A Hardy Plant Facilitates Nitrogen Removal via Microbial Communities in Subsurface Flow Constructed Wetlands in Winter

The plants effect in subsurface flow constructed wetlands (SSF-CWs) is controversial, especially at low temperatures. Consequently, several SSF-CWs planted with Iris pseudacorus (CWI) or Typha orientalis Presl. (CWT) and several unplanted ones (CWC) were set up and fed with secondary effluent of sewage treatment plant during the winter in Eastern China. The 16S rDNA Illumina Miseq sequencing analysis indicated the positive effects of I. pseudacorus on the bacterial community richness and diversity in the substrate. Moreover, the community compositions of the bacteria involved with denitrification presented a significant difference in the three systems. Additionally, higher relative abundances of nitrifying bacteria (0.4140%, 0.2402% and 0.4318% for Nitrosomonas, Nitrosospira and Nitrospira, respectively) were recorded in CWI compared with CWT (0.2074%, 0.0648% and 0.0181%, respectively) and CWC (0.3013%, 0.1107% and 0.1185%, respectively). Meanwhile, the average removal rates of NH4+-N and TN in CWI showed a prominent advantage compared to CWC, but no distinct advantage was found in CWT. The hardy plant I. pseudacorus, which still had active root oxygen release in cold temperatures, positively affected the abundance of nitrifying bacteria in the substrate, and accordingly was supposed to contribute to a comparatively high nitrogen removal efficiency of the system during the winter.

Scientific RepoRts | 6:33600 | DOI: 10.1038/srep33600 wetlands 6,7,12,13 . On the one hand, plant roots can provide a substratum for attached microorganism growth 1,8 . On the other hand, the plant litter can serve as carbon source for heterotrophic bacteria 6,13,14 . In addition, the emerging plants growing in SSF-CWs can also work as oxygen transfer passage for aerobic microorganism in the substrate 6,15,16 . However, several studies posited that plants had little statistical effect on the community abundance or structure of the overall bacteria or particular microbial functional groups [17][18][19][20] . In fact, the interaction between the macrophyte and microorganism community depends on a variety of factors such as plant species, vegetation growth status, season and even random chance 17,[20][21][22] . Therefore, the effects of different plants on microorganism groups in CWs, especially in the cold season, need deeper exploration. Further study on the complicated process occurring in CWs would be beneficial to optimizing design parameters and improving the overall performance in engineering application practices.
In the present research, yellow flag (Iris pseudacorus) and oriental cattail (Typha orientalis Presl.) were chosen. The two different species have different physiological properties and growth characteristics; for example, the yellow flag can remain active, but the acrial part of the oriental cattail will wither away in the winter. Furthermore, to investigate the microorganism community in the substrate, the V3-V4 regions of the bacteria 16S rDNA were sequenced via an Illumina MiSeq 2500 platform. Because of its higher integrity and broader range of applications 23,24 , 16S rDNA Illumina Miseq sequencing has recently been used frequently to study the microbial diversity in various environments [25][26][27][28] . Recently, this technology is also used for the analysis of microorganism communities in CWs [29][30][31][32][33][34] .

Results
Overall performance. The influent characteristics and the time for water sample determination were described in the second section (Experimental Design and Operation). The overall performance of the two types of SSF-CWs is shown in Fig. 1. CW I had the highest removal rates of NH 4 + -N (80.50%), TN (47.50%) and COD (81.07%) and was superior to the control, which had removal rates of 72.25% (NH 4 + -N), 39.33% (TN) and 73.57% (COD), respectively. For CW T , only NH 4 + -N removal (75.75%) presented a slight advantage compared to the control, while the removal rates of TN (34.17%) and COD (71.43%) were the lowest among the three systems. The removal efficiency of NO 3 − -N in the three systems was consistent at approximately 33.00%. NO 2 − -N was not detected in the effluent.
Growth and physiological characteristics of the plants. Plant    Microbial community composition. In addition to the richness and diversity of the microbial communities, the community composition in the substrate also plays a critical role in the pollutants removal in CWs. The abundance of different phyla of bacteria in the three types of SSF-CWs is shown in Fig. 2. A total of 24 identifiable phyla, in which Proteobacteria and Bacteroidetes were the two dominant species, were detected in the three systems. In CW I , Proteobacteria and Bacteroidetes accounted for 51.16% and 33.91%, respectively, followed by Cyanobacteria (6.05%) and Verrucomicrobia (4.23%). In CW T , Proteobacteria and Bacteroidetes accounted for 72.73% and 22.59%, respectively. In the control, Proteobacteria (64.29%) and Bacteroidetes (27.26%) constituted the primary phyla, and Cyanobacteria (3.39%) and Verrucomicrobia (2.55%) followed.
The primary genera (relative abundance > 0.50%) with the addition of three nitrifying bacteria (Nitrosomonas, Nitrosospira and Nitrospira) in the three systems are shown in Table 3. A total of 40 genera were listed, and they mainly belonged to two classes, β-proteobacteria (17 genera) and γ-proteobacteria (10 genera). The bacterial community composition showed a significant difference at the genera level among the three systems. In CW I , Flavobacterium (18.28%), Albidiferax (10.02%) and Ohtaekwangia (7.72%) constituted the three dominant genera.

Discussions
It was widely acknowledged that the transformation and removal of nitrogen in CWs were mainly attributed to nitrification and denitrification, followed by vegetation assimilation, substrate adsorption and the other pathways 3,4,6-8, 35 . Because the sorption of ammonia by gravel or sand was not the primary means for ammonia removal 22 , only the nitrogen contents in the water and plants were determined in this research. Our results showed that the nitrogen accumulation in the vegetation was very limited during the variables determination stage (Stage III). This result was related to the weak growth of I. pseudacorus and the dormant aboveground part of T. orientalis during the cold season. Before the variable determination, the current systems had operated normally for 6 months, which was considered a long enough time for the microbial community in the substrate to reach a steady state 13,36,37 . Compared to traditional molecular methods, 16S rDNA Illumina Miseq sequencing is a more effective approach for studying microbial diversity due to its higher throughput and integrity as well as its broader application range 23,24 . PCR-DGGE is often restricted by an insufficient resolution to characterize the microbial communities in the complex samples 13 , while 454-pyrosequencing has a low throughput and high error rate and is costly 24 . In the present study, approximately ten thousand OTUs were identified in each sample, which were clustered with 97% similarity. This result was 4-5 times higher than that in several previous studies on this issue 22,38 .  Some classes (relative abundances < 0.15%, except for Nitrospira) are grouped into "Others". the three types of CWs in the present research, a detailed comparison was carried out at different taxonomic levels. Firstly, the results indicated that Proteobacteria was the dominant phyla followed by Bacteroidetes in the three systems. This finding was consistent with a plenty of studies that reported Proteobacteria as the dominant phylum in various microorganism samples from the substrate or rhizosphere in wetlands 22,38,40,41 . Proteobacteria, which displayed a remarkably high level of bacterial metabolic diversity involved in global carbon, nitrogen and sulfur cycling, played an important role in pollutants removal 40,42 . Further analysis at the class level showed an obvious difference in bacterial community compositions among the three systems. The bacterial community mainly consisted of β-proteobacteria, Flavobacteria and γ-proteobacteria in CW I , while it was highly enriched with β-proteobacteria in CW T and the dominant class was γ-proteobacteria in CW C . Due to the amount of different influential factors, such as wetland types, substrate characteristics, wastewater components, operational parameters, environmental conditions and the different estimations by different approaches used in previous CW studies, the dominant class in CWs has been a long debated subject 13 Table 3. The main genera (relative abundance >0.50%) with the addition of three nitrifying bacteria (Nitrosomonas, Nitrosospira and Nitrospira) in CW I , CW T and CW C . Primary genera (relative abundance >2.00%) in each sample were bolded. different plants with different growth characteristics and physiological properties had a considerable effect on the bacterial community compositions in CWs. According to the list for genera in which at least one member had been characterized as a denitrifying strain reported by Heylen, et al. 45 and Philippot, et al. 46 , it was found that a considerable proportion of the bacterial genera detected in the present research had a close relationship to denitrification. The compositions of denitrifying bacteria presented a remarkable difference among the three systems. The system grown with a hardy plant showed superior evenness of denitrifying bacteria, while certain genus such as Flavobacterium or Halomonas presented a fairly high abundance in the freezing-sensitive plant system and the unplanted system. However, it seemed that all the three types of SSF-CWs had adequate denitrification capability.
The difference in the overall efficiency of nitrogen removal among the three systems might result from the nitrification process. The bacteria involved in nitrification mainly consisted of two groups, the ammonium-oxidizing bacteria (AOB), which converted ammonium to nitrite, and the nitrite-oxidizing bacteria (NOB), which converted nitrite to nitrate 11 . A higher ratio of NOB to AOB populations was considered indicative of a higher nitrification capacity and more complete ammonia oxidation 47 . In this research, higher relative abundance of nitrifying bacteria and higher ratio of NOB to AOB populations accompanied by higher removal efficiency of NH 4 + -N were recorded in CW I compared to CW T and CW C . This phenomenon could be attributed to more oxygen being released from the hardy plant I. pseudacorus 6,15 . Zhong,et al. 11 also reported that enhanced levels of oxygen and nitrite favoured the growth of NOB in CWs. In the present experiment, Nitrosomonas and Nitrosospira were the two dominant AOB lineages. This finding was consistent with a previous study in a vertical flow CW by Tietz, et al. 48 . This study also found that Nitrosomonas in the three systems was more abundant than Nitrosospira. The result might be related to the high influent NH 4 + -N concentration, which contributed to the formation of a community dominated by Nitrosomonas, because Nitrosomonas had a lower substrate affinity but a higher maximum activity than Nitrosospira 49,50 . Moreover, Nitrospira was the only NOB detected in the present systems. These results suggested that the three nitrifying bacteria, Nitrosomonas, Nitrosospira and Nitrospira, shared the most responsibility for nitrification.
The diversity of the microbial composition suggested a diversity of the functional characteristics that were relevant to pollutant removal in wastewater 51 . Any difference in microbial functional groups might lead directly to the difference in the purification performance of the CWs 11 . Alpha diversity analysis suggested that the presence of a hardy plant induced higher diversity and evenness of bacterial community composition at different taxonomic levels ranging from class to genus. This might be related to the multiple influencing mechanisms of plants on microorganisms such as complicating the attachment surface, providing organic carbon, as well as changing the oxidation-reduction conditions in the wetlands 1,6,15 . The results revealed by the ACE estimator and Shannon index indicated that the system grown with hardy plant had higher microbial richness. This finding was consistent with several earlier studies 22,38,52,53 . The oxygen released from plant roots could affect the microorganism richness in the substrate 39 . Previous studies have shown that the bacterial communities were grouped according to an oxygen concentration gradient 40,54,55 . In the current study, the two plants grown in the CWs have different physiological properties and growth characteristics. During the experiment period, I. pseudacorus remained active, but the acrial part of T. orientalis withered away. The rate of root radial oxygen loss (ROL) was also higher in CW I compared to CW T . This might contribute to a higher richness of aerobic microorganism in CW I . Moreover, the carbon source supply is also an important factor that influences the microbial communities in the substrate 6,8 . The senescent or dead plant could release a variety of labile organic compounds such as amino acids, sugars, volatile fatty acids, etc. 56 . In this research, T. orientalis, a dormant plant during the winter, showed a decline in shoot biomass because of the senescence of leaves during the determination period. Chen et al. 38,57,58 suggested that the enhanced litter decomposition might contribute to increased microbial richness in the substrate of the CW. However, in this study, the effect of plant litter on denitrification efficiency enhancement was limited in the T. orientalis system. This could be because quite a large proportion of plant litter, such as highly crystalline lignocellulose, was recalcitrant and the complete degradation of litter needed a relatively long time 7,59 . Therefore, the present results suggested a positive effect of the hardy plant on the nitrogen removal in SSF-CWs during the winter, while the freezing-sensitive plant showed slight differences on the pollutants removal compared to the control.
In addition, some other factors may also have significant impact on the contaminants removal in CWs. Plant root exudates can also be used as the dissolved organic matter for the metabolism of heterotrophic bacteria 13 . In the current research, due to the limitations of experimental conditions, the root vitality was measured as a proxy indicator. The results seem to indicate that the presence of I. pseudacorus have more active effects towards the microbial community compositions in the substrate. Moreover, the root activity and growth status of the macrophytes in CWs also influence the richness and enzyme activity of microbial populations 10,52,60 . Wang, et al. 61 reported that plant ROL played an important role in nutrient removal by significantly affecting microbial abundance and activity in CWs. Additionally, anammox was also recognized as another approach for nitrogen removal apart from nitrification-denitrification 62 . However, the anammox bacteria, placed in the order Brocadiales in the phylum Planctomycetes, has not been observed in this study. Besides, the coexistence of some autotrophic bacteria such as Cyanobacteria in wetlands was also reported benefit to the pollutants degradation, because they could provide the heterotrophic bacteria with oxygen, which played a similar and complementary role in addition to the macrophytes to a certain degree 22,63 . Furthermore, some archaea with specific activity of ammonia oxidation also play an indispensable role in nitrogen removal processes 64,65 . All these factors should be taken into consideration in further researches.
In conclusion, 16S rDNA Illumina Miseq sequencing revealed detailed information about the interaction between the plant and the bacterial community composition in SSF-CWs. The hardy plant I. pseudacorus had a positive effect on the bacterial abundance and eventually on the community composition in the substrate. The presence of a hardy plant with active root oxygen release enhanced the relative abundance of nitrifying bacteria in the substrate, and consequently was supposed to contribute to the increase in the nitrogen removal efficiency of the system during the winter. However, some other factors, such as the bacteria population and activity, the presence of Cyanobacteria and ammonia-oxidizing archaea may also have an important influence on the nitrogen removal in SSF-CWs. Hence, further researches will be needed to understand detailed mechanisms referring to the biodegradation process of pollutants in CWs. Plant Sampling and Analysis. Plant samples were harvested and separated into roots and shoots, dried at 65 °C to a constant weight, and then ground into powder. The N content was determined by an Elemental analyzer (CHN-O-Rapid, Heraeus, Germany) at the beginning and end of the experiment 67 . The rate of root radial oxygen loss (ROL) was measured using the titanium (III) citrate buffer method 68,69 . The root vitality was quantified via the triphenyl tetrazolium chloride (TTC) method 70 . Microbial Sampling and Analysis. Preparation of microbial samples. As shown in Fig. S1, microorganism sampling points of each system was divided into three layers according to the substrate. Each layer comprised four sampling points. The four points formed a circle with a radius of 0.4 m around the system center, and the connections between each point and the system center formed a cross. 100 g of sand from the top layer and 100 g of gravel from the middle layer were obtained by a cylindrical sampler with 2.5 cm diameter, while 200 g of gravel were obtained from the bottom layer by a sampling scoop on Feb 2 nd 2015. Then, the samples in each system were mixed well and vigorously shaken at 200 rpm for 3 h in sterile glass bottles to isolate the biofilm from the substrate surface. After centrifuging twice (6,000 × g, 15 min), the precipitate was collected for subsequent DNA extraction, 16S rDNA PCR amplification and Illumina MiSeq sequencing.

Methods
DNA extraction and PCR amplification. DNA was extracted from the samples using a QIAamp Fast DNA Stool Mini Kit (QIAGEN, Chatsworth, CA, USA) according to the manufacturer's instructions. DNA yields were determined using a SpectraMax 190 (Molecular Devices, California, USA), and the integrity was evaluated via 1.0% agarose gel electrophoresis. Then, DNA was diluted to 1 ng·μ L −1 in sterile water. The universal primer sets 341F (5′ -CCTAYGGGRBGCASCAG-3′ ) and 785R (5′ -GACTACHVGGGTATCTAATCC-3′ ) were used for amplification of the V3-V4 regions of 16S rDNA. A 10 ng template, 0.5 μ L of forward primers and 0.5 μ L of reverse primers were added into a 25 μ L reaction system for the PCR reaction. Thermal cycling consisted of denaturation at 94 °C for 3 min, which was followed by 30 cycles of 94 °C for 10 s, 55 °C for 15 s and 72 °C for 30 s, and finally held at 72 °C for 7 min. AmpureBeads (Beckman Coulter, Inc., CA, USA) were used for the PCR product purification.
16S rDNA Illumina MiSeq sequencing. The sequencing libraries were constructed with an NEB Next Ultra DNA Library Prep Kit for Illumina (New England Biolabs Inc., Boston, MA, USA), and then a Qubit 2.0 Fluorometer (Life Invitrogen, Inc., Carlsbad, CA, USA) was used to assess the libraries quality. Then, the libraries were sequenced on an Illumina MiSeq 2500 platform.
Quality filtering, OTUs picking, annotation and diversity analysis. The FASTX-Toolkit (version 0.0.14) and Mothur program (version 1.34.0) were used for the analysis, merging and quality filtering of the raw data sequences. All of the reads were quality filtered using an average quality value of 20 (Q20) during demultiplexing. Short reads (length < 40 bp) and chimeras were excluded. Reads were clustered by degree similarity levels with the UCLUST program (version 1.2.22q). Sequences with ≥ 97% similarity were assigned to the same genus. A RDP classifier (version 2.2) was used to annotate the taxonomic information. The Mothur program was used for alpha diversity analysis. ACE estimated the richness of the species, while the Shannon index estimated the diversity of the species.
Scientific RepoRts | 6:33600 | DOI: 10.1038/srep33600 Statistical Analysis. All statistical tests were performed with the statistical program SPSS 17.0 (SPSS Inc. Chicago, USA). The data were analysed with one-way analysis of variance to compare the performance of each mesocosm. Duncan's test was performed to detect the statistical significance of differences (p > 0.05) between the mean values of the treatments.