Trace Metal Availability Affects Greenhouse Gas Emissions and Microbial Functional Group Abundance in Freshwater Wetland Sediments

We investigated the effects of trace metal additions on microbial nitrogen (N) and carbon (C) cycling using freshwater wetland sediment microcosms amended with micromolar concentrations of copper (Cu), molybdenum (Mo), iron (Fe), and all combinations thereof. In addition to monitoring inorganic N transformations (NO3–, NO2–, N2O, NH4+) and carbon mineralization (CO2, CH4), we tracked changes in functional gene abundance associated with denitrification (nirS, nirK, nosZ), dissimilatory nitrate reduction to ammonium (DNRA; nrfA), and methanogenesis (mcrA). With regards to N cycling, greater availability of Cu led to more complete denitrification (i.e., less N2O accumulation) and a higher abundance of the nirK and nosZ genes, which encode for Cu-dependent reductases. In contrast, we found sparse biochemical evidence of DNRA activity and no consistent effect of the trace metal additions on nrfA gene abundance. With regards to C mineralization, CO2 production was unaffected, but the amendments stimulated net CH4 production and Mo additions led to increased mcrA gene abundance. These findings demonstrate that trace metal effects on sediment microbial physiology can impact community-level function. We observed direct and indirect effects on both N and C biogeochemistry that resulted in increased production of greenhouse gasses, which may have been mediated through the documented changes in microbial community composition and shifts in functional group abundance. Overall, this work supports a more nuanced consideration of metal effects on environmental microbial communities that recognizes the key role that metal limitation plays in microbial physiology.


INTRODUCTION
Wetland microbes are important for removing anthropogenic pollutants from surface waters, effectively preventing the contaminants from entering downstream coastal and marine ecosystems. Human-derived industrial nitrogen (N) input to the environment has been estimated to 150 Tg N y −1 (Schlesinger, 2009), and is of particular concern because it is linked to water quality degradation, eutrophication, and increased emissions of nitrous oxide (N 2 O), which is a strong greenhouse gas (GHG) that also contributes to ozone depletion (Ravishankara et al., 2009;IPCC, 2013). Biological production of N 2 O is mainly through denitrification, nitrification, and nitrifier-dentrification; in addition, some denitrifying microbes can remove N 2 O by reducing it to inert dinitrogen (N 2 ). Annually, ∼10% of the human-derived N input is returned to the atmosphere as N 2 via denitrification from wetlands and terrestrial ecosystems globally (Schlesinger, 2009). Wetlands are equally important for their role in carbon (C) cycling. Despite their relatively small coverage (∼10% of global land area), these ecosystems store large amounts of organic C and emit considerable amounts of methane (CH 4 ) at an estimated rate of 144 Tg CH 4 y −1 (IPCC, 2014). These net CH 4 emissions represent a balance between the microbial processes of methanogenesis and methanotrophy.
Microbial production and consumption of N 2 O and CH 4 is catalyzed by oxidoreductases that utilize metal co-factors such as molybdenum (Mo), copper (Cu), and iron (Fe) (Glass and Orphan, 2012). For denitrification, the pathway includes the reduction of nitrate (NO 3 − ) to nitrite (NO 2 − ), catalyzed by a periplasmic or membrane bound Mo nitrate reductase (NAR) (Schwarz et al., 2009), and the further reduction of NO 2 − to nitric oxide (NO) by either a heme-Fe cytochrome cd 1 or a Cu co-factor nitrite reductase (NIR), expressed by the genes nirS and nirK, respectively (Adman et al., 1995;Einsle et al., 1999). NO is further reduced to N 2 O by nitric oxide reductase (NOR), which contains a heme-Fe cofactor (Shiro et al., 2012). Finally, N 2 O is reduced by a Cu-containing nitrous oxide reductase (NOS) (Rosenzweig, 2000). Similar metal complexes are key in methane cycling. For example, Fe-, Ni-, and Zn-dependent ferredoxins and dehydrogenases catalyze the early steps of methanogenesis, and all methanogens utilize a common Ni methyl-coenzyme M reductase (MCR; mcrA) to catalyze the rate-limiting and final step of CH 4 production (Glass and Orphan, 2012;Wongnate and Ragsdale, 2015). Cu also has an integral role in CH 4 oxidation. For example, aerobic methanotrophs scavenge Cu with methanobactin, a high-affinity chaperone (Chang et al., 2018), for use in Cu-dependent CH 4 monoxygenases. Some anaerobic methanotrophs, such as members of the NC10 phylum (e.g., Methylomirabilis oxyfera), utilize a Cu-dependent particulate methane monoxygenase (pMMO) and an Fe-rich cd 1 nitrite reductase to couple CH 4 oxidation to denitrification via nitrite-dependent methane oxidation (N-DAMO) (Deutzmann et al., 2014;Cheng et al., 2019). Because of these sorts of interconnected pathways and processes, the abundance and bioavailability of trace metals in the environment can impact N and C biogeochemistry and exert a control on associated GHG emissions.
Although extensive research has considered the relative contribution of key environmental factors (e.g., pH, NO 3 − , NO 2 − , O 2 , and C/N ratio) in regulating the microbial processes associated with N and C biogeochemistry, our understanding of the impact of metal availability is more limited. Metals in the environment have traditionally been seen as unwanted pollutants, and trace metal accumulation has been linked to toxicity and inhibition of ecosystem processes (Samanidou and Papadoyannis, 1992). For example, concentrations exceeding the mg L −1 range for Cu, Mo, Fe, Zn, and Pb severely inhibit denitrification in soils, sediments, surface water bodies, and waste waters (Labbé et al., 2003;Magalhães et al., 2007;Liu et al., 2016). However, given the dependence of many microbial enzymes on metal co-factors, it also is possible for the opposite effect to occur -wherein enzyme function is limited due to an inadequate supply of trace metals. This scenario has received limited attention in environmental studies, but is well documented in case studies with model organisms. In fact, in vitro studies have shown that lack of Cu, Mo, or Fe severely inhibits denitrification or methane cycling due to the formation of non-functional enzymes typically lacking the respective metal co-factor. For example, in the soil bacteria Paracoccus denitrificans and Pseudomonas stutzeri, Cu is required to express a functional NOS dimer and reduce N 2 O to N 2 (Granger and Ward, 2003;Felgate et al., 2012;Black et al., 2016). In methanotrophs oxidizing CH 4 to methanol (CH 3 OH), the switch between a Cudependent pMMO or Fe-dependent soluble MMO (sMMO) is regulated by the availability of each metal (Murrell et al., 2000;Bollinger, 2010).
Despite great progress in understanding metal ecotoxicity and metalloenzyme biochemistry, our knowledge of how trace metal availability affects microbial activity in the environment is generally limited to selected processes such as ammonium (NH 4 + ), NO 3 − , and CO 2 assimilation in aquatic ecosystems (Twining et al., 2007;Moore et al., 2013;Romero et al., 2013;Schoffman et al., 2016). A broader understanding of how metal availability regulates microbial processes, especially in soils and sediments, is necessary if we are to fully comprehend environmental controls on ecosystem N and C cycling. In this study, we investigated the effects of trace metal additions on the microbial biogeochemistry of freshwater wetland sediments focusing primarily on NO 3 − /NO 2 − reduction and GHG kinetics, and used quantitative polymerase chain reaction (qPCR) to assess changes in the abundance of key microbial functional groups associated with these processes. We hypothesized that greater bioavailability of Mo, Fe, and Cu in the sediments would increase denitrification and that greater Cu availability would reduce N 2 O emissions (i.e., by stimulating NOS functioning). We also predicted that CO 2 emissions would increase, driven by higher denitrification rates, and CH 4 production would decrease because stimulated denitrifiers would outcompete methanogens for C substrates under anoxic conditions.

Sampling and Sediment Properties
Wetland soil was collected from a tidal-fluvial bar deposit in Pamunkey River (Virginia, United States;N 37.557451, W −76.972521) in October 2016. Five samples from the 5-15 cm depth interval (∼350 g each) were collected ∼2 m apart using a PVC core (diameter, 10 cm). Porewater was collected from a small pit, which was emptied and allowed to refill naturally with porewater prior to collection. Samples were kept on ice in a cooler during transport to the lab and stored overnight (4 • C) until experimental setup and characterization using standard methods (SSSA, 1996). Total metal content was analyzed according to United States Environmental Protection Agency (EPA) standard methodology at the Soil Testing, Insect ID and Plant Diagnostic Lab, Cooperative Extension of the University of the New Hampshire (Durham, NH, United States), and porewater Fe(II) concentrations were measured spectrophotometrically following Viollier et al. (2000). Sediment and porewater properties are summarized in Table 1. Previous studies at this site had identified these sediments to be conducive of denitrification, dissimilatory nitrate reduction to ammonium (DNRA), and methanogenesis (Berrier, 2019;Dang et al., 2019).

Experimental Set-up
After removal of visible root and stone fragments, the five soil samples were combined in equal parts to form a composite sample for subsequent experiments. The soil (1 kg) was amended with 1.5 L of filtered porewater (Cellulose, 10-µm pore-size, Millipore, Burlington, MA, United States) and homogenized using a commercial blender (30 s at max speed) to a final soilto-liquid ratio of 1:3.5 (dry w/v). Soil slurry aliquots (50 mL) were then transferred to sterile 125-mL glass bottles (Wheaton, Millville, NJ, United States). The bottles were crimped with rubber septa (#224100-180, Weaton, Millville, NJ, United States) and incubated for 3 days in the dark (25 • C) to allow residual O 2 to be consumed. Each bottle then received 250 µL of 2 M KNO 3 , yielding a final nominal NO 3 − concentration of 10 mM, ensuring non-limiting N conditions corresponding to typical NO 3 − concentrations used in DEA protocols (SSSA, 1994;Murray and Knowles, 1999  Data are shown as mean ± standard error (n = 4) except for total metal content, which was analyzed using a composite sample.
Cu. Metal additions were based on a preliminary assessment of N 2 O emissions derived by a model denitrifying bacterium, P. denitrificans, incubated with different Mo, Fe, and Cu levels (Supplementary Table S1). Seven experimental treatments were established to test the effect of each metal individually (Mo, Fe, or Cu) and in all possible combinations (Mo + Fe, Mo + Cu, Fe + Cu, and Mo + Fe + Cu). In addition, control microcosms were prepared, to which no metals were added. Four replicate microcosms were prepared for each control and treatment. Each bottle was then vortexed briefly (30 s, 6000 rpm), flushed with N 2 (60 min), and incubated without shaking in the dark (25 • C). Gas (5 mL) and slurry supernatant (1.5 mL) samples were collected using a sterile syringe after 0, 6, 12, 24, 48, 72, and 96 h. Withdrawn gas volumes were replaced with equal N 2 gas volumes. Following the 96-h sampling, bottles were opened in an O 2 -free chamber, and 0.3 g dry weight of soil was removed and immediately frozen (−20 • C) for molecular analyses.
Gas samples were stored in 3 mL Exetainer vials (Labco, United Kingdom) that previously were flushed with N 2 and vacuumed with a gas-tight syringe by removing four volumes, i.e., resulting in a vial pressure of ca. 6 kPa. Concentrations of N 2 O, CO 2 , and CH 4 were determined via gas chromatography (Shimadzu GC-14A using Porapak-N and HayeSep-D and Molecular Sieve MS13 columns and equipped with ECD (N 2 O), TCD (CO 2 ), and FID (CH 4 ) detectors (Shimadzu, Columbia, MD, United States) as described in Morrissey and Franklin (2015). For each gas, total gas production was determined as the sum of the gas accumulated in the headspace and the gas dissolved in the liquid slurry (e.g., Robertson et al., 1999) using the Ideal gas law, Henry's gas solubility law, and Bunsen coefficients (at 25 • C) of 0.545, 0.032, and 0.772 for N 2 O, CH 4 , and CO 2 , respectively (Felgate et al., 2012;Sander, 2015). To compare overall GHG emissions across treatments, we estimated the combined global warming potential (GWP) of CO 2 , N 2 O, and CH 4 in terms of CO 2 -equivalents (g CO 2 -eq) using the 100 years GWP factors of 298 for N 2 O and 28 for CH 4 as calculated on a mass to mass basis (IPCC, 2013).

Molecular Techniques
DNA was extracted from frozen soil samples (∼0.3 g dry weight) using the DNEasy Kit from Qiagen (Germantown, MD, United States) following the manufacturer's instructions The retrieved DNA was visualized on a 1.2% agarose gel and quantified using a Nanodrop spectrometer (Thermo Fisher Scientific, Wilmington, DE, United States). Quantitative PCR (qPCR) was used to determine the abundance of microbial groups typically found in wetland sediments: total bacteria (eub), nitrite reducers (denitrification: nirK and nirS; DNRA: nrfA), nitrous oxide reducers (nosZ-clade I and II), and methanogens (mcrA). SensiFAST TM SYBR R No-ROX Kit Polymerase 2× mix (Bioline, United Kingdom) was used for all reactions except nosZ-clade II, which used SYBR green Go-Taq R qPCR Master Mix (Promega, Madison, WI, United States). Primers were purchased from IDT (Integrative DNA Technologies, Skokie, IL, United States) and the DNA for the standard curves was extracted from isolates obtained from ATCC (American Type Culture Collection, Manassas, VA, United States). Data were collected using a Bio-Rad CFX-384 Real Time System (Bio-Rad, Hercules, CA, United States) and analyzed using CFX Manager software (Ver. 3.1), except nosZ-clade II quantification was conducted using QuantStudio 6 Flex (Thermo Fisher Scientific, Wilmington, DE, United States). Primers, qPCR reaction conditions, and efficiencies are summarized in Table 2.

Statistical Analyses
Data did not comply with the normality and variance homogeneity assumptions for ANOVA [RStudio (base); Shapiro's test, Bartlett's test]; therefore, non-parametric Kruskal-Wallis tests with Bonferroni correction were applied to assess differences in the mean of ranks among the treatments [RStudio (agricolae)], without any data transformation. For all statistical tests, p ≤ 0.05 was considered significant. Summary statistics were calculated with RStudio (dyplr) (Boston, MA, United States) and plotted with SigmaPlot 14 (Systat Software Inc., San Jose, CA, United States). Central tendency and measures of dispersion are shown as mean ± standard error (SE) with n = 4, unless otherwise specified.

RESULTS
The Effect of Trace Metal Additions on NO 3 − , NO 2 − , and NH 4 + Kinetics Kinetics of NO 3 − and NO 2 − transformations in microcosm pore-water were similar across all treatments. Nitrate reduction commenced at the same time for all the treatments (12-24 h, Figure 1), and the NO 3 − pool (10-11 mM) was generally consumed within the first 48 h of the incubation. Nitrite was produced concurrently with NO 3 − depletion and, on average, 0.25 ± 0.02 mM NO 2 − remained across all treatments at the end of the experiment (Figure 1). An increasing trend in pore-water NH 4 + concentration was observed for all treatments during the incubation, and treatments receiving additional Mo + Fe (p < 0.01) and Mo + Fe + Cu (p < 0.01) had significantly higher NH 4 + than the control at the end of the experiment (Figure 1). On average, 2.7 ± 0.1% and 4.6 ± 0.7% of NO 3 − was recovered as NO 2 − or NH 4 + , respectively, at the end of the incubation.
The Effect of Trace Metal Additions on N 2 O, CH 4 , and CO 2 Cumulative Gas Kinetics Concurrent with NO 3 − consumption in the slurries, N 2 O production increased rapidly between 24 and 48 h, especially in treatments without added Cu (Figure 2A). The control and treatments with Mo, Fe, and Mo + Fe accumulated significantly more N 2 O (approximately 79, 55, 62, and 94% of added NO 3 − , respectively), than the treatments with Cu, Fe + Cu, Mo + Cu, and Mo + Fe + Cu (approximately 13, 14, 4, and 11% of the added NO 3 − , respectively; Figure 2D). Total N 2 O concentrations reached a plateau between 48 and 96 h, most likely due to exhaustion of available C and NO 3 − . CH 4 accumulated quickly in the microcosms (6-12 h) and continued to be produced for the duration of the incubation, reaching a final concentration of  ∼0.8 mmoles (Figure 2B). By the end of the incubation, all metal additions caused an increase in CH 4 production relative to the control, though the differences were only significant between the control and the Cu, Mo + Cu, and Mo + Fe + Cu treatments ( Figure 2E). CO 2 production also increased throughout the incubation period, most rapidly between 12 and 24 h, and reached ∼2 mmoles CO 2 at the end of the incubation ( Figure 2C). As with CH 4 , the only significant treatment differences for CO 2 production were between the control and the Cu, Mo + Cu, Mo + Fe + Cu treatments ( Figure 2F).

The Effects of Trace Metal Addition on Targeted Microbial Groups
Bacterial abundance in the control treatment was 5.4 × 10 8 (±4.5 × 10 7 ) 16S rDNA copies g −1 sediment dry weight, which did not differ significantly from the abundance estimates obtained for the Cu, Mo, Mo + Fe, or Mo + Fe + Cu treatments. Abundance in the three remaining treatments (Fe, Fe + Cu, and Mo + Cu) was slightly higher (1.4fold) and significantly different from the control (p = 0.005; Supplementary Figure S1). Trace metal addition affected the abundance of nirK (p < 0.001), nirS (p < 0.001), nosZ-I (p = 0.001) and nosZ-II (p = 0.013) denitrifying microbial groups (Figure 3). Reducers of NO 2 − utilizing Cu-NIR (nirK) were more abundant in all treatments when compared to the control, though the increase due to Fe addition was not significant. Microbial groups having a cytochrome cd 1 -NIR (nirS) were relatively more abundant in the Mo and Fe + Cu treatments. The nosZ clade I and II microbial groups responsible for the reduction of N 2 O to N 2 were in general more abundant in the treatments that received Cu. We observed no effect of trace metals on DNRA NO 2 − reducers (nrfA, p = 0.24; Figure 3) but found that methanogens were more abundant in the treatments containing additional Mo (Mo, Mo + Cu, Mo + Fe, and Mo + Fe + Cu; mcrA (p = 0.001); Figure 3) when compared to the control.

DISCUSSION
We observed that trace metal (Mo, Fe, Cu, and their combinations) addition to wetland sediments regulated GHG emissions (Figure 2) and altered the abundance of microbial functional groups typically associated with N removal and C cycling (Figure 3). Previous studies have examined the effects of trace metal availability on these processes, often reporting inhibitory effects (Giller et al., 1998;Magalhães et al., 2007Magalhães et al., , 2011Deng et al., 2018;Keller and Wade, 2018) however, the range of metal concentrations previously tested far exceeds the levels applied in the current study. The key interpretation of the current experiment is that when available at trace levels, metals, specifically Cu, enhanced the reduction of N 2 O to N 2 and increased CH 4 and CO 2 emissions from wetland sediments. When N 2 O, CH 4 , and CO 2 were combined and expressed as g CO 2 -eq per microcosm, all treatments receiving Cu addition (Cu, Mo + Cu, Fe + Cu, and Mo + Fe + Cu) had significantly lower CO 2 -eq emissions (p < 0.001; Figure 4), indicating a major role of Cu in anaerobic respiration (denitrification and methanogenesis) as a regulator of nutrient cycling and GHG emissions.

Denitrification
In our experiments, none of the trace metal additions had an effect on NO 3 − removal (Figure 1), which is somewhat surprising since all bacterial nitrate reductases contain a Mo cofactor at their active sites (Moreno-Vivian et al., 1999). This may indicate that Mo was not limiting to the denitrifiers in this set of soils. Also, we observed a small delay in NO 3 − reduction (∼24 h) in all treatments that could be attributed to the lag time required to induce the denitrifying pathway within a population less accustomed to substantial concentrations of this alternative electron acceptor (Dodla et al., 2008). Concurrent with the loss of NO 3 − , we observed minor transient accumulation of NO 2 − ; this pattern has been observed frequently (Cooper and Smith, 1963;Taghizadeh-Toosi et al., 2020;Wang et al., 2020) and is thought to be a result of stochastic transcriptional regulation of the nir operon (Hassan et al., 2016). Following NO 2 − reduction, NO is produced. Though we did not analyze NO concentrations, it is unlikely that NO accumulated in our microcosms but was instead quickly reduced to N 2 O. Since NO has potent cytotoxicity at µM levels (Chaudhari et al., 2017;Hartop et al., 2017), any accumulation would have had a negative effect on other aspects FIGURE 3 | Gene copies per g sediment (dry weight) at the end of incubation (96 h) in microcosms amended with molybdate (Mo), iron (Fe), copper (Cu), and combinations thereof. Genes were targeted representing nitrite reducers (nirS, nirK, and nrfA), nitrous oxide reducers (nosZ-I and II), and methanogens (mcrA). Data are shown as mean ± standard error (n = 4). Letters within each bar graph indicate significant differences as determined by Kruskal-Wallis and Bonferroni post hoc testing.
of microbial metabolism, particularly CO 2 production, which we did not observe.
Regardless of what other metals were included in the microcosms, we observed a dramatic effect of Cu addition, leading to enhanced reduction of N 2 O to N 2 (Figure 2) and an increased abundance of nitrite (nirK) and nitrous oxide (nosZ clade I and II) reductase genes (Figure 3), which both code for Cu-containing enzymes. Overall, gene abundance for nosZ clade II was ∼10-fold greater than for clade I, which is consistent with recent reports that clade II is the dominant form in many soil ecosystems (Jones et al., 2013Orellana et al., 2014;Ligi et al., 2015;Graves et al., 2016). Clade II is also affiliated with a broader diversity of organisms than Clade I including several non-denitrifiers (Graf et al., 2014;Hallin et al., 2018), so it is important to recognize that our Clade II estimates may reflect changes in community composition that do not necessarily directly relate to N 2 O reduction. Subsequent analysis using primers that target subclades within Clade II (sensu Chee-Sanford et al., 2020) or metagenomic sequencing could help resolve this and provide a better understanding of the ecophysiology of various nosZ populations. The overall strong effect of Cu on N 2 O reduction was prevalent because the last step of denitrification is catalyzed by a Cu-containing reductase and no alternative pathways for N 2 O transformation have yet been discovered (Thomson et al., 2012). In other laboratory experiments, Cu has similarly been found to be a strong driver of denitrifying metabolism. For example, Cudeficient cells can have more transcripts of nosZ and Cuscavenging genes to compensate for the loss of N 2 O reduction due to limited Cu availability (Felgate et al., 2012). Also, strains deficient in Cu-transporters and chaperones may be unable to reduce N 2 O to N 2 , forming a nonfunctional NOR, even though nosZ is expressed (Sullivan et al., 2013). In our microcosms, denitrifying bacteria were able to rapidly take advantage of excess NO 3 − due to copious bioavailable Cu for synthesis of NOS. Even though concentrations of NO 3 − and Cu are usually lower in environmental scenarios, similar trade-offs may exist, which could affect ecosystem emissions of N 2 O and require further investigation.

DNRA
Over the course of the study, small accumulations of pore-water NH + 4 occurred (Figure 1) depending on the type of metal added, and DNRA could be another possible factor affecting the fate of added NO 3 − . The accumulation of NH 4 + was greatest in treatments that contained both Mo and Fe (i.e., Mo + Fe and FIGURE 4 | Global warming potential in g of CO 2 equivalents (g CO 2 -eq) of cumulative gas concentrations (96 h) of nitrous oxide, methane, and carbon dioxide in anoxic slurries amended with molybdate (Mo), iron (Fe), copper (Cu), and combinations thereof. Data are shown as mean ± standard error (n = 4). Letters indicate significant differences as determined by Kruskal-Wallis and Bonferroni post hoc testing.
Mo + Fe + Cu), which may indicate a synergistic response of the enzymes that reduce NO 3 − (all nitrate reductases are Mo-dependent) and the DNRA-specific nitrite reductase (nrfA), which includes a multi-heme complex. Overall, the accumulated NH 4 + (<1 mM) was much lower than the amount of NO 3 − removed (∼10 mM), and the abundance of the DNRA marker gene nrfA was low (Figure 3), so it is likely that negligible DNRA occurred in our microcosms. Indeed, anoxic organic matter mineralization (as indicated by consistent CO 2 production) could contribute to the accumulation of NH4+ and the observed differences due to metal addition could be indirect responses. DNRA and N mineralization commonly co-occur in wetland sediments (White and Reddy, 2009), but DNRA seems to be important mainly in environments with high C availability and limited NO 3 − availability (Hill, 2019). More conclusive results on DNRA activity in the present wetland setting require additional analyses, e.g., using 15 NO 3 − to trace the production of 15 NH 4 + from DNRA. Additional analyses using alternate nrfA primers could also be informative. In particular, the qPCR primers we used have poor coverage of NrfA Clade I and thus limited ability to detect members of the family Geobacteraceae, which have recently been found to be potentially significant for DNRA across a range of different soil habitats (Nelson et al., 2016). Given the recognized importance of Geobacteraceae in metal cycling (Röling, 2014), further consideration of this group using novel primers such as those recently developed by Cannon et al. (2019) would better assess the genetic potential for DNRA and effects of trace metal availability.

Effect on Carbon Mineralization
Wetland soils are large C sinks because anoxic conditions constrain organic matter mineralization and oxidation. In freshwater wetlands, the lack of electron acceptors such as O 2 , NO 3 − , and SO 4 2− , may lead to fermentative and reductive conditions where organic C will be reduced to CH 4 (Kim et al., 2015;Cheng et al., 2019). In this study, the pattern of CO 2 production primarily reflected the magnitude of denitrification, as discussed earlier. We assumed that the fraction of total CO 2 as bicarbonate (HCO − ) was likely constant, because at the end of the incubation pH increased approximately 1 pH-unit due to denitrification and no substantial differences were found in pH among the treatments (Supplementary Table S2). Equally, we did not quantify and identify forms of dissolved organic carbon as this was out of the scope of this study. We hypothesized that denitrifier activity would outcompete methanogen activity similar to the findings of Klüber and Conrad (1998); rather, CH 4 accumulated throughout the incubations (Figure 2B). This result was unexpected because denitrification should suppress methanogenesis as a result of the higher Gibb's free energy ( G 0 ). This set of sediments originates from a freshwater wetland that is known to produce significant amounts of CH 4 and host diverse methanogenic communities (Morrissey and Franklin, 2015;Berrier, 2019;Dang et al., 2019). It appears that the inhibitory effect of NO 3 − addition was transient or partial. It is possible that the rapid NO 3 − consumption and subsequent removal of denitrification intermediates alleviated inhibitory effects on methanogenesis. Since both processes were occurring, it is evident that denitrifiers and methanogens were competing for available C.
Overall, the higher abundance of methanogens (mcrA) and noticeable accumulation of CH 4 in all metal treatments versus the control indicates an important role of trace metal availability in CH 4 cycling in our system. As expected, Mo addition enhanced the abundance of mcrA genes (Mo, Mo + Cu, Mo + Fe, and Mo + Fe + Cu treatments; Figure 3). This is because a Mo co-factor is typically required for formylmethanofuran dehydrogenase (fmd) that catalyzes the reduction of CO 2 to formyl-methanofuran in the first step of methanogenesis by reduction of CO 2 with electrons from H 2 (Glass and Orphan, 2012). The subsequent step in methanogenesis is catalyzed by MCR, which is a Ni-containing enzyme. Because we did not add Ni in any of our treatments, we assumed that MCR would not change (or would decrease due to competition with denitrification). Instead, Mo addition and the increasing partial pressure of CO 2 in the bottles during the incubation seem to have triggered a shift toward a Mo-based methanogenesis pathway. In general, CH 4 production in Pamunkey River sediments have been found to proceed by hydrogenotrophic, acetoclastic, and syntrophic pathways (Berrier, 2019;Dang et al., 2019).
Though no Cu-dependent enzymes have yet been identified in the methanogenesis pathways (Glass and Orphan, 2012), the effect of Cu concentration on CH 4 emissions from wetlands is an area of active research. Prior work by Keller and Wade (2018) found a strong suppression of CH 4 emissions following Cu amendments to peatland soils, and argued broadly that Cu may inhibit methanogenesis in wetland environments. This contradicts our finding that Cu addition elicited more CH 4 than the control (Figure 2), as well as similar results from Thomas and Pearce (2004). Interestingly, the discrepancy across these three studies cannot be explained by simply looking at the concentration of Cu added; our Cu amendments were approximately 10-fold lower than those made by Keller and Wade (2018), whereas the amendments made by Thomas and Pearce (2004) were 10-fold higher. Understanding the effects of Cu on net CH 4 fluxes requires further study and should consider potential mitigating factors such as soil texture and cation exchange capacity (Thomas and Pearce, 2004), the impact of Cu amendments on dissolved organic carbon availability (Jiao et al., 2005), and metal effects on methane oxidation (Mohanty et al., 2000).

Environmental Implications
Previous studies have examined the effects of trace metal abundance (or addition) on N and C cycling in sediments, peat, and agricultural soils. Typically, those studies report an inhibitory effect; however, the ranges of trace metal concentrations in those studies far exceed the levels of the current study. For instance, Keller and Wade (2018) amended peat slurries with ∼200 µM trace element solution and found a significant decrease in CH 4 emission, most likely due to Cu-induced toxicity. Similarly, Magalhães et al. (2007) observed 85% inhibition of denitrification accompanied by accumulation of substantial N 2 O and NO 2 − at 79 µg Cu g −1 sediment (equivalent to 1200 µM Cu) in estuaries. In a follow-up study, lower denitrification rates due to Cu (∼900 µM) were accompanied by a decline in the abundance and β-diversity associated with the nirK, nirS, and nosZ microbial groups (Magalhães et al., 2011), while keeping in mind that the coverage of those primers has improved greatly since then. Elsewhere, the particularly low concentrations of dissolved Cu, Fe, and Mo (∼5, 1, and 0.2 µg L −1 , respectively) and the copious dissolved organic C (∼80 mg L −1 ) (Raudina et al., 2017) in peatlands triggered substantial emissions of CH 4 and N 2 O (Basiliko and Yavitt, 2001;Basiliko et al., 2013;Voigt et al., 2017). A more recent study that tested the effects of Cu pollution (addition of 100 µg L −1 , ∼1.6 µM) on urban freshwater wetland sediments found significantly reduced CH 4 emissions, but no change in N 2 O emissions (Doroski et al., 2019). In the latter case, the lack of a N 2 O response could be due to low NO 3 − availability (18 µM) relative to C.
The abundance of bioavailable metals in the environment, typically in their ion forms, is often several orders of magnitude lower than the routinely reported total metal content. Further, the ion form of trace metals could be unavailable to microorganisms due to physicochemical interactions (release and sorption kinetics with soil particles and organic matter) and plant uptake (Giller et al., 1998). At neutral and alkaline pH levels, metals tend to be effectively immobilized as inorganic compounds (metal-oxides, -hydroxides, andcarbonates). Additionally, organic complexes such as humic ligands are known to bind metal cations thus lowering their availability to microbes and plants. Furthermore, hydrogen sulfide (H 2 S), naturally produced due to dissimilatory SO 4 2− reduction in salt marshes, may limit the availability of trace metals through the production of insoluble metal-sulfide complexes with potential implications for N cycling and N 2 O emissions (Gauci et al., 2004;Butterbach-Bahl et al., 2013). Sulfate-reducing bacteria may outcompete methanogens and thus suppress CH 4 emissions in salt marshes. In the current study, the added SO 4 2− as CuSO 4 and FeSO 4 could potentially have been reduced to H 2 S; however, the initial and final porewater SO 4 2− concentrations remained at comparable levels (average change over time across treatments was only 2.8%, Supplementary Table S2) and no sulfide odor was detected when microcosms were opened.
Out of the three metals tested here, Fe 2+ addition does not appear to influence N and C cycling in this set of sediments. Fe 3+ is an abundant element in our biosphere including wetlands, however not directly biologically available, as it has to be reduced to Fe 2+ . Fe 2+ was added in our microcosms due to its role in the reduction of NO 2 − to NO with cytochrome cd1 NIR. Apparently, this set of sediments has the capacity to supply enough Fe 2+ . Another, Fe-dependent N-cycling process besides denitrification and DNRA is anaerobic ammonium oxidation (anammox), which relies heavily on cytochromes and oxyreductases containing Fe (Ferousi et al., 2017), and its relative prevalence could be determined by isotopic assessments (Brunner et al., 2013).
Ex situ experimentation offers key advantages in disentangling factors in a controlled environment. It is acknowledged that longer incubation times could shift the microbial community toward a fitter structure for trace metal or high levels of NO 3 − utilization. The aim of this study was to investigate the prospective of trace metal additions stimulating mixedcommunity microbial systems, as seen previously on model organisms. Equally, longer incubations, frequent sampling and non-limiting substrates could help us illustrate the potential of trace metals to enhance microbial functioning and microbial groups. Follow-up studies should consider shorter incubation times and lower substrate concentrations to understand the temporal shift in microbial function and diversity upon trace metal additions.
The present results suggest that trace levels of Cu exert a more important effect on N 2 and N 2 O emissions from the environment than previously thought. These findings are important not only for understanding fundamental controls on N 2 O production, but also for making predictions about potential future emissions given increasing nutrient and metal loads associated with urban pollution. Likewise, the effect of trace metal availability on denitrification enzymology also is of considerable interest in the context of agriculture and food production, and it has been suggested that metal additions could be a possible strategy to mitigate high soil N 2 O emissions associated with crop production (Richardson et al., 2009;Shen et al., 2019).

CONCLUSIONS AND FUTURE CHALLENGES
Our results demonstrate that trace metal availability affects microbial processes, resulting in greater GHG emissions and C mineralization. We found that even short-term (96 h) manipulation of trace metal availability led to changes in community structure (functional group abundance) and potential increases in GHG emissions. In particular, our results suggest that trace metal bioavailability may, directly or indirectly, regulate or co-limit denitrification kinetics and favor more active microbial communities that are able to quickly acquire the bioavailable metals and incorporate them in functional oxidoreductases. Disentangling the potential controls of various metals on denitrification and other N-cycling processes such as DNRA and anaerobic ammonium oxidation (anammox) and covarying effects on C cycling, requires additional experimentation across a range of metal concentrations, substrates [e.g., lower (NO 3 − )] and biochemical processes in different soil types and ecosystems. Future studies should also investigate metagenomic and metatranscriptomic profiles to understand and evaluate the ecological importance of trace metal limitation. This would allow a more comprehensive examination of potential impacts beyond the small group of functional genes that we considered, and would avoid issues such as primer bias, poor coverage, and inefficient amplification often associated with qPCR of phylogenetically diverse groups. Such studies should be accompanied by high-throughput metabolite or process rate analysis either in the lab or in the field.

DATA AVAILABILITY STATEMENT
All datasets presented in this study are included in the article/Supplementary Material.

AUTHOR CONTRIBUTIONS
GG designed the experiments, performed lab analysis, statistically tested and interpreted the data, and wrote and reviewed the manuscript. KRH performed lab analysis, and wrote and reviewed the manuscript. BLB, BS, and LE performed lab analysis and reviewed the manuscript. RBF performed lab analysis, interpreted data, and wrote and reviewed the manuscript. All authors contributed to the article and approved the submitted version.

FUNDING
This work was supported by a grant from the National Science Foundation (DEB 1355059) and is contribution (#92) from the VCU Rice Rivers Center.

ACKNOWLEDGMENTS
We thank Dr. Ioannis Ipsilantis for his comments on an earlier version of this manuscript. We also thank Colin McKenny for assistance in field and lab, Gabriella Balasa and Renia Passie for assistance with qPCR, Olivia De Meo for setting up the GC and HPLC procedures, and Dr. Scott Neubauer for GC and HPLC equipment access. This manuscript has been released as a pre-print at bioRxiv (#515809) (Giannopoulos et al., 2019).

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2020.560861/full#supplementary-material FIGURE S1 | Average eub gene copies g −1 sediment (dry weight) at the end of incubation (96 h) in microcosms amended with molybdate (Mo), iron (Fe), copper (Cu), and combinations thereof. Letters above each bar graph indicate significant differences as determined by Kruskal-Wallis and Bonferroni post hoc testing.  Giannopoulos, G. 2015).