A total of 1,000 piglets (Yorkshire × Landrace) were weaned at the age of 21 days with an average body weight (BW) of 6.09 ± 0.26 kg; they were assigned to 1 of 5 treatments with 20 replicate pens (50 piglets per replicate pen) for 42 days. Water and feed were provided ad libitum. The body weight and feed intake of pigs were recorded weekly.

The compositions of the basal diets are shown in Supplementary Table 1. An experiment diet (Table 1) was formulated to provide varying dietary calcium concentrations from calcium carbonate, calcium citrate, or multiple calcium. The formula of the diet should meet or exceed the nutritional needs of weaned piglets (26). B = A (control) plus 5/1,000 calcium citrate to replace calcium carbonate; C = B plus 1/1,000 organic trace minerals; D = C minus half of the acidifier; E = A (control) plus 5/1,000 multiple calcium to replace calcium carbonate. All piglets were fed five different diets for 6 weeks before being euthanized and samples collected. Piglets were housed on a 12-h light-dark cycle with free access to water, and the barn temperature was maintained at 30°C.

All piglets and feeds were weighed on days 0, 14, 28, and 42 to calculate the average daily feed intake (ADFI), average daily gain (ADG), and feed to meat ratio (F:G). On day 42, two blood samples were collected using heparin tubes from the front cavity veins of eight weaned piglets in five groups separately. Collected plasma samples were centrifuged at 1,000 × g for 15 min at 4°C and stored at −20°C for further analysis. Four weaned piglets were sacrificed in each group, and intestinal samples were subsequently collected.

Two piglets were selected in each repeat of five groups (40 piglets in all), and a total of 40 plasma samples were used for analysis. The plasma biochemical components include malondialdehyde (MDA), glutathione peroxidase (GSH-Px), superoxide dismutase (SOD), immunoglobulin-A (IgA), immunoglobulin-G (IgG), immunoglobulin-M (IgM), alanine aminotransferase (ALT), aspartate aminotransferase (AST), amylase (AMY), alkaline phosphatase (ALP), and lactate dehydrogenase (LDH) were measured using commercially available kits (Jiangsu Meimian Industrial Co., Ltd, Yancheng, China) following the manufacturer's instructions.

One piglet was selected randomly to execute (20 piglets in all) in each repeat of five groups and a total of 20 ileal samples and 20 samples of cecum for microbiota analysis. Microbial DNA was extracted from approximately 0.25 g of each intestinal chyme using a QIAamp DNA Stool Mini Kit (Qiagen, Germany), according to the manufacturer's instructions (27). Successful DNA isolation was performed by 2% of agarose gel electrophoresis. The bacterial universal V3–V4 region of the 16S rRNA gene was amplified according to PCR-barcoded primers 515F (5′-ACTCCTACGGGAGGCAGCAG-3′) and the reverse primer 806R (5′-GGACTACHVGGGTWTCTAAT-3′). The specific sequencing method was done as previously reported (28). The thermal cycle procedure is as follows: initial denaturation step, 95°C, 3 min; denaturation, 27 cycles, 95°C, 30 s; annealing, 55°C, 30 s; elongation, 72°C, 45 s; and final extension, 72°C, 10 min. Briefly, paired-end was sequenced on an Illumina MiSeq platform (PE300) platform (Illumina, USA) at the Majorbio Bio-Pharm Technology (Shanghai, China). The 16S rRNA amplicon sequences have been deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) (http://www.ncbi.nlm.nih.gov/bioproject/815951) under accession number PRJNA815951.

Quality filters were applied to trim the original sequences according to the criteria: (I) reads with an average quality score <20 over a 10-bp sliding window were re-moved, and truncated reads smaller than 150 bp were discarded. (II) Truncated reads containing homopolymers longer than eight nucleotides in length, more than 0 bases in barcode matches, or more than two different bases in primers were removed from the dataset. Checking and removing possible chimeras by USEARCH using the chimera layer “gold” database described by Edgar et al. (29). Clustering of OTUs with a similarity cutoff of 97% using USEARCH (30) and abundance-defining representative sequences for each OTU were identified using PyNAST (31) and the SILVA bacterial database (32). The rarefaction analysis was performed using Mothur version 1.39.5 (33) to reveal diversity indices, including Chao index and Shannon index. PCoA was performed using Canoco 4.5. Venn diagrams were implemented by Venn Diagram, and community diagrams were generated using R tools from the data in the files “tax. phylum.xls, tax.family.xls, and tax.genus.xls.”

PICRUSt has been used as a bioinformatics tool to predict the functional potential of metagenomes using 16S rRNA genetic data (34). Subsequently, by referencing the Kyoto Encyclopedia of Genes and Genomes (KEGG) database, the OTU table was imported into PICRUSt for functional gene prediction. PICRUSt utilizes 16S copy number prediction to normalize the OTU table so that OTU abundance more accurately reflects the true abundance of the underlying organism. We then looked for the precomputed genome content of each OTU, multiplied the normalized OTU abundance by each KEGG abundance in the genome, and summed these KEGG abundances for each sample to predict the metagenome. This prediction calculates the KEGG abundance for each metagenomic sample in the OTU table. For those optional organism-specific predictions, each organism's abundance per KEGG is kept and annotated. We focused our exploration of metagenomes on levels 2 and 3. These pathways related to organismal systems, human disease, and drug development are filtered out because they do not reflect microbial function.

Statistical analyses between the means of each group were analyzed using one-way ANOVA (one-way analysis of variance) followed by multiple comparisons using a post hoc test of S-N-K through SPSS 22.0. Statistical significance was set at p < 0.05.

Initial BW of weaned piglets did not differ among the 5 groups (Table 3). There were no significant effects of including different sources of calcium in five diets with normal or halved acidifier-fed piglets on BW in 6 weeks (p > 0.05). But the diet of multiple calcium dramatically decreased the ADFI compared to the C and D diet on days 15–28 (p < 0.05). Meanwhile, no significant differences in ADFI were noted between groups on days 1–14 and 29–42 (p > 0.10). As shown in Table 4, ADG increased significantly when multiple calcium instead of calcium citrate was added on days 1–14 (0.18 ± 0.03 vs. 0.22 ± 0.04). And F:G was not affected (p > 0.10) by the different sources of calcium in the diets at different time points for weaned piglets.

Variation of the antioxidant index in the plasma is shown in Table 5. It was obvious that whatever piglets were fed, the calcium carbonate, calcium citrate, multiple calcium, or different amount of acidifier, none of them changed the content of MDA, SOD, and GSH-Px (p > 0.10). A similar situation occurs in the immune function of blood, and it has no significant difference in the immune index (IgA, IgM, and IgG) among different treatments after weaning 6 weeks for piglets (p > 0.10).

Data of enzymes in the plasma of weaned piglets by feeding different calcium sources are displayed in Table 6. Comparing to the normal diet, the ALT concentration decreased significantly by feeding calcium citrate (B) in the diet for weaned piglets (165 ± 13 vs. 141 ± 23), but by adding additional organic trace minerals (C), the ALT concentration returned to the normal level (165 ± 13 vs. 164 ± 21). However, reducing the amount of acidifier in the diet by half (D) or replacing it with multiple calcium (E) does not change AST concentration (p > 0.05). Meanwhile, there was no statistically significant change in AST (p = 0.935), AMY (p = 0.242), and LDH (p = 0.524) concentrations in plasma among the five-diet fed piglets. Regarding ALP concentration in the plasma, no significant difference was observed when calcium citrate was in the diet instead of calcium carbonate (p > 0.05). In comparing diets based on calcium citrate, after half of the acidifier in the experimental diet for piglets, the concentration of ALP decreased dramatically (151 ± 24 vs. 125 ± 20) in the blood of weaned piglets. But when multiple calcium replaces the same amount of calcium citrate, there is no statistically significant change in the concentration of ALP (p > 0.05).

As shown in Figure 1A, the overall OTU numbers classified on the distance level of 0.03 were 652 detected in the ileal samples, most abundance 419 OTUs in the control group observed and the E group has the least 151 OTUs, and 49 were shared among five groups. The cecal OTUs numbers were more than ileum and owned 1,461 OTUs (Figure 1C). On the contrary, the cecum control group had fewer OTUs than the ileum control group. And 764 OTUs were shared in the cecum by five diet treatments. Besides, the PCoA showed that 20 samples from the ileum in five groups were not separated completely (Figure 1B). In the cecal samples, we could see the E group stay away from the other four groups, and the other four sets of samples were mixed together (Figure 1D).

The bacterial composition of the control group had a higher Chao1 estimator and Shannon diversity index than the other four groups, and the E group showed the lowest in the ileum (Figures 2A,B). However, the difference between the two indicators in the five groups was not statistically significant (p > 0.05). In contrast, the control group showed the lowest Chao1 estimator and Shannon diversity index in the cecum than the other four diet treatment groups (Figures 2C,D), and there was also no dramatic difference in the two indicators between the five groups (p > 0.05).

As shown in Figure 3A, Proteobacteria, Firmicutes, Verrucomicrobia, Campilobacterota, and Actinobacteriota were dominant phyla in the ileum of weaned piglets, accounting for more than 90% of the total number of ileal bacteria. As shown in Figure 4A, only the abundance of Actinobacteriota at the phyla level was significantly greater in the ileum of the A group compared to the other groups (p < 0.05). Compared with in the ileum, the most abundant bacterial community at the phylum level was Firmicutes, followed by, from most to least, Bacteroidota, Actinobacteriota, Proteobacteria, and Spirochaetota (Figure 3B). Although the phylum proportions of some bacterial communities vary according to the different sources of calcium added to the diet, there was no statistical difference between the main bacterial communities at the phyla level in the cecum (p > 0.05).

Downward to genus levels, Escherichia-Shigella, Streptococcus, Lactobacillus, Klebsiella, and Actinobacillus were the dominant genus in the ileum in the five groups (Supplementary Figure 1). In Figure 4B, the relative abundance of Rothia was significantly higher in the ileum of the A group compared to the other groups (p < 0.05). Analogously, the Globicatella (Figure 4C) and Subdoligranulum (Figure 4D) were significantly greater in the ileum of the A group compared to the D and E groups, respectively (p < 0.05). In the cecum, the top five genera in bacterial abundance were Streptococcus, Lactobacillus, Clostridium_sensu_stricto_1, Subdoligranulum, and UCG-005 (Supplementary Figure 2). In Figure 4E, the relative abundance of Streptococcus was significantly greater in the cecum of the E group in weaned piglets by multiple calcium diet than in the A, B, and C groups (p < 0.05). On the contrary, the abundance of Subdoligranulum (Figure 4F) decreased significantly when weaned piglets were fed the D and E diets than A, B, and C groups (p < 0.05). Besides, the abundance of unclassified_f_Lachnospiraceae (Figure 4G) was lower in the cecum of the A group than in the other groups (p < 0.05). And the abundance of Dorea (Figure 4H) decreased significantly when weaned piglets were fed the D and E diets than A, B, and C groups (p < 0.05).

In the study, PICRUSt was used to analyze the microbiota function of the ileum and cecum. The 16S rRNA sequencing results combined with genomic databases could be used to predict macrogenomic information (34). The predictable outcomes can be enriched at 2 and 3 levels of the KEGG pathways in the ileum (Supplementary Figure 3). In the top 10 KEGG pathways shown in Supplementary Figure 3a, membrane transport and signal transduction pathways were associated with environmental information processing. Six other pathways, including the metabolism of carbohydrates, global and overview maps, amino acids, energy, nucleotides, cofactors, and vitamins were associated with nutrients metabolism. The prokaryotic cellular community was associated with cellular processes. Totally, 320 pathways were predicted, and the top 10 pathways consisted of two pathways related to carbohydrate metabolism (as shown in Supplementary Figure 3b), including starch and sucrose metabolism, amino sugar, and nucleotide sugar metabolism. The highest abundance was found in ABC transporters, and it belonged to membrane transport. In addition, both purine metabolism and pyrimidine metabolism belong to nucleotide metabolism. And 327 pathways were totally predicted in the cecum. As shown in Supplementary Figure 4a, level 2 of the KEGG pathway in the cecum was only replication and repair different in the ileum, and it was related to genetic information processing. The most abundant KEGG pathway in the cecum was the biosynthesis of amino acids (Supplementary Figure 4b). Ribosome (ko03010) and quorum sensing (ko02024) belonged to translation and cellular community-prokaryotes, respectively. Finally, there was no significant difference in the 2 and 3 levels of the top 10 KEGG pathways in the ileum and cecum (p > 0.05).