60 perinatal Holstein cows were selected and divided into two groups according to the expected delivery period (2 ± 0.69). 30 heads per group. The trial period began from 21 days before the expected date of delivery of each cow to 21 days after delivery, for a total of about 42 days. Perinatal cows were fed total mixed ration (TMR) at 7:00 and 19:00 daily during the trial. At 7:30 and 19:30, the TMR was fed to the cows in the perinatal period and milked. They were free to eat and drink freely 24 h a day. Saccharomyces cerevisiae and its ferment were added to the TMR, and the two test groups were fed 0, 100 g per day. The cows rushed into the perinatal field 21 days before the expected date of delivery, entered the delivery room 7 days before the expected date of delivery, rushed into the care shed 1 day after delivery, and entered the new born cowshed 7 days after delivery.

Table 1 shows the formulation and composition of the diet.

Note: NEL = GE × [94.280 − 61.5730 (NDF/OM)] × 0.5501 − 0.395.

The amount of TMR and the amount of remaining material fed per day were recorded. The TMR sample was collected weekly and placed in an electric constant temperature blast dryer at 65˚C to a constant weight, and pulverized after 24 hours of resurgence. The cows were taken from the rectum before the morning feeding on the 19th, 20th, and 21st days after the cows were placed in the refrigerator, and stored in a refrigerator. The milk production was recorded for three days, and the milk composition was measured after mixing the milk. In the morning of 21 days after the postpartum, the cows took 10 ml of blood from the tail vein, and centrifuged to take the serum for cryopreservation. 100 ml of rumen fluid was taken through the esophagus. After filtering through 4 layers of gauze, some of them were cryopreserved by metaphosphoric acid and the other part was stored in liquid nitrogen. Transfer to a −80˚C refrigerator for long-term storage.

The dry matter determination of TMR and fecal samples was carried out according to the method of GB6435-86; the crude protein was determined according to the Kjeldahl method (GB/T 6432-94); the determination of coarse ash was carried out according to the method of GB/T 6438-86; The crude fat content is carried out according to the method of GB/T 6433-2006; the neutral detergent fiber and the acid detergent fiber are carried out according to the method proposed by Van Soest (1991); the acid-insoluble ash is carried out according to the method of GB/T 23742-2009; Calorimeter (6200, PARR) assay; blood hormones and bioactive substances were determined using the Elisa kit.

Data basic processing was performed using Excel 2007 software, and the results were analyzed by one-way ANOVA using SAS 8.2.

Table 2 shows the apparent digestibility of various nutrients in peripartum dairy cows. By detecting the apparent digestibility of nutrients, we can see that feeding Saccharomyces cerevisiae and its fermentation products during the perinatal period had no significant effect on the apparent digestibility of nutrients in dairy cows, and only increased the digestibility of crude fat (P = 0.08).

As shown in Table 3, the probiotic supplementation during the perinatal period

significantly increased the DMI (P < 0.01) and the postpartum lactose content (P < 0.01), with a trend of increased milk production (P = 0.052), but had no significant effect on other milk components.

As can be seen from Table 4, the β-hydroxybutyric acid content of the test group was significantly increased (P = 0.01). At the same time, the probiotic preparation increased the content of serum glucagon extremely significantly (P < 0.01), significantly reduced insulin content (P = 0.0225), but had no significant effect on other hormones.

As can be seen from Table 5, the contents of butyric acid and isovaleric acid in the test group increased significantly (P < 0.01, P = 0.02). At the same time, the valeric acid content decreased significantly.

There were 1726 OUTs detected in total from the two groups. The control group (PC) had a total of 1639 OTUs, of which 293 OTUs were unique to the control group. The test group (PE) had a total of 1433 OTUs, of which 87 OTUs were unique to the test group. A total of 1346 OTUs were common to both treatment groups (Figure 1).

The PCA analysis shows (Figure 2) that the two groups were distributed at two locations. The distance is far and almost without overlap, suggesting that their floral compositions were quite different. Relatively speaking, the data distribution of the red control group (PC) was relatively discrete, indicating that there was a greater deviation in the control group.

Figure 3 shows that the two treatment groups contained many common OTUs, and this part of OTUs was abundant in each group of samples. Relatively speaking, the samples of the control group were mostly distributed above the test group. Its curve appeared to settle earlier. This suggests that supplementation of Saccharomyces cerevisiae and its fermentation products to perinatal dairy cows may reduce microbial abundance in the rumen.

Through the comparison with the database, the OTUs were classified into species and profiling histograms for each sample in several classification levels: phylum, class, order, family, genera and species. The phylum level histogram was shown in column for all species horizontally. Starting from the class level, species with abundance of less than 0.5% in all samples were combined into the “Others” category.

Figure 4 is a histogram of phyla detected. A total of 14 bacterial phylum were found in this experiment, of which 4 were more than 1%: Bacteroidetes, Firmicutes, Proteobacteria and Spirochaetes. Hard-walled bacteria in phyla Bacteroides and Proteobacteria accounted for about 90% of total microorganisms, and 8.4% (control group) and 11.5% (test group) of microorganisms were unidentified. Among the 14 phyla, the test group showed significant decreases in the

content of Candidatus, Saccharibacteria (P = 0.01), Planctomycetes (P < 0.01), Verrucomicrobia (P < 0.01), while there was a significant increase in the content of Fibrobacteres (P = 0.01).

Figure 5 is a histogram of classes. A total of 24 classes of bacteria were found in the experiment. A total of 5 classes contributed greater than 1%, including Bacteroidia, Clostridium, γ-proteobacteria, Negativicutes and Spirochaetia. In

the 24 classes, the test group showed significant decreases in the contents of Coriobacteriia (P < 0.01), Planctomycetia (P < 0.01), Verrucomicrobiae (P < 0.01), while the content of Erysipelotrichia was significantly increased (P = 0.01), and the content of Fibrobacteria was also significantly increased (P = 0.04).

Figure 6 is a histogram of orders. A total of 4 orders were detected in this test. A total of 6 classes contributed more than 1%, including Acidaminococcales, Bacteroidales, Clostridiales, Lactobacillales, Pseudomonadales and Spirochaetales. Among the 4 orders, the test group showed significant decreases in the contents of Bacillales (P = 0.04), Planctomycetales (P < 0.01), Verrucomicrobiales (P < 0.01), while there were significant increases in the content of Aeromonadales (P < 0.01), Corynebacteriales (P = 0.01), Erysipelotrichales (P < 0.01), Fibrobacterales (P = 0.04), Pasteurellales (P = 0.02), and Veillonellales (P < 0.01).

Figure 7 is a histogram of families. A total of 7 families of bacteria were detected in this experiment. A total of 8 contributed greater than 1%, including Acidaminococcaceae, Carnobacteriaceae, Lachnospiraceae, Moraxellaceae, Porphyromonadaceae and Prevotellaceae, Ruminococcaceae, Spirochaetaceae. Among the 7 families, significant decreases in the test groups were observed in the contents of Bacillaceae (P = 0.04), Clostridiales_Family_XIII, Incertae_Sedis (P < 0.01), Eggerthellaceae (P < 0.01), Ruminococcaceae (P = 0.01), Streptococcaceae (P = 0.04), Verrucomicrobia_subdivision_5 (P < 0.01). There were significant increases in the content of Corynebacteriaceae (P = 0.01), Erysipelotrichaceae (P = 0.01), Fibrobacteraceae (P = 0.04), Pasteurellaceae (P = 0.02), Succinivibrionaceae (P < 0.01), and Veillonellaceae (P < 0.01).

Figure 8 is a histogram of taxa at the genus level. Compared with the control group, the test group showed decreased contents of Bacillus (P = 0.03), Butyrivibrio (P = 0.01), Denitrobacterium (P = 0.01), and Mogibacterium (P < 0.01), Porphyromonas (P = 0.05), Saccharofermentans (P < 0.01), Sphaerochaeta (P = 0.02), Streptococcus (P = 0.04) and other genera.

There were significant increases in the contents of Acidaminococcus (P = 0.03), Allisonella (P < 0.01), Bulleidia (P < 0.01), Corynebacterium (P = 0.01), Dialister (P < 0.01), Faecalibacterium (P = 0.02), Faekalitalea (P = 0.03), Fibrobacter (P = 0.04), Flavobacterium (P = 0.03), Kandleria (P < 0.01), Paraprevotella (P < 0.01), Pyramidobacter (P = 0.05), Roseburia (P < 0.01), Succinivibrio (P < 0.01) and other genera.

As seen in Figure 9, the observed species (obs), chao values, and ace values of the control group were all above the test group, indicating that the species abundance in the test group was lower than that of the control group. In terms of the Shannon Index and the Simpson Index, there was no significant difference between the two groups. However, it can be seen that the curves of the test group are distributed on both sides of the test group. That is to say, compared with the test group, the control group had significant variations in the species abundance. We can also get the same conclusion from the data and statistics provided in Table 6. Supplementation of probiotics during the perinatal period significantly reduced the abundance of rumen microbes in dairy cows, but had no significant effect on microbial diversity.

Mass spectrometry data was acquired using a mass spectrometer Xevo G2-XS QTOF (Waters, UK). Statistical analysis of the mass spectrometry data was performed using commercialization software Progenesis QI (version 2.2) (Waters, UK). The metabolite identification was based on the database KEGG. Differentially expressed metabolites were screened using VIP values of the first two principal components of the multivariate PLS-DA model, and in combination with univariate analysis of fold-change and q-value values. Screening conditions were: 1) VIP ≥ 1; 2) fold-change ≥ 1.2 or ≤0.8333; 3) q-value < 0.05. The ions meet these three criteria were considered differential ions. Metabolic pathway analysis was based on the database KEGG (Table 7).

This test was analyzed by T test and Fold change analysis (FC analysis). In the

statistical analysis, the p-value generated by the statistical test was subjected to FDR correction to obtain a q-value. The final result presents two indicators, the fold change and the q-value, in the form of a Volcano plot. Metabolites with a fold difference ≥ 1.2 or ≤0.8333, and a q-value < 0.05 was considered differential. Figure 10 shows the volcano maps in cation and anion modes, respectively.

Partial least squares discriminant analysis (PLS-DA) is a supervised discrimination analysis statistical method that can best reflect the differences between classification groups. The method uses partial least squares regression to establish a relationship model between metabolite expression and sample categories to model and predict sample categories. It can be seen from Figure 11 that the metabolites of the control group and the test group had significant differences.

Figure 12 shows a differential ion clustering analysis of cation and anion modes. The screening revealed 639 differential cations, among them, 401 cations were upregulated and 238 were downregulated. A total of 657 differential ions were identified in anion mode, of which 469 were up-regulated and 188 were down-regulated.

Metabolic pathway analysis can help understanding the major biochemical metabolic pathways and signal transduction pathways in which the metabolites are involved. The annotation of metabolite metabolic pathways in this project was

based on the KEGG database. A total of 109 differential metabolic pathways were detected in this experiment. The metabolic pathways with consistent results in cation and anion modes, and annotated number of metabolites greater than 10 were selected as follows.