Nitrogen Retention in Mesocosm Sediments Received Rural Wastewater Associated with Microbial Community Response to Plant Species

Vegetated drainage ditches (eco-ditches) have drawn much attention in recent years for the ability to remediate diffuse contaminants in rural wastewater through sediment retention, plant uptake and interception, and microbial metabolic activities. However, the effect of plant species on microbial community structure and nitrogen (N) retention in ditch sediment remains poorly understood. In this study, mesocosm plastic drums were planted with eight plant species commonly found in ditches and nurtured with wastewater for 150 days. Sediment total nitrogen (TN) was greatly increased after 150-day nurturing with rural wastewater, from 296.03 mg·kg−1 (Iris japonica Thunb) to 607.88 mg·kg−1 (Acorus gramineus O). This study also presents the effect of different plant species on sediment microbial communities, thus providing insight into N removal mechanisms in eco-ditch. Fifty-eight differentially abundant taxa were identified, and sediment microbial community structure for no plant (CK), Acg, Canna indica (Cai), and Typha latifolia L. (Tyl) was primarily linked to sediment NH4-N and TN. Extremely small proportions of ammonia oxidizing bacteria (AOB) and nitrifying bacteria were detected for all treatments, but large proportions of Crenarchaeota, which comprises the widely existent ammonium oxidized archaea (AOA), were found in CK, Acg and Cai. The abundance of Nitrosotalea from Crenarchaeota presented positive correlations with sediment NH4-N contents and ammonia oxidation function predicted by Faprotax, indicating Nitrosotalea might be the dominant ammonium-oxidizing microbes in sediment samples. The probable NH4-N removal pathway in wastewater sediment was through a combined effect of AOA, nitrifying bacteria, and anammox.


Introduction
Rural wastewater derived from agricultural and domestic activities is a major source of water pollution. The pollutants from rural wastewater include high concentration of organic matter and a certain amount of N, which may cause eutrophication in natural water bodies when the wastewater is discharged directly without treatment [1]. Conventional wastewater treatment processes are prohibitively expensive and not entirely feasible for widespread application in rural areas [2,3]. Thus, ecological technologies have gained increasing attention in the past two decades for the successful and

Mesocosms Plastic Drum Sediment (MPDS) Experiment
The MPDS was designed to evaluate the effect of different plant species on pollutants removal from rural wastewater. The drums were 55 cm (h) × 28.5 cm (d) with a vertical perforated PVC pipe installed in the centre of each plastic drum to pump wastewater [9]. The cultivation experiment of MPDS was conducted in a greenhouse constructed of polyethylene film at Yanting Agro-Ecological Station of Purple Soil, Chinese Academy of Sciences. The selected plant species were nurtured with pond water for approximate 2 weeks until the plants developed root systems. Healthy and uniformly sized young plants with an average height of 10-15 cm were then rinsed and transplanted into the MPDS drums at a density of 10-12 plant/m 2 . The drums were filled with 30-cm depth of sediment collected from natural ditch around the station. Rural wastewater obtained from the ditches was pumped through pipes into each drum. The wastewater was added once per week during the experiment to compensate for evaporation and maintain a constant water level (35 cm). The properties of pond water, Water 2020, 12, 3035 3 of 13 rural wastewater, and sediment are shown in Table 1. Each plant treatment has three replicates, and two control (CK) drums were left unplanted. Air temperatures in the plastic house were maintained between 14-38 • C during the experimental period. The plants flourished and the sediment microbes settled after being acclimated for 150 days.

Sediment Sampling and Chemical Analyses
Sediment samples were collected at the end of the cultivation experiment from each MPDS system, and vegetable matter was discarded. One subsample was stored at 4 • C for chemical analyses, while another subsample was stored at -80 • C for molecular analysis. NO 3 -N and NH 4 + -N were extracted with 2 M KCl and analyzed using a continuous-flow analyzer (Model AA3; Bran + Luebbe, Norderstedt, Germany). TN was determined by semi-micro Kjeldahl digestion using Se, CuSO 4 , and K 2 SO 4 as catalysts. Dissolved organic carbon (DOC) was extracted with 0.5 M K 2 SO 4 and analysed using a TOC-5000 analyzer (Shimadzu, Kyoto, Japan). Organic Carbon was determined based on dichromate oxidation and titration with (NH 4 ) 2 Fe(SO 4 ) 2 ·6H 2 O. The pH was measured from fresh soil-water suspensions (1:2.5 w/v). The sediment moisture content was calculated by determining the weight loss of sediment sample after drying at 105 • C for 24 h. TP was analyzed using the alkaline potassium persulfate digestion-ultraviolet spectrophotometric method.

DNA Extraction and MiSeq Sequencing of 16S rRNA Gene Amplicons
DNA was extracted from sediment samples using the MOBIO Power Soil DNA Extraction kit (MoBio Laboratories, Carlsbad, CA, USA). DNA concentration and quality were checked using a NanoDrop Spectrophotometer. Sediment DNA was diluted to 10 ng·µL −1 and stored at -40 • C for downstream use. The universal primers 515F-909R with 12 nt unique barcode were used to amplify the V4 region of the 16S rRNA gene [21,22]. The PCR reaction mixture (25 µL) contained 10 ng sediment DNA, 1 × PCR buffer, 1.5 mM MgCl 2 , 0.4 µM of each deoxynucleoside triphosphate, 1.0 µM of each prime, and 0.5 U Ex Taq (TaKaRa, Dalian, China). The PCR cycling profiles were as follows: initial denaturation at 94 • C for 3 min; 30 cycles of 94 • C for 40 s, 56 • C for 60 s, and 72 • C for 60 s; followed by 72 • C for 10 min. Two individual PCR reactions were conducted for each sample, and the amplification products were combined and examined using 1.0% agarose gel electrophoresis. The correct-sized amplicons were excised from the gel and purified using the SanPrep DNA Gel Extraction Kit (Sangon Biotech, Shanghai, China). Then the purified products were quantified and equal molar amount from each sample were pooled together. Sequencing libraries were prepared using a TruSeq DNA kit. Then sequencing was performed using an Illumina Miseq system with the Reagent Kit v22 × 250 bp, as described by the manufacturer' instructions.

Pyrosequence Data Analysis and Putative Function Prediction
Raw sequencing data were processed using QIIME Pipeline [23]. All sequence reads were trimmed and assigned to each sample based on their barcodes. The sequences with low quality and Water 2020, 12, 3035 4 of 13 shorter lengths were removed, and the qualified sequences were assigned to operational taxonomic units (OTUs) at a 97% similarity level. All the samples were randomly resampled to 10,458 reads. Alpha-diversity (chao1 estimator of richness, Shannon's diversity index, and Simpson index) and beta-diversity (principal coordinates analysis (PCoA)) were calculated based on weighted UniFrac analyses. OTUs were classified taxonomically using the Ribosomal Database Project classifier. Obtained nucleotide sequences were deposited with the European Nucleotide Archive (ENA) under accession numbers ERS2573254-RS2573279. The putative ecological functions of the prokaryotic OTUs were assessed using Functional Annotation of Prokaryotic Taxa (FAPROTAX) and visualised using TBtools (v 0.665) [24].

Statistical Analyses
One-way analysis of variance (ANOVA) was used to test differences in sediment properties, relative abundance of taxonomic units, and alpha diversity among plant treatments. Significance of taxonomic differences among treatments was further examined using Linear discriminant analysis (LDA) Effect Size (LEFse) [25]. The method employs the Kruskal-Wallis sum-rank test (α = 0.015) to identity taxa with significantly different abundances between treatments based on against all comparisons, followed by LDA to estimate the effect size of each differentially abundant taxa (LDA score > 3). The correlations between sediment properties, decreased N and the 50 most abundant microbial genera were assessed by Pearson's correlation. Redundancy analysis (RDA) were performed using the vegan package in R (R Project 3.1, v.2.3-1) to determine the environmental variable that best explained the variation of microbial community composition in plant-treated sediment.

Nutrient Retention in Sediment
Nutrient retention in sediment samples from the MPDS system varied with different plant treatments after 150-day nurturing with rural wastewater ( Table 2). NH 4 + -N concentrations in sediments were lower than 0.05 mg·kg −1 for all treatments. NO 3 -N concentrations were significantly lower in Tyl (0.04 mg·kg −1 ) compared with Acg (6.09 mg·kg −1 ) and Phh (8.64 mg·kg −1 ). Sediment TN content increased for all treatments after nurturing with rural wastewater; the highest TN content was found in Acg (1.18 g·kg −1 ), followed by CK, Rum, Cai, Phh, Myv, Tyl, Eic, and Ijt, ranging from 0.87-1.09 g·kg −1 . TP concentrations also increased for all plant treatments and were significantly higher for Ijt, Myv and Eic than for CK. However, DOC, organic C, and pH did not differ significantly among treatments.

Taxon Richness and Diversity Coverage Assessment
Based on a 16S rRNA gene sequence similarity cut-off value of 97% for 271,908 reads, a total of 36,344 OTUs were observed in the 26 sediment samples. Rarefaction curves enabled the assessment of differences in species richness among sediment samples with different plant treatments ( Figure 1). Sediments treated with plants had steeper rarefaction curves with higher taxon richness, compared with the CK treatment, indicating that the number of observed species was greater in sediments with plants. The Chao 1 estimator demonstrated that microbial community richness ranged from 8133 to 9467 taxa in the plant-treated systems compared with 5963 taxa in CK, which illustrated the significant impact of plants on microbial community richness ( Figure 1). Moreover, the average Shannon's diversity index values of microbial communities in Ijt (11.35), Tyl (11.34), Phh (11.25), Myv (11.09), and Cai (10.75) were significantly higher than Eic (9.24) and CK (7.63) (p < 0.05). The Simpson diversity index demonstrated a similar pattern with the Shannon's diversity with values ranging from 0.979 to 0.999 in plant-treated sediment samples compared with 0.955 in CK. Data (mean ± SD) in the same column with different small letters indicate significantly different (p < 0.05).

Taxon Richness and Diversity Coverage Assessment
Based on a 16S rRNA gene sequence similarity cut-off value of 97% for 271,908 reads, a total of 36,344 OTUs were observed in the 26 sediment samples. Rarefaction curves enabled the assessment of differences in species richness among sediment samples with different plant treatments ( Figure 1). Sediments treated with plants had steeper rarefaction curves with higher taxon richness, compared with the CK treatment, indicating that the number of observed species was greater in sediments with plants. The Chao 1 estimator demonstrated that microbial community richness ranged from 8133 to 9467 taxa in the plant-treated systems compared with 5963 taxa in CK, which illustrated the significant impact of plants on microbial community richness ( Figure 1). Moreover, the average Shannon's diversity index values of microbial communities in Ijt (11.35), Tyl (11.34), Phh (11.25), Myv (11.09), and Cai (10.75) were significantly higher than Eic (9.24) and CK (7.63) (p < 0.05). The Simpson diversity index demonstrated a similar pattern with the Shannon's diversity with values ranging from 0.979 to 0.999 in plant-treated sediment samples compared with 0.955 in CK.

Prokaryotic Community Composition
Microbial community composition also reveals important information regarding the impact of different plants species on contaminant removal. The relative abundance of prokaryotic 16S rRNA gene in the sediment samples at phylum level is shown in Figure 2. A total of 77 different phyla were detected in the 26 sediment samples. Proteobacteria, Crenarchaeota, Chloroflexi, Acidobacteria, Bacteroidetes, Planctomycetes, Actinobacteria, Euryarchaeota, Nitrospira, Firmicutes, [parvarchaeota], Spirochaetes, Gemmatimonadetes, WS3, and Chlorobi were the 15 most abundant phyla (in descending order of abundance). A total of 529 orders were detected, of which 110 orders were shared by all
PCoA based on weighted Unifrac distances was conducted to statistically assess the similarity of prokaryotic communities in different plant-treated sediments. The results of PCoA analysis with maximum variation of 58.48% (PC1) and 15.19% (PC2) are shown in Figure 2b. PCoA analysis revealed that prokaryotic communities were separated by plant treatment. Phh, Myv, Tyl, and Ijt were separated from other treatments on the PCoA 1 axis.

Effect of Plant Species on Putative Functions of Sediment Prokaryotic Communities
A total of 4439 OTUs (out of 36,343 OTUs) were classified into 91 specific functional categories according to the FAPROTAX prediction. The significant correlation between predicted functional composition and taxonomic composition (Mantel r = 0.301, p = 0.003) indicated the prediction is significative for all samples. The functional capacity of N metabolism was significantly affected by plant species, including N fixation, nitrification, and denitrification ( Figure 4). The N fixation process was enhanced by Ijt treatment (p < 0.05) and positively correlated with NO 3 --N (r = 0.484, Table S2). The predicted nitrification was reduced in Ijt, Phh, Rum, and Myv (p < 0.05) and demonstrated positive correlations with TN (r = 0.577, p < 0.01) and NH 4 + -N (r = 0.301, p = 0.136).

Correlations between Environmental Variables and Microbial Composition
RDA was performed to correlate the relative abundance of prokaryotic community (the top 50 abundant genera in total classified sequences were selected) with sediment parameters to determine the major environmental variables that impact the microbial community structure. As shown in Figure 5, this model attributed 52.9% of variance to species-environment correlations. The first and second axes represented 20.5% and 17.9% of variation, respectively. NH 4 + -N and TP were significant variables (p < 0.01), suggesting they were major environmental factors influencing the microbial community structure. Microbial community in CK, Acg, Cai, and Tyl were primarily linked to NH 4 + -N and TN, while those in Ijt, Myv and Phh treatments were associated with TP. The abundances of  process was enhanced by Ijt treatment (p < 0.05) and positively correlated with NO3 --N (r = 0.484, p< 0.05, Table S2). The predicted nitrification was reduced in Ijt, Phh, Rum, and Myv (p < 0.05) and demonstrated positive correlations with TN (r = 0.577, p < 0.01) and NH4 + -N (r = 0.301, p = 0.136). Further, anaerobic ammonium oxidation (anammox) was also positively related to NH4 + -N (r = 0.552, p < 0.01). Denitrification and nitrite ammonification were positively related to NO3 --N (r = 0.135, p= 0.547; r = 0.445, p < 0.05, Table S2) with the lowest values in Cai.

Correlations between Environmental Variables and Microbial Composition
RDA was performed to correlate the relative abundance of prokaryotic community (the top 50 abundant genera in total classified sequences were selected) with sediment parameters to determine the major environmental variables that impact the microbial community structure. As shown in Figure 5, this model attributed 52.9% of variance to species-environment correlations. The first and second axes represented 20.5% and 17.9% of variation, respectively. NH4 + -N and TP were significant variables (p < 0.01), suggesting they were major environmental factors influencing the microbial community structure. Microbial community in CK, Acg, Cai, and Tyl were primarily linked to NH4 + -N and TN, while those in Ijt, Myv and Phh treatments were associated with TP. The abundances of SAGMA-X (no rank), Nitrosotalea (family SAGMA-X), Ellin6513 (Acidobacteria), and

Discussion
Eco-ditches have been successfully employed to treat rural contaminants [26,27]. Plants, one of the most vital components for N uptaking from the eco-ditch system, provide large surface areas for microbial colonisation that improve the removal ability of the rhizosphere [9]. Our previous study suggested Tyl, Myv, and Rum had the highest efficiency for N accumulation, followed by Eic, Phh, Cai, Acg, and Ijt [9]. The results of the current study demonstrated that N retention in sediments of MPDS systems increased after 150-day treatment with wastewater, exhibiting 90-106% increases in Acg and CK, with 51-68% increases observed in other plant treatments ( Table 2 and Table S1). These results indicated that the N retention pathway in sediment was significantly impacted by plant species. Different plant species vary in radial oxygen loss, root exudates, and their subsequent impact on the sediment environment. The specific microbial populations of different kinds of plants are associated with specific root exudates and rhizodepositions [28]. Results of 16S rRNA gene pyrosequencing demonstrated that sediment microbial abundance was significantly impacted by plant species (Figures 2 and 3). Thus, the difference in sediment N retention among plant treatments was linked to microbial community structure in the sediment, as expected.
Conventional nitrogen removal from wastewater is known to comprise two key microbial processes: nitrification and denitrification. Nitrification is the microbially-mediated oxidation of ammonia (NH3) to nitrite (NO2 -) and nitrate (NO3 -), which may ultimately be removed from the

Discussion
Eco-ditches have been successfully employed to treat rural contaminants [26,27]. Plants, one of the most vital components for N uptaking from the eco-ditch system, provide large surface areas for microbial colonisation that improve the removal ability of the rhizosphere [9]. Our previous study suggested Tyl, Myv, and Rum had the highest efficiency for N accumulation, followed by Eic, Phh, Cai, Acg, and Ijt [9]. The results of the current study demonstrated that N retention in sediments of MPDS systems increased after 150-day treatment with wastewater, exhibiting 90-106% increases in Acg and CK, with 51-68% increases observed in other plant treatments ( Table 2 and Table S1). These results indicated that the N retention pathway in sediment was significantly impacted by plant species. Different plant species vary in radial oxygen loss, root exudates, and their subsequent impact on the sediment environment. The specific microbial populations of different kinds of plants are associated with specific root exudates and rhizodepositions [28]. Results of 16S rRNA gene pyrosequencing demonstrated that sediment microbial abundance was significantly impacted by plant species (Figures 2  and 3). Thus, the difference in sediment N retention among plant treatments was linked to microbial community structure in the sediment, as expected.
Conventional nitrogen removal from wastewater is known to comprise two key microbial processes: nitrification and denitrification. Nitrification is the microbially-mediated oxidation of ammonia (NH 3 ) to nitrite (NO 2 -) and nitrate (NO 3 -), which may ultimately be removed from the system by reduction to dinitrogen by denitrification or anammox. The rate-limiting step in the nitrification process is the oxidation of NH 3 to NO 2 -, which is performed by AOB including only a few special genera, such as Nitrosomonas, Nitrosospira and Nitrosococcus [29]. In this study, extremely small proportions of these three genera were detected in all samples. However, we found large proportions of Crenarchaeota in CK, Acg and Cai, which includes the widely existing ammonium-oxidising archaea (AOA) and is implicated in the process of oxidising ammonia to nitrite. Quantitive PCR of gene copies in a local eco-ditch sediment also showed that the absolute abundance of archaeal amoA was 10-100 fold higher than bacterial amoA genes ( Figure S2). Nitrosotalea (from Crenarchaeota) abundance demonstrated positive correlations with sediment NH 4 + -N contents ( Figure 5) and ammonia oxidation function predicted by Faprotax (r = 0.860, p = 0.000), indicating that Nitrosotalea might be the dominant ammonium-oxidising microbes in sediment samples. Previous studies reported that Nitrososphaera was the dominant AOA genus in natural freshwater wetland [30,31]. In the present study, Candidatus Nitrososphaera occupied 0.52-1.48% for planted treatments and 0.30% for CK. Plant root exudes organic carbon compounds, which are essential for the growth of Nitrososphaera and could be conductive to the high proportion of Nitrososphaera [32]. Nitrospira is generally considered to be the major contributor for oxidation of nitrite to nitrate and occupied 0.56-0.70% of all detected sequences in Myv, Cai, and Tyl, which was significantly higher than other treatments. Nitrospira abundance was positively related to nitrite oxidation predicted by Faprotax (r = 0.541, p < 0.01) and negatively related to NO 3 -N (r = −0.451, p < 0.05). We also found a positive relationship between NH 4 + -N and anammox (r = 0.552, p < 0.01).
Therefore, the key contributors for NH 4 + -N removal from wastewater sediment probably include ammonia-oxidising archaea, nitrifying bacteria and anammox. Denitrification is the reduction process from nitrification product(s) to gaseous nitrogen compounds (mainly dinitrogen) and is widely used in bioecological wastewater treatment to remove NO 3 --N.
Especially in the anoxic sediments of aquatic systems, denitrifying bacteria play the most significant part in long-term N removal [33]. The abundances of the core families responsible for denitrifying processes, Rhodocyclaceae (Rhodocyclales) and Comamonadaceae (Burkholderiales) [34,35], varied significantly with different plant treatments. The order Rhodocyclales exhibited the highest score in Tyl treatment, in which the genus Dechloromonas (Rhodocyclaceae) was also predominant ( Figure 3). Dechloromonas is linked to denitrification and is frequently found in wastewater treatment plants. Thaurea, another important denitrifying genus in the family Rhodocyclaceae, exhibited significantly higher abundance in Tyl (0.21%) than in Eic, Myv, and CK (0.03-0.06%) (p < 0.05). Higher gene copies of denitrifiers (nosZ) were also detected in Tyl than that in Phh-and Cai-treated ditch sediment using quantitive PCR ( Figure S2). The family Comamonadaceae also comprises anaerobic denitrifiers [36]. A dominant genera of Comamonadaceae (no rank) found in Ijt is reportedly capable of denitrification [37,38]. A previous study reported that DA052 (affiliated with Acidobacteria) had the potential to utilise oxidised N, such as NO 2 and NO 3 -, as additional N sources, to carry out assimilatory sulfate reduction [39]. This study reported that DA052 abundance was higher in CK (12.84%) and Eic (7.87%) than other plant treatments (0.01-4.48%). Plant treatments improved the O 2 content in sediment and the survival of Bacteroidetes, which exhibited higher abundance in plant-treated sediments (especially Phh, Ijt, Rum, Tyl, and Myv) compared with CK ( Figure 2, p < 0.05). Members of Flavobacterium from Bacteroidetes have been reported to be capable of denitrification [40][41][42][43]. In summary, unlike the autotrophic nitrifying bacteria responsible for nitrification, denitrifying bacteria are composed of ubiquitous, heterotrophic organisms [44]. The low NO 3 --N content in Tyl and CK might be associated with denitrifier differentiation, including heterotrophic denitrifiers, aerobic denitrifiers, facultative autotrophic bacteria, etc. However, the predicted denitrification was not significantly associated with NO 3 --N content (r = 0.135, p = 0.547). It should be noted the FAPROTAX prediction is based on cultured strains and these results warrant further verification.

Conclusions
In this study, we analysed the prokaryotic community composition in sediments from MPDS systems treated with different plant species. The results demonstrated that N retention in sediments, as well as microbial richness and diversity, were significantly affected by plant species. A total of 58 differentially abundant taxa were identified by LEfSe analysis. This study indicated that specific functional microbial differentiation among plant treatments was associated with removal of certain N contaminants from the sediment of eco-ditches. The probable NH 4 + -N removal pathway in wastewater sediment was through a combination of AOA, nitrifying bacteria, and anaerobic ammonium oxidation. Additionally, NH 4 + -N and TP concentrations were identified as major environmental factors influencing microbial community structure, according to RDA. Ijt treatment significantly enhanced the ability of the microbial community to remove N contaminants from the sediment.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4441/12/11/3035/s1. Table S1: Amount and percentage of increased TN in sediment during the 150-day experiment in MPDS systems treated with different plant types. Table S2: Correlations between soil chemical properties and N cycling processes predicted by FAPROTAX.

Conflicts of Interest:
The authors declare no conflict of interest.