Winter N removal of microcosms under different substrate addition
The inlet concentrations of TN, NH4+–N and NO3––N during the experiment period were 5.8–20.7, 0.3–14.4 and 0.1–1.2 mg·L–1, respectively (Fig. 1A). The outlet concentrations of each treatment were decreased significantly in all treatment groups and remained relatively stable at a low level compared to the fluctuating inlet concentrations, indicating that all soil treatments resulted in the effective sequestration of the N pollutants within the microcosms. Regarding their relative effectiveness against the different nitrogenous compounds, significant differences were observed in TN and NH4+–N removal rates among the groups (P < 0.05), while no significant differences among the different treatments were observed in the NO3––N removal (P > 0.05, Fig. 1B). The natural riparian soil itself (DS1) displayed a good capacity for the removal of N pollutants, with average removal rates for TN, NH4+–N and NO3––N of 88.3%, 89.0% and 71.0%, respectively. However, the alterations to the initial soil by the addition of gravel (DS2), gravel + biochar (DS3), ceramsite + biochar (DS4) and modified ceramsite + biochar (DS5) showed further improved capacities for the removal of N pollutants at winter wetland levels, while, the relative ineffectiveness of DS5 (and to a lesser extent, DS4) was also observed compared to DS2 and DS3.
The highest TN removal rate was observed in DS3, which was relatively stable at 97.2%, an increase of 8.9% compared to DS1 (P < 0.05). DS2 and DS4 treatments also promoted a higher rate of TN removal (5.5% and 5.9%, respectively), but these were not significant relative to DS1 (P > 0.05). Relatively higher and stable NH4+–N removal rates was observed in DS2 (97.2%) and DS3 (97.9%), which relative to DS1, correspond to increases of 8.3% and 8.9%, respectively (P < 0.05). Increases were also seen in DS4 and DS5, but these were not significant relative to DS1. Taken together, the results demonstrated that the addition of gravel (DS2) and gravel plus biochar (DS3) produced the largest and most significant improvements in N removal.
Compared to biochar, gravel has a lower capability for N adsorption and microorganism attachment (Yang et al., 2018), and its adsorption capacities can decrease over time (Kizito et al., 2017). This suggests that the physical adsorption of substrates by gravel was not the dominant mechanism behind the improved N removal observed here. The addition of gravel can enhance the aeration of soil substrates, with accelerated air and water transfer (Mukherjee et al., 2014), thus providing an altered micro-environment for the recruitment of root-microbes with potential affects on N assimilation (Zhou et al., 2018). Similarly, compared to ceramsite, the irregular geometry of gravel might increase the complex structure of the rhizospheric environment with potential affects on N assimilation by the microbiome. The addition of biochar plus gravel to riparian soil produced a higher NH4+–N removal rate. The high capacity for NH4+–N adsorption of biochar (Hou et al., 2016) might have contributed to this. However, biochar has surface fissures and holes, and this porous environment may facilitate to abundant microbial attachments (Deng et al., 2019). Furthermore, dissolved organic matter released from biochar may supply carbon sources to enhance microbial denitrification (Li et al., 2018), thereby promoting increased wetland NH4+-N removal rates.
Over the entire experiment, no significant differences (P > 0.05) were observed in in-situ water T and DO among the soil treatments, while significant differences were observed in pH, ORP, EC and TDS (P < 0.05, Table S1). DS2, DS3, DS4 and DS5 had significantly higher pH, EC and TDS (P < 0.05), with a significantly lower ORP (P < 0.05), compared to DS1. However, RDA analysis showed that in-situ water parameters only explain 20.9% of the total variation of wetland N removal rate (Fig. S3), of which the impacts of T (negative) and pH (positive) were significant (P < 0.05). It has been reported that higher temperatures promote better wetland N removal rates (Wu et al., 2017; Wang et al., 2021). However, our data indicates that higher water temperatures cannot be achieved by the any of the alterations to riparian soil substrate tested here. pH is a key factor influencing wetland N removal, the same as the previous report (Shamim and Joann, 2016). It has been proved that higher pH can promote the rapid conversion of NH4+–N to NO3––N through nitrification (Shamim and Joann, 2016), indicating that the increase of pH might be an important factor in the choice of additives for wetland substrate improvement.
Root Metabolite Characteristics Under Different Substrate Addition
We determined root exudate metabolites in rhizosphere soil using untargeted metabolomics. A total of 584 metabolite compounds were detected in the soil treatments in both the positive (289) and negative (295) polarity mode. PCA showed that the metabolite profiles of DS1 and DS2 were close and clearly separated from the other treatments (Fig. S5). In addition, the metabolite profiles between DS3 and DS4 were closer than with DS5, which was the most distinctive profile. Hierarchical clustering also showed the same treatment clustering as PCA (Fig. S5), and seemed that the broad reprogramming might not be limited to the metabolites associated with root exudates, which might also contributed from the microbiome. The data suggests that the impacts of the three treatments involving biochar addition on plant metabolite profiles were more pronounced than that from the addition of gravel alone.
The association of the different quantified metabolites with known metabolic pathways in KEGG allows the differential deployment of such pathways to be examined for significance (Cesco et al., 2021). Among the detected metabolites, 76 (36 and 40 compounds in positive and negative polarity mode, respectively) were associated with 11 metabolic pathways (Table S2; Figs. S6, 2A). There were no significant differences on the abundances of metabolites in xenobiotics biodegradation and metabolism or nucleotide metabolism between the soil treatments (P > 0.05). The abundances of metabolites in carbohydrate metabolism were lower in DS2 and DS5 while higher in DS4 as compared to DS1 (P < 0.05). Carbohydrates produced by photosynthesis (Blackstock, 1989), whose decrease might be caused in microcosms containing gravel or modified ceramsite + biochar. The increase of carbohydrate metabolism-related metabolites observed after ceramsite + biochar addition (DS4) could directly influence the production of various compounds (Toledo et al., 2017). Relative to DS1, the abundances of metabolites in lipid metabolism were lower while in metabolism of cofactors and vitamins, amino acid metabolism, biosynthesis of other secondary metabolites, metabolism of other amino acids, and metabolism of terpenoids and polyketides were higher in the three treatments involving biochar addition (DS3, DS4 and DS5) (P < 0.05). Plant lipid metabolism is active in the production of root exudates, leading to a significant content of fatty acids and lipids, thereby contributing to exudate hydrophobicity (Döll et al., 2021). The decrease metabolites related with lipid metabolism in the rhizosphere observed after biochar addition, might be conducive to a better water and nutrient uptake by the roots, thereby promoting the growth and development of plants (Table S3). The categories of ‘metabolism of cofactors and vitamins’, ‘amino acid metabolism’, ‘biosynthesis of other secondary metabolites’, ‘metabolism of other amino acids’, and ‘metabolism of terpenoids and polyketides’ were all increased, indicating that the root release of a variety of carbon sources is increased under biochar involved additions.
Metabolites which showed differential abundances in response to the different soil treatments. A heatmap displays the 30 of the 76 metabolites which are both most abundance and show significant differences in response to the different soil treatments (Fig. 2B). The associated dendrogram reflects the same clustering of treatments cluster as that observed for all metabolites (Fig. S5), i.e. between DS1 with DS2 and DS3 with DS4. Interestingly, DS5 behaved more as an outlier, but was more associated with DS3 and DS4 than DS1 and DS2, indicating that biochar addition had a great effect on metabolites, but compared with ordinary gravel and ceramsite addition, this effect was weakened by the addition of modified ceramsite. For more detailed information of metabolite compounds, melibiose and sucrose were the main metabolites within carbohydrate metabolism. L-isoleucine, involved in four metabolic pathways (chemical structure transformation maps, biosynthesis of other secondary metabolites, amino acid metabolism and metabolism of other amino acids), and phenylglyoxylic acid, involved in two pathways (amino acid metabolism and xenobiotics biodegradation and metabolism), were significantly increased under gravel + biochar addition (P < 0.05). KAPA, involved in the metabolism of cofactors and vitamins, significantly increased under modified ceramsite + biochar addition (P < 0.05). The number of compounds within lipid metabolism constituted the largest group (17 out of 30 metabolites), and the abundances of minority [testosterone glucuronide, methyl jasmonate, estrone 3-sulfate and lysoPC(18:3(6Z,9Z,12Z))] were higher, while majority [cortisone acetate, (S)-10,16-Dihydroxyhexadecanoic acid, hexadecanedioic acid, lysoPC(18:0), 9-OxoODE, estrone glucuronide, 9(S)-HpOTrE, lysoPC(16:1(9Z)/0:0), lysoPC(18:1(9Z)), lysoPC(16:0), leukotriene E4 and 9(S)-HpODE] were lower (P < 0.05) in the biochar addition treatments except where modified ceramsite was added (DS5).
Regardless of the specific processes affected by the different soil treatments in the present study, the effects induced by the treatments involved critical aspects of root metabolism, with potential consequences to N removal in wetlands. RDA analysis showed that the different abundances of metabolites in different metabolic pathways explains 69.0% of the total variation of wetland mean N removal rate (Fig. S6), of which terpenoids and polyketides, other secondary metabolites, amino acids and cofactors and vitamins had significant positive impacts (P < 0.05), while metabolites of lipid metabolism had a significant negative impact. Combined with the above results, the positively affected metabolites all increased in soil treatments with added biochar, while the negatively affected fatty acid derived and lipidic metabolites decreased, suggesting that the biochar treatment of wetlands might provide for a more abundant and varied carbon source in the rhizosphere for microbial consumption (Chen et al., 2016; Wu et al., 2017) and simultaneously, a lower hydrophobic barrier to root uptake of water and nutrients, leading to a higher rate of N removal. Furthermore, there is evidence that the lipidic content of root exudates can influence the growth of certain microorganisms in the rhizosphere (Fries et al., 1985). It could therefore the alterations to the exudate-derived lipidic contents of the rhizosphere among the treatments could reflect either a) differences in the consumption lipids in the rhizosphere by recruiting different microorganisms to the microbiome or b), the recruitment of different microorganisms by the different treatments could alter the signals from the rhizosphere to the plant, with consequences to the production of lipids in the root exudates and c), a mixture of a) and b). In either case, the differences in the microbiome’s constitution could clearly affect its contribution to N removal.
Rhizosphere Soil Microbial Community Compositions Under Different Substrate Addition
After being resampled, 9,505 OTUs were obtained from effective sequences and were assigned to 52 phyla, 134 classes, 367 orders, 643 families and 1341 genera. From the Shannon and Simpson indices, no significant differences were observed among the treatments (P > 0.05; Table S4), indicating that the diversity of the microbial community was not affected by the different additives. However, relative to DS1, a higher ACE index was observed in DS2, while a lower Chao index was observed in DS5 (P < 0.05), suggesting that gravel addition promoted a higher richness of microbial community, while modified ceramsite + biochar addition suppressed microbial community richness compared with the natural soil.
Based on relative abundances, the top five phyla across all treatments were Proteobacteria, Chloroflexi, Actinobacteria, Acidobacteria and Firmicutes, accounting for more than 80% (81.9–83.7%) of all microbes detected (Fig. 3A), which indicated that these phyla were the dominant players in the winter N removal of riparian reed wetlands. The top three most abundant phyla in all soil treatments were Proteobacteria, Chloroflexi and Actinobacteria. Proteobacteria is a ubiquitous and abundant phylum that plays important roles in the nutrient cycling (Sun et al., 2019), and is the most generally contributes to denitrification under different water environments (Miao et al., 2015). Chlorofloxi are common in anaerobic ammonia oxidation systems, some studies have found that Chloroflexi significantly enriched with biochar addition (Deng et al., 2019), but this phenomenon was not observed in this study. Actinobacteria are generally widely distributed in various ecological environments, which have the ability to form “mycelia” to reach water and nutrients, and play an essential role in the decomposition of soil recalcitrant matter (Rehákov et al., 2015). Actinobacteria were also found to be dominant in the rhizophere soil of many plants (Praeg et al., 2019; Liu et al., 2020). However, there were no significant differences observed between treatments on the abundances of these top three phyla (P > 0.05). Clear and significant differences were seen in the negative and positive affects of DS5 on the relative abundances of Acidobacteria and Firmicutes, compared to DS1 (P < 0.05). Acidobacteria are typically classified as oligotrophic microbes and are associated with habitats with more recalcitrant carbon (Praeg et al., 2019). Firmicutes are also active members in carbon and nitrogen cycles (Wang et al., 2022). Furthermore, DS5 resulted in a declined relative abundance of Gemmatimonadetes, Rokubacteria and Latescibacteria, and an enrichment of Bacteroidetes, indicating that the microbial community structure had been greatly adjusted under modified ceramsite + biochar addition.
There was a total of 30 dominant genera with relative abundance > 1% across all treatments (Fig. S7), which accounted for 47.1% (the highest) of the total sequences in DS1, while 35.9% (the lowest) in DS5 (P < 0.05), indicating that the microbial communities in the DS5 rhizosphere soil changed significantly compared to those in the initial soil. A heatmap showed that DS1 and DS4 were clustered together, indicating these treatments resulted in the most similar microbial communities, but with subtle variations in their contents of genera (Fig. 3B). This similarity was decreased progressively in DS3 then DS2. This confirmed that modified ceramsite + biochar addition exerted a stronger influence on the microbial community structure of rhizosphere soil than those of other soil treatments. A closer inspection of the effects of different soil treatments on dominant revealed that norank_o_SJA-15 and norank_o_SBR1031 within Chloroflexi, and norank_c_Thermodesulfovibrionia within Nitrospirae were significantly higher in DS2 than DS1, and that Ralstonia within Proteobacteria was significantly higher in DS3 than DS1. These genera all play important roles in nutrient (especially N) turnover (Wang et al., 2018; Yan et al., 2020).
Rhizophere microbes are the main force involved in wetland N removal and play an important role in maintaining the balance of the ecosystem (Feng et al., 2012). To examine which of the microbial phyla from all treatments were most likely associated with the observed N removal rates in microcosms, the study explored their relationship in RDA analysis. The results showed that microbial communities explain 63.1% of the total variation in the wetland N removal rate (Fig. S8). It appears that Rokubacteria had the most positive affect on wetland TN removal rates, whose relative abundance increased after the addition of different substrates except DS5. Bacteroidetes had the most negative affect on wetland TN removal rates, whose relative abundance decreased in the different substrate addition treatments except for DS5. Bacteroidetes are well known degraders of polymeric organic matter (Wang et al., 2009), which might have little impact on the role of wetland N removal.
Examining the relationship of root metabolite profiles with microbial community compositions under different substrate treatments
Plants release a blend of metabolites from the roots, which fuels the substrate-driven assembly process of the plant-specific rhizosphere microbes (Bulgarelli et al., 2013). The colonization of microbes in the rhizosphere has been described to be linked to root exudates through specific host-microbe interactions and recognition processes (Compant et al., 2010). Conversely, rhizosphere microbes can adjust the metabolic processes of plants to adapt to the soil environment (Kosova et al., 2018) and affect the metabolite profile of root exudates (Zhao et al., 2021). The metabolic reprogramming of plants induced by the addition of different substrates to the soil might affect the metabolite profile of root exudates, by their influence on the constitution of the rhizosphere microbial community. The alterations in the root exudate can subsequently lead to further alterations in the microbial community until an equilibrium is reached. While the existence of crosstalk between plants and microbes in the rhizosphere is undisputed, this process is not well understood.
The results of this study have shown that soil treatments affect the composition of both root metabolites and the rhizosphere microbial community (Figs. 2A, 3A). However, cluster analysis of these showed differing patterns of clustering to soil treatments for both (Figs. 2B, 3B), suggesting the interactions of soil treatments with microbial community recruitment and microbial effects on exudate metabolite production are complex. Correlation analysis showed that there were complex correlations within plant metabolism and microbiota themselves (Fig. 4), indicating that the interrelations among plant metabolic pathways existed and this occurred with strong inter-phyla cooperation in the microbiome. Considering together, there were no significant correlations observed between metabolites in lipid metabolism with the relative abundance of any microbe phyla in the both initial soil and substrate addition treatments, indicating that these metabolites had little effect on the microbial community and vice verse. Lipid metabolites in the rhizosphere might be deployed in the exudate to establish hydrophobic barriers rather than as an energy source to recruit members of the microbial community, so that these metabolites would remain independent of the community composition. However, all phyla could be related to some metabolic pathways of exudates in the different treatments (Fig. 4). There was a significant and positive correlation between the abundance of Proteobacteria and metabolites in nucleotide metabolism, and Actinobacteria with metabolites in xenobiotics biodegradation and metabolism, while a significant negative correlation of Nitrospirae with metabolites in metabolism of cofactors and vitamins. Chloroflexi, Acidobacteria, Rokubateria and Latescibacteria were all positively correlated to metabolites in carbohydrate metabolism and energy metabolism. Of these, Acidobacteria and Latescibacteria were also negatively correlated to metabolites in nucleotide metabolism. Firmicutes and Bacteroidetes were both negatively correlated to metabolites in carbohydrate metabolism and energy metabolism. Gemmatimonadetes showed a significant negative correlation with metabolites in metabolism of cofactors and vitamins, but a significant and positive correlation with metabolites in energy metabolism. Cyanobacteria was positively correlated to metabolites in both biosynthesis of other secondary metabolites and metabolism of other amino acids.
The chemical nature of root exudates can shape the microbial community by secreting specific root metabolite compounds that select for different species (Huang et al., 2014). In order to further explore the more detailed and internal relationships, we made a correlation analysis between the top metabolites in the correlated metabolic pathways (Fig. 2B) and the top genera of correlated phyla (Fig. 3B) and removed all the irrelevant, non-correlated members of both (Fig. 5A). We identified six major metabolites that were correlated with microbial genera. There were significant negative correlations between sucrose content with norank_o_SJA-15 and norank_c_Thermodesulfovibrionia, and sucrose was significantly lower in DS2 than DS1, but in equal or greater quantities in other treatments (Fig. 5B), while norank_o_SJA-15 and norank_c_Thermodesulfovibrionia were significantly higher in DS2 than DS1 or other treatments (Fig. 3B). Considering wetland N removal was increased in DS2 compared to DS1, we suggest that norank_o_SJA-15 and norank_c_Thermodesulfovibrionia are strong candidates for the microbial genera responsible for N removal process after gravel addition alone to soils (DS2). The majority of the microbial genera were positively intercorrelated with melibiose except for Bacillus, Geobacter and Erysipelothrix, indicating that melibiose was an important exudate carbon source for most of the microbes in the microcosms. Although 4-Nitrophenol is a metabolite in the xenobiotics biodegradation and metabolism pathway, it is also environmentally toxic (Tang and Zhang, 2021), and significant decreased in response to all substrate additives compared to the control. These decrease occurred with a higher N removal rate and was correlated with the abundance of several microbial genera, which might contribute to wetland N removal. Both L-Urobilin and uridine significantly increased in the three biochar addition microcosms compared to the natural riparian soil. L-Urobilin was only significantly negatively correlated with norank_c_Thermodesulfovibrionia, whereas uridine was only significantly and positively correlated with Bacillus. These alterations in metabolite profiles and rhizosphere bacterial may represent specific mechanisms operating after the addition of biochar involved to riparian soils. Although there was no significant difference in the abundance of KAPA in response to the different treatments, it was positively or negatively correlated with some microbial genera. We expect that the correlations observed between root metabolites and microbial taxa above could be further investigated about regarding their potentials as bioindicators of optimal conditions for wetland N removal.
The SEM was used to examine the relationship between rhizospheric microbiota and metabolites, and wetland N removal (Figs. 6). The interactive model was well-fitted, and the evaluation parameters were χ2/df = 2.924, RMSEA = 0.258, P = 0.000. The SEM results showed that the bacterial community was negatively correlated with metabolites (path coefficient = -0.596), while metabolites were positively correlated with the bacterial community (path coefficient = 0.681). Both bacterial community and metabolites positively affected wetland N removal, and compared with bacterial community, metabolites had a stronger effect on wetland N removal (relatively larger path coefficient). The results revealed that the bacterial communities, metabolites and their crosstalk in the rhizosphere were correlated with wetland N removal, which may be an important way for promoting winter N removal in riparian wetland by substrate improvement. In this study, root exudates were extracted from rhizosphere soil, so it is possible that soil microbial metabolites might have made a major contribution to the variation in metabolite abundances (Zhou et al., 2019). Further research is required to determine the contributions of root-related and microbe-related contribution to metabolite in the rhizosphere (Wang et al., 2021). In addition to the winter performance of microcosm systems, more work is needed to investigate the long-term performance of practical applications under the optimized condition.