Heterotrophic N2-fixation contributes to nitrogen economy of a common wetland sedge, Schoenoplectus californicus

A survey of the ecological variability within 52 populations of Schoenoplectus californicus (C.A. Mey.) Soják across its distributional range revealed that it is commonly found in nitrogen (N) limited areas, but rarely in phosphorus limited soils. We explored the hypothesis that S. californicus supplements its nitrogen demand by bacterial N2-fixation processes associated with its roots and rhizomes. We estimated N2-fixation of diazotrophs associated with plant rhizomes and roots from several locations throughout the species’ range and conducted an experiment growing plants in zero, low, and high N additions. Nitrogenase activity in rhizomes and roots was measured using the acetylene reduction assay. The presence of diazotrophs was verified by the detection of the nifH gene. Nitrogenase activity was restricted to rhizomes and roots and it was two orders of magnitude higher in the latter plant organs (81 and 2032 nmol C2H4 g DW-1 d-1, respectively). Correspondingly, 40x more nifH gene copies were found on roots compared to rhizomes. The proportion of the nifH gene copies in total bacterial DNA was positively correlated with the nitrogenase activity. In the experiment, the contribution of fixed N to the plant N content ranged from 13.8% to 32.5% among clones from different locations. These are relatively high values for a non-cultivated plant and justify future research on the link between N-fixing bacteria and S. californicus production.


Introduction
Growth of terrestrial as well as wetland plants in non-agricultural settings is often limited by the availability of nutrients, specifically nitrogen (N) and phosphorus (P) [1,2]. Plants have evolved two broad strategies to deal with nutrient-limiting environments: (1) conservation of use; and (2) enhanced acquisition [3,4,5]. Resorption of nutrients from senescing to newly growing or storage organs is a typical example of a conservation of use strategy. The enhanced acquisition involves production and secretion of hydrolytical enzymes such as phosphatases in case of P limitation, or the utilization of rhizosphere bacteria that can increase the bioavailability of N through N 2 -fixation [6]. Bacteria capable of performing N 2 -fixation can colonize both root surfaces (and are referred to as "epiphytes" here) as well as the internal tissues ("endophytes") (for review see [7]. The reduction of N 2 to ammonia during its biological fixation is an energetically expensive process and the input of easily available C from roots into the plant rhizosphere can sustain high activity of root associated diazotrophs [8,9,10,11]. Epiphytic diazotrophs establishing loosely organized associative relationships in the rhizosphere have been documented frequently for tropical grasses [8,12,13] and among wetland plants for littoral macrophytes [14,15,16]. Endophytic relationships of diazotrophs have been reported in a variety of plant roots and rhizomes including sugar cane, Sorghum, Miscanthus, and others [17,18,19,20,21,22]. Here we will focus on the potential role of both epiphytic and endophytic nitrogen fixation in the nutrient economy of a giant bulrush, Schoenoplectus californicus. Schoenoplectus californicus (Cyperaceae) is a large, perennial, rhizomatous wetland sedge reaching up to 6 meters in height and often forming monospecific stands [23]. Two varieties with similar ecology, var. californicus (C. A. Meyer) Soják and var. tereticulmis (Steud.) Vegetti, are present in the southern part of its distributional range. As a dominant producer of biomass, S. californicus can impact biogeochemical cycles by providing a source of organic material and by oxygenating the rhizosphere [24]. In many regions, it plays an important role in the human economy providing raw materials for the construction of boats, all-purpose mats and handicrafts [25,26]. In the survey of the ecological variability within 52 populations of S. californicus throughout the Western Hemisphere (Fig 1) we noticed that availability of phosphorus (P) appears to be important and this species is rarely found in P limited soils [23]. While common in the P-rich soils of Chile and the Central American highlands, the species is absent from the predominantly P limited ecosystems of the Yucatan peninsula and Cuba. Contrastingly, it is commonly found in areas known to be N limited, such as the Orinoco delta of Venezuela, the delta of the Paraná River in Argentina, and the Central Valley of California. In the entire data set from Carpenter [23], the average total soil P was 0.74 mg g -1 , which is well over 0.5 mg g -1 regarded as a sufficient amount of P for wetland sediments [27]. In contrast, N availability was found to be variable but generally on the low side (average total soil N of 5.5 mg g -1 ), with about half of the locations containing < 4 mg g -1 . No differences were found in biomass production among the populations from sites with N-limited sediments, nor did there seem to be less N in plant tissue (Table 1), therefore positive association with N-fixing diazotrophs was suspected.
Biological nitrogen fixation, BNF, belongs to the most essential biological processes and its knowledge is critical to our understanding of biogeochemical ecosystem functioning [28,29,30]. It occurs via two primary pathways: symbiotic and nonsymbiotic, and is performed by prokaryotic organisms called diazotrophs that can be either autrotrophic (cyanobacteria), or heterotrophic (numerous genera of bacteria such as Azospirillum, Herbaspirillum and others [31]). Symbiotic N 2 -fixation is defined as the biological reduction of N 2 occurring through mutualistic relationships between microorganisms (e.g., rhizobia) and plant roots (e.g., legumes). Nonsymbiotic nitrogen fixation includes fixation by the true free-living diazotrophs (e.g.,heterotrophic N 2 -fixation in leaf litter and soil or water) [31,32], as well as by autotrophic and heterotrophic organisms of intercellular and epiphytic growth associated with certain species of bryophytes [33]. In the past, the majority of studies dealt with the symbiotic N 2 -fixation in legumes, however more recently, the important role of free-living and epiphytic N 2 -fixation is being acknowledged and current evidence suggests that free-living N 2 -fixation represents a critical N input to many terrestrial and aquatic ecosystems, particularly those lacking large numbers of symbiotic N 2 -fixing plants [11,21]. Most of the studies on endophytic N 2 -fixation have been focused on crop plants such as sugar cane and rice, while only a few studies on natural plant populations are available and none, to our best knowledge, reports endophytic diazotrophs from any Cyperaceae family. Field experiments have demonstrated that 60% or more of plant N may be provided by plant-associated biological N 2 -fixation in sugarcane [34,35].  With the rapid development of molecular methods, many laboratories are now reporting on the actual bacterial composition of both epiphytic and endophytic diazotrophs, or at least the detection of nifH gene encoding the Fe protein polypeptide of the nitrogenase enzyme. The detection of the nifH gene in the genomic DNA sample labels that particular isolate as a potential diazotroph [36]. Expression of genes associated with N 2 fixation (nifH) has been demonstrated multiple times [37,38].
Here we report the N 2 -fixation activities associated with roots and rhizomes of a common macrophyte, Schoenplectus californicus. Our goal was to answer the following questions related to both the basic understanding of S. californicus associated N 2 -fixation processes as well as applied aspects aimed at potential economical utilization of this plant: a. Are there any differences between nitrogenase activity of roots (epiphytic) and rhizomes (endophytic) associated diazotrophs and what proportion of plant N budget has been derived from N 2 -fixation? We predicted that root associated N 2 -fixation will be higher due to the higher population densities of rhizobacteria as compared to endophytic bacteria.

Material and methods
The study has three components: 1. Preliminary survey of epi-and endophytic N 2 -fixation associated with S. californicus populations from different parts of its wide range of distribution.
2. Regional study of epi-and endophytic N 2 -fixation associated with S. californicus populations from the littoral zones of Lake Atitlán, Guatemala, assessing the variability of the two processes at nutrient poor and nutrient rich sites.
3. Controlled growth experiment with two varieties of S. californicus, originating from eight different locations throughout its distribution range, at three different N levels with a labeled 15 N source to determine the contribution of N 2 -fixation by epi-and endophytic diazotrophs to the plant's growth.
The field permit to do research at Lake Atitlán was issued by the Autoridad para el Manejo Sustenable de la Cuenca del Lago Atitlán y su Entorno (AMSCLAE); no other field permits were required.
Nitrogenase activity, delta 15 N signature of shoots and/or rhizomes, and the presence of the nifH gene (see below) were used as indicators of a fixing capability of diazotrophs associated with the respective S. californicus populations.  Table 2 also includes description of study sites around Lake Atitlán in Guatemala and few additional sites from Central California and Belize.  and Belize (P-limited) were surveyed [27] (Table 2). ARA assays were run in hours following sample collection. Component 2: Case study of Schoenoplectus californicus populations at Lake Atitlán. To assess the spatial and temporal variability of S. californicus associated N 2 -fixation and the potential impact of nutrient enrichment, we collected samples from spatially separated populations in the littoral zone of an oligo-to mesotrophic Lake Atitlán, Guatemala in August 2010 and February 2011 ( Table 2)-one of the N-limited regions included in the initial Carpenter's study [23]. Three of the locations were in the unpolluted parts of the lake, while the remaining three were at the proximity to the runoff from human settlements or other nutrient rich sources. The polluted locations were characterized by about five-and two-fold higher concentrations of total N and total P in water, respectively, compared to the unpolluted locations (polluted: TN 525 μg L -1 , TP 93 μg L -1 ; unpolluted: TN 116 μg L -1 , TP 50 μg L -1 ). The polluted locations had also higher TN and TP content in sediments ( Table 2). Component 3: Cultivation experiment. From the S. californicus clone collection, we selected five clones originating in the locations with low N and three in locations with high N ( Table 2; Note: Soil TN < 4 mg g -1 and TP < 0.5 mg g -1 in wetland sediments indicate potential N or P limitation [27]). The selection encompassed the genetic (variety tereticulmis and californicus) and geographic ranges of the species (Fig 1, Table 2). From now on the two varieties, tereticulmis and californicus, will be abbreviated as "tere" and "cali". The plants were propagated and transplanted for 5 weeks into sterilized sand and tap water (NO 3 -N = 3 ppm; PO 4 -P = 0.2 ppm). At the beginning of the experiment, individual plants comprised of a 7-12cm long shoot with a short piece of rhizome were planted in 3 replicates for each treatment in 3L pots filled with sterilized sand. Since the plants have been in the common garden cultivation for several years, there were no differences among the plants from low vs high group in either δ 15 N or % N content (δ15N 4.4 ‰ +/-0.99 SD and 4.8 ‰ +/-1.04 SD; N content 1.4% +/-0.30 SD and 1.5% +/-0.27 SD for low and high respectively). All pots received 0.25% Hoagland nutrient solution minus N. Hoagland solution was changed three times during the duration of the experiment. The zero, low, and high N treatment received biweekly 0, 20, and 200 mg/pot of N, resulting in the total addition of 0, 120, and 1200 mg N/pot (corresponding to 0, 4, and 40 g N m -2 ). Nitrogen was applied as KNO 3 with a value δ 15 N of 67.74‰. All pots were placed under ambient environmental conditions in Davis, CA, in a large tub filled with water to prevent overheating. The experiment lasted 87 days (June 3-August 29; during this time, there is typically no rainfall in the Central Valley of California, thus there were no concerns regarding potential uncontrolled N addition by rain). At the end of the experiment, the cumulative length of shoots was recorded, shoots, rhizomes, and roots were separated, washed in DI, and the aliquots were used for measurement of nitrogenase activity. The remaining samples were freeze-dried, weighed, and ground for subsequent determination of δ 15 N, total N, and molecular analyses for nifH gene presence and abundance among samples.

Collection of samples for nitrogenase activity measurements
Four plants at each location (components 1 and 2) and one plant per pot (component 3) were extracted carefully from sediments to prevent extensive root damage and roots and rhizomes were rinsed in surface water to remove adhered larger particles and sealed in a plastic bag. The samples were transported to the laboratory, where the live roots were identified by color and structure, rinsed in distilled water, and an equivalent of~20-40 mg DW was transferred to 40 ml glass test tubes with three replicates per plant. A preliminary test was conducted to verify that this sample manipulation did not impact nitrogenase activity. In the test we compared ARA in root samples treated as described above with those that were collected, quickly rinsed in deoxygenated water and immediately placed to the fixation vials where the oxygen was lowered by exchange of part of the headspace for N 2 gas. There were no significant differences in the nitrogenase activity measured by ARA for the two treatments over 2, 5, 10, 18 and 25 hours of incubation (t-test, P = 0.7). Rhizome and stem tissues were surface sterilized sequentially in sterile distilled water, 95% ethanol, and 1.6% hypochlorite for 30s each; between each step the tissues were rinsed three times with sterile water [39]. The outer, coarse surface layer of rhizomes was removed after surface sterilization, leaving only tissue that did not come in contact with the soil. Because of the complexity of the root structure, it is difficult to reliably surface sterilize these organs; we decided to consider the root-associated nitrogenase activity as a result of joint activity of both the endo-and epiphytic microorganisms. The tissue was transferred to 40 ml glass test tubes (each in three replicates) with 200 μL (for rhizomes) and 20 mL (for roots) of distilled water and fitted with Teflon septum lids.

Nitrogenase activity
The acetylene reduction technique, ARA [40], was employed to estimate N 2 -fixation by the reduction of acetylene to ethylene by nitrogenase. Ten percent of the headspace were replaced with acetylene gas, freshly generated from calcium carbide, and the bottles were incubated for 24 hours at 28 degrees C. At the end of the exposure, 7-8 mL of headspace was withdrawn with an airtight syringe (Alltech) and analyzed by gas chromatograph (Shimadzu 14 GC) with a flame ionization detector and a Porapak-T column at 80˚C. The results are reported as the nitrogenase activity in nmol C 2 H 4 g -1 d -1 of dry weight. Controls run as samples without acetylene addition as well as blanks (tubes without plant tissue incubated with acetylene) showed no endogenous ethylene production. Samples were kept for dry weight determination after terminating the exposure (see [11] for more detailed description).

Calibration of ARA through 15 N 2 reduction assay
On a subset of 13 root and 7 rhizome samples from the experiment, the nitrogenase activity, measured using the ARA, was calibrated by 15 N 2 reduction assay. The measurements were conducted at the same time as ARA but 2 mL of 15 N 2 (99atom%, Cambridge Isotope) was added instead of C 2 H 2 . At the end of the incubation, the content of the containers was frozen, freeze dried and then ground in a Wiley mill. The initial 15 N natural abundance of the sample was determined from the ARA samples. The delta 15 N (in relation to atmospheric N 2 as the reference standard material) was measured by an isotope ratio mass spectrometer (see below). Biomass specific N 2 fixation rate, normalized to organic N, was calculated as isotopic balance [41,42]: Where PN is N concentration in the sample, AP is 15 N enrichment (atom% 15 N) of the sample or substrate (N 2 ) pool at the beginning (initial) and end (final) of incubation; Δt is the length of incubation. N 2 fixation rate expressed in terms of fixation of molecular N 2 to organic material was then calculated:

Isotope and tissue nutrient analyses
Tissue N and P concentration of shoot and rhizome samples were assessed. Stable isotopes of N were measured by continuous flow isotope ratio mass spectrometry using a PDZ Europa ANCA-GSL elemental analyzer interfaced to a PDZ Europa 20-20 isotope ratio mass spectrometer (Sercon Ltd., Cheshire, UK). Dried samples containing approximately 20-150 μg N (2-3mg of sample) were packaged in tin capsules (Elemental microanalysis, Manchester, MA) and combusted at 1000˚C in the elemental analyzer. The ratio of 15 N/ 14 N (R15) was measured for the sample and for an injection of standardized N 2 gas introduced into the mass spectrometer in each sample cycle. The δ 15 N was calculated from: and expressed on "per mil" basis. Total P was measured spectrophotometrically using ascorbic acid reduction of phosphomolybdate complex after combustion and consequent acid digestion [43].

Proportion of N derived from N 2 -fixation
Percent of N derived from N 2 -fixation was calculated for the experiment using the two-endmember linear mixing model [44] formulated from the mass balance equations: where δ M , δ A and δ B represent the mean isotopic signatures of the mixture M and sources A and B respectively, in our case mixture M was the isotopic signature of plants at the end of the experiment, source A had the isotopic signature of 67.74‰ (δ 15 N of the nitrate-N); source B was 0‰ (δ 15 N of the N from the air). For the calculation of the means and confidence intervals we used the ISOERROR 1.04 Excel spreadsheet by Phillips and Greg [44].

Quantifying the presence of diazotrophs in rhizomes and roots by qPCR
The presence of diazotrophs was assessed in Atitlán rhizome samples from 2011 and in selected root and rhizome samples from the cultivation experiment. Total DNA was extracted from Schoenoplectus rhizomes or roots using the Power Soil DNA Isolation Kit (Mo Bio Laboratories), according to manufacturer's instructions. DNA quality and quantity was determined by electrophoresis and total DNA concentration in the samples was measured fluorometrically. Total bacteria (16S rDNA gene) were quantified using universal primers 341f (CCT ACG GGAGGCAGCAG) and 515r (ATTCCGCGGCTGGCA) as described by [45]. The qPCR reactions were set up using the FastStart Universal SYBR Green Master Mix (Roche). First denaturation at 95˚C for 10 min was followed by 30 cycles of denaturation (95˚C, 45s), annealing (60˚C, 30s), and extension (72˚C, 30s). For the quantification of diazotrophs (nifH) the IGK3 (GCIW THTAYGGIAARGGIGGIATHGGIAA) and DVV (ATIGCRAAICCICCRCAIACIACRTC) primers were used [46]. First denaturation at 95˚C for 10 min was followed by 40 cycles of denaturation (95˚C, 15s), annealing (58˚C, 30s), and extension (72˚C, 60s). Standard curves were obtained with serial 10 fold dilutions of a known amount of amplicon of the 16S rDNA and nifH genes, respectively. Amplicons were prepared from genomic DNA of E.coli for 16S RDNA gene and from Methylocystis heyeri for nifH gene. Each extraction, no-template control, and standard curve dilution was replicated three times. Average copy number per μl of reaction qPCR mixture was converted into copies of the gene per ng of total extracted DNA. Standard deviation was determined by the StepOne Software v2.3 (Thermo Fisher Scientific). According to threshold cycles (C T ) of standards and the NTC values, a detection limit of approximately 10 to 100 gene copies per assay was achieved for nifH and 16S rDNA quantification, which corresponds to 10 2 to 10 3 gene copies per gram of dry rhizome or root dry weight.

Data analyses
Because most of the data sets exhibited variance heterogeneity (F max -test; [47]), we used nonparametric Kruskal-Wallis Test (Statview 5 software package), to evaluate the effect of different treatments and the effects of clones (separate for each treatment) and Schoenoplectus variety (separate for each treatment) on response variables. For the same reason, non-parametric Mann-Whitney U-test was used instead of Student's t-test to test the difference between means of two samples.

Preliminary survey
Tests of N 2 -fixation capability of epiphytic and endophytic diazotrophs, providing preliminary data for future research revealed that four randomly selected Schoenoplectus californicus clones from our clone collection, as well as the same species from two locations in the Central Valley of California did display nitrogenase activity. The nitrogenase activity for rhizomes ranged from 13.0 to 40.9 nmol C 2 H 4 gDW -1 d -1 , while root-associated activity varied from 523 to over 2000 nmol C 2 H 4 gDW -1 d -1 (Table 3). No nitrogenase activity was found in any of the shoots (data not shown). Schoenoplectus acutus and Typha domingensis both showed nitrogenase activity related to roots, but contrary to S. californicus, we did not find any nitrogenase activity in rhizome tissue of these two species.

Case study of BNF associated with Schoenoplectus californicus populations at Lake Atitlán
Nitrogenase activity of both rhizome and root associated diazotrophs was found in all tested samples (Table 4). Due to the large variability, only the endophytic rhizome fixation was  Table 4). The nitrogenase activity associated with roots was on average more than three times higher in the unpolluted zones, although this difference was not statistically significant. There did not seem to be any trend related to time (August 2010 -rainy season vs February 2011 -dry season). No correlation was found between rhizome and root N 2 -fixation. Delta 15 N of shoots ranged from 2.0 ‰ to -8.2 ‰ averaging -2.1 ‰; the rhizome δ 15 N ranged from 1.6 ‰ to -5.8 ‰ averaging -1.3 ‰ (Table 4). There was correlation between rhizome nitrogenase activity and rhizome δ 15 N (R 2 = 0.69; P = 0.04; Fig 2A), but not the shoot δ 15 N; this despite the fact that shoot δ 15 N and rhizome δ 15 N were well correlated (R 2 = 0.9; P = 0.004; Fig 2B).
Diazotrophs (based on the nifH gene presence) were detected in all analyzed samples ( Table 5). The proportion of nifH gene copies among the total bacterial DNA was low, but well above the method detection limit. It varied from 0.01 to 0.04% (Table 5). While there was a trend of increasing nitrogenase activity with increasing nifH proportion, the correlation was not significant, however, it became significant when all availbale data were included (see the text in the Cultivation experiment subchapter).

Cultivation experiment
Growth response to nitrogen treatments. Biomass of shoots, rhizomes, and roots as well as cumulative shoot length was recorded at the end of the experiment (87 days). In each treatment, all these response variables were closely correlated (R 2 >0.9), thus we present total dry mass only ( Table 6). As expected, the effect of the N treatment was highly significant; the plants in zero N treatment grew very slow with the average biomass per pot of 8.8 g, however, they did form new healthy shoots. The low N and high N treatments resulted in correspondingly higher biomass with 14.1 and 54.4 g per pot, respectively. The tere variety produced Table 4 significantly less biomass than cali variety in each of the N treatments. Plants in zero N treatment invested more growth effort into roots as documented by a low shoot/root ratio of 0.69 as compared to 1.63 and 2.68 for low N and high N treatments respectively. (Note, that "root" in S/R ratio stands for both roots and rhizomes). The tissue N content was low in zero N and low N treatment, 0.60 and 0.85% respectively, and significantly higher, 1.71%, for high N treatment. In low and high N treatment, the tere variety contained more N in shoot tissue than cali. Nearly all the N added to the low N treatment pots was recovered in the biomass of plants in the low N treatments. Response of nitrogenase activity to nitrogen treatments. We found evidence of endophytic (rhizome) N 2 -fixation in all treatments and all clones. The activity fluctuated considerably, ranging from 22.7 nmol C 2 H 4 g DW -1 d -1 to 460 nmol C 2 H 4 g DW -1 d -1 (Table 6, Fig 3). There were no differences between zero and low N treatments because of the large variability, Relationship between rhizome nitrogenase activity and rhizome δ 15 N (A) and shoot and rhizome δ 15 N (B) in Schoenoplectus californicus from Lake Atitlán; full diamond-unpolluted sites, empty diamonds-polluted sites. Each nitrogenase activity value is a mean of 3-4 replicates, isotope data were measured on pooled samples. Nitrogenase activity expressed as ethylene production (nmol C 2 H 4 gDW -1 d -1 ).

. Means ± standard deviations of nitrogenase activity of epiphytic (roots) and endophytic (rhizomes) diazotrophs measured as ethylene production (nmol C 2 H 4 gDW -1 d -1 ), shoot and rhizomes δ 15 N, and tissue N and P of Schoenoplectus californicus from unpolluted and polluted locations in the littoral of Lake
https://doi.org/10.1371/journal.pone.0195570.g002 Heterotrophic N 2 -fixation and nitrogen economy of Schoenoplectus californicus but rhizome nitrogenase activities in low and high N treatment were closely correlated (R 2 = 0.81; P = 0.002). Rhizome fixation was significantly lower in the tere variety. Root (epiphytic) nitrogenase activity was also highly variable and did not differ between zero and low N treatment, while it was on average about 4x higher in high N treatment. Root nitrogenase activity was positively correlated among all three treatments. Endo-and epiphytic fixations were not correlated except for the low N treatment (positive correlation; R 2 = 0.51; P = 0.05).

Proportion of N from N 2 -fixation (N dfa ).
For the low N and high N treatments, we used the mixing model to calculate the proportion of N derived from N 2 -fixation (N dfa ). In the low N treatment, we found differences in the contribution to the plant N content among clones Table 5 The presence of nifH genes is expressed as copies per ng DNA or as a proportion of diazotrophs (nifH gene copies) in the total bacterial DNA pool (%).

Roots/ Rhizomes
https://doi.org/10.1371/journal.pone.0195570.t005 Heterotrophic N 2 -fixation and nitrogen economy of Schoenoplectus californicus from different locations ranging from 13.8% for Guillermo to 32.5% for San Juan (Fig 4). Variety cali utilized the N from N 2 -fixation significantly more than tere (N dfa 26.3% and 18.6%, respectively; Mann-Whitney U-test; P = 0.01). In the high N treatment the N dfa was very low (0.5 to 6.4%) and the differences between clones were not significant, although San Juan location, SJ, had the highest proportion of N from N 2 -fixation in both treatments. No differences between varieties were found in the high N treatment.  Similarly to the Atitlán case study, there was a negative correlation between rhizome nitrogenase activity and rhizome δ 15 N in both zero and high N treatments (Fig 5). Not enough data on rhizome δ 15 N were available to calculate correlations for low N treatment. Rhizome and shoot δ 15 N values were significantly positively correlated in zero treatment (R 2 = 0.66; P = 0.05), while non-significant positive trend was found in high N treatment (R 2 = 0.43; P = 0.15).
On average 40x more nifH gene copies were found in the root material than in the rhizomes (Table 5). When the nitrogenase activity was plotted against the proportion of nifH gene for all samples (rhizomes from the Atitlán case study and roots and rhizomes from the cultivation experiment), the correlation was highly significant (Fig 6).
Calibration of ARA through 15 N 2 reduction assay. In both rhizomes and roots, the nitrogenase activity measured by ARA was correlated closely to the 15 N 2 reduction (rhizomes: y = 2.785x; r 2 = 0.98, roots: y = 2.38x; r 2 = 0.87) denoting that the C 2 H 4 reduction: N 2 reduction ratio was 2.78: 1 and 2.38: 1 for rhizomes and roots respectively.

Discussion
Our hypothesis that endophytic and epiphytic diazotrophs are associated with rhizomes and roots of Schoenplectus californicus throughout its distributional range and that the plant is capable of utilizing the fixed N has been confirmed by several lines of evidence (nitrogenase activity, presence of nifH gene, and δ 15 N data).
There is very little information available on N 2 -fixation associated with members of the Cyperaceae family and none that would relate to endophytic N 2 -fixation. Rhizosphere diazotrophs associated with Schoenoplectus americanus, a close relative of S. californicus, were reported to fix 367 ± 46 ng 15 N per plant per hour [9]. A direct comparison is difficult because of differences in the experimental setting, but using our C 2 H 4 : N 2 ratio of 2.38 (see results) and shoot: root ratio of~1 [9], their value would translate to some 20,000 nmol C 2 H 4 g -1 d -1 , i.e., an order of magnitude higher value compared to our data. This is potentially explainable by the fact that Dakora and Drake used intact plants for measurements and thus diazotrophs had an unrestricted access to root carbon exudates (see below). Root associated diazotroph activity in the same range as Schoenoplectus californicus has been reported for Cyperus papyrus, a dominant species of tropical swamps in Africa [16]. Eleocharis spp. from marshes in Belize displayed root associated N 2 fixation on the order of 3000 to 4000 nmol C 2 H 4 g -1 d -1 under conditions unlimited by carbon [11]. As for the wetland plants other than Cyperaceae, Eckard and Biesboer [14] reported the nitrogenase activity of 217 and 226 C 2 H 4 g -1 d -1 for Typha latifolia and T. angustifolia, respectively, and they concluded that populations of Typha may receive as little as 1-2% of their annual N requirement from N 2 -fixation.

Are there any differences between N 2 -fixation of roots (epiphytic) and rhizomes (endophytic) associated diazotrophs?
Our prediction that the root associated N 2 -fixation will be higher than endophytic N 2 -fixation was correct. In the Atitlán data set as well as in the cultivation experiment, the epiphytic N 2fixation was 6 to 60 times higher than the endophytic N 2 -fixation. This is in agreement with the general consensus that endophytic bacteria occur at lower population densities than surface associated epiphytic bacteria [22]. This difference can also result from the way we present the results, i.e., as nitrogenase activity per grams of dry weight. Because of their bulky structure, Schoenoplectus rhizomes have relatively large proportion of a "ballast" biomass, while fine roots provide a large surface area for diazotrophs to attach to. In the case of S. californicus with its extensive root structure, it seems possible that the plants benefit more from the epiphytic bacteria, but at this point it is still just a speculation. We assume that the diazotrophs in the root samples are epiphytic, but since we did not sterilize the roots, we cannot really exclude the possibility of endophytic root diazotrophs being present.

The importance of carbon limitation
The presence of a constant C supply as energy source is an important criterion to be satisfied for the diazotrophs [10,48]. It is quite probable that the N 2 -fixation of both rhizomes and roots in our measurements was underestimated, because in our experimental setting the diazotrophs on excised roots/in rhizomes did not have access to the natural and continuous input of C. A recent trial showed that, with the addition of glucose, the root fixation was on average 6x Heterotrophic N 2 -fixation and nitrogen economy of Schoenoplectus californicus higher (Rejmánková, unpublished data). Dalton et al. [37] reported a 9-fold increase in the nitrogenase activity associated with roots of Elymus mollis after the roots have been immersed in 1% glucose prior to the incubation treatment. Lower N 2 fixation was reported on excised roots of Cyperus papyrus compared to fixation on intact roots [16] and the C limitation has been further demonstrated by other authors [22,49]. Based on the increased N 2 fixation caused by the enhanced photosynthate supply to root-associated diazotrophs, Dakora and Drake [9] suggest that, as in legumes, a direct relationship exists between C supply and N yield from N 2 -fixation associated with the Cyperaceae. The relationship between C exudates and the diazotroph activity associated should be explored further.

N dfa
Estimates of nitrogenase activity provide important information but it may not reflect the amount of N actually utilized by plants. The proportion of N a plant obtains from N 2 -fixation is more meaningful. In our low N addition treatment, the N dfa ranged from 13.8% to 32.5% and was higher in the cali variety as compared to the tere (26.3% and 18.6%, respectively). The available data on N 2 -fixation contribution vary greatly and generally seem to be higher in plants artificially inoculated with a single strain or a mix of N-fixing bacteria. Field studies with "wild", non-inoculated plants report N 2 -fixation contributions ranging from 0 to 30% [13,50]. The "engineered" plants, on the other hand, often obtain more than 50% of N from BNF (e.g., sugarcane: 34.8-58.8% [51]; rice variety BAS-370 close to 70% [52]; Pennisetum 50% [12]; poplar 65% [20], etc.). Presently, S. californicus is an important economical plant in several regions, where it is used for boat construction or/and mat weaving [26], but the demand is still covered by natural production. Should the larger demand for this species occur, inoculation trials with endophytic diazotrophs may become important.

Constitutive endophytic N 2 -fixation
N 2 -fixation with its very high energy requirement should, theoretically, down-regulate where there is high N availability in the soil [29]. In our experiment, we surprisingly found a larger nitrogenase activity in both rhizomes and roots in high N treatments than in zero N treatment. This could be potentially explained by localized depletion of available N in parts of the rhizosphere. Towards the end of the experiment, plants in high N treatment were growing vigorously and some could have used most/all of added N. This would activate diazotrophs, which would likely perform well due to high quality carbon released as exudates from well growing plants. It would be in principle similar to what Hedin et al. [53] call the nitrogen paradox, where in the case of tropical forests, BNF from free living diazotrophs can occur in N-limited areas of the forest separated from areas of abundant N, allowing N 2 fixation to continue in these specific areas, despite the ecosystem being N-rich as a whole. Field data from Atitlán indicated that there was a trend towards lower BNF in the rhizosphere diazotrophs and a significantly lower BNF in the endophytic diazotrophs among populations from polluted areas. This confirmed our prediction that the N 2 -fixation will be higher in the unpolluted areas, while in the enriched locations, plants will be utilizing N available in sediments.

Are there any differences in N 2 -fixation between the two varieties of S. californicus?
The finding that S. californicus var. californicus consistently displays higher nitrogenase activity as well as higher N dfa than the variety tereticulmis was unexpected. Considering that the clones used for this experiment had been in cultivation in Davis for several years prior to the experiment, the result could not have been due to priming of the original locations. It demonstrates that the taxonomic identity can have an important effect on the N acquisition strategy and is in agreement with Wurzburger and Hedin's [54] finding that taxonomic identity was the major determinant of fixation across a broad soil P gradient in lowland tropical forests, and that soil nutrients had no or only limited influence on N 2 fixation. The fact that one of the varieties seems to be capable of utilizing larger proportion of N from BNF may become important if there ever is a need to increase the production of this species.

Is the presence of nifH gene a good predictor of N 2 -fixation?
The presence of nifH gene means that that bacterial community has the potential to perform N 2 -fixation, not necessarily that the process itself is present. Although the proportion of nifH gene copies among the total bacterial DNA was relatively low, in the samples from the cultivation experiment, the nifH gene quantity in both the root surface associated and endophytic bacteria was well correlated with the nitrogenase activity measured. The correlation was much weaker in (a much smaller) data set from the Atitlán case study. However, this correlation need not always be strong, because nifH genes are also present in DNA of inactive or non-living microbes (for example [55]). To the extent that nifH gene copy number reflects diazotroph abundance, the S. californicus strategy may be that of maintaining relatively low abundances of highly efficient diazotrophs on their root systems. Our data suggest that many of the diazotrophs on S. californicus roots or rhizomes were indeed alive and active, despite the fact that, in both types of experiments, they likely experienced a limitation by available plant-derived C. Measurements of N 2 -fixation using excised roots may underestimate the activity on intact roots. Unfortunately, other methods of assessing N 2 -fixation activity on intact roots are too technically challenging to be used routinely or under field conditions. Because of these limitations, the nifH gene does seem to be a good additional marker for the presence of N 2 -fixation in S. californicus associated bacteria. Considering the difference among the two S. californicus varieties, it remains to be assessed how the bacterial community composition varies among them, as well as how it varies between the endophytic and epiphytic bacteria.

What can we learn from stable isotope signatures?
How does δ 15 N signature support our hypothesis that S. californicus utilizes N from BNF by its associated bacteria? The best answer comes from the field data from Atitlán, where the δ 15 N of S. californicus shoots averaged -2.1 ‰ (Table 4), as compared to the other, presumably non-fixing lake littoral species ranging from 3.4 ‰ (Typha domingensis) and 4.6 ‰ (submersed spp.) to 4.9 ‰ (Salix humboltiana) (Rejmánková, unpublished data). The average littoral sediment δ 15 N of 3.4 ‰ is close to the values measured in the non-fixing species. Similarly, in the field collection of the S. californicus clones, the δ 15 N clones from low N locations averaged 0.6 ‰, while in the N rich locations it was 12.0 ‰ (Table 2). Although, as many authors pointed out, the assumption that the δ 15 N of leaf tissues reflects that of the N source in the soil is not always valid [56,57], plant isotope composition is more likely to reflect that of the N source when plant demand exceeds N supply [58],-this is clearly the case of N-limited wetlands throughout the S. californicus distributional range.

Potential N budget
Data from our experiment demonstrated that in the low N addition treatment on average 24% N was derived from N 2 -fixation. For the SJ clone originating at Lake Atitlán, this proportion was 34%. How does this agree with a budget calculated from endophytic and epiphytic nitrogenase activity using Atitlán originated material? The average nitrogenase activities for plants from nutrient un-enriched locations were 37.1 and 2567.7 nmol C 2 H 4 gDW -1 d -1 , respectively.
If we assume the average biomass of 400 g m -2 , 300 g m -2 , and 300 g m -2 for the aboveground stems, rhizomes, and roots, respectively (the biomass proportion from Castle, unpublished data), the longevity of plants one year, and the tissue N of 1.8%, then the plant would require 18g of N m -2 y -1 . With the C 2 H 4 reduction: N 2 reduction ratio of 2.8: 1 and 2.4: 1 for rhizomes and roots, respectively (see results), the contribution of rhizosphere and endophytic N 2 -fixation would represent 19%. This value is lower than N dfa from the experiment and this is assuming all fixed N is available to plants. Several reasons may be responsible: the longevity of plants may be higher; we may have underestimated the fixation by measuring it on excised roots (see above); and the conditions during the experiment in Davis may have been more favorable.

Conclusion
Although the absolute contribution of N 2 -fixation is difficult to determine, our results show that the N budget of S. californicus is substantially subsidized by fixed N. In support of this, there have been multiple observations that throughout its range, in the areas heavily impacted by sewage inflow, Schoenoplectus is being outcompeted by Typha domingensis (Rejmánková, unpublished data). The S. californicus system represents a suitable model for future studies on the effects of non-symbiotic N 2 fixation on the geographical distribution of plant species and varieties, plant physiology, or inter-species competition. As S. californicus is also an important plant for many native communities throughout Central and South America and is a species commonly used in constructed wetlands and wastewater treatment, the information presented in this paper may also help to improve its more applied functional roles.