Different Dietary Protein Sources in Low-Protein Diets Induces Changes in Antioxidant Capacity, Immunity, Fecal Microbiota and Metabolites of Weaned Piglets

Background: The inclusion of high-quality proteins and the protein restriction are commonly used in swine production. Our study was conducted to evaluate the effects of hydrolyzed wheat protein (HWP), fermented soybean meal (FSBM), and enzyme-treated soybean meal (ESBM) in low-protein diets on antioxidant capacity, immunity, fecal microbiota and metabolites of weaned piglets. Methods: A total of 144 weaned piglets were randomly assigned to 3 dietary treatments with 6 replications of 8 piglets per pen. The experiment was divided into phase 1 ( days 0-14) and phase 2 (days 15-28). The dietary treatments contained 15.90% HWP, 15.80% FSBM, and 15.10% ESBM in phase 1, and 7.90% HWP, 7.80% FSBM, and 7.50% ESBM in phase 2, respectively. Results: The ADG of piglets in the ESBM group was higher (P < 0.05) than HWP and FSBM during days 1-28. Compared with HWP and FSBM, piglets in ESBM had higher (P < 0.05) serum levels of FRAP and SOD on days 14, as well as higher (P < 0.05) serum FRAP level on days 28. Piglets in ESBM had lower (P < 0.05) serum levels of DAO and IL-1β than HWP on day 28. ESBM signicantly increased the abundance of Bacteroidetes, Oscillospiraceae and Christensenellaceae and decreased the Clostridiaceae abundance in the feces compared with HWP and FSBM. The PICRUSt analysis revealed that the number of gene tags involved in valine, leucine and isoleucine degradation and lysine degradation in ESBM were lower (P < 0.05) than HWP and FSBM. Piglets in ESBM had higher (P < 0.05) fecal butyrate content compared with FSBM, and ESBM tended to decrease (P = 0.076) fecal cadaverine level. Conclusions: Overall, ESBM had advantages over HWP and FSBM in improving antioxidant capacity, immunity, and fecal microbiota and metabolites of weaned piglets.


Background
In pig industry, weaning induces intestinal disturbances and results in continuous impairment of intestinal barrier function, intestinal in ammation, and oxidative damage [1,2]. Weaned piglets are suddenly forced to undergo a transition from a highly digestible milk to less digestible, plant-based solid diets containing complex protein [3], which could result in maldigestion, diarrhea or even death. Therefore, it is of importance to supply high-quality proteins in diets to alleviate weaning stress in piglets. Animal protein sources including sh meal have been widely used in the diets of weaned piglets as a source of readily digestible, high quality protein. However, high price, limited supplies, and unstable quality of animal protein sources make the nutritionist explore alternative to animal protein sources for weaned piglets. Soybean meal (SBM) is a poorly tolerated protein ingredient for the gastrointestinal tract of weaned piglet as result of various anti-nutritional factors including glycinin, β-conglycinin and trypsin inhibitor, which leads to digestive disorders, immune responses and poor performance [3,4].
Fermentation or bioprocessing of SBM including fermented soybean meal (FSBM), enzyme-treated soybean meal (ESBM) has been shown to reduce the contents of antinutritional factors such as glycinin, β-conglycinin and trypsin inhibitor, as well as improve growth rate of weaned piglets [3]. Moreover, hydrolyzed wheat protein (HWP) is used as a practical, functional protein source to improve growth performance of weaned piglets due to its glutamine and a variety of biologically active-wheat peptides [5,6].
The shortage of dietary protein resources and serious environmental contamination have been the main restrictive factors in the sustainable development of pig industry. Recent studies have reported that lowprotein diets supplemented with crystalline amino acids could save protein resources, decrease nitrogen emission and feed costs of pigs as well as reduce the incidence of intestinal injury without in uencing pig performance in comparison with traditional diets [7][8][9]. There is an emerging interest on the relationship between protein ingredients and gut health. The gastrointestinal tract of pigs is colonized with a large and diverse group of microbial community that could confer multitudinous physiological functions to the host, such as metabolism of nutrients, production of short-chain fatty acids (SCFAs), and development of intestinal mucosa [10][11][12]. Dietary protein is mainly decomposed into peptides and free amino acids in the foregut under the catalytic reaction of proteases and peptidases, and the fermentation of undigested and endogenous protein takes place in the hindgut [13,14]. The degradation of undigested protein in the hindgut could produce a variety of bene cial metabolites including SCFAs or branchedchain fatty acids (BCFAs) [15], and some potentially toxic products such as ammonia, biogenic amines, phenolic and indolic compounds [16]. Therefore, low-protein diets and better digestibility of protein could decrease transfer of the undigested protein into the hindgut and reduce production of potentially toxic products from microbial metabolism.
Although plant protein sources including HWP, FSBM and ESBM have been widely used in the diets of weaned piglets, little work has been done on comparing the effects of these different protein sources in low-protein diets on immune function and protein fermentation in the hindgut of piglets. The objective of this study was to compare the effects of different dietary protein sources such as HWP, FSBM and ESBM in low-protein diets on immunity, fecal microbiota and metabolites of weaned piglets.

Materials And Methods
All animal procedures used in this experiment were reviewed and approved by the Institutional Animal Care and Use Committee of China Agricultural University (Beijing, China). The study was conducted at the China Agricultural University Animal Experimental Base (Hebei, China). The three protein sources used in this study were HWP (hydrolyzed wheat protein; JPC56, Joosten, Weert, The Netherlands), FSBM (fermented soybean meal; Yuanyao Biotechnology Co., LTD, Jiangsu, China) and ESBM (enzyme-treated soybean meal; HP300, Hamlet Protein, Horsens, Denmark).
Animals, diets, and experimental design A total of 144 Duroc × Landrace × Large White weaned piglets, with an initial body weight (BW) of 7.75 ± 0.96 kg, were randomly assigned to 3 treatments with 6 replications of 8 piglets (4 barrows and 4 gilts) per pen. The whole experimental period lasted for 28 d, and contained phase 1 ( days 0-14) and phase 2 (days 15-28). The dietary treatments included (1) HWP diet: 15.90% JPC56 replacing soybean meal in phase 1 and 7.90% JPC56 replacing soybean meal in phase 2; (2) FSBM diet: 15.80% FSBM replacing soybean meal in phase 1 and 7.80% FSBM replacing soybean meal in phase 2; (3) ESBM diet: 15.10% HP300 replacing soybean meal in phase 1 and 7.50% HP300 replacing soybean meal in phase 2. The experimental diets had similar level of crude protein (CP, 18.0%). All diets were formulated to meet or exceed the nutrient requirements of National Research Council (NRC 2012) [17] for nursery pigs ranging from 7 to 11 kg (or 11 to 25 kg) BW and fed as mash (Table 1).
All piglets were housed in an commercial at-deck pens with duckbill drinkers, adjustable feeders and plastic slatted oors. Feeds and water were provided ad libitum for piglets. The room temperature and relative humidity were maintained at 24°C to 26°C and 60% to 70%, respectively. Piglet were weighed individually after fasting on days 14 and 28 to calculate average daily gain (ADG), average daily feed intake (ADFI) and feed conversion ratio (FCR). Fecal score were determined by clinical signs of fecal consistency every day based on the methods described by Pan et al. [18] and a scoring system was applied to indicate the presence and severity of diarrhea as following: 1 = hard feces; 2 = slightly soft feces; 3 = soft, partially formed feces; 4 = loose, semiliquid feces; and 5 = watery, mucous-like feces.

Sample collection and preparation
All diet samples were collected and stored at 4°C for further chemical analysis. On days 14 and 28, fasting blood samples (approximately 10 mL) were collected from 1 piglet per pen (closest to the average body weight of each pen) via the jugular vein into vacutainer tubes (Greiner Bio-One GmbH, Kremsmunster, Austria). Serum samples were obtained by centrifugation at 3000 × g for 10 min at 4°C, and then stored at −20°C until analysis of serum parameters. From d 26 to 28, approximately 200 g of feces was collected from each pen for 3 d and stored at −20°C. Then the 3 d collection of feces was pooled by pen and dried at 65°C for 72 h to determine apparent total tract digestibility (ATTD) of nutrients. All samples were ground to pass through a 1-mm screen (40 mesh) before analysis. At the end of experiment, six fresh fecal samples (1 piglet per pen) in each treatment were rapidly frozen in liquid nitrogen, and then stored at −80°C for further analysis of intestinal bacteria, SCFAs and nitrogen metabolites.

Chemical analysis
Diets and feces were analyzed in duplicate for the contents of dry matter (DM, method 934.01) and crude protein (CP, method 990.03) according to the Association of Official Analytical Chemists (AOAC, 2006) [19]. Gross energy (GE) was determined by an Automatic Energy Analyzer (Parr 1281, Moline, IL, USA).
The chromium (Cr) content in the diets and feces was analyzed using an Atomic Absorption Spectrophotometer (Z-5000; Hitachi, Tokyo, Japan) according to the procedure of Williams et al. [20].
DNA extraction, PCR ampli cation, Illumina MiSeq sequencing and data analysis Microbial DNA was extracted from the fecal samples using the DNA Kit (Omega Bio-tek, Norcross, GA, USA). The integrity of DNA was assessed by agarose gel electrophoresis. Isolated DNA was used as a template for ampli cation of the 16S rRNA gene V3-V4 region using universal primers 338F (5'-barcode-ACTCCTRCGGGAGGCAGCAG-3') and 806R (5'-GGACTACCVGGGTATCTAAT-3'), where barcode is an eightbase sequence unique to each sample. After purification, amplicons were pooled and paired-end sequenced on the Illumina MiSeq platform (Illumina Inc., San Diego, CA, USA). Raw fastq files were demultiplexed, and quality-filtered using QIIME(version1.17). Operational taxonomic units (OTUs) were clustered with 97% similarity cutoff using UPARSE and chimeric sequences were removed using UCHIME. The Ribosomal Database Project (RDP) Classifier (http://rdp.cme.msu.edu/) was used to analyze the taxonomy of each 16S rRNA gene sequence with confidence greater than 70%. The Venn diagram was present based on the occurrence of shared and unique OTUs among different treatments. α-diversity including observed species, Chao, Ace, Shannon, Simpson, Good's coverage was performed using MOTHUR (v.1.31.2). The Unifrac metric was used to calculate β-diversity and UPGMA tree was displayed by QIIME (v1.80) based on Unweighted UniFrac distances of the OTUs community. Principal component analysis (PCA) was conducted by R software (v3.1.1) according to the OTUs among different treatments. LefSe analysis was performed to identify the biomarkers for microbial communities from phylum to genus based on the relative abundance of annotated taxonomic profiling.

Predictive metabolic functions of microbial communities
Predictive metabolic functions of the bacterial community in the fecal samples were conducted by Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) based on 16S rRNA gene sequencing [21]. OUTs were used to predict functional pro ling of microbial communities by referencing the Kyoto Encyclopedia of Genes and Genome (KEGG) orthology database [22].

Measurements of SCFAs and BCFAs by high performance ion chromatography
The fecal samples were prepared for analyzing SCFAs and BCFAs as previously described by He et al. [23]. Samples (0.5 g) were weighed and dissolved in 8 mL of ultrapure water. After ultrasound for 30 min, fecal samples were centrifuged at 3000 × g for 5 min. The supernatant was diluted (1:50) and then filtered through a 0.22 µm filter. Extracted sample solution (25 µL) was analyzed for SCFAs (acetate, propionate, butyrate), BCFAs (isobutyrate, valerate and isovalerate) using a high performance ion chromatography system (DIONEX ICS-3000, Thermo Fisher, Waltham, MA, USA). The SCFAs and BCFAs were separated by an AS11 analytical column (250 × 4 mm) and an AG11 guard column using potassium hydroxide. The concentration of SCFAs and BCFAs were expressed as mg/g of feces.

Analysis of ammonia nitrogen in feces
The content of ammonia nitrogen (NH 3 -N) was analyzed based on the method as described by Chen et al. [24]. In short, 0.5 g of samples was dissolved in 5 mL of ammonia-free water and centrifuged at 5000 × g for 15 min. Next, 19 mL of ammonia-free water and 1 mL of potassium sodium tartrate was added into 1 mL of collected supernatant in a 50-mL sterile tube, followed by the addition of 1.5 mL of Nessler's reagent. The absorbance of nal mixture was determined at 420 nm against ammonia-free water using a UV-vis Spectrophotometer (MAPADA, Shanghai, China).

Determination of biogenic amines by high performance liquid chromatography
The concentrations of biogenic amines (putrescine, cadaverine and spermine) were determined by high performance liquid chromatography (HPLC) as previously described by Li et al. [25] with modi cation.
Briefly, 0.5 g of fecal samples was added into a 2-mL centrifuge tube and mixed with 1 mL of trichloroacetic acid. The mixture was centrifuged at 3600 × g for 10 min, the supernatants were then vortexed with the same volume of n-hexane for 5 min. The aqueous phase was extracted again in the same manner. Internal standard (20 mL) was added into the extracts, followed by the addition of 1.5 mL of saturated sodium bicarbonate, 1mL of dansyl chloride, and 1mL of sodium hydroxide. The mixture was subsequently heated at 60°C for 45 min with occasional shaking. Next, 100 μL of ammonia was added into the mixture and kept at 40°C in a water bath under N 2 condition. Finally, the analyzed sample was produced by adding acetonitrile to the residue. Each sample (20 μL) was injected into Agilent HPLC 1200 series equipped with a reversed-phase ZORBAX 80 A Extend-C18 column (250 mm × 4.6 mm i.d.; 5 µm particle size, Agilent, Santa Clara, USA). The ow rate, wavelength and column temperature were set at 1.0 mL/min, 254 nm and 40°C, respectively. The concentration of biogenic amines were expressed as μg/g of feces.
Statistical analysis Data analyses of growth performance, serum parameters, SCFA concentrations, and nitrogen metabolites were analyzed by One-Way ANOVA using the GLM procedure of SAS 9.4 (SAS Inst. Inc., Cary, NC), with results presented as mean ± SEM. Differences in the fecal score were determined by the χ 2 contingency test. Microbiota diversity metrics were conducted from standardized OTU reads using R software (version 3.2.2). The relative abundance of fecal microbiota composition was analyzed by the Kruskal-Wallis method. Differences were considered significant at p < 0.05, and tendency was declared with 0.05 ≤ p < 0.10.

Results
Growth performance and fecal score As shown in Table 2, piglets in the ESBM group had higher (P < 0.05) BW on days 14 and 28 than that of piglets in the HWP and FSBM groups. The ADG of piglets in the ESBM group was increased (P < 0.05) compared with piglets in the HWP and FSBM groups during days 1-14 and 1-28, and the ADG of piglets in the ESBM group was increased (P < 0.05) compared with the HWP group during days 15-28. The ADFI of piglets in the HWP group was lower (P < 0.05) compared with the FSBM and ESBM groups. Piglets in the ESBM group had lower (P < 0.05) FCR than the HWP and FSBM groups during days 1-14, and had lower (P < 0.05) FCR than the FSBM group during days 1-28. No difference was observed for fecal score among all dietary treatments.

Nutrient digestibility
Piglets fed with ESBM or HWP had higher (P < 0.05) ATTD of CP compared with piglets fed with FSBM ( Fig. 1). No difference was observed for ATTD of GE among dietary treatments.

Serum oxidative status
Piglets in the ESBM group had higher (P < 0.05) serum FRAP level than that of piglets in the HWP and FSBM groups on days 14 and 28 (Table 3). Serum activity of SOD was highest in piglets fed with ESBM diet on days 14 and 28. No signi cant differences were observed for serum levels of AOPP, MDA, CAT and GSH-Px among all dietary treatments.

Serum DAO, endotoxin and D-lactate
Piglets in the ESBM group had lower (P < 0.05) serum DAO level than that of piglets in the HWP group on day 28 (Table 4). No differences were observed for serum levels of endotoxin and D-lactate among treatments.

Concentrations of inflammatory cytokines and immunoglobulins in serum
As shown in Table 5, ESBM tended to decrease concentrations of IL-1β (P = 0.060) and IL-6 (P = 0.070) in serum of piglets on day 14. Piglets in the ESBM group had lower (P < 0.05) serum IL-1β level compared with the HWP group on day 28. No signi cant differences were observed for serum levels of IgA, IgG and IgM among all dietary treatments.

Fecal bacterial community
The OUT Venn analysis identi ed 47, 34 and 63 unique OTUs in the HWP, FSBM and ESBM groups, respectively ( Fig. 2A). The Shannon index was higher (P < 0.05) in the fecal samples of the ESBM group compared with the HWP and FSBM groups, and the Simpson index was lower (P < 0.05) in the fecal samples of the ESBM group compared with the FSBM group (Fig. 2B). Principal component analysis (PCA) plot based on the OUT level showed that greater variations were observed in the fecal samples of the ESBM group compared with the HWP and FSBM groups, and the HWP and FSBM groups had similar microbial communities (Fig. 3A). The UPGMA tree showed that signi cant differences in the structure of fecal microbiota among different treatments, indicating that the protein sources had important impacts on fecal microbial communities (Fig. 3B).
At the phylum level, the dominant bacteria were Firmicutes and Bacteroidetes, which accounted for 90% (Fig. 4A). The abundance of Bacteroidetes in the fecal samples of the ESBM group was higher (P < 0.05) compared with the HWP and FSBM groups (Fig. 5A). At the family level, the predominant families within the Firmicutes phylum consisted of Christensenellaceae, Clostridiaceae, Erysipelotrichaceae, Lachnospiraceae, Lactobacillaceae, Oscillospiraceae, Peptostreptococcaceae, and Streptococcaceae, while Muribaculaceae was the dominant family in the Bacteroidetes phylum (Fig. 4B). The abundance of Clostridiaceae in the fecal samples of the ESBM group was lower (P < 0.05) than the HWP and FSBM groups, and ESBM increased (P < 0.05) the abundances of Oscillospiraceae and Christensenellaceae compared with HWP and FSBM (Fig. 5B).
Prediction on amino acid metabolism of bacterial communities using PICRUSt Ten pathways associated with amino acid metabolism were shown in Fig. 7, including tryptophan metabolism, cysteine and methionine metabolism, phenylalanine metabolism, glycine, serine and threonine metabolism, tyrosine metabolism, valine, leucine and isoleucine degradation, lysine degradation, arginine and proline metabolism, histidine metabolism, as well as alanine, aspartate and glutamate metabolism. The number of the genes associated with valine, leucine and isoleucine degradation and lysine degradation in the ESBM group decreased (P < 0.05) compared with the HWP and FSBM groups. No differences were observed for other pathways associated with amino acid metabolism among all dietary treatments.

Fermentation metabolites of fecal samples
The concentration of SCFAs in the fecal samples of piglets fed diets with different dietary protein sources was shown in Table 6. Butyrate concentration in the fecal samples of the ESBM group was higher (P < 0.05) than that of the FSBM group. No difference was observed for NH 3 -N among all dietary treatments ( Table 7). The content of cadaverine in the fecal samples of the ESBM group tended to be lower (P = 0.076) compared with other groups.

Discussion
Restriction on the use of animal products in livestock has aroused interest in replacing animal protein sources with plant protein sources, and this need becomes more pronounced when the diet is focused on weaned piglets. In pig industry, plant protein sources such as HWP, FSBM and ESBM have been widely used in the diets of weaned piglets to improve growth performance [3,5,6,26]. However, little information was available towards comparing the effect of HWP, FSBM and ESBM on growth performance of weaned piglets. In this study, we found that piglets fed with ESBM had higher ADG than that of piglets fed with HWP and FSBM, indicating that piglets in ESBM had a higher growth rate compared with other diets. However, there were different reasons for lower ADG in piglets fed with HWP or FSBM compared with ESBM. Our result showed that no difference was observed for FCR between HWP and ESBM groups, but piglets fed with HWP had lower ADFI than that of piglets fed with ESBM during the whole period, indicated that lower ADFI in HWP may contribute to lower ADG in HWP compared with ESBM group. The inclusion of wheat gluten could decrease ADFI by increasing digesta viscoelasticity and affecting the transit of digesta [27]. In addition, no difference was observed for ADFI between FSBM and ESBM groups, but piglets fed with FSBM had higher FCR than that of piglets fed with ESBM, indicated that higher FCR in FSBM may contribute to lower ADG in FSBM compared with ESBM group. However, Ma et al. [3] showed that there was no signi cant difference for FCR between FSBM and ESBM groups. The inconsistency may be explained by the different experimental condition, diet composition, and inclusion level. Feed is the main source of energy and protein for animal maintenance and growth requirement. The digested and absorbed nutrients by gastrointestinal tract provide energy and amino acid for animals maintenance and growth via biological oxidation and metabolism [28]. In this study, piglets fed with FSBM had higher ADFI than that of piglets fed with HWP. However, no difference was observed for ADG between HWP and FSBM groups. Interestingly, piglets fed with HWP had higher ATTD of CP compared with piglets fed with FSBM, which may explain no difference for ADG between HWP and FSBM groups.
Weaning could induce oxidative stress [3], which is considered to be an imbalance process between the production rate of reactive oxygen species (ROS) and their clearance via the antioxidant system [29]. Antioxidant capacities in serum re ect the host's capacity to respond to endogenous oxidative damage, and improving the antioxidant capacities has positive effects on oxidative stress [30]. Normally, excessive oxidative radicals are eliminated by antioxidant systems including nonenzymatic antioxidant defensive components and a series of antioxidant enzymes. FRAP is considered a useful indicator of "the total antioxidant power" of plasma, re ecting the combined antioxidant effect of non-enzymatic compounds of endogenous and exogenous origin in biological uids [31]. In this study, piglets fed with ESBM had higher serum FRAP level than that of piglets fed with HWP or FSBM, indicating that ESBM had an advantage over HWP and FSBM in improving non-enzymatic antioxidant defense. Weaning stress could decrease serum concentrations of SOD and GSH-Px, as well as improve serum MDA and hydrogen peroxide (H 2 O 2 ) contents in piglets [32]. Our results showed that piglets fed with ESBM had higher serum SOD concentration than that of piglets fed with HWP or FSBM. It has been known that SOD acts as the rst line of defense against excessive oxidative radicals via catalyzing the dismutation of the endogenous superoxide radical to H 2 O 2 , which is decomposed into H 2 O and O 2 by GSH-Px and CAT [33].
Our result suggested that ESBM could have higher ability to improve antioxidant capacity via activating the enzymatic and nonenzymatic antioxidant defensive systems compared with HWP or FSBM.
The intestinal epithelial barrier consists mainly of a single layer of enterocytes and intercellular multiprotein complexes, which can maintain intestinal homeostasis by preventing the ows of pathogens, toxins, and antigens from the gut lumen into the circulating system [8,34]. Weaning stress could induce imperfect and impaired gastrointestinal tract, which is associated with increased permeability, resulting in translocation of pathogens through the intestinal epithelial cells into systemic circulation [8]. DAO, endotoxin and D-lactate are intestinal permeability-related biomarkers and their levels in serum could re ect the intestinal barrier function [35]. In this study, HWP increased serum DAO concentration on day 28 compared with ESBM, indicating that piglets in the ESBM group had better intestinal barrier function. Weaning process leads to intestinal in ammation and promotes the release of pro-in ammatory cytokines, which can explain the stunted growth performance and intestinal disorders of piglets [36]. Numerous pro-in ammatory cytokines including IL-1β, IL-6, and TNF-α are involved in the regulation of the in ammatory response and their levels could re ect the in ammatory status in the body [35]. Our result showed that ESBM decreased serum levels of IL-1β and IL-6 compared with HWP and FSBM, indicating that ESBM was superior to HWP and FSBM in alleviating weaning stress-induced in ammatory response.
The intestinal microbiota of swine is the diverse and dense community of microorganisms that could confer multifarious physiological functions to the host such as nutrient metabolism and development of the intestinal mucosa [8]. Diet composition, particularly macronutrients such as carbohydrates and protein, re ects the substrates available for the intestinal microbiota and in uences microbial composition and metabolism, which in turn could improve animal performance and health [25]. Normally, the diet ingredients are digested and absorbed in the foregut, and the undigested residues as well as endogenous compounds ow into the hindgut for microbial fermentation, which modulates the intestinal microbiota and bacterial metabolites [8,37]. Carbohydrates fermentation is bene cial to intestinal epithelial cells due to the production of SCFAs [25,38], while protein fermentation has adverse impacts on intestinal health due to potentially toxic metabolites such as ammonia, biogenic amines, and aromatic compounds that are related to proliferation of pathogenic bacteria [8,25]. In this study, we focus on the fecal microbiota composition and metabolites in piglets fed with different dietary protein sources including HWP, FSBM, and ESBM. The α-diversity is regarded as indicators of functional resilience of microbial ecosystems in the intestine, including species richness (Observed species, Chao, and Ace) and species diversity (Shannon, Simpson, and Good's coverage) [39]. Our results showed that no changes were observed for species richness of the fecal samples in piglets fed with HWP, FSBM, and ESBM. However, ESBM increased Shannon index in the fecal samples of piglets compared with HWP and FSBM, and ESBM decreased Simpson index in the fecal samples of piglets compared with FSBM, revealing that ESBM had an advantage over HWP and FSBM in improving fecal bacterial diversity without affecting bacterial richness. The β-diversity by PCA and UPGMA tree analysis showed that fecal microbiota responded differently to dietary protein sources. The reason for greater variations in the fecal microbiota structure of ESBM compared with other groups was unclear, but fecal microbiota structure of HWP and FSBM was similar, which may partly explain similar growth performance of piglets. At the phylum level, Firmicutes and Bacteroidetes were the most predominant bacterial phyla in the fecal samples, which is similar to the results conducted by Zhao et al. [40]. In this study, the addition of ESBM to weaned piglets signi cantly increased the relative abundance of Bacteroidetes in the fecal samples of piglets compared with HWP and FSBM. The phylum Bacteroidetes are considered as effective degraders for plant polysaccharides and other recalcitrant organic carbon and nitrogen sources [41]. The higher abundance of Bacteroidetes are known to help shape the metabolic milieu of the intestinal ecosystem and prevent diarrhea because they could interact with the intestinal immune system, suggesting the relationship between early bacterial colonization and intestinal immune maturation after weaning [42]. Although no difference was observed for fecal score among different protein sources, higher abundance of Bacteroidetes in piglets fed with ESBM could contribute to improvement in the intestinal immunity and further promote growth performance of piglets. Down to the family level, piglets fed with ESBM had lower abundance of Clostridiaceae and higher abundance of Oscillospiraceae and Christensenellaceae in the fecal samples of piglets compared with HWP and FSBM. In our study, these ndings were also con rmed by LEfSe analysis, which can identify unique high-dimensional biomarkers for analyzed microbial communities [43]. Certain bacterial species, from the family Clostridiaceae, were implicated in poor performance [25]. In addition, Clostridiaceae comprises certain pathobionts. For instance, Clostridium perfringens, belonging to the family Clostridiaceae, may not necessarily compromise the health of the pig, but they could have the potential to zoonotic diseases if excreted in the form of feces [44]. In previous studies, the Oscillospiraceae and Christensenellaceae families have been shown to include microorganisms with potential health bene ts and contribute to the formation of secondary bile acids that are known to protect against infection with Clostridium di cile [45,46]. Butyrate is one of the important microbial metabolites that could provide energy for colonocytes and has bene cial in uences on epithelium proliferation and differentiation [47]. The family Christensenellaceae is also considered as a butyrate producer and play a key role in maintaining gastrointestinal tract structure and function [34], which may be an explanation for the increased butyrate content in the fecal samples of piglets fed with ESBM. Together, these ndings suggested that dietary protein sources could differentially in uence fecal bacterial community of weaned piglets.
Dietary carbohydrates and proteins can be fermented in the colon to produce SCFAs and BCFAs [14]. SCFAs are derived from fermentation of dietary carbohydrates and protein, and deamination of valine, leucine, and isoleucine could produce BCFAs [48]. In this study, no difference was observed for BCFAs among all dietary treatments. However, using the PICRUSt program to predict amino acid metabolism of bacterial communities, the number of gene tags involved in valine, leucine, and isoleucine degradation in the ESBM group were signi cantly decreased compared with HWP and FSBM groups. In this study, the reason for this result may be that the gene tags are primarily related to decarboxylation, and it is necessary to further investigate the relationship between dietary protein sources and the degradation of valine, leucine, and isoleucine. The microbial fermentation of amino acids in the hindgut generates some potentially toxic products including ammonia, biogenic amines, phenols, and indoles [16], which are closely related to the proliferation of pathogenic bacteria and depress the growth performance of pigs [14]. The low-protein diets supplemented with an adequate level of crystalline amino acids may deter the formation of protein fermentation products in the large intestine without diminishing the growth performance of pigs [7]. NH 3 -N is a toxic metablite produced by the deamination of amino acids, which can interfere with epithelial cell turnover and reduce growth performance of pigs [24,25]. Our results showed that no difference was observed for NH 3 -N content in the fecal samples among all dietary treatments, indicating that microbial deamination of dietary and endogenous protein was not altered by the addition of HWP, FSBM, and ESBM to low-protein diets. However, ESBM tended to decrease fecal cadaverine content compared with HWP and FSBM, which revealed a decline in protein fermentation. Cadaverine is produced by microorganisms through the decarboxylation of lysine in the large intestinal lumen [49,50]. The PICRUSt analysis revealed that the number of gene tags involved in lysine degradation in the ESBM group were signi cantly decreased compared with HWP and FSBM groups. The result may contribute to explain decreased cadaverine content in the fecal samples of piglets fed with ESBM.

Conclusions
Different protein sources in low-protein diets could differentially modulate antioxidant capacity, immune function, and fecal bacterial community and metabolites of weaned piglets. ESBM had advantages over HWP and FSBM in improving growth performance and health of weaned piglets, which is mainly re ected by the increased ADG, the improved antioxidant status and immunity via increasing serum FRAP and SOD levels and decreasing serum DAO and IL-1β concentrations, and the improved fecal microbiota composition and metabolites including lower abundance of Clostridiaceae, higher abundance of Bacteroidetes, Oscillospiraceae and Christensenellaceae, enhanced fecal butyrate content and reduced protein fermentation.

Availability of data and material
All data generated or analyzed during this study are included in this published article.  Values are expressed as the means ± SEM, n = 6. a,b Means within a row with different letters differ at P < 0.05. Values are expressed as the means ± SEM, n = 6. a,b Means within a row with different letters differ at P < 0.05.  Values are expressed as the means ± SEM, n = 6. a,b Means within a row with different letters differ at P < 0.05. Values are expressed as the means ± SEM, n = 6.