Samples of rumen fluid were obtained from three Holstein dairy cows with cannulas (weighing approximately 600 ± 50 kg) to serve as the bacterial source for rumen analysis, which is a part of our previous study and has received ethical approval (Liu et al., 2023). All the procedures involving the care and management of dairy cows were approved by the Animal Care and Use Committee for Livestock of the Institute of Animal Sciences, Chinese Academy of Agricultural Sciences (protocol no.: IAS201914). The dairy cows were fed a typical total mixed ration (DM base) consisting of 62.09% corn silage, 7.66% alfalfa hay, 10.25% corn powder, 17.87% soybean meal, and 2.13% premix. The rumen content was collected and subsequently filtered through a four-layered cheesecloth to separate the liquid. Immediately, the liquid samples were combined (in equal amounts) and injected into a penicillin bottle along with an equivalent volume of sterile glycerol (15%, v/v) using a syringe. Subsequently, they were preserved at −80°C until required.

The microbiota from the rumen were fractionated based on different cell sizes using polyvinylidene difluoride filters with 300, 120, 80, 40, 20, 8, 6, 2.1, and 0.6 μm pore sizes (Merck, Merck Millipore Ltd., Ireland). The rumen fluid was successively passed through the 300, 120, 80, 40, 20, 8, 6, 2.1, and 0.6 μm filters. Four replicates were set up for each experimental treatment. To increase the flow of filtering, we used an injector, and the volumes filtered were less than 5 mL. Retentates were washed three times using a PBS solution. Bacterial fractionation resuspensions were centrifuged (13,000× g, 2 min, 4°C), and the sediment pellets were reserved. Consequently, bacteria fractions with size ranges at ≥300 μm, 120–300 μm (120 μm ≤ * < 300 μm), 80–120 μm (80 μm ≤ * < 120 μm), 40–80 μm (40 μm ≤ * < 80 μm), 20–40 μm (20 μm ≤ * < 40 μm), 8–20 μm (8 μm ≤ * < 20 μm), 6–8 μm (6 μm ≤ * < 8 μm), 2.1–6 μm (2.1 μm ≤ * < 6 μm), 0.6–2.1 μm (0.6 μm ≤ * < 2.1 μm), and < 0.6 μm were obtained.

DNA was extracted using the cetyltrimethylammonium bromide (CTAB) method plus bead beating, as previously documented (Liu et al., 2023). The quality of DNA was assessed through the utilization of agarose gel electrophoresis (1%), while the measurement of DNA concentration was conducted using a Nanodrop spectrophotometer (Thermo Scientific, United States).

The microbial diversity was detected by sequencing the V3–V4 region of 16S rRNA, which was amplified using PCR with the primers 338F (5´-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5´-GGACTACHVGGGTWTCTAAT-3′). The reactions were conducted using a MyCycler Thermal Cycler (Bio-Rad, United States) following the protocol described by Liu et al. (2020). The PCR amplicons were subjected to sequencing using the Illumina MiSeq platform at Majorbio (Shanghai, China).

Sequencing raw data were analyzed using the QIIME package (Bolger et al., 2014). First, the raw reads were trimmed to remove sequencing adaptors, followed by filtering to exclude any bases containing ambiguous information. Next, paired-end reads were merged using FLASH (Magoč and Salzberg, 2011) with parameters set at an overlap of >10 bp and a mismatch rate of <20%. Chimera sequences were removed using the UCHIME de novo algorithm (Wang et al., 2007). Operational taxonomic units (OTUs) were grouped by applying a threshold of 0.97 similarity using the USEARCH algorithm (Caporaso et al., 2010; Edgar, 2010). Finally, the OTUs were assigned to taxa by the RDP classifier based on the Greengenes database (13_5) (Wang et al., 2007).

The bacterial relative abundance data were analyzed using the Kruskal–Wallis test in SAS. The difference in alpha diversity across various pore sizes was assessed using ANOSIM within the QIIME software (Liu et al., 2020). Heatmaps and heatmap diagrams were generated using MicrobiomeAnalyst (Lu et al., 2023).

Data are deposited in the China National Microbiology Data Center (NMDC)¹ with accession number NMDC10018589. ²

After obtaining a total of 2,041,394 raw sequence reads, with an average of 49,510 reads, we assigned all the reads to 7,168 OTUs. The Chao1 and Shannon indices of alpha diversity in the <0.6 μm fraction were significantly higher, and there were no significant differences in the >0.6 μm fraction (Figures 1A,B).

The bacteria from the various pore sizes were clustered into four branches: >120 μm, 40–120 μm, 6–20 μm, and 20–40 μm, <0.6 μm (Figure 2). The changes in community composition indicate that filtration through different cell sizes could lead to the preliminary differentiation of rumen bacteria.

The heatmap shows that the filtration method had effects on enriching specific groups of bacteria. The community of relative abundance >1% in the >40 μm fraction was unclassified Bacteria, unclassified Bacteroidetes, unclassified Firmicutes, unclassified Prevotellaceae, unclassified Anaeroplasma, unclassified RF16, unclassified Erysipelotrichaceae, unclassified Mitsuokella, unclassified Butyrivibrio, unclassified Mollicutes, unclassified Oscillospira, unclassified Mogibacteriaceae, unclassified S24_7, D168 Desulfovibrio, unclassified Christensenellaceae, unclassified Moryella, unclassified TM7, and unclassified Ruminococcus (Figure 3A). The 6–20 μm fraction was unclassified TM7, unclassified Ruminococcus, unclassified Coprococcus, unclassified Roseburia, unclassified SR1, unclassified Schwartzia, unclassified Elusimicrobiaceae, unclassified Veillonellaceae, unclassified Gammaproteobacteria, unclassified Prevotella, unclassified Selenomonas, unclassified Oribacterium, unclassified Sutterella, Lachnospira pectinoschiza, unclassified Pseudobutyrivibrio, unclassified Campylobacter, unclassified Shuttleworthia, unclassified Succinivibrio, and unclassified Anaerovibrio (Figure 3A). The <0.6 μm fraction was unclassified Bacteria, unclassified Firmicutes, unclassified Bacteroidetes, unclassified Prevotellaceae, unclassified RF16, unclassified Erysipelotrichaceae, unclassified Mitsuokella, Fibrobacter succinogenes, unclassified Treponema, unclassified Schwartzia, unclassified Elusimicrobiaceae, unclassified Veillonellaceae, unclassified Gammaproteobacteria, unclassified Prevotella, unclassified Selenomona, unclassified Oribacterium, unclassified Sutterella, Lachnospira pectinoschiza, unclassified Pseudobutyrivibrio, unclassified Campylobacter, unclassified Shuttleworthia, unclassified Succinivibrio, and unclassified Anaerovibrio (Figure 3A). The 2.1–6 μm fraction was unclassified Prevotella, Fibrobacter succinogenes, and unclassified Treponema (Figure 3A).

The community of relative abundance <1% in the <20 μm fraction was unclassified Desulfovibrio, unclassified R4_45B, unclassified Victivallaceae, unclassified Clostridium, unclassified Succinivibrionaceae, and unclassified RF32 (Figure 3B). The 20–120 μm fraction was unclassified Sphingomonas, unclassified Caulobacteraceae, unclassified Chryseobacterium, unclassified Flavobacterium, Methylotenera mobilis, unclassified Delftia, and unclassified Stenotrophomonas (Figure 3B). The >120 μm fraction was unclassified Ruminococcus, Syntrophococcus sucromutans, unclassified Opitutae, unclassified Coprococcus, and Sharpea azabuensis (Figure 3B).

The species of unclassified Flavobacterium were enriched in the >6 μm fraction, with significantly higher abundance observed in the 20–120 μm fraction (Figure 4A). The species of unclassified Chryseobacterium were enriched in the 6–300 μm fraction, with significantly higher abundance observed in the 20–40 μm and 80–120 μm fractions (Figure 4B). The species of unclassified Delftia were enriched in the >6 μm fraction, with significantly higher abundance observed in the 20–120 μm fraction (Figure 4C). The species of Methylotenera mobilis were enriched in the 20–300 μm fraction, with significantly higher abundance observed in the 20–120 μm fraction (Figure 4D). The species of unclassified Caulobacteraceae were not enriched in the 2.1–6 μm fraction, with significantly higher abundance observed in the 20–40 μm and 80–120 μm fractions (Figure 4E). The species of unclassified Oligella were not enriched in the 2.1–6 μm fraction, with higher abundance observed in the >300 μm fraction (Figure 4F). The species of unclassified Sphingomonas were not enriched in the <0.6 μm, 2.1–6 μm, and 20–40 μm fractions, with significantly higher abundance observed in the 40–120 μm fraction (Figure 4G).

The abundance of unclassified Stenotrophomonas was significantly higher in the 80–120 μm fraction (Figure 5A). The abundance of unclassified Shuttleworthia was significantly higher in the 6–20 μm fraction (Figure 5B). The abundance of unclassified Sutterella was significantly higher in the <0.6 μm fraction (Figure 5C). The abundance of unclassified Alphaproteobacteria was significantly higher in the 8–20 μm fraction (Figure 5D). The abundance of unclassified SR1 was significantly higher in the 8–20 μm fraction (Figure 5E).