Alpine meadow degradation depresses soil nitrogen fixation by regulating plant functional groups and diazotrophic community composition

Biological nitrogen fixation (BNF), a function performed by diazotrophic microbes, plays an essential role in nitrogen (N) bioavailability in terrestrial ecosystems. However, little is known about the effects of meadow degradation on soil BNF and diazotrophic communities in alpine meadows. We investigated changes in soil BNF and their potential drivers in alpine meadows along a degradation gradient on the Tibetan Plateau (non-degraded, lightly degraded, moderately degraded, and severely degraded meadows) using real-time quantitative PCR and amplicon sequencing. Soil BNF rates decreased significantly along the meadow degradation gradient with a range of 17.34–79.84 nmol C2H4 g−1 dry soil d−1 across all sites. The highest BNF rate in the non-degraded meadow was 1.5–4.6-fold higher than that in the degraded meadows. The abundance and diversity of diazotrophs measured by nifH abundance and Shannon diversity was also decreased in the degraded meadows, accompanied by decreases in plant biomass, soil moisture, and nutrient content (C, N). Soil BNF rate was correlated with plant biomass, soil nutrient content, and diazotrophic abundance (including Nostoc, Scytonema, Rhodopseudomonas, and unidentified genera within the Rhizobiales and Proteobacteria). The community composition of diazotrophs differed markedly among sites with different levels of degradation, with both autotrophic (Cyanobacteria) and heterotrophic (Proteobacteria) diazotrophs contributing significantly to BNF. The plant functional groups, particularly the sedge family, were the primary drivers of soil BNF rates via mediating soil moisture, nutrient content (dissolved organic C and N), nifH gene abundance, and diazotrophic community composition. Our results reveal the main drivers of decreased BNF during alpine meadow degradation and emphasize the importance of plant functional groups in shaping the diazotrophic community and regulating the BNF rate. This information can be applied to the restoration of degraded meadow ecosystems.


Introduction
Alpine meadows are widely distributed in the Tibetan Plateau-covering a total area of 700,000 km 2 . They play multiple roles in soil and water conservation, carbon (C) sequestration, climate regulation, and habitat conservation (Che et al. 2019;Wang et al. 2020a). However, in the past 30 years, large areas of alpine meadows have experienced severe degradation due to severe climate change and anthropogenic activities, posing an increasing threat to their resistance and sustainability (Fu et al. 2012;Luo et al. 2018;Wen et al. 2010). Meadow degradation is usually characterized by the destruction of plant communities, soil drought, soil nutrient depletion, and even desertification (Urak et al. 2017;Yang et al. 2013). In addition, the low soil nitrogen (N) content in alpine meadows is a critical factor limiting the restoration and development of these ecosystems (Colin et al. 2019;Zhang et al. 2020a). Meadow degradation can accelerate the loss of soil N, further limiting its availability in alpine ecosystems (Kou et al. 2019). Although the dynamics of N in the pedosphere have been widely studied, our understanding of how meadow degradation affects N bioavailability is still largely lacking, hindering the modeling of N dynamics in alpine ecosystems.
Biological nitrogen fixation (BNF) is a microbial process that reduces gaseous N 2 to bioavailable ammonium (NH 4 + -N) via catalysis by nitrogenase enzymes; it represents a dominant input of N in meadow ecosystems with extensive N limitation (Vicente and Dean 2017;Wang et al. 2017). BNF occurs in two ways: symbiotic and free-living (Che et al. 2018;Reed et al. 2011;Wang et al. 2021a); the former in root or stem nodules of legumes, and the latter occurs independently of legume-microbe mutualism (Lindstroem and Mousavi 2020;Smercina et al. 2019). Compared with symbiotic N 2 fixation, free-living N 2 fixation may be more sensitive to environmental changes, but also could be more resilient (Che et al. 2018;Smercina et al. 2019). Although the rate of free-living BNF is lower than that of symbiotic fixation, in N-limited ecosystems in which most plants are unable to fix N 2 , free-living N 2 fixation may be one of the most important pathways for soil N accumulation (Reed et al. 2011;Wang et al. 2021a). Hence, understanding the abundance, diversity, and composition of these communities is imperative to elucidating their ecological functions in alpine ecosystems.
Although the mechanisms of BNF are well documented, how diazotrophic communities respond to alpine meadow degradation is still not understood. The nifH gene has been widely used as a molecular marker to study the abundance and community composition of soil N 2 -fixing bacteria for its highly conserved nature (Fan et al. 2019;Levy-Booth et al. 2014;Mårtensson et al. 2009). Furthermore, the function of N 2 fixation in soil is determined by the diazotrophic community composed of multiple distinct N 2 -fixing taxa (Che et al. 2017;Dai et al. 2021;Wang et al. 2020b). Therefore, to better understand the driving mechanisms of soil BNF in alpine ecosystems, the community composition and diversity of diazotrophs should be considered. Moreover, the biotic and abiotic pathways that regulate BNF are also unclear, especially in alpine meadow ecosystems.
Several studies have reported the effects of meadow degradation or restoration on BNF Wen et al. 2010;Zhang et al. 2020a). Changes in plant community structure caused by meadow degradation are important reasons for soil BNF changes . Plant roots regulate the rhizosphere micro-ecosystem and affect diazotroph metabolism by releasing root exudates, thus directly influencing diazotroph activity (Perez-Montano et al. 2014;Wang et al. 2020a). Revegetation can also promote the formation of diverse diazotrophic communities by increasing the abundance and diversity of microbes, thus regulating BNF (Huang et al. 2011;Li et al. 2021). Indirectly, the decomposition of plant residues provides organic matter and other substances, that change soil physicochemistry (Yang et al. 2021). The activity, abundance, diversity, and composition of N 2 -fixing bacteria are highly sensitive to variations in the physicochemical properties of soil moisture, pH, temperature, nutrients, and C/N/ phosphorus (P) availability (Fan et al. 2019;Li et al. 2021;Reed et al. 2011).
Plant functional groups (PFGs) are important factors that connect plant communits and ecosystem processes; PFG variation can generate feedback regulation on soil-microbe interactions (Wang et al. 2021b;. Luo et al. (2018) confirmed that the community structure of PFG can be used to as a marker for degradation of alpine grasslands on the Tibetan Plateau. Species grouped in a PFG have similar strategies in response to environmental disturbance; therefore, PFGs are crucial for the study of ecosystem functions, especially BNF . Thus, assessing the linkages between PFGs, soil properties, and diazotrophs should improve our understanding of the regulation of soil BNF with meadow degradation.
In this study, we aimed to assess the impacts of meadow degradation on diazotrophic community structure and soil BNF on the Tibetan Plateau. Here, we sequenced the nifH gene and quantified its abundance to assess the soil diazotrophic communities of alpine meadows differing the degradation levels. We hypothesized that: (1) meadow degradation significantly reduces soil BNF rate and changes diazotrophic community structure; and (2) plant communities, especially PFGs, are the main influencing factors of soil BNF rate and diazotrophic community structure. To test these hypotheses, we measured: (1) the main variation in soil physicochemical properties caused by meadow degradation; (2) how meadow degradation affects soil BNF rate and abundance, as well as the composition and diversity of diazotrophic community; and (3) whether PFG variation causes concomitant variations in soil BNF rate and diazotrophic communities.

Study site
The study was conducted at the grassland station (91°05′E, 30°51'N, 4333 m ASL) of the Alpine Meadow Nature Reserve of Damxung County in the mid-south portion of the Tibetan Plateau. This area has a typical semi-arid alpine continental climate; its annual mean precipitation and air temperature are 477 mm and 1.3 °C, respectively (Wang et al. 2021c). Approximately 85% of the precipitation occurs from July to August during the summer monsoon season. The average low temperature is −10.4 °C in January and the average high temperature is 10.7 °C in July.
There are approximately 62 frost-free days per year, with an annual accumulated temperature (≥ 0 °C) of 1800 °C. Its soil is mainly classified as mountain meadow soil (Cambisol, FAO/UNESCO classification), with a sandy loam texture. The meadow is dominated by Kobresia pygmaea, Stipa capillacea, and Carex montis-everestii (Fu et al. 2012).

Experimental design and sample collection
Four meadows were established along a degradation gradient (non-degraded, ND; lightly degraded, LD; moderately degraded, MD; and severely degraded, SD meadows) by the Chinese Academy of Sciences in this nature reserve, with an area of 15-20 hm 2 (Table S1; Fig. S1). Based on a vegetation survey (Tables S2 and S3), an index was used to evaluate the degree of meadow degradation as described by Wen et al. (2010). Five 50 × 50 m plots were set up randomly with similar terrain conditions in each meadow, with a distance of more than 80-100 m between each plot. In total, 20 plots were established in a randomized complete block design.
Five quadrats (1 × 1 m) were set in each plot to vegetation surveys and soil sampling in August 2019. Soil sub-samples were randomly collected at a depth of 15 cm per quadrat along an S-shaped pattern (nine cores per quadrat) by a sterile soil auger with a diameter of 10 cm. A composite sample was formed by evenly mixing all sub-samples from five quadrats. All soil samples were preserved on dry-ice box in situ and brought to the laboratory within one day. After roots, litter, stone, and debris were removed, each soil sample was filtered through a 2-mm sieve and divided into two subsamples. One subsample was stored at −80 °C for DNA extraction, and the other was air-dried for physicochemical analysis. After soil sampling, the plants in each quadrat were dug up and sorted into functional groups (sedges, grasses, forbs, and legumes) as described by . The above-and below-ground parts of each plant species were carefully separated, cleaned, dried at 70 °C for 48 h, and weighed to obtain the aboveground and belowground biomass (AGB and BGB). The plant 1 3 Vol:. (1234567890) Shannon diversity index was calculated as described by Zhang et al. (2016).

Soil physicochemistry analysis
Soil moisture (SM) was determined by oven-drying at 105 °C for 24 h. As described by Zhang et al. (2018), the concentrations of NH 4 + -N and NO 3 − -N were determined by extracted with 2 M KCl (1:10 w/v) and analyzed using a segmented-flow autoanalyzer system (AutAnalyel, Bran+Luebbe GmbH, Norderstedt, Germany). The Kjeldahl digestion method was used to determine soil total nitrogen (TN). After extraction with distilled water, dissolved organic nitrogen (DON) and dissolved organic carbon (DOC) were analyzed using a TOC/TN analyzer (LiquiTOC II, Elementar Analysensysteme GmbH, Langenselbold, Germany). Soil organic carbon (SOC) was determined using potassium dichromate (K 2 Cr 2 O 7 ) oxidation. Total phosphorus (TP) was determined by meltmolybdenum, antimony, and scandium colorimetry. Soil available potassium (AK) was extracted with 1.0 M CH 3 COONH 4 at a pH of 7.0 and determined using a flame photometer. Soil available phosphorus (AP) was determined using a UV spectrophotometer (Camspec, Cambridge, UK) after extraction in 0.5 M NaHCO 3 at a pH of 8.5 and measured by the method of molybdate ascorbic acid. Soil pH was measured with a pH meter (Spectrum Technologies Inc., Bridgend, UK) using a soil-water ratio of 1:2.5 (w/v) prepared after air drying.

Biological N 2 fixation rate
The BNF rates were measured by the acetylene (C 2 H 2 ) reduction assay (Patra et al. 2006). In brief, fresh soil (10 g on dried basis) from each sample was transferred to sterile bottles (100 ml). In each bottle, 10% of the headspace air was replaced by C 2 H 2 (99.99%) using a gas syringe. Flasks were incubated for 7 days at 25 °C. A mixed gas sample (20 ml) was then extracted from the headspace of each bottle and placed into a 12-ml pre-evacuated, airtight vial using a gas syringe. The concentration of ethylene (C 2 H 4 ) in these samples was measured using a gas chromatograph (Agilent Technologies Inc., USA). Each sample was prepared and measured in triplicates, and the mean value for each sample was calculated. The BNF rates are expressed as the rate of C 2 H 4 production rate (nmol C 2 H 4 g −1 dry soil d −1 ). We also blank controls without soil to measure the background C 2 H 4 concentration of C 2 H 2 , in which C 2 H 4 concentrations were lower than the detection limit of the gas chromatograph.

Microbial DNA extraction and quantitative PCR
The DNA was extracted from 0.5-g soil samples using the FastDNA ® SPIN Kit, (MP Biomedicals, Cleveland, USA) following the manufacturer's instructions. The quality and concentration of DNA were assessed using a NanoDrop 2000 spectrometer (Thermo Scientific, Wilmington, DE, USA). All purified DNA samples were stored at −80 °C until analysis.
The gene abundance of nifH was determined by an ABI Prism 7500 Real-Time quantitative PCR (qPCR) system (Applied Biosystems, Foster City, CA, USA) using primer pairs nifH-F (5′-AAA GGY GGW ATC GGY AAR TCC ACC AC-3′) and nifH-R (5′-TTGTTSGCSGCR TAC ATSGCC ATC AT-3′) (Rosch and Bothe 2005). The 20 μL qPCR reaction mixture consisted of 10 μL SYBR® Premix Ex Taq™ (Takara Bio USA), 1 μL DNA template, 0.8 μL each of the forward and reverse primers, and 7.4 μL double-distilled water (ddH 2 O). The qPCR program ran was: denaturation at 98 °C for 5 min, then 35 cycles of denaturation at 95 °C for 30 s, annealing at 58 °C for 30 s, and extension at 72 °C for 60 s. Reactions were carried out in triplicate for each sample and with negative controls (without DNA template) run simultaneously. Standard curves were obtained with tenfold serial dilutions of nifH plasmids. The nifH gene was quantified based on average slopes obtained from standard curves.

Amplicon sequencing and phylogenetic classification
Community composition and diversity of diazotrophs were analyzed using Illumina MiSeq sequencing technology (Illumina Inc.) with the primer sets nifH-F/nifH-R (Rosch and Bothe 2005). The 20 μL PCR reaction mixture consisted of 4 μL 5 × FastPfu buffer, 2 μL dNTPs (2.5 mM), 0.8 μL of each primer (5 μM), 0.4 μL FastPfu polymerase, 0.2 μL bovine serum albumin, 10 ng template DNA, and ddH 2 O. The PCR reactions were performed using TransGen 1 3 Vol.: (0123456789) AP221-02 TransStart ® FastPfu DNA Polymerase (Transgen Biotech Co., Ltd.) on an ABI GeneAmp ® 9700 PCR System (Applied Biosystems). The PCR program was: initial denaturation at 95 °C for 3 min, followed by 35 cycles at 95 °C for 30 s, annealing at 55 °C for 30 s, 72 °C for 45 s, and a final extension at 72 °C for 10 min. Each sample was amplified as three technical replicates, and then pooled as a PCR product. PCR amplicons were purified and recovered by agarose gel electrophoresis (1.5% gel agar, 90 V, 45 min). The product was quantified using a QuantiFluor™-ST Solid Standard (Promega, Madison, WI, United States), then equimolar concentrations of PCR product for each sample were pooled. Samples were diluted to appropriate loading concentration for a MiSeq run, spiked with 20% (v/v) PhiX control library, and sequenced using the Illumina MiSeq PE300 platform (Majorbio Company in Shanghai, China), producing 2 × 300 bp long reads. Demultiplexing of the reads was performed on the instrument.
The QIIME-1.9.1 pipeline (https:// docs. qiime2. org/ 2021.4/ citat ion/) was used to analyze the nifH raw nucleotide sequences. Quality-controlled sequences were obtained after filtering the adaptors, barcodes, and low-quality sequences (Phred quality score < 20, containing ambiguous nucleotides or not matching the primer). Effective tags were obtained after chimeric sequences were removed using the Gold Database with the UCHIME algorithm (Haas et al. 2011). Operational taxonomic units (OTUs) were clustered using UCLUST with a 97% similarity cut-off (Edgar 2010). Taxonomic identity was determined using the RDP classifier (Version 2.2, http:// rdp. cme. msu. edu/) in the FunGene (Version 9.6, http:// funge ne. cme. msu. edu/) database (Fish et al. 2013). Shannon diversity and Good's coverage were calculated using Mothur software v. 1.30.2 (https:// mothur. org/ CITAT ION/). Good's coverage across all sites was >0.99, implying that the sampling depth was sufficient for estimation of the microbial diversity (data not shown).

Statistical analysis
One-way analysis of variance and Duncan's multiple tests were used to assess the differences in plant and soil properties, soil BNF rate, and α-diversity of diazotrophs among the four meadows (P < 0.05). Nonmetric multidimensional scaling (NMDS) ordinations and permutational multivariate analysis of variance ANOVA (PERMANOVA) were performed to compare differences in diazotrophic community structure among four meadows using Bray-Curtis dissimilarities in Hellinger-transformed relative abundances in the "vegan" package (Version: 2.5-7). A distance-based redundancy analysis (db-RDA) was conducted to test the relationships between diazotrophic community structure and environmental factors with the capscale function in the "vegan" R package. To identify collinearity among explanatory variables, the variation inflation factors (VIF) were used as the criterion. The variable of VIF with the single highest value was removed, then the process was repeated until all values were less than 10. The relationships between BNF rate, gene abundance, and α-diversity of diazotrophs and environmental factors were assessed using Spearman's correlation. A mantel test was used to test the correlation between diazotrophic composition and environmental factors in the "vegan" R package. Structural equation modeling (SEM) was employed to identify the potential drivers of soil BNF rate using the "lavaan" R package (Version: 0.6-7). Before beginning the SEM, we eliminated the redundancy of soil and plant properties by removing the variables with high correlations (Spearman's ρ 2 > 0.7) with the varclus function in the "Hmisc" R package (Version 4.4-2) due to the strong collinearity among soil and plant properties. Subsequently, NH 4 + -N, NO 3 − -N, AK, pH, and C/N were utilized to establish the SEM (Fig. S2). Variables of plant richness, AK and pH were excluded from the SEM for those variables showed narrow variation and had no significant correlation with soil biological nitrogen fixation (BNF) rates. Aggregated boosted tree (ABT) analysis was conducted to quantify the relative importance of PFGs for BNF rates and diazotrophic community composition using the "gbmplus" R package. All statistical analyses were performed with R v.3.6.2 (R core team).

Plant characteristics, soil properties, and BNF rates
Plant characteristics differed significantly among the four meadows (Table 1). The plant cover decreased significantly in the degradation meadows, 1 3 Vol:. (1234567890) most sharply in SD meadows with the lowest cover (39.4%). The BGB values of MD and SD were also significantly lower than those for ND and LD, although no significant difference was found between LD and ND. However, LD had the highest AGB and, whereas MD had the highest plant Shannon diversity and richness index.
The composition of plant communities varied significantly among the four meadows ( Fig. 2 and  Fig. S3). The dominant species in ND was Kobresia pygmaea with a cover of 78.5% and relative abundance of 66.2%, but shifted to Carex montis-everestii in LD and MD. Forbs were dominant in SD, such as Leontopodium nanum. Along the meadow degradation gradient, the dominant PFGs transitioned from sedges to forbs (Table 2). Sedge groups were dominant in ND and LD, with relative abundances of 82.1% and 48.4%, respectively. Compared with ND, the plant cover, AGB, BGB, and relative abundance of sedge decreased significantly in the degraded meadows. However, the relative abundance of forbs significantly increased in the degraded meadows and was highest in SD. The plant cover, AGB, BGB, and Shannon diversity of forbs were in the order of MD > LD > SD > ND. The grass and Legume groups accounted for a minor proportion of the plants with relative abundances of 5.1-22.7% and 0-7.2%, respectively. The soil nutrient levels, including the contents of NO 3 − -N and NH 4 + -N, DOC, DON, SOC, TN, and C/N, were also decreased significantly in the degraded meadows (Table 3). However, compared with those in ND, AP, TP, and SM were decreased in MD and SD, but there was no significant difference between ND and LD. The AK content was also showed no significant difference among four meadows. Soil pH showed a significant but narrow variation from 6.33 to 7.10 among the four meadows. Soil BNF rates varied significantly among the four meadows, with a range of 17.34-79.84 nmol C 2 H 4 g −1 dry soil d −1 across the gradient (Fig. 1A). The highest BNF was observed in ND and was 1.5-, 1.8-, and 4.6-fold higher than in those of LD, MD, and SD, respectively.
Gene abundance, diversity and community composition of diazotrophs Similar to the trend in soil BNF, the nifH copy number and α-diversity of diazotrophs significantly decreased in our decreased meadows ( Fig. 1B and C). The abundance of the nifH gene was 0.99-7.68 × 10 6 copies·g −1 dry soil, with an mean rate of 3.31 × 10 6 copies·g −1 dry soil. Compared with nifH abundance in ND, that in LD, MD, and SD decreased by 41.2%, 63.7%, and 72.4%, respectively. The Shannon diversity decreased by 22.1-25.1% in the degraded meadows compared with those in ND, but no significant differences were observed among LD, MD, and SD. The abundance of nifH and the Shannon diversity of diazotrophs were all positively correlated with AP, NO 3 − -N, NH 4 + -N, DOC, DON, SOC, TN, SM, AGB, BGB, and plant cover, but negatively correlated with soil pH and plant Shannon diversity (Table 4). However, the AK and plant richness indices showed no significant correlation. TP correlated with the nifH abundance and the Shannon diversity of diazotrophs.
At the phylum level, Proteobacteria and Cyanobacteria were the most abundant taxa (55.6% and 40.3%, respectively) in the soil diazotrophic community across all sites ( Fig. 2A). At the genus level, unclassified_f_Nostocaceae (25.9%), Scytonema (8.3%), Nostoc (5.3%), Skermanella (20.2%), unclassified_o_Rhizobiales (12.4%), unclassified_c_ Alphaproteobacteria (10.2%) and Scyronema (8.3%) were the most abundant genera (Fig. 2C). NMDS and PERMANOVA analyses showed that diazotrophic community composition differed significantly between four meadows (Fig. S4). The diazotrophic Table 3 Soil physicochemical properties in four meadows along the degradation gradient Values are means ± standard error (n = 5). Different letters indicate significant differences between different treatments (Duncan's test, P < 0.05). ND non-degraded meadow, LD lightly degraded meadow, MD moderately degraded meadow, SD severely degraded meadow, AP available phosphorus, DOC dissolved organic carbon, DON dissolved organic nitrogen, SOC soil organic carbon, TN total nitrogen, TP total phosphorus, AK available potassium, SM soil moisture, C/N the ratio of soil organic carbon and total nitrogen community structure in ND showed clear differences from that those in the degraded meadows, whereas the community structures of diazotrophs in LD, MD, and SD were similar.  Diazotrophic community composition was significantly associated with all measured soil physicochemical properties (except for AK), plant cover, and Shannon diversity, and was most closely associated with NO 3 − -N and C/N (Table 1). Specifically, the relative abundances of Nostoc, Rhodopseudomonas, p_Proteobacteria, and o_Rhizobiales were positively correlated with AP, NO 3 − -N, NH 4 + -N, DOC, DON, SOC, TN, TP, SM, C/N, BGB, and plant cover, but negatively correlated with soil pH and plant Shannon diversity (Fig. 3). However, the relative abundances of c_Alphaproteobacteria, f_Rhizobiales, and Skermanella showed the converse pattern in response to soil and plant characteristics. Furthermore, Trichormus, Rhodomicrobium, and Scytonema showed no significant correlation with any of the soil and plant properties we measured. RDA also confirmed the importance of plant and soil properties (Fig. 4), in which the first two axes explained 62.2% of the variation in diazotrophic communities.

Potential drivers of soil BNF rate
Soil BNF rate positively correlated with AP, NO 3 − -N, NH 4 + -N, DOC, DON, SOC, TN, TP, SM, C/N, AGB, BGB, and plant cover, but negatively correlated with soil pH and plant Shannon diversity (Table 4). However, AK and the plant richness index showed no significant correlation with soil BNF rate. Soil BNF rate was also closely associated with diazotrophic community composition, and increased with the absolute abundance of dominant diazotrophic groups, including o_Rhizobiales, Scytonema, p_Proteobacteria, Nostoc, and Rhodopseudomonas (Fig. S5).
The established SEM explained 86% of the variance in the BNF rate (Fig. 5), and the plant community had a direct effect on BNF by altering the plant cover. In addition, plant community composition affected soil DOC and NO 3 − -N, diazotrophic composition, and nifH gene abundance, and further regulated the soil BNF rate (Fig. 6A). Plant cover had the highest total effect on soil BNF rates (Fig. 6B). Among diazotrophic community attributes, nifH abundance and Shannon diversity of diazotroph were good predictors of soil BNF rate (Fig. 6B, C and D).
Given the predominant regulation of soil BNF rate by plants, the effects of specific PFGs on soil BNF rate and the structure of the diazotrophic community were further examined by ABT modeling (Fig. 6). The Sedge plant cover was the main driver for soil BNF rate, gene abundance, α-diversity, and diazotrophic community composition. In total, sedges had a major effect on soil BNF rate, which together accounted for 58.5% of the variation (Fig. 6A). Sedges were also the main factors affecting nifH abundance, composition, and diversity of diazotrophs (Fig. 6B). The forbs groups also had an important influence on soil N 2 fixation and related diazotrophic community structure, accounting for 21.0-39.8% of the variation. However, the legume and grass groups had little effect on soil BNF and diazotrophic community structure.

Discussion
We found that the BNF rate was 1.5-4.6 times higher in non-degraded higher than in the degraded meadows, suggesting that alpine meadow degradation depressed soil BNF rate on the Tibetan Plateau (Fig. 1). Similar results were also found for the processes of vegetation restoration or degradation, in which the soil BNF rate increased or decreased, respectively, in multiple ecosystems (Li et al. 2021;Lopez-Lozano et al. 2016;Zhang et al. 2020a). For example, during the degradation of alpine peatlands, soil BNF rates decreased by 42.7% from pristine marshes dominated by sedge (1.71 μmol N·g −1 ·d −1 ) to moderately degraded meadows (0.98 μmol N·g −1 ·d −1 ), and decreased by 88.3% in sandy meadows (0.20 μmol N·g −1 ·d −1 ) . In an abandoned farmland ecosystem, Fig. 3 Heatmap for Pearson's correlation between vegetation and soil properties and the relative abundance of dominant diazotrophs at the genus level. AP: available phosphorus; DOC: dissolved organic carbon; DON: dissolved organic nitrogen; SOC: soil organic carbon; TN: total nitrogen; TP: total phosphorus; AK: available potassium; SM: soil moisture; C/N: the ratio of SOC and TN; AGB: plant aboveground biomass; BGB: plant belowground biomass; H p : plant Shannon diversity, BNF: biological N fixation rate. *P < 0.05; **P < 0.01; ***P < 0.001

3
Vol.: (0123456789) the BNF rates increased during restoration and ranged from 323.9 to 654.1 nmol C 2 H 4 g − 1 dry soil d − 1 (Li et al. 2021), approximately 10-fold higher than the rates measured in our study. In addition, soil BNF rates in alpine meadow ecosystems were lower than those reported in most agricultural ecosystems, but higher than those reported in degraded grasslands and forests with human impacts (Horel et al. 2018;Li et al. 2021;Wu et al. 2021). This difference may be attributed to different plant-soil-microbe interactions in different environments.
Plant cover and biomass (above-and belowground) were significantly reduced along the degradation gradient (Tables 1 and 2), suggesting that meadow degradation reduced aboveground productivity and strongly impacted plant community structure. In our study, plant community composition and structure, particularly sedges, were very important factors driving the variation of BNF rates, with plant cover having the greatest effect on soil BNF rates (Fig. 5). Furthermore, sedge groups had a major effect on the variation in soil BNF rates (Fig. 6). Similar results were found by Luo et al. (2018), who found that grassland degradation dramatically reduced plant cover and productivity, confirming that PFG characteristics are useful for indicating the process of land degradation. In our study, the dominant PFGs shifted from sedges in ND to forbs in SD (Table 2). One possible reason is that relative to other PFGs, forbs better compete for resources in severely degraded alpine meadows-particularly in the Tibetan meadow with extensive N limitation ). The coverage of sedges, instead of AGB, BGB, or diversity of PFGs, was the most important factor influencing soil BNF and diazotrophic communities (Fig. 6). The possible reason is that sedge changes BNF by changing soil, water, and nutrient cycling by blocking sunlight. Zhang et al. 2020a).
The variables associated with plant characteristics can impact soil-diazotroph interactions by both directly or indirectly, consequently, regulating the soil BNF rate (Lopez-Lozano et al. 2016;Zhang et al. 2020a). Lower plant biomass indicates smaller rhizosphere volume and a smaller supply of root exudates (i.e., labile organic C and N) for diazotrophs, which are detrimental to the function of BNF (Yang et al. 2021;Zhang et al. 2020a). Indeed, we observed obvious effects of meadow degradation on the diazotrophic community composition (Fig. 3). Both nifH abundance and α-diversity significantly declined along the degradation gradient (Fig. 1). Our results are consistent with those from other studies showing that both absolute abundance and diversity of diazotrophs are reduced by grassland degradation and increase with restoration (Lopez-Lozano et al. 2016;Wang et al. 2017). A possible reason for this is that meadow degradation not only reduces soil nutrition level and nutrient availability, but also provides a poorer living environment for diazotrophs and further inhibiting their growth (Zhang et al. 2020b). The diversity of PFGs may have a stronger influence on soil microbial diversity and soil function by providing more carbon resources (Andruschkewitsch et al. 2014;Wang et al. 2021b). In our study, the sedge groups had a major effect on the soil BNF rate, nifH abundance, Shannon diversity index, and diazotrophic community composition (Fig. 6). The roots and rhizomes of sedges have been shown found to promote diazotroph growth (Rejmankova et al. 2018). In addition, the weak influence of legumes on soil BNF rates and diazotrophic community structure suggests that free-living diazotrophs contribute more  . Therefore, plant-diazotroph interactions may be the major controlling factor in determining BNF rate during alpine meadow degradation-especially for sedges rather than legumes.
The plant biomass indirectly determines the input of organic matter and other substrates into the soil through the decomposition of plant litter and root residue (Yang et al. 2021). In our study, both soil moisture and nutrient supply capacity in surface layers decreased significantly along the degradation gradient (Table 3). Other studies of Tibetan alpine meadows have shown that meadow degradation is accompanied by a sharp decline in soil nutrients and moisture. For example, the contents of SOC and TN in a pristine meadow were seven and six times higher, respectively, than those in a sandy meadow ). The decrease in soil moisture during meadow degradation could be attributed to the decrease in vegetation cover or other traits (e.g., root biomass, leaf area, and plant height) that increase evaporation and/or reduce soil water retention (Yang et al. 2013). The decrease in aboveground and belowground plant biomass in degraded meadows could reduce the input of organic matter into the soil through decomposition of litter and release of root exudates (Prieto et al. 2017).
We found that all measured soil properties (except AK) were strongly correlated with BNF rates and diazotrophic community composition (Table 4). Both nifH abundance and α-diversity of diazotrophs were positively correlated with soil nutrients (e.g., AP, NO 3 − -N, NH 4 + -N, DOC, SOC, and TP) in our study. Particularly, DOC and DON correlated better with BNF rates. Dissolved organic matter, which supplies these microbes with additional organic nutrients, is the major factor controlling for N cycling in soils (Levy-Booth et al. 2014). N 2 fixation by heterotrophic diazotrophs is highly dependent on organic substances from the environment, such as those provided by soil and plant systems; thus, it is necessary to maintain high N 2 fixation potential in substrates with high C quantities (Dai et al. 2021). In addition, the reaction of BNF is catalyzed by nitrogenase and has a high energy requirement, suggesting that the growth and functions of diazotrophs are not only controlled by C availability, but also by N, P, and K availability (Zheng et al. 2019). However, several studies have shown that high inorganic N content (e.g., NH 4 + -N and NO 3 − -N) significantly inhibits soil BNF capacity in agricultural ecosystems with N fertilization addition (Dai et al. 2021;Fan et al. 2019), which is inconsistent with our results. A possible reason for this difference is that the soil N concentration tended to be higher in those other studies, resulting in excess N limiting the microorganism growth and activity . In our study, the increase in soil N concentration may have been caused by BNF capacity, rather than its cause. This has also been observed by Zhang et al. (2020b) in a degraded alpine peatland on the Tibetan Plateau. Soil moisture content and pH are also important factors influencing BNF rates, as they affect microbial activity and determine the processes of N 2 fixation Zhang et al. 2020a). Zhang et al. (2020a) found that the decreased soil BNF rate in a degraded alpine peatland was primarily driven by soil moisture, which led to changes in soil physicochemical properties and plant communities, further leading to changes in soil BNF. Wang et al. (2017) reported that soil pH was a major driver of diversity and turnover of diazotrophic community in alpine meadows on the Tibetan Plateau. However, pH was only weakly correlated with soil BNF rate and diazotrophic communities in our study (Table 4). The reason for this difference is that the influence of pH in this context is scale-dependent (Martiny et al. 2011). The significant effect of soil pH on BNF can be determined at larger scales with stepper pH gradients, such those observed with along an altitudinal gradient.
In our study, both heterotrophic and autotrophic diazotrophs, including Proteobacteria and Cyanobacteria, performed the majority of N 2 fixation (Fig. 2). However, their responses to meadow degradation varied, indicating differences in substrate preferences and nutrient acquisition strategies (Zhang et al. 2016). The relative abundance of Proteobacteria decreased, while that of Cyanobacteria increased along the degradation gradient (Fig. 2). Meadow degradation affects diazotrophic composition by altering multiple soil and plant properties, such as C/N, SOC, and DOC, which are the most important factors in shaping the diazotrophic community composition (Table 4). Previous studies have also demonstrated that C substrates are the main drivers of soil microbial community composition (Yang et al. 2021;Zhang et al. 2016). Indeed, both f_Nostocaceae and Skermanella-two of the most dominant abundant genera-were significantly correlated with soil C/N, SOC, and DOC (Fig.4). The reduced input of labile organic C from plants caused by meadow degradation may further limit the growth of dominant diazotrophic taxa, especially heterotrophic diazotrophs (Che et al. 2019). The increase in the abundance of autotrophic diazotrophs with meadow degradation can be explained by the increased light caused by the decreased plant cover ). However, the relative contribution of heterotrophic and autotrophic diazotrophs to soil BNF is unclear, and further research is needed.
Although the nifH gene may not be expressed to produce nitrogenase and the enzyme activity is also affected by various factors (Li et al. 2021;Mårtensson et al. 2009;Wang et al. 2018), we have shown in this study that in its abundance can still be used as a sensitive bioindicator of soil BNF capacity. The abundance of diazotrophic genera, including Nostoc, Scytonema, Rhodopseudomonas, o_Rhizobiales, and p_Proteobacteria, is more indicative of soil BNF rates than their relative abundance ( Fig. 5 and S5), which confirms the importance of the nifH gene. Shannon diversities of diazotrophs are also suitable for predicting and characterizing BNF rates (Fig. 6). Loss of population and diversity of soil microorganisms can destroy multiple ecosystem functions, such as the maintenance of soil fertility and health (Wagg et al. 2014). Therefore, communities harboring more diverse diazotrophic taxa can provide more ecosystem functions than those with fewer species. Degraded alpine meadows may be particularly sensitive to the functional deterioration of BNF, as their primary source is BNF. Additionally, decreased BNF and altered community composition of diazotrophs may be coupled with other biogeochemical processes-such as greenhouse gas production-that pose a threat to climate change (Eberhard et al. 2018).
It is worth noting that our research results are based on field surveys. Although this approach has been used to investigate changes in plant-soil ecosystems and processes, it is not perfect because it cannot identify all potential influencing factors, such as the influence of sampling geographical distance (Ramette and Tiedje 2007;Zhang et al. 2018;Wang et al. 2021c). For that, we examined the contribution of sampling distance to the changes in soil BNF rates and found that sampling distance had no significant effect on BNF (Fig. S6). Therefore, the decrease in soil BNF rates caused by meadow degradation could be driven by plant and soil properties and their interaction. As for the effects of PFGs in driving the interaction of soil-diazotrophs, the next step is to adopt strict control experiments to verify their effects, such as PFG removal experiment (Fanin et al. 2019;Xiao et al. 2017).

Conclusion
Our results suggest that soil BNF decreased significantly along meadow degradation gradient. The reduced soil BNF was closely related to changes in the plant community (especially in the sedges family) that regulate soil moisture, nutrient levels (dissolved organic C and N), nifH gene abundance, and diazotrophic community composition. Both autotrophic (Cyanobacteria) and heterotrophic (Proteobacteria) diazotrophs contributed to BNF. Our results emphasize the importance of plant functional groups in shaping the diazotrophic community and regulating soil BNF rate, providing information to inform the restoration of alpine meadows. Data availability All raw sequences were submitted to the NCBI Sequence Read Archive database (No. SRP318670). Other data is available by request to the authors.
Code availability Not applicable.

Declarations
Conflicts of interest/Competing interests There are no conflicts of interest or competing interest.