Sulfide-Induced Dissimilatory Nitrate Reduction to Ammonium Supports Anaerobic Ammonium Oxidation (Anammox) in an Open-Water Unit Process Wetland

ABSTRACT

Open-water unit process wetlands host a benthic diatomaceous and bacterial assemblage capable of nitrate removal from treated municipal wastewater with unexpected contributions from anammox processes. In exploring mechanistic drivers of anammox, 16S rRNA gene sequencing profiles of the biomat revealed significant microbial community shifts along the flow path and with depth. Notably, there was an increasing abundance of sulfate reducers (Desulfococcus and other Deltaproteobacteria) and anammox microorganisms (Brocadiaceae) with depth. Pore water profiles demonstrated that nitrate and sulfate concentrations exhibited a commensurate decrease with biomat depth accompanied by the accumulation of ammonium. Quantitative PCR targeting the anammox hydrazine synthase gene, hzsA, revealed a 3-fold increase in abundance with biomat depth as well as a 2-fold increase in the sulfate reductase gene, dsrA. These microbial and geochemical trends were most pronounced in proximity to the influent region of the wetland where the biomat was thickest and influent nitrate concentrations were highest. While direct genetic queries for dissimilatory nitrate reduction to ammonium (DNRA) microorganisms proved unsuccessful, an increasing depth-dependent dominance of Gammaproteobacteria and diatoms that have previously been functionally linked to DNRA was observed. To further explore this potential, a series of microcosms containing field-derived biomat material confirmed the ability of the community to produce sulfide and reduce nitrate; however, significant ammonium production was observed only in the presence of hydrogen sulfide. Collectively, these results suggest that biogenic sulfide induces DNRA, which in turn can explain the requisite coproduction of ammonium and nitrite from nitrified effluent necessary to sustain the anammox community.

IMPORTANCE This study aims to increase understanding of why and how anammox is occurring in an engineered wetland with limited exogenous contributions of ammonium and nitrite. In doing so, the study has implications for how geochemical parameters could potentially be leveraged to impact nutrient cycling and attenuation during the operation of treatment wetlands. The work also contributes to ongoing discussions about biogeochemical signatures surrounding anammox processes and enhances our understanding of the contributions of anammox processes in freshwater environments.

INTRODUCTION

Constructed freshwater wetlands that receive wastewater effluent offer an opportunity to gain insight into the biological nutrient cycling process because they are subject to comparatively high nutrient loading rates and are optimized for steady-state operational treatment. These wetlands have historically been utilized as a comparatively low energy alternative or complement to more traditional wastewater treatment systems in order to mitigate nutrient pollution with the assumption that denitrification is the dominant nitrogen removal process (1). However, recent findings suggest that anammox, the oxidation of ammonium with nitrite to produce N₂, may play a more prominent role in these and other wetlands than initially thought (2). More broadly, increased understanding of anammox processes and drivers has led to wastewater treatment innovations (3) and a deeper understanding of nutrient cycling as it has been estimated that in the marine environment anammox accounts for 50% of marine N₂ production (4–6). In some freshwater systems anammox has been estimated to account for up to 35% of nitrate removal (7–9), and while there is evidence for anammox occurrence in natural treatment systems such as vegetated surface flow wetlands and soil infiltration systems (10, 11), the environmental parameters supporting anammox, the overall importance of the process to nitrate removal, and potential natural treatment applications are uncertain.

Dissimilatory nitrate reduction to ammonium (DNRA), a process that converts nitrate to nitrite and subsequently to ammonium, is gaining recognition as a significant component of nitrogen cycling in terrestrial and freshwater environments (12). In coastal environments it has been estimated that DNRA is responsible for 30% of nitrate reduction (13), with sulfide suggested as a controlling factor (14). Sulfide and organic carbon have profound effects on the nitrogen cycle, especially in relation to ammonium. In oxygen-minimum zones it was first hypothesized that sulfur cycling was linked to anammox via sulfide-driven DNRA (15). Anammox microorganisms were originally found in a sulfidic wastewater treatment bed (16), and DNRA has been demonstrated to support anammox organisms in an enrichment culture where sulfide was used as the electron acceptor for nitrate reduction (17). At high enough concentrations, sulfide can promote DNRA by diverting nitrogen away from the canonical denitrification pathway due to inhibition of NO⁻ and N₂O⁻ reductases (18) as well as by inhibiting nitrification (19). Sulfide has been demonstrated to stimulate DNRA in estuaries and in some circumstances may be pronounced enough to outcompete denitrification as the dominant nitrate reduction mechanism (14). H₂S has been shown to inhibit anammox activity at concentrations in excess of 10 to 30 μM in an enriched anammox culture (17, 20); however, 2 mM pulses of sulfide to a fluidized anammox bed reactor stimulated ammonium removal, potentially through biological nitrite production from nitrate (21). The species Kuenenia stuttgartiensis and “Candidatus Scalindua” are capable of self-generating anammox precursors from nitrate, but the overall importance of this process is unknown (22, 23).

The present study focuses on a shallow (20- to 30-cm) basin that receives nitrified municipal wastewater effluent (2). As a subclass, shallow open-water unit process wetlands were developed as a specialized component of larger engineered wetland systems that receive municipal wastewater effluent and effluent-dominated surface waters. These units were initially designed to enhance the removal of recalcitrant trace organic contaminants such as pharmaceuticals and personal care products (24–26) but were found to have ancillary benefits for nitrate (2) and pathogen attenuation (27, 28) through an interplay of photolytic and biological processes that rivaled or even surpassed their vegetated cousins. The open-water unit process wetland is lined with a geotextile fabric in order to prevent the growth of macrophytes that would shade the water column and contains a benthic photosynthetic biomat that when mature can be up to 10 cm thick. Biological activity in the biomat causes diurnal cycling due to competing photosynthetic and heterotrophic processes with oxygen supersaturation in the water column during the day and suboxic conditions at night. The microorganisms that colonized this system were initially investigated in Jasper et al. (26) and shown to contain a high abundance of a single species of diatom, Staurosira construens var. venter, in conjunction with a diverse bacterial community.

A combination of field measurements and microcosms was previously employed to establish denitrification as the primary nitrate removal process; however, anammox organisms were present in the wetland, and ammonium-spiked microcosms suggested that anammox activity was responsible for as much as 25% of total nitrogen attenuation (2). In the present study, we assess the underlying ecology and mechanisms enabling anammox in these freshwater systems to increase understanding of why and how anammox is occurring in an engineered wetland with limited exogenous contributions of ammonium and nitrite. Specifically, we hypothesized that sulfur cycling plays an important role in the mechanism of nitrate attenuation by providing a shunt to ammonia production, with a resultant shift from denitrification to anammox processes. Our investigation was grounded in an operational field-scale system where we used a synthesis-based inquiry that targeted nutrient cycling biomarker genes, geochemical analyses, and taxonomic and functional microbial profiling. As a complementary method, we utilized bench-scale microcosm experiments containing field-derived material to further explore functional processes. Collectively, our results bring further insight into how geochemical parameters could potentially be leveraged to impact nutrient cycling and attenuation during the operation of treatment wetlands as well as an increased understanding of the ubiquity of anammox in natural and engineered systems.

RESULTS AND DISCUSSION

Microbial ecology of biomat.

Microbial communities in the biomat were significantly different across both the sampling depth (Adonis R² = 0.259, P ≤ 0.001) and along the wetland flow path (Adonis R²= 0.374, P ≤ 0.001) as interpreted by results of high-throughput 16S rRNA gene sequencing at discrete points (Fig. 1). The depth trend was most evident in samples collected closer to the inlet (∼2 m), where the biomat thickness of approximately 10 cm was most pronounced. In addition to community shifts between sampling locations, community alpha diversity was found to decrease along the flow path (see Fig. S2 in the supplemental material). These shifts can be explained by nutrient or other resource limitations, such as the observed decrease in nitrate concentration from inlet to outlet.

FIG 1.

Microbial community similarity within the biomat as a function of depth and distance from the inlet. Significant community shifts were seen as a function of both depth and distance from the inlet. Principal-coordinate analysis was established using weighted UniFrac 16S rRNA gene sequences, and the percentages of the variation explained are indicated on the axes.

To better understand the mechanism of nitrate attenuation within the biomat, we focused on taxonomic distribution as a function of depth at the sampling region nearest to the wetland inlet, where the communities (Fig. 1) and water parameters, as discussed later in this paper, varied most with depth. Diatom (Stramenopile) plastids accounted for 20.2% ± 3.9% of the total sequences at the top and increased to 31.9% ± 2.5% of the sequences at the bottom of the biomat, suggesting either benthic accumulation or that they may play a role within the mat beyond that of photosynthesis. Given the large diatom presence, the sequences were removed from subsequent 16S rRNA gene community analysis in order to focus on bacterial and archaeal shifts. The majority of the remaining sequences from the top and bottom of the biomat were assigned to taxonomic classes as listed by rank (Fig. S3). Alphaproteobacteria and Cytophagia were more than twice as abundant at the biomat top than at the bottom, and Flavobacteria were almost exclusively located within the top portion of the biomat. In contrast, Anaerolineae, Deltaproteobacteria, and Bacteroidia roughly doubled, and the methanogens Methanomicrobia and Methanobacteria more than tripled in relative abundance from top to bottom.

To better understand the functional implications of these shifts, we utilized this differential abundance analysis to focus a literature synthesis of previously reported attributes of bacterial clades primarily at the taxonomic planes of family and genus where putative functionality can be more effectively inferred (29). This approach, with a caveat of imperfect associations between microbial structure and function, enabled us to identify microbial shifts and infer potential associations with carbon, nitrogen, and sulfur biogeochemical processes between the top and bottom samples of the biomat in an approximate 10-cm vertical stratification (Fig. 2 and Table S1). Our analysis revealed that the top of the biomat had a relatively higher abundance of phototrophic and aerobic heterotrophic microbial clades that consume simple sugars, which is in line with what would be anticipated in such a photosynthetic mat. The top also contained organisms in the ZB2 and Hydrogenophaga clades, which are putative aerobic methane (30) and hydrogen (31) oxidizers, respectively.

FIG 2.

Differential abundance (bars) between the top and bottom of the biomat is expressed as log₂ fold change, where positive values indicate a greater abundance in the top and negative values indicate a greater abundance at the bottom. Putative functionality assigned to clades (Table S1) supports a shift toward anaerobic respiration with depth. Error bars represent ±1 standard deviation of triplicate samples. Clades with overlapping error bars were removed for clarity.

Consistent with oxygen depletion at lower depths, the bottom region of the biomat community was enriched with anaerobic organisms, including the fermentative clades Anaerolineae and Bacteroidales (32–34). Within this region, methanogens, including Methanosaeta and “Candidatus Methanoregula,” which putatively oxidize both acetate and hydrogen/CO₂ to produce methane (35, 36), were more abundant. Anammox organisms from the family Brocadiaceae (37) were enriched 4-fold at the wetland inlet bottom, with a total relative abundance of 1.7% versus 0.4% at the top. A similar trend was observed for sulfate reducers, including Deltaproteobacteria and Syntrophobacter sp. (38, 39) at 1.00% and 0.91%, respectively, at the bottom of the biomat compared to 0.1% and 0.3% at the top. Another sulfate reducer, Desulfococcus sp., doubled in relative abundance from 2.8% to 5.7% from top to bottom but was not statistically significant according to the DESeq2 (differential expression analysis for sequences) algorithm. The increased presence of both sulfate-reducing bacteria and anammox bacteria suggests a positive relationship between sulfide production and anammox that was explored further.

Effect of sulfide on ammonium production.

To confirm the putative functionality of the organisms observed through sequencing with respect to nitrogen and sulfur biogeochemical processes, a series of biomat microcosms with various amendments were studied. Microcosms amended with 4 mM sulfate demonstrated the sulfate-reducing potential of the biomat, with conversion of 3.6 ± 0.6 mM sulfate and production of 2.60 ± 0.06 mM sulfide, while control microcosms without sulfate had no significant production of sulfide (Fig. 3). The incomplete stoichiometric recovery of sulfide may be attributed to precipitation of sulfide-containing minerals, adsorption to surfaces, or volatilization during analysis.

FIG 3.

The washed biomat converted sulfate to sulfide at near stoichiometric proportions, as evidenced by microcosms amended with sulfate (A) and without amendment (B), demonstrating a large sulfate-reducing capacity. Error bars represent ±1 standard deviation of triplicate incubations.

To understand the effects of sulfide on nitrogen cycling, specifically nitrate conversion to ammonium, anaerobic microcosms were amended with 3.4 mM nitrate, with or without gaseous additions of hydrogen sulfide equating to a 1 mM aqueous concentration. Microcosms containing amendments of both nitrate and sulfide converted 38% ± 1.6% of nitrate to ammonium (700 ± 23 μM) after approximately 1 day (Fig. 4A). During this period, 1,093 ± 112 μM sulfide was oxidized and 1,844 ± 152 μM nitrate was reduced, implying that DNRA was occurring. The observed sulfide oxidation and ammonium production ratio of 1.5:1 was greater than the theoretical ratio of 1:1 predicted by the equation below, which describes the DNRA reaction. One possible explanation for the discrepancy between these ratios could be the formation of elemental sulfur from sulfide oxidation. An analogous experiment by Burnet and Garcia-Gil reported a similar ratio of 30% of nitrate converted to ammonia in the presence of elemental sulfur production from sulfide; however, their sulfide consumption-to-ammonia production ratio was four time less (18) than observed here. Other possible explanations for this discrepancy include the consumption of produced ammonium due to anammox activity, as well as sulfide association with surfaces in the biomat or precipitation.

$${\text{HS}}^{-}+{{\text{NO}}_3}^{-}+{\text{H}}^{+}+{\text{H}}_2\text{O}\to {{\text{SO}}_4}^{2-}+{{\text{NH}}_4}^{+}$$

FIG 4.

Anaerobic microcosms amended with nitrate and sulfide (A), only nitrate (B), and neither nitrate nor sulfide (C). Microcosms reveal significantly increased ammonium production in the presence of sulfide. Error bars represent ±1 standard deviation of triplicate incubations.

Microcosms amended solely with a similar quantity of nitrate (1,963 ± 300 μM) resulted in a much lower yield of 127 ± 7 μM ammonium, which translated to just 6% of that produced in the sulfide-amended system (Fig. 4B). In contrast, microcosms amended with hydrogen sulfide but without supplemental nitrate generated negligible amounts of ammonium (Fig. S4). Control microcosms that received neither nitrate nor sulfide produced 210 ± 42 μM ammonium (Fig. 4C). The comparatively modest ammonium generation seen in the control and nitrate-only microcosms can be explained by cell death and ammonification associated with the harvesting and washing processes utilized to prepare the biomat. Collectively, these microcosm experiments support the potential of the biomat to facilitate sulfide-induced DNRA.

Spatial trends in biogeochemical processes.

Pore water profiles within the biomat from the pilot-scale wetland revealed spatial trends further supporting the hypothesis that sulfide-induced DNRA could be an important process near the inlet. Consistent with earlier characterizations of the wetland (2), concentration-dependent nitrate reduction was observed along the horizontal flow path and with depth in the biomat (Fig. 5A). Approximately 89% of the nitrate was removed between the 2-m and 70-m sample locations, assuming that surficial biomat pore water samples, 0.36 mM and 0.04 mM nitrate, respectively, were representative of concentrations in the overlaying water.

FIG 5.

Differences in gene abundance and chemical profiles reveal spatial trends for nitrogen and sulfur biochemistry. Abundances are shown for the anammox biomarker gene (hzsA) with ammonium concentration (A) and the dissimilatory sulfate reduction gene (dsrA) with sulfate concentration (B). Samples were taken in triplicate along the horizontal flow path of the wetland cell, discretized by depth (top, middle, and bottom), and normalized to 16S gene copies. Error bars represent ±1 standard deviation.

Nitrate concentrations were highest at the inlet surficial regions of the biomat (0.37 ± 0.05 mM) and decreased with depth (0.04 ± 0.016 mM) within the biomat. Mirroring this trend, the highest ammonium concentration was detected at the bottom of the mat near the inlet (2.4 ± 0.76 mM), which was about 6.5 times the concentration of nitrate in the water column, indicating a significant accumulation of ammonium. While DNRA is consistent with higher ammonium accumulation within regions fed by higher nitrate concentrations, the accumulation could also be explained by ammonification. If ammonification and nitrification were significant processes in our system, it stands to reason that ammonium concentrations and the relative abundance of anammox microorganisms would be similar at the biomat bottom throughout the cell. However, other sampling locations revealed a slight increase in ammonium concentration with depth, but concentrations were all less than 0.2 mM (Fig. 5A), suggesting that ammonification played a lesser role. Aerobic ammonium oxidation is also unlikely as methanogenic communities are observed throughout the bottom layers (Fig. 2), and it has previously been shown that oxygen penetrates less than 4 cm within this biomat (2). The accumulation of 1.9 mM ammonium in the bottom pore water was observed approximately 30 m from the inlet a year prior during the month of April (2), indicating that ammonium accumulation may not be limited to the inlet portion of the wetland throughout the year. Temperature and sunlight intensity affect nitrate removal rates within open-water wetlands (2), which could lead to seasonal spatial variation in nutrient profiles and explain variability in ammonium accumulation between the two studies.

Sulfate concentrations, like nitrate, decreased with depth within the biomat especially near the inlet. However, unlike for nitrate, net sulfate loss at the biomat-pore water interface was not observed across the wetland transect (Fig. 5B). Samples closest to the inlet showed the greatest decrease in sulfate concentration with depth (Δ1.5 ± 0.06 mM), with over 77% of the total sulfate present at the biomat surface reduced. The decrease in sulfate concentration with depth was more modest at other sampling locations, with losses of approximately 0.2 mM. The proximal concurrence of nitrate and sulfate reduction in the depth profile, combined with the accumulation of ammonium 2 m from the inlet, highlights the potential for sulfide-driven DNRA in our system.

In turn, we queried for the abundance of process-relevant genes in association with these trends. The sulfate reductase gene (dsrA) exhibited similar trends with depth, but was significantly elevated near the inlet (2- and 30-m sample points), with a maximum of 3.1% ± 0.44% relative abundance (Fig. 5B and S5B). Further from the inlet (50 m) abundances were approximately half those within the inlet region. This is consistent with 16S rRNA gene sequencing as the dominant putative sulfate reducer, Desulfococcus sp., harbored a relative abundance of 2.7 to 5.6% closest to the inlet and decreased by about half (1.2% to 2.3%) at the 50-m sampling point. Quantitative PCR (qPCR) results normalized to the 16S gene are approximately a 30% underestimate compared to relative abundances determined by sequencing as the diatom 16S plastid sequences were removed from sequencing analysis.

Identification and enumeration of microorganisms responsible for performing DNRA coupled to sulfide in this system were elusive due to their broad phylogenetic distribution (40). Unfortunately, there were limitations in amplification specificity/efficiency of the primers for nrfA in our system despite their prior success in an estuarine system (41) (Table 1). An octa-heme nitrite reductase, such as the reductases found in anammox organisms (42), could also be responsible for the conversion of nitrite to ammonium. Insights from a recent metagenomic inquiry in estuary sediments that reconstructed genomes of novel organisms involved in carbon, nitrogen, and sulfur cycling help to bridge this gap (43). Specifically, this investigation revealed that novel organisms within Gammaproteobacteria have the genes for both sulfide oxidation and nitrate/nitrite reduction. Interestingly, Gammaproteobacteria also exhibited dominance in our open-water wetland cell, with a bacterial population second only to Anaerolineae within the bottom layer of the biomat (Fig. S3). Furthermore, a broad range of diatom species have been reported to perform DNRA in the absence of light (44), which could explain their increased abundance in the bottom of the biomat, as described in the prior microbial ecology section. However, without a DNRA biomarker for diatoms and with the difficulty of culturing diatoms axenically, it could not be determined if diatoms are directly involved in ammonium production in our wetland biomat.

TABLE 1.

Primer pairs used for quantitative PCR analysis

Primer namea	Primer sequence 5′–3′	Primer concn (nM)	Amplicon size (bp)	Target gene	Thermal profile	Efficiency (%)	Reference
EUB338	ACTCCTACGGGAGGCAGCAG	1,000	180	16S	95°C for 3 min; 40 cycles of 95°C for 60 s, 53°C for 30 s, 72°C for 60 s	90	91
EUB518	ATTACCGCGGCTGCTGG	10,000
hzsA1597F	WTYGGKTATCARTATGTAG	400	260	hzsA	95°C for 3 min; 40 cycles of 95°C for 30 s, 55°C for 30 s, 72°C for 30 s	81	51
hzsA1857R	AAABGGYGAATCATARTGGC	400
Dsr1F	ACSCACTGGAAGCACGGCGG	500	221	dsrA	95°C for 10 min; 40 cycles of 95°C for 30 s, 58°C for 30 s, 72°C for 40 s; 80°C quantification	85	92
Dsr-R	GTGGMRCCGTGCAKRTTGG	500
nrfAF2aw	CARTGYCAYGTBGARTA	500	269	nrfA	95°C for 10 min; 40 cycles of 95°C for 15 s, 52°C for 45 s, 72°C for 20 s; 80°C quantification	∼50	93
nrfAR1	TWNGGCATRTGRCARTC	500

^aEUB primers were used to quantify 16S data to which all other genes were normalized. The following primers were used to show potential for each function: hzsA primers, anammox; dsrA primers, sulfate reduction; nfrA primers, DNRA.

Distribution of anammox bacteria.

Preliminary evidence for anammox activity in this engineered wetland was originally documented through the observation of ammonium removal by anaerobic microcosms by using the acetylene block method to estimate denitrification and anammox rates. In these experiments, it was estimated that up to 25% of nitrogen was removed via anammox (2); however, this could underestimate the actual contribution as acetylene can also inhibit anammox (45) in addition to the targeted denitrification process. In our current investigation, we further queried for potential anammox activity by quantifying the anammox-dependent hydrazine synthase gene, hzsA, throughout the wetland cell. The abundance of hzsA was significantly higher at the bottom and middle of the biomat relative to that at the top at both the 2-m and 30-m sampling points (Fig. 5A and S5A). The maximum relative abundance of 0.06% ± 0.01% hzsA decreased to below detection near the outlet (data not shown) where the biomat decreased in thickness and pore water nitrate concentrations were lower. As anammox organisms are oxygen sensitive (46, 47), this trend with depth could further be explained by oxygen production in the overlying waters in association with diatom primary productivity. Sequencing results also showed enrichment in the anammox family Brocadiaceae at ∼0.6% relative abundance within the biomat bottom (Fig. 2), which is consistent with other studies showing Brocadiaceae lineages present in soils and freshwater wetlands (48, 49). Relative abundances of qPCR and 16S sequencing differ by an order of magnitude, which is the same result observed by previous investigations (50, 51), possibly due to primer biases (52).

Comparatively less sulfate and nitrate reduction may have occurred in the upper portion of the biomat during the day due to oxygen-supersaturated conditions resulting from photosynthesis (2). This would in turn lead to fewer precursors for the anammox process. A similar trend in community distribution has been recently reported in freshwater marshes where a high abundance of anammox organisms was observed near the inlet of a vegetated wetland (53). Though not directly investigated in the current study, nitrite was detected only in trace quantities (<0.05 mM) near the top few centimeters of the studied biomat in a prior investigation (2). Its absence with depth presumably limits anammox processes, as has been documented in mangrove (54) and marine sediments (55), and provides an explanation for the accumulation of ammonium at the bottom of the biomat. Prior work in our wetland has revealed a significant nitrate gradient along the wetland flow path, with approximately 70% of the nitrate removed by the middle of the wetland in summer months (2); similarly, our present work shows 63% nitrate removal halfway through the wetland cell. Building on this theme, the decreasing anammox trend from inlet to outlet could have been due to nitrite and ammonium availability, which would decrease as nitrate decreases.

Environmental implications.

Our present study provides insights into eutrophic freshwater systems by documenting the interdependence of sulfide and nitrate, as well as the presence of anammox organisms in this engineered wetland. A recent study investigating nitrogen reduction pathways in estuarine sediments found no correlation between anammox and DNRA rates and sulfide (56); however, that finding was counter to results of a study focusing on the same location that found anammox bacterial abundance positively correlated to percent organic carbon and sulfide concentration (57). DNRA has been previously linked to sulfide oxidation in marine environments (58) and has been directly coupled to anammox in marine sediments and enrichment cultures (17, 59). While it is hypothesized that DNRA is coupled to anammox in terrestrial and freshwater environments (60), direct evidence in these types of systems has previously proved elusive.

More broadly, our analysis of this wetland biomat supports the importance of linkages between carbon, nitrogen, and sulfur cycling in both engineered and natural environments. The biomat within this wetland has an organic carbon content of 13% (2). Though it is generally accepted that organic content-rich environments favor faster-reproducing heterotrophs (21, 61), anammox microorganisms have been found in organic content-rich sediments in contact with nitrate-rich waters (9, 62) as well as periphyton-dominated aerobic sediments (63) analogous to the open-water wetland cell investigated here. More common surface flow vegetated treatment wetlands and analogous natural systems, such as wetlands and estuaries, receive nitrate-rich water and often have an active sulfur cycle (64–66) and large soil carbon pools that average 7.7% carbon content (67). These systems may also host DNRA-linked anammox processes, which should be considered in assessments and predictions of biogeochemical cycling and water quality. From an applied perspective, the diatomaceous and bacterial photosynthetic biomat formed in open-water cells could offer an alternative treatment system to systems present in conventional vegetated treatment wetlands (68). Unique and potentially functionally beneficial attributes include depth stratification within the biomat, diurnal cycling of oxygen content, and the autotrophic capabilities of the assemblage where organic production via photosynthesis could fuel reductive processes.

The impact of these systems on greenhouse gas emissions is promising but not yet clear. As autotrophic systems, they should represent a net carbon sink, and our microcosms demonstrated that DNRA can achieve nitrate removal rates similar to those of canonical denitrification (Fig. 4). DNRA linked to anammox has the potential to reduce N₂O emissions as electrons are shuttled away from N₂O generation toward ammonium (7, 69, 70), suggesting a potential N₂O control strategy by utilizing sulfate present in the water supply or adding it as a supplement. However, the biochemistry of organisms that perform DNRA and their potential for N₂O generation through competing processes such as nitrite/nitric oxide detoxification systems need further study. There is also the caveat of unintended side effects, such as the increased potential for methylmercury formation associated with sulfide production, that should be considered (71).

Our phylogenetic inquiry also revealed an increasing presence of methanogens in the lower layers of the biomat. As such, it is possible that denitrifying anaerobic methane oxidation (DAMO) could also be responsible for nitrate reduction or compete with anammox organism for nitrite, which appears to be a limiting resource in this system. Nitrite-dependent methane oxidizers have been cocultured with anammox bacteria (72) and naturally cooccur in various anaerobic environments (73–75). Organisms linked to anaerobic methane oxidation with nitrate or nitrite, such as anaerobic methane-oxidizing archaea (ANME) or NC10-type organisms (72, 76–78), were not identified by 16S sequencing; however, this dynamic needs further investigation in our system.

It has been hypothesized that ammonification or partial nitrification combined with incomplete denitrification at oxic/anoxic interfaces is responsible for providing the precursors for anammox (9, 53, 79, 80) and that sulfide production can inhibit the growth of anammox organisms as it prevents nitrification (76). However, our results provide field evidence for an alternate explanation previously reported in microcosms (17) in which anammox organisms increase in abundance in the most-reducing zones of the microbial environment and appear to thrive within and be influenced by sulfidic systems due to an increase of DNRA.

MATERIALS AND METHODS

Sample collection and processing.

Samples were obtained from the Discovery Bay, CA, pilot-scale open-water unit process wetland. This open-water cell receives nondisinfected, nitrified municipal wastewater effluent that typically contains approximately 1.4 mM nitrate, 2 mM sulfate, 0.5 mM total organic carbon, and ammonium and nitrite at concentrations below 0.05 mM (2). The organic matter content of the biomat in the wetland is 32% organic matter and 12% carbon content, with little variation throughout the wetland. The wetland cell is 20 m by 20 m and contains baffles to minimize short-circuiting by dividing the cell into four runs in series where run 1 receives influent and the end of run 4 is the outlet (see Fig. S1 in the supplemental material). Approximately 20 ml of slurry wetland biomat was collected for pore water and molecular analysis from four different locations during a period of active growth in late spring (5 May 2014). A depth profile was obtained by slowly subsampling in triplicate from the top (proximal to open surface water), middle, and bottom (proximal to geotextile membrane) regions of the biomat using a 30-ml pipette. The biomat decreased in thickness from 10 to 2 cm along the water's flow path and was sampled at approximately 2 m, 30 m, 50 m, and 70 m from the inlet along the flow transect. Samples were transported to the laboratory on ice and immediately centrifuged at 5,000 × g for 5 min to separate water and biomass.

Pore water samples from different depths and reaches within the biomat were analyzed the same day as collected, and biomass samples were frozen at −20°C until processed. Nitrate, chloride, phosphate, and sulfate were analyzed by ion chromatography (Dionex DX-120). NH₄⁺ was quantified in filtered (1-μm pore size) samples by colorimetric analysis (standard method 4500-NH₃ C) (81). Archived biomat samples were freeze-dried using a Labconco FreeZone 6. DNA was extracted from 0.025 g of freeze-dried sample using a MoBio Power-Biofilm DNA isolation kit according to the manufacturer's instructions. Extracted DNA was quantified on a Qubit 2.0 instrument and diluted by a factor of 10 prior to amplification, resulting in concentrations ranging from 10 to 30 ng/μl.

Quantitative PCR.

Amplification for quantitative PCR was performed with a Roche Light Cycler 480 II using Quanta Biosciences PerfeCTa SYBR green super mix in 25-μl reaction mixtures with 2 μl of template DNA. Primer sequences, PCR conditions, and efficiencies for 16S, hydrazine synthase (hzsA), dissimilatory sulfate reductase (dsrA), and nitrite reductase (nrfA) genes are reported in Table 1. Samples were amplified in duplicate in parallel with nontemplate and negative extraction controls. Standards were generated from amplicons that were purified using Beckman Coulter Agencourt AMPure XP magnetic beads (Brea, CA) and quantified using an Invitrogen Qubit 2.0 instrument. No nonspecific amplicon bands were observed when run on an Agilent Bioanalyzer 2100. Standard curves were generated from triplicate 1:10 serial dilutions of purified standards (82). Crossing-point values and efficiencies were determined using the LightCycler 480 II software, version 1.5.0, second derivative maximum method. Gene abundances were normalized to both 16S copy number and dry weight independently, and significance was determined using students t test with a P value of 0.05.

16S rRNA gene sequencing and analysis.

Amplification, purification, and normalization of samples for 16S rRNA gene sequencing were performed using primers for the V4 region in accordance with published methods, with minor modifications to the protocol which exclude the Bioanalyzer steps and include an additional concentration step (83). Briefly, 2 μl of extracted DNA was amplified with dual, indexed primers for 30 cycles (Phusion master mix; New England BioLabs). Amplicons were normalized and purified with a SequalPrep normalization plate kit. Normalized amplicons were pooled and concentrated with Amicon Ultra 0.5-ml centrifugal filters (30,000-molecular-weight cutoff). Pooled concentrates were quantified and sequenced by the Biofrontiers Institute at the University of Colorado (CU), Boulder, using an Illumina MiSeq platform with a version 2 reagent kit (2 by 250 bp).

Resultant 250-base-pair sequences were processed in QIIME, version 1.9 (84). Forward and reverse sequences of each sample were joined into contigs with the multiple_join_paired_ends.py script with a minimum base pair overlap of 100. Joined sequences were passed through multiple_split_libraries_fastq.py with default quality-control parameters. Operational taxonomic units (OTUs) were parsed using pick_open_reference_otus.py with Usearch, version 6.1 (85), and chimeras were filtered out using the Greengenes gold database (86). Representative sequences were aligned using PyNAST (87) and the Greengenes 13_8 aligned reference database, and alignments were filtered using the Lane mask (88).

Taxonomy was assigned using the Ribosomal Database Project (RDP) classifier with a confidence value of 0.5 and the Greengenes 13_8 OTU taxonomy database. Diatom plastid sequences were removed, and the OTU table was rarified to the minimum number of sequences for a sample, 15,926, for alpha and beta diversity measurements. OTUs with less than 0.1% relative abundance were filtered out for phylogenetic analysis. The DESeq2 method (89) was used within QIIME to calculate differential abundance between samples using the filtered taxonomic OTU tables. The principle component diagram was generated in R using the phyloseq package (90).

Anaerobic microcosms.

Freshly harvested biomat, collected on 27 January 2015, was shipped overnight on ice and stored refrigerated for no more than 2 weeks prior to use. Biomat material was washed two times with phosphate-buffered saline (PBS) and centrifuged to remove soluble constituents. This process was followed by a final suspension in PBS where 10 ml each of the slurry (0.5 g of biomat dry weight) was added to triplicate microcosms containing 70 ml of commercially purchased minimal freshwater diatom DY-V growth medium primarily composed of morpholineethanesulfonic acid (MES) buffer, KCl, H₃BO₃, Na₂ β-glycerophosphate, Na₂SiO₃, CaCl₂, trace elements, and f/2 vitamin solution (NCMA at Bigelow Laboratories) (subcomponents of the medium containing sulfur and nitrogen species, NH₄Cl, NaNO₃, and MgSO₄, were omitted); the microcosms were then amended in accordance with experimental variables, where appropriate, with nitrate (3.25 mM) and/or hydrogen sulfide at a predicted 1 mM aqueous-phase concentration (calculated with a dimensionless K_H air-water partitioning coefficient of 0.1). Incubations were conducted in 160-ml glass serum bottles purged with N₂ gas that were sealed with butyl stoppers and shaken at 90 rpm. Control microcosms without additional amendments were also included, and all microcosms were incubated in the dark to minimize photosynthesis. Nitrate, nitrite, ammonium, sulfate, and sulfide were monitored with Hach TNTplus kits 835, 839, and 831, SulfaVer 4 powder pillows, and methylene blue sulfide reagents, respectively, using a Hach DR5000 spectrophotometer.

Accession number(s).

Sequences were deposited in the National Center for Biotechnology Information Sequence Read Archive under accession number SRP069033.