Chemical composition changes during fermentation

Chemical composition changes during fermentation. The conventional chemical components of cigar tobacco leaf from three locations were analyzed before fermentation (F0) and at different fermentation stages (F1, F2, F3, F4, F5). The OriginPro 2023 was used to determine the mean difference between different fermentation stages using Tukey’s HSD. Two-way analysis of variance showed that the total nitrogen (F = 3.26, P = 0.001), total sugar (F = 4.038, P = 0.0025), reducing sugar (F = 12.591, P < 0.0001), starch (F = 8.726, P < 0.0001), total polyphenols (F = 3.000, P = 0.016), acid cellulose (F = 6.098, P < 0.0001), acid lignin (F = 4.800, P < 0.0001), pectin (F = 4.848, P < 0.0001), and protein (F = 7.699, P < 0.0001) in cigar tobacco leaves were significantly different during fermentation. The results showed that the TS, RS, SH, AC, AL, PT, TP, PN and pH in cigar tobacco leaves were significantly decreased after fermentation (Fig. 2). There were no significant changes in the contents of nicotine, potassium oxide, chloride ion, magnesium, petroleum ether extract and total amino acids.

Fig. 2

Changes in conventional chemical components and physical properties in cigar tobacco leaves during fermentation. The values in the Figure are the mean value of four biological replicates sourced from three locations (n = 12, P < 0.05)

Physical properties changed during fermentation

There were no significant changes in the physical properties including leaf filling value (LFV), leaf stem ratio (LSR), leaf equilibrium moisture content (water retention, LWR), leaf thickness (LTH), leaf tension (LTV), leaf density (leaf mass weight, LMW) during fermentation. However, with the prolongation of the fermentation period, the LWR showed a trend of gradual increase, and the LTV showed a trend of gradual decrease after repiling three times (Fig. 2).

Volative aroma compounds changed during fermentation

A total of 44 volatile aroma compounds (VACs) were further analyzed qualitatively and quantitatively based on GC–MS before and after the fermentation of cigar tobacco leaves. The VAC profiles were found to vary significantly during fermentation in leaves of Puer region (Fig. 3A). In the PCA diagram, the first two principal components (PC1 and PC2) were shown to explain 41.3% and 16.8% of the data variance in different fermentation stages, respectively. The 18 samples from the six fermentaion stages were divided into two distinct groups, suggesting that each group had a relatively distinct metabolic VAC profile (Fig. 3C). Group 1 included PEF0. Group 2 included PEF1, PEF2, PEF3, PEF4, PEF5 and PEF6 (Fig. 3D). These two groups could be easily distinguished from each other (Fig. 3A and B). The PCA and hierarchical clustering heatmap of VACs results showed that the VAC profiles were different various fermentation periods. In this study, orthogonal projections to latent structures discriminant analysis (OPLS-DA) analyses were performed to investigate the discriminatory VACs contributing to differences among before fermentation (F0) and after fermentation (F5) from three regions. we evaluated the differences using the OPLS-DA model between DHF0 and DHF5 (R2X = 0.703, R2Y = 0.909, Q2 = 0.843), between PEF0 and PEF5 (R2X = 0.521, R2Y = 0.98, Q2 = 0.905), between LCF0 and LCF5 (R2X = 0.419, R2Y = 0.629, Q2 = 0.471). According to an FC ≥ 1 or ≤ 0.5 and VIP ≥ 1, there were 19 differential metabolites between DHF0 and DHF5 (upregulated = 3, downregulated = 16), 26 between PEF0 and PEF5 (upregulated = 3 downregulated = 23), and 25 between LCF0 and LCF5 (upregulated = 6, downregulated = 20 (Fig. 3C). Among these, 9 volatile aroma compounds were potential markers among three locations, 3 were significantly up-regulated and 6 were significantly down-regulated, including furfural, neophytadiene, pyridine, benzyl alcohol, geranylacetone, 3-hydroxy-2-butanone, N-hexanal, 3-Methyl-1-butanol and 2,3-pentanedione (Fig. 3D).

Fig. 3

Multivariate statistical analysis of volatile aroma compounds (VACs) during fermentation of cigar tobacco leaves. (A) Score scatter plot for the principal component analysis (PCA) model (PE). (B) Hierarchical clustering analysis and heatmap visualization of VAC profiles at different fermentation stages of PE cigar tobacco leaves. (C) Important characteristics of different locations (DH, PE and LC) before and after fermentation identified by OPLS-DA. The colored boxes on the right indicate the relative concentrations of the VACs in each group under study. (D) The potential markers responsible for the aroma differences among DH, PE and LC. The screening criteria were VIP > 1 and FC > 1.

Diversity of the microbial community in the cigar Tobacco leaves

This work aimed to evaluate the changes in the microbial community of cigar tobacco leaves during fermentation stages. A total of 4,091,288 bacterial 16 S rRNA and 4,197,729 fungal ITS sequences were obtained from 54 samples of cigar tobacco leaves, and the rarefaction curve tended to be flat, indicating that deep sequencing provided good overall operational taxonomic unit (ASV) coverage (Fig. S1). Among all the ASVs identified in this study, the bacteria shared by all three locations had 111 ASVs, mainly at the genus level of Acinetobacter, Sphingomonas, Enterobacterales, Stenotrophomonas, Bacillus, Corynebacterium, Pseudomonas and Staphylococcus. The PE, LC and DH had 4152, 743, 89 unique ASVs, respectively (Fig. S2A). For fungi, three locations shared 111 ASVs, mainly at the genus level of Aspergillus, Penicillium, Alternaria, Cladosporium and Fusarium. PE, LC, and DH had 337, 454 and 149 unique ASVs, respectively (Fig. S2B).

Microbial succession during fermentation

The diversity and similarity of the microbial community composition in the cigar tobacco leaves varied with the fermentation stages and locations. The α-diversity results showed that the diversity (Shannon index) of bacteria initially increased and then decreased. The fungi communities in our study tended to fluctuate continuously in α-diversity (Fig. 4A). The fungi α-diversity of the PE (F (5,12) = 4.188, P = 0.02) and the bacterial α-diversity (F (5,12) = 41.93, P < 0.001) was significantly reduced (Table 1). but the fungi and bacteria α-diversity of the LC and DH did not change significantly. The fungi and bacteria α-diversity did not change significantly in three locations (Table 1). With the extension of fermentation period, the α-diversity of fungi increased gradually and then reached their maximum value at the fourth repiling (F4) stage (F (2,6) = 36.29, P < 0.001). The α-diversity of bacterial increased gradually and reached their maximum value at the third repiling(F3) stage (F (2,6) = 326.8, P < 0.001). The α- and β-diversity of fungi and bacteria of all stages was greater than all regions (Table 1). These results showed that the fermentation stages had greater impacts on the microbial diversity than locations. To evaluate the compositional differences of the microbial communities in the samples between locations at each stage, a principal coordinate analysis (PcoA) was performed using the weighted UniFrac metric (bacterial) and Bray-Curtis metric (fungal). Significant differences were observed in the β-diversity of fungi (R2 = 0.566, P = 0.001) and bacterial (R2 = 0.411, P = 0.001) communities at every single stage, and the locations had a little influence on the beta diversity (fungi, R2 = 0.071, P = 0.071; bacteria, R2 = 0.101, P = 0.383) (Table 1). The microbial community structures of both bacteria and fungi had a significant temporal succession pattern between the before(F0) and after(F5) fermentation stages, but there was no obvious separation among the F1, F2, F3 and F4 fermentation stages, or in the community structure among DH, PE and LC (Fig. 4B). The succession of the bacterial community was greater than the succession of the fungal community during fermentation (Fig. 4; Table 1).

Fig. 4

The changing trend of fungal and bacterial alpha diversity (Shannon index) during fermentation (A). The PCA graphs of the fungal community and the bacterial community at different fermentation stages and from different locations (B)

Table 1 Experimental factors predicting α- and β-diversity of microbial communities in cigar tobacco leaves sourced from different locationsDynamic changes in the core microbiota during fermentation

We first identified the core microorganisms of cigar tobacco leaves from three locations at different fermentation stages, which consisted of the most important taxa based on the abundance and occupancy distribution. There were significant differences in the number of core microbial communities in different producing areas, and the relative abundance and occupancy of these core members in different fermentation periods also varied (Table S2). The bacteria mainly included Staphylococcus, Pseudomonas, Ralstonia, Sphingomonas, Bacillus, Massilia, Fibrobacter, Acinetobacter and Massilia (Table S2 a, c, e). The main fungi were Aspergillus, Cladosporium, Trichomonascus, Alternaria, Penicillium, and Fusarium (Table S2 b, d, f). More bacteria than fungi were observed in these core members.

In tracking temporal dynamics and succession of core taxa during cigar tobacco leaf fermentation. The relative abundances of some core core bacteria and fungi taxa had similar variation trends among PE, LC and DH, generally showing a trend of first increasing and then decreasing. However, the accumulated relative abundances were significantly different at different fermentation stages (Fig. 4). For example, the dominant bacterial genera Pseudomonas and Staphylococcus were present in high abundance in all fermentation stages of PE, LC and DH (Fig. 5A, B, C). Figure 5D-F reveals fungal dynamics and succession during fermentation. After fermentation commenced, the relative abundance of many core fungal taxas including Aspergillus gradually increased, reached their maximum at the F3 and F4 fermentation stages, and then sharply decreased.

Fig. 5

The relative abundance of core microbiota at different fermentation stages. The top 10 relative abundances are shown. A, B, and C represent the relative abundance of core bacterial communities in PE, LC, and DH, respectively. D, E, and F represent the relative abundance of core fungal communities in PE, LC, and DH, respectively

To confirm the stability of the core taxa in different fermentation cycles and to establish a model to correlate the composition of the microbiota with the fermentation cycles, the random forest supervised learning model was used to regress the relative abundance of bacteria and fungi at the genus level against the fermentation time. The results showed that these core groups had good ability to discriminate samples at different stages. Bacteria were stronger than fungi, and could accurately identify 50% of PE samples, compared with 33.33% for fungi (Fig. 6A, C). Discrimination of the LC and DH was low, reaching only 33.33% and 16.67%, respectively (Fig. S3A, S4A). The fungal core taxa had lower resolutions for the PE, LC and DH samples, reaching 33.33%, 25.00% and 16.67%, respectively (Fig. 6C, Fig. S3C, Fig. S4C). These results indicated that the core taxa of different fermentation stages were significantly different (Fig. 6A and C, Table S2a-f). The Gini index model showed that many core taxa were important features of the fermentation period model. For example, Sphingomonas and Penicillium could explain the largest variation of bacterial and fungal communities in PE locations, respectively (Fig. 6B, D). Cupriavidus and Trichomonascus could explain the largest variation in bacterial and fungal communities in LC locations, respectively (Fig. S3B, D). Corynebacterium and Aspergillus could explain the largest variation in bacterial and fungal communities in DH locations, respectively (Fig. S4B, D). Taken together, these results confirmed that these microorganisms might be biomarkers associated with certain stages of fermentation and may reflect the functional characteristics at different stages.

Fig. 6

The core taxa can distinguish the fermentation stages of cigar tobacco leaves. Classification of random forest models of the fermentation stage of the core taxa of PE bacteria (A) and fungi (B). The important features (top 20) based on Mean Decrease Gini (MDG) of random forest models of the core taxa of PE bacteria (C) and fungi (D)

Relationship between physicochemical characteristics and microorganisms during fermentation

To investigate the relationship between metabolic enzymes, physicochemical properties and microbial community composition. Canonical correspondence analysis (CCA) was performed based on microbial abundance, metabolic enzymes and physicochemical properties. The results showed that bacterial and fungal communities had significant effects on metabolic enzymes, physicochemical properties and volatile components during leaf fermentation. The measured metabolic enzymes, physicochemical properties and volatile components could explain 35.25%, 42.77% and 31.19% of the bacterial community variation, respectively (Fig. S5A, B, C). The measured metabolic enzymes, physicochemical properties and volatile components could explain 45.88%, 42.00% and 36.65% of the fungal community variation, respectively (Fig. S5D, E, F). The metabolic enzymes: AL, NAD-MDH, PAL, LiP, GDH, NR and CAT were the most important factors closely associated with the bacterial and fungal community (Fig. S5A, D). The pH, PEE, TAA, PN, NT, TS, TN, Mg, K were the most important physicochemical properties closely associated with the bacterial and fungal community (Fig. S5B, E). The volatile components: VAC2、VAC3、VAC6、VAC7、VAC8、VAC11、VAC16 and VAC17 were the most important factors closely associated with the bacterial and fungal community (Fig. S5C, F).

To further explore the co-occurrence patterns between core taxa and metabolic enzymes, physicochemical properties and volatile compounds, molecular ecological networks were constructed. In general, a more densely connected module was observed in the bacteria than in the fungi (Fig. 7, Fig. S6). The network of bacteria and volatile aroma components, physicochemical properties, and metabolic enzymes consisted of 181, 38, and 21 associations, respectively, in which 113 edges were positive associations. Most core taxa belong to Aureimonas, Ralstonia, Skermanella, Methylobacterium-Methylorubrum, Ensifer, Steroidobacter, Staphylococcus, Bacillus and Microvirga. In this network, some microbial taxa had significant correlations with various volatile aroma compounds, especially cibai trienediol (VAC41-43), 1-penten-3-one (VAC1), 2-butanone, and 3-hydroxy (VAC4), which had high connectivity (Fig. 7A). Some microbial taxa had high connectivity with acid lignin (AL), nicotine (NT), amylase (AL), glutamate dehydrogenase (GDH), and malic dehydrogenase (NAD-MDH) (Fig. S6). The network of fungi and volatile aroma compounds consisted of 99 associations (Fig. 7B), 69 edges of which were positive associations. Most core taxa belonged to Aspergillus, Penicillium, Sampaiozyma, Trichomonascus, Fusarium, Wallemia, Cercospora, Pallidocercospora, Mortierella, Xeromyces, Filobasidium, Vishniacozyma, Sporobolomyces and Podospora. In general, some bacteria and fungi jointly participated in the metabolism of the fermentation process and played different roles.

Fig. 7

Network analysis based on the cooccurrence of volatile components (A, B), physicochemical properties (C, D), metabolic enzymes (E, F) and the bacterial and fungal communities. Purple vee nodes represent volatile components. Grey ellipse nodes represent microbial members. Direct connections between nodes indicate strong correlations (Pearson correlation coefficient, P < 0.05). Red lines represent positive interactions between nodes, and blue lines represent negative interactions. The sizes of vee nodes represent the interconnected degree. The sizes of the circle nodes represent the average relative abundances of bacteria and fungi