What drives biological nitrogen fixation in high arctic tundra: Moisture or temperature?

,


INTRODUCTION
Nitrogen is usually the primary limiting nutrient for plant growth in subarctic and arctic tundra (Shaver andChapin 1980, Michelsen et al. 2012).While atmospheric N deposition is one major source of plant available N in many ecosystems, N deposition in pristine ecosystems, such as subarctic and arctic tundra, is low (<2 kg NÁha À1 Áyr À1 ; Van Cleve and Alexander 1981, Peñuelas et al. 2013) and is likely not sufficient to cover plant-N demand.Here, fixation of atmospheric N 2 is a large source of plant available N and is performed by free-living N 2 fixing bacteria (diazotrophs), and diazotrophs associated with lichens and mosses (Hobara et al. 2006, Rousk et al. 2016).In boreal forests, subarctic and arctic tundra, moss-associated N 2 fixation can exceed N deposition rates, mostly due to the high coverage and biomass of mosses (Van Cleve et al. 1983, Turetsky 2003, Gavazov et al. 2010, Michelsen et al. 2012).However, N 2 fixation is limited by the availability of molybdenum and phosphorus (Rousk et al. 2017a), as well as by extreme temperatures (Gundale et al. 2012, Stewart et al. 2014) and low moisture conditions (Rousk et al. 2014, Stewart et al. 2014).While it has been shown in laboratory (Rousk et al. 2014(Rousk et al. , 2017b)), field (Rousk et al. 2015), and modeling (Rousk and Michelsen 2017) approaches that moisture is the major factor driving moss-associated N 2 fixation, those studies have been conducted almost exclusively in boreal forests and subarctic tundra, and attempts to extend this observation to arctic ecosystems have been limited.Studies that have been conducted in the high Arctic on freeliving-as well as on moss and lichen-associated diazotrophs have shown a strong moisture dependence of N 2 fixation (Zielke et al. 2005, Stewart et al. 2011, 2014).On the other hand, Hobara et al. (2006) found an increase in N 2 fixation in plant-soil cores in arctic tundra with temperature up to 30°C, while a positive effect of moisture was not found.Thus, the primary driver (temperature or moisture) of N 2 fixation in the high Arctic remains unspecified.
Arctic ecosystems are among the most sensitive systems to changes in temperature (Seddon et al. 2016, see also Schuur et al. 2015).Increased temperatures resulting in drying-out of mosses in a future climate will lead to limited moisture availability for cyanobacteria within the moss carpet.This will likely lead to reduced N 2 fixation rates (Rousk et al. 2014) and, with that, reduced N input via the N 2 fixation pathway.On the other hand, as a component of warming, summer rainfall is likely to double in the high Arctic by the end of this century (Bintanja and Andry 2017), which could counteract surface drying.In addition to direct temperature effects on N 2 fixation, a warmer climate will also lead to increased thaw depth of permafrost in high arctic settings (Schuur et al. 2015), freeing up N pools (e.g., Keuper et al. 2012) that could inhibit N 2 fixation (e.g., Rousk et al. 2013).
A changing climate will enhance air and soil temperatures on the one hand, but it will also lead to increased shading of lichens, mosses, and freeliving diazotrophs from faster growing vegetation and expanding shrubs (Hollister et al. 2005, Wahren et al. 2005).Light is another notable driver of N 2 fixation and can limit diazotroph activity in the tropics (Taylor et al. 2017) as well as in the subarctic (e.g., Sorensen et al. 2012).Given that few diazotroph-associated shrubs grow in the high Arctic, and these are unlikely to expand their range considerably also in the nearer future, the loss of the main N 2 fixing associations as a result of shading and exclusive competition by shrubs (Hartley et al. 2012) will decrease N 2 fixation potential in these systems.Yet, the effects of shading on N 2 fixation in the Arctic have hardly been addressed.Thus, to assess the effects of climate change factors (warming, shading) on N 2 fixation in the Arctic, we utilized a field experiment in a high arctic heath in which climate change was simulated by increasing summer air and soil temperatures using open top chambers as well as by a shading treatment, reducing light input and temperature.We hypothesized that (1) N 2 fixation is highest in the warming treatment if moisture is not limiting, and (2) N 2 fixation is lowest in the shading treatment.To address these hypotheses, we measured N 2 fixation in situ in the climate change field treatments at different times in the growing season.We further divided plots into treatments with added water and without added water before the N 2 fixation measurements to assess how increased precipitation, in combination with shading and warming, in a future climate will affect N 2 fixation.To further test whether the climate change treatments lead to changes in soil N availability via changes in N 2 fixation, we also assessed the soil N pools.

Field sites and experimental set-up
The field experiment was established in a mesic-dry heath dominated by Salix arctica on Zackenberg, Greenland, in 2004 and has been maintained since.The experimental site is located in the center of the valley of Zackenberg in North-Eastern Greenland, near the Zackenberg Research Station (74°30 0 N, 21°00 0 W).The climate is continental high Arctic.Average air temperature is À20°, 7.0°, and À9.0°C in January, July, and whole year, respectively, and total annual precipitation is 211 mm but ranging from 93 to 310 mm (data 1997(data -2014(data , Jensen et al. 2016)), with the most precipitation falling as snow (Fig. 1, see also Appendix S1: Fig. S1).In June, July, and August 2010, the period in which we conducted our measurements, air temperature was 1.9°, 5.3°, and 5.3°C (long-term average 2003-2014: 1.5°, 6.8°, and 5.6°C), and precipitation was 13, 1, and 2 mm (long-term average 2003-2014: 5, 13, and 20 mm).The bedrock is characterized by non-calcareous sandy fluvial sediments, and the soil is classified as a Typic Psammoturbels (Turbic Cryosol) with a pH between 5 and 7.The permafrost is continuous, and the active layer is <1 m deep (60-70 cm during the measuring campaigns, Jensen andRasch 2011, Elberling et al. 2013).The plots are rather exposed (Appendix S1: Fig. S2A), with a maximum snow cover of 80-95 cm.Besides the prostrate Salix arctica Pall., other low woody plants are Vaccinium uliginosum L., Cassiope tetragona D. Don (L.), and Dryas spp.and a few herbaceous species such as Luzula arctica Blytt, Polygonum viviparum L., and Arctagrostis latifolia (R. Br.) Griseb.(Campioli et al. 2013).
The field experimental plots consist of five replicate blocks each with the treatments (1) control, (2) shading with sackcloth, and (3) warming by open top chambers.Warming was achieved by small (1.2 9 1.2 m, 50 cm high) dome-shaped open top chambers of 0.05-mm polyethylene film, constructed with polyvinyl chloride (PVC) tubes, with an opening of ~50 cm diameter at the top.The chambers were in place each year from late June to the end of August.Late summer canopy temperature is enhanced by 1.3°C and soil temperature at 2 cm depth by 0.3°C (Campioli et al. 2013), a moderate increase simulating surface temperature projections for high arctic systems for the mid-21st century (IPCC 2013).
Shading with sackcloth in chambers of similar constructions with PVC tubes reduces light with 50% and air and soil temperature by 1.5-2.0°C(Fig. 2, see also Ellebjerg et al. 2008).The reduction in incoming light is similar to the reduction caused by taller shrubs such as Betula nana and simulates the expansion of canopy-forming shrubs into high-latitude ecosystems.

Vegetation cover
The vegetation cover in each plot was estimated with the pin-point intercept method 23 July 2010 (Jonasson 1983).For this, a pin was passed vertically at 100 points (2 cm spaced) in a 0.5 9 0.5 m frame and every vascular plant species, mosses, lichens, and litter touched by the pin was recorded.

Dependence of N 2 fixation on moisture
Nitrogen fixation was measured using the acetylene reduction assay (ARA) in situ in 2010 in the climate change field experiment described above, in which diazotrophs associated with mosses, and free-living diazotrophs are the dominant N 2 fixers (Fig. 3; Appendix S1: Fig. S2).The ARA is a measure of the nitrogenase enzyme reducing N 2 to ammonia, as well as reducing acetylene to ethylene.To assess the effects of increased precipitation on N 2 fixation in the field, we established two cores with a diameter of 5 cm and a depth of 7 cm in each experimental field plot (n = 5, Appendix S1: Fig. S2B).The top of the cores were dominated by free-living diazotrophs (organic crust), including lichens and mosses (Appendix S1: Fig. S2B).We placed the cores in nylon mesh bags (50 mm mesh size; Sintab, Oxie, Sweden) and set them carefully back into the hole in the ground, from where the core originated, whenever we were not doing assays and ensured close contact between the core and the surrounding substrate.One core in each plot was with no further treatment, and one core was watered with 10 mL deionized water the day before each ARA measurement, and again immediately before the ARA incubations.This is similar to incidents of 5 + 5 mm precipitation and hence comparable to input in observed precipitation events in mid-June and early September, and the cumulative addition of 50 mm makes up slightly more than a doubling of average current June-September precipitation of 38 mm.For the ARA, samples were placed in closed, transparent PVC containers with a volume of 0.5 L (Appendix S1: Fig. S2).A 20-mL vial containing calcium carbide was placed in each container, and water was added with a syringe through a rubber septum to induce the development of 10% (vol.)acetylene in the container.The containers were kept in the field plots during the ARAs.Ten milliliters of gas samples was taken 1 min and 2 h after incubation with acetylene and analyzed for acetylene and ethylene on a gas chromatograph (Shimadzu GC-14B, Tokyo, Japan).Acetylene reduction was measured five times during the growing season from mid-July to August 2010 in control, shaded, and warmed plots.Soil temperature at 2 cm depth was measured just before the ARA incubations.Soil moisture was determined gravimetrically.The temperature in the containers during the ARA incubations was not different between the three field treatments and between the water addition treatments (no water addition: 18.4 AE 1.9, 15.6 AE 1.8, 17.9 AE 1.6°C for the control, shading, and warming treatment, respectively; water addition: 18.6 AE 1.7, 15.5 AE 1.6, 17.9 AE 1.6°C for the control, shading, and warming treatment, respectively).After the last measuring campaign in August, the top 2 cm (including organic crust) and the lower part (2-7 cm depth, including roots) of the cores were dried and analyzed for total N and carbon (C) content on an Isoprime isotope ratio mass spectrometer (Elementar UK Ltd., Cheadle Hulme, Stockport, UK) coupled to an Eurovector CN analyzer.

Soil sampling and analyses
Soil was sampled (two cores per plot) with a 4 cm diameter auger to a depth of 5 cm.The soil was homogenized and subjected to chloroform fumigation-extraction and analyzed as in Sorensen et al. (2008).In brief, 10 g of the sorted, fresh soil was fumigated with chloroform for 24 h to release the N and C in the soil microbial biomass, after which the soil was extracted for 1 h in 50 mL demineralized H 2 O.The extracts were filtered through Whatman GF-D filters and frozen until analysis.Another 10 g fresh soil was extracted with 50 mL demineralized H 2 O to recover soil inorganic N with a Fiastar 5000 Flow Analyzer and dissolved organic carbon (DOC) with a Shimadzu Total Organic Carbon Analyzer.To obtain microbial N, 5 mL of fumigated and unfumigated, demineralized H 2 O-extracts was digested in persulfate and analyzed with a Fiastar 5000 Flow Analyzer.Microbial N and C contents were calculated by subtracting the N and C in digested, unfumigated extracts from that in the digested, fumigated extracts.The microbial N and C contents were calculated by assuming an extractability of 0.4 (for N) and 0.45 (for C).Ten grams of fresh soil was dried at 70°C to constant weight to determine the soil moisture content.

Statistical analyses
To test for significant differences in vegetation ground cover, soil temperature, soil moisture, and other soil and soil microbial measures between the climate change field treatments, we ran one-way ANOVAs followed by Tukey's post hoc test.Differences in ARA between measuring campaigns and the climate change field treatments were tested with repeated-measures ANOVA, with water and field treatment as factors, followed by Tukey's post hoc test.ANCOVAs were run to test for the effects of the added water on the relationship between ARA and soil moisture and soil temperature.Finally, to assess the importance of soil moisture, soil temperature, and the water addition treatment on ARA, we performed backwards stepwise multiple regression analyses.All statistical analyses were performed in R 3.0.3(R Developmental Core Team 2014).

Vegetation ground cover
No differences in ground cover of different functional plant groups between the climate change field treatments were found.However, vascular plants taken together and litter made up most of the ground cover in all treatments (F 7,32 > 16 for all treatments, P < 0.0001, Fig. 3).

Nitrogen fixation during the growing season and the effects of water addition
Acetylene reduction was higher in the first two measurements in July (17 July, 21 July) than later in the season (30 July, 3 August).The highest activity was found at the first measuring campaign (F 4,120 = 13.55,P < 0.0001, Fig. 4).The plots receiving water before the ARA measurements had higher N 2 fixation rates than the plots without any water added in the control and warming treatments throughout the measuring campaigns and in the shading treatment only at the last measuring date (F 1, 120 = 7.64, P = 0.007, Fig. 4).No differences in ARA between the climate change field treatments were found.

The moisture and temperature dependence of N 2 fixation
Moisture content in the field did not change throughout the season, and soil moisture was higher in the shading plots when no water was added compared to the other plots (F 2,11 = 1.44,P = 0.007, Fig. 2a; Appendix S1: Fig. S3A).The water additions lead to increased moisture content in all treatments (F 2,11 = 1.90, P < 0.0001, Fig. 2a).Nitrogen fixation (ARA) increased with increasing moisture content in all climate change field treatments (R 2 = 0.31, P < 0.0001 for all measuring campaigns and treatments, Fig. 5).In the warming treatment, the added water and the moisture content further interacted to influence N 2 fixation (P = 0.002, t = À2.2,Fig. 5).
The climate change field treatments lead to differences in the soil temperature: The lowest soil temperatures were recorded in the shading treatment and the highest in the warming treatment (averaged across the season and water treatments: 12.3 AE 0.02°, 11.0 AE 0.1°, 14.0 AE 0.5°C for the control, shading, and warming plots, respectively, F 2, 118 = 37.52, P < 0.0001, Fig. 2b; Appendix S1: Fig. S3B).The water additions did not lead to changes in soil temperature between the field treatments (Fig. 2b).The lowest temperatures were recorded at the first two sampling time points in July (F 4, 118 = 30.15,P < 0.001, Appendix S1: Fig. S3B).
In the control plots, soil moisture was the most important factor affecting ARA, irrespective of adding water or not (P < 0.0001, R 2 = 0.41).In the shading plots, the added water seems to be the most important driver for ARA, followed by the interactive effects of the added water and soil moisture and soil temperature, and the interactive effects of all three factors (P = 0.007, R 2 = 0.23).In the warming plots, the pattern was similar.The most important factor influencing ARA was the added water, followed by the interactive effect of the added water and soil moisture, and the added water and soil temperature.When the model was run without the added water as factor, soil moisture was the most important factor controlling ARA in all treatments (P control < 0.0001, R 2 = 0.41, P shading = 0.008, R 2 = 0.15, P warming = 0.0001, R 2 = 0.34).In addition, in the warming treatment, the interaction between soil temperature and soil moisture was also significantly affecting ARA (P = 0.0001, t = À4.29).

Soil microbial and chemical characteristics
The soil had low total C and N concentration, and low amounts of available N and P. The high microbial C/N ratio pointed toward fungal dominance of the microbial community.No significant differences were found between the climate change field treatments (control, shading, warming) and between the water additions and no water addition treatments in any of the measured soil parameters (Table 1).Total C and N were higher in the top part of the ARA-cores that ) in relation to gravimetric moisture content (%) in the climate change field treatments control, shading, and warming, in cores with water added (black symbols) and without water added (white symbols).Shown are the rates from the five measuring campaigns during the growing season in high arctic tundra, NE Greenland.Significant relationships are indicated with fitted lines and R 2 for the warming treatment.The other two treatments did not show a significant interaction between acetylene reduction, soil moisture, and added water.Soil temperature (°C) Fig. 6.Acetylene reduction (lmolÁm À2 Áh À1 ; mean AE SE) in relation to soil temperature in the climate change field treatments control, shading, and warming, in cores with water added (black symbols) and without water added (white symbols).Shown are the rates from the five measuring campaigns during the growing season in high arctic tundra, NE Greenland.Significant relationships are indicated with fitted lines and R 2 for the warming treatment.The other two treatments did not show a significant interaction between acetylene reduction, soil temperature, and added water.

DISCUSSION
Nitrogen fixation increased with moisture content in all climate change field treatments and was even higher in the warming treatment when water was added.Our analyses also reveal that moisture is more important than temperature for N 2 fixation activity in our study system.Hence, also in the high Arctic, biological N 2 fixation is controlled in a hierarchical way, with moisture being the most important factor for nitrogenase activity, followed by temperature.This confirms previous findings for free-living (Belnap 2001) and moss-associated diazotrophs in the Subarctic (Rousk and Michelsen 2017) and high Arctic (also reviewed in Stewart et al. 2014).For instance, in Sphagnum mossdominated habitats in the Arctic, moisture content explained at least 50% of the variation in N 2 fixation rates in the field (Stewart et al. 2011).Nonetheless, moisture content is an indirect effect of temperature, lichens, mosses and free-living diazotrophs dry out quickly when temperatures increase.Thus, temperature and moisture interact strongly to affect N 2 fixation (Stewart et al. 2011, Rousk et al. 2017b).Further, while moisture affects the immediate activity of diazotrophs, temperature shapes the bacterial community composition over long time periods (see, e.g., Rinnan et al. 2009), thereby driving N 2 fixation rates.
In our study, the highest N 2 fixation activity was found at temperatures below 14°C.This stands in sharp contrast to previous findings on the temperature dependence of N 2 fixation, with a temperature optimum around 25°C for symbiotic as well as moss-associated N 2 fixation (Smith 1984, Houlton et al. 2008).Then again, most previous findings are from ecosystems outside the Arctic.Abiotic conditions, such as air temperature can be extreme in arctic habitats, and plant and microbial communities are adapted to those conditions (Billings and Mooney 1968, Bliss 1971, Robinson 2001, Rinnan et al. 2009).Organisms adapted to low temperatures, as found in the Arctic, are also more tolerant to still colder temperatures and can maintain high activity at temperatures that are too low for warm-adapted organisms (Rinnan et al. 2009, Birgander et al. 2013).Free-living cyanobacteria fix N 2 and photosynthesize at low temperatures in Antarctic and Arctic settings (À4°C, Davey 1983, Liengen 1999), and achieve light saturation at low light levels and can fix C (and N 2 ) when shaded by clouds and potentially by vegetation (Zielke et al. 2002).
Our results further show that, although higher moisture content promoted N 2 fixation, if the soil is already very moist, additional moisture does not necessarily increase N 2 fixation more, as seen in the shading treatments (Fig. 5).Soil temperatures in the shaded plots were lower than in the other treatments, and more moisture did not increase N 2 fixation in these plots.This indicates that the long-term exposure to lower temperatures might have shifted the diazotroph communities in these plots, adapted to decreased temperatures.This suggests that the long-term exposure to (lower) temperatures drives N 2 fixation.Further, N 2 fixation activity was highest in the warming plots, hinting to a shift in the diazotroph community adapted to higher temperatures (Rousk and Michelsen 2017).This is also reflected in the maximum N 2 fixation activity found above 12°C in the warming, around 12°C in the control, and below 10°C in the shaded plots.These temperature optima are much lower than previously reported.For instance, Hobara et al. (2006) report increasing N 2 fixation rates by free-living and lichenassociated diazotrophs in high arctic tundra up to 30°C, and maximum temperature for N 2 fixation in mosses from the Antarctic have been found between 25°and 27°C (Smith 1984).On the other hand, N 2 fixation rates in different legume species in the high Arctic differ in their temperature optimum ranging between 10°and 25°C (Schulman et al. 1988).In mosses from boreal forests, temperature optima of 13°and 22°C have been reported (Gentili et al. 2005), which was attributed to different cyanobacterial communities with different temperature optima, thereby occupying separate niches on mosses.However, we cannot rule out that N 2 fixation in our study system might be even higher at higher temperatures.A negative relationship between N 2 fixation and air temperature has been reported previously for lichens and mosses (Rousk et al. 2015).This was ascribed to the drying-out effect of the diazotroph associations by increased temperature.The large range in temperature optimum reported previously (e.g., Smith 1984, Hobara et al. 2006) might have been recorded while the N 2 fixer associations have been optimally moist.Further, mosses differ in their capacity to hold water (Elumeeva et al. 2011), which might also explain differences in N 2 fixation activity between studies investigating different moss species.
Our study further highlights the large differences between measured N 2 fixation rates in situ, and potential N 2 fixation under optimal (moisture) conditions (Fig. 4).Field measurements likely underestimate N 2 fixation rates and should take into account the abiotic conditions prevailing during and shortly before measurements.Most laboratory studies assess N 2 fixation under optimal conditions, making direct comparisons between field and laboratory based measurements ambiguous.
The climate change field treatments did not translate into differences in vegetation cover and plant community composition.This could explain the lack of a treatment effect on N 2 fixation rates, which stands in contrast to our hypotheses.If mosses and other N 2 fixing associations (lichens, free-living diazotrophs) are not outcompeted by faster growing vascular plants, then N 2 fixation rates may remain similar.The absence of a treatment effect on plant coverage is consistent with a recent synthesis showing that some plant communities in high arctic ecosystems respond little to moderate warming (1-3°C, Elmendorf et al. 2012).It is also consistent with unchanged relative growth rate and gross ecosystem production in warmed plots compared to controls in the willow heath at Zackenberg investigated in the current study (Campioli et al. 2013) and with the unchanged normalized differential vegetation index in the same warmed or shaded plots (data from mid-August 2012, not shown).Along those lines, the climate change field treatments did not result in differences in a range of soil measures after 6 yr of initiation of the treatments (Table 1).While small additions of N (0.5 gÁm À2 Áyr À1 ) can lead to significant and rapid (2 yr after N application) shifts in vegetation composition in the high Arctic (Arens et al. 2008), our climate change field treatments did not alter soil N pools.Changes in vegetation composition, N 2 fixation rates, soil N pools, etc., might emerge after a longer period of initiation of field treatments.For instance, an increase in shrub cover at the expense of mosses and lichens has been observed in longer term warming experiments (5-10 yr, Graglia et al. 2001, Hollister et al. 2005, Wahren et al. 2005), and a reduction in mossassociated N 2 fixation has been found in shaded plots in the Subarctic 21 yr after initiation of the field treatments (Sorensen et al. 2012).

CONCLUSIONS
Biological N 2 fixation in the high Arctic is strongly affected by moisture conditions.Thus, the consequences of changing precipitation patterns in a future climate will depend on the prevailing moisture conditions.Further, the diazotroph communities in arctic habitats are adapted to the historically low temperatures and will likely shift in a warmer, future climate.Increasing temperatures will have direct effects on N 2 fixation as well as indirect, long-term effects via shrub expansion into high arctic ecosystems at the expense of lichens, mosses, and free-living diazotrophs.Indirect effects of increased temperatures will likely reduce ecosystem N input via associative N 2 fixation in the long term.However, the net effects of climate change on N 2 fixation are ambiguous and will depend strongly on the degree and interactive effects of temperature, precipitation, and shrub expansion.

Fig. 2 .
Fig. 2. (a) Soil moisture (%) and (b) soil temperature at 2 cm depth (°C) in the climate change field treatments control, shading, and warming in high arctic tundra, NE Greenland, averaged across the five measuring campaigns in 2010.Given are means AE SE (n = 5) for the plots with water added (black bars) and the plots where no water was added (white bars).Different lowercase letters indicate significant differences between the water treatments and the climate change field treatments.

Fig. 3 .
Fig. 3. Coverage of vegetation and litter in the climate change field treatments control, shading, and warming in high arctic tundra, NE Greenland (mean AE SE, n = 5).Different lowercase letters indicate significant differences between plant species.

Fig. 4 .
Fig. 4. Acetylene reduction (lmolÁm À2 Áh À1 ) in three field treatments (control, shading, and warming) from July to August in high arctic tundra, NE Greenland.Shown are mean rates AE SE (n = 5) in cores that received water (black bars) and in cores that did not receive water prior to measurements (white bars).Asterisks indicate significant differences between the treatments with and without added water.Capital letters indicate significant differences in acetylene reduction between dates.

Fig. 5 .
Fig.5.Acetylene reduction (lmolÁm À2 Áh À1 ) in relation to gravimetric moisture content (%) in the climate change field treatments control, shading, and warming, in cores with water added (black symbols) and without water added (white symbols).Shown are the rates from the five measuring campaigns during the growing season in high arctic tundra, NE Greenland.Significant relationships are indicated with fitted lines and R 2 for the warming treatment.The other two treatments did not show a significant interaction between acetylene reduction, soil moisture, and added water.

Table 1 .
Soil microbial and chemical characteristics in the climate change field treatments control, shading, and warming in high arctic tundra, NE Greenland.