Rumen ecological distribution of Pichia yeasts and their effects on rumen fermentation and microbial community Authors: Xu Liang, Jichao Li, Pengyu Liu, Zheng Lai, Shana Huang, Fei Xie, Wei Jin, Shengyong Mao Published: 2026 ABSTRACT AbstractYeast supplementation has been widely studied to enhance rumen fermentation and feed efficiency, yet developing efficient yeasts adapted to the rumen environment remains a challenge. In this study, two rumen-derived Pichia strains (Pichia membranifaciens M12 and Pichia kudriavzevii Y4) were evaluated using in vitro rumen fermentation experiments, including a control and three supplementation groups (2 × 105, 2 × 106, and 2 × 107 CFU/mL) for each strain. Results indicated that the two strains did not affect pH but significantly reduced concentrations of ammonium nitrogen (NH3-N) and microbial crude protein (MCP). At 24 h, NH3-N decreased by up to 13.3% and MCP by 18.5%, while at 48 h, NH3-N showed a reduction of up to 22.0% and MCP decreased by up to 5.7%. P. membranifaciens significantly increased the concentration of total volatile fatty acids by 15.4% and elevated the proportions of acetate and propionate at 48 h. Microbial community analysis revealed that these shifts in fermentation parameters were associated with an altered bacterial community structure. Specifically, P. membranifaciens enriched cellulolytic bacteria (Ruminococcus), while reducing amylolytic and proteolytic taxa (Prevotella), and promoted the propionate‑producer (Succiniclasticum). These findings suggested that P. membranifaciens has the potential to influence rumen microbiota. Further examination of the in vivo prevalence of Pichia yeasts species via ITS (n = 72; average parity 2.8 ± 1.1) revealed a lower prevalence and relative abundance for P. membranifaciens compared to P. kudriavzevii. Metagenomic analysis (n = 8; average parity 2.7 ± 0.9) detected both species at low abundances. Overall, this study indicated that rumen-derived Pichia yeasts have the capacity to modulate rumen fermentation, with P. membranifaciens warranting further in vivo evaluation.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12866-026-05281-4.Keywords: Pichia, Yeasts, Rumen fermentation, Microbial community, Dairy cows Introduction The rumen harbors a complex microbial ecosystem comprising bacteria, archaea, protozoa, and fungi. These communities act synergistically to degrade plant fibers and convert feed into volatile fatty acids (VFAs) and microbial biomass, which is vital for the host [1–3]. However, maintaining the stability and metabolic efficiency of this intricate ecosystem, especially under intensive production systems, remains a significant challenge in ruminant nutrition. Consequently, substantial researches have focused on strategies to support or modulate rumen microbial function for optimal fermentation. Among these strategies, yeast-based additives, primarily Saccharomyces cerevisiae, have demonstrated considerable benefits. Extensive studies indicate that S. cerevisiae supplementation can help stabilize rumen pH, optimize the profile of VFAs, and enhance fiber degradation efficiency [4–6]. As a result, yeast supplementation has become a common practice in ruminant production to improve animal performance [7, 8]. However, the ecological adaptability of commercial S. cerevisiae strains is inherently limited, as they are not native residents of the rumen. Their functional effects are highly dependent on dietary composition and environmental conditions, leading to inconsistent outcomes [9–12]. Furthermore, yeast strains exhibit substantial functional diversity, with rumen persistence and specific metabolic activities varying considerably across species and strains [13]. These limitations suggest that the efficacy of conventional yeast additives may be constrained under certain rumen conditions, highlighting the need to explore microbial resources with inherent rumen compatibility. Recent attention has turned toward non-Saccharomyces yeasts for their potential to modulate rumen fermentation and microbial metabolism. Moreover, as native components of the rumen microbiota, endogenous yeasts offer an ecological advantage through long-term adaptation to the rumen environment. Among these, yeasts within the genus Pichia often exhibit tolerance to acid and bile, alongside the capacity to secrete hydrolytic enzymes such as cellulases and proteases, which can facilitate the degradation of fiber and nitrogenous compounds [14–16]. Previous studies suggest that certain rumen-derived Pichia strains can enhance rumen fermentation efficiency and microbial metabolic activity, potentially by improving redox balance and supplying growth-promoting factors [17, 18]. The ecological and metabolic modulations, such as the stimulation of cellulolytic bacteria and the stabilization of anaerobic conditions, have been extensively described for S. cerevisiae [19, 20]. However, whether these mechanisms can be extended to rumen-derived yeasts remains largely unknown, particularly because rumen fermentation is strongly influenced by dietary structure and feeding management. These factors affect pH stability, fermentation dynamics, and microbial activity [21], thereby further complicating the evaluation of microbial additives under varying conditions. Current knowledge of non-Saccharomyces yeasts (e.g., Pichia spp.) in the rumen is still limited and largely confined to the strain isolation and basic characterization. Their functional roles in rumen fermentation, as well as their ecological distribution and relevance, remain poorly understood [14, 17, 18]. In this study, two Pichia strains (P. membranifaciens M12 and P. kudriavzevii Y4) isolated from cow rumen were evaluated for their effects on fermentation parameters and bacterial community composition using an in vitro rumen fermentation system. Furthermore, ITS amplicon sequencing of rumen contents from 72 Holstein dairy cows (average parity 2.8 ± 1.1) and metagenomic sequencing of an independent cohort of 8 Holstein dairy cows (average parity 2.7 ± 0.9) were conducted to profile rumen yeast communities and assess the ecological prevalence of Pichia yeasts in vivo. By integrating controlled in vitro assays with in vivo microbial analysis, this research revealed the fermentation-modulating functions and ecological distribution of rumen-derived Pichia strains and characterized the ecological diversity of rumen yeast communities. Together, these findings provide insights into the ecological relationships of rumen yeasts and supported the development of rumen microbial additives. Materials and methods Isolation and identification of rumen-derived yeasts Rumen fluid was collected via oral stomach tube from three healthy Holstein dairy cows before morning feeding. These animals were sourced from the Nanjing Agricultural University Animal Experiment Farm. Serial dilutions (10–1 to 10–6) were prepared in sterile saline (0.9%) with three replicates, plated on yeast malt (YM) agar (yeast extract 3 g, malt extract 3 g, peptone 5 g, glucose 10 g, agar 20 g, distilled water 1,000 mL, streptomycin sulfate 2,000 U and potassium penicillin 1,600 U). Plates were incubated at 30 ℃ for 48 h with three replicates. From an initial pool of 73 yeast isolates, two representative Pichia strains (M12 and Y4) were subsequently selected based on their growth performance and morphological characteristics. Distinct colonies were purified and cultured in YM liquid medium for 48 h with continuous growth monitoring for subsequent identification. Genomic DNA was extracted using a fungal DNA extraction kit (Soleibao, Beijing, China), and the internal transcribed spacer (ITS) region was amplified with universal primers ITS1 (5’-TCCGTAGGTGAACCTGCGG-3’) and ITS4 (5’-TCCTCCGCTTATTGATATGC-3’). The resulting sequences were compared against the NCBI database using BLAST [22] to determine taxonomic identity. Based on the ITS sequencing and morphological observations (Fig. 1a), the isolates were identified as Pichia membranifaciens and Pichia kudriavzevii, showing 100% and 99.6% sequence similarity to their respective strains in the NCBI database. A phylogenetic tree was constructed for the two strains using the maximum-likelihood method in MEGA software (version 11.0; Fig. 1a) [23]. Isolation of two rumen-derived Pichia strains and supplementation in vitro. a Flowchart of isolation and identification of two rumen-derived Pichia yeasts. b Schematic diagram of the in vitro rumen fermentation process with supplementation of two Pichia yeasts. CFU, Colony Forming Unit; VFAs, Volatile Fatty Acids; M12, P. membranifaciens M12; Y4, P. kudriavzevii Y4 Preparation of yeast cell suspensions Both Pichia strains were cultivated in YM medium (yeast extract 3 g, malt extract 3 g, peptone 5 g, glucose 10 g, and distilled water 1,000 mL) in 250 mL Erlenmeyer flasks containing 100 mL medium and incubated for 48 h under appropriate growth conditions (30 ℃, 120 rpm) with three replicates. The cultures were initiated by inoculating 1% (v/v) of an overnight pre-culture into fresh medium. Each strain was prepared in three replicates. At the end of stationary phase, cell viability and density were assessed using 1% methylene blue staining followed by microscopic counting. The cultures were harvested by centrifugation at 8,000 rpm for 10 min, after which the cell pellets were washed twice with sterile physiological saline to remove residual medium components (Fig. 1b). The washed yeast cells were resuspended in sterile saline to obtain suspensions with the specified concentrations, which were subsequently used as inocula for the in vitro fermentation experiments. Rumen fermentation experiment Each in vitro fermentation system had a total volume of 50 mL. Each fermentation bottle contained 500 mg basal substrate (forage: concentrate = 6: 4) and 1 mL yeast suspension. The ingredient composition and nutritional levels of the basal substrate are presented in Table S1. Two incubation time points (24 and 48 h) were included. The experiment followed a 2 × 3 factorial design with two independent yeast strains (P. membranifaciens M12 and P. kudriavzevii Y4) and three supplementation levels (2 × 105, 2 × 106, and 2 × 107 CFU/mL). Accordingly, six treatment groups were established. A control (CON) group received the basal substrate supplemented with 1 mL sterile saline, while a blank control contained no substrate and 1 mL sterile saline only (Fig. 1b). The treatments were designated as M12-5/M12-6/M12-7 and Y4-5/Y4-6/Y4-7 according to supplementation levels. Each fermentation bottle was considered as the experimental unit for statistical analysis. The in vitro fermentation experiment was independently repeated three times, with each treatment performed in four replicates within each run. This sample size is consistent with standard practices in in vitro rumen fermentation studies, and the observed effect sizes with low replicate variability support adequate statistical power for the primary outcomes [24]. Rumen fluid was collected from three Holstein dairy cows before morning feeding using an oral vacuum sampler and immediately transferred into pre-warmed flasks flushed with CO2. These animals were sourced from the Nanjing Agricultural University Animal Experiment Farm. The fluid was filtered through four layers of sterile gauze and used as the inoculum. Synthetic saliva was prepared following the method of Menke et al. [25] and mixed with rumen fluid at a 2:1 ratio. Each fermentation bottle received 50 mL of this mixture and 500 mg of substrate, then was sealed with rubber stoppers and aluminum caps under CO2 to maintain anaerobic conditions. The bottles were incubated at 39 ℃ and shaken at 80 rpm. The fermentation was terminated in batches at 24 and 48 h, and samples of the fermentation broth were collected. The entire procedure was performed according to previously established methods [26]. At the end of each incubation, fermentation was terminated by immersion in an ice bath. The pH of the fermentation fluid was measured immediately, and samples were collected for the determination of VFAs, ammonia nitrogen (NH3-N), and microbial crude protein (MCP). Concentrations of VFAs were analyzed according to Jin et al. [27]. Briefly, 1 mL of fermentation sample was mixed with 0.2 mL of deproteinized-acidified solution [25% metaphosphoric acid (w/v) and 0.65% crotonic acid (w/v)] and analyzed using a gas chromatograph (GC-14B, Shimadzu, Japan). The detection conditions were as follows: column temperature 135 ℃, injector temperature 180 ℃, detector temperature 210 ℃, with crotonic acid as the internal standard and nitrogen as the carrier gas. NH3-N concentrations were determined using the phenol-hypochlorite method [28], with absorbance measured spectrophotometrically and quantified against a standard curve. MCP was quantified using the Coomassie brilliant blue G-250 method [29], using a standard curve prepared from known protein concentrations. All measurements were performed in three replicates to ensure analytical repeatability. Bacterial community profiling based on 16S rRNA sequencing The fermentation broth samples from all the time points were processed. DNA was extracted from 1 mL of fermentation liquid using the E.Z.N.A.® Soil DNA Kit (Omega Bio-Tek, USA). The bacterial 16S rRNA gene V3-V4 regions were amplified using the primers 341 F (5’-CCTAYGGGRBGCASCAG-3’) and 806R (5’-GGACTACNNGGGTATCTAAT-3’). Sequencing was performed on an Illumina MiSeq PE300 platform. Paired-end reads were merged using FLASH software (version 1.2.7) [30] and then processed with DADA2 (version 1.8) to assign reads with 100% similarity to amplicon sequence variants (ASVs) [31]. Representative ASV sequences were identified using the SILVA database (version 138.1), and α-diversity indices were calculated in QIIME2 (version 1.9.0) [32]. Principal coordinate analysis (PCoA) based on Bray–Curtis distances was performed in R (version 4.5.3). Rumen yeast community composition in cow rumen Animals were sourced from the Nanjing Agricultural University Animal Experiment Farm. Rumen contents were collected from 72 Holstein dairy cows (average parity 2.8 ± 1.1, days in milk 121 ± 48 d, and milk yield 47 ± 12.6 kg/day) that had been fed a common diet for six weeks. Samples were collected within 1 h prior to morning feeding using an oral sampler, immediately frozen in liquid nitrogen, and stored at −80 ℃ for subsequent ITS amplicon sequencing. The ingredient composition and nutritional levels of the diet were presented in Table S2. DNA was extracted from 200 mg of rumen contents using the E.Z.N.A.® Soil DNA Kit (Omega Bio-Tek, USA). Fungal ITS regions were amplified using primers ITS5F (5’-GGAAGTAAAAGTCGTAACAAGG-3’) and ITS1R (5’-GCTGCGTTCTTCATCGATGC-3’). Sequencing was conducted on the Illumina MiSeq PE300 platform. DNA fragments were merged for paired-end reads using FLASH software (version 1.2.7) [30]. DADA2 (version 1.8) was used to assign reads with 100% similarity to ASVs [31]. Representative ASV sequences were identified using the UNITE database (version 10.0). Metagenomic DNA libraries were prepared from rumen content samples of another herd of 8 Holstein dairy cows (average parity 2.7 ± 0.9, days in milk 154 ± 11 d, and milk yield 38.5 ± 6.2 kg/d) following the Illumina protocol, and sequenced on a NovaSeq 6000 platform. Raw reads were quality-filtered and adapter-trimmed using Trimmomatic (v.0.33) [33]. Host, human, and feed-derived contaminants were removed by aligning the reads to a comprehensive reference database using BWA-MEM (v.1.1.1) [34]. This reference database included the genomes of cattle (Bos taurus; GCA_002263795.2), human (Homo sapiens; GCA_000001405.28), corn (Zea mays; GCA_003185045.1 and GCA_000005005.6), soybean (Glycine max; GCA_000004515.4), wheat (Triticum aestivum; GCA_002220415.3), barley (Hordeum vulgare; GCA_900075435.2), and alfalfa (Medicago truncatula; GCA_000219495.2). Clean reads were assembled using MEGAHIT (v.1.1.1) [35] with the following parameters: -min-contig-len 500 -presets meta-large. Open reading frames were predicted using Prodigal (v.2.6.3; parameters: -p meta) [36] and clustered using CD-HIT (v.4.6.7) [37] at 95% nucleotide sequence identity to generate a non-redundant gene catalog. The representative gene catalog was then translated into protein sequences using EMBOSS Transeq (v.6.6.0.0) [38]. Taxonomic and functional annotations were performed using DIAMOND (v.0.9.22) [39] based on BLASTP searches against the NCBI-NR and KEGG (v.111.0) [40] databases. Taxonomic assignments were determined based on the NCBI-NR alignment results. Gene abundances were estimated by mapping clean reads to the gene catalog using BWA-MEM (v.1.1.1) [34] and calculated in transcripts per million (TPM), with corrections for gene length and total mapped reads per sample. The relative abundances of specific taxonomic levels and functional categories were calculated by summing the TPM values of their respective annotated genes. Statistical analysis Rumen fermentation parameters were analyzed using one-way analysis of variance (ANOVA) in SPSS 27.0 (IBM Corp., Armonk, NY, USA). At each time point, treatments were analyzed independently. Duncan’s multiple range test was applied for post hoc pairwise comparisons. Prior to analysis, microbial community data were tested for normality using the Shapiro–Wilk test and for homogeneity of variance using Levene’s test. Microbial community data were analyzed using Kruskal–Wallis tests with Wilcoxon pairwise comparisons, and differences in community structure were assessed using ANOSIM based on Bray–Curtis dissimilarities. Microbial co-occurrence networks were constructed at the genus level based on significant Spearman correlations (P < 0.05), and correlations between key microbial genera and fermentation parameters were also examined. Statistical significance was defined as P < 0.05. Results Effects of Pichia yeasts on in vitro rumen fermentation parameters To evaluate the time-dependent effects of rumen-derived Pichia strains on fermentation performance, in vitro rumen fermentation was conducted at 24 and 48 h (Fig. 2; Tables S3 and S4). After 24 h of incubation (Fig. 2a; Table S3), fermentation pH remained stable at ~ 6.2 and did not differ among treatments. The concentration of NH3-N was reduced in P. membranifaciens treatments (10–15% lower than the CON group), whereas no significant difference was observed in P. kudriavzevii groups. The concentration of MCP showed a treatment-specific response, with decreases of up to 20% in several treatments, while the highest supplementation level of strain Y4 showed an increase. Regarding VFA composition, acetate concentrations were significantly reduced in the M12-5 (58 vs. 60.6 mM; P = 0.008), M12-7 (58.6 vs. 60.6 mM; P = 0.036), Y4-6 (55.6 vs. 60.6 mM; P < 0.001), and Y4-7 (57.8 vs. 60.6 mM; P = 0.006) groups compared with the CON group. Propionate concentrations were significantly reduced in the M12-5 (28.6 vs. 30.1 mM; P = 0.027) and M12-7 (28.9 vs. 30.1 mM; P = 0.043) groups compared with the CON group. In contrast, butyrate concentrations were not significantly affected. Concentrations of total VFAs were significantly decreased in the M12-5 (112.6 vs. 117.8 mM; P = 0.022) and Y4-6 (112.9 vs. 117.8 mM; P = 0.028) groups compared with the CON group. Compared with the CON group, the molar proportion of acetate was significantly lower in the Y4-6 group (49.3% vs. 51.4%; P = 0.012), while the molar proportion of isobutyrate was significantly higher in the Y4-5 (1.1% vs. 0.9%; P < 0.001) and Y4-6 (1.1% vs. 0.9%; P = 0.006) groups. No significant differences were observed for the remaining molar proportions of VFAs compared to the CON group. Overall, after 24 h of incubation, supplementation with both Pichia strains was associated with reduced VFA concentrations and nitrogen-related parameters, whereas fermentation pH and most VFA molar proportions remained unchanged across treatments. Supplementation effects of two rumen-derived Pichia strains on fermentation parameters. a Effects of two rumen-derived Pichia strains on in vitro fermentation parameters at 24 h. b Effects of two rumen-derived Pichia strains on in vitro fermentation parameters at 48 h. CFU, Colony Forming Units; VFA, Volatile Fatty Acids, MCP, Microbial Crude Protein. Different superscript letters indicate significant differences among groups (P < 0.05, Duncan’s multiple range test), the absence of alphabetic annotations indicates that no significant differences were detected among treatments (P > 0.05). Data in the bar charts are presented as means with SEM. SEM, Standard Error of the Mean After 48 h of incubation (Fig. 2b; Table S4), fermentation pH remained stable at ~ 5.8, with no significant differences among treatments. The concentration of NH3-N was reduced across most treatments (by 5–20% relative to the CON group), and MCP concentrations were also lower (reductions of 5–10%). Compared with the CON group, acetate concentrations were significantly increased in the M12-5 (73 vs. 63.1 mM; P < 0.001) and M12-7 (69.5 vs. 63.1 mM; P = 0.013) groups; propionate concentrations were significantly increased in the M12-5 (36.1 vs. 31.8 mM; P < 0.001) and M12-7 (34.1 vs. 31.8 mM; P = 0.032) groups; and butyrate concentration was significantly increased in the M12-5 group (24.3 vs. 20.9 mM; P = 0.008). Total VFA concentrations were significantly higher in the M12-5 (144.5 vs. 125.2 mM; P < 0.001) and M12-7 (136.6 vs. 125.2 mM; P = 0.005) groups relative to the CON group. The molar proportion of acetate was significantly reduced in the Y4-6 group compared to the CON group (48.5% vs. 50.7%; P = 0.046). No significant differences were observed in the molar proportions of the other VFAs. After 48 h of incubation, fermentation responses differed between the two Pichia strains, with P. membranifaciens showing significant increases in acetate, propionate, and the total VFA concentrations accompanied by reduced NH3-N concentration, whereas changes associated with P. kudriavzevii were more limited. Effects of Pichia supplementation on rumen bacterial community structure Sequencing depths were normalized prior to relative abundance analysis. Bacterial α- and β-diversity were analyzed to assess whether Pichia supplementation altered overall community structure during in vitro fermentation (Fig. 3). After 24 h, there were no significant changes in the observed ASVs, Shannon, and Simpson indices between the CON and treatment groups (Fig. 3a). For β-diversity, no significant separation was observed between the CON and treatment groups (ANOSIM: R = 0.221, P = 0.461). After 48 h, there were no significant changes in observed ASVs, Shannon, and Simpson indices between the CON and treatment groups (Fig. 3b). For β-diversity, no significant separation was observed between the CON and treatment groups (ANOSIM: R = 0.244, P = 0.117). Collectively, supplementation with either P. membranifaciens or P. kudriavzevii did not result in significant changes in bacterial α-diversity or overall community separation. Supplementation effects of two rumen-derived Pichia strains on bacterial community structure. a Effects of Pichia supplementation on rumen bacterial community structure α- and β-diversity at 24 h in vitro. b Effects of Pichia supplementation on rumen bacterial community structure α- and β-diversity at 48 h in vitro. ASV, Amplicon Sequence Variant. Different superscript letters indicate significant differences among groups (P < 0.05, Wilcoxon rank-sum test), the absence of alphabetic annotations indicates that no significant differences were detected among treatments (P > 0.05) Effects of Pichia supplementation on bacterial composition Although the microbial community structure (PCoA) and α-diversity were unaffected, taxonomic composition was further examined to identify treatment-associated variations. After 24 h of fermentation (Fig. 4a; Tables S5 and S6), four bacterial phyla were identified based on a relative abundance of > 1% in at least one group, including Firmicutes, Bacteroidetes, Proteobacteria, and Actinobacteria. No significant changes were observed in the relative abundance of these phyla compared to the CON group. At the genus level, 21 genera were identified based on a relative abundance of > 1% in at least one group. The relative abundance of Prevotella was significantly reduced in the M12-7 (20.4% vs. 27.1%; P = 0.038) and Y4-7 (19.3% vs. 27.1%; P = 0.049) groups compared to the M12-5 group. Compared to the CON group, the relative abundance of Succiniclasticum was significantly increased in the M12-5 (8.4% vs. 6.8%; P = 0.037) and Y4-5 (8.8% vs. 6.8%; P = 0.029) groups, while the relative abundance of Lachnobacterium was significantly reduced in the M12-5 group (1.0% vs. 1.4%; P = 0.047). After 24 h of in vitro fermentation, supplementation with the two Pichia strains did not alter the relative abundance of dominant bacterial phyla but was associated with changes in several genera, including increased Succiniclasticum and reduced Lachnobacterium and Prevotella. Supplementation effects of two rumen-derived Pichia strains on bacterial composition. a Effects of Pichia supplementation on bacterial composition at the phylum and genus levels at 24 h. b Effects of Pichia supplementation on bacterial composition at the phylum and genus levels at 48 h. c Heatmap of Spearman correlations between dominant bacterial genera and fermentation parameters across treatments at 24 h and 48 h. Different superscript letters indicate significant differences among groups (P < 0.05, Wilcoxon rank-sum test), the absence of alphabetic annotations indicates that no significant differences were detected among treatments (P > 0.05). Asterisks denote significant differences relative to the CON group (*P < 0.05, **P < 0.01, ***P < 0.001, Wilcoxon rank-sum test) After 48 h of incubation (Fig. 4b; Tables S7 and S8), significant shifts in bacterial composition were observed. First, the four dominant bacterial phyla were also identified. Compared to the CON group, the relative abundance of Firmicutes was significantly increased in the M12-6 (52.1% vs. 45.9%; P = 0.03) and M12-7 (52.3% vs. 45.9%; P = 0.027) groups. Compared to the CON group, the relative abundance of Bacteroidetes was significantly decreased in the M12-6 (40.8% vs. 47.1%; P = 0.021) and M12-7 (40.6% vs. 47.1%; P = 0.034) groups. At the genus level, a total of 20 genera were identified. Compared to the CON group, the relative abundance of Prevotella was significantly reduced in the M12-7 group (11.9% vs. 17.6%; P = 0.046), and the relative abundance of Lentimicrobium (2.8% vs. 4.6%; P = 0.038) and Paraprevotella was significantly reduced in the M12-6 group (2.7% vs. 2.1%; P = 0.042). In contrast, the relative abundance of Ruminococcus was significantly increased in the M12-6 (1.9% vs. 2.5%; P = 0.035), M12-7 (1.9% vs. 2.4%; P = 0.047), Y4-6 (1.9% vs. 2.4%; P = 0.042), and Y4-7 (1.9% vs. 2.4%; P = 0.044) groups compared to the CON group. Compared to the CON group, the relative abundance of Succinivibrio was significantly reduced in the M12-6 group (1.4% vs. 0.9%; P = 0.026). After 48 h of in vitro fermentation, Pichia supplementation was associated with increased relative abundance of Firmicutes, decreased Bacteroidetes, and enrichment of Ruminococcus compared with the CON group. Correlation analysis of microbial community and fermentation parameters Correlation analysis was performed to explore associations between dominant bacterial genera and fermentation parameters (Fig. 4c). After 24 h, only a few significant correlations were detected between bacterial genera and fermentation parameters. Butyrivibrio and Lachnobacterium showed positive correlations with the concentrations of propionate and valerate. In contrast, Ruminococcus and Prevotella exhibited weak or non-significant relationships with fermentation parameters. By 48 h, the number and strength of correlations increased markedly. Ruminococcus, Lachnoclostridium, Colidextribacter, and Syntrophococcus displayed positive correlations with the concentrations of multiple VFAs (including propionate, butyrate, and valerate). Moreover, Succinivibrio and Lentimicrobium exhibited significant positive correlations with the concentration of NH3-N. Butyrivibrio showed positive correlation with pH and negative correlations with the concentrations of acetate and total VFAs. Overall, the results at 48 h exhibited a higher number of significant associations between bacterial genera and fermentation parameters compared to the 24 h. Yeast community structure and distribution in the rumen To investigate the diversity of rumen yeasts and the prevalence of Pichia, ITS amplicon sequencing analysis was conducted on 72 rumen samples collected from Holstein dairy cows (Fig. 5). According to the fungal ITS gene classification (Fig. 5a), three fungal phyla were identified based on a relative abundance > 1% in at least one group (Fig. 5b), including Ascomycota (95.7%), Chytridiomycota (2.2%), and Basidiomycota (1.2%). At the class level, five fungal classes were identified based on a relative abundance > 1% in at least one group (Fig. 5c), including Sordariomycetes (78.6%), Dothideomycetes (7.5%), Saccharomycetes (6.2%), Eurotiomycetes (2.7%), and Neocallimastigomycetes (2.2%). Within Saccharomycetes, a total of 23 yeast genera were detected, with relative abundances calculated as percentages of total Saccharomycetes sequences (Fig. 5d). The dominant genera (relative abundance > 1%) included Debaryomyces (41.7%), Kazachstania (36.8%), Candida (9.4%), Pichia (4.5%), Wickerhamomyces (2.7%), and Cyberlindnera (2.4%). Saccharomyces, a genus widely used commercially, had a relative abundance of 0.7%. Further analysis of Pichia-related sequences suggested distinct distribution patterns among species-level assignments (Fig. 5f; Fig. S1). ITS sequencing showed that P. membranifaciens was detected at low frequency (2.8%) with an average relative abundance of < 0.1%, whereas sequences assigned to P. kudriavzevii were detected in 97.2% of samples with a mean relative abundance of 4.5%. To achieve higher taxonomic resolution, shotgun metagenomic sequencing was performed on rumen samples collected from an independent cohort of 8 Holstein dairy cows. The results showed that 52 yeast genera were detected, with relative abundances calculated as percentages of total Saccharomycetes sequences (Fig. 5e). The dominant genera (abundance > 4%) included Hanseniaspora (13.5%), Ogataea (9.7%), Diutina (5.7%), Pichia (4.8%), Kazachstania (5.0%), and Brettanomyces (4.4%). At the species level, P. membranifaciens and P. kudriavzevii were detected in all samples (100% detection rate), with mean relative abundances of 2.9% and 1%, respectively (Fig. 5g; Fig. S2). Yeast community structure and distribution in the rumen. a Flowchart to analyze rumen yeast diversity and distribution via ITS gene sequencing. b Average rumen fungal phyla in 72 Holstein cows. c Rumen fungal classes in 72 Holstein cows. d ITS sequencing reveals rumen yeast genera in 72 Holstein cows. e Metagenomic sequencing reveals rumen yeast genera in 8 Holstein cows. f Prevalence of two Pichia yeast species based on ITS sequencing. g Prevalence of two Pichia yeast species based on metagenomic sequencing Discussion This study evaluated the functional potential of two rumen-derived Pichia strains using an integrated approach combining strain isolation, in vitro fermentation, and in vivo prevalence analysis. Results showed that rumen-derived Pichia, particularly P. membranifaciens, enhanced rumen fermentation and promoted fiber degradation. Unlike conventional S. cerevisiae, P. membranifaciens originates from the rumen ecosystem, suggesting inherent adaptation to its physicochemical conditions and microbial environment. In vitro fermentation over 24 and 48 h revealed time-dependent shifts in VFA profiles despite unchanged pH. Within the first 24 h, total VFA concentrations decreased (e.g., 112.6 vs. 117.8 mM) compared with the CON group, especially with P. membranifaciens, primarily due to reductions in the concentrations of acetate (58.0 ~ 58.6 vs. 60.6 mM) and propionate (28.6 ~ 28.9 vs. 30.1 mM), a transient inhibition likely reflecting competition for readily fermentable carbohydrates between Pichia yeasts and native microbes [7, 41]. Although absolute VFA levels declined, molar proportions remained stable, indicating a temporary reduction in overall fermentation activity [42, 43]. By 48 h, the system stabilized, and P. membranifaciens significantly increased VFA production compare with the CON group, with total VFA rising from 125.2 to 144.5 mM and accompanied by significant increases in acetate (73.0 vs. 63.1 mM), propionate (36.1 vs. 31.8 mM), and butyrate (24.3 vs. 20.9 mM). In contrast, P. kudriavzevii induced similar but weaker responses. Previous studies have consistently shown that yeast supplementation, particularly S. cerevisiae, enhances rumen fermentation by increasing the concentrations of total VFAs, acetate, propionate, and butyrate, while also stimulating specific bacterial populations and improving fiber degradation [19, 20, 44]. However, some studies have reported limited or no changes in total VFA despite improved rumen fermentation function, indicating that yeast effects are influenced by multiple factors such as strain, dosage, diet, and experimental conditions [21, 45]. In this study, VFA concentrations in the Pichia-supplemented groups decreased at 24 h but subsequently increased at 48 h compared to the CON group. This pattern may reflect distinct ecological interactions of rumen-derived Pichia, involving initial substrate competition and subsequent stimulation of fibrolytic and VFA-producing bacteria. Notably, both yeasts reduced NH3-N and MCP concentrations, possibly due to microbial competition for nitrogen, but these changes warrant cautious interpretation regarding in vivo utilization [46, 47]. Bacterial community analysis revealed no changes in microbial diversity, likely due to the short inoculation period, but modest genus-level shifts occurred with P. membranifaciens. After 48 h, there was a higher abundance of Ruminococcus and Succiniclasticum, genera linked to fiber degradation and propionate production [48–50], and a lower abundance of Prevotella, Succinivibrio, and Lentimicrobium, which metabolize non-cellulosic polysaccharides [51–54]. Correlation analysis indicated potential changes in the associations among fibrolytic and VFA-producing bacteria after 48 h, suggesting that P. membranifaciens may promote microbial cooperation in fiber degradation and energy metabolism. These findings suggest that the yeast-induced changes of these two strains in bacterial community structure are relatively limited. Although P. membranifaciens showed greater efficacy than P. kudriavzevii, the effects were not strictly dose-dependent, with lower inoculation levels sometimes eliciting stronger responses, highlighting the need for in vivo studies to define optimal supplementation. To better understand the biological context of these findings and determine whether the superior efficacy of P. membranifaciens observed in vitro stems from its natural dominance or a specific ecological gap, we further investigated the distribution of these two species within the rumen. The ITS sequencing analysis of rumen samples from 72 dairy cows revealed a striking contrast in their ecological prevalence. The results (within the class Saccharomycetes) revealed that P. kudriavzevii was widely distributed (97.2% detection rate; 4.5% mean abundance), whereas P. membranifaciens was rare (2.8% detection rate; < 0.1% mean abundance). In addition, other yeast genera such as Debaryomyces, Kazachstania, and Candida were also detected. The high abundance of these yeast genera could be attributed either to continuous dietary introduction or to adaptation to the rumen environment, or a combination of both [55–59]. Metagenomic sequencing provided validation of the species-level assignment of Pichia. Therefore, Pichia as a low-abundance strain that exhibited promising functional properties (e.g., greater enhancements in total VFAs and fibrolytic bacteria) under the tested conditions points to a potential opportunity for the development of probiotic interventions aimed at improving rumen fermentation efficiency [60–62]. It should be noted that this study was conducted under in vitro conditions, which do not fully replicate the complexity of the in vivo rumen environment, including continuous nutrient flow, host-microbe interactions, and dynamic microbial adaptation. Therefore, the observed effects should be interpreted with caution when extrapolating to practical applications. Future in vivo studies are needed to validate the functional effects of rumen-derived Pichia strains, determine optimal supplementation strategies under different dietary conditions, and assess the colonization potential of these strains in the rumen. Moreover, this study did not investigate the combined effects of the two Pichia strains; future research should explore the interactions among multiple strains. The mechanisms underlying the effects of these yeasts also warrant further investigation. Conclusions This study presents an evaluation of two rumen-derived Pichia strains using an in vitro rumen fermentation system, supplemented by ITS and metagenomic sequencing. Compared with the other strain evaluated, P. membranifaciens showed a more pronounced and consistent modulatory effect on overall rumen fermentation parameters by positively regulating carbohydrate degradation, VFA profiles, and the associated bacterial community. Although naturally scarce in the rumen, its exogenous supplementation effectively amplified its ecological contribution. Based on its performance, P. membranifaciens demonstrates promising potential to be developed as a probiotic, with an optimal in vitro supplementation level of 2 × 106 to 2 × 107 CFU/mL under the conditions tested. Further in vivo studies, ideally combined with multi-omics approaches, are required to validate its effects and clarify its ecological interactions before practical application. Supplementary Information Additional file 1: Fig. S1 and S2; Table S1, S2, S3, S4, S5, S6, S7, and S8. Acknowledgements Our project was supported by the high-performance computing platform of Bioinformatics Center, Nanjing Agricultural University. Abbreviations Amplicon Sequence Variant Branched-Chain Fatty Acids Colony Forming Unit Internal Transcribed Spacer Microbial Crude Protein National Center of Biotechnology Information Ammonia Nitrogen Principal Coordinate Analysis Polymerase Chain Reaction Volatile Fatty Acid Yeast Malt Authors’ contributions S.M., W.J., and F.X. conceived, designed, and supervised the project. X.L., J.L., P.L., and Z.L. conducted the experiments and collected samples. X.L. and J.L. carried out the bioinformatic analyses. X.L. and J.L. visualized and interpreted the data. X.L. and J.L. drafted the paper. F.X., W.J., and S.M. revised the paper. All authors read, edited, and approved the final manuscript. Funding This work was supported by the National Natural Science Foundation of China (32272896), the National Natural Science Foundation of China (32361143788), and the Fundamental Research Funds for the Central Universities (QTPY2026017). Data availability The sequencing data for the in vitro fermentation fluid and rumen content samples have been deposited in the NCBI Sequence Read Archive (SRA) under the accession numbers PRJNA1359379 (bacteria) and PRJNA1359369 (fungi). Metagenomic sequence reads for all samples in this study are available in the European Nucleotide Archive (ENA) under study accession number PRJNA1347380 (CON group). Other datasets used and analyzed during the current study are available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate All procedures were approved by the Animal Care and Use Committee of Nanjing Agricultural University (Nanjing, China) and conducted in accordance with the Regulations on the Administration of Laboratory Animals issued by the National Science and Technology Commission of China (1988). The ethical approval numbers were NJAU.No20230416N06 for donor animals used in this study. The donor animals are provided by the Nanjing Agricultural University Animal Experiment Farm, and informed consent was obtained from the farm management for their use in this research. The study does not involve any form of anesthesia or euthanasia. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Xu Liang and Jichao Li contributed equally to this work. Contributor Information Fei Xie, Email: xiefei_njau@163.com. Wei Jin, Email: jinwei@njau.edu.cn. Shengyong Mao, Email: maoshengyong@njau.edu.cn. References Associated Data Supplementary Materials Additional file 1: Fig. S1 and S2; Table S1, S2, S3, S4, S5, S6, S7, and S8. Data Availability Statement The sequencing data for the in vitro fermentation fluid and rumen content samples have been deposited in the NCBI Sequence Read Archive (SRA) under the accession numbers PRJNA1359379 (bacteria) and PRJNA1359369 (fungi). Metagenomic sequence reads for all samples in this study are available in the European Nucleotide Archive (ENA) under study accession number PRJNA1347380 (CON group). Other datasets used and analyzed during the current study are available from the corresponding author upon reasonable request.