content of total alkaloid, total nitrogen, starch decreased, while reducing sugar and total sugar increased (Table 1). There was the high level of reducing sugar and starch reduction in B. amyloliquefaciens LB monoculture and co-cultivation, and the minimum content of total nitrogen in B. kochii monoculture. Sugar-base ratio, the ratio of total sugars to total alkaloid, is one indicator used to characterize the quality of FCT. According to Han et al. (2010), FCT with sugar-base ratio is close to 10, implying excellent quality. It means that the quality of FCT was well improved by bioaugmentation. Meanwhile, volatiles played an important role in the avor of FCT (Yin et al. 2016). We pro led volatiles of FCT with untargeted metabolomics techniques, that got extensive data to pro le the chemical diversity and identify differential compounds. 163 compositions were detected, including 40 heterocyclic compounds, 38 carbonyl compounds, 28 alkanes or ole ns, 22 alcohol or phenols, 12 esters, 13 benzenoids and 6 acids and others (Table S1). The three-dimensional scatter plot of PCA (principal component analysis) demonstrated the rst three groups explain 40.5%, 23.9% and 12.9% of the total variance, respectively (Fig. 4A). The spread of data distribution suggests that a signi cant difference in compositions between bio-inoculation and control group. According to the OPLS-DA (orthogonal partial least-squares-discriminant analyzes) models ( Fig. S1 A-B), the 38 difference compounds were identi ed out by VIP (pred.) value > 1.0 (variable importance for predictive components) (Fig. S1 C) and p value < 0.05. The difference compounds included 10 heterocyclic compounds, 7 carbonyls, 8 alcohols, 3 alkanes, 2 ole ns, 3 acids, 2 esters, and 3 benzenoid (Fig. 4B), in which there were 4 compounds related to the metabolism of mono terpenoids, 9 compounds related to the phenylalanine degradation and 10 compounds associated with the Maillard reaction. The bioaugmentation of B. amyloliquefaciens LB and B. kochii, could promote Maillard reaction (Gong et al. 2021. There were higher content of Maillard reaction products in co-culture than ones in monoculture and the control. Four compounds were exhibited in Fig. 4-c. 2-pentyl-furan was the major aromatic contributors possessed fruity, roasted and sweet aroma (Ni et al. 2021). 1-(2-furanylmethyl)-1H-pyrrole was formed from cysteine and ribose by Maillard reaction and with coffee, green, hay aroma (Xu et al. 2013). Although the content of furfural was not the highest in the co-cultivation, the difference was not signi cant between B. kochii SC monoculture and co-cultivation. Furfural is a potent aroma compound leading to changes in the avor, texture, and color of food (Gong et al. 2021) and is a common present in coffee, peated malt, Bourbon vanilla, and others (Bueno et al., 2016). 2,5-dimethyl-pyrazine with a cocoa, grass, and medicinal tasting, had the high level in B. amyloliquefaciens LB monoculture and cocultivation.
In tobacco, terpenoids and their degradation products are also very important avors, especially the degradation products of carotenoids (Lewinsohn et al. 2005). Linalool is a sweet and earthy tasting compound and afforded oral and fruity odors (Gong et al. 2021). Dihydrokiwi lactone gave deep and mild odors and can mask bad avor (Mitsuya S., Hiroko S., et al. 1995). 3-hydroxy-β-damascone is also main aromatic compounds with intense aromas of fruit, wood, and violet , and could be biotransformed by microorganisms (Schwab, E. & Schreier, P. 1991). However, most of the alcohols and acids presented the highest levels in the monoculture samples. In benzenoids, benzyl alcohol, phenylethyl alcohol and benzeneacetaldehyde presented the high level and other benzenoids were degraded or transformed in the bio-cultivation. Summary, the high concentration of avor-related compounds was bene cial for the improvement of sensory quality in bioaugmentation, especially in the co-culture of B. amyloliquefaciens LB with B. kochii SC.

Conclusions
To improve the sensory quality of FCT, functional microorganisms were screened and performed bioaugmentation with co-cultivation. There were several ndings in our study. First, with the technology of ow cytometry single-cell sorting in 96-well plate, B. amyloliquefaciens LB with high alpha-amylase and B. kochii SC with high neutral protease were obtained e ciently from the FCT micro ora. Second, via bioaugmentation, B. amyloliquefaciens monoculture at 2 days and B. kochii monoculture at 2.5 days achieved the optimum quality by improving the different sensory indicators. B. amyloliquefaciens-B. kochii co-culture at a ratio of 3:1 demonstrated a more comprehensive quality enhancement than monoculture at 2 days. Finally, we found that the co-cultivation has the higher functional enzyme activity along with avor-related volatile compoundsn, which was implied that there was a good synergy on quality improvement of FCT. 38 differential compounds were analysis through OPLS-DA model. In co-cultivation, most of Maillard reaction products and terpenoid metabolites were at a higher level than others, which con rmed the hypothesis of screening protocol. However, the interaction mechanism between two strains needed further study. In this study, it provided a new reference for the function expansion that B. kochii was screened and application in FCT. In this study, we presented an effective bioaugmentation technique for changing the sensory attributes of FCT, which will be signi cance for improving raw material utilization and reducing costs in tobacco industry.

Declarations
Author contribution statements Xinying Wu: design of the study, analysis and interpretation of data, writing -original draft. Pengcheng zhu: providing sensory evaluation data. Dongliang Li: design of the study.  Sensory quality scores were evaluated by seven quali ed professional assessors. Figure 3