This experiment was conducted at the experimental base of the Institute of Animal Husbandry, Heilongjiang Academy of Agricultural Sciences (45°42′N, 126°38′E). All animal-based experiments were conducted in accordance with the principles and responsibilities outlined in the Northeast Normal University’s (Changchun, China) guidelines for animal research. Twelve male small-tailed Han sheep of similar weight (30.49 ± 4.31 kg, mean ± SD) were selected for this study. They were each assigned to 1 of 3 dietary treatments which represent different degrees of degraded grasslands classified by the biomass of grassland plants (Areas with 40–60% vegetation cover were deemed as moderately degraded, and those with less than 40% vegetation cover were regarded as heavily degraded) (25): (1) undegraded grassland plants (UG); (2) moderately degraded grassland plants (MG); or (3) severely degraded grassland plants (SG). Plant samples from grasslands degraded to different levels were obtained from the Songnen grassland in Jilin, China (44°45′N, 123°45′E) and plants were cut and baled for feeding to sheep. The grasslands in the area have already shown saline-alkaline degradation to different levels as a result of climate change and overgrazing. The climate of the area is continental with average annual temperatures ranging from 2.4°C to 2.7°C and precipitation ranging from 300 mm to 500 mm, most of which occurs between June and August (data from Changling County Climate Station, Jilin Province).

We established three 400-m parallel transects at 50-m intervals in grasslands with different levels of degradation (UG: undegraded grassland plants, MG: moderately degraded grassland plants, and SG: severely degraded grassland plants). Next, we determined the locations of 100 cm × 100 cm quadrants every 40-m along each transect. We used these quadrants to measure the composition and proportion of plants, and then plant samples were harvested at a cutting height of approximately 10 cm. The diets of the sheep were formulated according to the types and proportions of plants in actual grasslands degraded to different levels, as presented in Figure 1. Throughout the 8-week feeding trial, the sheep were fed individually in cages, which were spacious enough for them to stand and lie down in. The sheep were weighed on the final day of the trial period to determine average daily weight gain (ADG) which calculated by dividing the difference in body weight by the number of days. The sheep were fed twice per day, and allowed to feed and drink ad libitum throughout the experiment period.

The feed offered and refused for individual sheep was weighed each day to calculate the daily dry matter intake (DMI). Samples of various plants and formulated diets were collected once a week and later pooled by week. Subsequently, all feed samples were dried at 65°C for 72 h, milled to pass through a 1-mm mesh screen, then stored in sealed bags (150 mm × 220 mm) at 4°C until chemical composition was determined.

Ruminal fluid samples were collected on the final day of the experiment before the morning feeding via an oral stomach tube equipped with a vacuum pump. Approximately 10 mL of ruminal fluid initially collected was discarded to reduce the chance of contamination of the fluid sample in the stomach tube with saliva. After that, 4 layers of cheesecloth were used to strain the ruminal fluid, and the pH was measured immediately. Then, 10 mL of the filtrate is acidified by mixing with 2 mL of metaphosphoric acid (25%, wt/vol) and the mixture was centrifuged at 3,000 × g for 15 min. The supernatant was separated and stored at −20°C until volatile fatty acids (VFA) was determined. Another 4 mL of filtrate was immediately stored in liquid nitrogen until the microbial community was analyzed.

Samples of each plant and diets were sent to the Animal Nutrition Laboratory of Northeast Agricultural University (Harbin, China) for nutrient analysis using wet chemistry methods. Contents of the dry matter (DM, method 934.01), ash (method 942.05), ether-extract (EE, method 920.39) and crude protein (CP, method 988.05) in feeds were assayed in accordance with the procedures of AOAC International (26). The neutral detergent fiber (NDF) and acid detergent fiber (ADF) contents were measured based on the method described by previous study (27), which the heat-stable α-amylase was used to treat the feeds. The mineral element content was measured using an inductively coupled plasma-optical emission spectrometer (ICP-6800S, Shanghai Meixi Instrument Co., Ltd., China).

Concentrations of acetate and propionate in the rumen fluid samples were determined using gas chromatography (Shimadzu GC-2010, Japan) (28).

Total DNA in the rumen fluid was extracted using a DNA extraction kit (Shanghai Shengye Biotech, China). The V3 and V4 region of bacteria and archaea was used for 16S rRNA gene sequencing while the fungus was used for 18S rRNA gene sequencing, with all amplicon libraries preparation and sequencing performed on a MiSeq platform (Illumina, San Diego, CA, United States). The fluorescent quantitative polymerase chain reaction (PCR) amplification reaction conditions were pre-denaturation 98°C (30 s); denaturation 98°C (15 s), annealing temperature 50°C (30 s); extension temperature 72°C (30 s), 30 cycles, and final extension 72°C (5 min). After being recycled from a 1.8% agarose gel, the PCR products were purified using an OMEGA DNA purification column (Gene Company Limited). The purified products were quantified using a Quant-iT PicoGreen dsDNA Assay Kit (Gene Company Limited) in accordance with the kit’s instructions. The DNA fragments were paired-end sequenced by an Illumina miseq/novaseq. The DADA2 method was used to perform the steps of de-priming, mass filtering, denoising, splicing, and de-chimerization. Sequencing data were processed using MacQIIME version 1.9.1. Primers and homopolymer runs (maximum length, 8) of the joined sequences were trimmed after paired-end forward and reverse reads were joined. Only sequences ≥400 bp in length, with both an average quality score ≥ 25 and ambiguous bases ≤6 remained for downstream analysis. UCHIME (29) was used for de novo chimera checking, and USEARCH (30) was used for operational taxonomic unit (OTU) identification in order to identify similar sequences that had >97% similarity. BLAST (31) was used to assign representative sequences for bacterial and archaeal OTUs to the Greengenes16S rRNA gene database [version gg_13_8; (32)] and RIM-DB database (33), respectively.

The in vitro fermentation device adopted in this study was the artificial rumen simulating system MC-ABSF-II (Beijing Mancang Technology Co., Ltd., China). The protocol used for the in vitro incubation was described by Li et al. (34). Sheep rumen fluid was collected prior to morning feeding, filtered through four layers of cheesecloth, and combined with artificial saliva (39°C) at a 2:1 ratio (buffer: ruminal fluid, v: v) (35). Dispensing the 150 mL of buffered ruminal fluid into 200 mL incubation flasks that had been preheated. In each incubation flask, 2 g of each substrate were mixed with the buffered ruminal fluid. The mixture was then incubated for 24 h at 39°C in a hot water bath shaker. Using a real-time in vitro fermentation system (made by Jilin Academy of Agricultural Sciences, model Qtfxy-6), the milliliter (mL) of CH₄ from each flask was monitored.

R Statistical Software (v4.1.2; (36)) was used to analyze all of the experiment’s data. One-way ANOVA (LSD) were used to analyze the effects of grassland degradation on nutrient intake, rumen fermentation parameters and methane emission of sheep. To analyze the effects of nutrient intake to rumen microbial relative abundance, we used linear mixed effects model (LMMs). In this model, nutrient intakes were taken as fixed factors and grassland types (undegraded grassland plants, moderately degraded grassland plants and severely degraded grassland plants) were taken as random factors. Specifically, the equation of model is Yᵢ_j = X_ᵢjβ + Z_ᵢjμ_i + ϵ_ᵢj, where Yᵢ_j is the observed response for the j-th observation within group i, X_ᵢj is a vector of covariates associated with the fixed effects for the j-th observation in group i, β is the population mean which represents fixed-effect factors, Z_ᵢj is a vector of variables associated with the random effects for the j-th observation in group i, μ_i is intercepts relative to population mean across different groups which means variability between groups, ϵ_ᵢj is the error term for the j-th observation in group i representing the unexplained variability within the group. Statistically significant differences among treatments were analyzed using Tukey’s test. Significant differences were declared at p ≤ 0.05, and marginally significant differences were defined at 0.05 < p ≤ 0.10.

The results (Figure 2a) for in vitro fermentation showed that the CH₄ production in the SG group was higher than that of UG group (p < 0.05). Conversely, the sheep from MG group produced significantly less CH₄ than that of UG group (p < 0.05; Figure 2a). However, no difference for H₂ production was observed among the three groups (p > 0.10), indicating variation in carbon was driving differences. In addition, we further observed that ratio of acetate concentrations to propionate concentrations, ruminal total volatile fatty acids and acetate concentrations in the SG group was significantly decreased, but their levels significantly increased in MG group compared with the UG group (p < 0.01; Figure 2c). However, the pH (Figure 2b) and propionate concentration (Figure 2c) in the rumen were not altered among the three groups (p > 0.10). Similarly, we did not observe any differences in the average daily gain of sheep among the groups (Table 1).

We next focused particularly on the variation in microbes involved in the competition for carbon nutrition. Notably, we observed the relative abundance of Treproema that capable of converting carbon nutrition into acetic acid in the rumen of sheep in the MG group was significantly higher than that of sheep in the UG and SG groups (p < 0.05; Figure 3a). In addition, sheep in the SG group had higher ruminal Methanosphaera, which is a crucial role in the methane production by using carbon nutrition in relative abundance compared with the UG and MG groups (p < 0.05; Figure 3b). Among the different fungal genera, the relative abundance of Macrophoma, Moesziomyces and Erythrobasidium in the MG group increased significantly compared with the UG and SG groups, whereas the relative abundance of Nigrospora in sheep fed MG and SG diets decreased significantly compared with those sheep fed UG diet (p < 0.05; Figure 3c). Meanwhile, the Piromyces in relative abundance increased linearly with increasing levels of grassland saline-alkaline degradation (UG vs. MG vs. SG, all p < 0.05).

To investigate the underlying causes of the observed variation in carbon-nutrient competing microbes, we firstly conducted an analysis of the microbial network structure. Figure 4 showed that the Moesziomyces play the most important role among these key microbes, as it not only had a positive effect on Macrophoma, Erythrobasidium and Treproema, but also had a negative effect on Nigrospora (p < 0.05). Meanwhile, the Macrophoma and Erythrobasidium also had a positive effect on Treproema (p < 0.05). However, there was no significant effect of Nigrospora and Erythrobasidium on the Treproema (p > 0.10). Furthermore, our observations revealed that Methanosphaera was not integrated into the microbial network diagram, suggesting a potential insusceptibility of this genus to microbial interaction effects.

We further analyzed the significance of differences in nutrient intake among different groups of sheep. We found no significant differences in DM intake (Figure 5a)among the groups of sheep (p > 0.05), but the CP (Figure 5b) and EE (Figure 5c) intake of sheep in the SG group was higher (p < 0.05). Additionally, there were no significant differences in NDF intake (Figure 5d) among the groups of sheep (p > 0.05), while the ADF (Figure 5e) intake of sheep in the SG group was lower (p < 0.05). Notably, the Na intake (Figure 5f) of sheep increased linearly with the increasing levels of grassland saline-alkaline degradation (UG vs. MG vs. SG, all p < 0.05).

To identify the nutritional factors most likely to influence the variation in Methanosphaera, we next focused on analyzing the correlation between the nutritional resources consumed by those sheep with significant differences and Methanosphaera (Table 2). We found a significant effect of Na intake on Methanosphaera relative abundance (p < 0.01), but no significant effect of other nutrient intake on the Methanosphaera relative abundance was observed.