Legacy of contaminant N sources to the NO3− signature in rivers: a combined isotopic (δ15N-NO3−, δ18O-NO3−, δ11B) and microbiological investigation

Nitrate content of surface waters results from complex mixing of multiple sources, whose signatures can be modified through N reactions occurring within the different compartments of the whole catchment. Despite this complexity, the determination of nitrate origin is the first and crucial step for water resource preservation. Here, for the first time, we combined at the catchment scale stable isotopic tracers (δ15N and δ18O of nitrate and δ11B) and fecal indicators to trace nitrate sources and pathways to the stream. We tested this approach on two rivers in an agricultural region of SW France. Boron isotopic ratios evidenced inflow from anthropogenic waters, microbiological markers revealed organic contaminations from both human and animal wastes. Nitrate δ15N and δ18O traced inputs from the surface leaching during high flow events and from the subsurface drainage in base flow regime. They also showed that denitrification occurred within the soils before reaching the rivers. Furthermore, this study highlighted the determinant role of the soil compartment in nitrate formation and recycling with important spatial heterogeneity and temporal variability.

High nitrate concentrations in surface and ground waters remain a worldwide concern because of sanitary problems in drinking water and of ecological disturbances in aquatic systems such as eutrophication [1][2][3] . Hence, the determination of nitrate origin is the first step for effective management plans aiming at preserving surface water quality. Nitrate concentrations in rivers are controlled by spatial and temporal variability of the different nitrate sources and biogeochemical or physical reactions occurring from uplands to streams [4][5][6] (Fig. 1). The nitrogen and oxygen isotope ratios of NO 3 − (δ 15 N and δ 18 O) have been widely used to investigate the sources of NO 3 − in rivers and groundwater 4,[6][7][8][9][10][11][12][13][14] . Actually, at the catchment scale, some N sources, such as domestic and animal effluents, present overlapping δ 15 N ranges. The isotopic signature of nitrate results also from processes within the soils that modify its concentration (nitrification, denitrification, Fig. 1) [15][16][17] fractionate its isotopes, and can blur the initial signature of the N sources. These limits can be over passed by using complementary tracers.
The objective of the present study was to determine the origin of nitrate in rivers where multiple nitrogen sources co-exist (Fig. 1). To do so, isotopic (δ 15 N, δ 18 O and δ 11 B) and microbiological markers were combined for the first time at the watershed scale. This multi tracers approach was tested on two rivers of the southwest of France in an agricultural region (Fig. 2). Nitrate concentrations of Gabas River (GR, 150 km long) and Laudon River (LR, 15 km long) are a serious threat for the underlying karstic aquifer, used for drinking water supply 38 . To better integrate the spatial heterogeneity of N point and non-point sources of the whole watershed, a 2.5 years monitoring has been conducted in the karstified downstream part of the Gabas and Laudon catchments (Fig. 2). Water sampling was carried out every two months in base flow conditions but also during flood events (Fig. 3) to characterize NO 3 − dynamics in rivers. Additionally, samples of rain, waste water treatment plant (WWTP) effluents, manure and animal fresh dejections were collected to determine the isotopic and microbiological signature  of the local N sources (Fig. 2). Finally, samples of agricultural soil, water from buried drains and surface ditch were sampled to follow the N fluxes from topsoil to rivers (Fig. 2).

Results and Discussion
Results for each tracer are first presented and discussed separately in order to understand how they can trace the different water sources but also to highlight their limits. Then we discuss how the tracer's combination allows a global understanding of NO 3 − sources, reactions and pathways.  Fig. 5).
The large gap between δ 15 N and δ 18 O of nitrate extracted from topsoil and nitrate of the drainage network (ditches and buried drains) reflects that important processes occur within the soils before reaching the studied rivers. It has recently been shown that residual fraction of NO 3 − that is not immediately leached or consumed by plants is assimilated into soil organic matter and potentially recycled into NO 3 − 40 (Fig. 1). Some of the reactions involved in the N cycle, such as NH 4 + volatilization and NH 4 + nitrification are important isotope fractionating processes 16,17,41,42 . N availability and reaction rate of each process will thus control the large δ 15 N range of newly produced nitrate 16,43 . Additionally, in top-soils, water evaporation and the subsequent increase of δ 18  ] and rivers flow rates (Fig. 6), for both rivers, the lowest δ 15 N and δ 18 O values were mainly observed at high water stage (> 2 m 3 .s −1 for GR) whereas the highest were measured for base flow (< 2 m 3 .s −1 for GR, Fig. 6). These higher δ 15 N values could be due to a larger contribution of an enriched pool of nitrate such as WWTP effluents (δ 15 N = + 10.0‰ to + 17.3‰) (Figs 4 and 5) during base flow compared to high flow events. However, the δ 18 O of NO 3 − in the WWTP outlet (+ 6.4‰ to + 8.5‰) are lower than the maximum values measured in GR and LR and thus cannot explain the δ 18 O enrichment also observed for base flow samples (Fig. 5). Above all, considering the high [NO 3 − ] of WWTP effluents, a larger contribution of this of this pool should also have increased [NO 3 − ] in rivers, which is not observed (Fig. 4). A simple mixing of N sources thus cannot explain these high nitrate δ 15 N and δ 18 O during base flow and additional processes must be involved. Actually the nitrate δ 15 N and δ 18 O of LR samples, are pretty well distributed along the 2:1 slope expected for residual nitrate derived from denitrification 15,16,47 (Fig. 5) and are roughly inversely correlated with [NO 3 − ] (Fig. 4). For the GR samples, δ 15 N and δ 18 O values do not strongly follow the denitrification slope but a global positive trend exists between high-and base-flow samples. Chen et al. 6 , reported a similar pattern with seasonal distribution in the Beijang catchment and concluded that denitrification occurred within the watershed soils, before residual NO 3 − reached the river 6,7 . For the samples collected under high flow conditions (Fig. 6), nitrate δ 15 N vary largely while δ 18 O remain quite constant, which excludes denitrification processes. In this case, the low δ 15 N compared to base flow samples might be explained by a larger contribution of a 15 N depleted nitrate pool. The first possibility is that during these highly rain periods, the atmospheric nitrate (δ 15 However, if such mixing process had occurred, it would also have increased the δ 18 O-NO 3 − of rivers, which is not observed here   Fig. 7), usually encountered in uncontaminated water 19,48 . The δ 11 B of these samples (+ 13.5‰ and + 26.8‰) plots into the overlapping typical ranges of rain 40 and manure sources 19,42 (Fig. 7). The δ 11 B measured for local rain (27.7‰) and WWTP effluents (1.9‰) were in good agreement with the ranges reported in literature for atmospheric and domestic boron 19,22 (Fig. 7). However, the characterization of local animal pool (fresh dejection δ 11 B = − 3.3‰ and poultry manure δ 11 B = + 8.6‰) shows lower values than previously reported (+ 15.3‰ to + 27.6) 19 . Globally, [B] and δ 11 B of the GR and LR samples were really close to those measured in the underlying karstic aquifer (10.5 ± 2 ppb and 25.3 ± 1‰) 38 , and in rain, and were significantly different from the WWTP end-member (Fig. 7).   (Table 1). This domestic fecal contamination was corroborated by the detection of FRNAPHs of Group II at significant concentrations (proportion greater than 20% of total phages counted on more than 12 phages) in two of the 4 ditch samples, in 2 high flow LR samples and in both high-(n = 3) and base-flow (n = 2) GR samples (Table 1). These results are consistent with the fact that domestic rejects are constant and do not depend on hydrology or season. Animal contamination was pointed out in both rivers. Among the eleven samples from GR, cattle and pig markers were detected in respectively 7 and 4 samples, mostly collected under high flow conditions. Among the 15 samples of LR, cattle Bacteroidales were detected ten times in both high (6/10) and base flow samples (4/10) while pig markers were only detected in high flow samples (5/15). Finally, duck-chicken-goose marker was tested during the flood event of the last campaign and was detected in all GR and LR samples. These results reflect that runoff is the major pathway of microorganisms from animal sources to rivers through surface leaching during rainy events, but the markers do not allow to distinguish contamination from point sources, such as fresh dejections in farms, and non-point sources, such as manure spread on fields. However, as no animal marker was found (at significant concentration) in the ditches that drain maize fields, it could traduce that animal contamination measured in GR and LR rather arises from farming rather than from manure spread on fields. For LR, the regular presence of cows in the Laudon 200m upstream the sampling point can explain how cattle markers can reach the LR in absence of rain event.
The microbiological tracers undeniably indicate that domestic and animal fecal contaminations do impact the two rivers and thus can potentially contribute to NO 3 − contents of GR and LR.
Processes and pathways within catchment. In this study, anthropogenic contaminations have been evidenced by δ 11 B in the LR and GR rivers (without possible distinction between WWTP and animal sources), and both animal and human microbiological markers have been detected. Because the identified sources of microbiological tracers and of boron also contain high N levels, their combination with the δ 15 N and δ 18 O of nitrates offers a better understanding of the processes and pathways of nitrates to the rivers as the δ 11 B signatures of the rain and GW are significantly different from those of animal and domestic sources, they can be used for tracing and quantifying the proportions of B arising from these different groups of B sources. A simple mixing calculation (details in Methods) based on δ 11 B and [B] of three potential B sources (karstic groundwater, WWTP effluents and rainwater) was applied to each river sample. Results indicate that the atmospheric pool is the main source of boron to the rivers (50 to 96%), while groundwater brings a smaller proportion of boron (4 to 45%) and that domestic effluents (WWTP) contribute to a maximum of 10% of the total river boron (0 to 5% for GR and 0 to 10% for LR). This simplified model does not integrate the animal end-member because alkaline fusion procedure did not allow to characterize [B] arising from animal manure and fresh dejection leaching. However, considering the overlapping δ 11 B signatures of the animal and domestic effluents, it comes that at least a part of the calculated WWTP contribution could actually be of animal origin. Alternatively, the low [B] and high δ 11 B measured in GR and LR could also result from B adsorption onto soils particles before boron reached rivers with an isotopic 11 B enrichment 20,39,50 . Such a process could have blurred a larger contribution of domestic and/or animal effluents than previously deduced through the mixing calculation. B adsorption is more prone to occur within soils and GW (longer water residence time and higher rock/water ratio) than during surface leaching or within the rivers, and it is already expressed in the δ 11    contribution strongly increases. If microbiological tracers have shown that domestic and WWTP effluents are permanent contaminations for GR and LR, the denitrification signal identified by δ 15 N and δ 18 O of the GR and LR nitrates for base flow samples only associated with constant δ 11 B should rather reflect changes of NO 3 − pathways than changes of NO 3 − sources. During base flow regime, soils are not saturated, rain or irrigation water infiltrate within the soils carrying dissolved nitrate to the saturated zone where chemical and physical conditions (temperature, soil humidity, dissolved oxygen concentrations… ) and agricultural practices (N-fertilizer input periods) control the degree of denitrification. In absence of surface runoff, this shallow GW is the major source of water to GR and LR, bringing denitrified nitrate to rivers. This hypothesis is comforted by the high [NO 3 − ], δ 15 N and δ 18 O measured in the buried drains (Figs 4 and 5) but also by the close boron signatures of drain 1, LR and GR samples. On the contrary, when the soils of the catchment are saturated, runoff increases river's flows and surface leaching. Topsoil nitrates, that have not yet undergone denitrification, thus become an additional NO 3 − source for rivers. These nitrates arise, at least partially, from animal sources, as was deduced from the detection of cattle microbiological markers.

Conclusion
Isotopic and microbiological tracers have proven very powerful for the determination and characterization of contaminant sources to rivers. However, physical and chemical reactions of the N cycle within the soils blur the initial N signatures and soils become an additional source of newly produced (or transformed) nitrate, which contribution to the river depends on the hydrological stage of the catchment. That's why combining of these different tracers associated to temporal monitoring are required to explain the variations of nitrate concentrations and isotopic signatures measured in the two rivers. In the present agricultural catchment with nitrate pollution threat to the underlying karstic aquifer, anthropogenic contaminations were identified through δ 11 B measurements, microbiological tracers recorded animal dejections during high-flow stages but permanent human effluents. δ 15 N and δ 18 O of nitrate allowed to understand the N-cycle within the soils and its impact on the nitrate pathways to the rivers. It thus appears essential to monitor at least as much as for rivers themselves, the spatial heterogeneity and the temporal variability of nitrate concentrations and isotopic compositions in topsoil but also in drains and ditches in order to characterize the contribution of soil to the global nitrate content of rivers. This study also highlighted the crucial impact of hydrological conditions on nitrate contents and signature in rivers.

Methods
Study site. Located in the southwest of France, Gabas River and its tributary Laudon River, respectively drain catchment areas of 420 km 2 and 50 km 2 51 (Fig. 2). In the upstream part of its catchment, Gabas dug into sandy-clay molasses of Eo-Miocene. In the downstream part, Gabas and Laudon incise Cretaceous and Eocene karstic carbonate formations (anticline structure). Reliefs correspond to Miocene sandy formation and represent potential perched aquifers of little extension 51 . Gabas catchment and the sub-catchment of Laudon are mainly devoted to agriculture (80% of the total surface) with intensive maize cultivation and farms, evolving from cattle and pigs in the upstream part to poultry in the downstream part 51 . The two rivers are largely used for maize irrigation from April-May to August-September, depending on spring and summer rainfalls. Potential local sources of nitrate are thus, urea and manure applied on maize fields, nitrification of soil organic matter, livestock slurry and domestic effluents. The high [NO 3 − ] regularly measured in GR and LR, close to the European drinking limit of 50 mgNO 3 − .L −1 (European Directive 98/83/CE), represents a serious threat for the underlying karstic aquifer which is a strategic resource of drinking water 51 . Base flow of Gabas is around 1 to 3 m 3 /s (Fig. 3) and if it is globally inferior for Laudon with 0.5 m 3 /s, the variations of water levels are synchronous between the two streams. Flood responses are very rapid (a few hours), with maximum flows above 10 m 3 /s and 1 m 3 /s respectively for GR and LR.
Sampling strategy. Rivers water sampling has been carried out from October 2010 to January 2013. Sampling was realized at the very output of each watershed (Fig. 2) to better integrate the different point and non-point sources of nitrate occurring in the whole catchments and that potentially infiltrate to groundwater through the karstic outcrops. As agricultural activities evolve within the year whereas domestic inputs are more constant, sampling has been realized at different step of agricultural practices: before and after N inputs, under and without maize cover, in order to follow potential changes in agricultural/domestic contributions. Moreover, because biogeochemical processes affecting nitrate depend on parameters such as meteorology and hydrology, samples have been collected under different hydrological conditions: during base flow regimes and flood events (Fig. 3).
LR and GR were sampled 1 km before their confluence (Fig. 2). The lack of automatic measurements of LR flow rates forced us to use the Gabas chronicles for the interpretation of the LR data.
The different local potential sources of nitrate were collected to characterize their isotopic signature. Solid samples of fertilizers (urea pellets), cattle manure and fresh dejections (ducks) were provided by local farmers. Wastewater treatment plant effluents were sampled three times before its discharge in GR (Fig. 2). Rainwater samples have been collected in the downstream part of the GR catchment, less than 1 km from both river monitoring points (Fig. 2). An agricultural soil (maize field) located in the Laudon catchment (Fig. 2) has been sampled four times, between September 2011 and January 2013 under dry and rainy conditions and at different stages of maize growth. This type of soil is assumed to be representative of the whole Laudon's catchment and of the downstream part of Gabas catchment. Extraction of nitrate from these topsoil samples was performed to characterize their isotopic composition. To do so, soil was dried and crushed above 200 μ m, 70 g of soil were added to 140 ml of 0.5M KCl. The mixture was agitated for 2 hours (250 rotations per minute or rpm), centrifuged at 8000 rpm during 35 minutes and filtrated on 0.45 μ m nylon membrane. Finally, surface runoff and subsurface drainage have been collected from one ditch and two buried plastic drains (Fig. 2).
All water samples were filtered on 0.45 μ m nylon membrane and dispatched into three polyethylene bottles. A fraction (60 ml) of total filtered sample was stored frozen for nitrate and other major anions concentration Scientific RepoRts | 7:41703 | DOI: 10.1038/srep41703 measurement, another (60 ml) was poisoned with HgCl 2 (6%) for measurement of isotopic composition of nitrate and a third (250 mL) was acidified to pH = 2 with ultra-pure HNO 3 for analysis of boron concentration and isotopic ratio but also major cations concentrations. The physical and chemical parameters and concentrations of major anions and cations are available in a supplementary file (Supplementary Table S1 52 on an isotope ratio mass spectrometer (IRMS, DeltaVplus; Thermo Scientific, Bremen, Germany) in continuous-flow with a purge and trap system coupled with a Finnigan GasBench II system (Thermo Scientific). First step was nitrate reduction to nitrite. Sample, prepared in a salted buffer (NaCl  The [B] and δ 11 B values of the mixing end-members were measured in this study (rain and WWTP) or in a previous one (GW) 38  As discussed in the main text, the mixing doesn't take into account a possible B input from animal dejections and manure, due to the difficulty of defining an aqueous [B] for this solid end-member. However, its δ 11 B signature (8.6 and − 3.3‰) is intermediate between WWTP and river δ 11 B. Thus, the proportion of B arising from WWTP may in fact incorporate animal B input, which cannot be calculated.

Microbiological Analyses: FRNAPHs and Bacteroidales.
Microbiological samples were stored in sterile dedicated flask containing sodium thiosulphate salt (neutralizing agent effective against a wide range of oxidizing substances) used to preserve microorganism. FRNAPHs were enumerated after concentrating 1L of water sample using the membrane filtration-elution method 55 . Infectious FRNAPHs were counted (double agar-layer technique), collected, re-suspended in 1 mL of PBS with 15% glycerol and stored at − 20 °C (standard NF EN ISO 10705-1: 2001). Genotyping was performed by one-step real time, reverse transcription polymerase chain reaction (RT-qPCR) 56 . Research of Bacteroidales markers (HF183, Rum-2Bac, BacR, BacB2, Pig-1-Bac and Pig-2-Bac) was performed by filtering 1L of sample water through a 0.22 μ m pore size polycarbonate membrane. Filter was immersed in a GITC lyses solution and stored at − 80 °C. DNA extraction was performed with the Qiamp DNA minikit (Qiagen). Standard curves were calculated for plasmids containing the target sequence. PCR reactions were duplicated for each sample and measurements were performed using a Rotor gene 6000 thermocycler. The results are expressed as a number of copies in 100 mL of water.