Coupled molecular and isotopic evidence for denitrifier controls over terrestrial nitrogen availability

Abstract

Denitrification removes ecologically available nitrogen (N) from the biosphere and influences both the pace and magnitude of global climate change. Disagreements exist over the degree to which this microbial process influences N-availability patterns across Earth's ecosystems. We combine natural stable isotope methods with qPCR to investigate how denitrifier gene abundance is related to variations in nitrate (NO₃⁻) pool sizes across diverse terrestrial biomes and conditions. We analyze NO₃⁻ isotope composition (¹⁵N/¹⁴N, ¹⁸O/¹⁶O) and denitrifier gene nirS in 52 soil samples from different California ecosystems, spanning desert, chaparral, oak-woodland/savanna and forest. δ¹⁵N-NO₃⁻ correlates positively with δ¹⁸O-NO₃⁻ (P⩽0.03) and nirS abundance (P=0.00002) across sites, revealing the widespread importance of isotopic discrimination by soil denitrifiers. Furthermore, NO₃⁻ concentrations correlate negatively to nirS (P=0.002) and δ¹⁵N-NO₃⁻ (P=0.003) across sites. We also observe these spatial relationships in short-term (7-day), in situ soil-incubation experiments; NO₃⁻-depletion strongly corresponds with increased nirS, nirS/16 rRNA, and enrichment of heavy NO₃⁻ isotopes over time. Overall, these findings suggest that microbial denitrification can consume plant-available NO₃⁻ to low levels at multiple time scales, contributing to N-limitation patterns across sites, particularly in moist, carbon-rich soils. Furthermore, our study provides a new approach for understanding the relationships between microbial gene abundance and terrestrial ecosystem functioning.

Introduction

Microbial denitrification converts inorganic N (NO₃⁻ → NO₂⁻) to gaseous N forms (NO→N₂O→N₂), accounting for 26-40% of N outputs from natural terrestrial ecosystems worldwide (Houlton and Bai, 2009; Ciais et al., 2013). Such gaseous N losses directly modulate Earth's climate system via the production of the potent greenhouse gas N₂O, and, in N-limited ecosystems, denitrification has potential to restrict the availability of this nutrient for vegetation CO₂ uptake and carbon (C) sequestration. Denitrification is arguably the most poorly understood process in the terrestrial N cycle, posing a fundamental challenge to the fields of ecosystem- and microbial- ecology alike (Williams et al., 1992; Beckman and Koppenol, 1996; Davidson and Seitzinger, 2006).

Two basic problems have limited our understanding of terrestrial denitrification. The first is systemic to ecology if not all of science—quantifying the relationship between pattern and scale (Levin, 1992). Soil microbial processes are highly dynamic and variable in space and time; yet they control many different aspects of local ecological process and global biogeochemical cycles (Falkowski et al., 2008). In the case of denitrification, microbial ecologists and biogeochemists have pointed to variations in moisture, C and N availability controlling soil denitrification rates within and among global ecosystems (Bai et al., 2012; Groffman, 2012). The second challenge centers on the composition of the air: N₂ is the dominant product of terrestrial denitrifiers, but there are challenges with measuring N₂ from natural terrestrial denitrification directly. The Earth's atmosphere is ~78% N₂ by volume, thus posing a fundamental ‘signal to noise' barrier to measuring denitrification fluxes in the field. Studies using ¹⁵N tracers offer short-term snapshots of denitrification by tracking ¹⁵N-enriched NO₃⁻ into ¹⁵N₂ and ¹⁵N₂O; however, this approach is limited in naturally low N environments and is highly sensitive to the O₂ levels used during course of incubation experiments (Kulkarni et al., 2014). Extrapolating results from local soil experiments to entire ecosystems thereby represents a perennial challenge in terrestrial denitrification research (Davidson and Seitzinger, 2006).

Here, we combine measures of natural N and oxygen (O) stable isotope composition—which can reflect broader integrated patterns in ecosystem N-cycling (Robinson, 2001)—with quantitative real-time qPCR to advance our understanding of terrestrial denitrification from gene to ecosystem scales (Figure 1). As modern genetic tools have become mainstream, researchers have begun to explore the relationships between microbial community structure and ecosystem functioning (for example, Torsvik and Øvreås, 2002; Frey et al., 2004). Microbial diversity has been measured alongside such important steps in the N cycle as mineralization and nitrification, providing novel insights to the biogeochemical dynamics of ecosystems (Bardgett et al., 1999; Calderon et al., 2001; Jackson et al., 2003). Denitrification studies have similarly identified many microorganisms and enzymes involved in NO₃⁻ reduction (Knowles, 1982; Hochstein et al., 1988; Zumft, 1997). The nir (Nitrite Reductase encoding) genes are of particular interest, because they mark the crucial first gas-formation step in the process (Hochstein et al., 1988)—the point at which denitrification becomes a gaseous N loss from terrestrial environments.

Figure 1.

Detecting terrestrial denitrification (Concept). Microbial genes (small scale) can be considered alongside natural stable isotopic signature patterns in soil (large scale) for denitrification at the nitrite (NO₂⁻) reduction step. This first gas-formation step marks the point at which denitrification becomes a terrestrial N output.

We use qPCR techniques to quantify the relative abundance of nitrite reductase-encoding genes nirS and nirK (Hochstein et al., 1988; Zumft, 1997). These genes are functionally equivalent but structurally different—nirS encoding for heme (cytochrome cd1)-containing enzyme NirS and nirK encoding for copper-containing enzyme NirK or Cu-NIR (Hochstein et al., 1988; Zumft, 1997). Generally, denitrifiers possess either nirS or nirK rather than both (Heylen et al., 2006); a recent study found 10 rare organisms that contradicted this general observation, although the authors did not examine whether both genes were functionally active (Graf et al., 2014). We focus primarily on nirS given its widespread abundance across soil types, as supported by previous field and cultured strain studies (Braker et al., 1998; Throback et al., 2004). Furthermore, PCR primers for nirS capture most of the known diversity of organisms possessing the gene, whereas, due to nirK sequence divergence and taxonomic diversity, PCR primer coverage for nirK is more limited (Helen et al., 2016). Nir genes have been linked to denitrification rates in soil incubation experiments across different environmental conditions (Smith and Tiedje, 1979; Patra et al., 2005; Attard et al., 2011). However, whether an organism possesses such genes does not necessarily indicate whether they will be expressed and active at a given point in time, with little field-based inquiry into denitrifier gene-abundance vs function relationships.

Natural stable isotope composition offers a complementary, non-disruptive and integrative tool for investigating soil-denitrification across different spatial and temporal scales (Robinson, 2001). Denitrifiers preferentially consume the lighter, more abundant ¹⁴N (99.7% of all N) isotope at the expense of the heavier isotope ¹⁵N (0.3% of all N), thereby enriching the ¹⁵N/¹⁴N of NO₃⁻ substrates vs gaseous N products (Mariotti et al., 1981). Over time, such isotopic discrimination (‘fractionation') elevates the ¹⁵N/¹⁴N of soil nitrate pools relative to the air (Houlton et al., 2006), and ultimately, ecosystem to global-scale isotopic patterns in the terrestrial biosphere (Houlton and Bai, 2009; Mnich and Houlton, 2015). Other microbial processes could affect isotope signatures in field: nitrification has the potential to lower the ¹⁵N/¹⁴N of NO₃⁻; and heterotrophic NO₃⁻ consumption and other microbial processes (for example, in Dissimilatory Nitrate Reduction to Ammonium (DNRA)) could elevate the ¹⁵N/¹⁴N of NO₃⁻ substrates similar to denitrification (Fang et al., 2015). Denitrifiers have been shown to increase the δ¹⁵N and δ¹⁸O of NO₃⁻ within a small range of characteristic slopes (Lehmann et al., 2003; Houlton et al., 2006; Granger et al., 2008), allowing for dual-isotopic evaluation of denitrification's effect on ecosystem NO₃⁻ availability (Fang et al., 2015). Natural stable isotope budgets have implied utility of forecasting the effect of the terrestrial N cycle on global climate change (Houlton et al., 2015; Zhu and Riley, 2015); however, the importance of microbial-scale denitrification in driving larger ¹⁵N/¹⁴N patterns in ecosystems remains unclear. A particularly open question involves the extent to which denitrification expresses itself on N pools, given the potential for microsite consumption to eliminate the intrinsic isotope effect of denitrifiers on NO₃⁻ isotopes from local to ecosystem scales (for example, Houlton et al., 2006).

In this study, we ask whether the abundances of soil denitrifier genes are linked to natural stable isotope variations, and we ask whether these cross-scale data can clarify the role of microbial denitrification in regulating patterns of terrestrial NO₃⁻ availability. Specifically, we combine qPCR-derived nirS gene abundances with N and O isotopes of NO₃⁻ to investigate soil denitrification across redwood forest, oak woodland, chaparral and desert biomes. We complement these field observations with a series of in-situ incubation experiments, further isolating the temporal mechanisms behind the cross-system relationships examined. We use these approaches to test two overarching hypotheses.

The first hypothesis is that there is a significant positive correlation between nirS abundance and microbial consumption of lighter N and O isotopes. If supported, this hypothesis would point to a significant link between microbial denitrification at molecular scales and patterns of ecosystem isotopic pools at larger ones. In the absence of such a significant correlation, we would infer that there is no measurable relationship between the abundance of denitrifying genes and isotopic expression. This alternative would point to other processes driving natural isotope abundance patterns.

The second hypothesis is that there is a significant negative correlation among nirS/nitrate isotope composition patterns and soil NO₃⁻ concentrations. This finding would imply that denitrification plays an active role in controlling patterns of NO₃⁻ availability across ecosystems. However, if denitrification microbial gene abundances and isotopes increase only when NO₃⁻ is abundant, then we would infer the opposite—that denitrification increases in concert with NO₃⁻ abundance. If the gene abundances and isotopes do not correlate with NO₃⁻ levels, we would infer no directional effect either way.

Materials and methods

Study sites

We examined our overarching hypothesis and set of inferences across a matrix of terrestrial sites that occur over a short geographic domain in California (Table 1). Our diverse biome sites range only slightly in latitude (39^o 22' N to 34^o 47' N) yet represent substantial changes in water availability (Mean Annual Precipitation, MAP, ranged from 1103 to 292 mm across sites) and temperature (−3.9 to 12.0 °C minimum, 33.0–37.8 °C maximum) (Table 1).

Table 1. Description of field sites.

Biome	Sites	Latitude, longitude	Meana annual precip. (mm)	Max., min. air temp. (°C)	Meanaair temp. (°C)	Soil description	% Carbon by massa, top 0–10 cm (mean ±s.d.)	% Moisture by mass, top 0–10 cm (mean ±s.d.)	Incubation dates (m/d/y)	Weather during 7-day incubation
										Total precip. (mm)	Max., min. air temp. (°C)	Mean air temp. (°C)
Redwood forest	Jackson Demonstration State Forest's Caspar Creek watersheds: KJEb, M	39o 22' N, 123o 42' W; elev. 146–216 m	1103	37.8, −3.9	10.8	Ultic Hapludalfs & Typic Haplohumults from weathered sandstone and shale. Well-drained. pH 6.07–6.25 (Dahlgren, 1998)	8.22 ±0.17 control 7.15 ±0.23 clearcut	28.1 ±3.3 control 28.2 ±4.1 clearcut	11/8/14–11/15/14	14.2	17.0, 8.4	13.0
Oak woodland	UC Hopland Research & Extension Center (HREC) watersheds: Foster (2 sites), Foster Biological (2 sites)c,d	39o 01' N, 123o 05' W; elev. 667 m	1016	33.3, 12.0	14.95	Typic Haploxeralfs and Ultic Argixerolls from weathered graywacke, shale, siltstone or sandstone. Well-drained. Loam texture (40% sand, 35% silt, 25% clay) (Dahlgren et al., 2003). pH 6.19–7.16	3.29 ±0.05 control 2.05 ±0.06 burn only 4.71 ±0.50 burn and graze 4.51 ±0.09 graze only	22.3 ±6.6 control 13.5 ±3.3 burn only	No incubation	NA	NA	NA
Chaparral	HREC watersheds: James III (2 sites)b	39o 00' N, 123o 05' W	1016	33.3, 12.0	14.95	Typic Argixerollsfrom sandstone, shale and metasedimentary rock (Narvaez, 2007)	0.98 ±0.11 control 1.58 ±0.02 burn	12.7 ±3.9 control 12.9 ±5.5 burn	11/9/14–11/16/14	22.1	20.0, 7.8	13.9
Desert	Sweeney Granite Mountains Desert Research Center: under shrub, in interstitial spaces	34o 47' N, 115o 39' W; elev. 1252 m	292	33.0, −1.0	17.0	Granite-based, weakly developed, coarse. pH 7.0	0.90 ±0.12 under shrub 0.34 ±0.00 interstitial	6.6±9.5 under shrub 1.8±1.0 interstitial	2/2/14–2/9/14	0	13.3, 3.9	8.3

^aMean annual precipitation and air temperature data are long-term normals (1981–2010) from NOAA climate stations in Fort Bragg, Ukiah and Mitchell Caverns, CA, USA. Carbon data from UC Davis Stable Isotope Facility analysis of bulk N and C isotope data from dry soil with Micro Cube elemental analyzer interfaced to a PDZ Europa 20-20 Isotope-Ratio Mass Spectrometer.

^bClearcut disturbances occurred in the KJE redwood site in 1989.

^cFire disturbances occurred in one of the Foster and Foster Biological oak woodland sites in 2003, as well as in one of the chaparral sites in 2004.

^dPrescribed sheep grazing occurred at both Foster oak woodland sites.

Sites fell under a mesic to dry Mediterranean climate, characterized by cool, moist winters and warm, dry summers. Two coastal redwood sites were identified at the Caspar Creek Watershed in Jackson Demonstration State Forest, located near Fort Bragg, California. Four oak woodland and two chaparral sites were selected within the Coast Range of the Mayacamas Mountain foothills, in the University of California's Research and Extension Center. These sites have been exposed to various watershed-scale treatments, including clearcutting (1989 in one redwood site), burning (2003 in two of the oak woodland and 2004 in one of the chaparral sites), prescribed grazing (oak woodland), combinations of burning and grazing (oak woodland), and no disturbance (redwood, oak woodland, chaparral). Total dissolved nitrogen inputs from rainfall and dry deposition at these sites during the time of sampling did not differ significantly, although NO₃⁻ deposition was slightly higher in oak woodland and chaparral, dissolved organic N deposition more dominant in the redwood sites (Mnich and Houlton, 2015). The oak woodland and chaparral grazed sites may also receive NO₃⁻ inputs from grazer wastes; also, these sites are closer to farmland than the redwood sites and might receive NO₃⁻ inputs from fertilizers.

The desert site was located in the Sweeney Granite Mountain Desert Research and Extension Center in the Mojave National Preserve. Owing to patch-scale heterogeneity of the desert site, we divided our sampling scheme into two different categories: ‘under shrub/creosote' samples were collected directly under the canopy of the dominant creosote bush, Larrea tridentata, and ‘interstitial' samples were unvegetated. Any NO₃⁻ inputs would have to be wind-blown dry deposition, or else fixed from the atmosphere by soil microorganisms, as there was little rainfall at this site.

Field sampling approach

Surface soils (0–15 cm depth, at least five replicates per site) were sampled at all sites over a range of wet (winter) and dry (summer) seasons. This approach allowed us to examine denitrification in the organic soils where C and N concentrations were highest and heterotrophic and rhizosphere activity is most concentrated (Lennon, 2015). For homogenization, soil was sieved to the fine-earth fraction (<2 mm), and roots/organic debris were removed with tweezers. To minimize DNA or isotope cross-contamination, rubber gloves were worn and sampling tools (that is, shovels and trowels) were cleaned with ethanol between collections. Homogenized, fine-fraction, field-moist samples were mixed in half-gallon, resealable polyethylene bags, and subsamples were kept in 60 ml Falcon tubes and immediately put on ice in the field.

Week-long incubation experiments were conducted in-situ at redwood (n=10), chaparral (n=10), desert under-shrub (n=8) and desert interstitial (n=8) sites to isolate soils from leaching and plant uptake effects. At least 100 g surface soil (0–10 cm depth) was homogenized—using the methods described above—and brought to field capacity with high purity 14.6 M-ohm/cm water in half-gallon, resealable polyethylene bags. Prior to sealing the bags, subsamples were collected in 60 ml Falcon tubes and immediately placed on ice for analyses of initial nutrient, isotope and genetic conditions. The bags were then pressed to expel air, sealed to promote an anaerobic environment and buried in place. Field-incubated samples were collected after 7 days and analyzed for chemistry, isotopes and nir abundance.

Soil moisture

Pre-weighed tins were filled with approximately 15 g field-moist, fine-fraction soil and weighed. Soil-filled tins were placed in a drying oven set at 55 °C. Dry weights were recorded after 48 h or when soil was completely dry. Moisture was calculated as water weight of the total sample and was used to transform field-moist weights in future analyses into soil dry weights.

KCl extractable soil nitrate

Soil inorganic N was extracted in-field with a 2 M potassium chloride (KCl) solution within 4 h of sample collection. The method was adapted from Keeney and Nelson to the field-setting (Keeney and Nelson, 1982). HDPE Nalgene containers were triple-rinsed daily in high-purity water for three days prior to usage. The 2 M KCl solution was prepared with high-grade KCl and high-purity water in a clean Nalgene container sealed with a twist-on cap and parafilm to prevent leakage during travel. A 50 ml volume of 2 M KCl and approximately 10 g field-moist soil were both measured into 60 ml-capacity sterile polypropylene specimen containers, which were then tightly sealed with twist-on caps. Samples were shaken vigorously for 5 min, left to settle upright for 5 min, shaken again for 2 min, and then left to settle upright for 2 h. The supernatant was filtered through funnels lined with ashless Whatman No.1 paper filters, which had been pre-washed with the 2 M KCl solution. The filtrate was collected in clean 60 ml-capacity HDPE Nalgene containers, which were immediately stored in a freezer or kept on constant ice in a cooler for later quantitative analyses in our UC Davis lab.

Colorimetric methods were used to quantify KCl extractable soil NO₃⁻. A Griess reagent was added to samples in semi-micro cuvettes, with a 1:1 ratio in most cases, adjusted for greater sensitivity with lower concentrations (Doane and Horwáth, 2003). Vanadium(III) chloride reagent produced a red color indicating NO₃⁻, which was measured after 8 h at 540 nm in a spectrophotometer (Miranda et al., 2001; Doane and Horwáth, 2003). NO₃⁻ was calculated based on a standard curve created with stock potassium nitrate solution. A minimum 1 μM NO₃⁻ in the KCl extract was required for isotope analyses.

¹⁵N/¹⁴N and ¹⁸O/¹⁶O analysis

Natural abundance of ¹⁵N/¹⁴N and ¹⁸O/¹⁶O in soil NO₃⁻ extracts were analyzed via a ThermoFinnigan PreCon-GasBench interfacing a Delta V Plus isotope-ratio mass spectrometer (Rock and Ellert, 2007) at UC Davis' Stable Isotope Facility. Samples were stored at −20 °C and thawed and mixed thoroughly before analysis. The denitrifier method was used to convert NO₃⁻ (and any trace NO₂⁻) to N₂O, which was purged from vials, passed through an Ascarite CO₂ scrubber, and concentrated in two subsequent liquid N₂ cryo-traps (Sigman et al., 2001; Casciotti et al., 2002). A He stream (25 ml min⁻¹) carried the gas sample through this trace gas concentration system and then through an Agilent GS-Q capillary column (1.0 ml min⁻¹, 30 m × 0.32 mm, 40 °C).

International NO₃⁻ isotope standards USGS 32, USGS 34, and USGS 35 (National Institute of Standards and Technology, Gaithersburg, MD, USA) were used to calibrate the instrument, in addition to batch standards to account for drift. International reference standards were used to calculate isotope composition –N₂ in air for δ¹⁵N and Vienna Standard Mean Oceanic Water (VSMOW) for δ¹⁸O.

NO₃⁻ isotope composition was corrected for the small KCl blank as follows:

where X=N or O isotopes of nitrate, and f=fraction nitrate of total nitrate.

Soil DNA extractions

DNA was extracted from ~0.25 g fine-earth fraction (<2 mm) soil—corrected for moisture—into 100 μl aliquots using Mo-Bio PowerSoil DNA Isolation Kit (Mo Bio Laboratories, Carlsbad, CA, USA) following manufacturer instructions. To verify a successful extraction and to quantify double-stranded DNA (dsDNA), the Qubit fluorometer (Thermo Fisher Scientific, Carlsbad, CA, USA) assay was used with Quant-iT dsDNA Broad Range reagent and buffer. DNA extracts yielding detectable dsDNA were then stored at −20 °C according to the Mo-Bio PowerSoil DNA Isolation Kit.

qPCR to determine functional gene abundances

Real-time qPCRs were used to detect and quantify specific functional genes of interest: nitrite reductase genes nirK and nirS (NO₂⁻ → NO, the denitrification step with most isotopic fractionation), and bacterial 16S rRNA (the gene encoding for a highly conserved region in prokaryotic organisms, estimating total bacterial population). The genes of interest were amplified with specific target primers and thermal conditions and observed using fluorescent dye. The primers used for nirS were nirSCd3aFm and nirSR3cdm (Throback et al., 2004). The nirK primers were nirK876 and nirK1040 (Henry et al., 2004). The 16S rRNA primers were 341F and 534R (Lopez-Gutierrez et al., 2004). See Supplementary Materials for full primer sequence descriptions and thermal conditions.

Standards were pre-made with serial dilutions of target DNA cloned into a plasmid (Scow Laboratory, UC Davis, Davis, CA, USA). Each total reaction had 5 μl template DNA (various dilutions for optimal detection), 12.5 μl SYBR GreenER qPCR SuperMix (Invitrogen, stored at 4 °C), 0.5 μM forward and reverse primers, and nanopure autoclaved deionized water for a 25 μl total volume. All standards, environmental samples and controls were run in triplicate in Optical 96-Well Fast Thermal Cycling Plates via Applied Biosystems 7300 Real-Time PCR System.

Gene copies/g dry soil=[(mean quantity target gene detected)/5 μl sample] × (dilution factor × 100 μl original aliquot)/g dry soil. Mean quantity target gene detected=10^((n−b)/m), where n=the number of cycles required for the fluorescent signal to exceed the background level, and m and b are the slope and y-intercept of the standard curve. A slope of −3.5±0.4 is acceptable with R²>0.99.

Statistics and analysis

Means and standard errors of the natural logarithm of KCl extractable soil nitrate (NO₃⁻), δ¹⁵N of NO₃⁻, δ¹⁸O of NO₃⁻, and nirS abundance were regressed for sites for which > three parameter sets of values were detected. Using the natural logarithm of NO₃⁻ allowed us to perform linear regressions. Data distributions were normal or near-normal for all parameters based on normal quantile plots. Least-squares regression analyses determined the linear fit of all scatterplots, and we reported R-squared and P-values of regression slopes. P-values less than 0.05 were considered significant; at a 95% confidence level we reject the null hypothesis that there is no effect among variables.

We were unable to assemble full data sets of all target parameters for each site, owing to detection limits on NO₃⁻ concentrations (KCl sample extracts with <1 μM NO₃⁻ prevented samples from being analyzed for stable N and O isotopes). We focused results on sample sites wherein at least three samples were collected with detectable ensembles of NO₃⁻ concentrations, and abundance of isotopes and nirS. In the case of oak-woodland sites, criteria of triplicate nirS abundance data from at least three samples were not met; however, we were able to measure oak woodland N and O isotopes. Hence we include these data in analysis of isotopes but not the fully coupled analysis of isotopes, gene abundances and NO₃⁻ concentrations. Significant outliers detected via leverage and Cook's distance analyses were removed.

Stepwise regressions for δ¹⁵N- NO₃⁻ and nirS abundance were performed for incubation samples with the following data at a 95% confidence level: average field soil moisture for top 10 cm, average field %C for top 10 cm soil, mean air temperature, soil moisture in the incubation, nirS abundance, nirK abundance, 16S rRNA abundance, nirS/16S rRNA, nirK/16S rRNA, ln[NO₃⁻], O and N isotopes from NO₃⁻. These stepwise regressions were repeated for incubation samples from 7-day incubations which displayed decreased NO₃⁻ concentrations and had δ¹⁵N>0, representing samples with denitrification as the dominant NO₃⁻ loss process that had occurred. Results from these analyses steered the focus of the figures and discussion toward select relationships.

Results

Soil nitrate and isotope composition patterns

Soil NO₃⁻ concentrations differed substantially across sites, generally tracking changes in MAP and soil C concentrations among terrestrial biomes (Table 1,Figure 2). NO₃⁻ concentrations were lowest in the mesic redwood site, and increased into the drier ecosystems, with a peak concentration observed for the desert sites, particularly in samples collected beneath creosote vegetation (Figure 2). The δ¹⁵N of KCl extractable NO₃⁻ was negatively related to its concentration; highest mean δ¹⁵N-NO₃⁻ was observed for redwood sites (control 11.67±1.09‰, clearcut 10.79±0.53‰), intermediate values were apparent in the chaparral (1.56±1.37‰), and lowest values were observed for the desert sites (under shrub 0.85±0.32‰, interstitial −0.88±2.53‰) (Figure 2). Highest mean δ¹⁸O-NO₃⁻ values were observed in redwood control (11.03±2.19‰) and chaparral (11.61±1.21‰), intermediate in redwood clearcut (4.22±0.71‰), and lowest in desert sites (under shrub 0.47±0.69‰, interstitial −1.28±0.76‰) (Figure 2). δ¹⁵N-NO₃⁻ positively related with soil C (R²=0.22, P<<0.05).

Figure 2.

Surface soil NO₃⁻, δ¹⁵N-NO₃⁻ and nirS across sites. Includes sites with full study data set (N and O isotopes, extractable NO₃⁻, nirS). n=9 (redwood control, 13 for NO₃⁻), 3 (redwood clearcut), 11 (chaparral, 14 for NO₃⁻), 15 (desert under creosote), and 14 (desert interstitial). Bars are standard error.

Linear regressions indicated several statistically significant relationships among soil NO₃⁻ isotopes and concentrations. Field surface δ¹⁵N-NO₃⁻ and δ¹⁸O-NO₃⁻ were positively correlated, with a linear slope for all surface soil isotope values at 0.55 (R²=0.25, P=0.0002) (Figure 3). If excluding sites without sufficient corresponding nirS data (that is, no oak woodland), the slope lowered to 0.47 (R²=0.17, P=0.005). A significant, negative relationship was observed between soil NO₃⁻ concentrations and δ¹⁵N-NO₃⁻ across all sites examined (R²=0.19, P=0.003) (Figure 5).

Figure 3.

δ¹⁸O-NO₃⁻ vs δ¹⁵N-NO₃⁻. Bold symbols are mean values per site. Bars are standard error. Linear regression for field surface soil datapoints is y=0.55x +1.71 (R²=0.25, P=0.00026, n=49). Outliers identified via leverage analyses were removed. If excluding sites without sufficient corresponding nirS data (that is, no oak woodland), y=0.47x+1.86 (R²=0.17, P=0.005, n=44).

A positive relationship between δ¹⁵N-NO₃⁻ and δ¹⁸O-NO₃⁻ was observed in the in-situ 7-day surface soil (10 cm) incubations at the redwood and chaparral sites (Figure 6). In these ecosystems, decreased NO₃⁻ concentrations were strongly related to an increase in δ¹⁵N-NO₃⁻ (R²=0.50, P=0.01, n=11). In contrast, there was no clear linear relationship for N and O isotopes from desert incubations, despite declines in NO₃⁻ concentrations during the incubation experiments (R²=0.00, n=15; Figure 6).

nirS gene abundance patterns

Mean nirS abundance averaged ~10⁷ copies g⁻¹ dry soil across sites. Abundances were highest in the redwood site (7.43 × 10⁷±1.22 × 10⁷ copies g⁻¹ in redwood control, 6.04 × 10⁷±1.32 × 10⁷ copies g⁻¹ in redwood clearcut) and lowest in the desert sites (2.64 × 10⁶±1.22 x10⁵ copies g⁻¹ under creosote bush, 9.66 × 10⁶±2.65 × 10⁶ copies g⁻¹ in the interstitial spaces; Figure 2). nirS abundance positively correlated with soil C (R²=0.51, P<<0.05).

Relationships between isotope composition and nirS

nirS and δ¹⁵N-NO₃⁻ were significantly positively related across all sites (R²=0.35, P=0.00002) (Figures 2 and 4), and both parameters exhibited significant negative relationships with NO₃⁻ concentrations (nirS R²=0.20, P=0.002; δ¹⁵N-NO₃⁻ R²=0.19, P=0.003; Figures 2 and 5). Although the distribution of field nirS data appears bi-modal (Figure 2), the Breusch-Pagan Test revealed homoscedasticity of the error terms (P>0.05) for a linear regression model of ln[NO₃⁻] vs nirS (Figure 4). nirS abundances and δ¹⁵N-NO₃⁻ were highest in the undisturbed redwood site, which corresponded with the lowest soil extractable NO₃⁻ concentrations (Figure 2). The desert soils under the creosote vegetation exhibited the highest soil extractable NO₃⁻ concentrations, lowest nirS abundance and δ¹⁵N-NO₃⁻ near 0‰ (Figure 2).

Figure 4.

nirS vs δ¹⁵N-NO₃⁻. Bold symbols are mean values per site. Bars are standard error. Linear regression for all datapoints is y=3.27E^6x +1.20E^7 (R²=0.35, P=0.00002).

Figure 5.

nirS (top) and δ¹⁵N-NO₃⁻ (bottom) vs ln[NO₃⁻]. Soil extractable nitrate concentration prior to logarithmic transformation was in units of μM NO₃⁻ g⁻¹ dry soil. Bold symbols are mean values per site. Bars are standard error. Linear regressions for all datapoints: nirS y=−9.55E+06x + 3.28E+07 (R²=0.20, P=0.002); δ¹⁵N-NO₃⁻ y=−1.66x + 5.24 (R²=0.19, P=0.003).

Positive denitrifier gene-isotope relationships were likewise observed in the 7-day incubation experiments conducted in the redwood and chaparral sites. The decrease in NO₃⁻ was associated with both an increase in δ¹⁵N-NO₃⁻>0 (n=11) and increase in nirS (R²=0.15, P=0.24) in these ecosystems. Furthermore, nirS/16S rRNA exhibited a significant positive relationship with δ¹⁵N-NO₃⁻ (R²=0.70, P=0.001; Figure 7) in the short-term incubation experiments. From stepwise multiple regressions of desert incubations with NO₃⁻ loss and positive N isotopes, δ¹⁵N-NO₃⁻ responded negatively to ln[NO₃⁻], positively to nirS abundance, negatively to nirS/16S rRNA and slightly negatively to δ¹⁸O-NO₃⁻ (multiple R=0.86, standard error s.e.=7.4, n=15, significance F<0.05). For stepwise regressions of corresponding redwood and chaparral incubations, δ¹⁵N-NO₃⁻ responded negatively to ln[NO₃⁻], not significantly with nirS abundance, and positively to δ¹⁸O-NO₃⁻ (multiple R=0.95, s.e.=2.01, n=11, significance F<<0.05).

Other environmental and microbial variables included at the start (temperature, carbon, moisture, nirK, nirK/16S) did not have a significant relationship with δ¹⁵N-NO₃⁻ by the end of the stepwise regressions. These results were from stepwise regressions set at 95% confidence. Differences in linear response to δ¹⁸O-NO₃⁻ alone are shown in Figure 6. Denitrifier linear responses to δ¹⁵N-NO₃⁻ are shown in Figure 7.

Figure 6.

δ¹⁸O-NO₃⁻ vs δ¹⁵N-NO₃⁻ incubations. Includes initial and final subsamples from 7-day incubations for which [NO₃⁻] decreased and δ¹⁵N-NO₃⁻>0. Linear regression for redwood and chaparral incubations is y=0.58x +8.03 (R²=0.50, P=0.01, n=11). Linear regression for desert incubations is y=0.017x +2.05 (R²=0.00, n=15).

Figure 7.

Denitrifiers vs δ¹⁵N-NO₃⁻ in redwood and chaparral incubations. Data include any initial and final subsamples from redwood and chaparral 7-day in-situ incubations for which [NO₃⁻] decreased and δ¹⁵N-NO₃⁻>0 (n=11). Linear regression for nirS/16S rRNA is y=0.0075x + 0.0053 (R²=0.70, P=0.001). Regression for nirS is y=2.0 × 10⁶x +3 × 10⁷ (R²=0.15, P=0.24).

Discussion

Our findings support the first hypothesis—that variations in terrestrial δ¹⁵N are controlled primarily by biogenic gaseous emissions of N (Houlton and Bai, 2009), principally isotopic discrimination via denitrifying organisms' enzyme-systems (Mariotti et al., 1981). This result agrees with the elevated δ¹⁵N of N inputs compared to that of N leaching losses in our sites (Mnich and Houlton, 2015), and furthermore is consistent with global scale imbalance between input δ¹⁵N and the residual N of the terrestrial biosphere (Houlton and Bai, 2009; Vitousek et al., 2013). In addition, the negative relationships between ecosystem NO₃⁻ availability and δ¹⁵N/nirS support the second hypothesis—that microbial denitrification can consume NO₃⁻ to low levels in terrestrial ecosystems, even in temperate environments where N limitation is widespread (LeBauer and Treseder, 2008). The incubation portion of our study further identifies the capacity for denitrifiers to rapidly consume NO₃⁻ to low levels within a given site. Overall, these findings are at odds with general formulations of denitrification in contemporary ecosystem models (for example, CENTURY, Davidson and Seitzinger, 2006), which simulate an increase in this process with increasing soil NO₃⁻ supplies. Rather, the evidence points to direct control of denitrifiers over soil NO₃⁻ availability patterns, which, over time, help to explain the persistence of N limitation observed for grassland to forest ecosystems (Vitousek and Howarth, 1991, LeBauer and Treseder, 2008).

One possible explanation for the denitrifier-driven effect on soil NO₃⁻ is that plants in these sites rely on NH₄⁺ and/or direct uptake of amino acids to meet their N demands, thereby allowing denitrifiers to consume NO₃⁻ to relatively low levels. This interpretation is consistent with elevated δ¹⁵N and δ¹⁸O where NO₃⁻ is low vs high; plant reliance on NO₃⁻ would not impart such significant isotope effects. Moreover, denitrifiers may directly out-compete plants and other microbes for NO₃⁻ in moist and C-rich environments, perhaps taking advantage of strong redox potential gradients across soil aggregates in the forest and savanna sites (Sexstone et al., 1985; Ebrahimi and Or, 2016). Indeed, the significant positive correlations between nirS (R²=0.51, P<<0.05) and δ¹⁵N-NO₃⁻ (R²=0.22, P<<0.05) vs soil C concentrations across our terrestrial sites point to the likelihood of this mechanism. It is worthwhile to note that aggregate pore sizes can influence anoxic zones and greenhouse gas diffusion/emissions, even in unsaturated and variable field conditions (Ebrahimi and Or, 2016), although we did not study aggregates closely in these sites.

Temperature is another environmental factor to consider, as our sites exhibited a gradient in mean air temperatures (Table 1). Temperature increases have been shown theoretically and empirically to influence the kinetics of denitrification-driven isotopic fractionation (Maggi and Riley, 2015). The affinity for biological NO₃⁻ fractionation increases with temperature, and Maggi and Riley, 2015 accurately modeled this nonlinear phenomenon at a range of temperatures (20-35 °C), even with heterogeneous soils. While air temperature was not as closely related to isotopic data as other variables were in our study, it is not outside the realm of possibility that soil temperatures, which were not measured, influenced the patterns observed.

Furthermore, other N transformation processes could play a role in influencing available NO₃⁻ pools and potentially δ¹⁵N, including: nitrification, heterotrophic immobilization and potentially dissimilatory NO₃⁻ reduction to NH₄⁺ (DNRA). Nitrification is a fractionating process that acts to lower the δ¹⁵N of NO₃⁻ vs that of NH₄⁺ substrates, whereas both heterotrophic microbial assimilation and DNRA would be expected to impart an isotope effect on NO₃⁻ that resembles that of denitrification.

However, additional evidence for principal denitrifier controls over NO₃⁻ concentrations is apparent in dual-isotope relationships in our study (Figure 3). Specifically, a positive relationship between δ¹⁸O and δ¹⁵N with a slope near 0.6 is diagnostic of terrestrial denitrification, because denitrifying bacteria preferentially convert both light isotopes in NO₃⁻ to gaseous products (Lehmann et al., 2003). Heterotrophic NO₃⁻ immobilization also fractionates against O and N isotopes in NO₃⁻; though at a much higher slope, typically between 1.0 and 2.0 (Granger et al., 2010). Our observed 0.55 slope for field sites (R²=0.25, P=0.00026) is therefore consistent with the imprint of terrestrial denitrifiers on soil NO₃⁻ consumption (Figure 3), as seen in previous work in tropical and temperate forest soils (Houlton et al., 2006; Fang et al., 2015). We cannot specifically address why the slopes differ across studies, but past work on marine bacteria cultures indicates that chemolithotrophs fractionate a slope closer to 0.5, whereas heterotrophic denitrifiers appear to fractionate O and N isotopes at a higher slope (Frey et al., 2014). Differences between microbial effects on O and N isotope relationships generally reflect differences in kinetic rate transfers of isotopes from substrates to products and O exchange reactions with water and air. Indeed, variations in the isotopic composition of O source for NO₃⁻ could explain the δ¹⁸O-NO₃⁻ differences observed across our ecosystem sites (for example, redwood control vs clearcut in Figure 2).

Moreover, the spatial patterns observed across sites are similarly observed in our short-term incubation experiments (that is, 7 days), thus pointing to the importance of microbial denitrifiers in driving soil NO₃⁻ availability patterns in both space and time. Aerobic nitrification and even atmospheric deposition may have slightly diminishing effects on N and O isotopes of NO₃⁻ in an open system; and so the relationship between N and O isotopes of NO₃⁻ in anaerobic incubations is of particular interest. Indeed, δ¹⁵N and δ¹⁸O of NO₃⁻ in the soil incubation experiments reveal a slope that matches expectations for marine and freshwater denitrifiers (Lehmann et al., 2003; Granger et al., 2008) and observations for other terrestrial ecosystems (Houlton et al., 2006; Fang et al., 2015), similar to the spatial relationships observed for all samples across sites (that is, ~ 0.6) (Figure 6). Not only did we observe a positive relationship between nirS and δ¹⁵N-NO₃⁻ mirroring the cross-system results; but there was a significant positive correlation between nirS/16S rRNA and δ¹⁵N-NO₃⁻ in our soil incubation experiments as well (R²=0.70, P=0.001) (Figure 7). This demonstrates that the microbial community composition was rapidly enriched in organisms with denitrifying genes, which explains the increase in δ¹⁵N and δ¹⁸O that accompanied declining NO₃⁻ concentrations. Such community streamlining has been observed in aquatic environments (Jayakumar et al., 2009), but, to our knowledge, these are the first observations for terrestrial soil-systems. In sum, these findings support our conclusion that denitrifiers played a major role in determining NO₃⁻ pool sizes across sites, with isotopic and molecular evidence revealing complementary patterns across scales, from induced short-term consumption events to the determination of cross-site patterns among biomes.

Importantly, the inverse relationships between δ¹⁵N-NO₃⁻ and soil NO₃⁻ concentration is observed across sites (Figures 2 and 5); however, the calculated isotope effects were much lower than those observed in pure cultures of denitrifying bacteria. The isotope effect of denitrification ranges from 0‰ to ~−33‰ (Högberg, 1997), with an average of approximately −20±1.0‰ observed for laboratory studies (Delwiche and Steyn, 1970; Granger et al., 2008; Houlton and Bai, 2009). The magnitude of the isotope effect can be lower in the field, particularly in environments where non-homogenous interactions allow for locally complete NO₃⁻ consumption, as in a ‘closed system' Rayleigh model (Brandes and Devol, 2002; Houlton et al., 2006). In this case, NO₃⁻ can be completely consumed in micro-environments leaving little or no residual NO₃⁻ to express denitrification's isotope effect (Houlton et al., 2006). Furthermore, non-fractionating sinks such as plant uptake and microbial assimilation can consume NO₃⁻ without altering ¹⁵N/¹⁴N. In our sites the relationship between δ¹⁵N and ln[NO₃⁻] suggests an integrated isotope effect—a net isotope effect reflecting all NO₃⁻ sources and sinks—that varies from -2.12 in the redwood site to 1.92 in the desert interstitial site, with an overall effect across sites of −1.66.

One exception to the inverse relationship between δ¹⁵N-NO₃⁻ and soil NO₃⁻ is evident in the desert soil, particularly beneath vegetation canopies. The soil devoid of vegetation had notably lower [NO₃⁻] than those collected under L.tridentata ‘islands of fertility' and yet the interstitial sites had the same, if not lower, δ¹⁵N-NO₃⁻ as the vegetated ones (Figure 2). The weakly positive relationships in interstitial spaces suggest that nitrification, ammonia volatilization or other abiotic processes could be responsible for gaseous N production in such arid environments (McCalley and Sparks, 2009). It is possible that in arid and wind-prone desert ecosystems, uptake of NO₃⁻ by heterotrophic microbes and plants could mask any denitrification-driven enrichment of heavy N and O isotopes. Any substantial non-fractionating sinks for NO₃⁻ could greatly reduce the isotope effect of denitrifiers. We cannot rule out the role of episodically driven effects on denitrification in the desert, where precipitation occurs over very short, high-intensity events.

The field portion of our study provides useful, integrated snapshots of O and N isotopes of NO₃⁻ in a variety of field settings—taking into account the collective effects of ecosystem N cycling. The incubation experiments help to isolate the field-based evidence for denitrifier-driven controls over soil NO₃⁻, albeit under optimal conditions of low O₂ availability. Future work in such dynamic systems could benefit from long-term, daily field studies coupled with field-sensors of soil moisture, pH, temperature and redox.

Implications for linking microbial gene abundance to ecosystem functioning

Our results have implications for understanding linkages among microbial functional gene abundances and ecosystem pattern and process. As modern genetic techniques have tremendously increased, questions over the utility of microbial gene abundance measures have become paramount, representing a key area of active research in the microbial ecology and ecosystem biogeochemistry. Our finding of a positive correlation between gene abundance and isotope composition provides a new, integrative tool for connecting DNA to variations in ecosystem NO₃⁻ pools and natural isotope composition; it reveals a link between genes encoding for a given enzyme (for example, nirS) and its role in a key ecosystem process (for example, denitrification) across biomes. We suspect that such a ‘gene abundance to isotope' relationship is likely best observed in wet seasons or generally moist sites, owing to denitrifiers' ability to thrive in high moisture and anaerobic soil environments where C is abundant (Dawson and Murphy, 1972; Burgin et al., 2010; Szukics et al., 2010; Groffman 2012).

Previous studies on relationships between nirS/K and terrestrial denitrification rates have been mixed, which could be explained by mismatches in the scale of measurements or artifacts of disturbance (Wallenstein et al., 2006). Structure and abundance of microbial communities bearing nirS appear less sensitive to long-term fertilizations than those with nirK (Chen et al., 2010), and actual transcription of nirS is less sensitive to changes in pH than nirK (Liu et al., 2014). Thus, our focus on nirS likely minimized the chances of localized factors from obscuring relationships between community structure, gene abundance and actual gene expression.

Terrestrial denitrification has remained a poorly understood aspect of N cycling research (Davidson and Seitzinger, 2006). Our results demonstrate a significant link between nirS and natural isotope abundance, over scales ranging from short-term soil incubation experiments to the factors structuring longer-term soil NO₃⁻ pools across forest to desert ecosystems. Kinetic isotopic discrimination is the result of denitrifying organisms' enzymatic preferences for light isotopes in NO₃⁻. Thus, while gene abundance data alone do not represent the diversity or activity of soil microbial communities, the relationships between nirS and heavy isotope enrichment point to tight connections between denitrifier gene abundances and microbial enzyme activity across our sites. Where N availability is highest in the desert biome, the imprint of denitrification appears weakest; where N availability was lowest, coupled molecular and isotopic data point to substantial NO₃⁻ consumption by denitrifying bacteria. These findings suggest that models of denitrification should be re-formulated to include a more direct influence of denitrifiers in determining terrestrial N availability.