Located on the Loess Plateau in Northwest China, the Wei River is the largest tributary of the Yellow River, with a total length of 818 km, flowing through Gansu and Shaanxi provinces (Su et al., 2019). The Shaanxi section of the Wei River is 502 km long, which provides water for 60% of the population and 56% of the cultivated land in Shaanxi Province (He et al., 2020). The water flow in this section is from west to east, and the lithology of the riparian sediment gradually becomes finer from the upstream Baoji section (pebbles, gravels) to the downstream Weinan section (fine sand, silt), and the flow rate becomes slower. Affected by the input of excessive nitrogen fertilizer, inorganic nitrogen load (mainly NO₃⁻) has become the main source of non-point source pollution in the Shaanxi section of the Wei River, and the nitrogen pollution problem in the downstream Weinan section is more serious (Wang et al., 2018; Shi et al., 2019).

The riparian zone of the Weinan section was selected as a typical heterogeneous profile and sampling points were set up (shown in Figure 1). According to the natural stratification of the riparian zone profile, the sediment samples of different layers were collected using a portable sediment sampler (Rigo Co., Ltd., Saitama, Japan) after disinfection with anhydrous alcohol. One group of samples (about 10 kg) was collected from each layer, and a total of 5 groups of samples were collected. The samples numbered from top to bottom were W1 (0–50 mm), W2 (50–150 mm), W3 (150–200 mm), W4 (200–300 mm), and W5 (300–350 mm). Among them, W1, W3 and W5 layers were mixed with black sludge. In order to avoid pollution during storage and transportation, the collected samples were packed into sterile self-sealing bags and sterile tube tubes (50 mL), respectively. The samples of sterile self-sealing bags were used for experiments and physical and chemical properties analysis, and were stored at −4°C. The samples of sterile tube tubes were used to determine microbial diversity and abundance, and were stored in an ultra-low temperature refrigerator at −80°C.

As the sediment samples were collected, the pH value of the sediment was immediately measured using a pH400 portable soil pH meter (Spectrum Technologies, United States), the conductivity and total dissolved solids of the sediment were measured using a 2265FS conductivity meter (Spectrum Technologies, United States), and the water content (WC) and redox potential (ORP) of the sediment were measured using a soil multi-parameter tester (Delta-T WET-2, United States). In addition, some sediment samples were taken from each group to remove debris such as trunks, and then dried naturally and screened by a 2.0 mm filter. The treated samples were analyzed by laser particle size analyzer Mastersizer2000 (Malvern Instruments Ltd., United Kingdom), and the mineral structure of the samples was analyzed by SEM scanning electron microscopy (Thermo Fisher QUANTA 250PEG, United States) (Zhou et al., 2016). The dissolved organic carbon (DOC) in sediments was determined by total organic carbon analyzer (Shimadzu TOC-VCPH, Japan), and the dissolved inorganic carbon (DIC) was measured using a dissolved inorganic carbon analyzer (ApolloAS-C3, United States). The Fe(II) was determined by phenanthroline spectrophotometry (Zheng et al., 2020). After extraction with 2 mol L⁻¹ KCl, the contents of total nitrogen (TN), ammonia nitrogen (NH₄⁺-N), nitrate nitrogen (NO₃⁻-N) and nitrite nitrogen (NO₂⁻-N) in sediments were determined by UV-1800 ultraviolet spectrophotometer (Shimadzu, Kyoto, Japan), with detection limits of 0.015, 0.015, and 0.03 mg/kg, respectively. Dissolved organic nitrogen (DON) was determined by continuous flow analyzer (Skalar SAN++, Netherlands). The sample test was carried out in parallel for three times. The test results were expressed as the mean of three tests, and the error range was less than 5%.

After thawing and mixing the sediments at each sampling point, 0.25 g was weighed, and then DNA was extracted in strict accordance with the instructions of PowerSoil ^® DNA separation kit (Mo Bio Carlsbad, United States). The concentration and purity of DNA were detected by NanoDrop2000 spectrophotometer (Thermo Scientific, United States), and the qualified DNA samples were analyzed by fluorescence quantitative (qPCR). Sequencing was performed using the Illumina Miseq platform (Personal Biotechnology Company, Shanghai, China). The abundance of genes (nirS, norB, Amx-16S, nrfA) related to nitrogen removal processes (i.e., denitrification, anammox and DNRA) in sediment samples was determined by SYBR-Green dye using ABI7500 rapid real-time fluorescence quantitative PCR instrument (Applied Biosystems, Foster City, CA, United States). The Table 1 provides detailed information on primers, and three qPCR analyses were performed for each sample for error analysis.

In order to quantitatively identify the nitrogen removal process in riparian zones, ¹⁵N isotope tracing experiments were carried out using isotope pairing technique (IPT) (Nielsen, 1992). The sediment and sterile water were mixed at a mass ratio of 1: 7 to make a uniform mud and filled with helium for 30 min, and then the mud was filled into a series of airtight serum bottles with syringes (Exetainer, Labco, United Kingdom). The bottles were placed in a dark environment for 24 h, and the remaining NO₃⁻-N and NO₂⁻-N in the bottles and DO in deionized water were consumed. 3 serum bottles were randomly selected after culture, 200 μL 50% ZnCl₂ solution was added to inhibit microbial activity, and the remaining NO₃⁻-N and NO₂⁻-N concentrations in the bottles were extracted and determined. The remaining serum bottles were divided into 4 groups, numbered E1, E2, E3 and E4. The 100 μL helium-aerated isotope reagents were injected into the first three groups with syringes: E1 (¹⁵NH₄⁺ standard solution with 99.6% ¹⁵N), E2 (¹⁵NH₄⁺+¹⁴NO₃⁻ standard solution, at a volume ratio of 1,1) and E3 (¹⁵NO₃⁻ standard solution with 99% ¹⁵N) (Zhang et al., 2020), and the final concentration of standard solution in each group was 100 μmol/L. Then, 200 μL of 50% ZnCl₂ solution was injected into half of the bottles in each group as the initial sample, and 200 μL of 50% ZnCl₂ solution was injected into the remaining serum bottles after 8 h of culture as the termination sample. The concentration of dissolved N2 (²⁸N₂, ²⁹N₂, ³⁰N₂) in the supernatant was determined by MAT253 stable isotope ratio mass spectrometer (Thermo Fisher Scientifie Inc., MA, United States) to calculate the denitrification and anammox rates.

The E4 was used to determine the DNRA reaction rate, and the injection solution was K¹⁵NO₃ (99.0 atom%, final concentration approximately 100 μmol / L) aerated by helium, in addition, the starting and ending samples were filled with high-purity helium for aeration for 30 min to remove N₂ (²⁹N₂, ³⁰N₂) produced by denitrification and anammox in the bottle (Deng et al., 2015). Then 200 μL HBrO was injected into the aerated serum bottle to completely oxidize ¹⁵NH₄⁺ to ²⁹N₂ and ³⁰N₂, and the concentration of N₂ and ¹⁵NH₄⁺ in E4 was determined by mass spectrometer and ultraviolet spectrophotometer (Soonmo and Wayne, 2002; Hou et al., 2013; Yin et al., 2014), all samples needed three parallel experiments. Denitrification (D_total), anammox (A_total) and dissimilatory ammonium reduction (R_DNRA) rates (μmol·N·kg⁻¹·h⁻¹) were calculated from the following equations (Nils Risgaard-Petersen et al., 2003; Trimmer et al., 2006):

Where Q₃₀ is the formation rate of ³⁰N₂ measured by mass spectrometer, F_N is the proportion of ¹⁵NO₃⁻ in total NO₃⁻ after adding ¹⁵NO₃⁻, Q₂₉ is the formation rate of ²⁹N₂ measured by mass spectrometer, [¹⁵NH₄⁺]₁ is the concentration of ¹⁵NH₄⁺ in the system at the initial time, [¹⁵NH₄⁺]₂ is the concentration of ¹⁵NH₄⁺ μmol/L in the system after 8 h of culture, W is the sample mass kg, and T is the culture time h. Analysis of variance (ANOVA) was used to examine the spatial and temporal differences in denitrification, anammox and DNRA rates.

In order to further explore the role of Fe(II) and DOC in the N-loss process of riparian zone, batch static culture experiments were carried out in anaerobic glove box (COY-7150220, United States) by controlling variables. The samples were evenly mixed with each layer of sediment samples in riparian zone. Anaerobic deionized water (Ar/N₂ for 20 min) was mixed with sediment samples at a ratio of 3: 1, and 12 g of the above homogeneous mixed slurry was taken into a 100 mL serum bottle (in triplicate). After standing, 80 mL of pure water was added, and then Ar/N₂ was introduced for aeration for 20 min, and the bottle was filled with anaerobic deionized water. Cultured under anaerobic dark conditions for 2 days to remove the sample background to remove nitrate, nitrite and oxygen from the sample background. Different exogenous substances were added to each bottle (details in Supplementary Tables S1, S2), and anaerobic dark culture was performed for 31 days. The supernatant was collected regularly with a 1 mL syringe (interval of 24 h on the first 15 days, and interval of 48 h on the 16th day) to determine the concentration of Fe(II), nitrate nitrogen, nitrite nitrogen and ammonia nitrogen by the phenanthroline spectrophotometry and UV-1800 ultraviolet spectrophotometer (Shimadzu, Kyoto, Japan), all samples need three parallel experiments. After each sampling, it was sealed with a sealing film to prevent oxygen pollution.

In this study, the OriginPro 2022 and R language were used for graphic drawing and analysis. One-way analysis of variance (ANOVA) was used to test the differences in nitrogen conversion rate, microbial functional genes and sediment physical and chemical indicators (95% confidence interval). The SPSS software package (version; 25 IBM, United States) was used to estimate the relationship between different variables through pearson correlation and linear regression equation. The significance level of all statistical analyses was set as p < 0.05.

The lithology structure of the riparian zone in the study area is heterogeneous in the vertical direction (Table 2). The soil particle size ranged from 1.98 to 215.9 μm, mainly silt (4–63 μm), with clay (>63 μm), silt, and fine sand (<4 μm) contents accounting for 0–0.03%, 72.15–98.1%, and 1.86–27.85%, respectively. There were abundant minerals in the sediments of the riparian zone, mainly quartz (55.9–64.1%) and plagioclase (14.8–20.0%). Followed by calcite (6.0–8.6%), chlorite (4.7–6.6%), illite (3.4–5.4%), potassium feldspar (1.7–4.4%), dolomite (1.3–4.1%) and a small part of amphibole, which indicated that the parent rock of the sediment was completely weathered and the mineral maturity was high.

The soil particles of the riparian zone sediments had a high degree of polymerization, and micro-aggregates were observed in the sediment W1, W3, and W5 layers containing sludge, and there were obvious bridging substances between the soil particles (Supplementary Figure S1), which greatly hindered the uptake of organic matter by microorganisms, protected the ‘comprehensiveness’ of soil organic matter, and formed natural organic carbon. Therefore, the DOC content in this area was high (Table 2). In addition, the pH range of the sediment environment in the riparian zone was 7.77–8.02, which was weakly alkaline overall. Along the vertical direction of the riparian zone, conductivity (EC) and total dissolved solids (TDS) showed a decreasing trend, Fe(II) gradually increased, and oxidation reduction potential (ORP) gradually decreased, and the reducibility of the sedimentary environment in the riparian zone gradually increased.

As shown in Figure 2, the distribution of nitrogen content in different depths of sediments in the riparian zone of the Weinan section of the Wei River varied greatly, with the total nitrogen (TN) content ranged from 11.8 to 15.31 mg/kg, and the DON content ranged from 7.21 to 11.42 mg/kg, which was the main form of nitrogen. With the increase of sediment depth in the riparian zone, TN showed a downward trend, DON showed an increasing trend (Figure 2A), and dissolved organic nitrogen (DIN) content gradually decreased (Figure 2A), which confirmed that the riparian zone had a removal effect on inorganic nitrogen pollution.

It had been shown that the migration and transformation of nitrogen in the riparian zone is mainly completed by the geochemical biological effects of NH₄⁺-N, NO₃⁻-N and NO₂⁻-N (nitrification, denitrification, anaerobic ammonium oxidation, etc.) (Zhao et al., 2021). In the riparian sediments of Weinan section of Wei River, the NO₃⁻-N, NO₂⁻-N and NH₄⁺-N contents were 0.06–1.96 mg/kg, 0.03–1.71 mg/kg and 0.42–3.12 mg/kg, respectively. In layers W1, W3, and W5, which contained sludge, the DIN in the sediments mainly existed in the form of NO₂⁻-N and NH₄⁺-N, and the NO₃⁻-N content was low and reached a minimum value of 0.06 mg/kg in the lowest layer W5, while in the sediment layers W2 and W4, the content of NO₃⁻-N was significantly higher than that of NO₂⁻-N and NH₄⁺-N, which indicated that there was a complex nitrogen cycle in the sediments of the riparian zone in Weinan section.

Metagenomics can determine the composition and function of microbial communities (Daims et al., 2015; Chen et al., 2022). In this study, genes associated with denitrification (nirS, norB), anammox (Amx-16S) and DNRA (nrfA) reactions were detected in the sediments of the riparian zone of the Weinan section of the Wei River by qPCR, which indicated that the riparian zone may remove nitrogen from the environment through denitrification, anammox and DNRA processes. Among them, denitrification combined with anammox removed NO₃⁻-N ‘permanently’ from the riparian zone of the Wei River, while DNRA converted NO₃⁻-N into inorganic nitrogen and retained it in the riparian zone for other organisms (Crowe et al., 2012; Deng et al., 2015). The Coverage index of each sample in this study was greater than 0.99, indicating that the determination results of sediment samples had high confidence (Supplementary Table S3). There were significant differences in the diversity and abundance of microbial communities involved in the nitrogen removal process along the sediment depth (Figure 3).

The diversity of denitrifying bacteria community in the riparian zone of Weinan section was the highest, and Proteobacteria was the most important denitrifying bacteria (Figures 3A,B), which was the same as the results of other freshwater system sediments (Song et al., 2012; Dong et al., 2021). Aerobic denitrifying bacteria such as Azoarcus, Ochrobactrum, Pseudomonas, Paracoccus, and Azospira are the main dominant genera (Figures 3A,B). Which was attributed to the interaction between river water and groundwater. The transport of dissolved oxygen to sediments provided an oxidizing environment for microorganisms. The high redox potential in sediments also confirmed this view (Table 2). These aerobic microorganisms use oxygen/NO₃⁻-N/NO₂⁻-N as an electron acceptor for denitrification under aerobic conditions (Huang et al., 2021; Zhang et al., 2022).

In addition, the abundance of nirS gene was 2.01 × 10⁴–6.92 × 10⁵ copies g⁻¹, and the abundance of norB gene was 1.74 × 10⁵–3.68 × 10⁵ copies g⁻¹ (Figure 4). As a nitrite reductase marker gene nirS (Wang et al., 2019), the abundance of nirS gene was the same as the distribution of NO₂⁻-N content (Figures 2A, 4A). With the increase of sediment depth, the abundance of nirS gene decreased first and then increased, and then decreased gradually, and decreased to the minimum in W5 layer (Figure 4A). As a marker gene of nitric oxide reductase (Liu et al., 2017), the abundance of norB gene was similar to the distribution of N₂O content (Figure 4B and Supplementary Figure S2). The abundance of norB gene decreased sharply from W1 layer, and decreased to the minimum value in W2 layer and then increased gradually. It was shown that sediment denitrification capacity was positively correlated with the abundance of the nirS gene (Mosier and Francis, 2010). We also found that the spatial distribution pattern of nirS gene abundance and denitrification rate in the riparian zone of Weinan section was the same (Figures 4A, 5), which was higher in W1 and W3 layers and lower in other layers, which further indicated that the abundance of nirS gene could be approximated as an indicator of denitrification capacity in sediments.

The dominant genera of anammox bacteria in the riparian zone of Weinan section of Wei River were Candidatus Brocadia, Candidatus Jettenia, Candidatus Kuenenia and Candidatus Scalindua (Figure 3C). Most of these dominant bacteria were autotrophic, which used inorganic carbon as a carbon source in the anaerobic environment, and were anaerobic ammonia-oxidizing bacteria commonly found in terrestrial ecosystems (Humbert et al., 2010). The lowest diversity and relatively homogeneous composition of anammox bacteria were found in the sludgy layers W1, W3 and W5 (Figure 3C), which could be attributed to the fact that the high DOC content in the sludgy layer inhibited the growth of anammox bacteria (Guven et al., 2005; Chamchoi et al., 2008). The abundance of anammox bacteria 16S rRNA gene was the highest in riparian sediments, which was two orders of magnitude higher than the abundance of denitrifying functional genes, ranging from 2.03 × 10⁶ to 3.85 × 10⁷ copies g⁻¹ (Figure 4C), and gradually decreased along the sediment depth.

The dominant genera of DNRA bacteria in riparian zone were mainly Sorangium, Luteitalea, Lacunisphaera and Corallococcus (Figure 3D), which were similar to the reports of DNRA bacterial diversity in other rivers and lakes (Yuan et al., 2022; Zhao et al., 2022), and most of these microorganisms were heterotrophic (Wang et al., 2020a). Among them, Sorangium is an aerobic bacterium belonging to Myxococcales, which can use NO₃⁻-N/NO₂⁻-N to produce NO (Banerjee et al., 2018), and is considered to be a key microorganism for synergistic control of soil microbial composition. Luteitalea belongs to Acidobacteria, which are aerobic and chemoheterotrophic bacteria, and can use organic matter in the environment as a carbon source for life activities (Vieira et al., 2017; Qiao et al., 2020). Compared with other functional genes, the abundance of nrfA gene in DNRA bacteria was the lowest, ranging from 3147.43–6221.28 copies g⁻¹ (Figure 4D). Meanwhile, along the depth direction of the sediment, the abundance of nrfA gene was completely consistent with the distribution of DOC content (Figure 4D and Table 2). DOC, as an important carbon source of DNRA, can promote DNRA reaction (Broman et al., 2021), which is an important factor in regulating DNRA in riparian zone.

Microorganisms are important mediators for nitrogen migration and transformation in riparian sediments (Kuypers et al., 2018; Liu et al., 2021). Our observations showed that heterogeneous riparian zones can provide microhabitats for denitrifying, anaerobic ammonia oxidation and DNRA bacteria. The differences in the diversity and abundance of these communities in sediment layers with different lithologic structures may affect the nitrogen removal process of the riparian zone.

As shown in Figure 5, we calculated the denitrification rate (D_total), anammox rate (A_total) and DNRA rate (R_DNRA), which further confirmed that the natural attenuation process of nitrogen in the riparian zone of the study area was achieved by the combined action of these three processes. Unlike other regions where denitrification and anammox reactions are the main N-loss pathways in sediments (Choi et al., 2016; Liu et al., 2019; Zhang et al., 2020; Zheng et al., 2020), the contribution rate of anammox to nitrogen removal in the riparian zone of Weinan section of Wei River was extremely low (about 12.2%), and A_total was 4.14–5.77 μmol·N·kg⁻¹·h⁻¹ (Figure 5), which was related to the unique properties of the riparian zone. Under the interaction of river water and groundwater, the sedimentary environment of the riparian zone was oxidizing and heterogeneous, while anammox bacteria were extremely sensitive and had extremely high requirements for environmental stability, using inorganic carbon (DIC) as a carbon source for anammox reactions in low or even anaerobic environments (Jin et al., 2012; Huang et al., 2022), and the content of DIC in the riparian zone was low (Table 2), thus the activity of anammox bacteria was suppressed. Even though the oxidation of the environment along the sediment depth direction became weaker, the abundance of anammox bacteria also decreased (Figure 4) and A_total remained low. However, the anammox process contributed 14–19% of the N₂ production (Figure 5), and its nitrogen removal capacity cannot be ignored.

In contrast, denitrification and DNRA played a major role in the nitrogen removal process in the riparian zone of Weinan section of Wei River, and their nitrogen removal contribution rates were about 59.6 and 28.2%, respectively. Among them, D_total was 17.76–35.19 μmol·N·kg⁻¹·h⁻¹, R_DNRA was 9.37–12.47 μmol·N·kg⁻¹·h⁻¹ (Figure 5). DNRA and denitrification belong to the dissimilatory reduction process of NO₃⁻-N, and both of them use NO₃⁻ as the electron acceptor with substrate competition (Yin et al., 2017; Wang et al., 2020b; Wang Z. et al., 2022). It is generally believed that denitrification requires less free energy than DNRA action in the same environment, and thus preferentially occurs in most cases (Kartal et al., 2007; Chen et al., 2015). Meanwhile, the community diversity and abundance of denitrifying bacteria in riparian zone sediments were much higher than that of DNRA bacteria (Figures 3, 4), thus denitrification dominated the N-loss process in riparian zones (Figure 5).

In addition, we also observed that both D_total and R_DNRA showed significant spatial distribution characteristics. Along the sediment depth direction, D_total decreased first and then increased, and then gradually decreased (Figure 5), which was similar to the variation of denitrifying bacteria (especially nirS gene) abundance and its contribution rate to nitrogen removal (Figure 4A). The R_DNRA fluctuated slightly, but with the increase of sediment depth, the contribution rate of DNRA to nitrogen removal increased significantly, and reached the maximum value of 35.7% in the bottom layer W5 (Figure 5), which we speculated might be related to the competitive mechanisms of denitrification and DNRA. Studies have shown that DOC: NO₃⁻ (electron donor: acceptor) is a key factor in regulating denitrification and DNRA (Tiedje et al., 1983; Rütting et al., 2011), but in the sediment of a single medium, it was not DOC: NO₃⁻ but resource concentration (NO₃⁻) that determines the choice of DNRA and denitrification pathways, and when the content of NO₃⁻ becomes more limited, it is more conducive to DNRA (Vuono et al., 2019). Although the sediments in the riparian zone of the Weinan section were heterogeneous, the lithology structure of the bottom W5 layer was relatively simple, and the silt (accounting for about 98%) was the main lithology structure (Table 2). Meanwhile, the minimum NO₃⁻ content in the W5 layer was only 0.06 mg/kg, which might promote the NO₃⁻ to select the reduction path of DNRA and increase the contribution rate of DNRA in this layer.

Previous studies have shown that the abundance of functional genes is a good predictor of nitrogen transformation rate (Wang et al., 2017; Zhang et al., 2018). Interestingly, in this study, although the abundance of 16S rRNA gene of anammox bacteria was the highest and the abundance of nrfA gene of DNRA bacteria was the lowest, the anammox rate in riparian sediments was significantly lower than that of DNRA (Figure 5). Therefore, the abundance of microbial genes cannot systematically quantify the biogeochemical process of nitrogen. The nitrogen cycle in riparian zone is not only determined by microorganisms, but also related to the nature of riparian zone itself, environmental factors and specific nitrogen transformation process. This study shows that molecular biology techniques coupled with isotope tracing methods can better quantitatively reveal the N-loss process in the riparian zone.

To evaluate the nitrogen removal efficiency of the riparian zone of Weinan section of Wei River, we compared the nitrogen removal rate of sediments from different aquatic ecosystems in various regions of the world (Table 3). We found that the nitrogen removal rate of sediments in the Weinan section of the Wei River was 4.14–35.19 μmol·N·kg⁻¹·h⁻¹, which was higher than that of sediments in most regions, including wetlands, continental shelf and estuary sediments. This indicates that the riparian zone of Weinan section has a strong capacity to reduce inorganic nitrogen pollution. We also found that the nitrogen removal rate of riparian sediments in the study area was similar to that of estuarine turbidity maximum zone (TMZ) and wetland. In recent years, TMZ has been considered as an important place to remove reactive nitrogen from land to sea due to its large number of suspended particles that can easily adsorb reducing cations to consume oxygen, provide a growth environment for anaerobic microorganisms, and promote denitrification and anammox reactions (DeVries et al., 2012; Zheng et al., 2021), and wetland is considered to be a hot spot for nitrogen removal due to its high organic carbon and anaerobic environment prone to denitrification (Cheng et al., 2020). Therefore, the sediments in the riparian zone of the Weinan section share some common characteristics with TMZ and wetland, such as high organic matter content, low oxygen concentration and diverse microbial communities.

Iron and DOC are important components in the riparian zone, and they can regulate the nitrogen transformation process in environmental media (Wu et al., 2021; Lu et al., 2022), which plays an important role in the natural attenuation of nitrogen (Yang et al., 2018; Lu et al., 2022). Large amounts of Fe(II) and DOC were present in the riparian zone of the Weinan section of the Wei River (Table 2), and we investigated the effects of Fe(II) and DOC on the process of natural attenuation of NO₃⁻-N in the riparian zone through microcosmic incubation experiments.

We observed that compared with the in situ riparian zone sediments and the sediments supplemented with Fe(II) combined with sterilization, the riparian zone sediments supplemented with Fe(II) only had higher NO₃⁻-N reduction rates (Figure 6A), which proved that there may be a microbial-driven nitrate-dependent ferrous oxidation process (NDFO) in the riparian zone, which can reduce NO₃⁻ to NO₂⁻, N₂O, and N₂. Additionally, Supplementary Figure S3 showed that the riparian zone sediment consumed a large amount of DOC in the system after the addition of Fe(II), which further indicated that the microorganisms involved in the NDFO process were mainly chemoenergetic, inorganic allotrophic or parthenogenic and used DOC as a carbon source. In the first 10 days of incubation, the contents of NO₃⁻-N and Fe(II) in the system showed a sharp decline, and the NO₃⁻-N was completely consumed to the minimum (0.07 mg/L, Figure 6B and Supplementary Figure S4), while the contents of NO₂⁻-N and NH₄⁺-N showed a significant upward trend (Figure 6B). This result indicated that the NDFO reaction (Fe(II) + NO₃⁻ → Fe(III) + NO₂⁻) was carried out in the riparian zone at this stage, which may be accompanied by denitrification (NO₃⁻ → NO₂⁻), nitrogen mineralization (DON → NH₄⁺) and DNRA reaction (NO₃- → NH₄⁺). During the 11–30 days of culture, the content of NO₃⁻-N in the sediments remained at the minimum value, and the content of NO₂⁻-N showed a significant downward trend, while the content of NH₄⁺-N increased first and then decreased, and the content of Fe(II) decreased first and gradually increased after 20 days (Figure 6B and Supplementary Figure S4), indicating that NDFO (Fe(II) + NO₂⁻ → Fe (III) + N₂), iron ammonia oxidation (Feammox, Fe(III) + NH₄⁺ → Fe(II) + N₂), and anammox (NO₂⁻ + NH₄⁺ → N₂) may exist at this stage. Thus, the Fe-N cycle exists in the riparian zone sediments and Fe(II) accelerates the attenuation of NO₃⁻-N in the riparian zone sediments through the NDFO reaction driven by microorganisms, while the coupling of Feammox with the NDFO process achieved the recycling of Fe within the system.

We also observed that under different carbon-nitrogen ratios (DOC: NO₃⁻), the content of NO₃⁻-N in the reaction system showed a downward trend (Figure 6C), while the content of NH₄⁺ -N showed a trend of rapid increase and then gradual decrease (Figure 6D), and at the same time, the higher the value of DOC: NO₃⁻, the more dramatic the changes in the NO₃⁻-N and NH₄⁺-N content (Figures 6C,D). In the pre-culture period (first 10 d), the NO₃⁻-N content in the system decreased and the NH₄⁺-N content increased, which might be attributed to the promotion of DNRA reaction by high carbon-nitrogen ratios. Numerous reports have shown that DOC is the substrate for DNRA and denitrification processes, denitrification process is a competitive advantage when the carbon-nitrogen ratios in the soil is lower than 2.5, but DNRA reaction occurs preferentially when the value of carbon-nitrogen ratios is higher than 4, and the sufficient DOC is beneficial to promote the activity and growth of DNRA bacteria (Tiedje et al., 1983; Wang et al., 2020a; Chen et al., 2022). In the later stage of culture (after 10 d), the content of NO₃⁻-N and NH₄⁺-N in the system decreased gradually, which may be due to the anammox. At this stage, DOC and NO₃⁻-N in the system were consumed by denitrification and DNRA reaction, resulting in NO₃⁻-N and NH₄⁺-N, which provided sufficient substrate for anammox reaction and promoted the growth of anammox bacteria. Hence, in the riparian zone of Weinan section of Wei River, sufficient DOC increases the carbon-nitrogen ratios in the environment, which promotes the DNRA reaction and accelerates the attenuation of NO₃⁻-N.

As the riparian zone is an active biogeochemical region with naturally developed physicochemical and microbiological gradients that are key drivers of the nitrogen cycle (Hou et al., 2017). We observed that D_total was significantly positively correlated with A_total and negatively correlated with R_DNRA (Supplementary Figures S5, S6), which suggested that denitrification and anammox process in the riparian zone had a mutually reinforcing relationship. Denitrification could provide electron acceptors (NO₂⁻-N) for the anammox process, and the consumption of organic carbon by denitrifying bacteria could alleviate its inhibitory effect on anammox bacteria (Hou et al., 2015; Liu et al., 2019). On the other hand, denitrification and DNRA showed a competitive relationship, competing for the common electron donor DOC and electron acceptor NO₃⁻ in environmental media. Therefore, the synergistic and competitive relationship among denitrification, anammox and DNRA jointly promoted the nitrogen removal process in the riparian zone.

In the riparian zone of Weinan section of Wei River, DNRA and denitrification are the most active in the attenuation process of nitrate. For the denitrification process, TDS, EC, NH₄⁺ and NO₂⁻ were the main environmental factors affecting D_total, which were significantly positively correlated with D_total (p < 0.05, Figure 7A). This conclusion was consistent with the results of Magalhaes et al. (2005) in the intertidal sediments of Douro estuary in Portugal and Liu et al. (2018) in the sediments of Yangtze River lakes, these substances directly or indirectly provide electron acceptor/donor or energy for denitrifying bacteria to promote denitrification. A large number of studies have shown that nitrate is the reaction substrate of denitrification and is proportional to the denitrification rate (Mulder et al., 1995; Yao et al., 2018; Broman et al., 2021). Interestingly, the correlation between NO₃⁻ content and D_total in this study was not strong and negatively correlated (Figure 7A), which may be related to the NO₃⁻ content and the abundance of denitrifying bacteria in the riparian zone. The electron affinity of NO₃⁻-N and NO₂⁻-N was different, and there was a competitive electron relationship (Xing et al., 2016). In the W1, W3, and W5 layers containing sludge, the contents of DOC, NH₄⁺-N and NO₂⁻-N were significantly higher (Table 2 and Figure 2A). Therefore, the electrons provided by a large amount of DOC may preferentially flow to NO₂⁻-N for denitrification (nitrite reduction) or DNRA reaction. While in the W2 and W4 layers with higher NO₃⁻-N content, the DOC content and denitrification function (nirS, norB) microbial abundance were lower (Table 2 and Figures 4A,B), and the denitrification process might be inhibited. Accordingly, D_total in the riparian zone was significantly positively correlated with DOC and NO₂⁻ contents, but not with NO₃⁻ contents in sediments.

For the DNRA process, NO₃⁻, sediment lithology (Silt), DOC, Fe(II) and NO₂⁻ were the main environmental factors affecting R_DNRA (p < 0.05, Figure 7B). Among them, NO₃⁻ and Fe(II) were significantly positively correlated with R_DNRA, and the other environmental factors were significantly negatively correlated with R_DNRA. Fetzer and Li et al. found that the microbial community of riparian sediments is very sensitive to lithologic structure, coarse particles promote microbial activity and accelerate material transport, and fine particles will block exchange channels, reduce microbial activity and hinder solute migration and transformation (Fetzer et al., 2017; Wang et al., 2020b). The lithology of the sediments in the riparian zone of the Weinan section of the Wei River was dominated by silt, followed by fine sand, and these fine particles might inhibit the DNRA reaction. Microcosmic experiments showed that there was a coupled Fe-N cycle in the riparian zone, and Fe(II) would promote the attenuation of NO₃⁻-N through NDFO action, which reduced the reaction substrate of DNRA and inhibited the DNRA reaction. In contrast, NO₃⁻ and DOC provided electron acceptors and electron donors, respectively, for the DNRA reaction (Li Z et al., 2020; Chang et al., 2021), which facilitated the DNRA reaction.

In addition, D_total was significantly and positively correlated with the abundance of denitrification functional genes nirS and nroB, and R_DNRA was also significantly and positively correlated with the abundance of DNRA functional gene nrfA. This conclusion is consistent with the research results of China’s Yangtze River estuary (Zheng et al., 2021), Northeast China Sea (Chang et al., 2021), and high-altitude rivers (Zhang et al., 2020). This further proves that the nitrogen removal process of riparian sediments is mediated by the redox reaction of different microorganisms. In addition, in the study of soil nitrogen cycle, it was found that the microorganisms involved in denitrification and DNRA processes usually lived in the same environment, and the optimal pH range was relatively close, denitrification was 6–8, and DNRA was 5–9. High pH will promote nitrate reduction through DNRA reaction and weaken the intensity of denitrification (Stevens et al., 1998; Fernandes et al., 2012). Interestingly, in this study, the pH of the sediment environment was 7.77–8.02, and the correlation analysis showed that pH was negatively correlated with D_total and denitrifying bacteria abundance, and significantly positively correlated with R_DNRA and DNRA bacteria abundance (p < 0.05, Figure 7). It indicated that the pH of the sedimentary environment in the riparian zone was more suitable for the growth of DNRA bacteria.

In general, compared with the permanent submerged sediments, the riparian sediments in the Weinan section of the Wei River are in a dry-wet alternation state. The interaction between river water and groundwater causes the fluctuation of water level in the riparian zone, which changes the physical and chemical properties of the sediments, and then affects the microbial community structure and the distribution characteristics of inorganic nitrogen. The abundance of nitrogen-converting microorganisms, NO_X⁻, Fe(II) and DOC jointly drove the N-loss process in the riparian zone.