Fecal bacteria and metabolite responses to dietary lysozyme in a sow model from late gestation until lactation

Lysozyme (LZM) is a natural anti-bacterial protein that is found in the saliva, tears and milk of all mammals including humans. Its anti-bacterial properties result from the ability to cleave bacterial cell walls, causing bacterial death. The current study was conducted to investigate the effects of dietary LZM on fecal microbial composition and variation in metabolites in sow. The addition of LZM decreased the fecal short-chain fatty acids (SCFAs). Zonulin and endotoxin in the serum, and feces, were decreased with lysozyme supplementation. Furthermore, fecal concentrations of lipocalin-2 and the pro-inflammatory cytokine TNF-α were also decreased while the anti-inflammatory cytokine IL-10 was increased by lysozyme supplementation. 16S rRNA gene sequencing of the V3-V4 region suggested that fecal microbial levels changed at different taxonomic levels with the addition of LZM. Representative changes included the reduction of diversity between sows, decreased Bacteroidetes, Actinobacteria, Tenericutes and Spirochaetes during lactation as well as an increase in Lactobacillus. These findings suggest that dietary lysozyme supplementation from late gestation to lactation promote microbial changes, which would potentially be the mechanisms by which maternal metabolites and inflammatory status was altered after LZM supplementation.

weaning-to-estrus interval (WEI), and improved sow and offspring health as indicated by altered inflammatory cytokines and a reduction in the piglets' diarrhea rate 22 . However, there remains little knowledge of the maternal diversity of gut microbiota following lysozyme supplementation during the perinatal period. As a 1, 4-β-N-acetylmuramidase, lysozyme cleaves the β-1,4-glycosidic bond between the N-acetylmuramic acid and N-acetylglucosamine residues of the bacterial peptidoglycan, resulting in an incomplete cell membrane and leading to cell death 23 . This suggests that dietary lysozyme would lead to variation in the microbiota in the gut, which has been confirmed in piglets and infants 13,19,24 . Therefore, we investigated the effect on the maternal microbiota following lysozyme supplementation.
Thus, in this study we investigated the effects of adding lysozyme to maternal diets on gut permeability, gut inflammation, gut microbial and metabolite composition from the period of late gestation to lactation in a sow model. Table 1, LZM 150 or LZM 300 had no effect on the pH and concentration of SCFAs, in sow fecal material, on day 1 of lactation. In contrast, both LZM diets increased pH values on days 7 and 21 of lactation (P < 0.05) when compared with the control diet. However, no difference was found between the two LZM treatments. Sows fed with the LZM 150 and LZM 300 diets showed decreased (P < 0.05) fecal levels of acetate, butyrate and total SCFAs on day 7 of lactation and decreased fecal levels of butyrate on day 21 of lactation when compared to sows from the control diet group. Interestingly, concentrations of total and individual SCFA increased linearly from day 1 to day 21 of lactation ( Fig. S1).

Fecal changes in pH and concentration of short-chain fatty acids (SCFAs). As shown in
Difference in gut permeability and endotoxin concentrations. As a biomarker for gut permeability serum zonulin was measured. As can be seen in Fig. 1A1, no differences were found between the treatments in terms of the serum levels of zonulin on day 1 of lactation. Sows which were fed the LZM 150 and LZM 300 diets both showed decreased zonulin in their serum on days 7 and 21 of lactation (Fig. 1A2,A3, P < 0.05). This indicates that sows' gut permeability was decreased by the LZM treatment. In addition, when compared with the control groups, both LZM 150 and LZM 300 treatments decreased (Fig. 1B2,B3,C2,C3, P < 0.05) the serum and fecal concentrations of endotoxin on days 7 and 21 of lactation, although no difference was found on day 1 of lactation ( Fig. 1B1,C1).

Inflammation state of the sows' intestinal mucosal epithelium. Four biomarkers of inflammation
were measured to reflect effects of LZM on the inflammatory response in sows' intestines. As can be seen in Fig. 2, concentrations of fecal lipocalin-2 and TNF-α (pro-inflammatory cytokines) were decreased on days 7 and 21 of lactation in sows which had been fed the LZM 150 and LZM 300 diets ( Fig. 2A,B). However, no significant differences were found in fecal IL-6 between the treatments (Fig. 2C1-3). In contrast, concentrations of IL-10 (anti-inflammatory cytokine) increased on days 7 and 21 of lactation for the sows on the LZM 150 diets (Fig. 2D1-3). Differences in the fecal microbiota among the three treatments at the phylum and genus levels were identified on day 1, 7 and 21 of lactation. Sows which had been fed the LZM 300 diets showed decreased relative abundance of Bacteroidetes and Actinobacteria in the gut at the phylum level on days 7 and 21 of lactation (Fig. 4A,B). In addition, sows fed LZM 150 diets showed decreased relative levels of Tenericutes and Spirochaetes in the gut on days 1 and 7 of lactation (Fig. 4C,D). However, when compared with the control group sows which had been fed the LZM 300 diets had increased relative levels of Proteobacteria on day 1 of lactation (Fig. 4E). The relative abundance of Euryarchaeota was decreased in the LZM 300 treatment group compared with the LZM 150 treatment group on day 21 of lactation ( Fig. 4F). At the genus level, 23 genera relative abundances changed across the different treatments on day 1, 7 and 21 of lactation (Table 3). Sows which had been fed the LZM 150 and LZM 300 diets www.nature.com/scientificreports www.nature.com/scientificreports/ had decreased relative levels of Ruminococcaceae_UCG-005 on day 1 of lactation, decreased relative abundances of Parabacteroides, Prevotella_1, Cellulosilyticum and Succinivibrio on day 7 of lactation, and increased relative abundances of Lactobacillus and Ruminococcaceae_UCG-013 on day 21 of lactation. However, sows fed the LZM 150 diets had increased relative abundance of Methanobrevibacter and Ruminococcaceae_UCG-009 on day 21 of lactation. Additionally, the LZM 300 treatment increased the relative abundances of Romboutsia and Acinetobacter on day 1 of lactation, Ruminococcaceae_UCG-009 on day 7 of lactation, and Enterococcus, Staphylococcus on day 21 compared with the control group. Sows which had been fed LZM 300 diets had decreased relative abundances   Based on multiple-taxonomic LEfSe analysis of day 21 of lactation, at the phylum level Euryarchaeota was enriched in sows from the LZM 150 group, and Proteobacteria was enriched in sows from the LZM 300 group (Fig. 5A,B). At the family level, Prevotellaceae and Moraxellaceae were enriched in sows from the control group, while Lachnospiraceae, Lactobacillaceae and Methanobacteriaceae were enriched in sows from the LZM 150 group. Enterobacteriaceae, Streptococcaceae, Enterococcaceae, Bacillaceae and Staphylococcaceae were enriched in sows from the LZM 300 group. At the genus level, Acinetobacter was enriched in sows from the control group, however Lactobacillus, Lachnospiraceae_XPB1014_group and Methanobacteriales were enriched in sows from the LZM 150 group, and Streptococcus, Enterococcus and bacillus were enriched in sows from the LZM 300 group. In addition, LEfSe analysis at days 1 and 7 of lactation demonstrated a difference of enriched bacteria (Fig. S4). Therefore, these results demonstrate that sows' gut microbiota composition is profoundly changed by the supplementation with LZM during lactation.

Figure 3.
Comparison of the gut microbiota composition among treatments. Principal coordinate analysis to visualize the weighted UniFrac distances of fecal samples from individual sow. Con.d1 = control at day 1 of lactation, Con.d7 = control at day 7 of lactation, Con.d7 = control at day 21 of lactation, LZM 150 d1 = control diet + lysozyme 150 mg/kg at day 1 of lactation, LZM 150 d7 = control diet + lysozyme 150 mg/kg at day 7 of lactation, LZM 150 d21 = control diet + lysozyme 150 mg/kg at day 21 of lactation, LZM 300 d1 = control diet + lysozyme 300 mg/kg at day 1 of lactation, LZM 300 d7 = control diet + lysozyme 300 mg/kg at day 7 of lactation, LZM 300 d21 = control diet + lysozyme 300 mg/kg at day 21 of lactation.   Table 3. The relative abundances at genus level (%, >0.1% in at least one sample) of feeding sows diets supplemented with lysozyme. Data are expressed as mean ± SD. Sows were regarded as the experimental units, n = 6 for each treatment. LZM 150 = control diet + lysozyme 150 mg/kg, LZM 300 = control diet + lysozyme 300 mg/kg. a,b Within a row, means with different superscripts are different (P < 0.05).

Discussion
Lysozyme lyses a specific link in the peptidoglycan layer of bacterial cell walls resulting in cell decomposition 23 . It has been reported that piglets which consume LZM that had been highly expressed in transgenic goats' milk showed decreased Firmicutes and increased Bacteroidetes in their feces 25 . The abundant phyla Firmicutes and Bacteroidetes accounted for 85% of microbiota in the sow feces in this study, which is in good agreement with previous studies on pigs 26,27 . It was suggested by one study that increased Firmicutes means the body has a higher capacity for energy acquisition from their diets 28 . However, no difference was found between the Firmicutes and the ratio of Firmicutes/Bacteroidetes in feces depending on the lysozyme treatment. The difference between these studies may be due to differing lysozyme types, dosages, and physiological stages of pigs along with differences in the duration of feeding. Zou et al. 24 found that adding 100 mg/kg lysozyme (L100) to the diets of growing pig decreased cecal microflora evenness compared with control or L50 groups through 16S rRNA sequencing. In this study, we found microbial diversity decreased with lysozyme supplementation only in the LZM 300 treatment on day 21 of lactation. In addition, no difference was found in the richness of microbiota community between treatments. Interestingly, we also found that the microbiota community richness was increased with the lactation progress. This is in agreement with Cheng et al. 8 , who found the lowest gut microbial richness on day 3 of lactation. This may help explain how the physiological process of farrowing significantly changes the abundance of maternal gut microbiota, as well as how microflora continue to remodel and stabilize over the time. Fermentation of dietary fiber by microbiota in the hindgut produces many SCFAs, including acetic acid, propionic acid and butyric acid 29 . With intestinal absorption, acetate is most frequently used for lipid biosynthesis 30 . Propionate tends to be a precursor of gluconeogenesis 31 . Butyrate, mainly used as an energy source, is metabolized by colonocytes and regulates the growth and death of both epithelial and immune cells 32,33 . In addition, SCFAs are signal molecules for regulating lipid and glucose metabolism by combining with free fatty acid receptors 34,35 . These contribute to the key role of SCFA in the diet-gut microbiome-host metabolism axis. Acetate, a carbohydrate fermentation product, is produced by the majority of enteric bacteria. It was more than 50% of the total SCFA in feces in this study, which is in agreement with the findings of both Zhou et al. 2 and Cheng et al. 8 .
Additionally, we found that concentrations of total and individual SCFA increased from day 1 to day 21 of lactation, which aligns with the work of Cheng et al. 8 , who found an increase in butyrate from day 3 to day 14 in lactating sows. Diets supplemented with lysozyme decreased sows' acetate, butyrate and SCFAs on day 7 of lactation, while they decreased the butyrate on day 21 of lactation compared with the control treatment. Based on our previous data, sows on lysozyme diets didn't have any difference in back fat, weight increase during the gestation period or weight loss during the lactation period when compared with sows in the control treatment 22 . It seems that the decrease of the SCFAs didn't affect sows' lipid and glucose metabolism. Higher SCFAs in stool may indicate a greater amount of energy loss in feces, as suggested by Koren et al. 36 . The SCFAs Histogram of the LDA scores reveals the most differentially abundant taxa among different treatments. Con. = control, LZM 150 = control diet + lysozyme 150 mg/kg, LZM 300 = control diet + lysozyme 300 mg/kg. had higher colonic absorption, with less than 5% being excreted in feces in humans 37 . When compared to production or absorption of SCFAs per se, fecal SCFA concentrations are more suitable for reflecting the overall net production and absorption of SCFAs in gastrointestinal tracts 38,39 . In order to assess SCFAs potential production, SCFA producing-related phylum and genera were analyzed. At the phylum level, Bacteroides, which mainly produces propionate, was decreased by LZM supplementation at 300 mg/kg on day 7 of lactation. The concentration of propionate had a tendency (P = 0.06) to decrease with LZM supplementation on that day. Despite the butyrate-producing genus Romboutsia being significantly decreased, Lactobacillus was significantly increased by the addition of LZM on day 21 of lactation. This was consistent with previous studies, which found lysozyme is beneficial to enrich Lactobacilli in the intestine of breastfed infants [10][11][12] . This possibly means that LZM supplementation in this study decreased SCFAs-producing genera and decreased the SCFAs. However, we did not find any changes in the related phenotypic indicators. This may be attributed to the reduction of SCFAs, which had not reached the threshold that causes phenotypic changes.
The increase in zonulin concentrations in blood may reflect the impaired intestinal mucosal integrity 40 . It is also the case that low intestinal microorganism richness has been associated with increased intestinal permeability in overweight pregnant women 41 . In this study, we observed that sows which had been fed LZM 150 and 300 diets had both decreased levels of zonulin in their serum on days 7 and 21 of lactation, but although no difference was found on day 1 of lactation. These findings indicate that the gut permeability of sows decreased with the addition of lysozyme suggesting lysozyme may modulate the intestinal barrier integrity in sows. While microbiota community richness was increased by the lactation progress, we didn't observe a decrease in gut permeability in late lactation. This may due to the differences between human and sows, or could be because the microbiota richness variety hasn't reached a threshold which induces changes in gut permeability. Correlation analysis at the phylum level found that Bacteroidetes and Spirochaetes were positively correlated with serum zonulin. The 16 s rRNA results found that LZM reduced the abundance of Bacteroidetes and Spirochaetes, meaning that serum zonulin would decrease. Although Tenericutes was negatively correlated with serum zonulin, and the results of 16 s rRNA found that LZM reduced the abundance of Tenericutes, meaning that the concentration of serum zonulin may increase, but this was not consistent with the detected results. The above results indicate that there are many influencing factors to affect serum zonulin, and the change in its concentration was a result of multiple factors. www.nature.com/scientificreports www.nature.com/scientificreports/ At the genus level, consistent with a zonulin concentration decrease in serum, sows which had been fed LZM 150 and LZM 300 diets had decreased Cellulosilyticum on day 7 of lactation, while sows which had been fed LZM 300 diets had increased Staphylococcus on day 21 of lactation. Spearman's correlation found that Cellulosilyticum was positively correlated with serum zonulin, while Staphylococcus was negatively correlated with the serum zonulin (Fig. S5B). This possibly meant that sows which had been fed LZM diets had decreased phylum Bacteroidetes and Spirochaetes, genus Cellulosilyticum and increased Staphylococcus, resulting in a decrease in zonulin concentration, which in turn reflects a decline in gut permeability. Additional studies should be performed in order to demonstrate this hypothesis.
It also is reported that butyrate maintains an intestinal barrier integrity via a modulation of expression of tight junction proteins 42,43 . However, we found that sows which had been fed LZM 150 and LZM 300 diets both had a decreased concentration of butyrate and decreased endotoxin concentration in serum and feces. Endotoxin generally refers to lipopolysaccharide (lipid A) and it is a pyrogen, which is mainly present in the bacterial cell wall of gram-negative bacteria 44 . At the phylum level, we found that LZM reduced the gram-negative bacteria Bacteroidetes, but increased the gram-negative bacteria Proteobacteria. However, in terms of the proportion of microorganisms, Bacteroidetes was 11.95%, while Proteobacteria was 6.54%. This may be one important reason why LZM can reduce endotoxin levels in the serum and fecal material. In addition, phylum Spirochaetes decreased with the addition of LZM. One study demonstrated that many of the organisms within the Spirochaetes phylum cause prevalent diseases. Pathogenic members of this phylum include Brachyspira pilosicoli and Brachyspira aalborgi, which both cause intestinal spirochaetosis 45 . Based on these observations, we suggest that while the addition of LZM does not increase the content of butyrate, it reduces intestinal permeability and endotoxin content as well as reducing level of Spirochaetes, which may be beneficial to intestinal health.
At the phylum level, we observed that sows which had been fed LZM 150 and LZM 300 diets decreased Actinobacteria and Tenericutes on days 1 and 21 of lactation. The two kinds of phylum were thought to be associated with inflammation responses. Patients with inflammatory bowel disease 46 as well as obese Ossabaw minipigs 47 have high levels of Actinobacteria at the colon. In general, chronic or low-grade inflammation is a hallmark of obesity and metabolic disorders 48,49 . It has also been found that, along with gestational weight gain induced by the addition of fat, the phylum Actinobacteria increased. From the inflammatory indicators they found that the sows were in high-grade inflammation 2 . Similarly, high levels of phylum Tenericutes have been found in diet-induced obese mice 50 and obese Göttingen pigs 47 . We found that Tenericutes decreased with the addition of LZM. This effect of LZM was similar with the report of Everard et al. 51 who found that the addition of probiotic yeast decreased the abundance of Tenericutes. Given these observations, we hypothesized that sows fed LZM diets present with low-grade inflammation compared with control. Consistent with this hypothesis, lipocalin-2 and pro-inflammatory cytokines TNF-α in feces were decreased in the LZM 150 and 300 groups, while the IL-10 (anti-inflammatory cytokines) increased in the LZM 150 group on days 7 and 21 of lactation. Consistent with these findings, our previous study found that LZM 150 and LZM 300 increased IL-10 in serum on day 7 of sow lactation, and had a tendency to enhance IL-10 in serum on day 1 of lactation 22 , and previous studies also found that lysozyme are involved in modulating the inflammatory response 14,15 . Cytokines regulate both immune responses 52 and nutrient distribution which, characterized by proinflammatory cytokines, induce more nutrients to enter the immune response rather than being used for growth 53,54 . Lower TNF-α was found in pigs which had been fed lysozyme diets than in control groups 20 . Phylum Proteobacteria contains multiple pathogens and also has proinflammatory properties 55 . High levels of Proteobacteria are a potential diagnostic feature of epithelial dysfunction and dysbiosis in gut microbiota 56,57 . We found that LZM supplementation at 300 mg/kg enriched the Proteobacteria on day 1 of lactation. However, we didn't find any cytokine differences between treatments. Based on these findings, we propose that pro-inflammatory or anti-inflammatory factors are affected by changes in microbial flora; not only by a single microbe, but by the results of multiple microbial interactions. However, the possible mechanisms for how gut microbiota interact to cause intestinal inflammation in perinatal period maternal body require further study.
In conclusion, this study suggested that gut microbiota and their metabolites were altered in sows following the addition of LZM to their diets during the period of late gestation to lactation. Typical changes involve the expansion of diversity among treatments, an increase in Lactobacillus as well as a decrease in Romboutsia, Spirochaetes, Actinobacteria and Tenericutes in LZM 300 treatment. Microbial changes triggered by the additional of LZM during late gestation to lactation could potentially be the mechanism for the positive effects seen of LZM on gut permeability and gut health regulation, as well as maternal anti-inflammation. Until now, there has been little knowledge about the maternal microbial alterations induced by LZM supplementation, meaning that special attention should be given to this area in future research. Using lysozyme in maternal diets from late gestation to lactation may be a beneficial way of preventing gut permeability, related inflammation status and metabolites disturbances with a remodeled microbial ecosystem.

Materials and Methods
All animal procedures in this study were approved by the Animal Care and Use committee of Sichuan Agricultural University, under ethical approval number DKY-S20156137. All the experiments were performed in accordance with the guidelines and regulations of the Animal Care and Ethical Committee of the Sichuan Agricultural University.
Animals and experimental design. A total of 60 Yorkshire × Landrace sows (3-6 parity, 14.07 ± 2.58 mm backfat thickness) were used in this study, which was designed as a single factorial arrangement of treatments including the control group (basal diet, n = 20), LZM 150 diet group (basal diet + 150 mg/kg lysozyme, concentration in accordance with the results of antibacterial tests, data not shown, n = 20), and the LZM 300 diet group (basal diet + 300 mg/kg lysozyme, n = 20). The sows received treatment diets from day 85 of gestation (2020) 10:3210 | https://doi.org/10.1038/s41598-020-60131-1 www.nature.com/scientificreports www.nature.com/scientificreports/ (usually farrowing at day 114 of gestation) to the end of weaning (day 21 of lactation). Based on a corn-soybean meal, the diet was formulated to meet nutritional requirements recommended by the National Research Council 2012 (NRC 2012). The gestation basal diet included 3.04 Mcal digestible energy (DE)/kg, 14.65% of crude protein, 0.69% Lys, 0.85% calcium, and 0.67% phosphorus, while the lactation basal diet included 3.29 Mcal DE/kg, 17.54% of crude protein, 0.99% Lys, 0.99% calcium and 0.68% phosphorus. None of the diets included antibiotics, probiotics or other medicines. The LZM used in the study was given by Shanghai Longyou Biotechnology Co., Ltd. Enzyme activity is 5000U/mg.
During gestation (day 85 to 106), all sows were housed in individual gestation stalls. On day 107 of pregnancy, sows were transferred to individual farrowing crates. Before farrowing, sows were fed an average of 3.5 kg/d diet. On farrowing days, sows were fed 0.5 kg of feed, which was gradually increased by 1.0 kg/d until the maximum feeding amount was reached as described as before 22 . Sows were then allowed to feed ad libitum throughout the rest of their lactation days. Sows had free access to water in this study.
Sample collection. On day 1 of lactation, 10 sows were randomly selected from each treatment. Used the same sows to collect samples at days 7 and 21 of lactation. Fresh feces were collected by massaging the rectum of each sow 8 . Sows had no diarrhea or other disease before sampling. On day 1, samples were collected within two hours after farrowing. Fresh feces were stored in three sterile tubes and kept in liquid nitrogen before being transferred to −80 °C. The 5 mL blood sample was obtained from the same 10 sows' ear veins for each treatment group before morning meal on days 1, 7 and 21 of lactation. Serum samples were obtained by centrifuging blood for 10 min at 3,000 g at 4 °C, and were then stored at −20 °C for further analysis 22 .

Analysis of fecal pH value and SCFAs.
Fecal pH values (n = 10) were detected following the method of Topping et al. 58 . Briefly, distilled water was used to dilute 0.5 g of feces at a ratio of 1:2 (weight/volume) by a blender. Fecal homogenate was then centrifuged (20 °C) for 15 min at 3,000 g, after which the supernatant was measured by pH meter 58 (PHS-3C pH, Shanghai, China).
Concentrations of the fecal SCFAs (acetate, propionate and butyrate, n = 10) were analyzed as described by Chen et al. 59 . A 2 g feces sample was suspended in distilled water (5 mL) and allowed to stand for 30 min, then it was centrifuged for 10 min at 12,000 g at 4 °C. The 2 mL of sample supernatant was then transferred and mixed with 0.4 mL of metaphosphoric acid. The sample placed at 4 °C for 30 min, and centrifuged again for 10 min at 12,000 g at 4 °C before being transferred to 1.2 mL of supernatant and mixed with 15.2 μL of crotonic acid (210 mmol/L, internal standard), after which 0.3 mL liquid was mixed with 0.3 mL methanol 59 . Finally, 1 μL of supernatant was used to analyze of SCFAs by a gas chromatograph 59  Carlsbad, CA, USA) was used to extract the microbial DNA of thawed fecal sample (0.5 g, n = 6, 6 samples were randomly selected from the 8 sows of which were randomly selected for the metabolic biomarker analysis of each treatment at day 1 of lactation. Samples came from the same 6 sows of days 7 and 21 of lactation of each treatment.). The nucleic acid/protein analyzer (Beckman DU-800, Beckman Coulter, Inc., CA, USA) was utilized to examine the concentration and purity of DNA. Then the DNA samples were sent to perform amplicon pyrosequencing on Illumina HiSeq PE250 platforms and bioinformatics analyses at Novogene Bioinformatics Technology in Beijing, China. The V4 hypervariable region of the 16S rRNA gene was amplified using a forward primer 515 f (5′-GTGCCAGCMGCCGCGGTAA-3′) and a reverse primer 806r (5′-GGACTACHVGGGTWTCTAAT-3′) as described as before 2 .
The linker sequence, barcode sequence in the reads, homopolymer runs exceeding 6 bp, primer mismatches, and sequence lengths shorter than 100 bp were removed. Paired-end reads from the original DNA fragments were merged using FLASH version 1.2.7 60 . Then Quantitative Insights Into Microbial Ecology 61 (QIIME) version 1.7.0. was applied to filter out sequences containing ambiguous bases (N) or low-quality bases of the merging data. Following removed the chimeric sequences in the splicing sequence. High-quality tags were clustered into OTUs utilizing Uparse v7.0.1001 (http://drive5.com/uparse/) at 97% sequence similarity. The Ribosomal Database Project (RDP) classifier Version 2.2 (http://github.com/rdpstaff/) was applied to assign taxonomy for 16S rRNA gene sequences. The relative abundance of each OTU was examined at different taxonomic levels. A Venn diagram was generated for comparison among the OTUs of the three treatments. Alpha diversity values for each sample were assessed by Qiime 1.7.0. To identify bacterial taxa differentially represented among different treatments at genus or higher taxonomy levels, the linear discriminant analysis coupled with effect size (LEfSe) was applied 62 .
Statistical analysis. The original data were checked using Grubbs' test method. If |Xp − − X | å λ (α, n) S, the Xp was considered as the outlier. Before using parametric analyses, descriptive statistics were performed to check the normality and homogeneity of variances. Data of relative abundance at the phylum and genus level in fecal matter were log-transformed before statistical analysis. The data were analyzed using the General Linear Model (GLM) procedures of SAS (V9.3, SAS Institute Inc., Cary, NC, USA) followed by a DUNCAN analysis for Scientific RepoRtS | (2020) 10:3210 | https://doi.org/10.1038/s41598-020-60131-1 www.nature.com/scientificreports www.nature.com/scientificreports/ multiple comparison when the F test in the analysis of variance table was significant. The statistical model was: Y ij = μ + t i + e ij where Yij is the analyzed variable, μ is the overall mean, t is the effect of treatment ( i = 1, 2, 3), and e is the residual error ( i = 1, 2, 3, j = 1…10 or 6). Normality of residual check was performed using PROC UNIVARIATE with Normal and Plot options in SAS. Data for the figures (Figs. 1,2,4, S1) were analyzed using a one-way ANOVA followed by Tukey's test for multiple comparisons using GraphPad Prism analysis software. All data are shown as mean ± SEM. Differences of p < 0.05 were considered significant when, whereas p < 0.10 was considered a trend.
To analyze correlations between microbiota and metabolic parameters, Spearman's correlation in R 3.0.2 with the Rstudio 0.97.310 package and gplots package for the heat map were used. Differences of p < 0.05 were considered significant. Data were corrected by false discovery rate analysis according to the Benjamini-Hochberg method with an α of <0.05 63 .