Functional traits define species by their ecological role in the ecosystem. Animals themselves are host–microbe ecosystems (holobionts), and the application of ecophysiological approaches can help to understand their functioning. In hard coral holobionts, communities of dinitrogen (N2)-fixing prokaryotes (diazotrophs) may contribute a functional trait by providing bioavailable nitrogen (N) that could sustain coral productivity under oligotrophic conditions. This study quantified N2 fixation by diazotrophs associated with four genera of hermatypic corals on a northern Red Sea fringing reef exposed to high seasonality. We found N2 fixation activity to be 5- to 10-fold higher in summer, when inorganic nutrient concentrations were lowest and water temperature and light availability highest. Concurrently, coral gross primary productivity remained stable despite lower Symbiodinium densities and tissue chlorophyll a contents. In contrast, chlorophyll a content per Symbiodinium cell increased from spring to summer, suggesting that algal cells overcame limitation of N, an essential element for chlorophyll synthesis. In fact, N2 fixation was positively correlated with coral productivity in summer, when its contribution was estimated to meet 11% of the Symbiodinium N requirements. These results provide evidence of an important functional role of diazotrophs in sustaining coral productivity when alternative external N sources are scarce.
Hermatypic corals fulfil important functions as key ecosystem engineers of tropical coral reefs, providing the habitat for some of the most productive, diverse and economically important biological systems on Earth . They do so by controlling a wide range of biogeochemical processes important to coral reef functioning, and by intensively generating and transforming inorganic and organic materials . These sessile cnidarians owe their evolutionary success to an endosymbiosis with photosynthetic dinoflagellate microalgae of the genus Symbiodinium that contribute a substantial fraction of the total gross primary productivity—that is, the amount of inorganic carbon (CO2) photosynthetically fixed per unit of time—in coral reef ecosystems [3,4]. Up to 50% of the net fixed CO2 in the coral is subsequently released as organic carbon (C) in the form of mucus , and provides a nutritious food source to other reef organisms via the microbial and sponge loops .
Besides endosymbiotic eukaryotic algae, hermatypic corals are also associated with a diverse array of prokaryotes . The prokaryote–dinoflagellate–coral assemblage is often referred to as the coral holobiont . In the resulting host–microbe ecosystem, the interactions among the partners determine the functioning and ecological success of the whole group of organisms . Evidence is accumulating that among these prokaryotic communities dinitrogen (N2)-fixing microorganisms (termed diazotrophs) form highly specific associations with their cnidarian hosts and endosymbiotic algae, and may play an important role in holobiont functioning [9–15].
Symbiotic corals show a remarkable degree of nutritional plasticity, being able to not only autotrophically fix CO2 at high rates, but also to rapidly assimilate organic C and nitrogen (N) through heterotrophy [16,17]. Additionally, corals have evolved strategies to exploit any source of inorganic N , enabling them to survive in oligotrophic tropical waters where N is the most limiting nutrient to primary productivity [19,20]. However, the abundance of symbiotic Symbiodinium within the coral host is limited by N concentrations [4,21]. N deficiency and starvation of Symbiodinium can cause severe depletion of the main N-containing photopigment [22,23], chlorophyll a (chl a), and may result in a decline in productivity. Although very efficient internal recycling of N prevents disintegration of the symbiosis and loss of Symbiodinium, recycled N cannot account for new biomass production and growth, and therefore external N sources are necessary [4,24].
In organismal and community ecology, morpho-physio-phenological traits that indirectly affect individual performance are defined as ‘functional traits’ . In oligotrophic environments, where ambient concentrations of inorganic and organic N are scarce, diazotrophs may thus confer a functional trait to the holobiont, as microbial N2 fixation could potentially provide an important supplementary source of a limiting element for coral productivity . This holds even more so because ammonium (i.e. the direct product of N2 fixation) was recently shown to increase carbon acquisition and translocation to the coral host compared with other forms of N such as nitrate . However, the functional significance of N2 fixation in sustaining productivity of hermatypic corals is still unclear.
Here, we present results from an annual study examining the relationship between N2 fixation and primary productivity in hermatypic coral holobionts at a fringing coral reef in the northern Red Sea (Gulf of Aqaba). We made use of the characteristic seasonality of the sampling location, which harbours some of the northernmost warm water coral reefs on Earth. These reefs are exposed to deep water column mixing during early spring followed by a strong thermal stratification in summer  (figure 1a; electronic supplementary material, figure S1). Stratification leads to depleted inorganic N concentrations approximately one order of magnitude lower in summer compared with spring  (electronic supplementary material, tables S1 and S2). We thus repeated our measurements of the four locally dominant hermatypic coral genera during all four seasons. Our sampling strategy allowed us to assess the contribution of associated diazotrophs to changing N requirements of the coral holobiont in response to pronounced seasonality.
(a) Collection and maintenance of corals
All research described was conducted at the northern Red Sea (Gulf of Aqaba) during 2013. Sample collection was carried out along the slope of the fringing reef in front of the Marine Science Station (MSS) in Aqaba, Jordan (location: 29°27′ N, 34°58′ E) using scuba. In February (winter), April (spring), September (summer) and November (autumn), eight coral fragments (5–6 cm in height) of each of the hermatypic coral genera Acropora, Pocillopora, Stylophora and Goniastrea, identified morphologically, were sampled haphazardly at 10 m water depth from individual colonies that were at least 5 m apart to avoid sampling clones. These genera were chosen as they were the dominant hermatypic coral genera at our reef site, representing a large part of the total coral cover. Fragments were cut underwater with clean bone cutters, placed in plastic bags (taking care not to cause any abrasion of the tissues) and transported to the laboratory facilities within 30 min from sampling. In the laboratory, fragments were immediately glued on ceramic tiles and maintained in an 800 l flow-through aquarium directly supplied with untreated reef water (from 10 m water depth) at a rate of approximately 4000 l h−1. Natural levels of light intensity inside the maintenance aquarium were adjusted to seasonal in situ conditions at 10 m water depth using layers of black mesh. Measurements of temperature, light intensity and inorganic nutrient concentrations in the flow-through aquarium were routinely conducted, and differences from in situ conditions were undetectable (see electronic supplementary material, table S1).
(b) Environmental monitoring
Temperature, light intensity, seawater inorganic nutrients, chl a, particulate organic C (POC), particulate N (PN) and dissolved organic C (DOC) concentrations were monitored at the coral sampling location on the reef (10 m depth) during all seasons. Temperature and light intensity were monitored at 1 min intervals by data loggers, whereas seawater samples were collected once weekly over four weeks per season using canisters (n = 4) and analysed for inorganic nutrients, chl a, POC and PN concentrations using the procedures detailed by Bednarz et al. . In addition, one 50 ml subsample was collected from each canister for DOC analyses (n = 4) and gently vacuum filtered (max. suction pressure 20 kPa) through a precombusted GF/F filter directly into acid-washed 30 ml HDPE sample bottles using a custom set-up with an in-line polycarbonate filter holder. Prior to filtration, the syringes, the HDPE sample bottles and the filtration set-up were soaked in 0.4 M HCl for 24 h and then washed with MQ water and with the sample. DOC samples were immediately acidified with 80 µl of 18.5% HCl and stored in the dark at 4°C until analysis. Samples were analysed by the high-temperature catalytic oxidation method on a Shimadzu TOC-VCPH total organic carbon analyser. The instrument was calibrated with a 10-point calibration curve of serial dilutions from a potassium hydrogen phthalate certified stock solution (1000 ppm Standard Fluka 76067-500ML-F). Consensus reference material provided by DA Hansell and W Chan of the University of Miami (batch 13, lot no. 08-13, 41–45 µmol C l−1) was used a positive control between every 10 samples. Each sample was measured with five replicate injections, and analytical precision was less than 3% of the certified value.
(c) Physiological measurements
Coral fragments (n = 8 for each coral genus during each season) in the incubation chambers were maintained in the flow-through aquarium described above during measurements of net productivity (Pn), dark respiration (R), N2 fixation, and POC and DOC release to ensure in situ temperature and light intensity in the chambers. Fragments were allowed to recover from fragmentation for at least one week prior to the measurements. Ceramic tiles were cleaned of sediment and epibionts with a fine brush immediately prior to the start of the incubations. Only visually healthy and entirely healed fragments were used for incubation experiments. Incubation chambers were filled with the seawater from the aquarium and corals were transferred into the chambers taking particular care to prevent any air exposure. Each chamber contained a stir bar powered by a submersible magnetic stirrer (600 r.p.m., Cimarec i Telesystem Multipoint Stirrers, Thermo Scientific) to ensure water mixing (and gas equilibration with the headspace, if applicable). Temperature and light intensity were monitored during all measurements by placing a data logger in one additional chamber. Following each incubation, the dissolved oxygen (O2) concentration in the incubation water was measured to ensure that oxic conditions in the chambers were maintained. Rates of Pn, R, N2 fixation, organic matter (OM) release as well as Symbiodinium density and chl a content were normalized to skeletal surface area of the coral fragments, measured using the advanced geometry protocol of Naumann et al. . Pn and R of the coral fragments were assessed by their O2 fluxes in closed-cell respirometric glass chambers (1 l) according to Bednarz et al.  and gross productivity (Pg) was calculated (Pg = Pn + |R|) for each specimen. Coral fragments were returned to the maintenance aquarium and after approximately 4 h N2 fixation was quantified by using an adapted acetylene (C2H2) reduction technique [32,33], as fully described by Bednarz et al. . N2 fixation incubations lasted for 24 h, starting and ending before sunset (approx. at 17 : 00), and the incubation time was kept shorter than in previous studies on corals  in order to minimize stress for the organisms. Coral fragments were returned to the maintenance aquarium and after approximately one week organic matter release by the four coral genera was quantified using the established beaker incubation method [34–36]. Details on the method can be found in the electronic supplementary material.
(d) Tissue parameters
For Symbiodinium density and chl a content analysis of the corals, a subset of the incubated coral fragments (n = 4) was stored at −20°C prior to analysis in spring and in summer. Coral tissue was removed from the skeleton using a jet of pressurized air and 0.2 µm-filtered seawater. The resulting tissue slurry volume was brought to 50 ml with 0.2 µm-filtered seawater and homogenized using a vortex. One aliquot of 9 ml was subsequently subsampled, fixed in 3 ml 16% formaldehyde solution (final concentration 4%) and preserved at 4°C until Symbiodinium density analysis. Subsequently, the homogenate was centrifuged at 5000 r.p.m. for 5 min, the supernatant removed and the pellet resuspended with a known volume of 0.2 µm-filtered seawater. The number of Symbiodinium cells was counted using an improved Neubauer haemocytometer . The total number of Symbiodinium cells in the initial 50 ml slurry was calculated and normalized to coral surface area (cells cm−2). Another aliquot of 5 ml was taken from each homogenate and immediately centrifuged at 5000 r.p.m. for 5 min. The supernatant was discarded and the pellet frozen at −20°C overnight to break the cells. The following day the pellet was resuspended in 10 ml 90% acetone, and chl a was extracted for 24 h in the dark at 4°C. After another centrifugation run, chl a was measured using the non-acidification fluorometric method  on a Trilogy fluorometer equipped with the non-acidification chl module (CHL NA no. 046). Chl a content was normalized to coral surface area (µg chl a cm−2) and to the total number of Symbiodinium of each fragment (pg chl a cell−1).
(e) Data analyses
To identify differences in Pn, R, Pg and N2 fixation for the different coral genera during the four seasons, we used a two-way ANOVA design with the factor ‘season’ (fixed and orthogonal, four levels) and the factor ‘genus’ (fixed and orthogonal, four levels) and n = 8. POC, DOC and TOC release rates were tested for differences with the same design but with n = 6. Symbiodinium density, areal chl a content and chl a per Symbiodinium cell were tested for differences with the same design but with the factor ‘season’ having two levels only (spring and summer) and n = 4. Data were visually inspected for normality using q–q plots, tested for homogeneity of variances using Cochran's C-test and transformed if necessary. In case of a significant interaction term, Student–Newman–Keuls comparison tests were applied a posteriori to identify significant differences. To visualize multivariate changes in environmental variables among the four seasons, a principal coordinates analysis was performed on a Euclidean distance matrix of previously normalized environmental data resulting from weekly averages. Differences in response to seasonality were tested using a PERMANOVA test . The analysis was conducted using the Euclidean distance as coefficient of dissimilarity on previously normalized data. Type 3 (partial) sum of squares was used with unrestricted permutation of raw data (9999 permutations). All multivariate analyses were run using the PERMANOVA tool included in the Primer 6+ package. Linear regressions were used to explore univariate relationships between coral productivity and N2 fixation over the four different seasons. CO2 and N2 fixation rates were established using proxies (O2 for CO2 and C2H4 for N2). However, there is a direct relationship between these measures and the actual CO2 and N2 fixation rates [32,40,41], and the proxies chosen are routinely considered accurate surrogates for measurements of the latter rates in hermatypic corals. Linear models were fitted to the data, and the significance of relationships was tested using the Pearson r-test in Statistica v. 12 software. Linear regressions were performed for each season separately to avoid spurious correlations induced by the dependence of the two biological processes on the different seasonal environmental conditions (e.g. temperature), and the presence of outliers was tested with box and whisker plots for each variable in Statistica v. 12 software. Given that patterns in physiological measurements were similar across coral genera, regressions were performed with the four genera combined in single seasonal datasets.
(a) Physiological measurements
Average gross primary productivity ranged from 17.7 ± 1.9 in Stylophora in winter to 31.8 ± 0.9 µg O2 cm−2 h−1 in Goniastrea in summer. Coral productivity showed no consistent trend according to the season or to the coral genus (figure 1b), although lower coral productivity was found in winter in Stylophora and Goniastrea (electronic supplementary material, tables S3 and S4), also compared with Acropora and Pocillopora (electronic supplementary material, table S5). N2 fixation ranged from 0.01 ± 0.01 in Acropora in winter to 0.56 ± 0.12 nmol C2H4 cm−2 h−1 in Goniastrea in summer. There was a strong seasonal pattern (figure 1c; electronic supplementary material, table S3 and S4), with N2 fixation rates in summer being approximately five- to 10-fold higher than in spring in all coral genera, and up to 20-fold higher in Acropora. N2 fixation was also significantly higher in autumn compared with winter and spring in Goniastrea, whereas no consistent pattern was detected among the different coral genera (electronic supplementary material, table S5). N2 fixation rates varied on a diurnal basis, with the highest rates generally occurring in the early morning (electronic supplementary material, figure S2). POC release by corals was significantly lower in winter compared with summer (electronic supplementary material, table S3 and S6). However, dissolved (DOC) and total organic C (TOC) fluxes induced by the four hard coral genera were not different among genera and seasons (electronic supplementary material, figure S3 and table S3).
(b) Symbiodinium and photopigment content
Symbiodinium density and areal chl a content were significantly lower in summer compared with spring in all coral genera except for Symbiodinium density in Pocillopora (figure 2a,b; electronic supplementary material, tables S3, S4 and S6). Conversely, chl a content per Symbiodinium cell was significantly elevated in summer in all coral genera except Acropora (figure 2c; electronic supplementary material, table S3 and S4).
(c) Relationships between N2 fixation and gross productivity
There was a significant relationship between N2 fixation and gross primary productivity (figure 3) in summer and autumn (summer: Pg = 21.076 + 15.417x, r = 0.551, p = 0.001; autumn: Pg = 21.565 + 16.940x, r = 0.475, p = 0.006) but not in winter and spring (winter: Pg = 22.488 + 30.553x, r = 0.205, p = 0.261; spring: Pg = 27.517 + 8.890x, r = 0.128, p = 0.484).
In this study, we provide the first measurements of N2 fixation in scleractinian corals subject to seasonally changing environmental conditions. Our finding of consistent N2 fixation activity reveals that diazotrophs were active and associated with all four hermatypic coral genera (Acropora, Pocillopora, Stylophora and Goniastrea) throughout the year, despite the pronounced seasonality. This association has previously been described in physiological and molecular studies [10–15,42], and N2 fixation rates measured in this study are in the range of those reported for corals in the literature , supporting the hypothesis of a role of diazotrophs for coral holobiont functioning. N2 fixation increased significantly in all corals during summer (five- to 10-fold), similar to the seasonal pattern recently found for N2 fixation activity in soft corals  in the Gulf of Aqaba. Conversely, N2 fixation remained low in all other seasons and particularly in winter and spring, probably owing to the higher availability of combined N during the water-mixing season . In summer, corals exhibited a significant decrease in Symbiodinium density and areal chl a content, whereas a significant increase in chl a per Symbiodinium cell was detected. The observed pattern of seasonal variability in Symbiodinium abundance represents an established phenomenon in tropical and subtropical scleractinian corals caused by an increase in photosynthetically active radiation during summer, which can enhance Symbiodinium loss owing to concomitant warmer temperatures [44,45]. Moreover, growth efficiency and mitotic index of the symbiotic algae are temperature-dependent and can decrease with temperatures exceeding 25°C . Despite the observed seasonal variability in Symbiodinium population density, all corals maintained high gross productivity and organic matter release throughout the year, indicating high metabolic plasticity to seasonal variation in environmental conditions.
For tropical coral reefs, it is widely accepted that N is the most limiting nutrient to coral primary productivity . Phosphorus (P) may also be limiting to coral metabolism , particularly in the case of an imbalanced nutrient supply with excess inorganic N (e.g. of anthropogenic origin) . At our study site, the reef-surrounding seawater exhibits particulate organic C to N (POC : PN) and dissolved inorganic N to phosphate (DIN : PO43−) ratios that are respectively higher and lower than the Redfield ratio (106 : 16 : 1) during all seasons (electronic supplementary material, table S1), suggesting that N rather than P is the limiting nutrient in the Gulf of Aqaba. The lack of effect of phosphorus additions on planktonic rates of N2 fixation corroborates this suggestion . N limitation causes decreased pigment content in Symbiodinium , with N deficiency resulting in up to 86% reduction in chl a per Symbiodinium cell after only one week . Conversely, corals investigated here showed higher chl a per Symbiodinium cell during the highly N-depleted stratified summer period, indicating that the endosymbiotic dinoflagellates successfully photoadapted, overcoming ambient limitation of N.
The positive linear relationship between N2 fixation and gross productivity in summer and autumn points to the N fixed by diazotrophs as a source of N for photosynthesis. However, in order for both processes to occur in the coral holobiont, a mechanism is required to protect nitrogenase (the enzyme responsible for N2 fixation) from molecular and reactive oxygen species accumulating during photosynthesis. Our results are in line with the work by Lesser et al. , who reported the highest N2 fixation occurring during a window of time in which subsaturating irradiances are still high enough to sustain photosynthesis but oxygen levels are likely to be at their lowest diurnally. This suggests that a temporal separation may be the prevalent N2 fixation strategy of diazotrophs in corals. However, rates measured here are probably the result of the activity of a diverse diazotrophic community, potentially employing different N2 fixation strategies to prevent oxygen inhibition of nitrogenase. In fact, recent studies have shown that cyanobacteria are only part of a diverse community of diazotrophs in the coral tissue and mucus surface layer [12,13,50], and in this study we estimated the ecological significance of N2 fixation at the diazotrophic community level, without selecting coral colonies according to their putative association with cyanobacteria.
There is debate in the literature whether the fixed N is subsequently acquired by Symbiodinium in the coral host. Lesser et al.  provided evidence supporting its utilization through an analysis of δ15N stable isotope signatures, whereas recently Grover et al.  found no incorporation using the 15N2-tracer addition technique. In our study, we used a modified acetylene reduction method that has been demonstrated to accurately measure N2 fixation rates . Although we do not provide direct evidence of a physiological mechanism linking these two processes, we found a positive linear relationship between N2 fixation and coral productivity in summer and autumn. This strongly suggests that active diazotrophs provided the required N to sustain Symbiodinium photosynthesis during the nutrient-depleted season, whereas the energy required for N2 fixation was obtained in the form of photosynthates (i.e. products of photosynthesis) from Symbiodinium within the coral host. This result warrants further investigation in future studies to determine the nature of this relationship. Similar scenarios of mutualistic association are well documented in land plants where heterotrophic diazotrophs in roots, such as rhizobia, fuel the plant with bioavailable N, while relying on plant photosynthesis to carry out the energetically demanding process of N2 fixation. Unique to the coral system is that photosynthetic eukaryotic algae (Symbiodinium), prokaryotes (N2 fixers) and the animal host (coral) cooperate, facilitating the nutritional success of the entire group of partners. While diazotrophic cyanobacteria were at first suggested to play a key role in corals [10,11], recent studies have shown that rhizobia (α proteobacteria) are also common partners in the coral holobiont [12,13,50], and that actinobacteria capable of N2 fixation are found intracellular within the endosymbiotic dinoflagellates in corals , further supporting the hypothesis of a mutualistic association. However, these studies also outline a more complicated picture than previously thought. It is clear that more research is needed that aims at identifying the key microbial players involved in N2 fixation and their functional roles in corals. This should combine molecular approaches with ecophysiological measurements, and further investigate the potential for shifts in coral-associated diazotrophic communities owing to laboratory conditions during coral maintenance and incubation measurements differing from in situ conditions.
Based on our measurements and established literature values (see electronic supplementary material), we propose here a mechanistic C and N flux model of the coral holobiont to estimate the importance of N2 fixation relative to other N sources in contributing to the Symbiodinium N demands (figure 4). In spring, when ambient DIN concentrations and uptake rates are highest, the coral host is replenished with N from the surrounding water (figure 4a) and actively limits algal population growth by removing excess nutrients from the intracellular milieu surrounding the Symbiodinium [21,46]. Conversely, during the warm, high-irradiance, nutrient-depleted summer (figure 4b), environmental conditions trigger a dynamic expulsion of symbiotic algae, leading to smaller Symbiodinium populations with increased chl a content per unit cell [44,46]. However, N fixed and transferred by the diazotrophic community to the Symbiodinium prevents N deficiency within the holobiont. Symbiodinium are thus provided with sufficient N to sustain biosynthesis of the chlorin ring of chl a, the light-capturing engine of photosynthesis. In summer, the contribution of N2 fixation to Symbiodinium N demands (CZND) is 11%, comparable with the contribution of DIN uptake from reef-surrounding waters (figure 4b). According to our model, these two N sources together with heterotrophic N uptake facilitate 98% of the total CZND, implying an almost complete N sufficiency of Symbiodinium (figure 4; electronic supplementary material, table S1). Heterotrophy always represented the major N acquisition pathway in our model, in line with previous studies on the ecological relevance of this process in corals . However, although diazotroph-derived N may only constitute a smaller fraction of the total CZND compared with heterotrophy, our results suggest that it may represent a crucial one in oligotrophic waters.
The dynamic equilibrium between the associated eukaryotic and prokaryotic communities in the coral holobiont (figure 4) sets the basis for high functional stability, enabling high gross productivity despite changes in environmental conditions. Corals may achieve this functional stability by adjusting population densities of symbiotic Symbiodinium and by benefiting from N fixed by symbiotic diazotrophs (figure 2a and figure 1c, respectively). As a consequence of stable year-round primary productivity, hermatypic corals are able to sustain high production and regeneration of their mucus surface layer and its concomitant release as organic matter (electronic supplementary material, figure S3). On a daily basis, we estimate that 10–12% of the total C and 14–28% of the total N acquired by the holobiont are lost via organic matter release (LOC and LON, respectively; figure 4). This released organic matter fuels biogeochemical cycles and provides a food source for other reef organisms, thereby importantly contributing to ecosystem functioning and productivity [5,6].
While experimental evidence tracking the fate of N in the symbiosis will be essential to confirm this model of coral holobiont functioning, recent research from various reef locations has identified diazotrophic assemblages in different coral species to be spatially and temporally consistent [12,13,50]. Concurrently, numerous tropical reef systems worldwide that are far from coastal influence experience extremely low seawater N concentrations, comparable with those measured in summer in the Gulf of Aqaba . N2 fixation may therefore be of widespread importance for sustaining coral productivity in these oligotrophic reef systems, and could explain why no lower thresholds for dissolved nutrient concentrations have been yet found to limit coral reef growth . However, many coral reef areas worldwide are increasingly exposed to nutrient enrichment and eutrophication resulting from human influence on the marine environment . In these reef systems, the functional trait of N2 fixation may become increasingly redundant if dissolved nutrients are readily available.
The care of animals used in experimentation was in accordance with institutional guidelines.
Additional data supporting this article have been uploaded as part of the electronic supplementary material.
U.C. and C.W. wrote the manuscript and directed its implementation. U.C., V.N.B., N.v.H., L.R. and M.S.N. conducted the field activities and performed the analyses. U.C. and M.S.N. developed the C and N flux model. R.A.F. assisted in the design of the acetylene reduction assays. M.M.D.A. assisted in advising on field activities. All authors helped in drafting the article and revising it critically for important intellectual content.
The authors declare that they have no competing interests.
This study was supported by the German Research Foundation (DFG) through grant Wi 2677/6-1 to C.W. V.N.B. was supported by a stipend of ‘Evangelisches Studienwerk Villigst e.V.’. R.A.F. was supported by the Knut and Alice Wallenberg Foundation (Sweden).
We thank F. Al-Horani, S. Helber and the MSS scientists and staff for fieldwork assistance and logistical support. We also thank S. Wilson and D. G. Capone for advice on experimental design and data analyses, N. Rädecker, M. Birkicht and D. Peterke for their help during sample analyses, and M. Huettel for his comments on the manuscript. We thank three anonymous reviewers and David Bourne from Axios Review for their time and effort in providing helpful suggestions and comments on the manuscript.
- Received September 17, 2015.
- Accepted September 30, 2015.
- © 2015 The Author(s)