Introduction

The net nitrogen (N) inventory in marine ecosystems is determined by the balance between N losses (e.g., denitrification and burial) and new N production (e.g., N-fixation, atmospheric deposition, and riverine inputs). However, the N cycle is more dynamic and intricate than observed historically, as new pathways and organisms become incorporated into our understanding. In addition to the major pathways of the N cycle—N-fixation, organic matter remineralization, nitrification (oxidation of ammonium to nitrate), and denitrification (the stepwise reduction of nitrate to N2 gas)—we now know that other processes play a key role in the marine environment. Additional nitrate reduction pathways—dissimilatory nitrate reduction to ammonium (DNRA; Koike and Hattori 1978) and N2 production via anaerobic ammonium oxidation (anammox, Mulder et al. 1995)—play a more significant role in the environment than previously thought (Burgin and Hamilton 2007; Lam et al. 2009). Previously unknown microorganisms mediating N pathways, such as unicellular N-fixers (e.g., cyanobacteria, Zehr and Turner 2001; heterotrophic bacteria, Welsh 2000) and ammonia-oxidizing archaea (Francis et al. 2005), may be significant or even dominant in the ocean. These new developments increase the complexity of measuring and modeling the N inventory. This complexity is compounded by anthropogenic N-loading ultimately from the Haber-Bosch process, which adds three times as much N globally as biological N-fixation (Galloway et al. 2013). Understanding, mitigating, and modeling the impacts of this N-loading requires quantification of these N transformation rates.

Over the last decade, new evidence has shown a wider occurrence of N-fixation and a more diverse community of N-fixers than previously thought. For example, Trichodesmium, a well-studied filamentous cyanobacterium, was considered one of the few significant sources of new N to marine environments (Capone et al. 1997). However, unicellular cyanobacteria can also contribute substantially to pelagic marine N-fixation (Zehr and Turner 2001). Additionally, rates of heterotrophic N-fixation in coastal sediments are higher than measured previously. In the past, acetylene reduction assays were the primary method for measuring heterotrophic N-fixation in sediments. Rates measured with this technique were low (8.1 × 10−3 to 13 μmol N m−2 h−1) and not important to balance N budgets, leading to the conclusion that sediment N-fixation was not a significant N source (see reviews by Capone and Carpenter 1982; Howarth et al. 1988; Seitzinger and Garber 1987). However, this method significantly alters the active microbial community (Fulweiler et al. 2015). Newer methods have the advantage of not using inhibitors and preserving the redox gradient at the sediment-water interface (e.g., stable isotopes and direct measurements of dissolved gasses in whole-core incubations). These methods generally yield higher rates of sediment N-fixation when it is detected. Recent studies indicate that N-fixation (reported as negative values, indicating sediment N2 uptake) can be a significant N source in coastal sediments (−12 to −250 μmol N m−2 h−1 reported in: Gardner et al. 2006; Fulweiler et al. 2007; Mortazavi et al. 2012; McCarthy et al. 2015), reservoirs (approx. −200 μmol N m−2 h−1; Grantz et al. 2012), and tidal flats (−402.7 μmol N m−2 h−1; Vieillard and Fulweiler 2012).

Why and how N-fixation occurs in sediments are not well understood. For cyanobacteria, the expression of the nitrogenase (nif) operon is repressed by ammonium (Lindell and Post 2001), which is often replete in sediment pore water (e.g., Froelich et al. 1979). However, heterotrophic diazotrophs can have a much higher ammonium threshold (>1000 μM) before nitrogenase inhibition occurs (Welsh 2000). Knapp (2012) compiled results from numerous studies in a variety of sediments and documented N-fixation at in situ concentrations of up to 30 μM NO3 − and/or 200 μM NH4 +. Therefore, other factors (e.g., oxygen, organic carbon availability) may regulate microbial composition and heterotrophic N-fixation rates in sediments. Sediment core incubations from Laguna Madre, TX, showed increased net N-fixation after glucose additions (An et al. 2001).

It is difficult to measure denitrification and N-fixation in sediments simultaneously, as denitrification produces and N-fixation consumes N2 gas. The geochemical factors determining the balance between denitrification and N-fixation (and, therefore, the net N2 flux) are not well understood. Denitrification is influenced by oxygen concentration, as well as carbon and nitrate availability (Cornwell et al. 1999; Ward et al. 2009). Field and mesocosm studies (Fulweiler et al. 2007, 2013; Eyre et al. 2013) in Narragansett Bay, RI suggest that both the quality and quantity of available organic carbon may control the balance between these processes. However, rates of these processes have not been measured separately and simultaneously, and heterotrophic N-fixation has not been proven to occur in sediments beyond the observation of N2 uptake. This study addresses both of those issues in Waquoit Bay, MA, a typical temperate, eutrophic estuary that has been shown to exhibit both net denitrification and net N-fixation in sediments (Foster and Fulweiler 2014).

We studied Waquoit Bay to investigate the balance between three major sediment N cycling processes (N-fixation, denitrification, and DNRA). We hypothesized that (a) N2 production (defined as denitrification, including any anammox) and consumption (defined as N-fixation) occur simultaneously in sediments and (b) changes in oxygen and dissolved inorganic N (DIN) concentrations determine the net N2 flux. The objectives of the study were to: (1) measure N-fixation, potential denitrification, and potential DNRA directly in heterotrophic marine sediments; (2) determine the relative contributions of denitrification and N-fixation to the net N2 flux; and (3) understand the mechanisms controlling N-fixation and loss processes and how they are reflected in the total N2 signal.

Methods

Waquoit Bay is a small (3.34 km2), shallow, temperate estuary on the south shore of Cape Cod, Massachusetts. The bay is adjacent to a dense residential area, and N-loading to the bay doubled from 1930-1980 (Bowen and Valiela 2001). The total annual N load is calculated as 26,781 ± 9.6 kg N y−1 (Valiela et al. 2000). The Metoxit Point (MP) site is located in the main basin (41° 34′ 08.80″ N, 70° 31′ 05.13″ W) near the mouth of the Quashnet River, which carries more than a third of the total mean annual N load at 9879 ± 11.0 kg N y−1 (Valiela et al. 2000). Despite this high N load, Foster and Fulweiler (2014) reported an average net N-fixation at MP in 2011 of −21 ± 12 μmol N m−2 h−1. The MP site is shallow (∼2 m) with flocculent, organic-rich sediments (C/N ratio 8.6 ± 0.2, Foster and Fulweiler 2014). Thus, Waquoit Bay is an ideal model system to investigate the dynamics of N production and loss in coastal sediments.

Sediment cores were incubated over 5 days and samples were collected for dissolved gas and major nutrients. Sediment cores were collected in October 2012 with a HYPOX pole corer and check valve system designed to keep the sediment-water interface (SWI) and overlying water (OLW) intact (Gardner et al. 2009). Sediment cores (7.6 cm diameter) were incubated in a continuous-flow system as described previously (e.g., McCarthy et al. 2013). Briefly, cores and overlying water were capped with gas-tight plungers fitted with gas-tight PEEK inflow and outflow tubing at ∼3 cm above the SWI (OLW volume ∼136 mL). Unfiltered bottom water (collected from ∼15 cm above the SWI in the field) in gas-tight 25 L Tedlar bags was pumped through the OLW at 1–2 mL min−1 (OLW residence time ∼97 min).

Isotope additions were performed, and the cores were pre-incubated overnight (∼12 h) in an environmental chamber at in situ temperature (20 ± 2 °C) until sampling began the next morning. Three treatments were conducted in triplicate: (1) control (no isotope addition); (2) 5 μM 30N2 (99 %, Cambridge stable isotope laboratories) and (3) 4 μM 15NO3 − (K15NO3, 99 atom%, Cambridge Isotope Laboratories). Tracer addition concentrations were kept low to mimic in situ concentrations as closely as possible. However, nitrate concentrations were very low (<0.2 μM to undetectable in the OLW), so the tracer addition represents >95 % of the nitrate pool, and the denitrification and DNRA rates are therefore considered “potential.” To make stock solutions, 30N2 gas was added to deionized water in a 1 L Tedlar bag and shaken gently on a shaker table overnight; 20 mL of stock solution was added to the site water in 25 L Tedlar bags, and final 30N2 concentration was determined by membrane inlet mass spectrometry (MIMS). Cores were incubated under continuous-flow conditions for 120 h after amendments. Initial sampling (time 0) was completed ∼12 h after the tracer additions. Water samples were collected from the inflow reservoir and outflow ports at 0, 5, 12, 24, 48, 72, 96, and 120 h for dissolved gasses (28/29/30N2, argon, oxygen) and dissolved inorganic nutrients (ammonium, nitrate, nitrate + nitrite, ortho-phosphate) (Table 1).

Table 1 Geochemical fluxes for each timepoint for all treatments in μmol N m−2 h−1
Full size table

Dissolved gas samples were collected in 12 mL Exetainers (Labco, UK). The samples were filled from the bottom and overflowed by at least one volume before preserving with 25 μL of saturated ZnCl2. Dissolved gas samples and cores were inspected carefully for bubbles, because bubble formation can strip N2 if samples are not collected and handled properly, causing the N2/Ar ratio to decrease and appear as N-fixation (Eyre et al. 2002). Additionally, oxygen levels remained below saturation, and incubations were conducted in the dark. All nutrient samples were filtered immediately with a 0.2 μm nylon filter (GE Osmonics) and stored frozen. The Exetainer vials were stored under water at in situ temperature (±2 °C) until analysis. At 0, 24, 72, and 120 h, water samples were also collected for 15NH4 + analysis using high-performance liquid chromatography (HPLC; Gardner et al. 1995), and one core from each treatment was sacrificed and subcored with duplicate, sterile 10 mL cutoff syringes for 15Norganic and molecular analysis. At time 0, one core was sacrificed. Due to sacrificing cores, three cores were incubated through 24 h, but only two cores remained at 48 and 72 h and one at 96 and 120 h for each treatment. The subcores for 15Norganic isotope measurements were sectioned into 1 cm depths down to 3 cm and dried overnight at 60 °C.

Dissolved gasses (28N2, 29N2, 30N2, O2, and argon (Ar)) were measured on a quadrupole MIMS (Kana et al. 1994) in the Fulweiler Lab at Boston University. The MIMS was tested to determine if low-oxygen conditions would result in production of NO+ (mass 30; Eyre et al. 2002; Kana and Weiss 2004) by using serial 1 mL 20 mM sodium dithionite additions to room temperature water standards (Kana and Weiss 2004; Hartnett and Seitzinger 2003). No significant change in the 28N2/Ar was observed with decreasing O2, and no additional mass 30 molecules were produced allowing the samples to be processed on the MIMS without a furnace to remove oxygen. Samples were normalized to air-equilibrated deionized water standards (18 and 29 °C). Precision (coefficient of variation) is <0.5 % for N2 concentrations and <0.05 % for N2/Ar, and accuracy is <1.5 % for N2 concentrations and <0.5 % for N2/Ar for the MIMS (Kana et al. 1994). The MIMS system has similar sensitivities for the three N2 gas species (An et al. 2001). The standard error for the 30N2 signal of the standards was 0.017 %, and the detection limit (calculated as three standard deviations of the standard values above the baseline) was 0.0054 μM.

Potential denitrification rates were calculated from the sum of 28N2, 29N2, and 30N2 production and any calculated N-fixation in 15NO3 −-amended cores using a total isotope technique (An et al. 2001). The An et al. (2001) calculations assume that 14N and 15N bind randomly when forming N2 gas (i.e., that 28N2, 29N2, and 30N2 would show binomial distribution). Assuming the proportion of 15NO3 − is p and 14NO3 − is 1 − p, the N2 production rates are calculated for 28N2 as (1 − p)2, for 29N2 as 2p(1 − p), and for 30N2 as p 2. The An et al. 2001 equations use the 28N2 flux to predict the expected 29N2 and 30N2 values, and then compare the measured 29N2 and 30N2 values to the predicted values. Any differences are attributed to N-fixation, and the N-fixation rate is calculated. The potential denitrification rate is calculated from the sum of the total N2 flux added to the predicted N-fixation rate.

Potential DNRA was calculated from 15NH4 + production in the 15NO3 −-amended cores (An and Gardner 2002). Note that this method may overestimate actual DNRA because 15NH4 + production represents a potential rate from increased substrate, but simultaneously may underestimate DNRA from coupled nitrification/denitrification of in situ 14N (Gardner and McCarthy 2009). N-fixation was calculated from 28N2, 29N2, and 30N2 concentrations and fluxes in the 15NO3 −-amended cores (An et al. 2001) and from the sum of 28N2, 29N2, and 30N2 uptake minus the 30N2 uptake due to diffusion in the 30N2-amended cores. The 30N2 diffusion flux (F) was calculated from Fick’s Law (Hamersley and Howes 2005):

$$ F=\left(-\varPhi \times {D}_{\mathrm{a}}\times \varDelta C\right)/x $$

where Φ is the sediment porosity, D a is the apparent diffusion coefficient, ΔC is the change in concentration of N2 across the sediment, and x is the thickness of the sediment layer. The diffusion flux was calculated at each time point by estimating the diffusion depth (x) at time (t) (Crank 1975):

$$ x=\sqrt{\left({D}_{\mathrm{a}}\times t\right)} $$

Sediment porosity (0.83 ± 0.01) for MP was reported by Foster and Fulweiler (2014, 2015), and the apparent diffusion coefficient (7.7 to 8.2 × 10−6 cm−2 s−1) was reported by An and Joye (2001), using the Boudreau (1996) calculations, for estuarine sediments.

Ammonium released from mineralization was estimated from sediment oxygen demand (SOD; the difference between oxygen concentrations in the inflow and outflow normalized for flow rate and surface area), assuming approximately 1 mol O2 is required to respire each mole of ∼CH2O, and assuming a C/N ratio of ∼9:1 (Foster and Fulweiler 2014). We acknowledge that this assumption may be an overestimate, as some oxygen may be used in sulfur (re)oxidation.

Dissolved inorganic nutrients were measured on a Seal Auto Analyzer 3 with segmented flow injection using standard colorimetric techniques (Solorzano 1969; Grasshoff et al. 1999).

14NH4 and 15NH4 concentrations and isotope ratios were measured using HPLC in the Gardner Laboratory at the University of Texas Marine Science Institute (Gardner et al. 1995), and enriched values were ∼1000 % greater than natural abundance of 15NH4.

Sediment 15Norganic measurements were performed by the Boston University Stable Isotope Laboratory on a EuroVector Euro EA elemental analyzer (Owens 1988). The elemental analyzer detection limit was 0.2 μg (the samples were three orders of magnitude greater), precision was within 1.5 %, and the samples were up to 128 % enriched. Triplicate samples were analyzed from the top 3 cm of sediment. The flux of N2 fixed as organic matter was estimated from the mass of 15Norganic as:

$$ \mathrm{Flux}=\frac{{}^{15}{\mathrm{N}}_{\mathrm{organic}}\mathrm{mass}\times \left(\mathrm{sample}\ \mathrm{mass}/\mathrm{sediment}\ \mathrm{density}\right)\times \mathrm{S}\mathrm{A}}{\mathrm{time}} $$

where SA is the surface area of the core, and sample mass is the total mass of the dried sediment analyzed by the elemental analyzer. The 15Norganic production is an underestimate of total production, as only ∼1 % of 28N2 pool was labeled.

Dissolved gas and nutrient fluxes were calculated from concentrations:

$$ \left(\mathrm{Outflow}\hbox{--} \mathrm{inflow}\right)\times \mathrm{flow}\ \mathrm{rate}/\mathrm{core}\ \mathrm{surface}\ \mathrm{area}=\mathrm{flux} $$

Statistical Analyses were performed using Excel and JMP software. One-way analysis of variance (ANOVA) or group ANOVA tests were used to determine significant differences at a 95 % confidence level. p values are reported throughout the text.

RNA was extracted from three depths of sediment subcores (0–1 m, 1–2, 2–3 cm) to quantify functional genes for N-fixation (the nitrogenase (nifH)) and denitrification (the nitrite reductase (nirS)). Extractions were performed using the MoBio RNA PowerSoil Total RNA Isolation and RNA PowerSoil® DNA Elution Accessory Kits per the manufacturer’s specifications. DNA free (New England Biolabs) removed DNA from RNA samples, and samples were tested for contaminant DNA via gel visualization and PCR. Complementary DNA (cDNA) was synthesized using the SuperScript III FirstStrand Synthesis kit (New England Biolabs), and cDNA from the three depths were combined and quantified using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies). A 361-bp nifH fragment was amplified using the Poly et al. (2001) PolF (5′-TGC GAY CCS AAR GCB GAC TC-3′) and PolR primers (5′-ATS GCC ATC ATY TCR CCG GA-3′). A 256-bp nirS fragment was amplified using primers NirS1F (5′-CCT AYT GGC CGC CRC ART-3′) and NirS3R (5′-TTC CTB CAY GAC GGC GGC-3′) (Braker et al. 1998). All quantitative polymerase chain reaction (qPCR) standards were made as described previously (Newell et al. 2011) by cloning and purification of the PCR product followed by serial dilution of the cleaned plasmid. For nifH quantification, each 20 μL reaction included 10 μL Quantitect SYBR Green (Qiagen), 0.5 μL of each primer (10 μM), 8.5 μL reverse transcription-PCR-grade water (Ambion), and 1 μL template cDNA (100 ng). For nirS quantification, each 20 μL reaction included 10 μL Quantitect SYBR Green (Qiagen), 0.2 μL of each primer (100 μM), 9.1 μL RT-PCR-grade water (Ambion), and 0.5 μL template cDNA (50 ng). Quantitative PCR was performed on an ABI 7900ht real-time qPCR machine. The nifH thermal profile consisted of 95 °C for 10 min, followed by 40 cycles of denaturation at 95 °C for 15 s, annealing at 60 °C for 30 s, and elongation at 72 °C for 30 s (Fan Lu 2013). The nirS thermal profile began with 15 min at 95 °C, then 40 cycles of 94 °C for 15 s, 55 °C for 30 s, and 72 °C for 30 s. After amplification, a melting curve was obtained by heating the products to 95 °C, cooling them to 65 °C, and then gradually heating them to 95 °C at a rate of 0.2 °C s−1. The specificity of the qPCR products was confirmed by melting curve analysis. Each assay was carried out within a single assay plate to permit direct comparison of copy numbers between environmental DNA samples. Each assay included triplicate, no template control (NTC) wells, triplicates for each of the five standards, and triplicate environmental samples. Automatic analysis settings were used to determine the threshold cycle (Ct) values. The efficiencies for each standard curve were within the acceptable range of 90–110 %, and R 2 values were above 0.99. The gene copy numbers were calculated according to:

$$ \mathrm{Gene}\ \mathrm{copy}\ \mathrm{number}=\left(\mathrm{ng}\times \mathrm{number}\ {\mathrm{mol}}^{-1}\right)/\left(\mathrm{bp}\times \mathrm{ng}\ {\mathrm{g}}^{-1}\times \mathrm{g}\ {\mathrm{mol}}^{-1}\mathrm{of}\ \mathrm{bp}\right) $$

and then converted to gene copy number per gram of sediment.

Results

Control and 15NO3 − cores exhibited decreasing net 28N2 uptake through 72 h and then switched to an increasing net 28N2 efflux (Fig. 1). Significant 30N2 production (greater than the control, one-way ANOVA, p = 0.014, n = 6) in the 15NO3 − cores began at 24 h, but a large, positive net 28N2 flux was not seen until 120 h (Fig. 2). Calculated potential denitrification rates (An et al. 2001) were low (<7 μmol N m−2 h−1) until 120 h, when the rate peaked at 40.5 μmol N m−2 h−1 (Fig. 3). nirS expression remained constant in the 15NO3 − cores until between 72 and 120 h, when it increased by four orders of magnitude (Fig. 4). In the control cores, nirS expression decreased from 0 to 72 h, then increased by two orders of magnitude from 72 to 120 h.

Fig. 1
figure 1

Net N2 fluxes in control (a), 15NO3 − (b), and 30N2 cores (c). The 28N2 flux is shown in black (left y-axis) and the 30N2 flux is shown in gray (right y-axis). All fluxes are in μmol N m−2 h−1. Note the difference in the y-axis scales. Error bars depict the standard error

Full size image
Fig. 2
figure 2

N-fixation rates, as determined by directly observed fluxes (the sum of 28N2, 29N2, and 30N2 uptake minus the 30N2 uptake due to diffusion) from 30N2-amended cores (gray) and calculated from 28N2, 29N2, and 30N2 fluxes in 15NO3 −-amended cores using An et al. (2001) equations (black). Error bars depict the standard error

Full size image
Fig. 3
figure 3

Denitrification rates (as the sum of 28N2, 29N2, 30N2 production and calculated N-fixation, An et al. 2001) in the 15NO3 − cores over time. Error bars depict the standard error

Full size image
Fig. 4
figure 4

Quantification of key functional gene transcripts for denitrification (nirS) and N-fixation (nifH) using RT-qPCR. Copies are per 100 ng template, nirS copies are shown in black, and nifH copies are shown in gray. Error bars depict the standard error

Full size image

Potential DNRA was calculated from production of 15NH4 + in the 15NO3 − cores. DNRA was observed as production and efflux of 15NH4 + from 15NO3 at 24 h (average ± SE, 5.7 ± 4.8 μmol N m−2 h−1) but not at 72 or 120 h (Table 2). The ammonium efflux from sediments was greater in the 15NO3 − cores than the control cores at 0 and 120 h (but not 24 h) (Table 1).

Table 2 15NH4 + flux (μmol N m−2 h−1) mean and standard error at times 0 and 24 h
Full size table

The 30N2 cores showed net 28N2 and 30N2 uptake over 120 h (Fig. 1). 30N2 uptake could not be explained by diffusion from the OLW to the sediments alone. The diffusion estimates for 30N2 uptake from Fick’s law ranged were on average 25 % of observed 30N2 uptake values (which ranged from −15.0 ± 2.3 to −52.6 ± 1.7 μmol N m−2 h−1) over time for diffusion into the entire core. The ratio of 30N2/28N2 uptake was lower than predicted by diffusion from their relative concentrations, but the net 28N2 flux is a balance between denitrification and N-fixation. 15NH4 + effluxed from the sediments in the 30N2 cores through at least the first 24 h of sampling (Table 2) but was not detected after 72 h. 15N-biomass in the sediments from the 30N2 cores was detected at 72 and 120 h (0.83 ± 0.47 and 0.55 nmol N m−2 h−1, respectively). nifH functional gene expression was detected in all cores of all treatments over the course of the incubation and were maintained at a consistent level (Fig. 4).

N-fixation also was calculated from the sum of 28N2, 29N2, and 30N2 uptake in the 30N2 cores (Fig. 2), although 29N2 fluxes showed no discernible patterns in any treatment, and most rates were not significantly different from zero. 28N2 and 30N2 fluxes in the 30N2 cores were negative at every time point. 28N2 uptake was greatest at the initial sampling time (time “0” = −124 ± 3.1 μmol N m−2 h−1) and then declined over the next 24 h. The 28N2 fluxes had more uptake in the 30N2-amended cores than the control or 15NO3 −-amended cores (ANOVA, p < 0.05) during the first 48 h. N-fixation was also calculated from 28N2, 29N2, and 30N2 concentrations and fluxes in the 15NO3 cores according to An et al. (2001). The calculated rates declined over the first 72 h, but dropped to zero when the net N2 signal was positive (Fig. 2).

Discussion

The control and 15NO3 − cores exhibited decreasing net 28N2 uptake through 72 h and then switched to an increasing net 28N2 efflux (Fig. 1), indicating that the system shifted from net N-fixation to net denitrification over the course of the 5-day incubation. This pattern may be explained in part by increased denitrification rates in response to nitrate inputs and oxygen decreases.

Nitrate Loss

DNRA was observed as production and efflux of 15NH4 + from 15NO3 at 24 h (average ± SE, 5.7 ± 4.8 μmol N m−2 h−1), but not at 72 or 120 h (Table 2). Large 14NH4 + efflux from all treatments suggested that some fraction of the tightly recycled in situ nitrate may be reduced via DNRA as well. The NH4 + efflux from sediments was greater in the 15NO3 − cores than the control cores at 0 and 120 h (but not 24 h), but the differences were much greater than can be accounted for by 15NO3 − reduction alone. However, no 15NH4 + flux was observed after 24 h. Higher total 14NH4 + flux in 15NO3 cores suggests that cation exchange mechanisms (Seitzinger et al. 1991; Gardner and St. John 1991) may be masking 15NH4 + release from sediments (as 15NH4 + replaces 14NH4 + between clay sheets). Regardless, DNRA is likely underestimated here, as the only definitive measurement is production of 15NH4 +, and the relative fraction of 14NH4 + flux from DNRA is unknown.

Denitrification was calculated using a total isotope approach (An et al. 2001), as the sum of 28N2, 29N2, and 30N2 production and calculated N-fixation. Calculated potential denitrification rates (An et al. 2001) peaked at 40.5 μmol N m−2 h−1 (Fig. 3), coinciding with the four orders of magnitude increase in nirS transcript abundance at 120 h (Fig. 4). Similarly, in the control cores, nirS transcript abundance increased by two orders of magnitude from 72 to 120 h, coinciding with the shift from net N2 uptake to efflux. There was a positive, linear correlation between the nirS transcript abundance and the potential denitrification rates (R 2 = 0.99, p = 0.003, 98 % confidence). Thus, denitrification increased rapidly over the last 48 h of the incubation. The peak potential rate of 40.5 μmol N m−2 h−1 falls within the denitrification rates previously measured for this system at other sites (October 1992, 190 ± 59 μmol N m−2 h−1; October 1993, 38 ± 0.5 μmol N m−2 h−1) reported by LaMontagne et al. (2002) using a different analytical method of measuring the N2 flux (a gas chromatograph connected to a modified gas stripping device). Both denitrification and N-fixation rates are similar to the N2 flux range measured in 2011 from nine cores at MP (−50.5 to 38.5 μmol N m−2 h−1, Foster and Fulweiler 2014).

From 0 to 24 h, sediment nitrate uptake was nearly balanced by DNRA and denitrification. However, at 120 h, potential denitrification was 33 % lower than the measured nitrate uptake (Fig. 6), and no DNRA was observed in the 15NO3 − cores. These results suggest that the potential denitrification rates may be 1.5× higher than the total isotope calculations estimate. We hypothesize that this difference is due to these calculations (An et al. 2001) underestimating the co-occurring N-fixation, as the denitrification rate is ultimately calculated by adding the N-fixation rate to the total N2 flux; therefore, if N-fixation is underestimated, so is denitrification.

Recent work in estuaries in Massachusetts (Koop-Jakobsen and Giblin 2010), Texas (Gardner et al. 2006) and Florida (Gardner and McCarthy 2009) showed that DNRA and denitrification can co-occur at similar rates. Additionally, a review by Giblin et al. (2013) reported that DNRA accounted for more than 30 % of the nitrate reduction at 26 of 55 coastal sites. The balance between these processes is not well understood, although DNRA and SOD were positively correlated, and denitrification was inversely correlated with both processes, in Florida Bay (Gardner and McCarthy 2009). An earlier study in Baffin Bay, TX (An and Gardner 2002) showed much higher rates of DNRA than denitrification, possibly due to the high sulfide concentrations in the sediments (which may inhibit denitrification, but enhance DNRA; Brunet and Garcia-Gil 1996).

Nitrate is tightly recycled in this system, as the in situ nitrate concentration in this study remained very low (<0.2 μM) or not detectable (Table 1). We observed DNRA in the first 24 h, but very low denitrification rates. Waquoit Bay also has very high organic carbon (620 μM dissolved organic carbon in Childs River groundwater, Hopkinson et al. 1998), supporting Tiedje’s (1988) hypothesis that DNRA is more important in high-carbon, low-nitrate systems (Burgin and Hamilton 2007). While nirS expression shows denitrifying bacteria were present and actively expressing the nitrite reductase gene throughout the incubation, these facultative anaerobes use oxygen preferentially when it is plentiful. DNRA was not observed after 24 h (although it may have been masked), while 30N2 production from denitrification increased. There was a strong correlation between 30N2 production in the 15NO3 − cores and oxygen decline (R 2 = 0.69, p < 0.0001, 98 % confidence). While no nitrate was added to the control cores, we observed a similar ∼40 μM decrease in OLW oxygen concentration during the incubation (Fig. 5). As oxygen declined, both the control and the 15NO3 − cores shifted to net denitrification (N2 efflux) by 96 h (oxygen in the 30N2 cores was generally greater than but not significantly different from the control cores through the first 48 h, Fig. 5). Denitrifiers may compete more effectively with DNRA for nitrate when low-oxygen favors nitrate respiration (Körner and Zumft 1989).

Fig. 5
figure 5

Oxygen (O2) concentrations (μM) from the outflow lines throughout the incubation. Oxygen in the control cores is shown as black, filled diamonds, in 15NO3 −-amended cores as black, open squares, and in 30N2-amended cores as gray triangles. Oxygen concentrations in the inflow did not decrease below 90 % of saturation. Error bars depict the standard error

Full size image

Anammox typically accounts for <6 % of the total N2 loss from shallow (<20 m) systems (Dalsgaard et al. 2005). Our results suggest that anammox was not a significant pathway in this study, as no 29N2 (e.g., significantly different from the control cores, p = 0.61, ANOVA) was produced in the 15NO3 − cores, even though >1 μM ambient NH4 + was available. Likewise, anammox contributed ≤4 % of the total N2 production in nearby Narragansett Bay, RI (Brin et al. 2014). However, while anammox bacteria are capable of reducing nitrate to nitrite reduction (Kartal et al. 2007), they ultimately can provide only ∼10 % of the NO2 − needed to oxidize ammonium.

N-fixation

Sediment N-fixation occurred throughout the experiment in the 30N2-amended cores. Multiple lines of direct evidence support this conclusion: (1) 15NH4 + effluxed from the sediments in the 30N2 cores within at least the first 24 h of sampling (Table 2), (2) the 30N2 cores showed net 28N2 and 30N2 uptake (greater than diffusion) within 120 h (Fig. 1), (3) 15N-organic matter in the sediments from the 30N2 cores was detected at 72 and 120 h, and (4) nifH functional gene expression was detected in all cores of all treatments over the course of the incubation, indicating that N-fixers were both present and expressing nitrogenase genes (Fig. 4). N-fixation occurred at similar rates at the same station in Waquoit Bay in 2011 (Foster and Fulweiler 2014). The nifH-containing community was characterized by Newell et al. (2011) and dominated by δ-proteobacteria. N-fixation was determined from the sum of 28N2, 29N2, and 30N2 uptake minus the 30N2 uptake due to diffusion (~25 %) in the 30N2 cores (Fig. 2) and also calculated from 28N2, 29N2, and 30N2 concentrations and fluxes in the 15NO3 cores (after An et al. 2001). We assume that N2 uptake is attributed to sediment N-fixation and diffusion. However, although there are no reported values of N-fixation in the Waquoit Bay water column, work by Bentzon-Tilia et al. (2015) has highlighted the potential importance of simultaneous N-fixation by photoheterotrophic and heterotrophic bacteria and cyanobacteria in estuarine water columns. Although the cores were incubated in the dark and the OLW remained aerobic, we cannot rule out the possibility that some of the N2 uptake could be attributed to water column N-fixation.

The 30N2 uptake rates were adjusted to account for diffusion within the core. This estimate may be a conservative estimate if meiofauna bioturbation speeds up diffusion. However, even in the extreme scenario that diffusion accounted for 100 % of the 30N2 uptake (which is unlikely due to the observation of 15N-organic matter and 15NH4 + production), the 28N2 flux in those cores accounted for 80 % of the measured N-fixation rate. These calculated rates followed the same pattern initially as the direct rates: N-fixation was greatest at the initial sampling time (time “0” h, −83.4 ± 19.1 μmol N m−2 h−1) and then decreased, but the calculated rates were not detected by 96 h (Fig. 2). The N-fixation rates measured directly from the 30N2-amended cores were significantly (group ANOVA, p = 0.00012, and paired t tests, p < 0.0001) different from and approximately double the An et al. 2001 calculated rates from the 15NO3 − cores. The calculated rates dropped to zero when the net N2 signal was positive, despite their design to estimate simultaneous N-fixation and denitrification.

In situ dissolved inorganic N concentrations in the OLW were not high enough to inhibit N-fixation. Heterotrophic N-fixers, such as sulfate-reducing bacteria, can have a much higher threshold of ammonium concentrations required to (indirectly) repress expression of the nif operon than cyanobacteria (Knapp 2012). Indeed, sulfur and sulfate reducers have been linked to sediment N-fixation in Narragansett Bay (Brown and Jenkins 2014; Fulweiler et al. 2013). In marine environments, N-fixation continued to occur in the presence of up to 30 μM NO3 − and/or 200 μM NH4 + (Knapp 2012). Millimolar concentrations of ammonium are required to shut off N-fixation by sulfate reducers (Welsh 2000), and 100 μM ammonium only slightly impairs N-fixation in Pseudomonas stutzeri A1015 (Desnoues et al. 2003). Pore water ammonium was not measured in this study, but the OLW concentration never exceeded 10 μM; thus, we believe that N-fixation was not inhibited by DIN concentrations in these incubations.

Despite the presence of ammonium, sediment conditions may favor N-fixation. Low-oxygen conditions are conducive to the obligatory anaerobic N-fixation pathway, and each N2 fixed requires 8H+ and 8e-, and sediments have a high redox potential. In fact, enhanced shuttling of protons and electrons through the nitrogenase may be a response to strongly reducing conditions in sediments.

Coastal Nitrogen Budget

Coastal ecosystems inherently vary over time and space. Physical and geochemical conditions fluctuate daily with light and tides, as salt water exchanges with freshwater, and seasonally with temperature and river input (among other factors). The results presented here and in other studies (Table 3) indicate that sediment N2 fluxes respond to changing conditions dynamically as the balance of denitrification and N-fixation is influenced by the supply of labile organic carbon, nitrate, oxygen, etc. (Burgin and Hamilton 2007). N2 fluxes also vary on daily and yearly timescales. For example, the N dynamics in our core system shifted from net N-fixation to net denitrification over the course of the 120-h incubation. This may be in response to changing O2 concentrations in the OLW (Fig. 5), which mimics the diel change in in situ oxygen, which declines (sometimes into hypoxia) at night in the summer. In Narragansett Bay, RI, interannual shifts occurred in net sediment N2 flux, oscillating between average net denitrification and net N-fixation (Fulweiler and Heiss 2014). In Weeks Bay, Alabama, a year-long study showed a sinusoidal net N2 flux, with N-fixation or denitrification as the dominant process varying on a monthly scale (Mortazavi et al. 2012). Individual sampling efforts produce snapshots, but longer-term sampling efforts across all seasons are required to determine how net N2 fluxes fluctuate temporally.

Table 3 Comparison of sediment core incubation studies reporting both N-fixation (N2 uptake) and denitrification (N2 production)
Full size table

Many recent studies indicate that N-fixation rates in coastal sediments can equal or exceed N2 loss rates at times. Using whole-core incubations and direct measurements of dissolved gasses, sediment N-fixation rates similar to or exceeding denitrification have been reported in temperate and tropical estuaries, wetlands, and near-shore marine ecosystems (see Table 3), although denitrification rates may be higher averaged over time. Likewise, N-fixation and denitrification co-occurred within the same order of magnitude in our experiments. This co-occurrence of N-fixation and denitrification was supported by simultaneous measurement of nifH and nirS expression, and the control and 15NO3 cores showed an increase in nirS corresponding with the shift in N2 influx to efflux (15NO3, R 2 = 0.99, p = 0.003, 98 % confidence). These results are similar to those reported in a large mesocosm using Narragansett Bay sediments where both nifH and nirS expression were detected concurrently despite overall net denitrification rate measurements (Fulweiler et al. 2013).

Balancing the N budget in aquatic systems is desirable for nutrient management and modeling efforts. Here, estimated ammonium production from remineralization ranged from 103 to 180 μmol N m−2 h−1 from 24 to 120 h in the 15NO3 − and 30N2 treatments (Fig. 6). This rate is insufficient to balance the ammonium efflux measured in the OLW (328 ± 24 to 345 ± 46 μmol N m−2 h−1 at 24 h and 276 to 429 μmol N m−2 h−1 at 120 h). Ammonium production approaches the measured ammonium efflux when N-fixation rates from the 30N2 cores (106 ± 13 μmol N m−2 h−1 at 24 h and 122 μmol N m−2 h−1 at 120 h) are included.

Fig. 6
figure 6

N budget at the sediment-water interface at a 24 and b 120 h. Rates are reported in μmol N m−2 h−1. Directly measured rates are shown in white: NH4 + and NO3 − fluxes, N-fixation, DNRA, 15N-assimilation to organic matter (OM), and calculated rates are shown in black: denitrification and N-fixation (An et al. 2001) and NH4 + remineralized from oxic respiration (calculated from SOD). Blue represents the OLW and brown the sediment. Nitrification (gray) was not measured, although significant rates are not required to balance the budget. (30N2) or (15NO3 −) following a rate indicates which core treatment the value was calculated/reported from

Full size image

N-fixation can also explain underestimated denitrification rates. Nitrate concentrations were very low (<0.2 μM) to undetectable in the OLW. When ∼5 μM nitrate was added to the inflowing water, nitrate concentrations in the outflow decreased over time (from 3.2 ± 0.8 μM at 12 h to 0.61 ± 0.05 μM at 120 h, Table 1). At 120 h, nitrate uptake flux was −64 μmol m−2 h−1, but DNRA was not observed, and the highest measured potential denitrification rate was 40 μmol N m−2 h−1. This result leaves approximately 24 μmol of N unaccounted for and may have underestimated the denitrification rate. The potential denitrification rate is calculated based on the sum of the 28N2, 29N2, and 30N2 fluxes and the calculated N-fixation rate, which may be underestimated by the same calculations, and cause an underestimate of denitrification. Thus, direct N-fixation estimates in sediment N budgets should be included to estimate a N mass balance accurately.

N inputs for a coastal N budget typically include river discharge, sewage, groundwater, urban runoff, atmospheric deposition, and pelagic N-fixation, while outputs include N loss via denitrification and anammox, or export to the ocean (Billen et al. 1985; Nixon et al. 1995). When sediment N-fixation is included, it usually refers to N-fixation in sediments in seagrass beds (Eyre and McKee 2002; Wulff et al. 2011), and those rates can be quite high, exceeding 410 μmol N m−2 h−1 (Eyre and McKee 2002). N-fixation in bare sediments is often excluded as an input, given the low rates measured indirectly (e.g., acetylene reduction) in the 1970s and 1980s (Howarth et al. 1988) or because it was not clearly required to balance the budget (Nixon et al. 1995). However, more recent rates reported in the literature and here suggest that sediment N-fixation may be on par with or a large percentage of N2 loss pathways when averaged over time (Table 3). When the net N2 flux varies between net N-fixation and denitrification, the ability of an estuary to mitigate excess nitrogen loading can vary greatly on an annual scale, making it critical to accurately measure the rates of both processes, as well as the N-loading. In Narragansett Bay, a nine year study showed that in some years, up to 25 % of the incoming N is removed through denitrification, while in other years, up to 30 % of the N is added through N-fixation (Fulweiler and Heiss 2014). The mean and maximum rates for N-fixation and potential denitrification measured in this paper (extrapolated to kg N ha−1 year−1 over the 3.34 km2 of Waquoit Bay) are the same order of magnitude as the total N load measured by Valiela et al. (2000) of 26,781 ± 9.6 kg N year−1/N-fixation could be 48,000 kg N year−1 (mean) to 75,000 kg N year−1 (maximum) and denitrification could be 1500 kg N year−1 (mean) to 16,000 kg N year−1 (maximum). While the fluctuation in daily rates (from undetected to maximum rates) and the seasonal variability reported in Foster and Fulweiler (2014) suggested that the annual loss or load from denitrification and N-fixation are much lower than these maximum rates, these numbers emphasize both the potential contribution of N-fixation to an already eutrophic estuary, as well as the importance of the balance between these processes.

With a preponderance of evidence indicating dynamic sediment N2 fluxes, constraining coastal marine N budgets may require a more thorough evaluation of N2 flux pathways. Net sediment N2 flux measurements throughout the year and over multiple years are critical to determining the balance of N2 production and consumption. Quantifying production and consumption individually may be necessary to model that balance over time. This study highlights the need to evaluate both N-fixation and denitrification, either using net N2 fluxes or direct measurements of opposing pathways, to accurately model the coastal marine N budget.