In this study, 23 fermented seafood samples were collected. Nine fermented fish samples were collected from the cities of Lingshui, Baoting and Qiongzhong; 5 fermented crab samples were collected from Wanning City; and 9 fermented shrimp samples were collected from the cities of Baoting, Ledong, Lingshui and Wanning in Hainan Province in China. The specific sampling areas were in the Supplementary Table S1. The sampling time was October 2017. Fermented seafoods were produced by local populations through natural fermentation and were collected from the local population after the fermentation ended. Fermentation usually ended in 10–15 days. After collection, these samples were placed in a cooler and immediately transported to the laboratory and stored at -20°C for DNA extraction. The pH value was measured for a uniform mixture of one gram of fermented seafood sample with 10 mL of a sterile NaCl solution (0.85%, w/v).

Ten grams of fermented seafood sample was mixed uniformly with 90 mL of a sterile NaCl solution (0.85%, w/v). On the one hand, the solution was used to dilute the mixture, which was then coated on MRS solid agar medium; On the other hand, the diluent of mixture was used to extract DNA directly for 16S rDNA sequencing.

A single colony was selected, and DNA was extracted after the second-generation of the culture. After being oscillated evenly, the supernatant was used for DNA extraction with a QIAGEN DNA Mini-Kit (QIAGEN, Hilden, Germany) and a bead-beating method (Tanaka et al., 2009). The quality of the extracted DNA was detected by 1% agarose gel electrophoresis, and DNA samples were stored at -20°C for further processing. The universal forward primer A27F (5′-GCAGAGTTCTCGGAGTCACGAAGAGTTTGATCCTGGCTCA-3′) and the reverse primer A1495R (5′-AGCGGATCACTTCACACAGGACTACGGCTACCTTGTTACG-3′) were used to amplify the gene. Sequencing of the single colony was performed in Shanghai Personal Biotechnology Company. The sequencing results of the single bacteria were compared with the NCBI database after being aligned with MEGA software and the identification results were obtained. After identification, the strains were stored at -80°C in the laboratory for later experiments (SRA accession: PRJNA517547).

The DNA of fermented seafood was extracted with a QIAGEN DNA Mini-Kit. The V3-V4 region of the 16S ribosomal RNA (rRNA) genes was amplified (Fouhy et al., 2016), and a set of 6-nucleotide barcodes was added to the universal forward primer 338F (5′-ACTCCTACGGGAGGCAGCA-3′) and the reverse primer 806R (5′-GGACTACHVGGGTWTCTAAT-3′) (Wang et al., 2018). The PCR products were quantified by using an Agilent DNA 1000 Kit and an Agilent 2100 Bioanalyzer (Agilent Technologies, United States). The Illumina MiSeq high-throughput sequencing platform was used to sequence the amplified products, which were pooled to a final concentration of 100 nmol/L at equimolar ratios. The 16S rDNA of fermented seafood samples were sequenced and the sequences of fermented seafood were submitted to the NCBI database (SRA accession: PRJNA507916).

We performed microbial community analyses in the QIIME platform (v1.7) using high-quality sequences (Caporaso et al., 2010b). PyNAST (Caporaso et al., 2010a) and UCLUST (Edgar, 2010) were used to align sequences and cluster under 100% sequence identity to obtain the unique V3-V4 sequence. Operational taxonomic units (OTUs) were classified using UCLUST after representative sequences were selected with a 97% threshold identity. The taxonomy of each OTU representative sequence was assigned using the Ribosomal Database Project (RDP) classifier with a minimum bootstrap threshold of 80% (Cole et al., 2007). A representative set of OTUs checked for chimeras and established in FastTree (Price et al., 2009) was used to construct a taxonomic tree for downstream analyses, including alpha and beta diversity calculations. To evaluate alpha diversity, the Shannon-Wiener and Simpson’s diversity indices were calculated, and the Chao1 and ACE indices were measured to estimate community richness. UniFrac metrics, which are used to measure phylogenetic distance, were calculated to evaluate the beta diversity between the sets of sequences collected from many different microbial communities, and both weighted and unweighted calculations were performed prior to principal coordinate analysis (PCoA). PICRUSt (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States) was used to predict the metabolic pathways of the microbiota (Langille et al., 2013). All statistical analyses were conducted using the R program. The PCoA results were visualized with the ggplot2 package (Ito and Murphy, 2013). The relative abundance of taxa was compared by the Kruskal-Wallis test based on the rarefied OTU subset (Elliott and Hynan, 2011). The network was drawn by Cytoscape software (v3.6.0) (Killcoyne et al., 2009).

In our previous research, a diverse selection of fermented vegetables samples [including fermented Chinese cabbages (FCC), fermented bamboo shoots (FBS) and fermented watermelons (FW)] was analyzed and various bacterial strains were isolated (Peng et al., 2018). A variety of bacterial strains was also isolated from fermented seafood, and 16s rDNA was sequenced and compared with the NCBI database. Of which, the 16S rDNA sequences of 124 Lactobacillus plantarum and 158 Lactobacillus fermentum strains were chosen for a comparison of SNPs (Qi et al., 2013).

We measured the pH value in the samples (Table 1). Interestingly, the average pH value of fermented shrimp was dramatically higher than in the other samples (P < 0.001), and no significant differences were observed among FBS, FCC, FW, FC and FF. The Shannon (Figure 1A), Simpson (Figure 1B), Chao1 (Figure 1C) and ACE (Figure 1D) indices were measured to evaluate the alpha diversity of the fermented seafood samples [including fermented fish (FF), fermented crabs (FC) and fermented shrimp (FS)]. The higher alpha diversity of fermented seafood indicated that the microbial species found were more abundant than those found in fermented vegetables (p < 0.001). The alpha diversity of FS was highest in fermented seafood, but in the fermented seafood samples, no significant differences existed among between FS, FC and FF. This result meant that a different alpha diversity existed among the different raw materials of the fermented species.

Based on principal coordinate analysis (PCoA) of the weighted UniFrac distance (Figure 2A,C) and the unweighted UniFrac distance (Figure 2B,D), we compared the intergroup differences between the fermented vegetables and fermented seafood samples to evaluate the microbial β diversity. Each point represents the microbial structure of one sample. Based on the weighted UniFrac distance, in the fermented seafood samples, the orange points of FF, the yellow points of FS, and the green points representing FC were not clearly separated (Figure 2A). However, in Figure 2C, blue points, representing fermented seafood (including FF, FS, and FC), clustered in the lower left section of the coordinate axis and green points, representing fermented vegetables (including FBS, FCC and FW), clustered in the upper right section of the coordinate axis, which showed that the structure of the microbial community was significantly different among the fermented seafood and fermented vegetable samples, and a significant separation (P < 0.001) in PC1 was observed based on both the weighted (Figure 2E) and unweighted (Figure 2F) Wilcoxon rank-sum tests.

The bacterial genera whose average relative content was more than 0.1% in all samples and were present in more than 85% of the samples were selected. We identified the dominant bacterial genera in samples (Figure 3), which included Lactobacillus, Pediococcus, Weissella, Bacillus, Lactococcus, Pseudomonas, Acinetobacter, Rummeliibacillus, Enterobacter, Enterococcus and Sphingomonas. The relative abundances of Lactobacillus, Pediococcus, Weissella, Bacillus, Lactococcus, Pseudomonas, Acinetobacter and Rummeliibacillus were more than 1%. In fermented seafood, the average abundance of Lactobacillus was only 19%, and in fermented vegetables, the relative average content of Lactobacillus was 67% (Peng et al., 2018). The Lactobacillus, Pediococcus, Weissella, Caproiciproducens, Bacillus and Staphylococcus contents were high in fermented seafood, and Caproiciproducens was a unique and dominant genus of fermented seafood, whereas only Lactobacillus played an absolutely leading role in fermented vegetables.

We also compared the bacterial genera found among the FC, FF and FS. Caproiciproducens was the most abundant genus in FS (Table 2), and Lactobacillus was the most abundant genus in FC and FF. Lactobacillus and Pediococcus were dominant genera in FF and FC; interestingly, the presence of these two genera was lower in fermented shrimp, and Lentibacillus, Ochrobactrum, and Bacterium were more abundant in FC and FF.

After the different microbial genera had been identified, the metabolic pathways in the samples were predicted. We selected and compared the metabolic pathways that had significant differences (P < 0.05) among all samples (Figure 4). The abundance of Membrane Transport, Replication and Repair, Translation, and Nucleotide Metabolism were higher than other pathways. Membrane Transport plays an important role in the endocytic pathway and transport proteins, and the metabolic pathway of Membrane Transport in FS was the highest.

We performed a correlation analysis of climatic conditions, core bacterial genera, microbial metabolic pathways and pH value in fermented seafood (Figure 5A) and fermented vegetables (Figure 5B). The relative abundances of Pediococcus, Caproiciproducens and Weissella were highest in fermented seafood; however, in fermented vegetables, only Lactobacillus had a higher abundance than other microbial genera. Lactobacillus had an obvious strong negative correlation with pH value, as shown in Figure 5B, which meant the high abundance of Lactobacillus would reduce the pH value, and this finding also demonstrated that fermented vegetables had a low pH value. The relative content of the metabolic pathways of Amino Acid Metabolism was the highest in fermented seafood, and Membrane Transport and Replication and Repair were the metabolic pathways with the highest abundance in fermented vegetables.

Based on the 16S rDNA sequencing, 178 strains of bacteria were isolated and identified (Table 3). Although the microbial diversity between fermented vegetables and fermented seafood was different, Lactobacillus fermentum and Lactobacillus plantarum were still dominant. Lactobacillus fermentum had a higher abundance in FS, whereas Lactobacillus plantarum had a higher abundance in FF.

From our previous research, we found that Lactobacillus fermentum and Lactobacillus plantarum were the predominant Lactobacillus species in fermented vegetables. We wanted to determine the evolutionary distance of Lactobacillus fermentum and Lactobacillus plantarum in the different fermented foods. Thus, we compared SNPs of Lactobacillus plantarum and Lactobacillus fermentum. We found 107 SNPs from the 16S rDNA sequences of Lactobacillus plantarum and 134 SNPs of Lactobacillus fermentum in the samples (Figure 6, 7). The evolutionary tree of Lactobacillus plantarum and Lactobacillus fermentum is illustrated on the left of the figure. In Figure 6, the mutations are concentrated in the lower part, and in Figure 7, the mutations are concentrated in the middle part. Mutagenesis has been shown to occur in strains that are close to each other in evolution, and by further comparison, we found that most of the mutations occurred in fermented vegetables.