The Liangtan River (29°26′–29°52′ N, 106°18′–106°24′ E) is a first-class tributary of the lower right bank of the Jialing River (Figure 1A). In a preliminary investigation, we found more than 15 barriers that led to fragmentation of the river. According to the barrier height, the Liangtan River can be divided into four habitats: pond (P) and stream (S) for low barriers (less than 2 m high, LB) and lake (L) and river (R) for high barriers (between 2 m and 5 m, HB; Figure 1; Fuller et al., 2015; Belletti et al., 2020).

Samples were collected from the four habitats of the Liangtan River (December 2020 and August 2021). Sediment samples were obtained at a depth of 10 cm below the interface sediment: water, whereas water samples were obtained from a total of 20 sampling sites in the river channels. Sediments were immediately placed in sterile bags and sealed, placed on ice bags, shipped back to the laboratory within 12 h, and stored in a −80°C freezer. The water samples were divided into two parts. The first part was supplemented with HgCl₂ (20 μg/mL) for subsequent determination of water quality indicators. The other part was returned to the laboratory within 12 h and filtered through a 0.22-μm glass fiber membrane to remove trapped microorganisms for subsequent high-throughput sequencing. Characteristics of the 20 sites studied (5 sites per habitat) in the Liangtan River are shown in Supplementary Table S1.

Electrical conductivity (EC), water temperature (T), and DO were determined by a YSI^® ProODO DO meter, and the pH was determined by a YSI^® 63 pH meter. Total organic carbon (TOC) was determined by a Shimazu^® TOC-VWP analyzer (Shimadzu^®, Japan). Chlorophyll a (Chl a) was extracted by ethanol and determined by spectrophotometry (Ortega et al., 2019). Total nitrogen (TN) was measured by the potassium persulfate oxidation method (Zhang L. Y. et al., 2021). Total phosphorus (TP) was extracted by HClO₄-H₂SO₄ and measured by the molybdenum blue method. Nitrite nitrogen (NO₂⁻-N) was determined by N-(1-)-ethylenediamine spectrophotometry. Nitrate nitrogen (NO₃⁻-N) was measured by potassium peroxodisulfate solution spectrophotometric methods. Ammonium nitrogen (NH₄⁺-N) was measured by spectrophotometry with a Nessler reagent (Ministry of Environmental Protection, 2012).

Total DNA was extracted from 0.5-g sediment samples using a Rapid Soil DNA Isolation Kit. The concentration of the extracted DNA was then quantified using an ND-2000C spectrophotometer (Nanodrop, Thermo Scientific, United States), and the extracted DNA was used for high-throughput sequencing. The V3-V4 hypervariable region of the 16S rRNA gene was amplified with 338F/806R universal primers (Supplementary Table S2), and each sample was identified. Specific PCR amplification conditions of the 16S rRNA gene were based on those described in previous studies (Junfeng et al., 2015). The resulting PCR products were then sequenced on an Illumina HiSeq 2000 platform. Data processing and statistical analysis were carried out (see the attachment for specific methods). The raw data of Illumina Miseq sequencing were submitted to the NCBI under the BioProject accession number PRJNA904922.

The concentrations of N₂O and CH₄ in water were measured using the static headspace method. In this study, the traditional method (Donis et al., 2017) was improved. Firstly, a syringe is used to extract 200 mL water sample from the water sampler. The sample should be extracted slowly during the collection process to avoid bubbles. Extract 100 mL nitrogen from an air bag containing high purity nitrogen (99.999% purity) (500 mL air bag) to form an air chamber above the syringe (the same syringe used to collect water sample). Hold the syringe and shake it up and down for 3 min to achieve a balance between the gas–liquid phase. Push the gas in the syringe into the prepared vacuum air bag (300 mL air bag) for preservation. Record the temperature of the water after the shake. The preserved headspace samples were returned to the laboratory and the concentration of N₂O and CH₄ was determined by gas chromatography.

16S rRNA and the abundance of functional genes of denitrifying bacteria were detected by qPCR using the Majorbio Cloud Platform (Shanghai). The functional genes were mainly divided into nitrite reductase (nir), NO reductase (nor), N₂O reductase (nos), and particulate methane monooxygenase (pmo). Three repeats of qPCR were performed on each sample using an ABI 7500 sequence detection system with the SYBR method (Applied Biosystems, Canada). Primer sequences are listed in Supplementary Table S2. qPCR-specific tests were performed using a melting profile and gel electrophoresis to reduce the possibility of overestimating gene abundances. In addition, each qPCR reaction consisted of negative control, in which no DNA template was added. The abundance of functional genes was converted to the number of copies of functional genes per gram of dry sediment or per milliliter of water. The efficiency was assumed to be 100% for DNA extracted from sediment or water samples (Zheng et al., 2019).

One-way ANOVA and Tukey’s post hoc tests were used to evaluate the differences in microbial diversity and abundance, structure, and environmental factors among habitat types. The above statistical analyses were conducted using PASW Statistics 18 software (IBM SPSS Inc., Chicago, United States). The molecular ecological networks of habitats S, P, R, and L were constructed to analyze the microbial community structure. Linear discriminant analysis (LDA) was used to estimate biomarkers of the four habitats based on analysis software (LEfSe). Based on Origin, a heatmap was drawn to analyze the microbial abundance related to the production of N₂O and CH₄ in the four habitats. To further elucidate the direct and indirect effects of environmental factors and microbial communities on N₂O concentrations and CH₄ concentrations, we conducted path analyses using the maximum likelihood estimation method. Gephi version 0.8.2 was used for network visualization and modular analysis. According to Banerjee et al. (2018), keystone species were identified using the following thresholds: genera with high mean degree (>30) and low betweenness centrality (<700) in the network. Path coefficients, R², direct and indirect effects, and model fit parameters were calculated using R studio.

Samples were taken from five randomly selected points in each habitat (Figures 1A–C). Through principal component cluster analysis of physical and chemical indexes of four habitats, it was found that the four habitats could be distinguished well, which indicated that it was appropriate to divide the different habitats of Liangtan River (Figure1D). The abundance and composition at the phylum level of microbial communities from water and sediment was analyzed for each habitat (S, P, R, and L; Supplementary Figure S1). A total of 14,481,930 high-quality bacterial 16S rRNA gene sequences and 30,863 operational taxonomic units (OTUs) were obtained by high-throughput sequencing.

The relative abundances of the phyla Proteobacteria, Bacteroidota, and Cyanobacteria were higher in water (17.6–76.5%, 2.8–57.9%, and 0.2–41.4%, respectively) than in sediment (7.0–45.8%, 1.3–9.2%, and 0.2–4.2%, respectively). In contrast, the relative abundances of Chloroflexi, Firmicutes, Acidobacteriota, Verrucomicrobiota, and Desulfobacterota were higher in sediment (4.3–35.3%, 2.0–31.1%, 2.2–14.8%, 1.5–10.0%, and 0.1–9.0%, respectively) than in water (0.05–6.5%, 0.7–13.5%, 0.03–0.9%, 0.1–4.5%, and 0.06–1.0%, respectively, Supplementary Figure S2). Most of these biomarkers were abundant in the sediment of S and R. There were 31 and 102 biomarkers (p < 0.05, LDA > 2.0) in water and sediment, respectively (Figure 2). The classes Acetobacteraceae, Blastocatellales, Microtrichales, and Pseudomonadales were the main biomarkers in sediment samples from S. Species belonging to Rhodobacterales, Xanthomonadales, and Chitinophagales dominated in sediment samples from R. Flavobacteriaceae and Fusobacteriaceae species were most abundant in water samples from S. Species affiliated with Veillonellales, Prevotellaceae, Nocardioidaceae, and Lachnospirales were abundant in water samples from R (Figure 2). The number of biomarkers was lower in L and P. In L, the classes Ncardioidaceae and Micrococcales were the main biomarkers in sediment, and Saccharimonadales was dominant in water. In P, the classes Norank, Fusobacteriales, and Prolixibacteraceae were the main biomarkers in sediment, whereas RBG-13-54-9, Chitinophagales, and PeM15 were dominant in water.

The topological structures of the bacterial communities at the OTU level are shown in Figure 3. Regarding the sediment samples, the percentage of Chloroflexi nodes increased in P compared with that in S, while the node ratio in Proteobacteria and Actinobacteriota decreased. Similarly, the percentage of Chloroflexi nodes increased in P compared with that in S, while the node ratio in Proteobacteria and Actinobacteriota decreased.

Then we further analyzed the variation of ecological network parameters in the four habitats. The two networks were defined according to the same threshold, and the corresponding statistical results are shown in Supplementary Table S3. In water, the modularity of the S bacterial network (1.734) was lower than that of the P bacterial network (3.147). In S, the resulting networks consisted of 334 nodes linked by 2,809 edges, which consisted of 1,816 positive correlations and 993 negative correlations. In P, the resulting networks consisted of 287 nodes linked by 2,105 edges, which consisted of 1,210 positive correlations and 895 negative correlations. Compared with R, the modularity of L was lower. In R, the resulting networks consisted of 320 nodes linked by 2,812 edges, which consisted of 1,759 positive correlations and 1,042 negative correlations. In L, the resulting networks consisted of 300 nodes linked by 2,816 edges, which consisted of 1,764 positive correlations and 1,052 negative correlations. The presence of barriers will reduce the number of nodes, and positive or negative correlations in the microbial network, leading to the estrangement of microbial connections in the water.

In sediments, the modularity of S was higher than that of P, and that of L was higher than that of R. In S, the network consisted of 370 nodes and 3,756 edges. The network showed 2,293 positive correlations, which was higher than the negative correlations (1,463 green lines). In P, the network consisted of 388 nodes and 4,789 edges. The network showed 2,967 positive correlations and 1,822 negative correlations. In R, the network consisted of 409 nodes and 4,128 edges. The network showed 2,730 positive correlations and 1,428 negative correlations. In L, the network consisted of 373 nodes and 4,733 edges. The network showed 2,735 positive correlations and 1,999 negative correlations. This suggested that the presence of obstacles may lead to closer connections of microbial networks in sediments.

In water, the abundances of anammox bacteria and nitrifiers were low, and denitrifiers (Pseudomonas) and methanotrophs (Methylocystis) were more abundant (Figure 4A). The denitrifying bacterial abundance of P was higher than that of S, and that of L was higher than that of R. The methanotrophs abundance of S was higher than that of P, and that of L was higher than that of R. In the sediment, the abundances of anammox bacteria and denitrifiers were low, the nitrifiers (Nitrosospira) and methanotrophs (Methylophilus, Methylocystis, Methylocystis) were more abundant (Figure 4B). The nitrifying bacterial abundance of S was higher than that of P, and that of R was higher than that of L. The methanotrophs abundance of S was higher than that of P, and that of L was higher than that of R.

The concentrations of dissolved N₂O in S, P, R, and L were 0.014–0.029, 0.014–0.037, 0.019–0.078, and 0.012–0.11 μmol/L, respectively (Figure 5A). Compared with S, the concentration of N₂O in P showed an increasing trend. The concentration of N₂O in L was higher than that in R. LB and HB can increase the N₂O concentration in rivers by 1.13 and 1.29 times, respectively. The concentrations of dissolved CH₄ in S, P, R, and L were 0.30–4.18, 0.04–0.67, 0.03–2.4, and 0.03–34.0 μmol/L, respectively (Figure 5B). The CH₄ concentration in S was higher than that in P, and that in L was higher than that in R, it can be inferred that that LB reduces CH₄ accumulation and HB increase it. LB reduced the CH₄ concentration by 0.118 times, while HB increased it by 2.76 times.

Next, we conducted path analyses to further elucidate the direct and indirect effects of environmental factors and microbial communities on N₂O concentrations and CH₄ concentrations. DO and NO₃⁻-N could regulate the concentration of N₂O in water both directly and indirectly (Figure 6A). The regulation of DO to N₂O is reverse, while that of NO₃⁻-N is positive. DO (β = −0.18, standardized coefficient) directly impacted the concentration of N₂O. NO₃⁻-N was the most significant parameter (β = 0.76, standardized coefficient) influencing the concentration of N₂O. NO₃⁻-N indirectly through water denitrifying communities (β = 1.32, standardized coefficient) impacted the concentration of N₂O. Furthermore, TN indirectly through sediment nitrifying bacterial community (β = 0.41, standardized coefficient) impacted the concentration of N₂O. Moreover, DO directly (β = −0.35, standardized coefficient), and indirectly through soil pH (β = −0.31, standardized coefficient), impacted the concentration of CH₄ (Figure 6B).

As can be seen from the above analysis, the LB and HB promote the enrichment of denitrifiers in the P and L habitat. To further verify our hypothesis, the abundances of functional enzyme genes of denitrifiers of water were analyzed in four habitats. The abundances of nirS, nirK, and nosZ in water and sediment were counted (Supplementary Tables S6–S9). To estimate the abundance contribution of different genes in the denitrification pathway to N₂O accumulation, the nisS/nirK and (nisS + nirK)/nosZ copy number ratios were analyzed (Figure 7). The nirS/nirK of P and L was higher than that of S and R, suggesting that the LB and HB promote the enrichment of nirS-type denitrifiers, and then affect the accumulation of N₂O. In sediments (nirS + nirK)/nosZ copy number ratios were decreased in P and L compared with S and R, suggesting that the LB and HB promoted the nosZ enrichment in sediments.

Then, the abundances of pmoA in water and sediment were counted (Supplementary Tables S6–S9). To estimate the contribution of functional genes to the abundance of CH₄ accumulation in CH₄ oxidation, we analyzed the pmoA/16S rRNA copy number ratio in water and sediment of four habitats (Figure 7). pmoA is a key enzyme gene for CH₄ oxidation, which reduce CH₄ accumulation (Lieberman and Rosenzweig, 2004; Vrieze and Verstraete, 2016). In water, the pmoA/16S rRNA copy number ratio of R was the highest, while in sediment, the pmoA/16S rRNA copy number ratio of P was the highest. These results further suggest that the presence of small barriers affects CH₄ accumulation by altering the activities of enzymes involved in methane oxidation.

LB and HB lead to lower flow rates and increased water depth in fragmented rivers, which change environmental variables and affect N₂O and CH₄ accumulation. The small barrier leads to an increase in the concentration of NO₃⁻-N (Supplementary Tables S4, S5), which promotes the accumulation of N₂O in P and L. Both LB and HB reduce river velocity, leading to an increase in the water bed contact area per unit of water volume, and an increase in the diffusion efficiency of NO₃⁻-N at the river-sediment interface (Mulholland et al., 2008). This is consistent with previous work showing that nitrate concentrations strongly affect the production of N₂O (Meyer et al., 2008). With few exceptions, the nitrate concentration in sediment and surface water is positively correlated with the production of N₂O (Quick et al., 2019). The positive correlation between N₂O and NO₃⁻-N might stem from incomplete denitrification (Beaulieu et al., 2011; Abed et al., 2013; Jung et al., 2019). LB and HB promoted the enrichment of denitrifiers in the water of P and L respectively, which may lead to an increase in incomplete denitrification under unit nitrogen load (Mulholland et al., 2008).

Moreover, in most rivers, the overlying water above the surface of the sediment is anoxic and carries dissolved oxygen that continuously seeps into the sediment (Xia et al., 2018). The depth of dissolved oxygen penetration into sediment is affected by sediment roughness, porosity, and connectivity (Xia et al., 2018). LB and HB lead to increased water depth and decreased DO concentration in the river (Supplementary Figure S3), which is conducive to incomplete denitrification.

In addition, the higher abundance of nitrifiers (Nitrosospira) in the lotic habitats (S and R) suggest that the LB and HB would inhibit the growth of nitrifiers. The main reason is that the barriers reduce the DO in the river, thus inhibiting the growth of nitrifiers (aerobic bacteria). At present, there are still many uncertainties about the relationship between nitrifiers and N₂O accumulation (Beaulieu et al., 2010). According to our results, the decrease of nitrifiers promotes N₂O accumulation to a certain extent in river surface sediments.

According to the CH₄ concentration in four habitats, the CH₄ concentration in P was lower than in S. The main reason was that the abundance of methanotrophs in P was higher than that in S, resulting in a large amount of CH₄ consumption. It can be seen that the LB reduces CH₄ accumulation. Compared with R, the CH₄ concentration in L was higher. The potential reason was that the concentration of DO in L was reduced, which was conducive to the generation of CH₄ under anaerobic conditions. It can be seen that the HB increases the accumulation of CH₄, these results were consistent with other literature reports (He et al., 2021).

Denitrifiers and methanotrophs were abundant in the four habitats and significantly affected N₂O production and CH₄ oxidation, respectively. Hence, a co-occurrence network analysis was conducted to identify the cooperative and competitive relationships among denitrifiers, methanotrophs, and other microorganisms. In networks of P, Proteobacteria was still the first dominant phylum, and the second dominant phylum was Cyanobacteria. Cyanobium_PCC-6307 is a genus with high content of Cyanobacteria phylum. It has been reported that the abundance of denitrifiers has a competitive relationship with non-diazotrophic Cyanobium sp. (Song et al., 2022), so it is further speculated that LB increases the abundance of Cyanobium, further limits complete denitrification and increases N₂O accumulation. Interestingly, denitrifiers (Pseudomonas) and methanotrophs (Methylocystis, Methylophilus, Methylotenera), were dominant nodes in water. Previous studies have also reported low N₂O and high CH₄ fluxes in rivers, which they suggest are due to competition (Deemer et al., 2016). The potential cause is the production of a copper chelator (methanobactin) by methanotrophs during the CH₄ cycle (Spirito et al., 2016), which has been shown to effectively compete for copper from denitrifiers, thereby increasing N₂O production while reducing CH₄ emissions (Chang et al., 2018). Similarly, the availability of copper is particularly important for nosZ expression and activity, since copper is required for the active site of the enzyme (Gaimster et al., 2018). nosZ will compete with methanotrophs for copper ions, thus reducing N₂O production and increasing CH₄ accumulation. This is consistent with our results that the ratio of nosZ and concentration of CH₄ in S and L is higher than that of P and R (Supplementary Figure S4).

In sediments, the Chloroflexi was the dominant bacteria in P and L habitats, it was speculated that the accumulation of Chloroflexi may promote the accumulation of N₂O. It has been reported that bacteria attached to Chloroflexi may be highly active protein degradation, breaking down the extracellular peptides bound to the extracellular polymer matrix and simultaneously breathing nitrate to produce nitrite (Lawson et al., 2017). The increase of nitrite can promote denitrification (Quick et al., 2019), and thus increase the accumulation of N₂O, which further proves that the barriers can increase the accumulation of N₂O. Unlike in water, nitrifiers (Nitrosospira) and methanotrophs (Methylocystis, Methylophilus, Methylotenera), were dominant nodes in sediments. It has been reported that nitrifiers can be inhibited by methanotrophs due to competition between nitrifiers and methanotrophs for available oxygen and inorganic nitrogen (Megmw and Knowles, 1987). As LB and HB reduce DO concentration, nitrifiers in P and L were more severely inhibited compared with S and R, resulting in the stronger activity of methanotrophs. On the other hand, ammoxidation microorganisms may consume CH₄, thereby reducing the amount of CH₄ oxidized by methanotrophs (Bodelier and Laanbroek, 2004), and thus competing with methanotrophs for CH₄. Therefore, the decrease of CH₄ content in habitats P and R was also related to the nitrifiers.

In order to further analyze the accumulation patterns of N₂O and CH₄ in fragmented rivers, the potential metabolic pathways of N₂O and CH₄ in different habitats were further analyzed. Compared with microbial functional groups, the metabolism between functional enzyme genes can better reveal the coupling relationship of element metabolic processes in environmental media. First, the potential N₂O metabolism pathways of microorganisms in four habitats were analyzed. The analysis of the metabolic processes of nitrogenous microorganisms in four habitats revealed that the content of enzymes ([EC:1.7.7.2]) involved in nitrate transformation processes was higher in P compared with S in water (Figure 8), and this promoted the generation of N₂O. This further proves that LB promotes the generation of N₂O. The expression of enzyme genes involved in the dissimilatory nitrate reduction to ammonium (DNRA) process (Li et al., 2019) ([EC:1.7.1.15], [EC:1.7.7.1], [EC:1.7.2.2]) was higher in S than in P in sediment. The results indicated that more NO₂-was converted into NH₄⁺ in S, which reduced the possibility of N₂O formation. The content of enzymes ([EC:1.7.2.6]) involved in nitrate transformation processes was higher in L compared with R in water and sediment, and indirectly promotes the generation of N₂O. The expression of enzyme genes involved in the DNRA process ([EC:1.7.1.15]) was higher in R than in L in sediment. The results indicated that more NO₂⁻ was converted into NH₄⁺ in R, which reduced the possibility of N₂O formation. Some studies have shown that the growth rate of a new class of non-denitrifying N₂O reductants may be slow, but the isolated strains have great metabolic flexibility, enabling them to grow through DNRA (Sanford et al., 2012; Jones et al., 2014), which plays an important role in N₂O reduction. It indicates that both LB and HB can promote the DNRA process, thus increasing the accumulation of N₂O.

The analysis of the metabolic processes revealed that the content of enzymes (EC:1.17.1.10; Li et al., 2019) involved in CH₄ oxidation processes was higher in P compared with S in sediment (Figure 8), and directly reduced the generation of CH₄. It further shows that LB inhibits CH₄ accumulation by promoting CH₄ oxidation. The enzymes (EC:1.14.18.3; Zhang et al., 2020) involved in CH₄ oxidation processes were higher in R compared with L in water and sediment, and directly reduced the accumulation of CH₄. It further shows that HB increases CH₄ accumulation by inhibiting CH₄ oxidation. Some studies have speculated that aerobic CH₄ oxidation coupled to denitrification process and anaerobic nitrite-dependent CH₄ oxidation processes in river sediments (Zhang et al., 2020). It was speculated that the barriers may affect these two processes and thus the accumulation of N₂O and CH₄.

In summary, we preliminarily sorted out the accumulation patterns of N₂O and CH₄ in fragmented rivers (Figure 9). It was found that N₂O accumulated continuously with the increase of barriers height, and the potential reasons were related to enrichment of Cyanobium, Chloroflexi, and nirS-type denitrifiers, as well as the increase of NO₃⁻-N concentration and decrease of DO concentration in water (Figure 9). The reason for the decrease of CH₄ concentration caused by the LB may be related to the compete with Pseudomonas in water. The HB can lead to the accumulation of CH₄, which was largely related to the methanotrophs to compete with Nitrosospira in sediment and the decrease of DO concentration.