Effectiveness of Exogenous Fe2+ on Nutrient Removal in Gravel-Based Constructed Wetlands

A group of microcosm-scale unplanted constructed wetlands (CWs) were established to evaluate the effectiveness of exogenous Fe2+ addition on ammonium nitrogen (NH4+-N), nitrate nitrogen (NO3−-N), and total phosphorus (TP) removal. The addition of Fe2+ concentrations were 5 mg/L (CW-Fe5), 10 mg/L (CW-Fe10), 20 mg/L (CW-Fe20), 30 mg/L (CW-Fe30), and 0 mg/L (CW-CK). The microbial community in CWs was also analyzed to reveal the enhancement mechanism of pollutant removal. The results showed that the addition of Fe2+ could significantly (p < 0.05) reduce the NO3−-N concentration in the CWs. When 10 mg/L Fe2+ was added and the hydraulic retention time (HRT) was 8 h, the highest removal rate of NO3−-N was 88.66%. For NH4+-N, when the HRT was 8–24 h, the removal rate of CW-Fe5 was the highest (35.23% at 8 h and 59.24% at 24 h). When the HRT was 48–72 h, the removal rate of NH4+-N in CWs with 10 mg/L Fe2+ addition was the highest (85.19% at 48 h and 88.66% and 72 h). The removal rate of TP in all CWs was higher than 57.06%, compared with CW-CK, it increased 0.63–31.62% in CWs with Fe2+ addition; the final effluent TP concentration in CW-Fe5 (0.13 mg/L) and CW-Fe10 (0.16 mg/L) met the class III water standards in Surface Water Environmental Quality Standards of China (GB3838-2002). Microbical diversity indexes, including Shannon and Chao1, were significantly lower (p < 0.05) in Fe2+ amended treatment than that in CW-CK treatment. Furthermore, phylum Firmicutes, family Carnobacteriaceae, and genus Trichococcus in Fe2+ amended treatments was significantly (p < 0.05) higher than that in CW-CK treatment. Fe3+ reducing bacteria, such as Trichococcus genus, belonging to the Carnobacteriaceae in family-level, and Lactobacillales order affiliated to Firmicutes in the phylum-level, can reduce the oxidized Fe3+ to Fe2+ and continue to provide electrons for nitrate. It is recommended to consider adding an appropriate amount of iron into the water to strengthen its purifying capacity effect for constructed artificial wetlands in the future.


Introduction
Diffuse pollution caused by agricultural production is one of the main causes of water pollution. In the United States, non-point source pollution accounts for more than 80% of total pollution; in Europe, more than 50% of river pollution is caused by agricultural nonpoint source pollution [1,2]. In China, non-point pollution from agriculture has also become the dominant source and key problem of water pollution control, as point source pollution has been effectively controlled [3]. Agricultural non-point source pollution sources are scattered and hidden, spatio-temporal patterns of the time and space of pollution are the bacterial community structure in biofilms from substrates under different Fe 2+ concentrations, and to (3) determine the optimal Fe 2+ input that can enhance the efficiency of constructed wetlands.

Experimental Description
A group of microcosm-scale unplanted constructed wetlands (CWs) made of polyethylene plastic with a thickness of 0.2 cm, a height of 40 cm, and 15 cm inner diameters were established. The specific structure and design are shown in Figure 1. The two outlets of the reactor are separated by 15 cm. The reactors are 15 cm apart with two water outlets from the bottom to the top. Each reactor is filled with 2.5 L of fine gravel (1-2 cm in diameter, the specific surface area is 0.03 g/m 2 , and the pore size is 82.99 nm). The reactors were placed in Greenhouse No. 3, Northeast Institute of Geography and Agroecology, Chinese Academy of Sciences, at a temperature of 25-30 °C.

Synthetic Wastewater and System Operation
The synthetic wastewater with a pH of 6.8-7.2 was from tap water. According to the nutrient concentration of paddy field drainage in the Sanjiang Plain, the concentrations of NH4 + -N, NO3 -N, and TP were set to 5 mg/L, 7 mg/L, and 1 mg/L, respectively. NO3 − -N, NH4 + -N, and TP in the synthetic wastewater were prepared with KNO3, NH4Cl, and KH2PO4, however, the existence of Fe 2+ requires acidic conditions, and the simulated wastewater needed to be configured with tap water. Tap water contains a large number of chloride ions, which will quickly oxidize Fe 2+ to Fe 3+ . A metal chelating agent can use the strong binding effect of complex agent molecules and metal ions. The metal ions are encapsulated inside the chelating agent to become stable compounds with higher relative molecular weight, thus preventing the precipitation of metal ions and controlling the amount of free ferrous ions in the solution. Therefore, the oxidation of Fe 2+ can be controlled, to a certain extent, through the use of an appropriate complexing agent. Citric acid is one of the most widely distributed small molecule acids in nature. Liang et al. and Zhang et al. [17,18] found that the addition of citric acid can effectively control the amount of free ferrous ions in the system and ensure the continuous generation of free ferrous ions. The Fe 2+ solution was made by adding FeSO4·7H2O and C6H10O8C6, according to the preliminary test; when the molar ratio of the Fe 2+ to citric acid was 1:7, Fe 2+ is oxidized the slowest. The experiments included two phases, in the first phase, it was a 45-day stabilization period, the bacterial communities were fed by 1/4 Hoagland nutrient solution, and

Synthetic Wastewater and System Operation
The synthetic wastewater with a pH of 6.8-7.2 was from tap water. According to the nutrient concentration of paddy field drainage in the Sanjiang Plain, the concentrations of NH 4 + -N, NO 3 -N, and TP were set to 5 mg/L, 7 mg/L, and 1 mg/L, respectively. NO 3 − -N, NH 4 + -N, and TP in the synthetic wastewater were prepared with KNO 3 , NH 4 Cl, and KH 2 PO 4 , however, the existence of Fe 2+ requires acidic conditions, and the simulated wastewater needed to be configured with tap water. Tap water contains a large number of chloride ions, which will quickly oxidize Fe 2+ to Fe 3+ . A metal chelating agent can use the strong binding effect of complex agent molecules and metal ions. The metal ions are encapsulated inside the chelating agent to become stable compounds with higher relative molecular weight, thus preventing the precipitation of metal ions and controlling the amount of free ferrous ions in the solution. Therefore, the oxidation of Fe 2+ can be controlled, to a certain extent, through the use of an appropriate complexing agent. Citric acid is one of the most widely distributed small molecule acids in nature. Liang et al. and Zhang et al. [17,18] found that the addition of citric acid can effectively control the amount of free ferrous ions in the system and ensure the continuous generation of free ferrous ions. The Fe 2+ solution was made by adding FeSO 4 ·7H 2 O and C 6 H 10 O 8 C 6 , according to the preliminary test; when the molar ratio of the Fe 2+ to citric acid was 1:7, Fe 2+ is oxidized the slowest. The experiments included two phases, in the first phase, it was a 45-day stabilization period, the bacterial communities were fed by 1/4 Hoagland nutrient solution, and were allowed to adapt nitrogen and phosphorus concentration in the wetland environment. The second phase (from 27 December 2020 to 4 February 2021) was divided 10 trials and each trial was 3 days. In the process of the experiment, firstly, 2 L of synthetic wastewater was added into the reactor in advance, and then 2 L of different concentrations of Fe 2+ solution were poured into each reactor, and the control group was only tap water.
To avoid causing a disturbance in the reactor, synthetic wastewater and Fe 2+ solutions were prepared in two water tanks (70 L). The concentration before the simulated wastewater and simulated iron-containing groundwater entered the reactor is shown in Table 1. Moreover, before each inflow event, CWs were flushed with synthetic wastewater three times. The composition of the synthetic wastewater used in the stabilization period was the same as for the second phase.

Sample Collection and Determination
The effluent was sampled at the upper outlet and was measured at 20 min and 1, 4, 8, 24, 48, and 72 h during the first four trials of the second phase. Water quality indicators were not measured in the last six trials. The experimental result is the average of the previous three trials. All of the samples were filtered with a 0.45 um membrane (Navigator, Shanghai, China). The NH 4 + -N, NO 3 − -N, TP, Fe 2+ , total iron, and Fe 3+ concentrations were determined using an Automatic Chemical Analyzer (Mode Smart Chem 200, Rome, Italy). At the end of the last trials, gravel substrate samples from the core of reactor were collected using a pre-sterilized 50 mL centrifuge tube and were frozen at −20 • C.
Species richness and diversity statistics, including Coverage, Chao1, Ace, Simpson, and Shannon ever, were also calculated using Mothur V1.43.0 (Department of Microbiology & Immunology at The University of Michigan, Annaburg, MI, USA). Finally, all effective bacterial sequences without primers were submitted for downstream analysis.
Biological classification is based on the characteristics of organisms in terms of their morphological structure and physiological functions. The basic unit of classification is the species. The higher the classification level, the less the organisms have in common; the lower the classification level, the more the organisms have in common. The classification system is a hierarchical system, usually consisting of seven main levels: kingdom, phylum, class, order, family, genus, and species. Species (species) are the basic units, closely related species are grouped into genus, closely related genera are grouped into families, families belong to orders, orders belong to classes, classes belong to doors, and doors belong to kingdoms. The higher the classification level, the less the organisms have in common; the lower the classification level, the more the organisms have in common.

Statistical Analysis
Statistical analyses were performed using the SPSS 19.0 (SPSS Inc., Chicago, IL, USA). Data were presented as means ± standard deviation (SD). Means between different treatments were compared by one-way analysis of variance (ANOVA) with Tukey HSD test at a significance level of 0.05. All figures were designed and plotted by Origin 8.5 (OriginLab Inc., Northampton, MA, USA).

Nutrients Removal Efficiency
The effluent concentrations of NH 4 + -N in CWs with different concentrations of Fe 2+ additions are shown in Figure 2a. With the extension of HRT, the NH 4 + -N concentrations in the effluent water sharply decreased for a 72 h-HRT in all reactors. The removal efficiency of NH 4 + -N in the CW-Fe10 reactor was the highest, which reached 88.66%, it increased by 11.91% compared with the CW-CK, followed by CW-Fe20, CW-Fe30, and CW-Fe5, respectively. CW-Fe5 had the lowest removal rate (62.54%
As shown in the Figure 2c, with the extension of HRT, the TP concentrations in effluent water sharply decreased in all of the reactors. It is worth noting that the effluent concentrations of CW-Fe5, CW-Fe10, and CW-Fe20 were significant (p < 0.05) lower than CW-CK after 4 h. There was no significant difference (p > 0.05) between CW-Fe30 and CW-CK, indicating that the addition of 30 m/L Fe 2+ had little effect on the removal of TP. The highest TP removal rate was observed in CW-Fe5 (88.68%), followed by CW-Fe10 (85.92%), CW-Fe20 (70.46%), CW-Fe30 (57.69%), and CW-CK (57.06%) at 72 h, and the effluent concentration of TP in CW-Fe5 and CW-Fe10 reached class III of the Surface Water Environmental Quality Standards of China (GB3838-2002) at 72 h, and their concentrations were 0.13 mg/L and 0.16 mg/L, respectively. The TP removal efficiency of the reactor was expressed as CW-Fe5 > CW-Fe10 > CW-Fe20 > CW-Fe30 > CW-CK after 4 h. It shown that the removal rate of TP decreased with increasing the concentration of Fe 2+ in the reactors. The optimal concentration of Fe 2+ was 5 mg/L. The removal rate of TP in CW-Fe5 reached 80.08% with 24 h, so, in a short time, adding low concentrations of Fe 2+ was beneficial to TP removal from the CWs.

Effluent Dissolved Fe 2+ , Total Iron and Fe 3+ , and pH Value Concentrations
Furthermore, the concentrations of Fe 2+ , total iron, and Fe 3+ were monitored (Figure 3

Microbial Richness and Diversity
Alpha diversity could reflect the abundance and diversity of microbial communities in locally homogeneous habitats, including a series of statistical analysis indexes, as shown in Table 2. OTUs were used to classify microbial species (e.g., species, genus, and phylum) artificially. The Chao1 richness index could estimate the number of microbial species. As a common indicator for assessing microbial diversity, the Shannon diversity index often reflects the species diversity and evenness of the community at the same time. The more evenly distributed the individuals, the higher the Shannon value. The higher the coverage rate, the lower the probability of undetected sequences. This index actually reflects whether the sequencing results represent the real situation of the samples. The Alfa diversity in all the reactors was well captured, with an estimated coverage of 0.99. The microbial coverage rate in the reactor without Fe 2+ addition was 1.00. The diversity indexes of OTUs of the Shannon index for CW-Fe5, CW-Fe10, CW-Fe20, and CW-Fe30 were significantly (p < 0.05) lower than that of CW-CK. Chao1 in CW-Fe10, CW-Fe20, and CW-Fe30 were significantly (p < 0.05) lower than that of CW-CK. The diversity indexes of OTUs and Shannon in CW-Fe5 and CW-Fe10 were significantly higher (p < 0.05) than those of CW-Fe20 and CW-Fe30. It suggested that the addition of Fe 2+ reduced the microbial diversity in the CWs. There was no significant difference of OTUs and Shannon between CW-Fe5 and CW-Fe10, and there was a significant difference (p < 0.05) of Shannon between CW-Fe20 and CW-Fe30. It can be inferred that when Fe 2+ was higher than 20 mg/L in CWs, it could significantly reduce the microbial diversity in the reactor, thus indicating that 20 mg/L is the critical concentration for improving the microorganisms in CWs.

Microbial Richness and Diversity
Alpha diversity could reflect the abundance and diversity of microbial communities in locally homogeneous habitats, including a series of statistical analysis indexes, as shown in Table 2. OTUs were used to classify microbial species (e.g., species, genus, and phylum) artificially. The Chao1 richness index could estimate the number of microbial species. As a common indicator for assessing microbial diversity, the Shannon diversity index often reflects the species diversity and evenness of the community at the same time. The more evenly distributed the individuals, the higher the Shannon value. The higher the coverage rate, the lower the probability of undetected sequences. This index actually reflects whether the sequencing results represent the real situation of the samples. The Alfa diversity in all the reactors was well captured, with an estimated coverage of 0.99. The microbial coverage rate in the reactor without Fe 2+ addition was 1.00. The diversity indexes of OTUs of the Shannon index for CW-Fe5, CW-Fe10, CW-Fe20, and CW-Fe30 were significantly (p < 0.05) lower than that of CW-CK. Chao1 in CW-Fe10, CW-Fe20, and CW-Fe30 were significantly (p < 0.05) lower than that of CW-CK. The diversity indexes of OTUs and Shannon in CW-Fe5 and CW-Fe10 were significantly higher (p < 0.05) than those of CW-Fe20 and CW-Fe30. It suggested that the addition of Fe 2+ reduced the microbial diversity in the CWs. There was no significant difference of OTUs and Shannon between CW-Fe5 and CW-Fe10, and there was a significant difference (p < 0.05) of Shannon between CW-Fe20 and CW-Fe30. It can be inferred that when Fe 2+ was higher than 20 mg/L in CWs, it could significantly reduce the microbial diversity in the reactor, thus indicating that 20 mg/L is the critical concentration for improving the microorganisms in

Bacterial Community Composition
The Figure 5 shown the microbial community composition and structure of each reactor substrates at the phylum (Figure 5a), family (Figure 5b), and genus (Figure 5c) levels. Phylum Proteobacteria accounted for the highest relative abundance at each sampling point, ranging from 40.89 to 66.57%. Compared with CW-CK, the addition of Fe 2+ significantly (p < 0.05) increased the relative abundance of phylum Firmicutes in the reactors. The proportions of phylum Bacteroidetes found in CW-CK, CW-Fe5, CW-Fe10, CW-Fe20, and CW-Fe30 were 6.23, 3.08, 4.01, 4.64, and 3.04%, respectively. The relative abundances of phylum Proteobacteria, phylum Firmicutes, and phylum Bacteroidetes were the highest in CW-Fe5, CW-Fe20, and CW-CK, respectively.

Influence of Fe 2+ Addition on Nitrogen Removal in CWs
NH4 + -N concentrations in the effluent water sharply decreased for a 72 h-HRT in all reactors with the extension of HRT, and the removal rate of NH4 + -N in CW-Fe10 was higher than for the other reactors. Some studies indicated that ammonia could be directly oxidized by O2 to NO2 − , namely through nitrification [12]. Nitrification was the main reason for the reduction of NH4 + -N concentrations in all reactors in this study, and it is a reason the NH4 + -N concentrations in effluent water sharply decreased for a 72 h-HRT in all of the reactors. As an active element, iron participates in the nitrogen biogeochemical cycle [19]. Although the iron cycle only involves the conversion of two main chemical valences, they have significantly different influences on the biogeochemical cycle of nitrogen. A previous study demonstrated that the increase of ammonia monooxygenase (AMO) enzyme and nitrite oxidoreductase (NXR) in an Fe-added sludge system was conducive to nitrification [20]. In addition, the biochemical processes of oxidizing NH4 + -N At the family level, the major genera across the substrate samples of were detected as family Carnobacteriaceae (0.28-45.85%), family Pseudomonadaceae (1.20-19.15%), and family Comamonadaceae (3.32-11.67%). The proportion of family Carnobacteriaceae was the highest in CW-Fe20 (45.85%) and the lowest CW-CK (0.28%). Compared with the CK, the addition of Fe 2+ significantly (p < 0.05) increased the relative abundance of family Carnobacteriaceae in the reactor. CW-Fe5 had the most family Pseudomonadaceae (19.15%), while it was difficult to identify in CW-Fe30 (1.20%). The relative abundance of family Comamonadaceae in CW-Fe5 (11.11%) and CW-Fe10 (11.66%) was significantly more than that in CW-CK (5.46%).

Influence of Fe 2+ Addition on Nitrogen Removal in CWs
NH 4 + -N concentrations in the effluent water sharply decreased for a 72 h-HRT in all reactors with the extension of HRT, and the removal rate of NH 4 + -N in CW-Fe10 was higher than for the other reactors. Some studies indicated that ammonia could be directly oxidized by O 2 to NO 2 − , namely through nitrification [12]. Nitrification was the main reason for the reduction of NH 4 + -N concentrations in all reactors in this study, and it is a reason the NH 4 + -N concentrations in effluent water sharply decreased for a 72 h-HRT in all of the reactors. As an active element, iron participates in the nitrogen biogeochemical cycle [19]. Although the iron cycle only involves the conversion of two main chemical valences, they have significantly different influences on the biogeochemical cycle of nitrogen. A previous study demonstrated that the increase of ammonia monooxygenase (AMO) enzyme and nitrite oxidoreductase (NXR) in an Fe-added sludge system was conducive to nitrification [20]. In addition, the biochemical processes of oxidizing NH 4 + -N with Fe 3+ (anaerobic ammonium oxidation coupled with ferric iron reduction (Feammox)) and reducing NO 3 − -N with Fe 2+ (nitrate-dependent anaerobic ferrous-oxidation (NDAFO)) in an anaerobic environment are well known and have been found to coexist with Anammox. Feammox may occur in experimental reactors with Fe 2+ addition. Studies have found that the process of Feammox has a better effect for sewage treatment in constructed wetlands, including a high removal rate in NH 4 + -N [12]; Scholars have also observed a positive correlation between dissolved iron, ammonium, and nitrite in three experimental wetlands [4]. Feammox needs a large amount of Fe (III) to improve nitrogen removal in wastewater treatment [12]. In this study, 53.62 to 78.81% Fe 2+ was oxidized to Fe 3+ in the CW-Fe20 and CW-Fe30 reactors after 24 h, resulting in higher Fe 3+ concentrations (18.90 to 18.84 mg/L) and a large consumption of oxygen (anaerobic condition) in the reactors. In addition, phylum Planctomycetes contain bacteria related to Anammox [21]. Phylum Planctomycetes not only contribute to anammox, but also play an important role in denitrification [22]. Filamentous bacteria were detected in all reactors with Fe 2+ addition, indicating that anammox may occur in the reactor, increasing the NH 4 + -N removal rate. It had a relatively higher concentration of Fe 3+ in the CW-Fe20 and CW-Fe30 reactors, and there was also relatively high Fe 2+ in the reactors. A high concentration of Fe 2+ was not conducive to the oxidation of iron and ammonia, and would change the activity of the microorganisms. Studies have shown that facultative Fe 2+ reducing bacteria may exist under both aerobic and hypoxic conditions in the transformation process of two iron forms. Fe 2+ oxidation would affect the subsequent Fe 3+ biological reduction by changing the bacterial activity. It is documented that microorganisms such as Gram-positive Bacillus subtilis variants, due to the production of reactive oxidants, under neutral conditions, air oxidation of Fe 2+ , and zerovalent iron could inactivate Aspergillus niger and Gram-negative Escherichia coli [23]. Therefore, Fe 2+ oxidation may reduce the coexisting Fe 3+ reducing bacteria, thus hindering the subsequent Fe 3+ biological reduction [23]. After 24 h, Fe 3+ was reduced to Fe 2+ , and Fe 2+ increased in I20-CW and I30-CW in the reactors, which is not conducive to the removal of NH 4 + -N, nitrification, and Feammox. In addition, the chemical oxidation process of Fe 2+ also consumed a large amount of O 2 , resulting in competition with nitrification of NH 4 + -N. Thus, it was difficult for microorganisms to oxidize NH 4 + -N, compared with chemical oxidation, which inhibited the NH 4 + -N removal [24]. In this study, the effluent removal efficiency of NH 4 + -N was 76.75% in CW-CK, and it was consistent with the CW-Fe30 reactor at 72 h. Some studies indicated that many processes, including nitrification, anammox, dissimilated nitrate reduction to ammonium (DNRA), substrate adsorption and desorption, denitrification, ammonia volatilization, and plant absorption, could affect NH 4 + -N removal in constructed wetlands [25]. In CW-CK, part of NH 4 + -N could be removed by nitrification, but Fe 2+ was oxidized to Fe 3+ in CW-Fe20 and CW-Fe30 and a large amount of oxygen was consumed, and nitrification was difficult to undertake. However, excessive Fe 2+ concentrations led to a decrease in DO in the CWs system, thus inhibiting the activity of anammox bacteria and other microorganisms [26]. So, Fe 2+ addition had less impact on the removal rate of NH 4 + -N, because more NH 4 + -N was generated during the denitrification process in the reactors. Some research has suggested that an increase in electron donors leads to a greater reduction of NO 3 − -N to N 2 and NH 4 + -N instead of NO 2 − -N [15,27,28]. Si et al. [15] found that adding iron could cause the accumulation of NH 4 + -N, because it by produced Fe 2+ , thus providing more electron donors. Compared with CW-CK, adding Fe 2+ could significantly reduce the concentration of NO 3 − -N in CWs (p < 0.05). This indicates that Fe 2+ promoted the denitrification process and participated in denitrification. This was consistent with previous relevant research conclusions [29]. Xu and Cai [30] added NO 3 − and Fe 2+ in subtropical soil for cultivation under anaerobic reduction conditions, and it was found that the NO 3 − concentration was significantly positively correlated with the Fe 2+ concentration, indicating that Fe 2+ participated in the electron transfer process of denitrification. Zhang et al. [31] indicated that the removal rate of NO 3 − -N was 45.63 ± 3.31% in anaerobic digestion sludge with a paper inoculation source and Fe 2+ .
There are two mechanisms through which Fe 2+ participated in the denitrification of CWs. On the one hand, Fe 2+ donated electrons directly to nitrate [11]. However, under general liquid phase conditions, chemical denitrification did not occur easily or was very slow [32]. It promoted the solid interface or some solid materials, such as metal ions and metal oxide, to exist in chemical denitrification progress [33]. On the other hand, Fe 2+ and NO 3 − -N depend on certain microorganisms for the reduction metabolism, which could reduce NO 3 − while oxidizing Fe 2+ . In most cases, Fe 2+ participating in the denitrification of NO 3 − -N was dominated by biological mechanisms. In addition, CWs was a comprehensive ecological treatment system with substrates and rich microorganisms, providing suitable conditions for chemical and biological denitrification [34]. With 20 min of influent, there was mainly chemical denitrification, and after 20 min, biological denitrification was dominant. If there was not sufficient organic carbon as an electron donor for heterotrophic denitrification in nitrate-dependent iron (II) oxidation, microorganisms could spontaneously utilize Fe 2+ as an inorganic electron donor. Because of low organic carbon in reactors, the microorganisms would improve NO 3 − -N removal efficiencies through cooperation with Fe 2+ . The specific process is when Fe 2+ is oxidized to Fe 3+ , NO 3 − -N is used as an electron donor, Fe 2+ is oxidized to Fe 3+ , and NO 3 − -N is reduced to N 2 . In addition, the removal rate of NO 3 − -N was faster after 24 h in this study. The main reason was that the denitrification time increased with the prolongation of HRT, and there were more denitrifying bacteria after 24 h, and most nitrate-dependent iron (II) oxidizing strains relied on organic cosubstrates, such as acetate [35]. In this study, some small molecular weight organic matter citric acid was added to maintain nitrate-dependent iron (II) oxidizing strains. Therefore, with the extension of HRT, the removal efficiency of NO 3 − -N increased. Third, the high concentration of Fe 2+ continuously provided electron donors for biological denitrification. After 24 h of influent, the Fe 3+ in the reactor was reduced to Fe 2+ . In these redox kinetic systems, Fe 2+ could be oxidized by O 2 to Fe 3+ when O 2 was depleted, the Fe 3+ could be reduced to Fe 2+ through the coexisting Fe 3+ reducing bacteria. These redox transitions formed the Fe 2+ /Fe 3+ cycles, continually providing electron donors for denitrification. This study also found that the removal rate of NO 3 − -N in CW-Fe20 and CW-Fe30 reactors was lower than that of CW-Fe10, which suggested that when the influent Fe 2+ concentration was 10 mg/L, the removal rate of NO 3 − -N was the highest. The main reason may be the high concentrations of iron affected the activity of the microorganisms. Some studies have shown that iron could cause oxidative damage to microorganisms, thus leading to cell death [15].

Influence of Fe 2+ Addition on Phosphorus Removal Performance in CWs
In this study, adding Fe 2+ could increase the removal efficiency of TP. This was consistent with the conclusions of other research [25], and the concentration of Fe 2+ was significantly negatively related with phosphorus concentrations in constructed wetland water. Iron enhanced the removal of phosphorus through abiotic and biological processes [36]. Similar research has shown that orthophosphate existed in domestic sewage and was removed by chemical precipitation and biological metabolism [20]. A low temperature has little effect on the removal of phosphorus, indicating that the non-biological process was the main way to remove phosphorus. From a non-biological point of view, the main reason was that iron promotes phosphorus removal through the following three mechanisms: coexistence and precipitation of Fe 2+ or Fe 3+ , adsorption of solid iron oxide/hydroxide, and flocculation and precipitation of Fe 2+ [25]. Firstly, when Fe 2+ was added to the reactor, in a short time, the Fe 3+ from the oxidation of Fe 2+ reacted with the phosphate to form FeOOH [37] and Fe 3 (PO 4 ) 2 with a large specific surface area. Due to its low solubility in water, it was immediately precipitated as a solid [20]). Secondly, iron oxyhydroxide could be produced in reactors with Fe 2+ addition, and the production of iron oxyhydroxide promoted the adsorption of phosphorus (iron oxyhydroxide−phosphate) [38]. The huge specific surface area of iron oxyhydroxide improved the adsorption capacity of iron filings. The adsorption of phosphorus by the substrate and plant roots are important ways to remove TP [14]. Both substrate and plant root absorption were found in the early stage of the experiments, and played an important role in TP removal [39]. Vymazal [40] found that most of the iron in the matrix was in the form of solid iron oxide, while ions were limited. Therefore, adsorption become the main method for phosphorus removal [41]. In this study, the phosphate was adsorbed by the matrix, which explains why the phosphate in CW-CK also had a certain removal rate. Finally, the phosphate could be removed by the formation of ferrous phosphate, deposited on the iron surface layer and not present in the precipitation agent [42]. The removal efficiency of TP decreased with the extension of HRT in this study. The results were consistent with the research of [14]. The main reason for this is that the gradual thickening of the biofilm on the substrate could hinder the effective contact and mass transfer between the pollutants and the substrate, resulting in a reduction of the removal efficiency of TP [43]. In addition, large Fe 2+ could quickly consume oxygen in the wastewater, causing phosphorus-accumulating bacteria to release phosphorus under anaerobic conditions [44]. Meanwhile, due to low oxygen levels, Fe 3+ was easily reduced to Fe 2+ , releasing complex phosphorus, thereby lowering the efficiency of phosphorus removal [45]. High concentrations of Fe 2+ inhibited the removal efficiencies of TP, which may be attributed to the three main reasons. One was that excessive Fe 3+ formed a large amount of iron oxide precipitation, covering the surface of the substrate, reducing the adsorption of phosphorus by the substrate. The other was that the addition of high concentration of Fe 2+ could quickly result in an anaerobic environment in the reactors, and phosphorus accumulating bacteria release phosphorus [44]. The third is that, under anaerobic conditions, Fe 3+ is easily reduced to Fe 2+ , which releases complex phosphorus, thereby reducing the efficiency of phosphorus removal [45]. Therefore, the higher concentration of Fe 2+ inhibits the removal of TP.

Influence of Fe 2+ Addition on the Features of Bacterial Communities
In this study, the iron addition reduced the microbial diversity in the reactors. This was consistent with the results of previous studies [46], which indicated that more Fe 2+ led to a decrease in the abundance of microorganisms and an increase in the abundance of tolerant species. Moreover, the addition of Fe 2+ could reduce the types of other bacteria and increase the relative abundance of specific dominant bacteria, thus enhancing nitrogen and phosphorus removal. Figure 5 shows the bacterial community structure of all five reactors at the phylum (Figure 5a), family (Figure 5b), and genus (Figure 5c) levels. Phylum Proteobacteria includes bacteria responsible for nitrification and denitrification activities and various metabolic bacteria. The circus diagram (Figure 5a) shows that phylum Pro-teobacteria accounted for the highest relative abundance at each sample, ranging from 40.89% to 66.57%. The existence of phylum Proteobacteria played an important role in the degradation of organic matter, nitrification and denitrification, and phosphorus removal in wastewater [47]. The Alpha-proteobacteria, Beta-proteobacteria, and Gamma-proteobacteria in the phylum Proteobacteria were also very important for sewage treatment. Phylum Proteobacteria was usually the largest phylum found in constructed wetland systems [48]. Iron-oxidizing microorganisms (Fe (II)-oxidizing microorganisms, FeOM) could oxidize Fe 2+ to Fe 3+ under aerobic or anoxic/anaerobic conditions, and thus synthesize organic matter [49]. From the perspective of the phylum, FeOM mainly included phylum Nitrospira, phylum Firmicute, phylum Chlorobi, and phylum Proteobacteria [49]. The related bacteria phyla were found, indicating that FeOM existed in the reactor. Previous research reported that nitrate-dependent iron oxidizing bacteria (NFeOB) were classified as autotrophic or heterotrophic microorganisms, and most of the isolated organisms belonged to phylum Proteobacteria and phylum Actinobacteria [50]. Phylum Actinobacteria played an indispensable role in the nitrogen cycle [47]. In this study, the abundance of phylum Actinobacteria and phylum Proteobacteria in CW-CK was higher, indicating that denitrification was the main pathway for the decrease of NO 3 − -N concentrations in CW-CK. Phylum Firmicutes were the dominant bacteria in most constructed wetlands. The proportion of phylum Bacteroidetes increased with the outbreak of cyanobacteria blooms, and gradually became the dominant bacteria in eutrophic waters [14]. Phylum Bacteroidetes were also related to denitrification, and were detected in all reactors. The abundance of phylum Bacteroidetes in CW-CK was higher than that in reactors with Fe 2+ addition, and it could be deduced that t Fe 2+ was not the key factor for the growth of phylum Bacteroidetes. This was consistent with the results of previous studies, where the relative abundance of phylum Bacteroidetes was inhibited by pyrite and ferrous sulfide substrates, whether in aeration or non-aeration areas [51]. The above studies show that adding Fe 2+ resulted in less autotrophic denitrification, and heterotrophic denitrification or iron-dependent denitrification became dominant. Figure 5b shows the bacterial community structure composition of the overall five CWs at the family level. The relative abundance of heterotrophic denitrifying bacteria (such as family Comamonadaceae and family Planctomycetaceae) were found in the reactors, and were higher in CW-Fe5 and CW-Fe10, closely related to the removal of nitrogen elements [52]. Some of the microorganisms were anammox bacteria, and it is suggested that, besides the conventional nitrogen element removal pathways, the anammox process may play an important role in the nitrogen transformation process of CWs [53]. Certain bacterial genera in the Comamonadaceae family (such as Hydrogenophaga) of the reactors have been proven to be a type of phosphorus accumulating bacteria, which could absorb the dissolved phosphorus in the water and convert it into polyphosphate in the body for energy accumulation. This is considered to be the main contributor to phosphorus removal [54]. A large number of studies have shown that some of the microorganisms in the Comamonas family (family Comamonadaceae) belong to denitrifying bacteria, and the relative abundance of such microorganisms is closely related to NO 3 − -N removal [55]. The higher NO 3 − -N removal efficiency in the CW-Fe5 and CW-Fe10 may be related to the relative abundance of the family Comamonadaceae and family Planctomycetaceae. Furthermore, a large number of phosphorus-accumulating bacteria and denitrifying phosphorus-accumulating bacteria also exist in the family Comamonadaceae [54]. It is beneficial to the removal of TP in CW-Fe5 and CW-Fe10 reactors. Family Planctomycetes not only had the function of anammox, but also contributed to denitrification [22]. Family Planctomycetes were detected in four CWs with Fe 2+ addition, indicating that anammox may occur in the reactors. Studies have shown that autotrophic denitrification is the main reaction to remove nitrogen from pyrite and ferrous sulfide substrates, which is not conducive to the growth of the family Planctomycetaceae [14]. This also revealed that the relative abundance of family Planctomycetes in CW-CK was higher than for the other reactors in this study.
The structure and composition of the bacterial community in CWs at the genus level is shown in Figure 5c. The genus Trichococcus, within the family Carnobacteriaceae, the order Lactobacillales, affiliated to phylum Firmicutes, accounted for 13.55-45.85% of bacterium in the reactors with Fe 2+ addition, while it only accounted for 0.28% of that in CW-CK. More specifically, genus Trichococcus was more abundant in the reactors with Fe 2+ addition, and it was particularly enriched in CW-Fe20, so it is believed that adding Fe 2+ was beneficial to genus Trichococcus growth. Tang et al. [56] suggested that genus Trichococcus belong to Fe (III)-reducing bacteria, and Baek et al. [57] indicated that genus Trichococcus was a kind of electroactive iron-reducing bacterium. They could reduce the oxidized Fe 3+ to Fe 2+ and continue to providing electrons for nitrate, further suggesting that genus Trichococcus may participate in nitrogen biodegradation. In addition, genus Trichococcus is a common acid-producing bacteria, mainly utilizing glucose to produce lactic acid, formic acid, acetic acid, and CO 2 [58]. The enrichment of the genus Trichococcus was conducive to organic matter degradation, contributing to the utilization of carbon sources by heterotrophic denitrifiers [59]. At the same time, previous studies have shown that the genus Trichococcus in the constructed wetland with iron addition could achieve more efficient denitrification. Genus Pseudomonas is a genus of denitrifying bacteria and a dominant group of nitrifying bacteria in constructed wetlands. It played an important role in heterotrophic denitrification [14]. The Geobacter and Pseusomonas genera, derived from iron reducing bacteria, were important for NH 4 + -N oxidation and contributed to nitrogen removal [60]. Pseudomonas was beneficial for removing COD cr and TN, and was the dominant genus for removing TP [9], which might be attributed to Pseudomonas as the weak Fe (III) reducers [60]. The Pseudomonas of CW-Fe5, CW-Fe10, CW-Fe20, and CW-Fe30 were 18.92%, 8.65%, 4.76%, and 1.15%, respectively, suggesting that it was an important factor in the removal of nitrogen and phosphorus.

Conclusions
By adding different concentrations of Fe 2+ to the CWs, whether Fe 2+ addition has a positive effect on nutrient removal was explored. The main conclusions are as follows: adding 10 mg/L Fe 2+ has the best nitrogen removal rate, and adding 5 mg/L Fe 2+ has the highest TP removal rate. The above low-concentrations Fe 2+ are within the range of the Fe 2+ concentrations in the groundwater in the Sanjiang Plain. Considering the field situation, a portion of groundwater can be pumped into wetlands in the Sanjiang Plain to reduce the nutrient concentrations in CWs. Compared with CW-CK, adding different concentrations of Fe 2+ significantly (p < 0.05) reduces the microbial diversity in the reactor and increases the relative abundance of the genus Trichococcus, within the family Carnobacteriaceae, the order Lactobacillales, affiliated to phylum Firmicutes, which belongs to Fe (III)-reducing bacteria, and can reduce the oxidized Fe 3+ to Fe 2+ and continue to provide electrons for nitrate and achieve more efficient denitrification with a small number of denitrifying bacteria after Fe 2+ addition.

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