Tremella aurantialba SCT-F3 (CGMCC No.23831) was isolated from Naematelia aurantialba (Bandoni & M. Zang) Millanes & Wedin and deposited at the China General Microbiological Culture Collection Center.

The fermentation medium was potato glucose broth containing 6 g/L potato extract powder and 20 g/L glucose and was autoclaved at 121°C for 15 min. T. aurantialba SCT-F3 was inoculated by sterile loops in 250-mL flasks with 75 mL fermentation medium and cultured at 150 rpm/28°C for 5 days. The fermented potato glucose broth was centrifuged (5,000 × g for 10 min) to remove mycelium pellets, and the supernatant was used as the fermentation medium. Unfermented potato glucose broth was used as the unfermented medium. Dexue No.1 from the original Chinese tobacco drying strips was used in this study. CFLs were loosened and spread flat on the table by hand. Mixed the medium and pure water in a ratio of 1:3 (v/v), then the mixture was evenly sprayed using an electronic sprayer on the surface of the leaves with 20% inoculation amount (v/g). After 2 h, the medium and water were absorbed entirely by the CFLs. The CFLs were transferred to an oak barrel for fermentation. The oak barrels were placed in the fermentation room, the temperature of the fermentation room was controlled at 37°C, the relative humidity of the fermentation room was controlled at 75%, and the fermentation was carried out for 40 days. CFLs added with T. aurantialba SCT-F3 fermented medium were sampled at 0, 20, and 40 days and marked as J0, J20, and J40; CFLs added with unfermented potato glucose broth fermented were sampled at 0, 20, and 40 days and marked as P0, P20, and P40; uninoculated CFLs were marked as Control. Samples were taken from the four corners and central position of the oak barrel, and the five positions of the CLTs were ground and mixed for testing. All samples were mixed well, transferred into sterile bags, and stored at −20°C until further analysis. All experiments were performed in triplicates.

The fermented CFLs were stemmed and dried to 15 ± 2% moisture content. The samples were hand-rolled to a length of 110 mm and a circumference of 47 mm, then placed in a temperature humidity chamber at a temperature of 20 ± 2°C and humidity of 62 ± 3% to balance the samples’ moisture to 13 ± 1%. Five assessors from the staff trained by the China National Tobacco Quality Supervision & Test Center evaluated the sensory quality of all samples to recognize the aroma, smoke, aftertaste, combustibility, and comfort level. The samples were identified and scored according to 0–9 points, ranging from weak to strong.

The chemical components of total alkaloids, total nitrogen, reducing sugars, and total sugars in the CFLs were determined according to the continuous flow (potassium thiocyanate) method of the Tobacco Industry Standard for the determination of total alkaloids (YC/T 468–2013), total nitrogen (YC/T 161–2002), and water-soluble sugar (YC/T 159–2019) (Hu et al., 2022).

VFCs in CFLs were assayed using headspace solid-phase microextraction coupled with gas chromatography–mass spectrometry (HS-SPME-GC–MS) as previously described (Zheng et al., 2022a,b). A 50/30 μm DVB/CAR/PMDS fiber (Supelco Inc., United States) was used for aroma extraction. The CFLs were pulverized using a grinder, and the powder (1.5 g) was placed in a 10–mL glass vial. The sample was extracted at 60°C for 30 min, and the fiber was inserted into the injection port of the GC device at 250°C for 1 min to desorb the analytes. The sample was analyzed on a DB-5MS column (60 m × 0.25 mm × 0.25 μm, Agilent Technologies, United States). The GC oven temperature was maintained at 40°C for 2 min, followed by an increase of 10°C/min to 250°C and then held for 5 min. Helium (purity: 99.999%) was used as the carrier gas at a constant flow of 1 mL/min. The mass spectrometer conditions were: EI voltage, 70 eV; ion source temperature, 300°C; mass range, m/z 35–350. The identification of volatile compounds in the extracts was based on a comparison of the mass spectra and retention times of the individual compounds with those of standard compounds deposited in the National Institute of Standards and Technology database in the MS device (SI and RSI ≥800). Peak areas were compared with those of the internal standard for quantification.

Samples (5.0 g) were suspended in 250 mL sterile phosphate buffer saline and shaken for 2 h at 200 rpm, after which the supernatant was centrifuged (10,000 × g for 30 min). The total genomic DNA of each sample was extracted using an EZNA^® Soil DNA Kit (Omega, United States) according to the manufacturer’s instructions. The bacterial genomic DNA were amplified and sequenced using primers 515F and 907R for 16S rRNA genes (Parada et al., 2016). The fungal genomic DNA was amplified using primers ITS1F and ITS2R for internal transcribed spacer (ITS) genes (Usyk et al., 2017). Amplicons were pooled in equal amounts and sequenced using a 2 × 300 paired-end configuration and an Illumina MiSeq sequencing system (Illumina, United States). Operational taxonomic units (OTUs) of qualified sequences were identified using the clustering program VSEARCH version 1.9.6 and the SILVA version 132 database (Rognes et al., 2016) with 97% similarity. Alpha diversity, including the Chao1, Shannon, and Simpon values, was analyzed using QIIME version 1.9.1 (Caporaso et al., 2010). Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt2) and Fungi Functional Guild (FUNGuild) were used to predict metagenomic functions based on normalized OTU tables (Douglas et al., 2019; Xie et al., 2021).

All experiments were repeated at least thrice. SPSS version 19 software (SPSS Inc., United States) was used to carry out a one-way analysis of variance and Duncan’s multiple comparison test (p < 0.05). Heat maps and cluster analyses were performed using R version 4.0.0. Galaxy 1 was used for the linear discriminant analysis of effect size (LEfSe) to assess significant differences in CFLs with different treatments. SIMCA-P version 13.0 software (Umetrics, Sweden) enabled a partial least-squares (PLS) analysis of the contributions of conventional chemical constituents and VFCs to the sensory evaluation. Additionally, to determine the correlation between representative microbes and core VFCs based on Spearman’s correlation coefficients (p < 0.05), network analysis was performed using the Gephi software.

Seven CFLs were evaluated and scored according to the cigar evaluation criteria. The detailed evaluation scores for each sample are shown in Figure 1. The scores of the fermented CFLs were higher than those of the control, and as fermentation progressed, the quality of the CFLs improved gradually. Except for combustibility, other qualities were significantly improved (p < 0.05) in those with T. aurantialba SCT-F3 fermentation medium (J0, J20, and J40) compared to those in potato glucose broth (P0, P20, and P40). Specifically, aroma mellow, richness, maturity, and sweetness increased; irritation decreased, and balance sense improved.

During fermentation, the contents of total alkaloid, total nitrogen, total sugar and reducing sugar fluctuated from 1.8–2.8%, 4.3–5.4%, 0.2–1.7% and 0.2–1.4%, respectively. The total alkaloid, reducing sugar, and total sugar contents significantly increased (p < 0.05) in samples supplemented with fermentation media on day 0 (F0 and P0), as shown in Figure 2. They then decreased to their lowest levels on day 20 (J20 and P20) and gradually increased (J40 and P40). The total nitrogen content at 20 and 40 days (J20, P20, J40, and P40) was higher than that on day 0 (J0 and P0) and in control.

In total, 188 volatile components were detected in the CFLs, including 32 ketones, 20 aldehydes, 11 alcohols, and 10 esters. Furthermore, 34 VFCs (Figure 3), five aromatic amino acid degradation products, six plastochrome degradation products, one cembrane degradation product, 10 Maillard reaction products, and four nicotine degradation products were selected for further analysis. These compounds have been reported to have different aromas that may play important roles in the aroma profiles of cigars. The contents of aromatic amino acid degradation products, plastochrome degradation products, and cembrane degradation product, decreases gradually with fermentation, whereas four Maillard reaction products and 3 nicotine degradation products increased.

Cluster analysis divided the unfermented (control, P0, and J0) and fermented (J20, P20, J40, and P40) CFLs into two clusters, and the fermented CFLs were subdivided into two groups: P20 and J20, P40 and J40. The addition of fermentation medium caused a decrease in eight VFCs and an increase in 14 VFCs. Compared to the control, the 12 components of J0 increased. Styrene and 2-methyl-propanal levels peaked at J20. The vast majority of the VFCs changed after fermentation. Compared to the control, 15 VFCs in J40 decreased, 13 VFCs increased, 18 VFCs in P40 decreased, and 11 VFCs increased. Specifically, 2,3-butanedione, methylpyrazine, 2,6-dimethylpyrazine, 2-ethyl-6-methylpyrazine, β-nicotyrine, D,L-anatabine, cotinine, and ethyl acetate were enriched on day 40.

PLS was performed on scaled values of chemical constituents and VFCs as X-variables and on scaled values of total sensory evaluation scores as Y-variables for correlations. Two significant principal components of the total variance in the data matrix were extracted. The R2X (cum), R2Y (cum), and Q2 (cum) were 0.522, 0.808, and 0.701, respectively, which meant that 52.2% variation was due to these two components, with a total of 80.8% dummy Y-variables per class and 70.1% overall cross-validated R² for these two components. The data indicated that the PLS model was suitable for this study. The distributions of the samples in the first and second components of the statistical analysis are shown in Figure 4. Groups J0 and P0 are located on the right side of the plot, J20 and P20 on the lower left, and J40 and P40 on the upper left. Table 1 shows that 14 variables with a VIP higher than 1.0 were important in the first and second most significant principal components. 2-ethyl-6-methylpyrazine, methylpyrazine, D,L-anatabine, β-nicotyrine, nicotinic degradation products, and total nitrogen were significantly positively correlated with the total sensory evaluation scores. In contrast, 1-phenylethanol, geranylacetone, megastigmatonone, D-solanone, 6-methyl-3,5-heptadiene-2-one, Maillard reaction products, 1,2-dimethoxybenzene, and γ-caprolactone were significantly negatively correlated with the total sensory evaluation scores.

Figure 5A shows alpha diversity indices, including the Chao1, Shannon, and Simpson indices, of bacterial and fungal communities in CFLs. The Chao1, Shannon, and Simpson values of the bacterial communities were higher than those of the fungal communities, indicating that the richness and diversity of the bacterial communities were generally higher than those of the fungal communities. The Chao1 values of J0 were higher than those of others, indicating that the richness of the bacterial and fungal communities increased after adding the T. aurantialba SCT-F3 fermentation medium. The Shannon and Simpson values of microbes in CFLs with medium without fermentation (P0 and J0) were higher than those in the other samples, indicating that adding fermentation medium increased the diversity of the microbial community. As fermentation progressed, fungal alpha diversity indices decreased.

Unconstrained principal coordinate analysis of weighted UniFrac distance revealed that the microbiota of CFLs with different treatments explained 77.4% of the bacterial variance (Figure 5B) and 98.1% of the fungal variance (Figure 5C). The addition of the fermentation medium altered the structure of the bacterial and fungal communities. The fungal community structure of sample J20 significantly differed from the other samples.

The abundance of bacterial and fungal taxa is shown in Figure 6. The dominant bacterial phyla were Bacillota and Pseudomonadota, and the dominant fungal phyla were Ascomycota and Basidiomycota (Figures 6A,B). Pseudomonadota and Ascomycota of unfermented samples (control, P0, and J0) were significantly higher than those of the fermented samples. The dominant bacterial genera, with relative abundances higher than 5.0% in at least one sample, were Staphylococcus, Burkholderia-Caballeronia-Paraburkholderia, unclassified Enterobacteriaceae, Pantoea, Aerococcus, Pseudomonas, and unclassified Burkholderiaceae, which represented between 44.6 and 96.1% of the total abundance in each CFL sample (Figure 6C). The relative abundance of Staphylococcus decreased significantly at P0 and J0. The dominant fungal genera were Aspergillus and Alternari (Figure 6D). The relative abundance of Aspergillus and Alternari significantly decreased in P0 and J0, whereas the relative abundance of Stemphylium showed the opposite trend.

LEfSe analyses revealed significant differences below the phylum level to explore the different microbes among the CFLs under different treatments (Figure 7). The circles from innermost to the outermost represent bacterial and fungal classifications from the phylum to genus levels, and the corresponding colors in every group denote bacterial and fungal taxa with a significant difference. Notably, 85 different bacteria appeared at the LDA threshold of 2, judging by statistically significant differences (p < 0.05), which consisted of two phyla, four classes, 10 orders, 22 families, and 47 genera (Figure 7A). In detail, 30 genera were significantly enriched in inoculated T. aurantialba SCT-F3 CFLs without fermented (J0), such as Escherichia-Shigella, Luteimonas, and Brachybacterium. Four genera were significantly enriched in P0: Brevundimonas, Clostridium sensu stricto1, Variovorax, and Moraxellaaceae. Two genera were significantly enriched in J20, including Oceanobacillus and Curvibacter. Eight genera were significantly enriched in J40, such as Aquabacterium, Burkholderia-Caballeronia-Paraburkholderia, and Prauserella. There were 12 fungi of two classes, one order, four families, and five genera (Figure 7B). Colletotrichum and Symmetrospora were significantly enriched in J0 and P0, Sagenomella was enriched in J20, and Aspergillus in J40.

Function predictions derived from the microbial community data were obtained using PICRUSt for bacterial functional genes and FUNGuild for fungal functional genes to analyze microbial function (Figures 8A,B). The abundance of functional gene sequences in fungi was much lower than in bacteria. The functions of bacteria and fungi in CFLs include nucleotide biosynthesis, amino acid biosynthesis, fatty acid and lipid biosynthesis, and fermentation. The abundance of many bacterial metabolic pathways increased significantly after fermentation, whereas fungal metabolic pathways showed the opposite trend. For samples with T. aurantialba SCT-F3 fermentation medium (J0, J20, and J40), the abundance of functional genes increased, peaked on the 20th day of fermentation, and then gradually decreased. The relative abundance of functional bacterial genes at P40 increased significantly. There were 65 functional genes up-regulated and 11 down-regulated in samples inoculated with T. aurantialba SCT-F3 (J0, J20, and J40) compared with samples inoculated with potato glucose broth (P0, P20, and P40) (Table 2 and Figure 8C). The abundance of functional genes related to amino acid metabolism, fermentation, and fatty acid degradation increased after adding the microbial medium.

Furthermore, a metabolic network, including nicotine degradation (M00810 and M00811) and nicotinate degradation (M00622) according to KEGG,¹ during fermentation was constructed using the annotated enzymes and their metabolic pathways (Figure 9). The types and abundances of bacterial enzymes were higher than those of fungi. The enzymes associated with nicotine degradation were less abundant than those associated with nicotinate degradation. In the nicotine degradation pathway, EC:1.5.99.4 (nicotine dehydrogenase subunit A/B/C, ndhA/B/C), EC:1.4.2.2 (nicotine dehydrogenase, nox/nicA2), EC:1.5.3.5 ((S)-6-hydroxynicotine oxidase, nctB), EC:1.4.2.3 (pseudooxynicotine dehydrogenase, pao), EC:1.5.3.6 ((R)-6-hydroxynicotine oxidase, 6-hdno), EC:1.2.1.83 (3-succinoylsemialdehyde-pyridine dehydrogenase, sap), and EC:1.5.99 (3-succinoylpyridine monooxygenase) were not detected in any sample. The abundance of EC:1.5.99.14 (6-hydroxypseudooxynicotine dehydrogenase subunit alpha, kdhA), EC:1.14.13.10 (2,6-dihydroxypyridine 3-monooxygenase), and EC:1.5.3.19 (4-methylaminobutanoate oxidase (formaldehyde-forming)) increased in J0 and gradually decreased during fermentation. In the nicotinate degradation pathway, the abundance of EC:1.13.11.9 (2,5-dihydroxypyridine 5,6-dioxygenase, nicX), EC:3.5.1.106 (N-formylmaleamate deformylase, nicD), and EC:3.5.1.107 (maleamate amidohydrolase, nicF) showed the same trend. The abundance of microbial enzymes annotated to succinate semialdehyde transferred into the gamma-aminobutyrate (GABA) shunt, dicarboxylate-hydroxybutyrate cycle, and hydroxypropionate-hydroxybutylate cycle, and fumarate transferred into the citrate cycle, urea cycle, tyrosine degradation, reductive citrate cycle, dicarboxylate-hydroxybutyrate cycle, hydroxypropionate bicycle, incomplete reductive citrate cycle, and arginine biosynthesis were higher than the abundance of enzymes involved in nicotine degradation and nicotinate degradation pathways.

Spearman’s correlation coefficients (p < 0.05) were used to establish significant associations between the 25 dominant microorganisms and 34 VFCs (Table 3). Acinetobacter (M1), Aerococcus (M2), and Oceanobacillus (M14) were not significantly correlated with VFCs. Moreover, 2-Ethyl-6-methylpyrazine, 2-phenylethanol, 4′-methylacetophenone, acetoin, acetophenone, and styrene were not significantly correlated with microorganisms. Graphs of the positive and negative-correlation networks were visualized using Gephi, as shown in Figures 10A,B, respectively. The modularity index of the positive correlation graph was 0.452 (>0.4), suggesting that the network had a modular structure. Twenty genera were significantly correlated with 21 VFCs values. A total of 41 notes and 72 edges (pairs of significant and robust correlations) were identified. There were six highly connected genera (degree ≥5), including Pantoea, Paracoccus, Stenotrophomonas, Mycosphaerella, Stemphylium, and unclassified Eurotiales, belonging to Pseudomonadota and Ascomycota. Seven highly connected VFCs, 1-octen-3-one, geranylacetone, sulcatone, 2,3-butanedione, benzeneacetaldehyde, 2-hexenal, and decanal, belong to plastochrome degradation products, Maillard reaction products, aromatic amino acid degradation products, and others. Nicotinic degradation products, cotinine, and 2,4,6-collidine were significantly and positively correlated with Burkholderia-Caballeronia-Paraburkholderia and Prauserella, Mycosphaerella, and Stemphylium, respectively. Gamma-valerolactone was significantly and positively correlated with Mycosphaerella and unclassified Eurotiales; geranylacetone with Pantoea; and unclassified Eurotiales, Alternaria, Paracoccus, and Stemphylium, neophytadiene, and megastigmatrienone with Pantoea.

The modularity index of the negative correlation graph was 0.605 (>0.4), suggesting that the network had a modular structure. Fourteen genera significantly correlated with 17 VFCs. A total of 31 notes and 35 edges (pairs with significant and robust correlations) were identified. There were three highly connected notes (degree ≥5), including Aspergillus, Paracoccus, and 2,3-butanedione. Aspergillus was significantly and negatively correlated with D-solanone, acetic acid, 2-hexenal, benzeneacetaldehyde, valeric acid, γ-caprolactone, and decanal, which are cembrane degradation products and Maillard reaction products. 2,3-Butanedione significantly and negatively correlated with Mycosphaerella, Stemphylium, unclassified Eurotiales, Paracoccus, and Stenotrophomonas, belonging to Ascomycota and Pseudomonadota. The nicotinic degradation products, β-nicotyrine, and D,L-anatabine were significantly and negatively correlated with Stemphylium, Paracoccus, Mitochondria, and Luteimonas. Gamma-valerolactone was significantly and negatively correlated with Bacillus; ethyl acetate with Pantoea and unclassified Enterobacteriaceae; 2,6-dimethylpyrazine with Paracoccus; methylpyrazine with Paracoccus, Stemphylium, and Stenotrophomonas; and tetramethylpyrazine with Escherichia-Shigella.