Analysis of nutrients removal performances
The purification efficiency of each CW in different seasons were summarized (Supplementary material, Fig.S1). The nutrients removal efficiency of CW-BK was significantly lower than experimental groups (ANOVA, P < 0.05), and CWs with different monoculture species were also performed significantly different in nutrients removal (P < 0.05). It was in accordance with previous publications (Lu et al., 2012; Vymazal, 2011), which compared the nutrients removal efficiency of CWs in different plant species, and the results showed that different plants differ in nutrients removal efficiencies. Further, CWs in mixed culture plants resulted in significantly increased nutrients removal efficiency compared monoculture plants (P < 0.05) (NH4+-N, NO3−-N, TP and COD increased by 4.6% − 28.7%, 1.2% − 20.3%, 2.5% − 6.4% and 5.9% − 12.7%, respectively). More importantly, CWs with different mixed plant species also showed significantly different nutrients removal efficiency (P < 0.05).
To better understanding the differences in purifying function between different mixed culture CWs and figure out the optimal plant configuration, annul average nutrients removal efficiency of each CW in mixed culture was analyzed (Fig. 2A, B and C). CW-G3 and CW-G2 achieved higher removal efficiency of TN (Fig. 2A), CW-G3 and CW-G1 achieved higher removal efficiency of TP (Fig. 2B), while CW-G3 and CW-G2 achieved higher removal efficiency of COD (Fig. 2C). Thus, CW-G3 performed well in all TN, TP and COD removal. Furthermore, the concentrations of N in different forms were detected after 4 days treatment (Fig. 2D, E and F). NH4+-N is firstly transformed into NO3−-N/NO2−-N by nitrifying bacteria with O2 and then converted into N2/N2O by denitrifying bacteria with the aid of electronic donors OC (Tan et al., 2020). NO3−-N concentrations of CW-G3 were significantly lower than CW-G1 and CW-G4 (Fig. 2D and E), while NO2−-N concentration of CW-G1 were significantly higher than other three CWs (Fig. 2F) (P < 0.05). These results indicated that CW-G3 were sufficient in O2 and OC to complete nitrifying-denitrifying process (Zhao et al., 2021).
Thereby, mixed culture of plants positively impacted purification efficiency of CWs, and different plant configurations differed in nutrients removal efficiency. Moreover, among the four CWs, CW-G3 showed better diversity nutrients removal efficiency (TN, TP and COD: 94.2%, 82.9% and 74.7%, respectively), which was the optimal plant configuration.
Analysis of plant activity
Plant roots could enhance nitrifying efficiency by releasing O2 and increase denitrifying efficiency by secreting OC to the rhizosphere. Thus, the healthy roots of plants are critical for N removal in CWs (Yang et al., 2020). The aeration tissues of plant roots were observed through SEM (Supplementary material, Fig. S2). Four species plants in CW-G3 maintained the complete root structures with tightly arranged stomas, large openings, and well-developed aeration tissues (Fig. S2 C). But root structure of several plant species in other three CWs were rot without complete aeration structure (Fig. S2 A, B and D). Kludze et al., (2013) proved that the well-developed roots aeration tissues could release O2 to the rhizosphere and improve DO content around rhizosphere. Further, the DO content in rhizosphere effluent of each CW were detected, and CW-G3 was significantly higher than that in other CWs (0.92, 1.42, 1.96 and 1.22 mg/L in CW-G1, CW-G2, CW-G3 and CW-G4, respectively), which was consistent with SEM results.
We also assessed enzyme SOD, POD, CAT and MDA levels of plant leaves (Fig. 3). After 4 days treatment of wastewater, SOD, POD, CAT and MDA activities increased 1.1 ~ 1.8 folds, 1.3 ~ 3.3 folds, 1.2 ~ 20.1 folds and 1.1 ~ 5.8 folds, respectively. Moreover, the increase value of POD and CAT enzyme activities of plants in CW-G3 was significantly higher than that of plants in other CWs (Fig. 3B and C), while its MDA enzyme activity was significantly lower than that of plants in other three CWs (Fig. 3D) (P < 0.05). As antioxidant enzymes, SOD, POD and CAT play critical roles in the reactive oxygen species scavenging system (Baxter et al., 2014). Increased antioxidant enzyme activities indicated that wastewater triggered stress responses in plant leaves. MDA is a negative product of stress response under adverse conditions, which is also a hidden danger to photosynthetic ability of plant leaves (Yang et al., 2020). When photosynthetic ability was destroyed, O2 and OC would not be successfully produced by leaves and released to the rhizosphere, which led to reduced nitrification and denitrification efficiency in the rhizosphere, thus inhibiting N removal process (Racchetti et al., 2017; Salgado et al., 2018). It indicated that plants in CW-G3 could reduce oxidative stress by increasing POD and CAT enzyme activities, thereby reducing the accumulation of MDA to maintain a high photosynthetic capacity and product high level of O2 and OC. Yang et al., (2020) reported that exposure of nanoplastics in CWs increased MDA level of plant leaves and reduced root activities, which further inhibited N removal efficiency. This is also consisted with our research. Meanwhile, well-developed roots aeration tissues of CW-G3 ensured O2 and OC release to rhizosphere. Therefore, CW-G3 performed well in nutrients removal.
Analysis of microbial community in rhizosphere and non-rhizosphere
Microbial community was analyzed to demonstrate the variation of N concentrations. Biodiversity indexes Chao1 and Shannon of microbial community in rhizosphere and non-rhizosphere were provided (Supplementary material, Table S2), which reflect the richness and diversity of microbial community, respectively (Fu et al., 2019). Shannon index of the rhizosphere of four experiment groups was significantly higher than of the non-rhizosphere, while no significantly difference was found in the blank group, which may indicate that root exudates benefit the microbial diversity. Both Chao1 and Shannon of CW-G3 were significantly higher than that of other three CWs, indicating that plant configuration in CW-G3 provided a suitable growth environment for microorganisms and positively affected the richness and diversity of the microbial community.
Furthermore, the distribution of microbial communities at phylum level was analyzed (Supplementary material, Fig. S3). Phylum Proteobacteria (26.47% − 57.4%), followed by Actinobacteria (1.2% − 40.4%), Bacteroidetes (2.9% − 22.9%), and Patescibacteria (1.8% − 19.2%) were dominant in four CWs, which have been reported to be widespread in aquatic environments (Pang et al., 2016; Sun et al., 2019b). Proteobacteria (Cheng et al., 2016), Actinobacteria (Lv et al., 2021) and Bacteroidetes (Jia et al., 2021) could significantly reduce N, thus, CWs performed well in N removal (higher than 89.1%) (Fig. 2). The relative abundance of Patescibacteria in CW-G3 (rhizosphere: 9.0%; non-rhizosphere: 19.2%) was significantly higher than that in other CWs (P < 0.05). Zhang et al., (2021c) reported that adding OC in CWs significantly increased the abundance of Patescibacteria, which may suggest that more OC can be utilized by Patescibacteria in CW-G3. At genus level (Fig. 4A and B), there were obvious differences in the microbial communities of each experimental group, which was resulted by different plant configurations. Moreover, the relative abundances of Saccharimonadales and Flavobacterium in the rhizosphere (8.5% and 17.7%, respectively) and non-rhizosphere (8.7% and 10.0%, respectively) of CW-G3 were significantly higher than those in other CWs. Saccharimonadales affiliates to Patescibacteria, which is beneficial to CWs denitrification (Chen et al., 2018; Zhang et al., 2021b). Chen et al., (2022) proved that extra OC added in CWs increased relative abundance of Saccharimonadales and improved denitrifying process. This result indicated that more OC was utilized by bacteria in CW-G3, which was consistent with phylum results. Flavobacterium was also widely reported to have the nitrate-reducing capability (Pishgar et al., 2019). Therefore, CW-G3 showed the highest N removal efficiency.
Moreover, principal co-ordinates analysis (PCoA) was employed to investigate the differences in microbial composition between two samples (Fig. 4C). The samples were clustered various in the plot since CW-G1 was placed top-right, CW-G2 was placed top-left, CW-G3 and CW-G4 were placed bottom-middle in the figure. The rhizosphere samples and non-rhizosphere samples in each CW were also clustered on two sides of the oval. These results indicated diversity of microbial communities because of root exudated by different plant configurations. Further, degradation/utilization/assimilation pathways of microbial community were analyzed by PICRUSt2 (Fig. 4D), which showed that carbohydrate degradation (6030.59 PWY/Million) was predominant, followed by nucleoside and nucleotide (5090.38 PWY/Million), and amino acid degradation also took a great portion. Carbohydrates and amino acids are low-molecular-weight compounds, which can be easily up-taken and metabolized by microorganisms (Koner et al., 2021). Nucleoside and nucleotide can be utilized by bacteria to favor their proliferation (Bugenyi et al., 2020). The results indicated that plants in CWs provided extra OC (sole carbon glucose was added into CWs) for bacterial function and proliferation.
Analysis of root exudates
Non-targeted metabolic profiling was conducted to analyze the differences in metabolic profiles among different CWs. In total, 1250 and 634 metabolites were detected in positive ion mode (POS) and negative ion mode (NEG), respectively. The metabolites were divided into seven major groups, including organic acids (e.g., salicylic acid, succinic acid), amino acids (e.g., alanine, glycine), carbohydrate (e.g., maltose, sucrose), fatty acids (e.g., linoleic acid, oleic acid), amines (e.g., triethylamine, tributylamine) alkaloids (e.g., xanthine, guanine), lipids (e.g., methyl palmitate), and others.
Moreover, metabolites were mapped to the reference annotation pathway in Kyoto Encyclopedia of Genes and Genomes (KEGG), and metabolic pathways between CW-G3 and other mixed culture CWs were compared, respectively (Fig. 5). Among these pathways, biosynthesis of unsaturated fatty acids (ath01040) and linoleic acid metabolism (ath00591) were significantly different and present in three compared groups (Fig. 5A, B and C) (P < 0.05). Their metabolic pathway was depicted further (Fig. 5D and E), which showed that levels of metabolites linoleic acid, oleic acid and arachidonic acid (Fig. 5D), as well as their fatty acid easter linoleate and arachidonate (Fig. 5E) were significantly different between CW-G3 and other CW. Relative contents of these different regulated metabolites were analyzed (Table 1), all linoleic acid, oleic acid and arachidonic acid in CW-G3 were significantly lower than that in other three CWs (P < 0.05). Subbarao et al., (2008) demonstrated that unsaturated fatty acids linoleic acid released by plant roots inhibited microbial nitrifying, so that NH4+-N is hard to convert into NO3--N in soil, while arachidonic acid showed no significantly inhibitory effect on nitrification. Similarly, Souri, (2017) also reported that linoleic acid, oleic acid significantly inhibited nitrification process, in which higher concentrations have led to less nitrite production. Thereby, lower N removal efficiency of other three CWs (Fig. 2A) indicated that higher concentrations of linoleic acid, oleic acid inhibited microbial nitrifying, rather than utilized by microbes as OC.
The other significantly different metabolic pathways included glycerophospholipid metabolism (ath00564), purine metabolism (ath00230), histidine metabolism (ath00340), tropane, piperidine and pyridine alkaloid biosynthesis (ath00960) (KEGG pathways were shown in Supplementary material, Table S3). Relative contents of significantly different metabolites were shown in Table 1. Contents of maltose in CW-G3 were 6.6 ~ 11.1 folds of that in other three CWs, which have been widely reported to be used as OC in wastewater purification system to improve N removal efficiency (Wang et al., 2016). Contents of lauramide, choline, triethylamine and urocanic acid in CW-G3 were also significantly higher than compared CW (P < 0.05). All these metabolites have been proved to be utilized by microorganisms as OC (Merkova et al., 2018; Gao et al., 2018; Fujii et al., 2014; Wargo, 2013), which indicated that plants in CW-G3 released higher contents of OC than in other CWs. Thereby, with lower nitrifying inhibitors and higher organic C, plant configuration in CW-G3 showed better N removal efficiency (Fig. 2A).
Table 1
Ratios of significantly different metabolites between compared CW groups.
Name | Ratio (RX-G3 / RX-G1) | Ratio (RX-G3 / RX-G2) | Ratio (RX-G3 / RX-G4) |
Arachidonic acid | 0.466 | 0.838 | 0.369 |
Oleic acid | 0.440 | 0.471 | 0.220 |
Linoleic acid | 0.267 | 0.681 | 0.182 |
Maltose | 7.216 | 11.100 | 6.645 |
Lauramide | 1.329 | 2.520 | - |
Choline | 2.764 | - | - |
Triethylamine | 3.063 | - | - |
Urocanic acid | - | 3.416 | - |
Notes: “-” represent the metabolites did not detected in CW. Ratio represent relative concentration, ratio > 1: metabolite contents of CW-G3 higher than of compared CW; ratio < 1: metabolite contents of CW-G3 lower than of compared CW. |
Co-occurrence network between bacteria and root exudates
The co-occurrence network between bacteria and eight differently regulated root exudates was analyzed (Fig. 6), which was composed of 31 nodes and 98 edges (linkages). Network diameter, average path length and modularity index was 4, 2.077 and 0.467 (> 0.4), respectively, indicated that the real-world network has a module structure (Ren et al., 2020). Unsaturated fatty acids arachidonic acid, oleic acid and linoleic acid were negatively correlated with bacteria (e.g., Pseudorhodoplanes, Geobacter, Nitrospira, Areomonas, Amaricoccus, Saccharimonadales and Flavobacterium), which means that the higher contents of these unsaturated fatty acids inhibited bacterial functional performance. Pseudorhodoplanes has been proven to be able to fix N and release plant growth promoting regulators (Şeker et al., 2017). Nitrogenase-encoding genes have been shown in nearly all Geobacter species (Jing et al., 2022). Nitrospira is a widely studied nitrifying bacteria, Areomonas and Amaricoccus are also take part in nitrification and help for N removal (Daims et al., 2015; Koch et al., 2019). Therefore, the N removal efficiency of other three CWs was lower than that of CW-G3 due to its high contents of N removal inhibitors (acids arachidonic acid, oleic acid and linoleic acid), which inhibited bacterial function.
Nevertheless, maltose, urocanic acid, lauramide, choline and triethylamine were positively correlated with bacteria (e.g., Thauera, Simplicispira, Pseudorhodoplanes, Thiothrix, Chloroflexus, Saccharimonadales and Flavobacterium), which means that the higher contents of these OC benefited bacterial functional performance. Thauera (Luo et al., 2020), Simplicispira (Siddiqi et al., 2020) and Thiothrix (Xia et al., 2019) were widely found in wastewater treatment systems as denitrifying bacteria that facilitated N removal. Chloroflexus (Sun et al., 2019a) degraded histidine coupled to the reduction of NO3−-N enhanced N removal in wastewater treatment systems. Saccharimonadales (Zhang et al., 2021b) and Flavobacterium (Pishgar et al., 2019) are denitrifying bacteria and their relative abundances were significantly higher in CW-G3 than that in other three CWs. These results indicated that higher contents of maltose, urocanic acid, lauramide, choline and triethylamine in CW-G3 could be utilized as OC by microbes, which was beneficial for N removal. Secretion of unsaturated fatty acids oleic acid, linoleic acid and arachidonic acid inhibited function of nitrifying bacteria, while maltose, urocanic acid, lauramide, choline and triethylamine enhanced function of denitrifying bacteria. Thereby, CW-G3 performed well in N removal than other three CWs.