The bacterial strain used in this study was isolated from the gut of the migratory locust, which was reared as an animal model system at the Institute of Zoology, Chinese Academy of Sciences, Beijing, China. The whole guts of 10 locusts were carefully dissected with sterile devices, homogenized in 10-ml of sterile phosphate-buffered saline (PBS) solution (pH 7.4), and then diluted by sterile PBS solution (1:10,000). Subsequently, 50 ml of the diluted homogenized solution was spread over de Man Rogosa and Sharpe (MRS) medium (Man et al., 1960). A single colony was used to prepare the broth and establish the glycerol stocks. The single colony was also used for 16S rRNA gene amplification by using the universal primers 27F (5′-AGAGTTTGATCCTGGCTCAG-3′) and 1492R (5′-TACGGCTACCTTGTTACGACTT-3′). Amplified products were sequenced and assigned to taxa based on NCBI public databases by using the basic local alignment search tool. Isolated strains were maintained at −80°C as frozen stock cultures in MRS broth.

Weissella confusa LM1 was cultured in MRS broth at 37°C for 12 h. Genomic DNA was extracted by using Qiagen DNeasy Blood & Tissue Kit (Qiagen) according to the instructions of the manufacturer. Harvested DNA quality was checked by agarose gel electrophoresis (0.8%), and DNA quantity was determined by using NanoDrop 1000 (Thermo Fisher Scientific). The whole-genome sequence of W. confusa LM1 was obtained through single-molecule real-time (SMRT) sequencing by using PacBio Sequel platform and through next-generation sequencing (NGS) by using Illumina NovaSeq platform. Libraries for SMRT sequencing were constructed with an insert size of 10 kb by using the SMRTbell Template kit (Pacific Biosciences). Libraries for NGS were constructed with an insert size of 350 bp by using Ultra DNA Library Prep Kit (NEB).

Preliminary assembly sequences were obtained on the basis of the raw data from PacBio Sequel platform by using the assembly software PacBio SMRT Link V5.0.¹ Low-quality reads of less than 500 bp were filtered to obtain clean data and to ensure the accuracy of the subsequent analyses. The automatic error correction function of SMRT portal was performed. Long reads of over 6,000 bp were selected as the seed sequence, the other reads were aligned to the seed sequence by using Basic Local Alignment with Successive Refinement (BLASR) to improve the accuracy of the seed sequence (Chaisson and Tesler, 2012). After obtaining the initial assembly result, the arrow algorithm was used to correct the variant sites in the initial assembly result by using the variant Caller module of the SMRT Link. NGS data were used to further correct the preliminary assembly sequence by BWA (Li and Durbin, 2009). Cyclization was then confirmed according to overlap (at least 1.5 kb). The initiation site was corrected by BLAST with the DnaA genes and confirmed through GC-Skew.

The genome of W. confusa LM1 was automatically annotated using the GeneMarkS-2+ based on the NCBI Prokaryotic Genome Annotation Pipeline (PGAP) (Tatusova et al., 2016). Each gene was assigned to a category within the Clusters of Orthologous Groups (COG) database (Galperin et al., 2021). Circular layouts were generated using Circos V0.69² (Krzywinski et al., 2009). Ribosome RNA (rRNA) was identified by using RNAmmer V1.2 (Lagesen et al., 2007), and tRNA genes were predicted by tRNAscan-SE V2.0 (Lowe and Chan, 2016). Carbohydrate-Active enZYmes (CAZy) database (Lombard et al., 2014) was used for carbohydrate metabolism analysis by using DIAMOND with a cutoff e-value of 1e-5 and minimum identity of >40% (Buchfink et al., 2015). The online server of the Antibiotics and Secondary Metabolite Analysis Shell (antiSMASH V6.0³) was used for the annotation and analysis of secondary metabolite biosynthesis gene clusters with “relaxed” detection strictness (Blin et al., 2021).

The virulence factor database (VFDB) core dataset (setA) was used to curate information about the virulence factors of bacterial pathogens by using DIAMOND with a cutoff e-value of 1e-5 and 70% identity based on the protein data (Chen et al., 2012). Based on the genome data, PathogenFinder V1.1⁴ was used to predict the pathogenicity of the bacteria toward human hosts (Cosentino et al., 2013). Prophage regions of W. confusa strains were detected by using the PHASTER server (Arndt et al., 2016). Phylogenetic tree of intact phages was created by using VICTOR⁵ (Meier-Kolthoff and Goker, 2017). ResFinder database was used to identify the acquired antibiotic resistance genes based on genome data by using the ResFinder 4.1⁶ with thresholds of 90% identity and 60% gene coverage (Zankari et al., 2012). The Comprehensive Antibiotic Resistance Database (CARD) was also used to identify antibiotic resistance genes by using the Resistance Gene Identifier (RGI) software to predict resistome from protein data based on homology and SNP models with the selection criteria of perfect and strict hits only⁷ (Alcock et al., 2020).

For the comparative genome analysis, the genomic sequences of each Weissella species and strains were downloaded from NCBI. Average nucleotide identity (ANI) value was calculated by using JSpecies V1.2.1 with BLAST (Richter and Rossello-Mora, 2009) under the default parameters (Goris et al., 2007). The dropoff value for gapped alignment was set to 150, the e-value was set to 1e-15, the identity was set to 30%, and alignment length was set to 1,020 bp. ANI is the average value based on the comparisons of all orthologous protein-encoding genes of pairwise genomes and is a classic method for judging whether strains belong to the same species. Strains with ANI values of over 95% are generally considered to be of the same species (Yoon et al., 2017). The obtained ANI value matrix was clustered and visualized by using R package pheatmap V1.0.12. Multiple genomic sequence alignment visualization was displayed by using Mauve V20150226 (Darling et al., 2004), which is used for multiple genome alignments to identify genome-wide rearrangements and inversions. Minimum locally colinear block (LCB) weight was set as 15 and Muscle V3.6 was used for full alignment (Edgar, 2004).

The pan-genome of W. confusa was constructed by using Bacterial Pan Genome Analysis (BPGA) V1.3.0 with 50% sequence identity cu-off for clustering (Chaudhari et al., 2016). Phylogenetic tree was constructed on the basis of the core genes of different strains through the neighbor-joining method. The effects of isolation ecological niches on the genetic characteristics of W. confusa strains were analyzed via principal component analysis (PCA) according to the gene presence–absence matrix generated from BPGA.

The complete genomic sequences of W. confusa LM1 have been deposited in GenBank under the accession numbers CP080582.1 (chromosome) and CP080583.1 (plasmid). The BioProject accession number for this project was PRJNA752246 (http://www.ncbi.nlm.nih.gov/bioproject/752246).

We used the MRS medium to screen and isolate LAB in the gut of the migratory locust. Approximately 10⁷ colony-forming units (CFUs) were found in the whole gut of locusts. Eighty colonies were randomly selected to identify species composition by sequencing 16S rRNA gene sequence. We found that 74 (92.5%) of the colonies were W. confusa, which stably existed in the locust gut.

We isolated the strain and sequenced the whole genome of W. confusa LM1. The W. confusa LM1 genome consisted of a circular chromosome and a circular plasmid (Figures 1A,B). The size of its genome was 2.53 Mb; the sizes of the circular chromosome and circular plasmid were 2.50 Mb and 25.85 kb, respectively (Table 1). A total of 2,320 protein-coding sequences (CDSs) were predicted by GeneMarkS, and 1,543 genes were annotated in the COG database, which were primarily mapped to the metabolism category, particularly carbohydrate transport and metabolism (Figure 1C). This strain from locust had 114 RNA genes, which are as follows: 84 tRNAs, 10 5S rRNA, 9 16S rRNA, 9 23S rRNA, and 2 sRNA genes.

There were 28 complete genomes of Weissella genus in the NCBI public databases. Based on the complete genome sequences, we calculated the ANI value between different species and strains (Figure 2). The ANI value reflected the evolutionary relationship between two species, and 95% sequence identity was considered as the species level threshold.

The 29 complete genome assemblies of Weissella were clustered into five clusters, namely, A, B, C, D, and E (Figure 2). Cluster A had 2 species and 14 strains of W. cibaria and W. confusa, whose ANI values were about 78–80% compared with each other. However, the ANI value was less than 75% compared with that of other species. Cluster B had four species and four strains, namely, Weissella soli, Weissella cryptocerci, Weissella viridescens, and Weissella diestrammenae, and all of them had lower ANI values compared with other strains. In particular, the ANI value of W. diestrammenae was only about 70% compared with that of other species, indicating its species-specific status. Cluster C had three species and four strains, namely, Weissella jogaejeotgali, Weissella paramesenteroides, and W. hellenica, and they were found to have a close evolutionary relationship with an ANI value of above 78%. In particular, the ANI value between W. jogaejeotgali and W. paramesenteroides was 84%. Cluster D had two species and four strains, namely, Weissella koreensis and the unidentified species W. sp. HDW19. The species of Weissella sp. HDW19 had a high ANI value of 80% with W. koreensis. Cluster E had 1 species, W. ceti, and three strains. W. ceti had a low ANI value with all other species.

By 2021, three complete and 35 draft genomes of W. confusa strains were deposited into the NCBI public database, and 85% of them were sequenced in the past 3 years. According to the information on source and region, W. confusa could be isolated from different niches, such as food, plants, and animals, especially in vertebrate animals (Supplementary Table 1). LM1 was the first insect-derived W. confusa strain to be sequenced. It had the largest genome size (2.53 Mb), which was at least 6% more than those of the other strains (2.18–2.38 Mb). Additionally, LM1 had a lower GC content than the other W. confusa strains.

To analyze the evolutionary relationships among different W. confusa strains, we calculated the ANI value according to the genomic information of 39 strains (Figure 3A) and performed an evolutionary tree analysis based on the CDSs (Figure 3B). Although all strains were named W. confusa, according to ANI values, the ANI values fluctuated between 90 and 100%. According to their ANI values. The 39 strains were classified into four clusters, namely, A, B, C, and D, on the basis of their genome sequences (Figure 3A). Cluster A had 3 strains, which had a relatively low ANI value compared with the 36 other strains (91–92%). The W. confusa VTTE-062653 was remarkably different from all the other strains. However, the ANI value between W. confusa VTTE-90392 and W. confusa 718955 was 99%. Cluster B had 1 strain, W. confusa VTTE-153457, whose ANI value was approximately 95% compared with those of other strains. Cluster C had 29 strains, and they all had high ANI values (98–100%). Cluster D had six strains, whose ANI values were in the range of 96–98%. W. confusa LM1 was the most similar to W. confusa 1001271B_151109_G12, which was isolated from human feces (United States).

To analyze the genomic structures among different W. confusa strains, we compared all four complete genomes and subjected them to similarity analysis by using the Mauve software (Figure 4). With W. confusa LM1 as the reference, the four genomes exhibited 11 LCBs, and W. confusa LM1 showed more uncolored regions.

To explore the whole genetic characteristics of W. confusa, we performed pan-genome analysis and obtained the core, accessory, and unique genes of the 39 strains of W. confusa (Figure 5). The 768 genes shared by all W. confusa strains formed the core genome. The 2,246 genes present in more than one strain formed the accessory genome, and 1,193 genes specific to one single strain formed the unique genome. The percentage of unique genes was divergent in different W. confusa strains (Figure 5). Some strains, such as W. confusa FS54 (32%), W. confusa LM1(11%), and W. confusa VTTE-062653 (7%), contained a high percentage of unique genes. By contrast, five strains, including W. confusa MGYG-HGUT-00155, W. confusa AF03-13, W. confusa NBRC 106469, W. confusa VTTE-133279, and W. confusa VTTE-153460, did not have unique genes. These different W. confusa strains were widely distributed in food, animals, and plants. However, little was known about the relationship between their genomic characteristics and their ecological niches. Among the 39 W. confusa strains, 15 strains were isolated from food, 15 strains were isolated from animals, 6 strains were isolated from plants, and 3 strains were isolated from other ecological niches (Supplementary Table 1). On the basis of the gene presence–absence matrix of different W. confusa strains, no close relationship was found between their genetic characteristics and ecological niches (Figure 6). Moreover, we performed the fitting curve f(x) = a × x^b (a = 2,055, b = 0.18) based on the correlation of total gene number and genome number. The curve parameter indicated that the genome of W. confusa was still open (Figure 7). The open pan-genome implied a great potential for discovering novel genes with the increasing number of W. confusa strains sequenced.

To compare genome functional categories, we used functional assignments from the COG database. In the pan-genome of W. confusa, we found that most core, accessory, and unique genes fell in the metabolism category (Figure 8). Meanwhile, genes with predicted functions and unknown functions were also more abundant (Supplementary Figure 1). Compared with the core genome, the majority of genes in the no-core genome (accessory and unique genomes) were involved in carbohydrate transport and metabolism (G), amino acid transport and metabolism (E), cell wall/membrane/envelope biogenesis (M), defense mechanisms (V), and cell motility (N). By comparison, the vast majority of genes in the core genome were involved in housekeeping functions, such as nucleotide transport and metabolism (F), coenzyme transport and metabolism (H), and translation, ribosomal structure, and biogenesis (J).

To further understand the carbohydrate metabolism in W. confusa, we identified the CAZymes based on the pan-genome. Most genes fell into glycosyltransferases (GTs), glycoside hydrolases (GHs), carbohydrate esterases (CEs), carbohydrate-binding molecules (CBMs), whereas a few genes were annotated in polysaccharide lyases (PLs). All of the genes were not annotated in auxiliary activities (AAs). A few core genome genes were on CAZymes. Compared with the core genome, the accessory genome had more genes on GT, GH, CBM, and CE, the unique genome had more genes on GT, GH, and CE, but fewer genes on CBM (Figure 9A). In addition, different W. confusa strains exhibited various numbers of CAZyme genes (Supplementary Table 2). There were 110 CAZyme genes in W. confusa 718955, but only 85 CAZyme genes in W. confusa MBF8-1. This finding implied that the metabolic abilities of carbohydrates were specialized among W. confusa strains. To further compare the specificity of different strains, we analyzed the CAZyme genes of different strains based on unique genome (Figure 9B). Twenty W. confusa strains contained unique CAZyme genes, and GT (56%) and GH (31%) genes were the most common in the unique CAZyme genes, such as GT2 (28%), GT4 (17%), GH23 (7%), and GH70 (6%).

The genomic repertoire for the secondary metabolite synthesis of the 39 W. confusa strains was assessed by using antiSMASH. W. confusa strains were found to contain a small number of genes that encode key secondary metabolism biosynthesis enzymes. Only 42 gene clusters encoded secondary metabolites in all the 39 W. confusa strains (Supplementary Table 3). Moreover, 42 gene clusters that encoded secondary metabolites were classified into three types, namely, T3PKS (39), arylpolyene (2), and lassopeptide (1). Each strain harbored one T3PKS. The strain W. confusa FS54 and W. confusa LM1 harbored one arylpolyene. The strain W. confusa WcL17 harbored one lassopeptide.

Based on the VFDB core dataset (setA), we scanned the W. confusa genomes for virulence-related factors. In total, we identified 107 putative orthologs involved in the production of virulence factors in 39 W. confusa strains (Supplementary Table 4). These 107 putative orthologs were classified into three virulence-related factors, namely, hyaluronic acid capsule, capsule, and EF-Tu. All W. confusa strains had three virulence-related factors except W. confusa FS54, which only had one capsule. There was no strain-specific virulence factor. According to the function of the three virulence-related factors (Supplementary Table 4), we found that none of the virulence-related factors was associated with toxin production and secretion system, thereby indicating that W. confusa strains were not sufficient to perform virulence. Moreover, we predicted the pathogenicity of W. confusa strains toward human hosts by using PathogenFinder. W. confusa strains were predicted to be non-human pathogens. We estimated the antibiotic resistance of W. confusa strains based on the antibiotic resistance databases, ResFinder and CARD, and no antibiotic resistance genes were recognized in W. confusa strains. We detected the phages in 39 genomes of W. confusa strains. There were 112 phages, including 40 intact phages, 55 incomplete phages, and 17 questionable phages (Supplementary Table 5). There were 30 W. confusa strains with intact phages, which could be divided into two major clusters (Supplementary Figure 2), and we did not find any virulence factors and antibiotic resistance genes in all phages.