Response of Sediment Microbial Communities to the Rural Wastewater in the Pond-Ditch Circulation System

Microorganisms played important roles in nutrient removal in Pond-ditch circulation system (PDCS). However, dynamics of microbial community in the PDCS, and responses of rhizosphere and non-rhizosphere microbial community to rural wastewater remains unclear. In this paper, average operational taxonomic units numbers of sediment microbial varied from 10,254 to 17,112, and values in rhizosphere were higher than those of the non-rhizosphere (p < 0.05). Bacillus, Clostridium sensu stricto 1, and Geobacter were the predominant genera in PDCS sediment with relative abundances of 0.52–17.61%, 0.26–8.08%, and 0.20–4.58%, respectively. However, Bacillus, Clostridium sensu stricto 1, and Geobacter genera in rhizosphere were more abundant than those in non-rhizosphere at day 30. Chao 1 index ranged from 10,225 to 17,033 and showed significant positive correlations with all sediment properties (p < 0.05). Chao 1 and Shannon indices in rhizosphere were significant positively related to tartaric acid and total organic carbon, respectively; while significant correlation between Shannon and Simpson indices in non-rhizosphere and oxidation-reduction potential were detected (p < 0.05). Redundancy analysis suggested that lactic acids, proteins, and amino acids had strong positive effects on Geobacter and Clostridiu sensu stricto 12 in the rhizosphere; while Bacillus and Clostridium in non-rhizosphere were significantly affected by sediment ammonia nitrogen and nitrate nitrogen. Environmental variables accounted for 66.9 and 60.3% of the total variation for the microbial community of non-rhizosphere and rhizosphere sediments, respectively. Our results highlight that root exudates and sediment available N alter predominant genera in the rhizosphere and non-rhizosphere, respectively, which is benefit for optimizing removal efficiency of PDCSs in large-scale applications.


INTRODUCTION
With the rapid development of economy and industrialization, the problem of water pollution in rural areas of China is getting worse. Large amounts of untreated domestic wastewater were discharged into the ponds, ditches, lakes, and rivers, which resulted in the deterioration of water environment and threatened the health of people (Bowes et al., 2015). In order to solve this problem, many types of wastewater treatment systems have been developed and widely used, such as constructed wetland system (Childers, 2020;Torrens et al., 2021), high-rate algal ponds (Evans et al., 2005), mineralbased small-scale active filter system (Gustafsson et al., 2008), sequencing batch reactor-biofilm system (Yin et al., 2015). Recently, our group have developed a new rural wastewater treatment system, pond-ditch circulation system (PDCS), which exhibits the advantages of high efficiency, cheap, and simple management and is an appropriate alternative for rural wastewater remediation (Ma et al., 2015a;Ma et al., 2015b).
In the PDCS, microorganisms played important roles in nitrogen (N) and phosphorus (P) removal. For example, approximately 79.5% of the total nitrogen removal was attributed to microbial process, especially nitrification and denitrification, which were mediated by a range of microbes, such as amoA, arch-amoA, nirS, and nirK genes (Ma et al., 2016). Meanwhile, anaerobic anammox bacteria have been widely used in nitrogen removal reactors due to their ability of converting ammonium into nitrogen gas under the anoxic conditions (Hu et al., 2010;Liu et al., 2020). The anaerobic anammox processes also performed a crucial role in nitrogen removal of the PDCS due to 41.3-50.3% contribution to nitrogen gas production (Ma et al., 2019). Similarly, some bacteria such as denitrifying phosphateaccumulating organisms can utilize nitrate (NO 3 − -N) or nitrite (NO 2 − -N) as the terminal electron receptor under the anoxic condition for the simultaneous removal of N and P Chen et al., 2021). Many studies reported that structures and diversities of microorganisms were affected by a number of abiotic and biotic factors, such as pH, temperature, oxygen concentration, carbon availability, hydraulic retention time, and plants (Khammar et al., 2005;Tomaszewski et al., 2017;Li et al., 2018;Liu et al., 2020).
Our previous studies have demonstrated that in the PDCS, aquatic plants participated in nutrient removal, were responsible for 10.1% N and 50% P reduction, respectively (Ma et al., 2016;Ma et al., 2019). Besides direct absorption, plants can also regulate the nutrient removal via root exudates such as sugars, polysaccharide, amino acids, and organic acids. These exudates can be used as the carbon and nitrogen sources, or stimulating signals by the rhizosphere microorganisms (Bais et al., 2006). This will in turn lead to the gathering of more microorganisms around the plant rhizosphere and result in "rhizosphere effect" (Egamberdieva et al., 2008). This "rhizosphere effect" is beneficial for the removal of nutrient in the rhizosphere microenvironment (Nie et al., 2015;Chen et al., 2016). and can exert significant impacts on the rhizosphere microorganisms . Therefore, the compositions, diversities of both rhizosphere and nonrhizosphere sediment bacteria vary greatly. Although several studies have correlated environmental factors to the microbial communities in decentralized treatment approaches, how rhizosphere and non-rhizosphere sediment bacteria in the PDCS responds to abiotic environmental factors still remain unclear. Based on our previous studies, we assumed that biotic factors such as plants and microbial communities rather than abiotic environmental factors were the dominant factors for nutrient removal in the PDCSs.
In this study, we have built two small-scale PDCS systems to treat domestic wastewater in rural areas of Southern China. These systems continuously operated for 2 months. This study aimed to elucidate dynamic changes of microbial community structures and diversities between rhizosphere and non-rhizosphere sediments and their responses to the environmental factors. The main contents of this study are as follows: (1) the total efficiency of N and P removal in the PDCS; (2) bacterial clustering information in rhizosphere and non-rhizosphere microenvironments; (3) relative abundances and diversities of bacteria in rhizosphere and non-rhizosphere environments at phylum and genus levels; (4) spearman's correlation coefficient analysis between rhizosphere and non-rhizosphere microbial diversities, water quality, sediment properties, and root exudates; (5) redundancy analysis (RDA) of dominant bacterial species and environmental factors in the rhizosphere and non-rhizosphere of the PDCSs. Based on this research, we have elucidated the dynamic changes of microbial communities in the PDCSs and the relations between environmental variables and microbial communities. These findings will be helpful in the design and the optimization of nutrient removal in PDCSs and other nutrient removal systems in large-scale applications.

Experimental Water and Sediments
The experimental sediments and water were collected and used as we previously described (Ma et al., 2015a). Briefly, four different types of water were collected from sites A, B, C, and D, respectively; two different types of sediments were sampled from sites A and C ( Figure 1A). Initial total nitrogen concentrations in these four types of water were 3.28 mg L −1 , 24.22 mg L −1 , 4.16 mg L −1 , and 0.81 mg L −1 , respectively; While initial total phosphorus contents in the overlying water of sites A, B, C, and D were 0.26 mg L −1 , 2.0 mg L −1 , 0.4 mg L −1 , and 0.07 mg L −1 , respectively. The initial physic-chemical indices of water and sediments in pond 1, ditch, pond 2, and water distribution storage apparatuses were analyzed and presented in Supplementary Tables S1, S2.

Construction of the PDCS Systems
Two sets of rural wastewater treatment system, was designed and operated as we previously described elsewhere (Ma et al., 2021b). Briefly, the small-scale system consisted of six parts: a water distribution bucket, pond 1, ditch, pond 2, a water-storage tank, and a water pump ( Figure 1B). The two ponds were laid with 10 cm sediments sampled from the sites A and C and planted with Vallisneria natans L., respectively. Moreover, in the ditch, a 10 cm thick gravel layer was evenly spread and planted with Iris tectorum L. and Acorus calamus L. Then, these five microcosms were filled with 45 L, 96 L, 35 L, 45 L, and 35 L water sampled from sites D, A, B, C, and D, respectively. In the PDCS system, the running parameters was circulating every other 4 h with a water flow of 3.6 Lh -1 , thus resulting in 0.30 m 3 m −2 d −1 , 0.43 m 3 m −2 d −1 , and 0.37 m 3 m −2 d −1 hydraulic loading rates for pond 1, ditch, and pond 2, respectively. Meanwhile, in the static group, the velocity of water flow was zero.
Approximate 50 g rhizosphere and non-rhizosphere sediments in pond 1, ditch, and pond 2 of these two systems were also collected at days 15, 30, and 60, representing the early, middle, and late stages of the experiment, respectively (Zou et al., 2021). About 10 g of fresh sediment samples was used to measure water contents (WC) of each sediment sample gravimetrically. The contents of organic matter were determined gravimetrically using approximate 5 g fresh sediment sample dried at 450°C for 3 h. The contents of total organic carbon (TOC) and sediment total nitrogen (STN) in air-dried and sieved sediment samples were determined using an elemental analyzer (Vario TOC cube, Hanau, Germany). The contents of NH 4 + -N, NO 3 − -N, NO 2 − -N in sediments were determined using an automatic nutrient analyzer (EasyChem plus, Systea, Italy). Sediment P fractions were measured using the SMT harmonized P extraction protocol, and the four main forms of P in sediment total phosphorus (STP), sediment inorganic phosphorus (SIP), NaOH-P, and HCl-P were analyzed as previously described (Ruban et al., 2001). The left sediments were mixed thoroughly and frozen in −20°C for future analysis. The dynamic changes of physic-chemical characteristics for the non-rhizosphere sediments for the PDCS systems were shown in Supplementary Tables S3, S4, respectively.

Analysis of Root Exudates
The contents of five types of root exudates proteins, polysaccharides, amino acids, lactic acids and tartaric acids were measured (Ma et al., 2021a). Briefly, 10 strains of Vallisneria natas L., 1 strain of Acorus calamus L. and Iris tectorum L. were sampled from the ponds and ditches at days 15, 30, and 60, respectively. However, in the pond 1 of the static group, all Vallisneria natans L. died due to the deterioration of water quality in the early stage of the experiment. Therefore, root exudates for this microcosm were missed. After careful rinsing in the water, plant roots were immersed in 500 ml ultra-pure water for 24 h. The values of pH for these solutions were measured. Then, 400 ml culture solution was concentrated to 20 ml using a vacuum circumgyration evaporator. The contents of proteins and polysaccharides were analyzed using the coomassie brilliant blue method (Spector, 1978) and the anthrone-sulfuric acid chromatometry method (Grandy et al., 2000), respectively. The contents of amino acids were determined by an ICS 5000 + amino acid analyzer (model 120A, Thermo Fisher, United States). Meanwhile, the contents of tartaric acids in root exudates were measured using a DIONEX ICS-5000 + DP chromatograph (Thermo Fisher, United States). The initial and changes of proteins, polysaccharides, amino acids, lactic acids, tartaric acids, and pH of root exudates of aquatic plants were presented in Supplementary Tables S5, S4, respectively.

Construction of 16S rDNA Library
Bacterial DNA were extracted and used for 16S rDNA gene sequencing as previously described with some modifications (Poret-Peterson et al., 2019). Briefly, the total DNA of sediment samples was extracted using the PowerSoil DNA Isolation Kit according the manufacturer's instructions (MoBio, Carlsbad, CA, United States). Then, the concentrations and qualities of DNA samples were spectrophotometrically measured (Nanodrop 2000, NanoDrop Technology, Wilmington, DE, United States). The hypervariable V3-V4 regions of bacterial 16S rDNA were amplified using the specific bar-coded forward primer 5'-ACTCCTACGGGAGGC AGCAG-3' and reverse primer 518R, 5'-ATTACCGCGGCTGCT GG-3'. Polymerase chain reaction (PCR) amplification procedures were as follows: denaturation at 95°C for 3 min, denaturation at 95°C for 30 s, annealing at 55°C for 30 s, extension at 72°C for 30 s, 35 cycles. Following the preamplification, the qualified DNA samples were used in the next PCR amplification. Then, the amplified PCR products were purified and used to construct the 16S rDNA library.

Sequencing of 16S rDNA Gene and Data Analysis
Sequencing of PCR products were performed using the Illumina paired-end sequencing technology (Mega Genomics Corp. Beijing, China). Briefly, the raw data of bacterial 16S rDNA gene sequencing were demultiplexed, qualify-filtered by Trimmomatic using the specific filtering criteria of QIIME (Caporaso et al., 2010) and merged using the FLASH software. Chimeric sequences were determined and deleted using the UCHIME algorithm (Edgar et al., 2011). The operational taxonomic units (OTUs) were clustered with a threshold of 97% sequence similarity using UPARSE (version 7.1, http:// drive5.com/uparse/), annotated and classified using the SILVA database. Sediment bacterial α-diversity was estimated via the observed chao1, Shannon, and Simpson indices.

Statistical Analysis
All data were presented as mean ± SEM. One-way ANOVA with Turkey's post hoc tests were performed to analyze the differences in water quality, sediment physic-chemical parameters, α-diversity and richness indices of sediment bacteria, root exudates in these two systems. Sediment bacterial community richness (presented as observed species and goods coverage) and α-diversity (presented as Chao 1, Simpson, and Shannon) in rhizosphere and non-rhizosphere microenvironments in these two systems were calculated in R version 3.5.2 using the vegan package (Zhou et al., 2020). Spearman's correlation coefficients were calculated to determine the associations among sediment microbial diversity, water quality, sediment properties, and root exudates. RDA was performed to determine the correlations between the top 10 sediments bacterial and abiotic environmental variables using the Canoco software (Version 5.0, Microcomputer, Ithaca, NY, United States). The RDA will be picked up and used as the appropriate ordination form when the longest gradient length of detrended correspondence analysis was <2. All statistical analyses were done in SPSS software (version 20.0, SPSS Inc., Chicago, United States). p < 0.05 was considered statistically significant.

Nutrient Removal Rates in PDCS Systems
The efficiencies of N and P removal in the two systems were demonstrated in Figure 2. In the PDCS, the concentrations of TN, NH 4 + -N, and NO 2 − -N in the three microcosms pond 1, ditch, and pond 2 declined from day 4 and reached a plateau level at day 30. Contents of NO 3 − -N in the PDCS remained at a low level before day 20, and then significantly increased, peaking at Frontiers in Environmental Science | www.frontiersin.org September 2021 | Volume 9 | Article 717458 FIGURE 2 | Changes of TN, NH 4 + -N, NO 3 − -N, NO 2 − -N, TP, and IP in overlying water for two systems within 60 sampling days. In the circulation PDCS system, the water was circulated 3.6 L/h every other 4 h, while the control group was static.
Frontiers in Environmental Science | www.frontiersin.org September 2021 | Volume 9 | Article 717458 5 day 40, and gradually declined to a low level. The final removal efficiencies for TN and NH 4 + -N in the PDCS system ranged from 72.7 ± 2.3 to 94.5 ± 1.5%, and 72.3 ± 2.6 to 99.1 ± 0.3%, respectively. For the P, in the PDCS, the concentrations of TP and IP in the three microcosm's pond 1, ditch, and pond 2 significantly declined from day 4 and reached a stable low level at day 30. The final removal efficiencies of TP in pond 1, ditch, and pond 2 in the PDCS system were 81.7 ± 2.3, 97.4 ± 1.1, and 77.8 ± 2.1%, respectively. Significant differences were detected in TN, NH 4 + -N, NO 3 − -N, NO 2 − -N, TP, and IP between these two systems (p < 0.05).
The values of the other six indices ORP, W-temp, pH, DO, turbidity, and salinity for water were presented in Supplementary Figure S1. The ORPs for the two systems increased from day 1 to day 15, and then decreased. The W-temp for these two systems fluctuated from 25.5 ± 0.3°C to 32.6 ± 0.2°C.The values of pH generally increased from 7.16 ± 0.13 from day 0 to 8.19 ± 0.15 at day 60. The values of DO also increased from 0.05 ± 0.01 to 5.4 ± 0.1 mg L −1 . For the turbidity, their values fluctuated two times and generally decreased. For the salinity, the change trend for the static system were increasing, the values in the PDCS remained at a stable level. Significant differences in pH, DO, salinity, and turbidity were observed between these two systems (p < 0.05).

OTU Analysis in Single Microcosms of the PDCS System
Both rhizosphere and non-rhizosphere sediments were sampled for analyzing the compositions of bacteria in two systems at days 15, 30, and 60, respectively. After quality control and filtering, twenty-four OTU numbers for eight microcosms were obtained ( Figure 3A). Due to the death of plants in the pond 1 of the static system and the existence of two different types of plants in the ditches of the two systems, there were 452,989 ± 3,426 bacterial OTUs observed in the 36 rhizosphere/non-rhizosphere sediment samples. The OTU numbers in the S-ditch, S-pond 2, ditch, and pond 2 microcosms were all significantly lower than those of their corresponding rhizosphere counterparts, respectively (p < 0.05). Meanwhile, the OTU numbers in the microcosm of R-pond 1 was statistically insignificantly higher compared with that in the microcosm of pond 1 (p < 0.05). In order to show the unique and shared OTUs between the static and PDCS systems, a petal diagram was drawn based on the OTU analysis results ( Figure 3B). The ten petals represented ten different microcosms in these two systems. There were 1,360 OTUs shared by the 10 microcosms. The numbers of unique OTUs in the S-ditch, ditch, and pond 2 microcosms were significantly lower than those of their corresponding counterparts (p < 0.05).

Heatmap of the Relative Abundances of Bacterial Communities
Relative abundances of top 35 bacteria at phylum and genus levels were statistically analyzed (Figure 4). The utmost dominant bacterial phyla in pond 2 of the two systems were Firmicutes, Deinococcus, and Armatimonadetes. The frequencies of these three bacterial phyla in the rhizosphere of pond 2 were more abundant in those of the non-rhizosphere. For the ditch of two systems, the dominant bacterial phyla were Bacteroidetes and Proteobacteria, their values in the rhizosphere were significantly higher than those of their corresponding non-rhizosphere groups. In addition, the frequencies of Nitrospirae, and Acidobacteria in the rhizosphere of pond 1 for two systems were significantly higher than those of their corresponding non-rhizosphere groups. At the genus level, Geobacter and Pseudomanas genera were in relatively higher abundances in the rhizosphere of pond 1 for the PDCS than those in the FIGURE 3 | OTU numbers and Venn diagram for two systems. (A) OTU numbers in the pond-1, R-pond-1, ditch, R-ditch, pond-2, R-pond-2 of the PDCS and S-ditch, S-R-ditch, S-pond-2, S-R-pond-2 of the control static group, respectively. Due to the death of plants in the pond-1 of the control static group died at the early stage, the data for the S-pond-1 and S-Rpond-1 were omitted. (B) Each circle represents a single microcosm of the PDCS systems. The numbers in the overlapping area of the circles represent the numbers of OTUs shared between the microcosms, while the numbers without overlaps represent the numbers of OTUs unique to these single microcosms.
Frontiers in Environmental Science | www.frontiersin.org September 2021 | Volume 9 | Article 717458 non-rhizosphere. The frequencies of Clostridium_sensu stricto 1, Romboutsia, and Clostridium_sensu stricto 13 genera in the rhizosphere of pond 1 for two systems were more abundant in those in non-rhizosphere. For the pond 2 of the PDCS, the dominant bacterial phyla were Methenoregula and Methanosaeta, their values in the rhizosphere were significantly higher than those of their corresponding nonrhizosphere groups.
Relative abundances of Firmicutes in two ponds of two systems increased from day 15 to day 30. At the middle of experiment (day 30), Firmicutes was found to be more enriched in the rhizosphere of ponds for two systems than those in the nonrhizosphere of two ponds. Meanwhile, the phylum Proteobacteria was more abundant in the rhizosphere of the ditch for two systems than those in the non-rhizosphere of the ditch at day 30. At the genus level, the most abundant bacteria in both rhizosphere and non-rhizosphere sediments of the PDCS were Bacillus, Geobacter, and Clostridium sensu stricto 1 genera. The relative abundances of these three bacterial genera in the rhizosphere of the PDCS ranged from 0.56 ± 0.02 to 19.52 ± 1.95%, 0.28 ± 0.03 to 8.50 ± 2.13%, and 0.13 ± 0.02 to 11.89 ± 0.36%, respectively. Meanwhile, the relative abundances of these three bacterial genera in the nonrhizosphere ranged from 0.12 ± 0.01 to 15.61 ± 0.68%, 0.20 ± 0.02 to 4.58 ± 0.32%, and 0.26 ± 0.01 to 8.08 ± 1.35%, respectively. Dominant genera in ditches of two systems were Geobacter. Relative abundances of ditches for two systems increased from day 15 to day 30, and then decreased ( Figure 5). Meanwhile, the frequencies of Geobacter in genus Proteobacteria in the rhizosphere sediments of these two ditches at day 30 were significantly higher than those of their non-rhizosphere sediments (p < 0.05). Moreover, in the microcosms pond 1 and pond 2, the most dominant bacterial genera were Bacillus and Clostridium sensus stricto 1, and relative abundances of these two genera in Frontiers in Environmental Science | www.frontiersin.org September 2021 | Volume 9 | Article 717458 FIGURE 5 | Relative abundances of top 10 bacteria at the genus level in the two systems. The relative abundances of top 10 bacteria at the genus level at days 15, 30, and 60 were calculated. Abbreviations: In the static systems, S-ditch, S-R-ditch, S-pond-2, and S-R-pond-2 mean the sediments collected from the ditch, plant rhizosphere in the ditch, pond 2, plant rhizosphere in pond 2, respectively. The plants in the pond 1 of the static system died at the early stage of the experiment. In the PDCS, Pond-1, R-pond-1, ditch, R-ditch, pond-2, and R-pond-2 represented the sediments collected from the pond 1, plant rhizosphere of the pond 1, the ditch, plant rhizosphere in the ditch, pond 2, plant rhizosphere in the pond 2, respectively.
Frontiers in Environmental Science | www.frontiersin.org September 2021 | Volume 9 | Article 717458 8 two ponds of the PDCS remarkably increased from day 15 to 30. The frequencies of Bacillus and Clostridium sensus stricto 1 in the rhizosphere of ponds for the PDCS were higher than those in the non-rhizosphere ( Figure 5).

Richness and α-Diversity Analysis of Rhizosphere and Non-Rhizosphere Sediment Bacteria
To further explore the effects of root exudates on bacterial community in the PDCS, the richness (presented as observed species, goods coverage) and α-diversity (presented as Chao1, Shannon and Simpson) of rhizosphere and non-rhizosphere sediment bacteria were performed. As shown in Figure 6, the values for the observed species in the rhizosphere and nonrhizosphere sediments in the PDCS ranged from 9,560 ± 1,235 to 12,500 ± 1,230 and 9,080 ± 1,254 to 12,100 ± 1,035, respectively. The numbers of observed species in the rhizosphere groups S-Rditch, S-R-pond-2, R-ditch, and R-pond-2 were significantly higher than those of their corresponding non-rhizosphere groups S-ditch, S-pond-2, ditch, and pond-2, respectively. Meanwhile, the values for goods coverage in the rhizosphere and non-rhizosphere sediments in the PDCS ranged from 0.90 to 0.94 and 0.9 to 0.94, respectively. Moreover, Chao 1, Shannon, and Simpson indices in the PDCS ranged from 11,600 ± 1,396 to 16,400 ± 1,543, 11.3 ± 0.3 to 12.0 ± 0.3, and 0.9971 ± 0.0021 to 0.9979 ± 0.0003 in the rhizosphere S-ditch and S-R-ditch meant non-rhizosphere and rhizosphere ditch groups in the static system, respectively; Similarly, S-pond-2, and S-R-pond-2 meant non-rhizosphere and rhizosphere pond-2 groups in the static system, respectively. The same naming rules were applicable in pond-1, R-pond-1, ditch, R-ditch, pond-2, and R-pond-2 in the PDCS.
Frontiers in Environmental Science | www.frontiersin.org September 2021 | Volume 9 | Article 717458 sediments and ranged from 11,300 ± 286 to 15,900 ± 879, 11.1 ± 0.2 to 12.0 ± 0.2, and 0.9970 ± 0.0023 to 0.9977 ± 0.0002 in the nonrhizosphere sediment, respectively. Chao1 values in the five rhizosphere groups S-R-ditch, S-R-pond-2, R-pond-1, R-ditch, and R-pond-2 were all significantly higher than their corresponding non-rhizosphere groups S-ditch, S-pond-2, pond-1, ditch, and pond-2, respectively (p < 0.05). Similarly, the values of Shannon in the S-R-ditch, R-ditch and R-pond-2 groups were significantly higher than those of their corresponding nonrhizosphere groups S-ditch, ditch, and pond-2, respectively (p < 0.05). Moreover, Simpson values in the S-R-ditch, R-pond-1, and R-pond-2 were significantly higher than those of their corresponding non-rhizosphere groups S-ditch, pond-1, and pond-2, respectively (p < 0.05). Spearman's correlation coefficients between microbial diversity indices and water quality were presented in Table   1. In the rhizosphere, Chao 1 index showed significant and positive correlations with NH 4 + -N and turbidity (p < 0.05). Meanwhile, turbidity was significantly positively correlated with the Shannon index (p < 0.01). In the non-rhizosphere, significant positive correlations between the ORP and the Shannon and Simpson indices were observed (p < 0.05); while Simpson index was significantly negatively correlated with W-temp, and was significantly positively associated with salinity (p < 0.05).

Relationship Between Sediment Bacteria and Environmental Factors
As demonstrated in Figure 7A, the first two eigenvalues could explain 48.97 and 13.32% of the total variation at the genus level in the non-rhizosphere sediments, respectively. The environmental variables accounted for 66.9% of the total variation for the microbial community of non-rhizosphere sediments. In the non-rhizosphere sediment groups, SNH 4 + -N and SNO 3 − -N exhibited significant positive effects on the dominant bacterial genera Clostridium sensu stricto1, Clostridium sensu stricto 13, and significant negative effects on Arthrobacter and Geobacter. Moreover, pH, DO, and ORP had strong positive effects on Clostridium sensu stricto 12 and Bacillus.
As demonstrated in Figure 7B, the first two eigenvalues could explain 43.63 and 11.18% of the total variation at the genus level in the rhizosphere sediments, respectively. Environmental variables accounted for 60.3% of the total variation for the microbial community of rhizosphere sediments. In the rhizosphere, lactic acids, proteins, and amino acids showed significant positive effects on Lactobacillus, Geobacter and Clostridiu sensu stricto 12. Moreover, Bacillus, Clostridium sensu stricto 1, and Clostridium sensu stricto 13 were significantly positively affected by W-temp and WC.

DISCUSSION
The Role of Bacteria in Nutrient Removal of the PDCS In the present study, the averaged final TN and TP removal efficiencies of the PDCS at day 60 were mostly higher than that in IVCWs (51.6%) (Chang et al., 2015), in line with the mean value in algal pond combined with constructed wetlands (69.74%) (Zhao et al., 2016), yet lower than that in SFCWs (96.14%) (Li et al., 2020), suggesting the PDCS system was a valuable choice for rural wastewater remediation. Our previous study have demonstrated that microorganisms play important role in N and P removal in the PDCS (Ma et al., 2019). The removal of N was mainly contributed to simultaneous nitrification, anaerobic ammonium oxidation, and denitrification processes via the nitrogen-related bacteria . In this paper, Bacillus was one of the three predominant genera in ponds and the ditch for the two systems, respectively. This finding was in line with the previous study reported that Bacillus was one of the six predominant bacterial genera for heterotrophic nitrification and aerobic denitrification in aquatic ecosystems (Qiao et al., 2020). Previous studies reported that Bacillus. sp. participated in aerobic denitrification in the PDCS, and Clostridium sp. was used for anaerobic ammonium-oxidizing in the sludge fermentation reactor (Yang et al., 2011;Ma et al., 2021b). A earlier study reported that rhizosphere bacteria played a pivotal role in regulating the P transformation and can utilize various P forms in wetland plants (Teng et al., 2018). Geobacter and Bacillus. sp. were known to act as polyphosphate-accumulating organisms, showing a strong capacity of poly-P accumulation and P storage (Schelfhout et al., 2015;Wang et al., 2019).

Comparison of Microbial Community Structure and Diversity Between Rhizosphere and Non-Rhizosphere Bacteria
The "rhizosphere effect" can lead to greatly variations in compositions and diversities of rhizosphere and nonrhizosphere sediment bacteria, where the impacts were influenced by seasons (Dewedar et al., 2009), soil types (Zhao et al., 2017), and plant species (Yin et al., 2018). The OTU numbers in the rhizosphere sediments of five microcosms were all higher than those of their counterparts in the non-rhizosphere, suggesting the presence of more bacteria and more complex microbial communities in the rhizosphere sediments. Moreover, α-diversities of rhizosphere groups S-R-ditch, S-R-pond-2, R-pond-1, R-ditch, and R-pond-2 were all significantly higher than their corresponding counterparts (p < 0.05). These results indicated that compared with the non-rhizosphere, both the richnesses and diversities of microorganisms in the rhizosphere were enhanced due to the exudates from the plant roots (Stringlis et al., 2018). These effects will indirectly promote the roles of plants in nutrient removal in the PDCS. This was in line with the previous study in wetland plants which reported that "rhizosphere effects" affected the removal efficiency of nutrient via regulating the density and diversity of the rhizosphere microbes (Chen et al., 2016). Similarly, a recent study have reported that "rhizosphere effects" can result in the increase of relative abundances of some beneficial bacteria such as Streptomycetaceae and Bacillaceae, which exhibited significantly positive correlations with the uptake of N (Wu et al., 2021).
Both non-rhizosphere and rhizosphere groups shared five predominant bacterial genera, Bacillus, Geobacter, Clostridium sensu stricto 1, Clostridium sensu stricto 12, and Clostridium sensu stricto 13. These findings were consistent with the previous research in tidal marsh soil reported that relative abundances of Geobacter, Bacillus, Clostridium and Shewanella in the rhizosphere were higher than those in the bulk soil (Luo et al., 2018). However, at day 60, the relative abundance of Geobacter in the ditch of the PDCS was richer than those of two ponds; while frequencies of Bacillus and Clostridium sensu stricto 1 in two ponds of the PDCS at day 60 were significantly higher than those in the ditch. This was mainly due to that differences in environmental factors such as soil types (Bakker et al., 2015) and plant species (Yin et al., 2018), resulting in varying root exudates composition and contents (Ma et al., 2021b), thereby affected the community structure of bacteria. Moreover, the frequencies of Bacillus and Clostridium sensus stricto 1, and Geobacter in the rhizosphere of ponds and the ditch for two systems at day 30 were higher than those in the non-rhizosphere ( Figure 5). This was consistent with our results that protein and amino acids contents of Vallisneria natans L. in PDCS's two ponds increased from day 15 to 30 (Supplementary Table S5). This suggested that plants could adjust the secretion to cope with abiotic environmental stress (Edayilam et al., 2018). Relative higher proteins and amino acids values were conductive to increment of rhizosphere bacteria, which would be helpful in promoting the removal of nutrients (Bakker et al., 2015).

Responses of Rhizosphere and Non-Rhizosphere Bacteria to Abiotic Environmental Stress
Root exudates such as amino acids, organic acids, and sugars could provide carbon source for microbial growth and can drive sediment bacterial population richness and activities (Raaijmakers et al., 2009;Berendsen et al., 2012). Meanwhile, root exudates excreted by plants distributed in strong gradient manners from root surface into the soil. These root exudate gradients exert selective pressures, which affected the local microbial community structure such as their abundance and composition (Nunes da Rocha et al., 2009). Meanwhile, WC and W-temp played important roles in modulating the community structure of bacteria (Liddell et al., 2007;Gaumont-Guay et al., 2008). This was confirmed by our observations that Chao 1 indices of the rhizosphere and non-rhizosphere bacteria were both significantly correlated with water content ( Table 2); while significant negative relationship between Simpson index and W-temp was only detected in the non-rhizosphere (Table 1). Bacillus and Clostridium sp. in the rhizosphere were significantly affected by water content and W-temp. Higher water content could enhance oxygen availability and diffusion rates in the sediment, which was benefit for Bacillus growth (Ma et al., 2021b). This was confirmed by our finding that the relative abundance of Bacillus was positively associated with DO in the non-rhizosphere ( Figure 7A). In addition, Chao 1 index showed significant positive correlations with all measured sediment properties in the rhizosphere and nonrhizosphere ( Table 2). Compared to the rhizosphere, Bacillus and Clostridium sp. in the non-rhizosphere were mainly affected by SNH 4 + -N and SNO 3 − -N ( Figure 7A). Nitrification and denitrification always occurred in adjacent area resulting in the NO 3 − formed by nitrification diffusing towards an anaerobic zone where it was the terminal electron receptor for the denitrification process (Lu et al., 2012).

CONCLUSION
The PDCS system exhibited great capabilities in reducing nutrients N and P from rural wastewater, with the removal efficiencies ranging from 72.7 to 97.4%. Microorganisms were largely responsible for N and P removal, however, the community structures and diversities of rhizosphere and nonrhizosphere sediment bacteria varied greatly. The average numbers of bacterial OTUs in the rhizosphere sediments were significantly higher than their corresponding counterparts (p < 0.05). In the PDCS, the different microcosms in both rhizosphere and non-rhizosphere sediments shared dominant bacterial genera, such as Bacillus and Clostridium sensu stricto 1 for the ponds, and Geobacter for the ditch, respectively. Meanwhile, the contents of these three bacterial genera in the rhizosphere were higher than those of the non-rhizosphere at day 30. Moreover, Chao 1 index in both rhizosphere and non-rhizosphere was significantly positively correlated with all measured sediment properties, such as STN, STP, and OM. In the rhizosphere, Chao 1 index showed significant positive associations with NH 4 + -N, turbidity, tartaric acid, and Shannon index was significantly positively correlated with turbidity and TOC (p < 0.05). Similarly, in the non-rhizosphere, the Shannon and Simpson indices were associated with ORP (p < 0.05), respectively. RDA analysis demonstrated that exudates such as lactic acids, proteins, and amino acids exhibited strong positive effects on Geobacter and Clostridiu sensu stricto 12 in the rhizosphere; while Bacillus and Clostridium were significantly associated with SNH 4 + -N and SNO 3 − -N in non-rhizosphere. Together, regulating root exudates and sediment available N will increase the richness and diversity of dominant microbial species in the rhizosphere and non-rhizosphere. These findings are beneficial for the optimization and design of PDCSs and other nutrient removal systems in large-scale wastewater treatment applications.

DATA AVAILABILITY STATEMENT
The data presented in the study are deposited in the National Genomics Data Center repository, accession number CRA004345.