Feeding a Low-protein Maternal Diet Affects Qinghai Bamei Piglet Jejunal Microbiome-metabolome Response

This experiment investigated the impacts of feeding a maternal low-CP concentration diet having iso-essential amino acids on new born suckling piglets intestinal microbial composition and metabolic proles. The Bamei swine breed was selected due to high meat quality and avor, but demonstrates slower growth rates which may be related to jejunal nutrient supply. Forty randomly selected purebred Bamei sows were divided into two groups and fed a low dietary CP (12%, LP) or a normal CP (14%, CON) diet, respectively, but formulated to contain similar (iso-) essential amino acid concentrations per current recommendations. At 21 days, 12 piglets were randomly selected from each treatment and euthanized with jejunum content samples collected. The 16S rRNA gene sequencing and mass spectrometry-based metabolomics proling were combined as an integrated approach for evaluating the functional impact of maternal CP concentrations on piglet intestinal microbiome. Even though piglets demonstrated similar 0 to 21 d ADG among treatments, the jejunum relative weight, villus width, crypt depth and muscular thickness were increased (P < 0.05), while villus height, and villus height:crypt depth were reduced (P < 0.05) for the material LP compared to the maternal fed CON diet. Maternal CP concentrations can modify the intestinal microbial composition of Bamei suckling piglets. The relative abundances of the bacterial species Escherichia-Shigella, Actinobacillus, Clostridium_sensu_stricto_1, Veillonella, and Turicibacter were increased (P < 0.05) in the maternal LP fed diet compared with the maternal fed CON diet. Jejunal digesta metabolomics analysis indicated that several amino acids were metabolized (i.e. cys, met, tyr phe and trp), biosynthesized (arg phe, tyr, and trp), or degraded (lys) were enriched (P < 0.05) for the maternal fed LP compared with the maternal fed CON. Correlation analysis demonstrated that certain intestinal bacterial genera were highly related to the histomorphology and altered intestinal microbiota metabolites.


Introduction
The Bamei swine breed is a local swine breed in the Qinghai Province of the People's Republic of China.
Bamei pigs are known for their meat quality and avor, but are known to be slow growing (Jin, 2006;Yang and Gun, 2007). The Qinghai plateau has used both natural and arti cial selection practices for developing Bamei pigs developing a strong adaptability to the plateau, fat deposition, and good meat quality characteristics. However, Bamei's slow growth rates combined with the plateau's low feed quality/digestibility are important constraints limiting the Qinghai's growth potential of the Bamei swine industry. This results in international implications for meat imports along with grains and protein sources that need to be addressed for food production. Bamei adipose growth rates increase dramatically after 35 kg, while muscle growth rates decrease dramatically after achieving 55 kg (Yang and Song, 1991).
China's continuous improvement of people's living standards has resulted in Bamei pork becoming more popular.

Material And Methods
The gastrointestinal tract's microbial ecosystem is dynamic and complex with the composition known to vary widely across healthy individuals (Huttenhower et al., 2012). In the human and animal gastrointestinal tract lives a large and diverse microbial community playing a vital role in host health (Kuang et al., 2019), muscosal immunological environment maturation (Pattaroni et al., 2018), precision medicine development (Kuntz and Gilbert, 2017), and assisting with intestinal barrier integrity (Martinez-Lopez et al., 2019). Over the last decade, numerous studies have reported that the intestinal microbiome composition plays an important role in regulating the metabolic health of both rodents and humans (Kreznar et al., 2017). Recent rodent work suggests the major dietary factors regulating intestinal microbiome taxonomic composition are protein and carbohydrate intake (Holmes et al., 2017).
The intestinal microbiome is a complex and dynamic ecosystem of bacterial species being in a continual state of ux and highly susceptible to numerous environmental factors, especially dietary nutrient supply.
Reducing CP by 2 to 4 percentage units by adding crystalline amino acids (AA) to meet NRC (2012) nutrient recommendations has increased nitrogen utilization, reduced feed costs and nitrogen excretion, while promoting intestinal health and meat quality with similar growth performance (Wang et al., 2018). The high-quality protein source shortage is a global problem, but especially for China's large population.
Since 2002, the world's largest soybean importer is China (Zhang and Reed, 2008). In 2016, China imported 8.391 billion tons soybeans, which is > 26% of global production. Decreasing dietary protein concentrations can effectively reduce pressure on protein source availability (Wang et al., 2018). Many studies demonstrate dietary CP concentrations versus CP source, have a greater impact on intestinal microbiota composition (Rist et al., 2013). Previous studies have focused on changes in large intestinal microbiota, while ignoring the bacteria's role for the small intestine (Dai et al., 2010). Moderate diet protein restriction may alter intestinal microbiota composition while improving adult pig ileal barrier function (Fan et al., 2017). Chen et al. (2018) reported that decreasing dietary CP concentration 3 % units reduced ileal Streptococcus spp., while increasing Lactobacillus spp. and Bi dobacterium spp. These ileal microbiota alterations improved intestinal stem cell proliferation and altered tight junction protein distribution resulting in similar intestinal barrier function. Therefore, feeding dietary LP concentrations has advanced while maintaining essential amino acid supply and has been applied to swine production.
The application of 16S rRNA high-throughput sequencing technology provides methods for determining if maternal dietary CP concentration can alter intestinal microbial composition at different physiological stages and intestinal locations.
The Bamei pig was selected due to its popularity with local consumers, but it is a slower growing swine breed, which needs to be addressed. He et al. (2016) reported that IGF-1, insulin, leptin and amino acids may be associated with slow growth. The hypothesis was that altering maternal dietary CP concentrations would alter the intestinal microbiota and metabolites for the suckling piglets. The 16S rRNA gene sequencing method was integrated with LC-MS metabolomics to analyze maternal dietary LP concentration on piglet intestinal microbiome and metabolite pro les. The relationships between metabolites and microbiota were explored as well.

Ethics Statement
The use of animals was kept to an absolute minimum required to achieve statistical signi cance for validation purposes; a total of 40 animals were used for the work described in this paper. All procedures involving the Huzhu Bamei sows were conducted at the Qinghai Province Huzhu County Bamei Pig Seed Breeding Farm (Huzhu, China). All procedures were conducted in accordance with China animal welfare law Act 2011, approved by institutional ethical review committees (the State Key Laboratory of Plateau Ecology and Agriculture, Qinghai University Animal Ethics Committee)and conducted under the authority of the Project Licence (IACUC permit number: 2016080301). All procedures involving the Huzhu Bamei sows were conducted at the Qinghai Province Huzhu County Bamei Pig Seed Breeding Farm (Huzhu, China) for Scienti c Purposes. These Huzhu Bamei sows were experimental animals, and animal feeding followed the recommendations in the ARRIVE guidelines, animals slaughtering followed the American Veterinary Medical Association (AVMA) Guidelines for the Euthanasia of Animals (2020) and was approved by the National Administration of Swine Slaughtering and Quarantine regulations (Qinghai, China).

Animals and Diets
Forty (40) purebred Huzhu Bamei well body condition (score 4) sows were sourced through the Qinghai Province Huzhu County Bamei Pig Seed Breeding Farm (Huzhu, China) having similar body weight (BW), health status, and 3 to 4 years of age being randomly assigned to one of two treatments (20/treatment). The LP treatment diet (12% CP) was balanced for the ve EAA Lys, Met, Thr, Trp, and Val for their standardized ileal digestibility (SID) concentrations and then decreased CP by 2% compared to a control (CON; 14% CP) diet balanced for the same SID EAA according to Chinese feeding standards for a 90 kg heavy body conditioned sow. The complete diet composition is given in Table 1. After 5 d of facility and diet acclimation, the sows were fed the assigned treatment diet while skipping one estrous cycle (21 days) during natural estrus and then mated. The newborn piglets were maintained with their mothers prior to weaning with litter size, live birth %, birth weights, and diarrhea rates being published previously (Jin et al., 2019). All sows at all time had ad libitum access to feed and fresh water.

Sample Collection
Randomly, 12 piglets were selected from each treatment group, fasted for 12-hour, weighed, and euthanized with 50 mg/kg sodium pentobarbital on day 21 of age. The small intestine was ligated at the pylorus, duodenum, jejunum, and ileum and dissected. The ligated jejunum was weighed. The jejunal contents were sampled at approximately the half-way point of the jejunal length, placed into 1.5 mL sterile polypropylene tubes, and stored in liquid nitrogen until analyses were conducted for intestinal microbiome and metabolome. An approximate 1.5 cm jejunal tissue sample was collected, washed, and placed in 4% paraformaldehyde for histomorphometric analysis at the same time.

Histomorphometric analysis
Jejunal tissue samples xed in 4% paraformaldehyde were embedded in para n (5 µm) and stained with HE (hematoxylin-eosin). In each jejunal section, 12 intact villi were randomly selected from each piglet. The jejunum villus height, villus width, crypt depth, and muscular layer thickness were measured using an image analysis system (Caseviewer 2.0 software, 3DHISTECH, Hungary).

gDNA Extraction, 16S rRNA Gene Sequencing and Microbial Function Prediction
The jejunal content samples were extracted to harvest total bacterial DNA using the PowerSoil® DNA Isolation Kit (MO BIO Laboratories, Inc., Carlsbad, CA, USA) according to the manufacturer's instructions.
The DNA samples were stored at -80℃ until outsourced for analyzing the 16s rRNA gene by BIOMARKER (Beijing, China). The 16S rRNA gene sequence (Illumina HiSeq 2500) was used to measure microbial diversity and bacterial community composition. The extracted DNA was used as a template and PCR was performed using barcode primers located on both sides of the V3-V4 hypervariable region of the bacterial 16S rRNA gene. The primer sequences used were: 338F: 5'-ACTCCTACGGGAGGCAGCA-3' and 806R: 5'-GGACTACHVGGGTWTCTAAT-3'. Ampli cation was performed for 30 cycles using a DNA thermal Cycler (Bio-Rad, Hercules, CA, USA). The rst cycle was at 98°C for 2 min followed by 30 subsequent cycles of 98°C × 30 s, 50°C × 30 s, then 72°C × 1 min, and the last cycle at 72°C for 7 min.
The original DNA fragments from the raw sequencing reads were merged using FLASH v1.2.7 (http://ccb.jhu.edu/software/FLASH/). The reads were assigned to each sample according to the unique barcodes. The selected high-quality reads were used for bioinformatic analysis. Each sample's unique read was clustered into operational taxonomic units (OTU) based on a 97% sequence similarity determined accordingly by UCHIME v4.2 (http://drive5.com/usearch/manual/uchime_algo.html).

Sample Processing for Metabolomics
The samples of jejunal contents were thawed at 4℃. Then, 60 mg were mixed with 200 µL ultrapure water to assist in homogenization, followed by adding 800 µL of methanol/acetonitrile (1:1, v/v). Then samples were vortexed followed by sonication on ice. The samples were incubated for 1 hour at -20℃ to remove protein followed by 15 min centrifugation (13,000 × g at 4°C), The supernatants were collected followed by vacuum drying followed by storage at -80°C until analyzed using ultra-high-performance liquid chromatography equipped with quadrupole time-of ight mass spectrometry (UPLC-Q-TOF/MS). The quality control (QC) samples were prepared following the same procedures as previously described. For the UPLC-Q-TOF/MS analysis, the samples were re-dissolved in 100 µL acetonitrile/water (1:1, v/v). Instrument stability and repeatability was monitored using QC samples prepared by pooling 10 µL of each sample and analyzed after every 10 experimental samples.

UPLC-Q-TOF/MS Analysis
Jejunal content samples metabolic pro les were measured using an Agilent 1290 In nity LC system (Agilent Technologies, Santa-Clara, California, USA) coupled with an AB SCIEX Triple TOF 6600 System (AB SCIEX, Framingham, MA, USA). An ACQUITY UPLC BEH Amide 1.7 µm (2.1 × 100 mm) column for both positive and negative models was used for chromatographic separation. The A mobile phase was 25 mM ammonium acetate and 25 mM ammonium hydroxide in water and the B mobile phase was acetonitrile. The solvent gradient was 85% B mobile phase for 1 min followed by linearly reducing to 65% by 11 min followed by further reduction to 40% in 0.1 min. This mobile phase concentration was maintained for 4 min followed by increasing to 85% in 0.1 min increments with a 5 min re-equilibration period.
The ESI source parameters were: Ion Source Gas1 = 60, Ion Source Gas2 = 60, curtain gas = 30, source temperature = 600℃, and IonSpray Voltage Floating ± 5500 V. In the MS acquisition only mode, the instrument was set to acquire data covering the m/z range of 60-1000 Da. The TOF MS scan accumulation time was set at 0.20 s/spectra and product ion scan accumulation time was 0.05 s/spectra. In auto MS/MS acquisition, the instrument was set to acquire data covering the 25-1000 Da m/z range, and the product ion scan accumulation time was set at 0.05 s/spectra. The product ion scan was acquired using information dependent acquisition by selecting the high sensitivity mode. The parameters were set as follows: collision energy xed at 35 V with ± 15 eV; declustering potential at 60 V (+) and − 60 V (−); isotopes were excluded within 4 Da and candidate ions were monitored at 10 per cycle.

Statistical Analyses
All data were checked for outliers before any statistical analyses were conducted. Data were either plotted or the box and whisker plots and the Shapiro Wilk Test were used to verify that the data were normally distributed (P > 0.15). All data were subjected to least squares analysis of variance (ANOVA) for a completely random design (CRD; Steel and Torrie, 1980) having 2 treatments using SPSS 21 software (SPSS Inc., Chicago, IL, USA). The statistical linear additive model was: Where Y i = depended variable, µ -overall mean, T = treatment of LP or CON and e i = residual random error.
Least squares means were separated using the Least Signi cant Difference (LSD) and signi cant was declared at P < 0.05.
Microbial Data Analysis. The OTU were rari ed based on several metrics for alpha diversity analysis including OTU rank curves, rarefaction, and Shannon, along with Shannon, Chao1, Simpson, and ACE calculated indices. Principal Coordinates Analysis (PCoA) and unweighted pair group method with arithmetic mean (UPGMA) were performed using QIIME based weighted unifrac distance for beta diversity analysis (Jin et al., 2019). Finally, PICRUSt (Parks et al., 2014) was used to predict microbial function. Bacterial domains, phyla, and genera were compared using Wilcoxon rank-sum test, with the FDR adjusted P value < 0.05 being considered as signi cantly different. Finally, Spearman's rank correlations among jejunal microbiome changes, histomorphometric, and shifted metabolome were calculated to examine functional impacts of material LP diet concentrations on the small intestinal microbiome.
Metabolomics Data Analysis. UPLC-Q-TOF/MS raw data were converted to mzXML les using Proteo Wizard MSconventer tool and then processed using XCMS online software (https://bioconductor.org/packages/release/bioc/html/xcms.html). The XCMS parameters were: feature detection centwave settings (Δm/z = 25 ppm, peakwidth = c (10, 60)); retention time correction obiwarp settings (profStep = 1); and minfrac parameters = 0.5, bw = 5 and mzwid = 0.025 for chromatogram alignment. After being normalized and integrated using support vector regression, the processed data were uploaded into MetaboAnalyst 4.0 software for further evaluation (www.metaboanalyst.ca). Orthogonal partial least square discriminant analysis (OPLS-DA) and 3D-Principal Component Analysis (3D-PCA) for both positive and negative models were performed after log transformation and pareto scaling. For each variable, the variable importance projection (VIP) value in the OPLS-DA model was calculated to determine the classi cation contribution. Metabolites having VIP values > 1 were further evaluated using Student's t-test at univariate level for each metabolite with P < 0.05 considered as statistically signi cant. Changes in microbial community metabolite pro le can re ect microbial community dynamic alterations. Therefore, de ning relationships between metabolic function and microbial community structure via microbial and metabolomics data using correlation analysis may provide insight for a comprehensive understanding of microbial composition and community function.

Piglet Performance
Piglet birth BW (day 0) was greater for sows fed LP compared with piglet birth BW for sows fed CON, while 21 d piglet BW tended (P < 0.07) to be greater for piglets from sows fed LP compared with sows fed CON (Table 2). However, these initial and nal piglet BW differences did not affect piglet ADG, which was similar (P > 0.36) among both treatments.

Jejunal Morphology
Intestinal HE staining demonstrated that piglets nursing sows fed a maternal LP diet demonstrated reduced (P <0.05) villus height and ratio of villus height to crypt depth, while jejunum relative weight, villus width, crypt depth, and muscle thickness were increased (P < 0.05) compared with piglets from sows fed the maternal CON diet (Table 3).

The Diversity and Composition of Jejunal Microbiota
The 16S RNA jejunal microbiota samples after data ltering, quality control, and low-con dence singletons removal resulted in an average of 42,718 V 3 -V 4 16S rRNA gene sequence reads being obtained for the 21 d samples (two piglet litters were not yet weaned due to late farrowing). The sequence lengths ranged from 415 up to 429 bp. The rarefaction curves resulted in new OTU diminishing identi cation rates with increasing number of reads per sample. This implies that the jejunum bacterial community has adequate sampling depth for identifying dominant members. Similarly, the Good's coverages exceeded 99% demonstrating excellent sequence accuracy and reproducibility (Table 4). Of the 482 total OTU numbers, 452 OTU were detected in both groups. Based on the Shannon (P < 0.001), and Simpson (P = 0.001) indices piglets from the maternal fed LP diet demonstrated more diversity and greater evenness compared with piglets from the material fed CON diet ( Table 4). The Chaol (P = 0.519) and Ace (P = 0.435) indices were similar for piglets from the maternal fed LP compared with the maternal fed CON. Taxonomic analysis revealed the predominant phyla Firmicutes and Proteobacteria being 67.21% and 24.97%, respectively of total reads identifying 16 bacterial phyla. (Figure 1A). At the genus level, 232 genera were identi ed in the jejunal samples. The predominant genera were Lactobacillus (51.11%), Escherichia-Shigella (9.00%), Actinobacillus (7.41%), Clostridium_sensu_stricto_1 (5.60%), Romboutsia (4.35%), and Buchnera (3.54%), respectively ( Figure 1B).
Furthermore, using a PCoA plot illustrated microbial community dissimilarity and revealed distinct structures between piglets from the maternal fed LP compared with maternal fed CON ( Figure 1C). The PCoA plot uses a weighted method for unifrac similarity, which revealed PC1 and PC2 explained 55.61% and 13.98% of sample variation, respectively. Similarly, the jackknifed beta diversity and hierarchical clustering analysis via the Unweighted Pair-group Method with Arithmetic Mean (UPGMA) demonstrated that different piglets fed different maternal CP diets were clustered in their individual groups ( Figure 1D). In addition, piglets from maternal fed CON diets in the PCoA plot were clustered into two subgroups ( Figure 1C) and UPGMA hierarchical clustering analysis ( Figure 1D), which was attributed to individual variations of jejunum microbiome pro les.
The receiver operating characteristic curve (ROC) predicted different microorganisms for piglets from maternal fed LP compared to maternal fed CON piglets for inducing jejunal development. The area under the curve (AUC) judged via diagnosis test (Xia et al., 2013) that Lactobacillus is the most likely biomarker (0.9 < AUC < 1.0) for piglets from both treatments, while Clostridium_sensu_stricto_1 and Turicibacter are more likely biomarkers (0.8 < AUC < 0.9) for piglets from maternal fed LP sows.

Predicted Function of Jejunal Microbiota
The PICRUSt analyzed pathway compositions for evaluating jejunal bacterial community functional capacity is a functional-gene-count matrix. Second level KEGG (levels) metabolism pathway analysis via global and overview maps demonstrated that biosynthesis of other secondary metabolites were enriching amino acid, cofactors, and vitamins metabolism (P < 0.05), while lipid and nucleotide metabolism were decreased (P < 0.05) for piglets when maternal sows were fed LP diet compared with piglets from the maternal fed CON (Figure 2).

Correlations between Intestinal Microbial Species and Jejunum Morphological Traits
Numerous correlations via Spearman's correlation analyses (correlation coe cient | > or < 0.4, P < 0.05, Figure 3) were investigated between the different genera (n=6) relative abundances and morphological parameters (n = 7). Clostridium_sensu_stricto_1 was positively correlated with villus width, crypt depth, and muscular thickness, while being negatively correlated with villus height, and ratio of villus height: crypt depth. Escherichia-Shigella was positively correlated with muscular thickness and negatively correlated with villus height. Turicibacter was positively correlated with crypt depth and muscular thickness, while Veillonella was positively correlated with villus width. Lactobacillus was positively correlated with villus height, and villus height: crypt depth, and negatively correlated with jejunum weight, villus width, crypt depth, and muscular thickness.

Jejunum Metabolites and Metabolic Pathways
The 3D-PCA and OPLS-DA multivariate statistical analysis models were applied to evaluate the different group classi cations via Score plots (Figure 4). The 3D-PCA Score plots were derived from the LC-TOF/MS jejunal metabolic pro les demonstrated separation between the LP and CON fed diets. A clear separation and discrimination were observed between the two groups, which indicated that the OPLS-DA model could be used to identify piglet differences between maternal fed LP and CON diets. In addition, the volcano plots highlight the 44 metabolites being altered (VIP > 1.0 and P < 0.05) for piglets from the maternal fed LP fed treatment (Table 7). Thirty-four (34) metabolites were increased and 10 metabolites were decreased in piglets from the maternal fed LP diet compared with piglets from the maternal fed CON (Table 7). These amino acids, nucleotides, lipids, organic acids, and numerous metabolites are involved in multiple jejunum biochemical processes of the Bamei piglet. The hierarchical clustering analysis (HCA) with a heat map was performed to visualize the Bamei piglet jejunum metabolome differences associated with two maternal CP concentrations. The positive ionization data ( Figure 5) and negative ionization data ( Figure 6) clearly demonstrate similar clustering patterns of molecular features within each treatment. The maternal CP concentration demonstrated an impact (P < 0.05) on jejunum metabolome, while cluster differences were clearly observable in the HCA generated heatmap plot. The AUC value for each metabolite was calculated, such that metabolites with an AUC > or equal to 0.85 were selected as potential signatures. Six metabolites determine via the AUC elimination step ( Figure 7A Figure  7B) demonstrated that feeding a maternal LP diet altered (P < 0.05; rich factor > 0.10) Arg, Cys, His, Met, Phe, Try, Tyr, and linoleic acid metabolism, biosynthesis of Phe, Tyr, and Try, and Lys degradation. Consistent with the PICRUSt function prediction pathway, amino acid metabolism was enriched for piglets from sows fed a maternal LP concentration diet compared with piglets from sows fed a maternal CON.

Correlations between Differential Genera and Metabolites
The functional correlation between intestinal microbiome changes and metabolite perturbations (VIP > 2, P < 0.05) was evaluated using a correlation matrix generated by calculating the Spearman's correlation coe cient. Clear identi able correlations between perturbed intestinal microbiome and altered metabolite pro les were found (r > 0.5 or < -0.5, P < 0.05). Clostridium_sensu_stricto_1 was negatively correlated with L-His; Lactobacillus was positively correlated with chenodeoxycholate, cholic acid, glycocholic acid, L-His, L-Leu, L-Met, L-Try, L-Val, taurochenodeoxycholate, taurodeoxycholic acid, and tauroursodeoxycholic acid, but was negatively correlated with hypoxanthine, linoleic acid, palmitoyl ethanolamide, PC (16:0/16:0), and uracil; Turicibacter was negatively correlated with the D-Pro; Veillonella was positively correlated with uracil ( Figure 8). In summary, dietary maternal CP concentrations induced a piglet intestinal microbiome taxonomic perturbation, which in turn substantially alters the intestinal metabolomic pro le as observed due to changes in the diverse intestinal microbiota-related metabolites.

Discussion
In this study, reducing maternal dietary protein concentrations by 2% units resulted in similar 21 d ADG. The main nutrient digestion and absorption site is the jejunum (Low, 1976). Maternal suckled milk enters the piglet's gastrointestinal tract, thereby promoting crypt cell proliferation and proliferation. Suckling piglet jejunal development directly affects post-weaning growth performance (Buddington and Sangild, 2011). The small intestinal growth rate before and after birth of the piglet is greater than the whole body (Cheng and Leblond, 1974). The small intestine relative weight 24 h after birth is 50% greater than at birth (Adeola and King, 2006). Intestinal crypt depth increases 40% and villus height increases 35% within 3 d The combined data using Bamei piglets demonstrated that maternal dietary LP concentrations resulted in signi cant changes in intestinal microbiome composition compared with CON piglets. The altered intestinal microbial compositions were strongly associated with numerous changes of intestinal microbiota-related metabolites. These data demonstrate that a 2% reduction in dietary CP with similar SID AA concentrations not only alters intestinal bacteria abundance levels, but alters intestinal microbiome metabolic pro le, which makes the homeostasis of host metabolites rebuilt. These ndings may provide useful insights into a mechanism of feeding a maternal diet for altering the piglet's intestinal microbiome as an alternative mechanism of using dietary intervention for disease treatment.
After the piglet's birth, there are 2 sources of gut microbes with one being the maternal microbes, which are vertically passed, while the 2nd source is environmental, which are horizontally passed. In agreement with previous pig studies (Fan et  Lactobacillus are bene cial bacterial members of the small intestinal microbiota that were reduced for piglets from sows fed the LP diet. The intestinal bacterial environment can protect the intestine from toxic dietary ingredients (Di Rienzi et al., 2018). The reduction of Lactobacillus spp. abundance may result from decreased oligosaccharide ingestion (less soybean meal inclusion), which reduces nutrient availability, which relates to reduced piglet weight (Drissi et al., 2017). These results indicate that maternal dietary LP concentration alters Bamei piglets intestinal microbiota through altering the bene cial bacterial colony structure (Bian et al., 2016). Therefore, it is reasonable to hypothesize that intestinal microbiota differences are the result of early dietary intervention, host-microbe interactions, and/or host physiological state. The most important host-microbe interaction may occur on or at the intestinal barrier.
The metabolomics data revealed that maternal dietary CP concentration alters jejunal metabolite concentration indicating that jejunal metabolism may be linked with jejunal microbiota activity. Maternal LP concentration altered numerous metabolite concentrations associated with piglet protein digestion and absorption, as well as, amino acid biosynthesis. Amino acids are key precursors for protein and polypeptide synthesis and regulators of some metabolic pathways mainly derived from jejunal microbiota and host enzymatic degradation of dietary proteins and microproteins. The piglet L-Met concentration from sows fed the LP diet was higher compared with piglets from sows fed the CON diet. L-Met is a precursor to other sulfur-containing AA, including homocysteine, which are important for protein synthesis, polyamine formation, and synthesis of many metabolites.

Conclusions
These data clearly demonstrates that maternal diet CP concentration alters the piglet's intestinal microbiome composition for altered metabolite concentrations used in various metabolic pathways when diet CP was reduced 2% units with similar EAA concentrations. Reducing maternal dietary CP demonstrated altered piglet histomorphology, microbiota composition and function, while modulating jejunum microbiota metabolic pro les that are associated with speci c intestinal bacteria genera. These alterations aid in understanding the bene cial impacts of feeding maternal LP diets without affecting SID EAA concentrations on piglet intestinal health. Furthermore, the regulated intestinal microbiome-related metabolites may be potential biomarkers to be used in the future to explore functional impacts of maternal dietary interventions on the piglet's intestinal microbiome.    Table 3 Jejunum weight and tissue morphology by 21-day old suckling piglets when feeding maternal diets containing 12% (LP) or 14% crude protein (CON).  Table 4 Alpha diversity measures of bacterial communities by 21-day old suckling piglets when feeding maternal diets containing 12% (LP) or 14% crude protein (CON).  Table 5 Phylum-level taxonomic composition of the jejunal bacterial communities by 21-day old suckling piglets when feeding maternal diets containing 12% (LP) or 14% crude protein (CON).  Table 6 Genus-level taxonomic composition of the jejunal bacterial communities by 21-day old suckling piglets when feeding maternal diets containing 12% (LP) or 14% crude protein (CON).  PCoA plot and UPGMA tree using the weighted unifrac similarity method.

Figure 2
Predicted microbial functions using PICRUSt by 21-day old suckling piglets when feeding maternal diets containing 12% (LP) or 14% crude protein (CON) when bacteria differed.

Figure 3
Correlations between differential genera and morphological traits at the jejunum by 21-day old suckling Bamei piglets when feeding maternal diets containing 12% (LP; N=12) or 14% crude protein (CON; N=12). Each row in the graph represents a genus, each column represents a morphological trait, and each lattice represents a Spearman correlation coefficient between a genus and a morphological trait. Red represents a positive correlation, while blue represents a negative correlation. *Significant correlation between the LP and CON groups (P < 0.05).
Page 30/34  Hierarchical clustering analysis for identification of different metabolites by 21-day old suckling Bamei piglets when feeding maternal diets containing 12% (LP) or 14% crude protein (CON) following positive mode ionization. Each column in the figure represents a sample, each row represents a metabolite, and the color indicates the relative amounts of metabolites expressed in the group; Red indicates that the metabolite is expressed at high levels, and blue indicates lower expression. Hierarchical clustering analysis for identification of different metabolites by 21-day old suckling Bamei piglets when feeding maternal diets containing 12% (LP; red) or 14% crude protein (CON; green) following negative mode ionization. Each column in the figure represents a sample, each row represents a metabolite, and the color indicates the relative amounts of metabolites expressed in the group; Red indicates that the metabolite is expressed at high levels, and blue indicates lower expression.

Figure 7
Metabolic pathway enrichment analysis following positive and negative mode ionization to provide metabolite overview by 21-day old suckling Bamei piglets when feeding maternal diets containing 12% (LP) or 14% crude protein (CON).

Figure 8
Correlation analysis between genera and metabolite concentrations (VIP > 2) by 21-day old suckling Bamei piglets when feeding maternal diets containing 12% (LP) or 14% crude protein (CON). Each row in the graph represents a genus, each column represents a metabolite, and each lattice represents a Pearson correlation coefficient between a component and a metabolite. Red represents a positive correlation, while blue represents a negative correlation. *Significant correlation between the LP and Con groups (P < 0.05).