This study was conducted at the pig breeding farm in Shandong Province. The animal use protocol for this research was approved by the Animal Care and Use Committee of Shandong Agricultural University (Approval Number: SDAUA-2019-019).

Twenty-six Large White × Landrace crossbred multiparous sows (2nd parity) with similar back fat thickness (BF, 15.28 ± 0.45 mm) and body weight (174.34 ± 2.72 kg) were used in this study. The BF at the last rib was measured using a HG 9300 digital diagnostic ultrasound device (Caresono Technology Co. Ltd., Nanjing, China). After artificial insemination, the individual sow was housed individually in a gestation stall (2.37 × 0.65 × 1.13 m) kept at 21 ± 1°C. All the sows were mated within 3 days and fed a common fortified corn–soybean meal gestation diet (Supplementary Table 1) which was formulated to meet or exceed National Research Council (18) nutrient requirements. All sows received a daily meal at 0900 h and were fed the same amount of feed (days 1 to 89 of gestation 2.46 kg/d; days 90 of gestation to farrowing, 2.89 kg/d) during the entire gestation. On day 110 of gestation, sows were moved from gestation to farrowing rooms and kept in individual farrowing crates measuring 2.40 × 1.80 × 0.90 m thereafter. Backfat thickness and body weight of individual sow were measured at breeding and within 24 h of farrowing. At farrowing, the numbers of total born piglets, live born piglets, and dead born piglets per litter, as well as litter weight, were recorded, and the averages were calculated. Thus, two groups were generated (Table 1): 13 sows with litter size lower than the average in this trial (12.7 piglets) were classified as the low-reproductive performance group (LP group), while 13 sows with litter size higher than the average in this trial (12.7 piglets) labeled as the high-reproductive performance group (HP group). Sows had free access to water throughout the experiment and did not receive vaccine, antibiotics, or other medication in the feed or for any therapeutic purposes after insemination.

Fasting blood samples (12 h overnight) and fresh fecal samples from all healthy sows were collected on day 30 and day 110 of gestation before feeding in the morning. Samples were grouped as follows: LP30 and LP110: sows with low-reproductive performance on day 30 and day 110 of gestation, respectively; HP30 and HP110: sows with high-reproductive performance on day 30 and day 110 of gestation, respectively. Blood samples (5 mL) from the ear veins were collected into a tube containing heparin sodium and centrifuged at 3,000× g for 15 min. Plasma samples was transferred to 200 μL centrifuge tubes and stored at −20°C until analysis. Fecal samples (about 5 g) were collected from the rectum by a sterilized fecal collection tube and then stored at −80°C immediately for the detection of short-chain fatty acids (SCFAs) and analysis of microbiota.

Plasma biochemical parameters, including glucose (GLU), cholesterol (CHOL), triglyceride (TG), high density lipoprotein cholesterol (HDL-C), low density lipoprotein cholesterol (LDL-C), and non-esterified fatty acid (NEFA), were determined with commercial kits (Nanjing Jiancheng Bioengineering Institute, Nanjing, China) using standard spectrophotometric methods on an Autolab-PM4000 Automatic Analyzer (AMS Co., Rome, Italy) as previously described (19).

Concentrations of interleukin-2 (IL-2), interleukin-6 (IL-6), interleukin-10 (IL-10), tumor necrosis factor-α (TNF-α), immunoglobulin A (IgA), immunoglobulin G (IgG), immunoglobulin M (IgM), progesterone, estrogen, lutropin, and prolactin in the plasma of sows were determined with commercial ELISA kits (Feiya Biotechnology Co. Ltd., Yancheng, China) as described in Supplementary Methods.

The fecal SCFAs of sows were measured by a Varian CP-3800 gas chromatograph (Palo Alto, CA, USA) equipped with a micro-injector, a flame ionization detector, and a capillary chromatographic column as described in Supplementary Methods.

Microbial composition and diversity were analyzed as previously described in Li et al. (20). Briefly, bacterial genomic DNA was extracted from frozen fecal samples with an E.Z.N.A.^TM Stool DNA kit (Omega Bio-Tek, Norcross, GA, USA) according to the manufacturer's protocol. After DNA concentration and purity monitoring, DNA was diluted to 1 ng/μL using sterile water, and the V4 hypervariable region of 16S rDNA was amplified with 515F and 806R primer (5′-GTGCCAGCMGCCGCGGTAA-3′ and 5′-GGACTACHVGGGTWTCTAAT-3′, respectively), on the Illumina HiSeq PE2500 platform by Novogene (Beijing, China). Filtered, non-chimeric high-quality sequences (tags) sharing over 97% sequence similarity were clustered into the same operational taxonomic units (OTUs) by Uparse software (21), and then classified to different taxonomic levels with SILVA database (22) based on Mothur algorithm to annotate taxonomic information. Operational taxonomic units abundance information were normalized using a standard of sequence number corresponding to the sample with the least sequences for subsequent analysis of alpha diversity and beta diversity. Shannon, Simpson, Chao 1, and ACE indexes were chosen to ascertain differences in alpha diversity based on different groups (23, 24), and Bray-Curtis distances were calculated and visualized using Principal Coordinate Analysis (PCoA) (25). The statistical differences in alpha and beta diversity of bacterial communities between the two groups were examined using the Wilcoxon rank-sum test. Significant difference among the microbial communities was accessed with the analysis of similarity (ANOSIM) test.

The individual sow was regarded as the experimental unit for all variables. Differences in the data including plasma biochemical parameters, inflammatory factors, and fecal SCFAs were evaluated using the independent t-test (LP vs. HP) or paired t-test (G30 vs. G110) procedure of SAS 9.0 (Institute Inc., Cary, NC, USA) following normal distribution assessment using a Shapiro-Wilk's statistic (W > 0.05). Multiple testing was corrected by using the Benjamini-Hochberg false discovery rate. Spearman's correlations were used to assess the associations between bacterial abundance and litter size, as well as plasma biochemical indices. Treatment differences were considered statistically significant at P < 0.05, and 0.05 ≤ P < 0.10 was considered a statistical trend. Values are expressed as mean ± standard error in tables and figures.

A total of 4,392,562 total tags, 4,118,486 taxon tags, and 274,447 unique tags were obtained from 52 sow fecal samples, with an average of 84,472 ± 730, 79,202 ± 720 and 5,278 ± 179 per sample, respectively (Figures 1A–C). Based on 97% sequence similarity, a total of 21,114 OTUs were found in the HP group on day 30 of gestation, with an significantly higher average of 1,624 ± 10 OUT per sample compared to an average of 1,589 ± 12 OUT per sample in the LP group, where 20,657 OTUs were found in total (Figure 1D); the HP group tended had lower unique tags than the LP group on day 110 of gestation (5,500 ± 248 vs. 6,397 ± 389; P = 0.064; Figure 1C). In addition, from gestation d 30 to d 110, the unique tags number was significantly increased on average (P < 0.05), but OTUs number was decreased (P = 0.047).

To determine whether the sample size was sufficient for OUT testing, the species accumulation curves (SAC) was used in the present study. The SAC (Supplementary Figure 1) tended to flatten as the sample number of analyzed sequences increased up to 52, suggesting that the sample size was enough for OTUs testing and could estimate the species richness of the habitat.

The results of fecal microbial community structures assessment are shown in Figure 2. On d 30 of gestation, the HP group had significantly higher Shannon index (P < 0.05) and tended to have a higher Chao 1 index compared with LP group (P = 0.069); on d 110 of gestation, no significant differences were observed in alpha diversity between the two groups (P > 0.05).

In addition, to measure the evolutionary distance between microbiotas (beta diversity), the PCoA profile for sow fecal samples based on the Bray-Curtis distance was used in the present study, and ANOSIM test was used to assess significant differences among the microbial communities (Figures 3A,B). The results suggested that LP and HP groups had close distance on gestation d 30 which showed that the two groups had no significant difference in the microbial community (P = 0.189, Figure 3C); an obvious separation was observed in PCoA between samples from HP group and LP group on d 110 of gestation. The ANOSIM test also indicated that the two groups had notably different microbiota structures on gestation d 110 (P = 0.006, Figure 3D), and sows in HP group had greater beta diversity compared to sows in LP group at 110 days of gestation (P < 0.05). Besides, the Bray-Curtis distance analysis showed a global shift in microbial community composition from gestation d 30 to d 110 (P = 0.001, Figures 3B,E).

As shown in Figure 4, Firmicutes and Bacteroidetes were the most predominant phyla which accounted for more than 75%, followed by Spirochaetes and Tenericutes, in both groups during gestation (Figure 4A). No significant differences were observed in the top 10 phyla which accounted for more than 99.5% of the total bacteria population between the LP and HP groups on d 30 of gestation (P > 0.05, Figure 5A). However, the relative abundance of Firmicutes in the HP groups was significantly higher (P < 0.05) than that in the LP group, while the relative abundances of Bacteroidetes, Spirochaetes, and Fibrobacteres were significantly lower (P < 0.05) than that in the LP group on d 110 of gestation. In addition, the relative abundances of Firmicutes and Actinobacteria were significantly decreased (P < 0.05), while Bacteroidetes and Verrucomicrobia were significantly increased from gestation d 30 to d 110 (P < 0.05).

The species phylogenetic tree evolution was constructed by multiple sequences alignments to obtain the representative sequence of the top 35 genera. As shown in Figure 4B, the relative abundance of Firmicutes was contributed by Clostridium_sensu_stricto_1, Streptococcus, Lactobacillus, Christensenellaceae_R-7_group, Ruminococcaceae_UCG-002, Ruminococcaceae_NK4A214_group, Ruminococcaceae_UCG-005, Ruminococcaceae_UCG-014, Ruminococcus_1, and Lachnospiraceae_XPB1014_group; Bacteroidetes mainly distributed with Rikenellaceae_RC9_gut_group, Prevotellaceae_UCG-001, and Prevotellaceae_NK3B31_group; Spirochaetes was dominated by Treponema_2. In the LP and HP groups, Treponema_2 and Clostridium_sensu_stricto_1 were the top two genera on d 30 of gestation, and Treponema_2 and Streptococcus were the most dominant on d 110 of gestation. Of the top 35 genera, compared with sows in the LP group, sows in the HP group had significantly higher (P < 0.05) relative abundances of Eubacterium_coprostanoligenes_group, Lachnospiraceae_XPB1014_group, Sphaerochaeta, Ruminococcaceae_UCG-010, Roseburia, Ruminococcaceae_UCG-002, and Family_XIII_AD3011_group, and significantly lower (P < 0.05) Succinvibrio on d 30 of gestation; sows in the HP group had significantly lower (P < 0.05) Treponema_2, Prevotellaceae_UCG-001, Prevotella_1 and dgA-11_gut_group, as well as significantly higher (P < 0.05) Lactobacillus, Christensenellaceae_R-7_group, Terrisporobacter, and Escherichia-Shigella on d 110 of gestation (Figure 5B). Besides, sow fecal samples from gestation d 110 had significantly higher abundances of Streptococcus, Prevotellaceae_NNK3B31_group, Oscillospira, Eubacterium_coprostanoligenes_group, and Ruminococcaceae_UCG-010, but significantly lower Prevotellaceae_UCG-001, Terrisporobacter, Escherichia-Shigella, Ruminococcaceae_NK4A214_group, Christensenellaceae_R-7_group, and Romboutsia than those of sow fecal samples from gestation d 30 (P < 0.05).

On day 30 of gestation, at the phylum level, the relative abundances of Firmicutes and Actinobacteria tended to be positively correlated with litter size (P < 0.10), while the relative abundance of Proteobacteria tended to be negatively correlated with litter size (P < 0.10); at the genus level, the relative abundances of Turicibacter and Ruminococcaceae_UCG-014 had significant positive correlations with litter size (P < 0.05), and the relative abundances of Clostridium_sensu_stricto_1 and Romboutsia tended to be positively correlated with litter size (P < 0.10, Table 2).

On d 110 of gestation, at the phylum level, significant positive correlation between the relative abundance of Firmicutes and litter size was observed (P < 0.05), and the relative abundances of Bacteroidetes, Spirochaetes, and Actinobacteria were all significantly negatively correlated with litter size (P < 0.05); at the genus level, Clostridium_sensu_stricto_1, Turicibacter, Terrisporobacter, Christensenellaceae_R-7_group, and Escherchia-Shigella exhibited the significantly positive correlations with litter size (P < 0.05), while Rikenellaceae_RC9_gut_group, Treponema_2, and Sphaerochaeta had significant negative correlations with litter size (P < 0.05). In addition, the phylum Proteobacteria and genus Lactobacillus displayed a tendency to be positively correlated with litter size (P < 0.10), and the genus Sphaerochaeta tended to be negatively correlated with litter size (P < 0.10).

The concentrations of fecal short-chain fatty acids on d 30 and d 110 of gestation in the two groups are listed in Figure 6. On d 30 of gestation, there were no significant differences in fecal acetate, propionate, and total SCFAs concentrations between the LP group and HP group (P > 0.05), but sows from HP group showed significantly higher butyrate concentration than those of sows from LP group (P < 0.05). On d 110 of gestation, sows in the HP group had significantly lower acetate and total SCFAs concentrations than sows in the LP group (P < 0.05), and the propionate concentration in the HP group tended to be lower than that in the LP group (P = 0.053). The fecal SCFAs concentrations of sows on d 110 of gestation did not differ with that of sows on d 30 of gestation (P > 0.05).

As shown in Figure 7, on day 30 of gestation, significantly lower plasma TG levels were observed in the HP group compared with those in the LP group (P < 0.05); sows in the HP group tended to have a lower plasma GLU concentration than sows in the LP group (P = 0.070). On d 110 of gestation, sows in the HP group had significantly higher plasma TG and NEFA levels (P < 0.05) and tended to have lower plasma HDL-C concentration compared with those of sows in the LP group (P = 0.055). Besides, the concentrations of CHOL (P = 0.018), HDL-C (P = 0.085), and LDL-C (P = 0.061) were decreased, and the levels of TG (P = 0.020) were increased from gestation d 30 to d 110.

The levels of plasma inflammatory factors and immunoglobulins are shown in Figure 8. There were no significant differences in the plasma concentrations of inflammatory factors and immunoglobulins on d 30 of gestation between the LP and HP groups (P > 0.05). However, significantly higher TNF-α and IgM concentrations were observed in the HP group compared with those in the LP group on d 110 of gestation (P < 0.05). In addition, the plasma IL-6 and IL-10 levels were significantly increased (P < 0.05) from gestation d 30 to d 110.

The plasma hormone contents are shown in Figure 9. There were no significant differences in the plasma hormone concentrations on d 30 and d 110 of gestation between the LP and HP groups (P > 0.05). The plasma hormone contents of sows on gestation d 110 did not differ from those of sows on gestation d 30 (P > 0.05).

At the phylum level (Figure 10A), the plasma levels of CHOL, HDL-C, and LDL-C showed positive correlations with the abundances of Firmicutes and Actinobacteria (P < 0.05) and negative correlations with the abundances of Bacteroidetes and Fibrobacteres (P < 0.05); the plasma TG level had significant positive correlation with Euryarchaeota abundance (P < 0.05); the plasma IL-2 concentration was significantly positively correlated with Proteobacteria abundance (P < 0.05); the plasma IL-6 concentration was significantly positively correlated with the abundance of Bacteroidetes (P < 0.05); the plasma IgA concentration had significant negative correlation with Verrucomicrobia (P < 0.05). In addition, the IL-10 concentration tended to be negatively correlated with the abundance of Tenericutes (P < 0.10), and the IgM concentration tended to be negatively correlated with the abundance of Euryarchaeota (P < 0.10).

At the genus level (Figure 10B), the plasma level of CHOL showed positive correlations with the abundances of Clostridium_sensu_stricto_1, Christensenellaceae_R-7_group, and Terrisporobacter (P < 0.05) and negative correlations with the abundances of Eubacterium_coprostanoligenes_group, dgA-11_gut_group, Ruminococcaceae_UCG-010, and Sphaerochaeta (P < 0.05). The plasma HDL-C concentration was significantly positively correlated with the abundance of Christensenellaceae_R-7_group (P < 0.05) and was significantly negatively correlated with the abundances of Eubacterium_coprostanoligenes_group, dgA-11_gut_group, and Sphaerochaeta (P < 0.05). The plasma LDL-C level had significant positive correlations with the abundances of Clostridium_sensu_stricto_1 and Christensenellaceae_R-7_group (P < 0.05) and had significant negative correlations with the abundances of dgA-11_gut_group, Ruminococcaceae_UCG-010, and Sphaerochaeta (P < 0.05). The plasma NEFA concentration displayed positive correlation with the abundance of Eubacterium_coprostanoligenes_group and negative correlations with the abundances of Turicibacter, Terrisporobacter, and Romboutsia (P < 0.05). The plasma IL-2 concentration was significantly positively correlated with the abundances of Clostridium_sensu_stricto_1 and Succinivibrio, and significantly negatively correlated with the abundance of Ruminococcaceae_UCG-010 (P < 0.05). The plasma IL-6 level showed positive correlation with the Prevotellaceae_NNK3B31_group abundance (P < 0.05). The plasma IL-10 concentration had significant positive correlations with the abundances of Prevotellaceae_NNK3B31_group and Sphaerochaeta (P < 0.05). The plasma TNF-α concentration was significantly positively correlated with the abundances of Anaerotruncus (P < 0.05). The plasma level of IgA showed positive correlation with the Lachnospiraceae_XPB1014_group abundance (P < 0.05). The plasma level of IgG indicated negative correlation with the Christensenellaceae_R-7_group abundance (P < 0.05). The plasma IgM concentration was significantly positively correlated with the abundances of Ruminococcaceae_UCG-014 and Lachnospiraceae_AC2044_group (P < 0.05), and was significantly negatively correlated with the abundances of Methanobrevibacter and dgA-11_gut_group (P < 0.05).