Effects of the Probiotic, Lactobacillus delbrueckii subsp. bulgaricus, as a Substitute for Antibiotics on the Gastrointestinal Tract Microbiota and Metabolomics Profile of Female Growing-Finishing Pigs

Simple Summary Lactobacillus delbrueckii subsp. bulgaricus (LDB) is an important candidate for antibiotic replacement in pig production. In this study, LDB and antibiotic diets were fed to the LDB and antibiotic groups of female growing-finishing pigs, respectively. 16S rRNA sequencing was used to identify different microbiota. Liquid chromatography-mass spectrometry-based non-targeted metabolomics approaches were used to identify different metabolites. The co-occurrence network of the fecal microbiota and metabolite was analyzed. The results contain information on pig growth performance, microbiota data, metabolite data and co-occurrence networks, supporting the possibility of LDB as an antibiotics replacement in pig production. Abstract Lactobacillus delbrueckii subsp. bulgaricus (LDB) is an approved feed additive on the Chinese ‘Approved Feed Additives’ list. However, the possibility of LDB as an antibiotic replacement remains unclear. Particularly, the effect of LDB on microbiota and metabolites in the gastrointestinal tract (GIT) requires further explanation. This study aimed to identify the microbiota and metabolites present in fecal samples and investigate the relationship between the microbiota and metabolites to evaluate the potential of LDB as an antibiotic replacement in pig production. A total of 42 female growing-finishing pigs were randomly allocated into the antibiotic group (basal diet + 75 mg/kg aureomycin) and LDB (basal diet + 3.0 × 109 cfu/kg LDB) groups. Fecal samples were collected on days 0 and 30. Growth performance was recorded and assessed. 16S rRNA sequencing and liquid chromatography-mass spectrometry-based non-targeted metabolomics approaches were used to analyze the differences in microbiota and metabolites. Associations between the differences were calculated using Spearman correlations with the Benjamini–Hochberg adjustment. The LDB diet had no adverse effect on feed efficiency but slightly enhanced the average daily weight gain and average daily feed intake (p > 0.05). The diet supplemented with LDB increased Lactobacillus abundance and decreased that of Prevotellaceae_NK3B31_group spp. Dietary-supplemented LDB enhanced the concentrations of pyridoxine, tyramine, D-(+)-pyroglutamic acid, hypoxanthine, putrescine and 5-hydroxyindole-3-acetic acid and decreased the lithocholic acid concentration. The Lactobacillus networks (Lactobacillus, Peptococcus, Ruminococcaceae_UCG-004, Escherichia-Shigella, acetophenone, tyramine, putrescine, N-methylisopelletierine, N1-acetylspermine) and Prevotellaceae_NK3B31_group networks (Prevotellaceae_NK3B31_group, Treponema_2, monolaurin, penciclovir, N-(5-acetamidopentyl)acetamide, glycerol 3-phosphate) were the most important in the LDB effect on pig GIT health in our study. These findings indicate that LDB may regulate GIT function through the Lactobacillus and Prevotellaceae_NK3B31_group networks. However, our results were restrained to fecal samples of female growing-finishing pigs; gender, growth stages, breeds and other factors should be considered to comprehensively assess LDB as an antibiotic replacement in pig production.


Introduction
Probiotics are defined as "live strains of strictly selected microorganisms that, when administered in adequate amounts, confer a health benefit to the host" [1]. Probiotics have been widely researched in humans [2], rats [3], chickens [4], cattle [5] and pigs [6]. The most used microorganisms belong to the genera Lactobacillus [7], Bifidobacterium [8] and Saccharomyces [9]. Probiotics in pigs play important roles, such as defending against viral infection [10], enhancing meat quality [11], improving immune function [12] and increasing growth performance [13]. Most importantly, probiotics (especially Lactobacillus spp.) [14,15] are used as substitutes for antibiotics in pig production [16]. A previous study showed that Lactobacillus enhanced the reproductive performance of sows and the growth performance of weaned piglets [17]. According to a study by Chen et al., Lactobacillus can provide some levels of protective effect against porcine epidemic diarrhea virus infections [18]. Tian et al. report that Lactobacillus can enhance meat quality by increasing the concentration of inosinic and glutamic acid concentrations, decreasing drip loss and shear force [19]. Moreover, Geng et al. proposed that Lactobacillus may promote the immunity of weaned piglets by regulating cytokine levels [20]. According to the feed additives list in China, eight Lactobacillus spp. can be used as feed additives in pig production, including L. acidophilus, L. casei, L. delbrueckii subsp. Lactis, L. plantarum, L. reuteri, L. cellobiose, L. fermentans and LDB (http://www.moa.gov.cn/nybgb/2014/dyq/201712/t20171219_6104350.htm, accessed on 10 May 2022). Most of them are used as an alternative to antibiotics, including L. reuteri, L. fermentums, L. acidophilus, and L. salivarius [21], L. casei [22] and L. plantarum [23]. According to the Chinese Center of Industrial Culture Collection, the biological hazard of LDB is level four, which means low risk, low pathogenicity, less chance of laboratory infection and no cause of human or animal disease (http://www.china-cicc.org/, accessed on 10 May 2022). LDB is usually used to produce probiotic health food [24]. LDB can inhibit Escherichia coli [25] and Helicobacter pylori infections [26]. LDB can eliminate Clostridium difficile-mediated cytotoxicity and reduce C. difficile colonization in colorectal cells [27]. However, few studies have addressed LDB as an antibiotic replacement in pigs.
It is known that bacteria are usually located in the digestive tract, especially in the colon, rectum and cecum. The rectum microbiota forms an extraordinarily complex system that plays a key role in animal physiology and health, including host nutrient metabolism and regulation of carbohydrate metabolism [28]. Microbiota comprises diverse bacteria and other microorganisms, whose abundance is influenced by the host's genetics [29], age [30], disease status [31] and environmental factors. Previous studies have shown that 16S rRNA technology is suitable for exploring the rectum microbiota [32,33]. Using 16S rRNA technology, Wang et al. reported that the L. reuteri effectively reduced E. coli in pigs [34]. In addition, Xu et al. showed that Saccharomyces cerevisiae regulated the abundance of Enterococcus, Succinivibrio and Ruminococcus, among others [35]. The GIT is where major nutrient metabolism and absorption occurs, and since bacteria metabolize the nutrients, the fecal metabolites become increasingly complex as the diet changes.
Metabolomics is an emerging omics technology that explains differences at the metabolic level and is suitable for identifying fecal biomarkers [36]. Mao et al. illustrated (using metabolomics technology) that L. rhamnosus GG substantially increased the concentrations of caprylic acid, 1-mono-olein, erythritol and ethanolamine [37]. Of note, there is a strong association between GIT microbiota and metabolites and 16S rRNA technology, coupled with metabolomics, is able to explain the link between gut microbiota and metabolites. Using 16S rRNA and metabolomics technology, Liang et al. revealed that a diet supplemented with Clostridium. butyricum changed 22 metabolites and specific microbiota (such as Oscil-lospira, Ruminococcaceae_NK4A214_group and Megasphaera) in pigs [38]. However, 16S rRNA technology combined with metabolomics has not yet been applied to LDB in pigs.
The microbiota changes with age in pigs, especially in piglets [39]. According to Wang et al., the microbiota in the growing-finishing stage is relatively stable and sex does not significantly affect swine GIT microbiota [40]. Han et al. reported that sows in the growing-finishing stage (93 d and 147 d) had a stable intestinal environment [41]. These studies suggest that growing-finishing pigs are suitable for analyzing the possibility of replacing antibiotics with probiotics.
In the current study, female growing-finishing pigs were used. 16S rRNA technology was implemented to determine bacteria abundance in the microbiota, while metabolomics technology was used to examine metabolite contents in fecal samples of pigs using a liquid chromatography-mass spectrometry-based (LC-MS), non-targeted metabolomics approach. The relationship between the microbiota and metabolites was explored. 16S rRNA technology and metabolomics technology were used to further explain the possibility of using LDB as an antibiotic replacement in pigs.

Animals and Sample Collection
A total of 42 female growing-finishing pigs (Duroc×Landrace×Yorkshire, 59.39 ± 2.29 kg) in the growing-finishing stage were provided by Nanning Xingda Pig Farm (Nanning, China). All pigs were randomly divided into control (G0) and experimental (G1) groups and were raised under similar feeding management regimes. Each group comprised 21 pigs and three replicates, with seven pigs in each replicate. In this study, G0 animals were fed the basal formula diet with 75 mg/kg aureomycin, while G1 animals were fed the basal formula diet supplemented with 3.0 × 10 9 colony forming units (CFU)/kg LDB. The LDB was provided by the Chinese Center of Industrial Culture Collection (CICC6098). The basal diet was prepared according to the nutritional needs specified by the NRC (1998). The basal dietary formulation and nutrient contents are shown in Tables 1 and 2, respectively. The experiment lasted 30 days and none of the pigs in G1 received antibiotic treatment during the study period. Individual fecal samples were collected following rectal stimulation on days 0 (D0) and 30 (D30). All fecal samples were immediately snap-frozen in liquid nitrogen and stored at −80 • C in the laboratory. The body weights of pigs were recorded on D0 and D30, and the average daily gain was calculated. The feed intake in all replicates was recorded and the feed efficiency was calculated. All data were statistically analyzed using t-tests in SPSS 19.0 software. Fecal samples of nine pigs from G0 on day 0 (G0D0), G1 on day 0 (G1T0), G0 on day 30 (G0D30) and G1 on day 30 (G1D30) were randomly selected for 16S rRNA amplicon sequencing and untargeted metabolomic analysis. Differences between G0D0 and G1D0 were used to explore the effects on GIT microbes and metabolites. The difference between G0D30 and G1D30 helped examine the impact of LDB on GIT microbes and metabolites.

16S rRNA Amplicon Sequencing and Analysis
Microbial genomic DNA from pig fecal samples was extracted using the MagPure Stool DNA LQ Kit (Magen, D6358-03, Guangzhou, China) according to the manufacturer's instructions. The 16S rRNA gene V3-V4 (341F-805R) region was amplified via PCR using 5 -CCTACGGGNGGCWGCAG-3 and 5 -GACTACHVGGGTATCTAATCC-3 as the forward and reverse primers, respectively. The Illumina MiSeq platform was used for sequencing at Benagen (Wuhan, China), and 16S rRNA gene sequence exploration was executed using QIIME2 [42]. Amplicon bioinformatic analysis was performed using EasyAmplicon v 1.0 [9]. The VSEARCH parameters used were min_unique_size 35 in dereplication and sintax_cutoff 0.1 in the removal of plastids and non-bacteria [43]. The samples were rarefied to the lowest sequencing depth of 94,170. The principal coordinate analysis (PCoA) was based on the UniFrac binary distance. Using STAMP software, a p-value < 0.05 was defined for significantly different microbes. Metagenomic predictions were completed using Phylogenetic Investigation of Communities by Reconstruction of Unobserved States 2 (PICRUSt 2) and summarized by the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways [44]. The differences in the KEGG pathways were identified by STAMP software using Storey's FDR multiple test corrections. The organism-level microbiota phenotypes were predicted using BugBase software [45].

Co-Occurrence Network of GIT Microbiota and Metabolites
The relationship between the relative abundance of significantly different microbes and the relative concentrations of differential metabolites in each sample was calculated using the Spearman correlation function implemented in R. The false discovery rate was applied to the p-values obtained from Spearman correlations. An adjusted p-value < 0.05 and absolute value > 0.6 were defined as a significant relationship pair. Cytoscape_v3.8.2 was used to construct the co-occurrence network between the GIT microbiota and metabolite significance relationship pair.

Growth Performance Analysis
The growth indices, including the average daily weight gain, average daily feed intake and feed efficiency, were similar among groups (p > 0.05) ( Table 5).

Fecal Metabolic Signatures
We explored major differences among metabolites in G0D0 vs. G1D0 and G0D30 vs. G1D30 using untargeted metabolomics in the same samples. A total of 17,941 peaks in the positive and negative ion modes were identified. After filtering, 3488 and 1908 metabolites were matched in the positive and negative ion modes, respectively. The model interpretation rates of X (R2X) were > 0.68 in both the positive and negative ion PCA analyses. Fourteen quality control (QC) samples were obtained under both ion mode conditions ( Figure 5A,B). In G0D0 vs. G1D0, the model interpretation rates of Y (R2Y) in the two ion modes were 0.58 and 0.63 ( Figure 6A,B), respectively, and the prediction ability (Q2) in the two ion modes was −0.46 and −0.91 ( Figure 6C,D), respectively. In G0D30 vs. G1D30, the R2Y and Q2 in the two ion modes were over 0.76 and 0.32 ( Figure 7A,B), respectively, and the intercept of the permutation test in the two ion modes was less than −0.53 ( Figure 7C,D). Similar metabolites were obtained in the positive and negative ion modes in G0D0 vs. G1D0, respectively (p < 0.05). A total of 38 ( Table 6) and 18 (Table 7) different metabolites were obtained in the positive and negative ion modes in G0D30 vs. G1D30, respectively (p < 0.05) (Figure 8). The 56 differential metabolites in G0D30 vs. G1D30 were enriched in 15 metabolic pathways, including de novo triacylglycerol biosynthesis, glycerol-phosphate shuttle, cardiolipin biosynthesis, purine metabolism, spermidine and spermine biosynthesis, mitochondrial electron transport chain, vitamin B6 metabolism, glutathione metabolism, glycerolipid metabolism, phospholipid biosynthesis, methionine metabolism, steroid biosynthesis, tryptophan metabolism, bile acid biosynthesis and tyrosine metabolic pathway (Figure 9).

Co-Occurrence Network of the Fecal Microbiota and Metabolite Signatures
The 17 differential microbiota and 56 differential metabolites from G0D30 vs. G1D30 in 36 samples were used to calculate the Spearman correlation coefficients. A total of 437 significant relationship pairs were obtained, including nine microbiota-microbiota, 23 microbiotametabolite, and 405 metabolite-metabolite relationships (p < 0.05). The Spearman correlation coefficients in microbiota-metabolite significant relationship pairs ranged from −0.81 to −0.60 and 0.60 to 0.99. Three major microbiota-metabolite clusters were found ( Figure 10). Cluster one in G0D30 vs. G1D30 was composed of Lactobacillus, Peptococcus, Ruminococcaceae_UCG-004, Escherichia-Shigella, acetophenone, tyramine, putrescine, N-methylisopelletierine and N1acetylspermine. Cluster two in G0D30 vs. G1D30 was composed of norank_f_Porphyromonadaceae, 2-monolinolenin, capsi-amide, stearoyl ethanolamide, stearamide and etretinate. Cluster three in G0D30 vs. G1D30 comprised Prevotellaceae_NK3B31_group, Treponema_2, monolaurin, penciclovir, N-(5-acetamidopentyl)acetamide and glycerol 3-phosphate. Figure 10. The co-occurrence network between differential microbiota and differential metabolites. The circle shapes were the differential microbiota; the diamond shapes were the differential metabolites. The red lines mean the significant positive correlation; the blue lines mean the significant negative correlation. The size of lines means the correlation size; the size of shapes means the number of networks; the p was the positive ion model, and the N was the negative ion model.

Discussion
Since 2020, the addition of growth-promoting antibiotics in pig diets has been banned throughout China, and microbial feed additives are being considered as an antibiotic replacement. Similar to previous studies [47,48], LDB yielded no adverse effect on feed efficiency. The average daily gain and feed intake in pigs between G0 and G1 did not significantly differ, which may have been related to the short experimental period (30 d) in our study. Nevertheless, these results suggest LDB is a candidate antibiotic replacement in pigs because of the lack of negative effect on growth performance. However, a longer experimental cycle and pigs of different ages are needed to comprehensively elucidate the function of LDB. To determine the possibility of LDB as an antibiotic replacement, we conducted 16S rRNA sequencing and metabolomics. Expectedly, the GIT microbiota and the metabolites were strongly correlated [49][50][51]. The current study revealed major differences between G0D30 and G1D30 in the microbiota and metabolites of the fecal samples from LDB-fed pigs.
The GIT microbiota was dominated by the phyla Firmicutes and Bacteroidetes, which is consistent with the results of prior research [52]. However, in the current study, Firmicutes abundance was increased, and that of Bacteroidetes was decreased in G1D30 compared to G0D30. Interestingly, the contribution of potentially pathogenic forms of Firmicutes, Bacteroidetes and Spirochaetes in G1D30 decreased. The abundance of potentially pathogenic bacteria in G1D30 was substantially reduced, indicating that an LDB-supplemented diet may inhibit the growth of potentially pathogenic bacteria by regulating the GIT function [53]. In G1D30, the abundance of Streptococcus, a pathogenic bacterium, was significantly enhanced. Streptococcus_gallolyticus_subsp._pasteurianus comprised 83.97% of the Streptococcus spp. and was substantially enhanced in G1D30. The high abundance of Streptococcus_gallolyticus_subsp._pasteurianus poses a health risk to pigs, especially piglets, because it can cause severe neonatal sepsis and meningitis [54]. The abundance of Strep-tococcus_gallolyticus_subsp._pasteurianus can therefore be controlled should LDB replace antibiotics. Similar to the observations of Bergamaschi et al., the predominant bacterial genus was Lactobacillus rather than Prevotella [55]. The high abundance of Lactobacillus in G1D30 indicates that its presence in the diet is conducive to its abundance in feces [56]. Conversely, the abundance of Limosilactobacillus in G1D30 was not increased in the current study, which may be related to the LDB content used in our study. Furthermore, LDB induced a significant increase in Lactobacillus abundance and a significant decrease in the abundance of Treponema_2 and Prevotellaceae_NK3B31_group in the top ten most abundant genera. Similar to the results obtained by Sampath et al. [57] and Pupa et al. [14], Lactobacillus abundance was increased in pigs fed an LDB-supplement diet in our study. Probiotic supplementation inhibits Treponema_2 in pig caecal digesta [12]. L. reuteri substantially reduced the abundance of Treponema sp. in the human mouth [58]. Similar to Xu et al., the abundance of Prevotellaceae_NK3B31_group was substantially decreased in pigs administered compound probiotic diets [59]. Treponema_2 and Prevotellaceae_NK3B31_group are Gram-negative bacteria [60] and can produce lipopolysaccharides [61]. Lipopolysaccharides can trigger acute inflammatory responses and the release of inflammatory cytokines and chemokines [62]. Furthermore, as a product of Lactobacillus spp., lactic acid plays a key role in antimicrobial, antiviral and immune regulation [63]. The high levels of Lactobacillus spp. in the animal GIT can inhibit the abundance of pathogenic bacteria and decrease lipopolysaccharide-induced inflammation [64]. Thus, the LDB diet affected the GIT function by improving the abundance of Lactobacillus and Streptococcus spp. and decreasing the abundance of Treponema_2 and Prevotellaceae_NK3B31_group spp.
The amino acid metabolic pathways for arginine, proline, beta-alanine, glycine, serine and threonine lysine, and phenylalanine were identified in the microbiota. Methionine, tryptophan and tyrosine metabolism pathways were enriched in metabolites. Previous studies have reported that dietary supplementation with Lactobacillus spp. influences amino acid metabolism [67][68][69]72]. The energy metabolism was substantially enhanced, and glycolysis/gluconeogenesis pathways were inhibited in G0D30 compared to those in G1D30 in the microbiota. De novo triacylglycerol biosynthesis, the glycerol-phosphate shuttle, mitochondrial electron transport chain, and glycerolipid metabolism pathways were enriched in metabolites. Wang et al. showed that L. frumenti promotes porcine energy production [73]. Tang et al. reported that L. acidophilus NX2-6 enhanced glycolysis and intestinal gluconeogenesis [74]. The LDB diet enhanced the concentration of pyridoxine, tyramine, D-(+)-pyroglutamic acid, hypoxanthine, putrescine and 5-hydroxyindole-3-acetic acid and decreased the concentration of lithocholic acid and regulated the amino acid metabolism and energy production in the pig GIT in a previous study.
Lactobacillus, norank_f_Porphyromonadaceae and Prevotellaceae_NK3B31_group spp. were the core microbiota, and N1-acetylspermine was the core metabolite in the co-occurrence network. Clusters one to three in G0D30 vs. G1D30 contained eight, five and five microbiota-metabolite pairs, respectively. The abundance of norank_f_Porphyromonadaceae and Prevotellaceae_NK3B31_group in G0D30 was substantially higher than in G1D30. Importantly, Lactobacillus and Prevotellaceae_NK3B31_group were the most abundant microbiota in our study, and the average abundance of norank_f_Porphyromonadaceae in G0D30 and G1D30 was 0.00089 and 0, respectively. Clusters one and three played a more important role in mediating the effects of the LDB on porcine GIT function.
Nonetheless, our study had some limitations. For instance, only female pigs were used, excluding the post-weaning period analysis, and non-inclusion of ileal samples and immunological parameters. However, the results could be useful for the swine industry and public health because of the possible effect of reducing antibiotic use in animal production. Thus, we shall consider genders, growth stages, immunological parameters and ileal samples to comprehensively explain the possibility of using LDB as an antibiotic replacement in our further studies.

Conclusions
In summary, this study analyzed the different microbiota, metabolite contents and the link between the GIT microbiota and metabolites to explore the possibility of LDB as an antibiotic replacement. Our results revealed that LDB did not adversely affect growth performance. The LDB-supplemented diet increased Lactobacillus spp. abundance and the concentration of pyridoxine, tyramine, D-(+)-pyroglutamic acid, hypoxanthine, putrescine and 5-hydroxy indole-3-acetic acid. The LDB diet decreased the abundance of Prevotellaceae_NK3B31_group and the concentration of lithocholic acid. The Lactobacillus and Prevotellaceae_NK3B31_group networks greatly impacted how LDB regulates GIT function in our study. Nonetheless, additional factors such as genders, growth stages and breeds should be considered in our further study to comprehensively explain the mechanism of the LDB-mediated effect on the GIT function of pigs.