Enhancing sheep rumen health with Przewalskia tangutica: boosting fermentation and microbial diversity in saline-alkali sown pastures

Abstract

Background

Livestock grazing on saline-alkali pastures frequently encounter nutritional challenges due to inadequate forage quality and imbalanced rumen function. Phytogenic feed additives, which are abundant in bioactive compounds such as tropane alkaloids and flavonoids, have demonstrated potential in enhancing rumen fermentation and improving animal performance. Przewalskia tangutica, a desert plant rich in these compounds, remains largely unexplored concerning its effects on microbiota and metabolism in sheep under saline-alkali conditions. This study aims to evaluate the impact of graded supplementation of Przewalskia tangutica on rumen fermentation, microbial dynamics, digestive enzyme activity, and growth performance in sheep grazing on degraded saline-alkali pastures.

Results

Supplementation with Przewalskia tangutica significantly increased the average daily gain (ADG, P < 0.001) and reduced the feed conversion ratio (FCR, P < 0.001). The high-dose group (HP) exhibited the highest ADG at 139.94 g/d. Additionally, the concentrations of total volatile fatty acids (VFAs), including acetate, isobutyrate, and butyrate, were significantly elevated (P < 0.001), peaking at 1.46 mmol/L in the HP group. Furthermore, the activities of amylases, lipase, cellulase, and protease increased significantly (P < 0.05). While the bacterial alpha diversity remained unchanged, beta diversity analysis revealed structural shifts characterized by increased abundances of Verrucomicrobiota, Proteobacteria, and Fibrobacterota (P < 0.05). Fungal diversity also increased under the HP treatment (P < 0.05), whereas a reduction in Cladosporium abundance was correlated with enhanced fiber degradation. Moreover, microbial co-occurrence networks demonstrated more balanced interactions in the treated groups, with significant correlations among taxa, enzymes, and VFAs.

Conclusions

Przewalskia tangutica improves rumen fermentation and sheep performance in saline-alkali environments through targeted modulation of microbial communities and enzymatic activity, particularly at the high dose, significantly improved sheep growth performance by increasing average daily gain and reducing the feed conversion ratio, suggesting a sustainable strategy for livestock production in challenging ecosystems.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s42523-026-00592-3.

Importance

This study highlights Przewalskia tangutica as a breakthrough phytogenic solution for livestock production in saline-alkali regions, where poor forage quality and microbial dysbiosis limit productivity. Supplementation significantly enhanced sheep growth (↑ADG, ↓FCR) and rumen fermentation (↑VFAs) by modulating microbial communities—enriching fiber-degrading Fibrobacterota while suppressing salinity-tolerant Cladosporium fungi. These shifts correlated with improved fiber utilization and enzyme activities, directly addressing rumen inefficiencies under saline stress. By leveraging native plant bioactives, this strategy revitalizes degraded pastures, reduces dependency on synthetic additives, and enhances climate resilience in arid ecosystems. These findings provide a sustainable blueprint for optimizing livestock performance in marginal environments and advancing food security and ecological restoration in salt-affected regions globally.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s42523-026-00592-3.

Introduction

Ruminant livestock plays a crucial role in global agriculture and food security due to their unique ability to convert lignocellulosic plant biomass, which is indigestible to monogastric animals, into high-quality protein and energy sources for human consumption [2, 15, 40]. This remarkable digestive capacity is housed within the rumen, a complex anaerobic ecosystem often described as a 'natural fermentation vat' [28, 75]. The rumen hosts a dense and diverse consortium of microorganisms, including bacteria, archaea, protozoa, and fungi, which collectively ferment ingested feed into volatile fatty acids, the primary energy source for the host, along with microbial protein and essential vitamins [5, 24]. The stability and efficiency of this microbial ecosystem are critical for animal health, nutrient utilization, and overall productivity.

The rumen microbiome is notably sensitive to a variety of factors, including host genetics, age, environmental conditions, and especially dietary composition [26, 33, 37]. Livestock production in challenging environments, such as saline–alkali pastures, presents unique nutritional challenges. Forages cultivated under these conditions often exhibit altered nutritional profiles and may contain elevated levels of salt, which can impact palatability, digestibility, and ultimately the rumen environment itself [1]. Anecdotal and observational evidence frequently suggests suboptimal rumen function and diminished livestock performance in these regions, potentially linked to imbalances within the rumen microbial community and fermentation processes [32]. Consequently, strategies to enhance forage utilization and improve ruminant productivity on saline–alkali pastures are of significant practical and ecological importance.

The use of functional native herbage, particularly phytogenic compounds derived from plants, has emerged as a promising strategy to sustainably modulate rumen fermentation and enhance animal performance [49, 63, 75]. Przewalskia tangutica, a plant belonging to the Solanaceae family and native to high-altitude regions such as the Qinghai-Tibetan Plateau, has been traditionally utilized in folk medicine. It is known to contain various bioactive compounds, including tropane alkaloids (e.g., scopolamine and anisodamine) and flavonoids (e.g., quercetin) [52, 67]. These compounds exhibit anticholinergic, antioxidant, anti-inflammatory, and antimicrobial activities [17, 31, 60, 78]. Scopolamine and other tropane alkaloids can inhibit salivary and gastric acid secretion, reduce gastrointestinal smooth muscle tone and motility, and prolong gastric emptying time [9]. Oral administration of scopolamine in rats has been shown to suppress gastric acid secretion and delay gastric emptying [8]. Flavonoid compounds such as quercetin have also demonstrated antioxidant, anti-inflammatory, and immunomodulatory activities in various animal models, including improvements in intestinal barrier function and modulation of gut microbiota composition [20, 46]. In in vitro systems simulating rumen fermentation, quercetin altered rumen fermentation parameters and bacterial communities, reduced methane production, and promoted nitrogen utilization and microbial protein synthesis in the rumen of beef cattle [45, 68]. These alkaloids may subtly modulate rumen motility and digesta passage rate, while flavonoids exert selective antimicrobial activity, suggesting a potential dual-action mechanism for optimizing fermentation efficiency in sheep consuming saline-alkali forages. However, excessive scopolamine intake has been associated with toxic effects, including hepatic and renal damage in mice [3], metabolic disturbances in zebrafish [27], and gastrointestinal toxicity (diarrhea, vomiting) in humans [39]. Some studies suggest that certain species, including poultry, rabbits, and specific small ruminant breeds, possess particular hydrolases that can effectively inactivate alkaloids and are thus less sensitive to toxicity [30]. Nevertheless, investigating the specific interactions between these compounds and rumen microbes under controlled application regimes represents a critical avenue for future research, particularly for livestock production in challenging saline-alkali environments.

Addressing this knowledge gap is vital for evaluating the potential of Przewalskia tangutica as a targeted feed additive to enhance digestive efficiency and productivity in sheep grazing on marginal lands. Therefore, this study aimed to investigate the dose-dependent effects of dietary supplementation with Przewalskia tangutica on rumen fermentation characteristics, microbial community structure (including bacteria and fungi), key digestive enzyme activities (e.g., cellulase, amylase, and protease), and microbial ecological networks in sheep grazing on natural saline–alkali pastures. It was hypothesized that supplementation with Przewalskia tangutica would positively modulate the rumen environment by increasing VFA production, altering the relative abundances of specific functional microbial groups, influencing digestive enzyme activity, and modifying intermicrobial relationships, thereby potentially improving the digestive capacity of sheep under saline–alkaline grazing conditions.

Materials and methods

The experiments were conducted according to the guidelines of experimental field management protocols (file No: 2010–1 and 2010–2) and were approved by the Animal Use and Care Committee of Lanzhou University. All procedures for handling and caring for animals conformed with China’s regulations on the protection and use of laboratory animals and were conformed with the guidelines of the Chinese Zoological Society.

Test site

This study was conducted at Linze Pratacultural Research Station of Lanzhou University, which is located in Linze County, Zhangye city, Gansu Province, China (100°02′E, 39°15′N; altitude, 1390 m; mean annual temperature, 7.16 °C; mean annual precipitation, 121.5 mm). The soil texture is sandy loam and classified as Aquisalids according to USDA soil taxonomy, the topsoil layer (0–15 cm) had a pH of 7.8, salinity content of 0.88% [23, 43]. Climate data of experimental period (June–August): mean temperature 23.4 °C (range 18.2–28.7 °C) and precipitation 87.3 mm. It focuses primarily on a specialized intensive cropping production system (SICP) and an extensively integrated crop‒livestock production system (EICL), both of which are integrated within a broader agricultural framework to support sustainable land use and productivity in arid and semiarid regions [18].

Experimental animals and group design

This study employed a Randomized Complete Block Design (RCBD) to control variability associated with the initial body weight of the animals. Twenty-four male hybrid lambs (Small-tailed Han sheep × Hu sheep; initial BW = 29.40 ± 1.25 kg; age = 6 months) were ranked by body weight and allocated to six blocks (n = 4 lambs per block). Within each block, lambs were randomly assigned to one of four dietary treatments (n = 6 per treatment): a control group (CON: 0% Przewalskia tangutica + 100 g soybean meal), a low-level Przewalskia tangutica group (LP: 1.5% Przewalskia tangutica + 100 g soybean meal), a medium-level group (MP: 3% Przewalskia tangutica + 100 g soybean meal), and a high-level group (HP: 4.5% Przewalskia tangutica + 100 g soybean meal). Throughout the trial, the basal diet consisted of fresh pasture forage from sown pastures (a mixture of alfalfa and tall fescue). Aside from the forage, the only other feed provided was the experimental supplement (Przewalskia tangutica and soybean meal). The Przewalskia tangutica supplement was prepared by harvesting the whole plant, air-drying it, and grinding it into a fine powder, which was then mixed with the soybean meal carrier for daily administration. The inclusion rates of Przewalskia tangutica were determined based on a preliminary in vitro fermentation experiment where these doses demonstrated a positive dose-dependent effect on VFA production without inhibiting overall gas production, the rationality of this inclusion level was further validated through ruminant digestion and metabolism trials [75]. The inclusion rates of Przewalskia tangutica (1.5%, 3.0%, and 4.5%) were calculated on a dry matter (DM) basis as a percentage of the estimated total daily dry matter intake (900 g/d), in accordance with the Chinese Standard for Meat Sheep Raising [35]. The daily supplementation levels were 13.5 g for LP, 27 g for MP, and 40.5 g for HP. A fixed amount of 100 g of soybean meals was provided to all animals, including the control group, to serve as a uniform and palatable carrier, thereby ensuring consistent intake of the experimental treatments. To control for potential confounding effects arising from the small variation in initial body weight, a Randomized Complete Block Design was utilized. This design statistically isolates the treatment effects from the influence of initial body weight, strengthening the validity of our comparisons.

The trial consisted of a 15-day adaptation period followed by a 32-day experimental phase. During the adaptation phase, sheep grazed together on the sown pastures, which consisted of a mixture of alfalfa and tall fescue, from 7:00 AM to 7:00 PM, with ad libitum access to water. Each evening, they were provided with their respective Przewalskia tangutica supplement in individual stalls to ensure complete consumption. The experimental phase included a digestion and metabolism trial conducted in four consecutive temporal batches, with six sheep per batch (n = 6, balanced across treatments), ensuring balanced representation from each treatment. Due to the logistical constraints of metabolic cage availability and intensive sampling procedures, not all 24 animals could be processed simultaneously. Therefore, the four batches were run sequentially under identical environmental and management conditions. The factor ‘batch’ (1–4) was subsequently included as a random effect in all statistical models to account for potential temporal variation (e.g., minor fluctuations in ambient temperature, humidity, or feed batch). Sheep in the trial were housed in individual metabolism cages and were provided with a basal diet of freshly cut grass from the same grazing area to standardize nutrient intake. The amount of grass offered was adjusted daily to maintain approximately 10% refusal, facilitating accurate calculation of individual dry matter intake (DMI). The Przewalskia tangutica supplement was administered once daily in the evening. The nutritional composition of the pasture and the supplement is detailed in Table 1.

Table 1.

Nutritional level of sown pasture forage and Przewalskia tangutica R.Br. (DM basis)

Item	Przewalskia tangutica	Sown pastures forage	Soybean meal
Nutritional level (as-fed basis)
DM	43.76	34.28	89.50
Nutritional level (%, as DM)
OM	87.68	93.21	93.64
CP	17.09	17.72	52.80
EE	3.84	3.57	1.65
NDF	61.17	63.05	12.52
ADF	37.75	38.09	7.31
GE (MJ/kg, DM)	16.98	16.92	18.75
Functional ingredient (%, as DM)
Scopolamine	0.46	/	/
Anisodamine	0.15	/	/
Anisodine	0.09	/	/
Quercetin	0.04	/	/

Note: DM, dry matter; OM, organic matter; NDF, neutral detergent fiber; ADF, acid detergent fiber; EE, ether extract; GE, gross energy

Sample collection and processing

During the sample collection period, the daily grass intake and refusal amounts for each lamb were recorded, and representative samples were collected, while precise individual forage intake during free-grazing was not measured, all sheep grazed together on the same pasture to standardize forage exposure. Importantly, during the digestion and metabolism trial, sheep were individually fed a known amount of freshly cut grass from the grazing area, allowing for accurate determination and standardization of DMI across this critical period. Body weight was recorded at trial initiation (Day 0) and on the final morning after overnight fasting (Day 32). Two hours after returning from grazing that evening, 50 ml of rumen fluid was collected from each lamb using a rumen fluid sampling tube, which consisted of an outer stainless-steel elastic hose and an inner transparent latex tube. The initial portion of the collected fluid was discarded to avoid contamination from saliva, gastrointestinal mucosal secretions, or residual distilled water. Within 10 min, the ruminal fluid was filtered through four layers of medical gauze and immediately analyzed for pH using a portable pH meter (Model 144 PB-10, Sartorius Co., Germany). The filtration was aliquoted into sterile centrifuge tubes and snap-frozen in liquid nitrogen. Within 30 min after the completion of ruminal fluid collection from all sheep, the liquid nitrogen-frozen samples were transferred to appropriate freezers: samples for DNA extraction were stored at -80 °C, while those for volatile fatty acid and enzyme analysis were stored at -20 °C.

Sample analysis

Nutritional levels of forage

Nutritional evaluation of forage and Przewalskia tangutica was conducted following standard forage analysis protocols [47]. The fresh weights of the collected forage samples were recorded, and the samples were subsequently dried in a 65 °C oven (DHG-9240 A, Jing Hong Instruments, China) for 72 h to determine the dry matter content (method 930.15). Following drying, the samples were ground and passed through a 1 mm sieve for nutritional analysis. Ash and organic matter contents were measured (method 942.05) by combusting the samples in a muffle furnace (SX-G03173, Zhonghuan Electric Furnace, China) at 550 °C for 5 h. The nitrogen content was determined (method 990.03) via a Kjeldahl nitrogen analyzer (Model K9840; Hanon Instruments, China), and the crude protein content was calculated as the nitrogen content multiplied by 6.25. Ether extract content was determined (method 920.39) via a fully automated fat meter (ANKOMXT15, ANKOM, USA). Neutral detergent fiber (method 2002.02) and acid detergent fiber (method 973.18) contents were determined via the filter bag technique with a semiautomatic fiber analyzer (A2000I, Ankom Instrument Co., Ltd., USA) [34]. Gross energy was determined via an automatic oxygen bomb calorimeter (PARR6400, PARR, USA) [72]. High-performance liquid chromatography‒tandem mass spectrometry (HPLC‒MS/MS) was used to quantify the bioactive compounds scopolamine, anisodamine, anisodine and quercetin. HPLC separation was performed via a Kinetex C18 column (150 mm × 4.6 mm, 5 μm particle size; Phenomenex), and the column temperature was maintained at 30 °C. The mobile phase consisted of gradient elution of solvent A (0.1% formic acid in water) and solvent B (acetonitrile/methanol, 80:20, v/v) at a flow rate of 1.0 mL/min. Detection was achieved via an Agilent 6460 Triple Quadrupole MS with electrospray ionization (ESI) in positive ion mode. Multiple reaction monitoring (MRM) transitions were optimized as follows: scopolamine 304.1→138.1; anisodamine 306.2→140.1; anisodine 318.2→156.1; quercetin 301.0→151.0. Atropine-d5 was used as an internal standard for alkaloids, and quercetin-d4 for quercetin quantification [29, 73].

Volatile fatty acids and digestive enzyme activity of rumen fluid

The concentrations of volatile fatty acids were determined using gas chromatography (GC–7890B, Agilent Technologies) as previously described by Liu et al., [25], with modifications. Briefly, rumen fluid supernatant was mixed with 25% (w/v) meta-phosphoric acid and 4-methylvaleric acid as an internal standard. The mixture was then centrifuged, and the supernatant was injected into the gas chromatography system, which was equipped with a flame ionization detector (FID) and a DB-FFAP capillary column (30 m × 0.32 mm × 0.25 μm, Agilent J&W). The carrier gas used was nitrogen at a flow rate of 2.0 mL/min. The oven temperature program was set as follows: an initial temperature of 80 °C held for 1 min, increased to 180 °C at a rate of 20 °C/min, further increased to 240 °C at a rate of 5 °C/min, and held for an additional 5 min. The injector and detector temperatures were maintained at 250 °C and 300 °C, respectively. VFA concentrations (mmol/L) were calculated based on the internal standard and external calibration curves. Additionally, a SpectraMax 190 Microplate Reader (MD, New York, NY, USA) was employed to evaluate the activities of digestive enzymes in the rumen fluid, following the manufacturer’s instructions provided by Shanghai Kexing Biotechnology Co., Ltd. (Shanghai, China).

DNA extraction and amplicon sequencing of rumen fluid

Total genomic DNA was extracted from rumen fluid samples via the TGuide S96 Magnetic Soil/Stool DNA Kit (Tiangen Biotech, Beijing, China) according to the manufacturer’s instructions. Samples were lysed using bead-beating at 30 Hz for 5 min with 0.1 mm and 0.5 mm zirconia beads (1:1 ratio). DNA concentration and purity were assessed via Qubit 4 Fluorometer (Thermo Fisher Scientific, Wilmington, USA) and NanoDrop 2000 UV-Vis spectrophotometer (A260/A280 ratio > 1.8, A260/A230 ratio > 1.5). Extracted DNA integrity was verified by electrophoresis on 1.5% agarose gels. Negative controls, including DNA extraction blanks and PCR no-template controls, were processed alongside the samples to monitor for potential contamination; any ASVs found to be significantly more abundant in negative controls were removed from the dataset. The V3–V4 hypervariable region of the bacterial 16S rRNA gene was amplified via the primer pair 338F (5’-ACTCCTACGGGAGGCAGCA-3’) and 806R (5’-GGACTACHVGGGTWTCTAAT-3’). Similarly, the fungal internal transcribed spacer (ITS) region was targeted with the primer pair ITS1F (5’-CTTGGTCATTTAGAGGAAGTAA-3’) and ITS2R (5’-GCTGCGTTCTTCATCGATGC-3’). Polymerase chain reaction (PCR) was performed in a total volume of 10 µl, containing 5–50 ng of DNA template, 0.3 µl of forward primer (10 µM), 0.3 µl of reverse primer (10 µM), 5 µl of KOD FX Neo buffer, 2 µl of dNTP mixture (2 mM each), 0.2 µl of KOD FX Neo polymerase, and sterile distilled water (ddH₂O) to adjust the volume. Thermal cycling conditions included initial denaturation at 95 °C for 30 s, annealing at 50 °C for 30 s, extension at 72 °C for 40 s, and a final extension at 72 °C for 7 min. Amplicons were purified via an Omega DNA purification kit (Omega Inc., Norcross, GA, USA) and quantified with the Qsep-400 system (BiOptic Inc., New Taipei City, Taiwan, ROC). Library preparation and paired-end sequencing (2 × 250 bp) were conducted on the Illumina NovaSeq 6000 platform (Beijing Biomarker Technologies Co., Ltd., Beijing, China).

Sequencing data processing and data analysis

Raw sequencing reads were pre-processed using fastp (version 0.19.6, [11]) to filter low-quality sequences (Q < 20) and trim adapters. High-quality reads were processed using the QIIME 2 platforms (version 2020.2, [6]). The DADA2 plugin [10] was employed for denoising, paired-end merging and chimera removal to generate amplicon sequence variants (ASVs). To control potential contamination, ASVs identified in negative control samples (PCR water and extraction blanks) were removed from the dataset. The feature table was further filtered to remove low-abundance noise; only ASVs present in at least 5% of samples with a total abundance count > 10 were retained. This resulted in an average of 75,934 high-quality reads per sample for bacteria and 67,370 for fungi, which were used for downstream analysis. Taxonomy was assigned to bacterial ASVs using the SILVA database (release 138.1, [55]) and to fungal ASVs using the UNITE database (version 8.2, [42] ) with a confidence threshold of 0.70.

All statistical analyses were performed using the R software environment (version 4.2.0). For all statistical analyses, the individual lamb was defined as the experimental unit, with six independent replicates (n = 6) evaluated per dietary treatment.

Growth performance, rumen fermentation parameters, and enzymatic activities were analyzed using linear mixed-effects models (LMM) with the lmer function from the lme4 package (version 1.1–34) in R (version 4.3.1) []. To strictly align our statistical analysis with the physical randomization of the experiment (RCBD) and preserve nominal Type I error rates, both ‘block’ (initial body weight) and ‘experimental batch’ were incorporated as random intercepts in the final models. The statistical model was specified as follows:

Where Y_ij represents the response variable; µ is the overall mean; T_i is the fixed effect of the Przewalskia tangutica supplementation ( i = control, treatment levels); Batch_j is the random effect of the experimental batch (j = 1, 2, 3, 4), Block_k is the random effect of the initial body weight block (k = 1, 2, 3, 4, 5, 6), and is the residual error.

Relative abundance of rumen bacterial and fungal taxa was analyzed using generalized linear mixed-effects models (GLMM) via the glmmTMB package (version 1.1.7). Treatment was defined as a fixed effect, while block and batch were included as random intercepts. For microbiome data exhibiting excessive zeros, we fitted these data using Zero-Inflated Beta GLMMs with a logit link function (via family = beta_family (link = “logit”) and ziformula = ~ 1 in the glmmTMB package). This mathematically sound approach successfully models both the continuous proportions and the zero-inflation process without requiring artificial pseudo-count offsets. If a model exhibited a singular fit due to a negligible variance component of a random effect, the effect was still retained in the model to preserve the integrity of the original physical design. For compositional data (e.g., VFA molar percentages, which were strictly bounded within the open interval (0, 1) with no absolute 0 or 1 boundary values), a logit transformation was applied prior to LMM analysis to satisfy the assumption of residual normality. Degrees of freedom for fixed effects were estimated using the Satterthwaite approximation. Pairwise comparisons of treatment means were performed using the emmeans package (version 1.8.7) with Tukey’s adjusted P-values. Statistical significance was declared at P < 0.05. Alpha diversity indices were analyzed using LMMs. Beta diversity was visualized via Principal Coordinate Analysis (PCoA) based on Bray-Curtis distances, with significance tested using PERMANOVA (999 permutations). Differential abundance analysis for biomarker discovery was performed using LEfSe (Linear Discriminant Analysis Effect Size). To control the False Discovery Rate (FDR) in this high-dimensional dataset, p-values from the Kruskal-Wallis test were adjusted using the Benjamini-Hochberg (BH) procedure. Significant biomarkers were defined by an adjusted P < 0.05 and a logarithmic LDA score > 2.0. Correlations between microbial genera (top 20 abundant) and phenotypic traits were assessed using Spearman’s rank correlation. Co-occurrence networks were constructed using genera present in > 50% of samples. To ensure statistical robustness, edges were defined only for strong correlations () that remained significant after FDR correction (BH-adjusted P < 0.05).

Functional potential was inferred from taxonomic data. For bacteria, Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt2) [12] was used to predict Kyoto Encyclopedia of Genes and Genomes (KEGG) orthologs from ASV representative sequences [58]. It is important to note that the KEGG database includes pathways from all domains of life. The prediction of pathways under categories such as ‘Human Diseases’ or ‘Organismal Systems’ reflects the presence of microbial gene families that are components of these broad KEGG modules and should not be interpreted as host-derived functions. Our interpretation focuses on differential abundance in core microbial metabolic pathways. For fungi, ecological guilds (e.g., pathotroph, saprotroph, symbiotroph) were assigned using FUNGuild (v1.0) [41]. The FUNGuild assignments include confidence rankings of “Possible”, “Probable”, and “Highly Probable”. For downstream statistical analysis and visualization, only assignments with “Probable” and “Highly Probable” confidence levels were retained. Differentially abundant functional pathways (PICRUSt2) and guilds (FUNGuild) were tested using GLMMs with FDR adjustment for multiple guild comparisons.

Results

Effect of the addition of Przewalskia tangutica on the growth performance of sheep

Throughout this 6-day sampling period, DMI did not differ significantly among treatment groups (P = 0.895, Table 2), confirming that differences in basal nutrient intake did not confound the results. Average daily gain (ADG) differed significantly among groups (P < 0.001). The HP group exhibited the greatest gain (139.94 g/day, 95% CI: 119.39–151.08), followed by MP (111.28 g/day, 95% CI: 96.11–126.45) and LP (106.33 g/day, 95% CI: 91.16–121.50), all exceeding CON (78.72 g/day, 95% CI: 63.55–93.89), post-hoc Tukey’s HSD test revealed that HP and MP groups were significantly higher than CON (P < 0.05). Feed conversion ratio (FCR) also showed a significant overall treatment effect (P = 0.002). The HP group demonstrated the most efficient conversion (8.54, 95% CI: 6.52–10.87), followed by MP (10.05, 95% CI: 8.03–12.06) and LP (10.25, 95% CI: 8.23–12.27), all superior to CON (14.83, 95% CI: 12.82–16.85), Tukey’s HSD indicated that CON differed significantly from each supplemented group (P < 0.05), whereas LP, MP, and HP did not differ among themselves (Table 2).

Table 2.

Effects of Przewalskia tangutica on the growth performance of sheep

Item	Group1	EMM2	95% Cl3	SEM4	P value5	Cohen’s d6
DMI, g/d	CON	1111.3	988.54-1234.05	83.22	0.895
	LP	1095.12	972.36-1217.87
	MP	1113.06	990.31-1235.82
	HP	1156.54	1033.78-1279.29
ADG, g/d	CON	78.72c	63.55–93.89	10.62	< 0.001
	LP	106.33bc	91.164–1221.50			-2.72
	MP	111.28ab	96.108–126.45			-3.21
	HP	139.94a	119.39-151.08			-5.6
FCR	CON	14.83a	12.82–16.85	4.58	0.002
	LP	10.25b	8.23–12.27			2.14
	MP	10.05b	8.03–12.06			2.24
	HP	8.54b	6.52–10.87			0.00

Note: ¹ HP, high-dose group (4.5% Przewalskia tangutica); MP, medium-dose (3%); LP, low-dose (1.5%); CON, control (0%); ² EMM, estimated marginal mean from mixed-effects model; ³ 95% CI, 95% confidence interval of EMM; ⁴ SEM, standard error of the mean from model residuals; ⁵ P-value for overall treatment effect (Type III ANOVA); ⁶ Cohen’s d effect size for HP vs. CON comparison; magnitude: |d| < 0.2 negligible, 0.2–0.5 small, 0.5–0.8 medium, > 0.8 large; DMI, dry matter intake; ADG, average daily gain; FCR, feed conversion ratio; a, b and c Means within a row with different superscripts differ significantly (P < 0.05) based on Tukey’s HSD test

Effect of the addition of Przewalskia tangutica on digestive enzyme activities in sheep

Supplementation significantly affected AMS activity (P = 0.035). The LP (159.73 U/L, 95% CI: 122.90–196.56) and HP (157.32 U/L, 95% CI: 120.49–194.15) groups exhibited elevated activities compared with CON (91.44 U/L, 95% CI: 54.62–128.27; P < 0.05), whereas the MP group (107.91 U/L, 95% CI: 71.09–144.74) was intermediate and not statistically separable. Cellulase activity also differed significantly (P = 0.012); the HP group showed the highest activity (13.20 IU/L, 95% CI: 9.30–17.11), which exceeded CON (3.22 IU/L, 95% CI: 0.69–7.12) and MP (4.79 IU/L, 95% CI: 0.88–8.69; P < 0.05), while LP (8.99 IU/L, 95% CI: 5.08–12.89) was intermediate. LIP was significantly increased by supplementation (P = 0.001). All treatment groups, LP (46.50 U/L, 95% CI: 32.77–60.23), MP (45.71 U/L, 95% CI: 8.02–19.43), and HP (55.26 U/L, 95% CI: 41.54–68.99) exhibited greater activities than CON (22.36 U/L, 95% CI: 8.63–36.09; P < 0.05). AP activity was significantly reduced (P = 0.003); the MP (188.99 U/L, 95% CI: 161.71–216.27) and HP (176.85 U/L, 95% CI: 149.57–204.14) groups exhibited lower activities than CON (263.79 U/L, 95% CI: 236.51–291.07; P < 0.05). PDH activity was significantly modulated (P = 0.019). The MP group showed reduced activity (98.95 U/L, 95% CI: 87.69–110.21) compared with CON (118.51 U/L, 95% CI: 107.27–129.77; P < 0.05), whereas LP and HP remained similar to CON. LDH was significantly decreased in all supplemented groups (P = 0.035); LP (28.99 U/L, 95% CI: 20.37–37.62), MP (30.62 U/L, 95% CI: 21.99–39.24), and HP (27.56 U/L, 95% CI: 18.94–36.19) were lower than CON (45.11 U/L, 95% CI: 36.48–53.73; P < 0.05), with a large effect size (Cohen’s d = 3.83) (Table 3).

Table 3.

Effects of Przewalskia tangutica on digestive enzyme activities in sheep

Item	Group1	EMM2	95% Cl3	SEM4	P value5	Cohen’s d6
AMS (U / L)	CON	91.44b	54.62-128.27	382.55	0.035
	LP	159.73a	122.90-196.56			-3.02
	MP	107.91ab	71.09-144.74			-0.73
	HP	157.32a	120.49-194.15			-2.92
Pepsin (U / L)	CON	88.8	62.83-114.78	190.34	0.064
	LP	129.2	103.23-155.18
	MP	126.77	100.79-152.74
	HP	136.91	110.94-162.89
Cellulase (IU / L)	CON	3.22b	0.69–7.12	4.30	0.012
	LP	8.99ab	5.08–12.89			-2.41
	MP	4.79b	0.88–8.69			-0.66
	HP	13.20a	9.30-17.11			-4.17
LIP (U / L)	CON	22.36b	8.63–36.09	53.15	0.001
	LP	46.50a	32.77–60.23			-0.00
	MP	45.71a	8.02–19.43			-1.98
	HP	55.26a	41.54–68.99			-3.91
PDH (U / L)	CON	118.51a	107.27-129.77	35.75	0.019
	LP	126.69a	115.43-137.94			-1.18
	MP	98.95b	87.69-110.21			-2.83
	HP	120.24a	108.99-131.51			-0.29
CMC (IU / L)	CON	200.5	156.95-244.06	535.14	0.493
	LP	236.44	192.88-279.99
	MP	233.08	189.52-276.63
	HP	237.44	193.89–281
Pectinase (U / L)	CON	60.64	49.08–72.21	37.71	0.151
	LP	56.01	44.45–67.58
	MP	42.62	31.06–54.19
	HP	52.23	40.67–63.79
BCAAD (U / L)	CON	28.12	22.54–33.70	8.79	0.112
	LP	25.85	20.27–31.43
	MP	19	13.42–24.58
	HP	22.22	16.64–27.81
Urease (IU / L)	CON	841.25	645.55-1036.95	120.02	0.404
	LP	696.94	501.24-892.64
	MP	629.03	433.34-824.73
	HP	704.22	508.52-899.92
Aps (U / L)	CON	263.79a	236.51-291.07	16.74	0.003
	LP	237.09a	209.81-264.38			1.60
	MP	188.99b	161.71-216.27			4.47
	HP	176.85b	149.57-204.14			5.20
SDH (IU / L)	CON	1310.61	1131.96-1489.27	109.57	0.903
	LP	1363.78	1185.13-1542.44
	MP	1381.91	1203.26-1560.56
	HP	1323.91	1145.25-1502.56
Cellobiase (IU / L)	CON	39.01	25.86–52.18	8.07	0.300
	LP	24.5	11.34–37.66
	MP	25.63	12.47–38.79
	HP	33.5	20.34–46.66
EG (IU / L)	CON	85.89	69.78-101.99	9.88	0.229
	LP	86.94	70.83-103.04
	MP	66.85	50.75–82.96
	HP	82.29	66.19–98.40
MDH (IU / L)	CON	156.12	128.82-183.42	16.74	0.863
	LP	156.12	128.82-183.42
	MP	144.45	117.15-171.75
	HP	148.27	120.97-175.57
Xylanase (IU / L)	CON	24.77	18.39–31.15	11.49	0.474
	LP	28.83	22.44–35.21
	MP	25.46	19.08–31.84
	HP	30.3	23.92–36.68
LDH (IU / L)	CON	45.11a	36.48–53.73	5.29	0.035
	LP	28.99b	20.37–37.62			3.52
	MP	30.62b	21.99–39.24			3.16
	HP	27.56b	18.94–36.19			3.83
Chymosin (IU / L)	CON	630.08	505.71-754.45	76.27	0.394
	LP	512.15	387.78-636.52
	MP	634.55	510.19-758.92
	HP	608.68	484.31-733.05

Note: ¹ HP, high-dose group (4.5% Przewalskia tangutica); MP, medium-dose (3%); LP, low-dose (1.5%); CON, control (0%); ² EMM, estimated marginal mean from mixed-effects model; ³ 95% CI, 95% confidence interval of EMM; ⁴ SEM, standard error of the mean from model residuals; ⁵ P-value for overall treatment effect (Type III ANOVA); ⁶ Cohen’s d effect size for HP vs. CON comparison; magnitude: |d| < 0.2 negligible, 0.2–0.5 small, 0.5–0.8 medium, > 0.8 large; AMS, amylase concentration; LIP, lipase; PDH, pyruvate dehydrogenase; CMC, carboxymethyl cellulase; BCAAD, branched-chain amino acid dehydrogenase; Aps, aminopeptidase; SDH, succinate dehydrogenase; EG, endoglucanase; MDH, malate dehydrogenase; LDH, lactate dehydrogenase; a and b Means within a row with different superscripts differ significantly (P < 0.05) based on Tukey’s HSD test

Effect of the addition of Przewalskia tangutica on rumen fermentation parameters in sheep

Rumen pH did not differ among treatments, ranging from 7.14 to 7.22 across groups (P = 0.757). In contrast, total volatile fatty acid (VFA) concentration exhibited a significant dose-dependent increase (P < 0.001, SEM < 0.001). The HP group demonstrated the highest concentration (1.46 mmol/L, 95% CI: 1.42–1.49), followed by MP (1.39 mmol/L, 95% CI: 1.36–1.42) and LP (1.34 mmol/L, 95% CI: 1.31–1.37), all exceeding CON (1.26 mmol/L, 95% CI: 1.24–1.30). Acetate proportion increased significantly with supplementation (P < 0.001, SEM < 0.001). The HP group reached 70.72% (95% CI: 70.62–70.85) compared with 70.32% (95% CI: 70.22–70.42) for CON, with a Cohen’s d of − 5.81. Propionate proportion remained stable across treatments (17.35–17.58%, P = 0.880, SEM = 0.094). Butyrate proportion was significantly elevated in a dose-responsive manner (P < 0.001, SEM = 5.14 × 10⁻⁵), increasing from 7.57% (95% CI: 7.56–7.58) in CON to 7.68% (95% CI: 7.67–7.69) in HP (Cohen’s d = − 14.71). Similarly, isobutyrate proportion increased significantly (P < 0.001, SEM < 0.001), with HP at 1.68% (95% CI: 1.67–1.68) versus CON at 1.64% (95% CI: 1.63–1.64; Cohen’s d = − 14.33). Isovalerate proportion did not differ among groups (P = 0.839, SEM = 0.004). Valerate proportion showed a numerical decrease with supplementation (CON: 1.40%; HP: 0.61%) but was not statistically significant (P = 0.062, SEM = 0.0814). The acetate-to-propionate ratio was unaffected by treatment (P = 0.944, SEM = 0.005) (Table 4).

Table 4.

Effects of Przewalskia tangutica on rumen fermentation parameters in sheep

Item	Group1	EMM2	95% Cl3	SEM4	P value5	Cohen’s d6
pH	CON	7.17	7.048–7.292	0.006	0.757
	LP	7.14	7.009–7.273
	MP	7.22	7.107–7.350
	HP	7.16	7.040–7.283
Total VFA, mmol/L	CON	1.26c	1.24–1.30	0.0004	< 0.001
	LP	1.34b	1.31–1.37			-3.52
	MP	1.39ab	1.36–1.42			-5.62
	HP	1.46a	1.42–1.49			-8.55
Acetate%	CON	70.32c	70.22–70.42	0.0003	< 0.001
	LP	70.51ab	70.41–70.61			-2.82
	MP	70.61ab	70.51–70.71			-4.28
	HP	70.72a	70.62–70.85			-5.81
Propionate%	CON	17.35	16.89–17.82	0.094	0.880
	LP	17.51	17.04–17.97
	MP	17.57	17.10-18.03
	HP	17.58	17.12–18.05
Isobutyrate%	CON	1.64c	1.63–1.64	0.0002	< 0.001
	LP	1.65b	1.65–1.66			-6.00
	MP	1.67a	1.67–1.68			-12.33
	HP	1.68a	1.67–1.68			-14.33
Butyrate%	CON	7.57d	7.56–7.58	5.14E-05	< 0.001
	LP	7.62c	7.61–7.63			-6.71
	MP	7.65b	7.64–7.66			-11.43
	HP	7.68a	7.67–7.69			-14.71
Isovalerate%	CON	1.72	1.62–1.81	0.004	0.839
	LP	1.72	1.62–1.82
	MP	1.77	1.67–1.87
	HP	1.73	1.63–1.82
Valerate%	CON	1.4	0.98–1.84	0.081	0.062
	LP	0.99	0.56–1.42
	MP	0.73	0.3–1.16
	HP	0.61	0.18–1.05
Acetate/Propionate	CON	4.06	3.95–4.17	0.005	0.944
	LP	4.03	3.92–4.14
	MP	4.02	3.91–4.13
	HP	4.02	3.92–4.13

Note: ¹ HP, high-dose group (4.5% Przewalskia tangutica); MP, medium-dose (3%); LP, low-dose (1.5%); CON, control (0%); ² EMM, estimated marginal mean from mixed-effects model; ³ 95% CI, 95% confidence interval of EMM; ⁴ SEM, standard error of the mean from model residuals; ⁵ P-value for overall treatment effect (Type III ANOVA); ⁶ Cohen’s d effect size for HP vs. CON comparison; magnitude: |d| < 0.2 negligible, 0.2–0.5 small, 0.5–0.8 medium, > 0.8 large; a, b and c Means within a row with different superscripts differ significantly (P < 0.05) based on Tukey’s HSD test

Effects of Przewalskia tangutica on microbial diversity and abundance in the rumen of sheep

ASV distribution analysis (Fig. 1A) revealed 550 shared bacterial and 199 shared fungal ASVs across groups. The CON group presented the most unique bacterial ASVs (3,734), whereas the HP group presented the most unique fungal ASVs (2,065). The alpha diversity indices did not significantly differ among the bacterial communities (Fig. 1B). However, for fungal communities, the HP group presented significantly higher ACE and Chao1 indices than the LP group (P < 0.001) and a greater Shannon index than did the MP group (P < 0.05). Beta diversity analysis (PCoA, Fig. 1C) revealed limited structural shifts among the bacterial communities (partial overlap; PC1 + PC2 explained 60.03% of the variance). In contrast, fungal communities displayed greater intergroup dispersion and clear separation in the HP group (PC1 + PC2 explained 59.07% of the variance).

Fig. 1.

Effects of Przewalskia tangutica supplementation on rumen microbial diversity. (A) Venn diagrams showing the number of shared and unique bacterial and fungal amplicon sequence variants (ASVs) among the control (CON), low-dose (LP), medium-dose (MP), and high-dose (HP) groups. (B) Alpha diversity indices (Chao1, ACE, Shannon, Simpson) for rumen bacterial and fungal communities. Asterisks denote significant differences (*P < 0.05, **P < 0.001). (C) Principal coordinate analysis (PCoA) based on Bray-Curtis dissimilarity for bacterial and fungal communities. Each point represents the microbial community of an individual sample. The percentage of variation explained by each principal coordinate is indicated on the axes. Ellipses represent 95% confidence intervals for each treatment group. The overall effect of treatment was significant (PERMANOVA, R² = 0.21, P = 0.001)

Effect of Przewalskia tangutica on the structure of microbial communities in the rumen of sheep

16S rRNA and ITS sequencing revealed significant impacts of Przewalskia tangutica supplementation on rumen microbial communities (Fig. 2). The bacterial communities were dominated by Firmicutes (47.13%) and Bacteroidetes (40.63%), LP increased Fibrobacterota; MP increased Verrucomicrobiota and Cyanobacteria; and HP increased Proteobacteria and Spirochaetota (all P < 0.05). Both LP and MP reduced the Actinobacteria (P = 0.006). Prevotella dominated at the genus level, with the treatment groups showing reduced NK4A214_group, Christensenellaceae_R-7_group, and related taxa (P < 0.05). LP increased Selenomonas, whereas LP/MP increased Succiniclasticum and unclassified_Selenomonadaceae but reduced Saccharofermentans (P < 0.05). Phylogenetic analysis revealed distinct community structures, with CON enriched in 30 clades (e.g., Rikenellaceae_RC9_gut_group), LP enriched in 20 clades (e.g., Succiniclasticum), MP enriched in 8 clades (uncultured_rumen_bacterium), and HP enriched in 6 clades (e.g., Treponema).

Fig. 2.

Effects of Przewalskia tangutica on differences in the species distributions and community structures of sheep rumen microorganisms. A, B, Histograms of species distributions at the bacterial and fungal phylum (A) and genus (B) levels. C, Linear discriminant analysis effect size (LEfSe) cladograms for different groups of bacterial and fungal community structures

Fungal communities did not differ at the phylum level (Ascomycota, Basidiomycota, and Neocallimastigomycota were dominant). At the genus level, MP reduced Cladosporium, whereas HP increased unclassified_Ascomycota/Basidiomycota (P < 0.05). Phylogenetic analysis revealed that CON was enriched in 8 clades (e.g., Cladosporium), LP was enriched in 9 clades (e.g., Piromyces), and HP was enriched in 14 clades (e.g., Trichosporon).

Microbial co-occurrence networks (Fig. 3) constructed using strong and significant correlations (|ρ| > 0.6, FDR-adjusted P < 0.05) reveal variations in network topology. The bacterial networks in CON and LP had 23 nodes and 231 edges, whereas those in MP and HP had 22 nodes and 210 edges. The CON bacterial network was dominated by positive associations, whereas LP, MP, and HP presented a more even distribution. The fungal networks in CON and LP had 23 nodes and 231 edges, whereas those in MP and HP had 22 nodes and 231 edges. Negative associations were more frequent in CON and MP, positive associations predominated in LP, and HP showed balanced interactions.

Fig. 3.

Effect of Przewalskia tangutica on the visualization of correlation networks for the structure of bacterial and fungal communities in the rumens of different groups of sheep

Correlation analysis of bacterial and fungal genera, rumen volatile fatty acids, and digestive enzyme activities

Comprehensive Spearman correlation analyses were performed to explore the relationships between rumen microbiota, fermentation parameters, and digestive functions (Fig. 4A-F), the full matrices of significant correlations (P < 0.05) are provided in Supplementary Table S1-S6.

Fig. 4.

Effects of Przewalskia tangutica on correlations between indicators. A, Genus-level (top 20 most abundant) correlations between the bacterial and fungal communities. B, Correlations between bacteria at the genus level (top 20 most abundant genera) and rumen fermentation parameters. C, Genus-level (top 20 most abundant) correlations between bacterial and enzymatic activities. D, Genus-level (top 20 most abundant) correlations between fungal and enzyme activities. E, Correlations between fungi at the genus level (top 20 most abundant genera) and rumen fermentation parameters. F, Correlations between rumen fermentation parameters and enzyme activities

Key patterns emerged from these analyses. Several bacterial genera known for fiber degradation (e.g., Uncultured_rumen_bacterium, unclassified_Selenomonadaceae) showed positive correlations with major cellulase and xylanase activities. In contrast, some taxa (e.g., Rikenellaceae_RC9_gut_group, UCG_002) were negatively associated with total VFA concentrations but positively correlated with specific enzymes like pectinase and cellobiase. Fungal genera also display distinct association patterns. For instance, the relative abundance of Cladosporium was negatively correlated with total VFA, whereas Mortierella showed a positive correlation. Several fungal taxa (e.g., Podospora, unclassified_Basidiomycota) were positively associated with the activities of proteolytic and ammonia-producing enzymes (e.g., Aps, cellobiase). Furthermore, significant positive correlations were observed between total VFA concentration and the activities of key enzymes involved in protein metabolism (e.g., pepsin, BCAAD, urease) and energy metabolism (LDH).

Predicted metagenomic and fungal guild profiles

KEGG pathway enrichment (Fig. 5) analysis predicted differential representation of functional categories across treatment groups (HP, MP, LP, and CON). Pathways associated with organismal systems (thermogenesis, endocrine signaling), human diseases (cancer, antimicrobial resistance, and infectious diseases), and various metabolic processes (lipid, glycan, amino acid, terpenoid, xenobiotic, and secondary metabolite metabolism) exhibited varying degrees of enrichment. The metabolic pathways of 2-oxocarboxylic acid metabolism, polyketide sugar unit biosynthesis, lysine biosynthesis, tryptophan metabolism, peptidoglycan biosynthesis, and thermogenesis were enriched in the LP and MP groups. In contrast, the pathways related to acarbose and validamycin biosynthesis, streptomycin biosynthesis, drug metabolism involving other enzymes, geraniol degradation, limonene and pinene degradation, lysine degradation, synthesis and degradation of ketone bodies, glycerophospholipid metabolism, pertussis, vancomycin resistance, microRNAs in cancer, the thyroid hormone signaling pathway, and the adipocytokine signaling pathway were enriched in the CON group.

Fig. 5.

Functional predictions for the rumen microbiota with significantly different KEGG pathways (GLMM, FDR-adjusted P < 0.05) for the different treatment groups

We employed FUNGuild to predict fungal guilds on the basis of taxonomic assignments. The analysis revealed substantial variations in the relative abundances of major functional guilds among the CON, LP, MP, and HP groups, as illustrated in Fig. 6A. Statistical comparisons revealed significant shifts in guild composition associated with the different treatments (Fig. 6B-F; Welch’s t test, 95% confidence intervals). Specifically, the relative abundances of the ectomycorrhizal, plant pathogen, and wood saprotroph guilds were significantly lower in the LP, MP, and HP groups than in the CON group (Fig. 6B, C, D). Further pairwise comparisons among the treatment groups revealed additional significant differences: compared with the CON group, the LP and MP groups presented a greater proportion of animal endosymbionts (Fig. 6B, C), whereas the HP group was notably enriched in soil saprotrophs compared with the other groups (Fig. 6D, E, F) and presented an increased abundance of animal parasites compared with the LP group (Fig. 6F).

Fig. 6.

Predicted functional profiles of rumen fungal communities. (A) Relative abundance of major fungal functional guilds predicted by FUNGuild across treatment groups. (B-F) Bar plots showing statistically significant differences in the relative abundance of specific guilds between treatment groups. Data was analyzed using Generalized Linear Mixed Models (GLMMs) followed by Tukey’s post-hoc comparisons. Significance was declared at Benjamini-Hochberg adjusted P < 0.05. The comparisons shown are: (B) CON vs. LP; (C) CON vs. MP; (D) CON vs. HP; (E) MP vs. HP; and (F) LP vs. HP

Discussion

A primary constraint on livestock productivity in saline–alkali grazing systems is the inherently low digestibility of available lignocellulosic biomass. In the current investigation, dietary supplementation with Przewalskia tangutica did not significantly alter the DMI of lambs (P > 0.05). The observed mean DMI of approximately 1.1 kg/d (representing 3.7% of body weight) aligns with the established nutritional requirements and intake expectations for growing sheep of this physiological stage [44, 77]. This consistency confirms that the inclusion of Przewalskia tangutica did not adversely affect diet palatability or voluntary appetite. Notably, our results demonstrate a significant, dose-dependent improvement in ADG and a concomitant reduction in the FCR in response to Przewalskia tangutica supplementation (Table 2). The observation that these substantial growth performance gains occurred independently of changes in DMI is a critical finding, strongly suggesting that Przewalskia tangutica functions primarily by enhancing feed utilization efficiency rather than stimulating intake. Several interconnected mechanisms likely underpin this enhanced efficiency. First, the supplementation significantly upregulated the activities of key digestive enzymes—including amylase, lipase, and protease (Table 3), alongside increased total VFA production (Table 4). Such enzymatic enhancement suggests a more robust degradation of dietary nutrients, thereby increasing the availability of energy and metabolic substrates essential for protein synthesis. Specifically, the elevated concentrations of total VFAs, acetate, and butyrate provide the primary energy substrates for ruminants [22]. This metabolic shift was strongly correlated with increased cellulase and amylase activities (Fig. 4F, Table S6), indicating that Przewalskia tangutica improves the efficiency of microbial fermentation rather than merely altering the fermentation pathways. Second, the bioactive profile of Przewalskia tangutica (Table 1) likely modulated the rumen microbial ecosystem toward a more efficient state. The presence of tropane alkaloids and flavonoids may have created a favorable microenvironment for the proliferation of fibrolytic and hyper-ammonia-producing microorganisms [36], thereby improving the digestibility of fiber fractions (NDF and ADF). Furthermore, specific bioactives such as quercetin are recognized for their potent antioxidant and anti-inflammatory properties [74]. By mitigating subclinical gastrointestinal inflammation common in saline-stressed environments, these compounds may facilitate the partitioning of nutrients toward growth and tissue deposition rather than being diverted to maintain immune responses. Finally, it is plausible that absorbed alkaloids exert post-absorptive effects on metabolism, potentially shifting nutrient partitioning to favor lean tissue accretion. While the current study highlights the role of Przewalskia tangutica in promoting fermentation efficiency and enzymatic activity, further research utilizing metagenomic validation and metabolic indicators is warranted to fully elucidate the direct interactions between specific plant bioactives and host physiological responses.

The rumen, renowned as a natural fermentation vat, harbors a complex and diverse microbial ecosystem essential for fiber degradation and maintaining host homeostasis through intricate interactions, including competition, mutualism, and antagonism [19, 21]. The host’s diet is a primary factor shaping the structure and function of the rumen microbiome [64, 71]. A defining characteristic of a healthy rumen microbiome is resilience. Our 16S rRNA analysis showed that Przewalskia tangutica did not disrupt the stability of dominant phyla or general alpha diversity. Consistent with previous studies [13, 75, 76], Firmicutes and Bacteroidota were the dominant bacterial phyla in the sheep rumen and play pivotal roles in digestion and nutrient metabolism. Instead, it exerted a nuanced, “precision-editing” effect on the microbiome. Specifically, we observed a selective enrichment of Fibrobacterota and Verrucomicrobiota—taxa specializing in the degradation of complex carbohydrates [61, 62], nitrogen cycling [66, 80], and fiber fermentation yielding volatile fatty acids (VFAs) [56], alongside a paradoxical reduction in several non-dominant genera [79]. The reduction of these non-dominant taxa, despite the increase in overall fermentation efficiency, implies a mechanism of “functional pruning.” In the control group (grazing saline forage), the rumen may harbor metabolically redundant or inefficient “generalist” bacteria that compete for substrates without contributing significantly to VFA pools [57]. Przewalskia tangutica appears to suppress these inefficient groups, reducing ecological noise and resource competition. This interpretation is supported by the co-occurrence network analysis, which shifted from predominantly positive (cooperative/dependent) interactions in the control group to a balanced positive–negative topology in treated groups (Fig. 3). Such a network structure often indicates increased functional specialization and stability [14], allowing specialized degraders like Fibrobacter to dominate the functional niche of fiber hydrolysis.

In contrast to the bacterial community, the fungal community exhibited a significant response in alpha diversity to Przewalskia tangutica supplementation (Fig. 1B). The dominant fungal phyla were Ascomycota, Basidiomycota, and Neocallimastigomycota, consistent with findings in other herbivores on the Qinghai-Tibetan Plateau [59]. One of the most novel findings of this study is the significant remodeling of the fungal community, specifically the suppression of Cladosporium. While fungi represent a smaller portion of the rumen biomass than bacteria, they play a disproportionately large role in the initial physical disruption of lignocellulose [16]. Cladosporium is a halotolerant mold often associated with environmental stress [7]. We hypothesize that in sheep grazing saline–alkali pastures, the high salt load favors the overgrowth of stress-tolerant Cladosporium at the expense of obligate anaerobic fibrolytic fungi [16]. The negative correlation observed between Cladosporium abundance and VFA concentration suggests that this genus acts as a “stress-opportunist” that contributes little to fermentation. The administration of Przewalskia tangutica significantly reduced Cladosporium abundance while concurrently increasing Fibrobacterota and enhancing fungal network complexity (Fig. 3). This suggests an alleviation of competitive exclusion: by suppressing the opportunistic fungal overgrowth, the additive restores the niche for efficient fiber-degrading consortia [38]. To our knowledge, this is one of the first studies to link the suppression of halotolerant rumen fungi to improved fiber fermentation in saline environments.

The observed ecological shifts are likely driven by the synergistic action of tropane alkaloids and flavonoids present in Przewalskia tangutica. Although this study did not directly measure alkaloid concentrations in the rumen or their physiological effects, the known bioactivities of these compounds offer a plausible, albeit speculative, mechanistic framework for our observations. Unlike tannins, which primarily act by binding proteins, or essential oils, which have broad-spectrum antimicrobial effects [50], tropane alkaloids (e.g., scopolamine, anisodamine) possess well-documented anticholinergic activity [70]. In the context of the rumen, we speculate that these compounds may subtly modulate rumen motility [53]. A moderate reduction in passage rate would increase the retention time of fibrous digesta, allowing for more complete microbial colonization and degradation, thereby explaining the higher digestibility and VFA yield [78]. Second, flavonoids such as quercetin exert selective antimicrobial pressure [4, 54]. Unlike broad-spectrum antibiotics, flavonoids target bacterial cell membranes and enzyme systems differentially [69], thereby modulating VFA profiles [51, 54, 65]. They have been shown to inhibit opportunistic pathogens and methanogens while sparing or promoting cellulolytic bacteria [48]. Furthermore, the antioxidant properties of flavonoids may mitigate the oxidative stress induced by saline–alkali diets, protecting the oxygen-sensitive anaerobic fungi and bacteria, thus preserving their metabolic function, the antioxidant and anti-inflammatory properties of quercetin may improve the ruminal environment for anaerobic microbes by mitigating oxidative stress, thus protecting metabolic functions and indirectly promoting VFA production [4, 45].

While 16S/ITS sequencing provides a robust taxonomic framework, it offers only predictive functional insights. The functional pathways identified here (e.g., animal endosymbiont enrichment) require validation via metagenomics or metatranscriptomics to confirm specific gene expression related to carbohydrate-active enzymes (CAZymes). Although our findings clearly demonstrate microbial interactions with phenolic compounds, the absence of direct measurements of key production parameters—such as in vivo feed digestibility and methane emission potential—limits the immediate applicability of these findings. Additionally, direct measurement of rumen motility and in vivo alkaloid pharmacokinetics would further elucidate the host-microbe interaction mechanism.

Conclusions

In summary, this study demonstrates that Przewalskia tangutica supplementation significantly enhances growth performance (increased ADG, improved FCR) in sheep grazing saline-alkali sown pastures, an effect driven by a targeted modulation of the rumen environment that boosts fermentation and digestive enzyme activities. These benefits appear driven by targeted microbial modulations specifically the enrichment of fibrolytic Fibrobacterota and the suppression of halotolerant Cladosporium, rather than broad ecosystem restructuring. However, limitations exist. The functional insights reported are based on predictive profiling and require metagenomic validation. Additionally, while we hypothesize that plant bioactives (alkaloids/flavonoids) mediate these effects, physiological parameters such as rumen motility were not directly measured. Despite these constraints, Przewalskia tangutica shows promise as a functional additive for improving livestock efficiency in challenging saline environments.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

VFA

Volatile fatty acids

ADG

Average daily gain

FCR

Feed conversion ratio

TP

Total protein

ALB

Albumin

GLB

Globulin

GLU

Glucose

UA

Uric acid

BUN

Blood urea nitrogen

TG

Triglyceride

SOD

Superoxide dismutase

MDA

Malondialdehyde

GSH-PX

Glutathione peroxidase

AMS

Amylase concentration

LIP

Lipase

PDH

Pyruvate dehydrogenase

CMC

Carboxymethyl cellulase

BCAAD

Branched-chain amino acid dehydrogenase

Aps

Aminopeptidase

SDH

Succinate dehydrogenase

EG

Endoglucanase

MDH

Malate dehydrogenase

LDH

Lactate dehydrogenase

Author contributions

Conceptualization: F.H. and X.Z.; data curation: F.H.; formal analysis: X.Z.; funding acquisition: F.H.; methodology: F.H. and X.Z.; Software: X.Z.; supervision: F.H.; writing—Original draft: X.Z. and F.H.; writing—Review and editing: X.Z., F.H., W.D., and W.Z. All the authors have read and agreed to the published version of the manuscript.

Funding

This study was supported by the Key Research and Development Program of Gansu Province (23ZDWA002), National Key Research and Development Program (2025YFE0212000), Innovation Platform Plan Program of Gansu Province (26JDWA001), Program for Innovative Research Team of Ministry of Education (IRT_17R50).

Data availability

The datasets presented in this study can be found in online repositories. The name of the repository and accession number can be found below: NCBI; PRJNA1264498. Submission ID: SUB15326431 and BioProject ID: PRJNA1264498.

Declarations

Ethics approval and consent to participate

The experiments were conducted according to the guidelines of experimental field management protocols (file No: 2010–1 and 2010–2) and were approved by the Animal Use and Care Committee of Lanzhou University.

Consent for publication

All authors have read and agreed to publication.

Competing interests

The authors declare no competing interests.