Symbiosis revisited: phosphorus and acid buffering stimulate N 2 fixation but not Sphagnum growth

Abstract. In pristine Sphagnum-dominated peatlands, (di)nitrogen (N2) fixing (diazotrophic) microbial communities associated with Sphagnum mosses contribute substantially to the total nitrogen input, increasing carbon sequestration. The rates of symbiotic nitrogen fixation reported for Sphagnum peatlands, are, however, highly variable, and experimental work on regulating factors that can mechanistically explain this variation is largely lacking. For two common fen species (Sphagnum palustre and S. squarrosum) from a high nitrogen deposition area (25 kg N ha−1 yr−1), we found that diazotrophic activity (as measured by 15 − 15N2 labeling) was still present at a rate of 40 nmol N gDW−1 h−1. This was surprising, given that nitrogen fixation is a costly process. We tested the effects of phosphorus availability and buffering capacity by bicarbonate-rich water, mimicking a field situation in fens with stronger groundwater or surface water influence, as potential regulators of nitrogen fixation rates and Sphagnum performance. We expected that the addition of phosphorus, being a limiting nutrient, would stimulate both diazotrophic activity and Sphagnum growth. We indeed found that nitrogen fixation rates were doubled. Plant performance, in contrast, did not increase. Raised bicarbonate levels also enhanced nitrogen fixation, but had a strong negative impact on Sphagnum performance. These results explain the higher nitrogen fixation rates reported for minerotrophic and more nutrient-rich peatlands. In addition, nitrogen fixation was found to strongly depend on light, with rates 10 times higher in light conditions suggesting high reliance on phototrophic organisms for carbon. The contrasting effects of phosphorus and bicarbonate on Sphagnum spp. and their diazotrophic communities reveal strong differences in the optimal niche for both partners with respect to conditions and resources. This suggests a trade-off for the symbiosis of nitrogen fixing microorganisms with their Sphagnum hosts, in which a sheltered environment apparently outweighs the less favorable environmental conditions. We conclude that microbial activity is still nitrogen limited under eutrophic conditions because dissolved nitrogen is being monopolized by Sphagnum. Moreover, the fact that diazotrophic activity can significantly be upregulated by increased phosphorus addition and acid buffering, while Sphagnum spp. do not benefit, reveals remarkable differences in optimal conditions for both symbiotic partners and calls into question the regulation of nitrogen fixation by Sphagnum under these eutrophic conditions. The high nitrogen fixation rates result in high additional nitrogen loading of 6 kg ha−1 yr−1 on top of the high nitrogen deposition in these ecosystems.

Abstract. In pristine Sphagnum-dominated peatlands, (di)nitrogen (N 2 ) fixing (diazotrophic) microbial communities associated with Sphagnum mosses contribute substantially to the total nitrogen input, increasing carbon sequestration. The rates of symbiotic nitrogen fixation reported for Sphagnum peatlands, are, however, highly variable, and experimental work on regulating factors that can mechanistically explain this variation is largely lacking. For two common fen species (Sphagnum palustre and S. squarrosum) from a high nitrogen deposition area (25 kg N ha −1 yr −1 ), we found that diazotrophic activity (as measured by 15−15 N 2 labeling) was still present at a rate of 40 nmol N gDW −1 h −1 . This was surprising, given that nitrogen fixation is a costly process. We tested the effects of phosphorus availability and buffering capacity by bicarbonate-rich water, mimicking a field situation in fens with stronger groundwater or surface water influence, as potential regulators of nitrogen fixation rates and Sphagnum performance. We expected that the addition of phosphorus, being a limiting nutrient, would stimulate both diazotrophic activity and Sphagnum growth. We indeed found that nitrogen fixation rates were doubled. Plant performance, in contrast, did not increase. Raised bicarbonate levels also enhanced nitrogen fixation, but had a strong negative impact on Sphagnum performance. These results explain the higher nitrogen fixation rates reported for minerotrophic and more nutrient-rich peatlands. In addition, nitrogen fixation was found to strongly depend on light, with rates 10 times higher in light conditions suggesting high reliance on phototrophic organisms for carbon. The contrasting effects of phosphorus and bicarbonate on Sphagnum spp. and their diazotrophic communities reveal strong differences in the optimal niche for both partners with respect to conditions and resources. This suggests a trade-off for the symbiosis of nitrogen fixing microorganisms with their Sphagnum hosts, in which a sheltered environment apparently outweighs the less favorable environmental conditions. We conclude that microbial activity is still nitrogen limited under eutrophic conditions because dissolved nitrogen is being monopolized by Sphagnum. Moreover, the fact that diazotrophic activity can significantly be upregulated by increased phosphorus addition and acid buffering, while Sphagnum spp. do not benefit, reveals remarkable differences in optimal conditions for both symbiotic partners and calls into question the regulation of nitrogen fixation by Sphagnum under these eutrophic conditions. The high nitrogen fixation rates result in high additional nitrogen loading of 6 kg ha −1 yr −1 on top of the high nitrogen deposition in these ecosystems.

Introduction
Nitrogen (N) availability is considered to limit or co-limit primary production in pristine Sphagnum-dominated ecosystems (Aerts et al., 1992;Lamers et al., 2000;Limpens and Berendse, 2003). Peat mosses (Sphagnum spp.) function as a filter that very effectively absorbs particularly ammonium (NH + 4 ) but also nitrate (NO − 3 ) from atmospheric deposition, leading to N limitation in the rhizosphere of vascular plants (Lamers et al., 2000;Bragazza et al., 2004;Fritz et al., 2014). Since the availability of N determines primary production, there appears to be a close link between the N and C cycles (Hungate et al., 2003;Vitousek et al., 2013). This link is especially important in peatlands, which, by storing substantial amounts of C, play an important role in global C cycling (Ruesch and Gibbs, 2008;Clymo and Hayward, 1982). Being ecosystem engineers in peatlands, Sphagnum spp. produce recalcitrant litter, rich in phenolic compounds (Verhoeven and Toth, 1995), and actively acidify their environment (Clymo and Hayward, 1982). This, combined with moist, anaerobic conditions results in the accumulation of peat with a high C content (Van Breemen, 1995). Recently, it has been shown that the high N 2 fixation activity of the Sphagnum microbiome could explain the discrepancy between low inputs of atmospheric N and high N accumulation rates in the peat of pristine Sphagnum peatlands (Vile et al., 2014), confirming the strong link between C and N accumulation. On the other end, high atmospheric N deposition may compromise the C sequestration function of peatlands by stimulating microbial processes such as overall decomposition (Bragazza et al., 2006) and denitrification (Gruber and Galloway, 2008). N 2 fixing microorganisms (diazotrophs) live on the surface and inside dead hyaline cells of Sphagnum (Opelt et al., 2007;Bragina et al., 2012;Larmola et al., 2014), forming a symbiosis with their host. A highly diverse microbial community, including Proteobacteria, Verrucomicrobia and Cyanobacteria, has been found to colonize peat mosses (Bragina et al., 2014), and many of these microorganisms have the capacity to fix N 2 (Bragina et al., 2013;Kox et al., 2016). Also, in other bryophytes, like Hylocomiaceae (feather mosses), such a symbiotic relationship can be found with N 2 fixing cyanobacteria, supplying up to 50 % of the total N input in boreal forests (Rousk et al., 2013). These phototrophic diazotrophs provide N to their host in exchange for C compounds (Bay et al., 2013;Leppänen et al., 2013), a process that we refer to as a direct mutualism, with reference to the direct transfer of chemicals between host and symbiont (Ho and Bodelier, 2015). In these moss symbioses, as well as in vascular plant symbioses, application of high rates of inorganic N were found to decrease N 2 fixation rates, with the host plant shifting to the use of this readily available inorganic N source (Gundale et al., 2011;Zackrisson et al., 2004;Rousk et al., 2014). There may also be a different, indirect type of interaction in which Sphagnum receives a flow of nutrients from dead and lysed microorganisms. Although the exact nature of the Sphagnum-microorganism symbiosis remains unknown, i.e., a direct mutualism or an indirect interaction, N fixed by cyanobacteria associated with Sphagnum was found to enhance Sphagnum growth . A high variation in rates of N 2 fixation has been found not only for different species and different systems, but also for similar ecosystem types at different locations. To our knowledge, the mechanistic explanation for this high variation of symbiotic N 2 fixation rates in Sphagnum peatlands is still lacking.
In areas with high N deposition like in our field site in the Netherlands, the necessity for microorganisms with diazotrophic capacity to actually fix N 2 can be expected to diminish, as NH + 4 availability usually leads to down-regulation of the expression of the nitrogenase enzyme responsible for N 2 fixation (Dixon and Kahn, 2004). Nutrients other than N have been suggested to influence N 2 fixation, especially phosphorus (P) (Vitousek and Field, 1999), which is generally the second nutrient limiting primary production (Bieleski, 1973;Vance, 2001). P limitation has been shown to play an important role in biomass growth and functioning of peatlands (Larmola et al., 2013;Hill et al., 2014;Fritz et al., 2012) and appeared to control N 2 fixation rates (Toberman et al., 2015;Vitousek et al., 2002;Chapin et al., 1991). Besides, isolated cyanobacteria were shown to be directly stimulated by P (Mulholland and Bernhardt, 2005) and, in Azolla spp., a fern species with symbiotic cyanobacteria within its leaves, P was shown to drastically increase plant growth and N content (Cheng et al., 2010). In peat mosses from N-rich sites, increased P availability can be expected to complement the high N supply  and lead to an increase in photosynthesis (by 14 %) (Fritz et al., 2012) and moss growth (by 42 %) (Carfrae et al., 2007). It is therefore expected that the addition of P will improve the performance of the Sphagnum-microorganism association in high N deposition areas.
Next to nutrient availability, the alkalinity and pH of the environment is known to be a key biogeochemical factor affecting Sphagnum presence and performance in peatlands (mires). Higher concentrations of bicarbonate (HCO − 3 ) and concomitantly higher pH values (from 7.5 and upwards), through the influence of minerotrophic groundwater or surface water in rich fens, have been shown to hamper Sphagnum growth (Clymo, 1973;Lamers et al., 1999). While the effect of environmental factors such as pH and nutrient availability on Sphagnum itself has been thoroughly studied (Clymo, 1973;Kooijman and Paulissen, 2006;Bragazza and Gerdol, 2002), it remains unknown how these environmental factors influence the activity of its diazotrophic community and how this in turn affects Sphagnum performance in peatlands. Information about the factors regulating the diazotrophic community is vital to understanding the high variation in N 2 fixation rates in Sphagnum-dominated wetlands that may strongly affect both nutrient and carbon cycling.
We therefore used a controlled, full-factorial setup to experimentally test the effects of P and HCO 3 addition on N 2 fixation rates of the diazotrophic community and on photosynthesis and growth of two common fen species, Sphagnum squarrosum Chrome and S. palustre L., from a Dutch poor fen. Our prime research question was whether P availability and alkalinity were key regulators of both diazotrophic and Sphagnum activity, with P increase having a positive effect on both partners and alkalinity increase a negative effect. In addition, in view of a direct mutualistic relationship between the moss and its diazotrophs, as with Azolla spp. and its cyanobacteria, we expect that higher N 2 fixation rates will provide additional N. Combined with higher P availability, this may increase Sphagnum photosynthesis and growth even further, as long as no other resource or condition becomes limiting. By testing this hypothesis, we are able to not only investigate the regulation of N 2 fixation by these abiotic factors, but also to explore the nature of the symbiotic interaction, i.e., which benefits or costs the diazotrophic microbial community experiences through the close association with their host, and vice versa.

Collection of Sphagnum and peat
Two common species of Sphagnum, S. squarrosum and S. palustre, were chosen for their widespread occurrence (Europe, America, Asia, Australia) and their differences in habitat preference. While both are typical fen species, S. squarrosum is known to withstand slightly more buffered (higher pH) conditions (Clymo, 1973;Rydin and Jeglum, 2006). Field conditions of the site where the mosses were collected are shown in Table 1. To mimic their natural habitat, including moist conditions and supply of substrate-derived CO 2 for Sphagnum development (Smolders et al., 2001), peat mosses were placed on Sphagnum peat monoliths. Both peat mosses and monoliths were collected from the Ilperveld peatland in the Netherlands (52 • 26 22.68 N; 4 • 56 54.8 E), where monoliths (25 × 12 × 20 cm depth) were placed in glass mesocosms (25 × 12 × 30 cm depth) and then transported to the lab. Soils were kept wet with demineralized water (1 cm above soil level) and allowed to acclimatize for 2 weeks. Patches of 70 (S. palustre) or 80 (S. squarrosum) capitula (top 2 cm of moss) representing similar fresh weights were placed on top of the monoliths. A total of 16 mesocosms were placed in a water bath maintained at 15 • C (using a cryostat) with a light regime of 16 h light using four 400 W lamps (Hortilux Schreder HS2000, Monster, the Netherlands) and one growth lamp with 120 deep red/white LED (light emitting diode) lamps (Philips, Green-Power LED, Poland), providing in total 150 µmol PAR (photosynthetically active radiation) m −2 s −1 and a temperature of 18 • C at vegetation level. The light level was chosen to mimic realistic field conditions where Phragmites australis and sedges in these fens create low but not limiting light levels for Sphagnum spp. (Bonnett et al., 2010;Kotowski and Diggelen, 2004).

Experimental setup
After acclimatization, there was a constant flow of different treatment solutions through the mesocosms, at a flow rate of 5.4 L per week using peristaltic pumps (Masterflex L/S tubing pump; Cole-Parmer, Schiedam, the Netherlands) to create constant conditions in a 1 cm water layer over the soils. The lower 1 cm of Sphagnum spp. was flooded, while capitula were just above the water layer. Four different treatment solutions were applied (N = 4 replicates per treatment), which were spatially distributed in a randomized block design. The treatments were applied in a full factorial design with a P treatment of 10 µmol L −1 P (as Na 4 P 2 O 7 ) and a HCO 3 treatment of 3 mmol L −1 NaHCO 3 . Also, 5 mg L −1 of sea salt with small amounts of trace elements (Tropic Marine, aQua united LTD, Wartenberg, Germany) was added to all treatment solutions (including control) to mimic rainwater quality and to prevent osmotic stress. A graphic figure of the experimental setup and pictures can be found in Fig. 1. Furthermore, each mesocosm was provided with an amount of rainwater equivalent to the mean annual rainfall in the Netherlands (750 mm) and with an N concentration equivalent to the Dutch atmospheric deposition of 25 kg N ha −1 yr −1 . Three times a week, 150 mL of artificial rainwater was sprayed on the peat mosses, containing 5 mg L −1 sea salt (Tropic Marine, aQua united LTD, Wartenberg, Germany), 19 µmol L −1 KCl, 10 µmol L −1 CaCl 2 , 10 µmol L −1 Fe-EDTA, 1 µmol L −1 KH 2 PO 4 , 0.7 µmol L −1 ZnSO 4 , 0.8 µmol L −1 MnCl 2 , 0.2 µmol L −1 CuSO 4 , 0.8 µmol L −1 H 3 BO 3 , 8 nmol L −1 (NH 4 ) 6 Mo 7 O 24 and 91 µmol L −1 NH 4 NO 3 . Treatment solutions were supplied during 10 weeks, after which plant, microbial and abiotic measurements were conducted.

Plant performance
Photosynthetic rates of the mosses were determined using a fast greenhouse gas analyzer (NIRS) with cavity ringdown spectroscopy (CRD) (GGA-24EP; Los Gatos Research, USA). From each mesocosm one individual of each moss species was taken and placed in a closed glass vile (100 mL) under similar light conditions as used in the experimental setup (150 µmol m −2 s −1 PAR), connected to the gas analyzer. Changes in CO 2 concentrations were measured over a time period of 5 min, in a closed loop with the NIRS-CRDS gas analyzer capable of measuring concentration changes at a very high resolution (Crosson, 2008) and of accurately measuring photosynthesis (Hunt, 2003). Additionally, dark measurements were carried out for each sample, and gross photosynthetic rates were calculated by cor- recting the slope of CO 2 decrease in light with the slope of the CO 2 increase in the dark. Also, capitula were counted and average lengths of Sphagnum individuals determined. The total fresh weight (FW) of Sphagnum biomass was measured, after which material was dried at 70 • C for 48 h to determine dry weight (DW) in order to calculate relative growth rates.

N 2 fixation rates and elemental composition of Sphagnum
Two subsamples (the top 2 cm of two individuals) of S. squarrosum and S. palustre from each mesocosm were placed separately in 30 mL glass serum bottles with rubber stoppers; 6 mL of headspace was removed with an injection needle and replaced with 15−15 N 2 gas (98 atom % 15 N, Sigma-Aldrich, Germany), leading to 20 % 15 N 2 labeling. Samples were incubated for 48 h with a light regime of 16 h of light (150 µmol m 2 s −1 PAR) at 18 • C. They were then dried at 70 • C for 48 h and ground using a mixer mill (MM301, Retsch, Germany) for 2 min at 30 rotations s −1 . Total N concentrations and isotopic ratios were determined using an elemental analyzer (Type NA 1500 Carlo Erba, Thermo Fisher Scientific Inc., USA) coupled online via an interface (Finnigan Conflo III) to a mass spectrometer (Thermo Finnigan DeltaPlus, USA). For every control and P -treated sample an additional incubation was carried out under similar but dark conditions. For every incubated subsample a control sample was taken that had not been incubated with 15−15 N 2 , to correct for background isotopic composition as influenced by the different treatments. The cor-rected increases in 15 N labeling were converted to N 2 fixation rates (nmol N 2 gDW −1 h −1 ), using the average of both labeled subsamples. These N 2 fixation rates were also converted to rates of N fixed per unit area with bulk density data from the field (dry weight of the upper 2 cm of each species in a 10 cm 2 plot (N = 4 replicates)). Fixation rates per hectare per year were calculated assuming N 2 fixation activity throughout the growing season (Rousk et al., 2015) during a growing season of around 250 days for peatlands in the Northern Hemisphere with mild winters (Helfter et al., 2015;Zhu et al., 2012) and corrected for an average seasonal temperature of 13 • C, assuming a Q 10 of 3 (Kravchenko and Doroshenko, 2003;Granhall and Selander, 1973;Alexander and Schell, 1973). Total P and potassium (K) concentrations were determined in digestates of dried and ground Sphagnum-microorganism tissue. Digestates were prepared by heating in 500 µL HNO 3 (65 %) and 200 µl H 2 O 2 (30 %) for 16 min in a microwave (m.l.s. 1200 Mega, Milestone Inc., Sorisole, Italy). After dilution with demineralized water, P and K concentrations were measured by inductively coupled plasma emission spectrometry (IRIS Intrepid II, Thermo Electron corporation, Franklin, MA, USA).

Soil and water chemistry
At the end of the experiment, two soil subsamples of a fixed volume were taken from each mesocosm. Homogenized subsamples were dried at 70 • C for 72 h and weighted to determine bulk densities. Organic matter concentrations were 573.33 28.60 C (mg g −1 ) 294.75 14.54 N (mg g −1 ) 18.02 0.60 P (mg g −1 ) 0.80 0.04 K (mg g −1 ) 2.00 0.16 determined through loss on ignition at 550 • C for 3 h. Dried soils were digested with 4 mL HNO 3 (65 %) and 1 mL H 2 O 2 (30 %) using a microwave and measured by inductively coupled plasma emission spectrometry as described above. C and N contents of dried soil were measured using an elemental analyzer (see above). Soil properties can be found in Table 1b. The pH of surface water was measured with a standard Ag / AgCl electrode (Orion Research, Beverly, USA) combined with a pH meter (Tim840 titration manager; Radiometer analytical, Lyon, France). Alkalinity was determined by titrating down to pH 4.2 with 0.1 N HCl using an auto burette (ABU901 Radiometer, Copenhagen, Denmark). Concentrations of PO 3− 4 , NO − 3 and NH + 4 were measured colorimetrically with a 3 Auto Analyzer system (Bran and Luebbe, Norderstedt, Germany), using ammonium molybdate (Henriksen, 1965), hydrazine sulfate (Kamphake et al., 1967) or salicylate (Grasshoff and Johannsen, 1972); Cl was determined with a Technicon Flame Photometer IV Control (Bran and Luebbe, Norderstedt, Germany). Concentrations of Al, Ca, Fe, S, Mg, Mn, Na, P and K were analyzed by inductively coupled plasma spectrometry (see above).

Statistical analyses
Values displayed in bar graphs are means ± SE (SEM) (N = 4). To test for the effect of P, HCO − 3 and different species on different parameters' three-way analyses of variance (ANOVAs) were used, using P, HCO − 3 and species as independent variables (fixed factors) with two categorical groups. All dependent variables were quantitative and at a continuous scale, i.e., nitrogen fixation rate, photosynthetic activity, relative growth rate, number of capitula, Sphagnum length increment, and pore water and tissue nutrient con-centrations. Normality was tested with a Shapiro-Wilk test on the residuals of the ANOVA and data that were not normally distributed were log-transformed prior to analysis to meet conditions of parametric tests. Homogeneity of the data was checked with Levene's test of equality of variances. No interaction effects were found for any of the parameters and significance was accepted at a confidence level of P <0.05. Statistical tests were performed using IBM SPSS Statistics 21.0 (IBM Corporation, 2012).

Results
From our full factorial experiment with additions of P and/or HCO 3 we took measurements on surface water (water quality changes) and on Sphagnum-microorganism tissue: N 2 fixation activity, plant performance parameters and nutrient contents.

Water quality changes
The addition of P (10 µmol L −1 ) resulted in an increase in total P in the surface water (F = 6.044; P <0.05) from 0.7 µmol L −1 to a concentration of 6.0 µmol L −1 , indicating net uptake and/or binding of P. Supply of HCO − 3 increased pH (from 4.3 to 8.0) and alkalinity (from 0.1 to 2.8 meq L −1 ) in the surface water (F = 2780.292; P <0.001). Furthermore, upon addition of HCO − 3 the concentrations of NH 4 , Ca, Mg, Cl, S, Fe, and Al in the water increased 2 to 5 times, and K concentration was increased by a factor of 1.4 (Table 2).

N 2 fixation
Under light conditions, diazotrophic activity was similar for both Sphagnum spp. Control incubations showed high average N 2 fixation rates of around 40 nmol N gDW −1 h −1 , translating to high area-based rates of around 6 kg N ha −1 yr −1 . When treated with HCO − 3 and/or P, however, S. squarrosum showed 40 % higher fixation rates compared to S. palustre (F = 4.510; P <0.05) (Fig. 2). Addition of P positively affected N 2 fixation for both Sphagnum species (F = 12.639; P <0.005), leading to at least 2 times higher fixation rates compared to their controls (Fig. 2). HCO − 3 addition had an even greater effect, and resulted in around 4 times higher N 2 fixation rates (F = 32.103; P <0.001) (Fig. 2). The combined P and HCO − 3 treatment increased the N 2 fixation rate to 300 nmol N gDW −1 h −1 in S. squarrosum.
In general, N 2 fixation rates were highest in light incubations and around 10 times lower under dark conditions (F = 65.642; P <0.001) (Fig. 3). However, a similar increase (1.5 times higher) in fixation rates upon P addition was found under both light and dark conditions (F = 18.588; P <0.001). Table 2. Surface water characteristics for the different treatments: control (C), addition of P (P) or HCO − 3 (HCO 3 ), or both (P + HCO 3 ). Displayed are means ± SE of the mean (N = 4). The unit for alkalinity (alk) is meq L −1 ; for all elements concentrations are expressed in µmol L −1 . In the effect row, significant differences of P or HCO 3 treatment are indicated by asterisks, where * represents P ≤ 0.05, * * represents P ≤ 0.01 and * * * represents P ≤ 0.001.  Figure 2. Rates of N 2 fixation of the diazotrophic communities of Sphagnum squarrosum (dark bars) and S. palustre (grey bars), under different treatments. Both P and HCO 3 treatment significantly increased N 2 fixation in both species, shown by letter combinations: p treatment (capital letters) and HCO − 3 treatment (lower case).

Plant performance
S. squarrosum and S. palustre had similar photosynthetic rates of around 65 µmol CO 2 gDW −1 h −1 and showed a strong negative response to HCO − 3 -rich water (F = 21.468; P <0.001), resulting in approximately 50 % lower photosynthetic rates (Fig. 4). HCO − 3 also resulted in 50-70 % lower relative growth rates (F = 29.339; P <0.001), relative decrease in the number of capitula (F = 86.090; P <0.001) and average length (F = 268.846; P <0.001) of both species (results not shown). The final biomass of HCO 3 treated mosses was around 10 % lower than that of the control group. Con- Figure 3. N 2 fixation rates of diazotrophic communities of Sphagnum squarrosum (dark bars) and S. palustre (grey bars), under light or dark conditions. Displayed is the mean ± SE (N = 4) of the control and p treatment (see text). Dark conditions significantly decreased N 2 fixation rates (shown by capital letters) and P treatment significantly increased rates (shown by lower case). trols of both species ended up with a final dry weight of around 3 g per Sphagnum patch, containing around 86 capitula with a length of around 73 mm per moss. This corresponds to a growth rate of 8.5 mg gDW −1 d −1 . In contrast, P treatment did not show an effect on any of the measured plant performance variables of the Sphagnum mosses.

Nutrient contents of Sphagnum-microorganism association
Concentrations of N, P and K in Sphagnum tissue, including their microbial community, were clearly influenced by sur- face water treatments ( Table 2). Addition of P-rich surface water increased the P content in Sphagnum-microorganism tissue by 75 % for both Sphagnum species (F = 11.549; P <0.005), while N and K concentrations remained unchanged. In treatments with HCO − 3 -rich water the N concentration increased by around 20 % (F = 6.955; P <0.05), and the concentration of K in the tissue decreased by around 25 % (F = 140.343; P <0.001), without affecting P concentrations, indicating K leakage. Individual N contents did not correlate with N 2 fixation rates (results not shown). N : P ratios differed between the two Sphagnum species (F = 4.673; P <0.05), with overall slightly higher ratios for S. palustre (mean of controls: 11.8), compared with S. squarrosum (mean controls: 7.9) (Fig. 5). These ratios decreased by 57-73 % after addition of P (F = 8.656; P <0.01) to 6.7 and 5.8 respectively, while HCO − 3 addition did not influence ratios at all. N : K ratios did not differ between the two Sphagnum species and were unaffected by addition of P. Addition of HCO − 3 however increased N : K ratios by 80 % (F = 143.049; P <0.001), due to leaking of K from Sphagnum tissue. Therefore the HCO 3 treatments were not included in Fig. 5.

Diazotrophic activity under high N conditions
Surprisingly, the diazotrophic communities of S. squarrosum and S. palustre showed appreciable N 2 fixation rates of around 40 nmol N 2 gDW −1 h −1 , even though they had been subjected to high (25 kg ha −1 yr −1 ) historical and experimental airborne N input. These rates are well in the range of N 2 fixation rates reported by Larmola et al. (2014) for Sphagnum spp. in Finnish peatlands (0-126 nmol gDW −1 h −1 ) and equal to the rates they found for mesotrophic fens, even though atmospheric N inputs were significantly lower in Finland (3 kg ha −1 yr −1 ; Mustajärva et al 2008). On an areal basis, N 2 fixation rates of our controls translated to an average N input of 6 kg N ha −1 yr −1 in the upper 2 cm of peat moss for a 250-day growing season (at an average temperature of 13 • C). This is on the same order of magnitude as the range of 12-25 kg ha −1 y −1 reported for pristine boreal bogs, although their growing season only lasts 140 days per year (Vile et al., 2014). Furthermore, similar to Markham (2009), we found Sphagnum-associated N 2 fixation rates to be at least 5 times higher than those found in feather mosses, which are around 1.5-3 kg ha −1 yr −1 (Rousk et al., 2014;DeLuca et al., 2002;Zackrisson et al., 2009;Leppänen et al., 2013). This could be due to morphological differences between the moss species (including hyaline cells of Sphagnum providing additional space and protection to microorganisms) and differences in microbial communities resulting from differences in habitat conditions and resources, i.e., availability of inorganic and organic nitrogen and carbon compounds, moisture content and presence of oxygen.
The tissue N concentration of around 11.8 mg g −1 in Sphagnum spp. appears to be high compared to a range of Sphagnum N contents for different N deposition sites (Lamers et al., 2000). Optimal growth conditions for Sphagnum balticum were found at an N content of 12.9 mg g −1 (Granath et al., 2009), suggesting that Sphagnum in our experiment is around the saturation point. Indeed, high amounts of inorganic N were still taken up from rainwater by Sphagnum spp., leaving the surface water nearly depleted of N (Table 2). These high N uptake rates, especially for NH + 4 , from surface water or rainwater are indeed typical for Sphagnum spp. (Fritz et al., 2014). Simultaneously, the associated diazotrophs were still fixing N 2 at appreciable rates under these N-rich conditions, even though N 2 fixation is an energy demanding process (Vitousek et al., 2002). The fact that N 2 fixation rates were high and all N present as NH + 4 in rainwater was taken up by the moss therefore suggests that dissolved inorganic N was not or hardly available for the microbial community and that diazotrophs were still experiencing N limitation. Next to this absolute limitation, the relative lack of N was also great, given the high concentrations of all other (micro)nutrients present in the surface water. So, even the high supply of 25 kg N ha −1 yr −1 by rainwater was rapidly taken up by Sphagnum, leaving insufficient N for the microbial community that, in this way, still experienced N limitation.

Both symbiotic partners strongly differ in optimal abiotic conditions
As expected, an increase in HCO − 3 concentration, resulting in a higher alkalinity and related higher pH, decreased Sphagnum performance. Photosynthetic rates and relative growth rates decreased by around 50 % for both species. Further- Figure 5. Means of (a) the N : P ratios and (b) N : K ratios for Sphagnum squarrosum (dark bars) and S. palustre (grey bars), displayed for control (C) and addition of P (P) to surface water. Given is the mean ± SE of the mean (N = 4). HCO 3 treatments were not included, because of leaking of nutrients from tissue (see text). Significant differences between treatments are shown with * * P <0.01 in the graph. more, HCO − 3 addition led to slightly higher surface water NH + 4 concentrations (Table 2), which most likely resulted from leakage from Sphagnum tissue. Increased N : K ratios indicated that K was also leaking from tissue, both pointing towards cell die-off. This is in accordance with earlier studies that showed sensitivity of Sphagnum spp. to buffered conditions (Clymo, 1973;Lamers et al., 1999), although the fen species used in this study are known to be more tolerant than typical bog species . Here, we showed that direct infiltration of HCO − 3 from mineralrich surface waters or groundwater into the moss layer negatively affects fen Sphagnum spp. performance, rather than Ca + , which does not directly affect pH .
To our surprise, the response of the diazotrophic community to high HCO − 3 levels was completely opposite to that of Sphagnum. Although Sphagnum biomass decreased by 10 % after treatment with HCO − 3 , the diazotrophic community was stimulated and showed around 4 times higher N 2 fixation rates. The increase in N 2 fixation may, therefore, have been a direct effect of leakage of C or other compounds from deteriorating Sphagnum tissue. However, a second plausible explanation for the increase in N 2 fixation activity is a direct beneficial effect of the increase in pH in the surface water on microbial growth rates and diazotrophic activity. It is indeed known for aquatic systems that dominant diazotrophs can be inhibited by a decrease in pH (Shi et al., 2012) and for agricultural soils that diazotrophic communities are larger in higher pH soils (Silva et al., 2013). In addition, the stimulated N 2 fixation may be explained by an indirect effect of increased decomposition rates as a result of buffering (Smolders et al., 2002), leading to the mobilization of additional organic compounds and nutrients from the soil to the surface water. This was also shown in a field gradient analysis at lower atmospheric N input, where nutrient-rich conditions correlated with increased N 2 fixation rates (Larmola et al., 2014). Since nutrient concentrations in surface water increased 2-to 5-fold in this study, increased N 2 fixation by increased decomposition is a likely third possibility.
Regardless of the effect of HCO − 3 being direct, indirect or both, it is still surprising that diazotrophic microorganisms associated with Sphagnum, a genus that requires a low pH and actively acidifies its environment, would thrive under more alkaline conditions. This strongly suggests that for the diazotrophic community the symbiosis with Sphagnum seems to be a trade-off, where a sheltered environment, including prevention of drought and predation (Jassey et al., 2013;Andersen et al., 2013), in hyaline cells outweighs the sub-optimal, acidic conditions and the competition with Sphagnum for nutrients.

Role of P availability
Sphagnum spp. and their diazotrophic microorganisms were found to respond in a remarkably different way to the addition of P. As hypothesized, based on N 2 fixation being a P demanding process (Vitousek et al., 2002), higher P availability doubled the N 2 fixation rates. This increase in N 2 fixation by P addition was 75 % higher in Sphagnum squarrosum compared to S. palustre, pointing out differences in the response of the microbiomes of both species. Even more surprising, however, was that the Sphagnum performance of both species was not at all affected by increased P availability. This implies that diazotrophs were stimulated directly by higher availability of P, rather than indirectly by additional supply of compounds obtained from the moss. This is also shown by the similar increase in N 2 fixation activity with P addition under dark conditions that we found (Fig. 3). Most of the diazotrophic activity in both Sphagnum species appeared to be light-related, as N 2 fixation rates went down by 90 % under dark conditions. This may have different reasons: (1) most of the diazotrophs are photoautotrophs; (2) most diazotrophs rely on other phototrophic microorganisms for their energy supply; or (3) most diazotrophs depend directly on products of Sphagnum photosynthesis. A high abundance of phototrophic microorganisms could be explained by the high availability of nutrients, since mutualistic interactions can be altered by nutrient loading in favor of phototrophic partners (Shantz et al., 2016). P addition did not, however, increase Sphagnum growth, raising the question of which other factor may have been limiting its growth. The low N : P ratios of Sphagnum tissue of controls (around 10) indicate relative N limitation (Wang and Moore, 2014;Bragazza et al., 2004). However, under these eutrophic conditions with high N availability and high tissue N concentrations, low ratios rather seem to be an effect of high P concentrations (Jiroušek et al., 2011). Concentrations of N, P and K in Sphagnum tissue (including their microbial community) were all high or on the high end for Sphagnum in minerotrophic peatlands, particularly for P (Aerts et al., 1999;Lamers et al., 2000;Bragazza et al., 2004) (Table 3). N : K ratios of around 1.6 for the controls in our experiment did not support the idea of K limitation (Bragazza et al., 2004). Other (micro)nutrients like molyb- Table 3. Concentrations of N, P and K (mg g −1 ) in Sphagnum for different treatments. Since no significant differences between species were found, data of both species were combined to display mean ± SE (N = 8). In the effect row, significant differences in P or HCO 3 treatment are indicated by asterisks: * P ≤ 0.05, * * P ≤ 0.01 and * * * P ≤ 0.001. N (mg g −1 ) P (mg g −1 ) K (mg g −1 ) HCO 3 effect * * * * * denum were also readily available from the surface water.
Since light conditions provided in the experiment resulted in at least 80-90 % of saturation of the Sphagnum photosystem (Harley et al., 1989) and drought was avoided, growth limitation by light or water also seems unlikely. The lack of additional growth with added P and additionally fixed N can therefore most likely be explained by the fact that control peat mosses were already at their physiological maximum. Biomass production rates (based on the average growth rate of 8.5 mg gDW −1 d −1 and a growth season of 250 days) corresponded to around 300 g m 2 yr −1 , which is indeed high (Rydin and Jeglum, 2006;Gunnarsson, 2005). With apparently no nutrient limitation for Sphagnum growth, P addition led to accumulation in Sphagnummicroorganism tissue. This lowered the N : P ratio, pointing towards unbalanced uptake of P or luxury consumption (increased nutrient accumulation without any gain in Sphagnum biomass). The amount of N fixed by diazotrophs under light conditions correlates with the N content of Sphagnum, including its microbiome tissue (Fig. 6). When we use the rate of N 2 fixation to calculate theoretical increases in N content for different treatments, these can explain the increase in N content (result not shown). The unbalanced uptake of P, relative to N, therefore calls into question the direct role of the high diazotrophic N 2 fixation rates we found here for Sphagnum growth, and rather suggests N accumulation in the associated microbial community. In conclusion, either the fixed N was not directly available for Sphagnum or it could not be used due to physiological constraints. In both cases, Sphagnum could not profit from the additionally fixed N and seemed to be competing for nutrients with its symbionts rather than regulating their activity by supplying additional C. This is in stark contrast to Azolla spp., where P addition is known to directly increase the growth rate and N content of the host plant (direct mutualism) (Cheng et al., 2010). Under the present environmental conditions, the symbiosis between Sphagnum and its microbial community seems to be based on the indirect transfer of nutrients after microbial die-off Figure 6. Linear regression of the average N content of Sphagnum, including its microbiome against the average N 2 fixation rates of both species. (Ho and Bodelier, 2015) rather than by a mutualistic interaction with Sphagnum directly benefitting from the additionally fixed N. More research is, however, needed to determine whether the symbiosis would change to a mutualistic interaction at low N conditions. At the ecosystem level, the increased N 2 fixation rates with the lack of additional biomass production of Sphagnum with added P led to remarkably high amounts of 15 kg ha −1 yr −1 of extra N input.

Importance of the symbiosis
We showed that in these N-rich fen systems, Sphagnum spp. still work as a filter monopolizing N and their microbial community still experiences N limitation. With all N taken up by Sphagnum, diazotrophs fix N 2 at appreciable rates despite high N deposition. N 2 fixation rates are even more increased by addition of P and by a higher HCO − 3 concentration, as an effect of increased pH or an increase in (micro)nutrients other than P. This may well explain the differences in N 2 fixation rates between fens and bogs (Larmola et al., 2014). The diazotrophic community seems to have different optimal environmental conditions than its host, and seems to trade off protection from herbivores inside Sphagnum hyaline cells against monopolization of N and active acidification by Sphagnum. As peat mosses did not benefit from the fixed N, active control of the diazotrophic community (e.g., by additional organic compound supply) seems unlikely. Given the high N 2 fixation rates and accumulation of N in Sphagnum peat, we hypothesize that the fixed N is available by reabsorption from decaying and dead Sphagnum tissue and dead microbial biomass, rather than by the direct transfer between diazotrophs and Sphagnum. Ho and Bodelier (2015) also suggested this alternative pathway of N transfer between Sphagnum and N 2 fixing methanotrophs, and feather mosses were suggested not to depend on their cyanobacterial community for N (Rousk and Michelsen, 2016). Since N loads (25 kg ha −1 yr −1 ) were high here, and N 2 fixation added 6 kg N ha −1 yr −1 or more with high P loads, peat mosses can be expected to not be able to reabsorb the mineralized N, which then leaches deeper into the peat. Here, it may become available to vascular plants (Lamers et al., 2000). In this way, the high N 2 fixation rates may speed up decomposition rates and invasion of vascular plants by supplying additional N to an already N loaded system. As high P input still increases N 2 fixation rates, this will not be able to balance out the high N loads.
Data availability. The data presented in this paper can be found in van den Elzen et al. (2017).

Conclusions
1. In N saturated fens with an N deposition of 25 kg ha −1 yr −1 the activity of diazotrophs can still be unexpectedly high (40 nmol N gDW −1 h −1 ). Since Sphagnum spp. monopolize all N in surface water, their microbial community still experiences N limitation.
2. Diazotrophs are stimulated by addition of P and HCO − 3 (buffer capacity), benefitting from additional organic compounds, nutrients and/or an increase in pH, which explains variations in N 2 fixation rates reported for peatlands differing in nutrient supply or buffering.
3. Sphagnum growth is -in stark contrast -hampered by the high HCO − 3 concentrations. This calls into question the concept of a direct mutualism and seems to point to a compromise for the diazotrophic community between a sheltered environment on the one hand and a sub-optimal pH and competition for nutrients with their host on the other.