Plant species identity and arbuscular mycorrhizal status modulate potential nitrification rates in nitrogen-limited grassland soils
Summary
1. Arbuscular mycorrhizal (AM) fungi and ammonia oxidizers (AO) represent key soil microbial groups regulating nitrogen (N) cycling in terrestrial ecosystems. Both utilize soil ammonium-N reserves for N assimilation, whilst the latter, through autotrophic nitrification, drive ammonia oxidation to highly mobile nitrate-N.
2. An incompatible interaction between root symbiotic AM fungi and AO was hypothesized and evaluated in plant–species-rich, N-limited Mediterranean grassland soils. Such an outcome would be manifested in a negative relationship between plant mycotrophy and local soil potential nitrification rates (PNR), a standard functional measure of ammonia-oxidizing activity in soils.
3. In three independent mesocosm experiments, grassland soils that supported monocultures of mycotrophic, as opposed to weakly and non-mycotrophic, plants exhibited significantly lower PNR. Under field conditions in a fourth experiment, we verified that soils from stands of weakly mycotrophic Agrostis capillaris sustained higher PNR than counterparts supporting highly mycorrhizal Prunella vulgaris and Fragaria vesca.
4. Discussion of mycotrophy-related modulation of AO activity centres on whether the observed relationships highlight evidence for either direct competition or a functionally important example of plant–microbial allelopathy.
5. Synthesis. Substantial evidence has been presented confirming (i) plant species identity-related regulation of PNR and (ii) negative relationships between plant mycotrophy and plant species-mediated impact on PNR in N-limited Mediterranean grassland soils. Likely mechanisms (i.e. competition and/or allelopathy) that underpin this functionally significant plant–microbe–soil relationship controlling the fate of ammonium-N require urgent elucidation in N-deficient ecosystems.
Introduction
Nitrogen (N) availability is a key regulator of net primary productivity (NPP) in terrestrial ecosystems (Chapin et al. 2004) through direct control of the growth of not only plants (Craine, Morrow & Stock 2008) but also soil microbes (Demoling, Figueroa & Bååth 2007) that underpin plant diversity and productivity (van der Heijden, Bardgett & van Straalen 2008). Plants and microbes are often believed to compete for available soil N resources (e.g. Song et al. 2007), but while in the short term, soil microbes are regarded as superior competitors (Hodge, Robinson & Fitter 2000), plants have the advantage of a longer life span that does allow short-term utilization of spikes in N availability (Kaye & Hart 1997; Hodge, Robinson & Fitter 2000). Terrestrial N cycling is, hence, governed by the complicated spatio-temporality of N availability and continues to represent a dynamic field for targeted exploration of plant–microbe interactions and consequences for NPP (Bardgett et al. 2005).
The two key forms of mineral N available to plants are ammonium (
) and nitrate (
) although plant species-specific differences with respect to the preferred form of inorganic N and the strength of the preference are clearly identifiable in the literature (e.g. Falkengren-Grerup 1995). As both these forms of inorganic N exhibit differential diffusion coefficients in soils, with the coefficient for 
being much lower than 
, a preference for one of the two available N forms may signify conformation to a particular adaptation strategy. Unlike 
, which is preferentially held on soil cation exchange sites (Sigunga, Janssen & Oenema 2002), 
, the anionic form, is highly mobile and can be rapidly lost from the soil system or further reduced to NOx and N2 through the activities of microbial denitrifiers (Prosser 2007). Moreover, assimilation of 
, although more expensive regarding the photons, water, Fe, Mn and Mo required for the extra step of reducing 
to 
(Raven, Wollenweber & Handley 1992), requires a less extended root system and thus supports a greater maximum specific growth rate (Raven, Wollenweber & Handley 1992; Rees 2007). Nitrification represents a significant step in the soil N cycle as the reduction of 
to 
, via an intermediate step that generates nitrite (
), results in increased N mobility and potential loss from the soil system through nitrate leaching to groundwater (Prosser 2007). The rate-limiting step in nitrification is the reduction of 
to 
which is mediated by a group of predominantly autotrophic bacteria and archaea that are collectively known as ammonia oxidizers (AO) (Prosser 2007). Potential nitrification rates (PNR) are commonly determined to assess impacts of fertilization and disturbance on soil N availability and productivity (Avrahami, Conrad & Braker 2002; Shen et al. 2008).
Arbuscular mycorrhizal (AM) fungi are ubiquitous symbionts of plant roots (Smith & Read 2008). A rough estimate suggests that over 80% of terrestrial plants engage in AM symbioses (Wang & Qui 2006). The most significant gain for the plant symbiont from these mycotrophic associations is an improved ability to mobilize, translocate and assimilate nutrients, particularly phosphorus (P), via the development of an extensive soil foraging extraradical hyphal network termed the mycorrhizosphere (Smith & Read 2008). An often overlooked aspect in relation to competition for soil N is the mediation of AM fungi, as the dominant view is that their direct contribution to plant mineral N nutrition is of minor importance (Hodge, Robinson & Fitter 2000; Smith & Read 2008). More recent investigations have shown that, in symbiosis, AM fungi either preferentially (Govindarajulu et al. 2005) or exclusively (Tanaka & Yano 2006) assimilate inorganic N in the form of 
. In this respect, engagement in symbiotic AM associations may have a critical impact on plant preference in relation to the form of inorganic N (Yoshida & Allen 2001). In N-limited ecosystems such as native grasslands, there is common engagement of plants in diverse AM associations (Smith & Read 2008) that result in the development of an extensive plant mycorrhizosphere-driven 
sink in soils. As AO are weak competitors for soil 
(e.g. Verhagen, Duyts & Laanbroek 1992; Bollmann, Bår-Gilissen & Laanbroek 2002), preferential demand and competition for 
in the AM mycorrhizosphere by participating AM fungi and other associated soil microbiota could lead to a rapid and prolonged decline in abundance of poorly competitive autotrophic AO (Fig. 1). Under such conditions of low autotrophic AO abundance, significantly reduced nitrification rates, and thus a decline in mobile 
N, would be expected in the mycorrhizosphere of host plants, which, by definition, display high levels of mycotophy.
Hypothetical view of nitrogen cycling in the rhizosphere versus mycorrhizosphere that incorporates both extracellular enzyme processes and microsite processes that emphasize competition for mineralized ammonia (ΝΗ4+) on a relatively N-rich microsite. Heterotrophic microbes (including AM hyphae) represent the most effective competitors for 
(Kaye & Hart 1997). By contrast, nitrifiers are believed to be weak competitors for soil ΝΗ4+ (Bollmann, Bår-Gilissen & Laanbroek 2002). Thus, the development of a high-density network of AM extraradical hyphae will result in reduced 
availability for AOs and nitrification. Based on a modified figure from Schimel, Bennet and Fierer (2005).
A series of four experiments were designed to test the hypothesis that plant species identity is an important regulator of PNR in N-limited grassland biomes. We further hypothesized reduced PNR in soils supporting mycotrophic, as opposed to non- or weakly mycotrophic, host plants because of the negative effects of higher densities of AM extraradical fungi on autotrophic AO activities. The three mesocosm-based experiments were aimed at assessing soil PNR and mycotrophy relations under manipulated conditions whereby control of AM extraradical mycelial density levels was achieved either through growth of monocultures of representative grassland plant species exhibiting differing levels of mycotrophy or through fungicide (Benomyl) application (Fitter & Nichols 1988) in containerized grassland soil. The fourth experiment was designed to assess and confirm identified plant mycotrophy-PNR relationships under field conditions. The experimental design stems from commonly observed positive correlations between AM root colonization and extraradical mycelial density (e.g. van Aarle & Olsson 2003). Moreover, there is increasing evidence of a strong positive correlation between evolutionary traits that promote high AM root colonization with those that promote increased extraradical hyphal exploration of surrounding soils (e.g. Powell et al. 2009). Despite plant species mycotrophy being regarded as a largely conserved attribute amongst plant species (e.g. Wilson & Hartnett 1998; Wang & Qui 2006), to strengthen inferences we carried out further evaluation of AM colonization status of the target plants. Criteria for plant selection were as follows: (i) plants species represented typical perennials of mesotrophic grasslands in Greece and (ii) excluded annual pioneers that are commonly non-mycorrhizal and could exhibit a distinct ecophysiology (e.g. Ripley 2002).
In the first experiment, soils supporting plant monocultures of three common species of perennial C3 grasses and two forbs were repeatedly sampled to assess extent and reproducibility of PNR over time (Experiment 1). For the second experiment, five perennial forb species were subjected to a factorial benomyl fungicide treatment to control AM colonization in the PNR assessment (Experiment 2). In this experiment, one of the forbs, Plantago lanceolata, was further subjected to factorial N and P fertilization to elucidate the nature of nutrient limitation. In the final controlled experiment, mycorrhizal and non-mycorrhizal forbs were investigated under manipulated N fertilization to confirm PNR relationships identified in the previous experiments (Experiment 3). The field-based study (Experiment 4) allowed further comparison of PNR activity in soils taken from under a grass and forb species present in factorially N-fertilized plots in grassland communities in northern Greece.
Materials and methods
Experiment 1
In the spring of 2005, the perennial C3 grasses Festuca ovina (L.), Lolium perenne (L.) and Poa pratensis (L.), and the forbs Plantago lanceolata (L.) and Rumex acetosa (L.) were sown as monocultures, each in four containers (40 × 30 × 25 cm), located in an experimental area inside the University Forest of Taxiarhis (40°23′ N, 23°28′ E, 850 m a.s.l.) as part of a larger experiment, presented in detail in Karanika et al. (2007). The soil was a sandy loam (43% sand, 22% silt and 35% clay), and 240 seeds of each species were sown per container into soil, collected from a neighbouring grassland. Containers were regularly fertilized with a final fertilizer application in April 2006. In August 2007, each container was divided into two sub-containers that resulted in eight replicates per plant species. Plant growth N and phosphorus (P) limitation during the subsequent 2-year experimental period, to July 2009, was revealed in the plant growth data. Means of residual soil 
-N, 
-N and 
: 
ratio under individual plant species monoculture treatments in December 2007 are provided in Table S1. Four harvests were carried out on the following dates: 20 April 2008, 5 July 2008, 12 April 2009 and 1 July 2009 to determine PNR of soils supporting the respective plant species monocultures. Two additional harvests were carried out on the following dates: 23 December 2007 and 20 December 2008 to determine soil nutrient status in the containers (Fig. S1). The aim of this first experiment was to test whether a consistent repression of AO activity could be detected in the mycorrhizosphere of the plants that invested more extensively in AM associations irrespective of the time of harvesting. In support of our hypothesis, soils supporting growth of the mycotrophic P. lanceolata and L. perenne (Karanika et al. 2008b) were expected to exhibit the lowest PNR in contrast to higher PNR in soils under the weakly mycorrhizal R. acetosa (Karanika et al. 2008b).
Experiment 2
Soil was collected from a second upland grassland site at the University Forest of Taxiarhis in northern Greece, sieved through a 5-mm mesh and homogenized before use. Soil pH and nutrient status of the soil are presented in Table 1. The experiment was set up in a glasshouse 10 km from Thessaloniki, Greece on 14 November 2008 in pots (13 cm height, 17 cm diameter) filled with 2 kg equivalent dry weight (EDW) soil under natural lighting conditions at a constant temperature of 20 °C. The five forb species (Plantago lanceolata L., Plantago major L., Prunella vulgaris L., Rumex acetosa L. and Rumex acetosella L.) were each grown as monocultures in sixteen replicate pots. Each pot was sown with 50 seeds and watered with tap water as required on a regular basis. Throughout the experiment, eight replicate pots received tap water only while the other eight also received 100 mg of the fungicide benomyl (Dupont Inc., Wilmington, Delaware, USA) in the form of a soil drench twice a month from 2 December onwards. Pots were regularly randomized on the bench until harvest on 14 April 2009. A further sub-experiment that lasted 71 days and comprised of 24 pots (8.5 cm height, 10.5 cm diameter) divided in three treatments (control, N-fertilized: 30 mg N per pot in the form of urea and P-fertilized: 20 mg P per pot in the form of superphosphate) was sown with P. lanceolata seeds to examine whether P. lanceolata growth was N or P limited. The aims of the experiment were (i) to verify the negative relationship between plant mycotrophy and PNR for a range of plants that formed a wider gradient with respect to their mycotrophy and (ii) to test whether a potential detrimental effect of benomyl manipulation on AM occurrence would result in a decline of inter-species differences with respect to PNR in soils of the examined plants. We expected that soil retrieved from monocultures of the three mycotrophic plants, P. lanceolata, P. vulgaris and P. major, would support significantly lower PNR in the absence of benomyl than soil retrieved from monocultures of R. acetosella and R. acetosa. Following the addition of the fungicide, we wanted to test for any increase in PNR because of fungicide-related decline in mycotrophy that would be expected to have been more pronounced for the most mycotrophic plant treatments. Thereby, we expected that, following benomyl addition, there would have been a decline in inter-species treatment differences with respect to PNR.
| Soil | pH | Organic C (%) | Total N (g kg−1) | Olsen P (mg kg−1) | Notes |
|---|---|---|---|---|---|
| Experiment 1 | 6.1 | 0.74 | 1.13 | 1.73 | Assayed in soil from the 1st harvest (December 2007) |
| Experiments 2 and 3 | 6.5 | 4.2 | 3.6 | 12.2 | Soil status was assayed before initiation of Experiment 2 |
| Polypotamos (Experiment 4) – Control | 5.6 | 4.41 | 6.7 | 11.7 | Assayed in July 2008 |
| Polypotamos (Experiment 4) – N treatment | 5.1 | 5.21 | 5.86 | 13 | Assayed in July 2008 |
Experiment 3
Soil was collected from the same upland grassland site as in Experiment 2, sieved through a 5-mm mesh and homogenized before use. The experiment, following the design of Experiment 2, was set up on 15 November 2009. The three forb species (Prunela vulgaris (L.), Silene vulgaris (Moench) and Rumex acetosella (L.)) were each grown as monocultures in eighteen replicate pots. Nine of the replicate pots received only tap water, whereas on 29 November 2009, the other nine, additionally, received a single application of 200 mg per pot NH4NO3 in the form of an aqueous solution. Pots were regularly randomized on the bench until harvest on 2 May 2010. The aims of the experiment were (i) to verify the negative relationship between plant mycotrophy and PNR through incorporation of a distantly related (to R. acetosella) perennial, non-mycorrhizal plant, S. vulgaris and (ii) to test whether N additions would induce a decline in the inter-species differences with respect to PNR similar to those that were noted in Experiment 2. In the absence of N fertilization, soil retrieved from P. vulgaris monocultures was expected to support the lowest PNR. It was expected that the addition of N would not result in a decline in inter-species treatment differences with respect to PNR similar to that noticed in Experiment 2 for benomyl treatments.
Experiment 4
Soil was collected as cores (5 cm internal diameter, 8 cm length) on 7 July 2009 from a separate grassland fertilization experiment close to Polypotamos (40°48′ N, 21°23′ E, 1340 m a.s.l.) in northern Greece. Details of the experimental site are reported elsewhere (Karanika et al. 2008a,b). In brief, the mesotrophic grassland (pH c. 5.6) hosted a plant community that resembled the MG5 plant community Cynosurus cristatus–Centaurea nigra grassland with a Galium verum sub-community (Rodwell et al. 2000). During the 2003–2004 growth season, plots (1.5 × 1.5 m) were established within areas that received factorial application of two levels of nitrogen (0 and 15 g N m−2 year−1 applied as NH4NO3) and two levels of P (0 and 10 g m−2 year−1 applied in the form of superphosphate). The fertilization regime had been continuously applied on a yearly basis since the 2003–2004 season. Research focused on PNR in the rhizosphere of Agrostis capillaris across the experimental site over 3 years had revealed that, commonly, a minimum of 45% of variance of PNR could be predicted when N fertilization status of the plots and pH were used as independent variables in multiple regression models (S.D. Veresoglou, O.K. Voulgari, A.P. Mamolos, R. Sen & D.S. Veresoglou, unpublished data). The cores (two per plot) were collected in a factorial manner from the control and N treatments of all eight blocks so that one core per plot was centred within a stand of A. capillaris plants and the other in a stand of a highly mycotrophic forb. As Prunella vulgaris does not occur in soils with a pH below 5, which was the case in some of the N-fertilized plots, the replacement plant species that was harvested in four out of the eight blocks (from both the control and the N fertilized plots) was Fragaria vesca. The aim of the experiment was to establish – following model comparison selection using the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC), which more readily punishes additional parameters – whether plant species identity could represent an additional predictive factor of PNR. For this reason, we attempted to rebuild the PNR models with incorporation of three dummy variables, one for each plant species. Models were built without any, with just one, or all possible combinations of two of these dummy variables and were subjected to model selection with AIC and BIC. It was expected that the optimal model would incorporate a minimum of one dummy variable that would signify that plant identity, also, has a significant effect on soil PNR under the tested field conditions. A further expectation was that A. capillaris (in the form of a dummy variable) in the models incorporated would have a positive effect (positive correlation coefficient) on PNR, whereas the dummy variables associated with the more mycotrophic P. vulgaris and F. vesca, a corresponding negative effect.
Analytical methods
In all cases, soil was harvested as cores (5 cm internal diameter, 8 cm length) and was sieved through a 2-mm sieve following transit at 4 °C to the laboratory. PNR assays were carried out within 48 h of harvest. Exchangeable 
and 
were assessed following extraction with 1 m KCl according to the salicylic method (Cataldo et al. 1979) and the colorimetric determination of 
(BSI 1984). PNR in soils was determined according to the protocol provided by Hart et al. (1994) with the addition of chlorate (Belser & Mays 1980). In brief, each sample consisted of 10 g of sieved (2 mm) soil that was mixed with 100 mL freshly prepared nitrification potential solution (1 mm potassium phosphate pH 7.2; 0.5 mm ammonium sulphate ((NH4)2SO4); 0.01 m ammonium chlorate in a 250-mL conical flask. The slurry was immediately agitated at 200 rpm in an orbital shaker (Stuart SSL1; Bibby Scientific Ltd, Stone, UK) to maintain good aeration. At 2, 8, 14 and 26 h after initiation of the procedure, 10 mL of the slurry was removed and filtered to allow concentration of residual 
to be determined. PNR for each sample was estimated from the slope of the regression on a four-point time course assay of nitrite (
) assayed according to Keeney & Nelson (1982). In Experiment 1, for the harvests on 20 April 2008 and 5 July 2008, no chlorate was added in the above-described PNR determinations, and the rate was calculated based on assay of 
. Kjeldahl N analyses of soil and plant tissue samples were carried out as described in Rowel (1994). Plant tissue P was assessed with Murphy & Riley (1962) following digestion of 0.5 g of plant tissue in the triple acid reagent 10:1:1 HNO3:H2SO4:HClO4 (Allen 1989). Assessment of AM fungal root colonization involved the staining of the roots with trypan blue (Phillips & Hayman 1970), and quantification of AM fungal colonization was achieved using the method of McGonigle et al. (1990).
Data analysis
Normality of the data sets was verified through Kolmogorov–Smirnov tests. In Experiment 1, to determine nitrification potential across all harvests, a repeated-measures anova approach was used for nitrification potential with the independent variables, harvest time and treatment. Specifically, the repeated-measures anova model comprised one factor between experimental units (treatment with five levels) and one factor within experimental units with repeated measures (harvest time with four levels). To verify equality of error variances at each harvest and equality of covariance matrices in the repeated-measures ANOVA design, Levene and Box tests were used, respectively.
In Experiment 2, an F test (Steel & Torrie 1980) revealed that benomyl manipulation had resulted in significantly higher variance in PNR (P = 0.007). Because homogeneity of variances is a prerequisite of parametric analysis of variance, we could not conduct a two-way anova. Instead, plant species impact on PNR in the controls and benomyl-treated samples were separately examined through two one-way anovas. To additionally address the impact of benomyl on PNR, individual t-tests were separately conducted for each treatment, between controls and benomyl-fertilized samples. In Experiment 3, no evidence against homogeneity of variances was detected, and PNR were examined with a two-way anova model.
In Experiment 4, PNR and pH values were analysed through a two-way anova with the factors N fertilization status and plant species functional group (grasses-A. capillaris or non-legume dicotyledon-forbs). With regard to PNR models, they all incorporated N fertilization status in the form of a dummy variable and pH as explanatory variables and additionally none, one, or all possible combinations of two of the dummy variables that were associated with plant identity. Model selection was carried out in R version 2.12.0 (R Development Core Team 2008) with AIC and BIC commands (library nlme). Preliminary testing of the independent variables was carried out to ensure no autocorrelation (Whittingham et al. 2006). For post hoc comparison, the least significance difference (LSD) statistic was utilized. Significance of individual tests was assumed for P < 0.05.
Results
Experiment 1
The effects of harvest (P < 0.001) and the interactive effects of harvest × treatment (P = 0.001) on soil PNR were identified in the applied repeated-measures anova. This confirmed plant species-specific, and medium-term seasonal, effects as being major determinants of the direction of plant-identity-related impacts on soil PNR. Pairwise treatment comparisons following repeated-measures ANOVA revealed significant contrasts between P. lanceolata–R. acetosa (P = 0.007) and L. perenne–R. acetosa (P = 0.003) while there was moderate evidence (P = 0.084) to indicate that the L. perenne treatment exhibited lower PNR than the P. pratensis treatment. A trend was detected of lower soil PNR in the L. perenne, as compared to the F. ovina, treatment (P = 0.111) (Fig. 2).
Soil potential nitrification rates following a repeated-measures anova of nitrification potential data in containerized grassland plant monocultures (Experiment 1). Data P values represent the level of significance of the one-way analysis of variance. Error bars represent 95% confidence intervals. Treatment means that do not share common letters are significantly different according to the LSD at P = 0.05. Abbreviations: Lp: Lolium perenne, Fo: Festuca ovina, Pp: Poa pratensis, Pl: Plantago lanceolata, Ra: Rumex acetosa. For all harvests n = 8.
Experiment 2
N-limitation on plant growth of P. lanceolata was confirmed in the sub-experiment (Fig. 3). With respect to AM root colonization, large inter-species differences were detected (Table 2). Benomyl, as expected, effectively reduced (P <0.001) AM root colonization in all plants except R. acetosella (Table 2). Addition of benomyl resulted in significantly (P < 0.05) higher plant tissue N content and increased dry weight in all treatments (Fig. S3) indicating – because the corresponding increase of plant P content, even for non-mycorrhizal plants, was disproportionately small – that benomyl impact was predominantly through addition of a considerable amount of benomyl-derived N and not mineralization of fungal biomass. Inorganic N in soils could be found mainly in the form of 
(Fig. S2).
Means ± SE total dry weight of Plantago lanceolata grown in grassland soil (sub-experiment of Experiment 2). The experiment lasted from 8 March until 18 May 2010 with 24 pots containing grassland soil that were each sown with 20 seeds. Treatment means that do not share common letters are significantly different according to LSD at P = 0.05 (n = 6).
| Plant species | Manipulation | Arbuscucles (%) | Vesicles (%) |
|---|---|---|---|
| Rumex acetosella | Control | 0 (±0)a | 0 (±0)a |
| Benomyl | 0 (±0)x | 0 (±0)x | |
| Rumex acetosa | Control | 18.2 (±1.0)b | 14.4 (±1.2)b |
| Benomyl | 13.9 (±1.3)y | 11.2 (±1.3)y | |
| Plantago lanceolata | Control | 43.1 (±2.1)d | 31.8 (±1.5)d |
| Benomyl | 24.8 (±1.4)w | 20.6 (±1.7)w | |
| Plantago major | Control | 26.4 (±0.6)c | 21.6 (±0.9)c |
| Benomyl | 18.8 (±2.0)z | 16.6 (±1.9)z | |
| Punella vulgaris | Control | 28.0 (±1.3)c | 23.5 (±0.9)c |
| Benomyl | 21.4 (±0.9)wz | 17.8 (±0.9)wz |
Plant species effect on PNR was assessed through two one-way anovas (Fig. 4a). In the one-way anova that addressed PNR in the controls, a species effect was found to be significant (P = 0.002). Non-mycorrhizal R. acetosella (Table 2) had significantly higher (P < 0.05) PNR than all three highly mycorrhizal plant treatments (Fig. 4a), whereas the significance of the differences among second-highest, poorly mycotrophic R. acetosa and the three known mycotrophic species (Table 2) were at P < 0.05 (P. major and P. vulgaris) and P < 0.1 (P. lanceolata) (Fig. 4a). By contrast, in the benomyl-treated samples, the plant species effect on PNR was only moderately significant (P = 0.059), and the only remarkable differences (at P = 0.1) were those between P. lanceolata on the one hand and P. vulgaris and R. acetosa on the other (Fig. 4b). Individual t-tests for each plant species, to check whether benomyl application had a positive impact on the rates of nitrification, were all significant, (P < 0.05) revealing a significant increase in PNR following benomyl addition as was originally hypothesized.
Means ± SE of soil potential nitrification rates (PNR) in target grassland plants (a) not exposed and (b) exposed to a benomyl treatment (Experiment 2). Plant species codes: Pl: Plantago lanceolata, Pm: Plantago mayor, Pv: Prunella vulgaris, Ra: Rumex acetosa, Rs: Rumex acetosella. Data P values represent the level of significance of the analysis of variance. Treatments means that do not share common letters are significantly different according to the LSD at P = 0.05. The impact of benomyl on PNR was significant in all cases at P < 0.05.
Experiment 3
In agreement with our hypothesis, soil originated from the P. vulgaris treatment, the only mycorrhizal plant species in the experiment, supported lower PNR than soil from the two other tested non-mycorrhizal plants (Fig. 5). Moreover, there was a trend relating N fertilization to decreases in level of AM root colonization (Table 3). Addition of N significantly (P <0.0001) increased PNR. In the absence of N fertilization, PNR were highest for non-mycorrhizal S. vulgaris (P < 0.05) (Fig. 5). N-fertilized samples responded in a similar way to water-treated controls, and PNR of N-fertilized P. vulgaris treatments were the lowest (P < 0.05) amongst the respective N-fertilized plant treatments (Fig. 5).
Means ± SE of soil potential nitrification rates of Prunella vulgaris (Pv), Silene vulgaris (Sv) and Rumex acetosella (Rs) in the non-fertilized-controls (white bars) and N-fertilized grassland soils (grey bars) (Experiment 3). Treatment means that do not share common letters are significantly different at P = 0.05. *Indicates significant (P = 0.05) difference between the two nutrient treatments within each species.
| Plant Species | Manipulation | Arbuscules (%) | Vesicles (%) |
|---|---|---|---|
| Rumex acetosella | Control | 0 (±0)a | 0 (±0)a |
| N-fertilized | 0 (±0)x | 0 (±0)x | |
| Silene vulgaris | Control | 0 (±0)a | 0 (±0)a |
| N-fertilized | 0 (±0)x | 0 (±0)x | |
| Prunella vulgaris | Control | 30.0 (±1.5)b | 25.7 (±1.9)b |
| N-fertilized | 27.4 (±2.9)y | 20.8 (±3.9)y |
Experiment 4
In the two-way analysis of PNR variance, the only significant (at P = 0.05) effect was that of the different blocks (P < 0.05) (Fig. 6). This result was indicative of the complex dynamics that have been driving PNR in this specific field site that necessitated simultaneous correction for soil pH variability before evaluating the effect of other factors (S.D. Veresoglou, R. Sen, A.P. Mamolos & D.S. Veresoglou, unpublished data). Analysis of pH values, on the other hand, revealed that pH was significantly lower in the N-fertilized treatments (P = 0.001) (Fig. 5).
Mean ± SE rates of (a) soil potential nitrification rate (PNR) and (b) pH in the non-fertilized control and the N-fertilized grassland plots under Agrostis capillaris and the forb species at Polypotamous (Experiment 4). Shaded and non-shaded bars represent respective PNR and pH means of soil from A. capillaris and two forbs Prunella vulgaris and Fragaria vesca. Codes are Ac: Agrostis capillaris Forbs: Prunella vulgaris and Fragaria vesca +N: Nitrogen-fertilized plots. Upper summary statistics represent significance of the F test.
In the multiple regression models generated (Table 4), the less mycorrhizal A. capillaris appeared to positively affect PNR, whereas more negative PNR were identified in soils supporting the two mycorrhizal forbs. Model selection criteria revealed that, with the exception of F. vesca, incorporation of plant identity, in the form of a single dummy variable, increased the accuracy of the models (Table 4). The optimal regression model, according to both AIC and BIC, was model 2, which incorporated as a sole dummy variable A. capillaris.
| Model no | Independent variables | P model | R 2 | Intercept sign of dummy variables | AIC | BIC |
|---|---|---|---|---|---|---|
| 1 | N, pH | <0.001 | 0.43 | – | 134.2 | 140.1 |
| 2 | N, pH, Agrostis capillaris | <0.001 | 0.51 | (+) | 131.3 | 138.6 |
| 3 | N, pH, Prunella vulgaris | <0.001 | 0.51 | (−) | 131.7 | 139.0 |
| 4 | N, pH, Fragaria vesca | 0.001 | 0.44 | (−) | 136.1 | 143.4 |
| 5 | N, pH, Agrostis capillaris, Prunella vulgaris | <0.001 | 0.53 | (+, −) | 132.1 | 140.9 |
| 6 | N, pH, Agrostis capillaris, Fragaria vesca | <0.001 | 0.53 | (+, +) | 132.1 | 140.9 |
| 7 | N, pH, Prunella vulgaris, Fragaria vesca | <0.001 | 0.53 | (−, −) | 132.1 | 140.9 |
Discussion
Data from four independent experiments targeting N-limited grassland systems and involving both ex situ and in situ experimentation have highlighted plant species identity as being a significant determinant in shaping soil PNR. PNR appeared to differ significantly between monocultures of different plant species in Experiment 1, in the control (non-benomyl treated) soil samples of Experiment 2 and in both control and N-fertilized soil samples of Experiment 3. Moreover, plant species identity was an indispensable explanatory variable in the models of PNR in Experiment 4. To our knowledge, the only available study to have demonstrated that plant identity may be important in shaping PNR is that of Olsson & Falkengren-Grerup (2000) who, for woody understorey plants, linked preferential 
or 
assimilation to PNR. In this specific study, the plant species tested were grassland species that, typically, grow best in the presence of both 
and 
, as has been demonstrated for P. lanceolata and P. major (Blacquière, Voortman & Stulen 1988). Thereby, it was demonstrated that factors other than 
and 
preference may drive plant-species linked variability in PNR.
To date, there have been two studies that have attempted to address the potential interaction of AM fungi with the AO community. Amora-Lazcano, Vázquez & Azcón (1998) detected an increase of culturable AO most-probable-numbers following inoculation of maize with AM fungi. However, the slow growth rates of AO, as revealed in the sequential harvesting approach, confirmed that the study was performed under non-equilibrium (artefactual) conditions for the AO community. By contrast, Cavagnaro et al. (2007) failed to detect differences between mycorrhiza-defective mutant tomatoes and their wild-type progenitors in either the composition or the density of ammonia-oxidizing bacteria. Cavagnaro et al. (2007) noted that this could have been due to the short (2-month) duration of the experiment, targeting slow-growing AO. No study, so far, has targeted the effect on the ammonia-oxidizing archaeal community that appears to be the key regulator of nitrification in soils (Leininger et al. 2006; Gubry-Rangin, Nicol & Prosser 2010).
This specific study attempted to shed light on potential links between mycorrhizal status of grassland plants and (mycor)rhizosphere PNR. Conservation of plant mycotrophy traits in the plant kingdom suggests that assessment of expected mycotrophy on the crude scale (i.e. non-mycorrhizal, weakly mycotrophic and mycotrophic plants) used here is possible from the literature (e.g. Pawlowska 2005; Wang & Qui 2006). However, because of high dependence of our hypothesis on plant mycorrhizal status, we wanted to, further, ensure that we could support that literature with further evidence from our own data. In Experiment 1, prior assessment of mycorrhizal status of the containers revealed higher mycorrhizal colonization for L. perenne amongst the three grasses and P. lanceolata amongst the two forbs (Karanika et al. 2008b). Mycorrhizal colonization rates were consistent with the literature where P. pratensis has been demonstrated to be one of the few grasses that exhibits negative mycorrhizal dependency, defined as the difference in plant biomass in the presence and absence of AM colonization (Wilson 1997; Wilson & Hartnett 1998). On the other hand, P. lanceolata is a highly mycorrhizal forb that has been used extensively in AM pot experiments investigating N assimilation (e.g. Blaszkowski 1997; Hodge 2003; Roesti et al. 2005; Leigh, Hodge & Fitter 2008; Veresoglou, Shaw & Sen 2011a), whereas R. acetosa, which was originally believed to be non-mycorrhizal (Harley & Harley 1987), is colonized weakly by AM fungi (Sanders & Fitter 1992; Pawlowska 2005). The mycotrophic status of test plants reported in the literature was in agreement with the observations from our experimental site reported in Karanika et al. (2008b). In Experiments 2 and 3, root colonization data verified the results of earlier reports on AM status of R. acetosella and S. vulgaris as being non-mycorrhizal, whereas P. lanceolata, P. vulgaris and P. major are extensively colonized by AM fungi (Pawlowska 2005; Wang & Qui 2006). For Experiment 4, assessment of root mycorrhizal colonization by Karanika et al. (2008b) in the experimental site investigated here revealed that the roots of A. capillaris were colonized more moderately by AM fungi than those of the two forbs. AM-related differences amongst the three plants were supported by data on phospholipids lipid fatty acids (PLFAs) assayed in soils from the Polypotamos experimental site in July 2008 that revealed a significantly lower concentration of the AM-attributed PLFA signature 16:1ω5c in cores taken from stands of A. capillaris than counterparts from P. vulgaris (Veresoglou et al. 2011b and unpublished data).
This specific study is not able to correct PNR data for the potential effect of phylogenetic signal (e.g. Felsenstein 1985). To assay nitrification potential, a procedure that lasts over 26 h is required and this imposes limits to the number of samples that may be processed simultaneously. As a result, we could not experiment with a broader range of plants simultaneously, as required to test for phylogenetic signal and possibly apply a phylogenetic correction. Moreover, there is a very limited range of non-mycorrhizal perennial plants in Mediterranean ecosystems that could be used to generate reliable contrasts. We may not have been able to exclude the likelihood of a phylogenetic effect, but it should be noted that what was positively tested in all experiments was a priori contrasts. Additionally, evidence presented is confirmative for both the two most phylogenetically divergent plant groups: grasses, where L. perenne was the most mycotrophic species, and forbs with mycotrophic Plantago lanceolata, Plantago major and Prunella vulgaris, that represented, mycotrophic plants. All were found to consistently support lower PNR compared to the other less mycotrophic species investigated. Finally, data gathered following benomyl application provide further evidence for the hypothesized relationship.
In Experiment 2, the PNR assay revealed a much higher intra-species variation in nitrification potential in the benomyl-treated soil samples than in those treated with water. Simultaneously, plant-species-specific effects on nitrification potential appeared inconclusive. Higher variability in PNR after benomyl manipulation could be justified in terms of non-homogenous addition of the fungicide. The facts that (i) benomyl addition tended to diminish the effect of plant species identity on PNR and (ii) higher PNR were identified in the benomyl-treated samples are in agreement with the original hypothesis. The fungicide had reduced the extent of plant AM root colonization (Table 2), and, as a consequence, the respective soil densities of AM extraradical mycelial hyphae directly exposed to benomyl in soil solution that were hypothesized to have caused the plant-species-specific differences. However, a plausible alternative scenario could be that as a result of fungicide application, there was a partial lifting of soil N-limitation because of release of fungal-derived mineral and organic N resources from senescing fungal mycelium and susceptible non-target organisms. Benomyl is a N-rich organic compound (c. 19% N), which, through partial mineralization could have increased N availability. Data on total N and P content of plants (Fig. S3) reveal that, following benomyl addition, there was a disproportional increase in N content of plants compared with the P increase that could not be justified by mineralization of fungal hyphae. The critical loss of soil-colonizing extraradical AM mycelium is of particular importance as under such conditions reduced competition for ammonium in the rhizosphere of the normally mycorrhizal plants could allow AO access to 
and thus an increased PNR. To address this question in Experiment 3, half of the pots received N in an inorganic form to lift N-limitation of the soil. Results from Experiment 3 revealed that N additions did not weaken the negative relationship between AM plant mycotrophy and PNR. The amount of N added was similar to the corresponding amount of N that was calculated to have been added following benomyl application in Experiment 2 (c. 173.5 mg per pot N). However, it should be noted that N in Experiment 3 had been applied as a single dose at the beginning of the experiment, which may have complicated comparisons with Experiment 2. Thereby, experimentation with benomyl and N additions, although not conclusive, yielded interesting evidence in support to the original hypotheses. However, taken together with data on plant species mycotrophy, the likelihood of a type I error, with respect to the original hypotheses, appears highly improbable.
In Experiment 4, there was an attempt to establish whether mycorrhizal status of plants could partially explain the variability that has been recorded in PNR under field conditions. Failure of analysis of variance to interpret PNR was in accordance with earlier unpublished work on PNR at this specific experimental site that had highlighted a need for incorporation of the N fertilization status and the pH of the plots to any models that may explain PNR in the specific grassland as chronic N application had led to acidification (S.D. Veresoglou & D.S. Veresoglou, unpublished data). As stated earlier in discussion, multiple regression modelling revealed that the impact of plant species identity was a significant regulator of PNR. More importantly, it highlighted that soil in the rhizosphere of A. capillaris sustained higher PNR than from either of the two mycotrophic forbs. Thus, the result was in agreement with our original hypothesis. It should be noted here that the only plant species whose identity did not improve model accuracy (Table 4) was F. vesca (incorporation into modelling did not improve either AIC or BIC criteria). This could be justified on the basis that F. vesca was harvested exclusively in plots with low pH, where P. vulgaris is excluded because of the low soil pH. Potentially, the linear effect of pH on PNR that was hypothesized in all models could more accurately describe the impact of pH at high pH values. As a result, the P. vulgaris and F. vesca effects on PNR were prone to respective over- and under-estimation.
Strong evidence supporting our allied hypothesis, that soil supporting monocultures of more heavily mycorrhizal plants may sustain lower PNR, has been presented in this study. Moreover, there was some evidence that such a relationship may be detected at micro-scales within field sites, in the (mycor)rhizospheres of mycorrhizal vs. non-mycorrhizal plants. An obvious follow-up question might concern whether the biological interaction between AM fungi and AO was simply a result of competition for the primary substrate 
, or more interestingly, a result of the involvement of allelochemicals, active phenolic acids that are capable of modifying plant or microbial behaviour at low concentrations (Blum 2004). The rhizosphere represents an environment rich in allelochemicals (Bertin, Yang & Weston 2004), and in the past, there have been records of cases where allelochemicals could impact nitrification. The grass, Brachiaria humidicola, has been documented to possess the ability to release inhibitors of nitrification (Subbarao et al. 2006).
AM fungi are believed to acquire N predominantly in the form of 
(Govindarajulu et al. 2005; Tanaka & Yano 2006; Smith & Read 2008). Many plants, however, preferentially assimilate 
(Aanderud & Bledsoe 2008). The model that was proposed by Helgason & Fitter (2009) states that AM fungal carbohydrate nutrition is based on hexoses scavenged from the apoplast while actively transferring phosphate to the plant. The model assumes that a similar mechanism exists for N transfer. In this respect, it may be crucial for the AM fungus to ensure that rates of nitrification in the soil are low as a mechanism to increase the dependency of the plant for N nutrition and ensures higher availability of ammonia in its vicinity that can be exchanged for plant-derived carbon. In the long evolutionary history of AM symbiosis, there has been adequate time (475 million years has elapsed from the evolution of AM fungi (Cairney 2000)) for the evolution of an AM-mediated allelopathic suppression of AO to maintain adequate symbiotic host-fungal 
-N assimilation. More integrated investigation is required to resolve this question.
However, there are two additional alternative scenarios that can be proposed here. First, AM status of plants might have had an impact on the ‘priming effect’ of C additions on N-transforming processes. De Nobili et al. (2001), as well as Ma, Dwyer & Gregorich (1999), have demonstrated that addition of negligible amounts of carbon (C) could lift the C limitation of the microbial community that resulted in a disproportional increase in the rates of N mineralization and possibly nitrification. The main source of C for the microbial community in the rhizosphere is root (and hyphal) exudation. AM colonization of plants results in large changes in the qualitative, quantitative and spatial parameters of exudation. A decline in sugar exudation (Jones, Hodge & Kuzyakov 2004) to the surrounding soil resulting from direct C allocation to AM fungi could, thus, have a detrimental effect on N-transforming processes.
An alternative scenario could be an indirect impact of plant species mycotrophy on nitrification through root structure. Considerable literature to date appears to support the idea of Baylis (1975) that plants with comparatively thick roots coupled to limited lateral or adventitious branching and scarce root hairs tend to have a greater mycorrhizal dependency than plants with more elaborate root systems (e.g. Schweiger, Robson & Barrow 1995; Wilson & Hartnett 1998). Plants conforming to the latter root architecture could support a more vibrant microbial community through a higher availability of root exudates. As discussed above through increased rates of resulting N-mineralization, microbial activity may permit an increase in nitrification potential.
In conclusion, evidence was presented supporting a hypothesis that plant species identity is an important regulator of PNR. Additional information presented here suggests that a key driver of plant species and community dynamics mediation is the biological interaction between AM fungi and AO in N-limited terrestrial ecosystems. The interaction could be detected both in established monocultures of plant species in soil mesocosm systems and under field conditions in natural stands of the three plant species investigated. Further research needs to be directed at elucidating mechanisms through integrated analysis of N-cycling bacterial, archeal and fungal communities in the rhizospheres and mycorrhizospheres on non-mycorrrhizal and mycorrhizal plant species under differing soil N status.
Acknowledgements
The authors would like to thank Mr Vasilis Fabrikis, Mr Kostas Athanasiadis, Mr Panagiotis Skederidis and Miss Olga Voulgari for technical support throughout the study and Dr Georgios Menexes for statistical help on the meta-analytical approaches. We also acknowledge valuable feedback provided by the editors and two anonymous referees during the revision of this manuscript. The project was partially funded, through a PhD fellowship to S.D.V. from the Chloros trust.
