The N-fixing legume Periandra mediterranea constrains the invasion of an exotic grass (Melinis minutiflora P. Beauv) by altering soil N cycling

Melinis minutiflora is an invasive species that threatens the biodiversity of the endemic vegetation of the campo rupestre biome in Brazil, displacing the native vegetation and favouring fire spread. As M. minutiflora invasion has been associated with a high nitrogen (N) demand, we assessed changes in N cycle under four treatments: two treatments with contrasting invasion levels (above and below 50%) and two un-invaded control treatments with native vegetation, in the presence or absence of the leguminous species Periandra mediterranea. This latter species was considered to be the main N source in this site due to its ability to fix N2 in association with Bradyrhizobia species. Soil proteolytic activity was high in treatments with P. mediterranea and in those severely invaded, but not in the first steps of invasion. While ammonium was the N-chemical species dominant in plots with native species, including P.mediterranea, soil nitrate prevailed only in fully invaded plots due to the stimulation of the nitrifying bacterial (AOB) and archaeal (AOA) populations carrying the amoA gene. However, in the presence of P. mediterranea, either in the beginning of the invasion or in uninvaded plots, we observed an inhibition of the nitrifying microbial populations and nitrate formation, suggesting that this is a biotic resistance strategy elicited by P. mediterranea to compete with M. minutiflora. Therefore, the inhibition of proteolytic activity and the nitrification process were the strategies elicited by P.mediterranea to constrain M.munitiflora invasion.

Leaf C and N concentrations. The leaf C and N concentrations ( Table 1) of P. mediterranea (T2) were significantly higher than those of the other native (T1) and invasive (T3 and T4) species. Consequently, the C:N ratio in P. mediterranea leaves (T2) was the lowest, followed by that of the native species in T1, contrasting with the highest values found in M. minutiflora leaves (T3 and T4). The leaf natural 15 N abundance (‰) of P. mediterranea showed a greater isotopic depletion than that of leaves of native and invasive species, suggesting a different nitrogen source among plants (Fig. 3).
Identification of rhizobial endosymbionts of P. mediterranea. The slow-growing isolated strains from P. mediterranea nodules formed mucoid colonies typical of rhizobia. The 16S rRNA gene sequences of the isolated strains were compared with those kept in EzBiocloud, a database containing the sequences of the 16S rRNA genes of the type strains of all bacterial species, and the results showed that they belonged to the genus Bradyrhizobium. Phylogenetic analysis of their 16S rRNA gene sequences showed that they clustered into three different phylogenetic groups within this genus, which also contained other species isolated in Brazil (Fig. 4). This is therefore the first report of the identity and diversity of Bradyrhizobium strains inhabiting nodules of P. mediterranea.
Soil N chemical species. Although the total C and N contents did not differ among treatments (Table 2), the soil C:N ratio was significantly lower in T2 plots, where P. mediterranea was the dominant plant species. Soil NH 4 + did not differ between the treatment with native plants (T1) and T2 (p = 0.13); T3 (p = 1,0) and T4 (p = 1.0). However, the NH 4 + content was reduced in T3 in relation to T2 (p = 0.01) at the beginning of the invasion (Fig. 5A), but there was no difference between T2 and T4 at advanced invasion (p = 0.9). Indeed, at a higher invasion degree (T4), the NH 4 + content was restored to similar levels as those found before the invasion (T1 and T2). In contrast, soil NO 3 − was significantly increased in T4 (Fig. 5B) when invasion by M. minutiflora was highest ( Fig. 1G), differing from T1 (p < 0.001), T2 (p = 0.02) and T3 (p < 0.001) (Fig. 5B). Nevertheless, such increase in nitrate did not occur at the initial invasion process (T3), and nitrate levels were similar to both T1 (p = 1.0) and T 2 (p = 1.0). Similarly, the presence of P. mediterranea in T2 did not facilitate nitrate formation when compared to the plots only containing other native species (T1) (p = 0.54). www.nature.com/scientificreports www.nature.com/scientificreports/ Soil enzyme activity and microbial nitrifying communities. Soil enzyme activity related to N metabolism ( Fig. 6) showed that protease activity was higher in the presence of the legume P. mediterranea (T2) than in the plots without this species T1 (p < 0.001), but was inhibited in the beginning of the invasion process (T3), which differed from T2 (p = 0.012). In the first stage of the invasion (T3), enzyme activity became similar to that in T1 (p = 0.1) plots without P. mediterranea. However, in fully invaded plots (T4), we observed a significant increase of soil protease activity in relation to T1 (p < 0.001), but not in relation to T2 (p = 1.0) and T3 (p = 0.19).  www.nature.com/scientificreports www.nature.com/scientificreports/  www.nature.com/scientificreports www.nature.com/scientificreports/ The abundance of functional genes of microbial communities that mediate the first step of the nitrification process, oxidation of NH 4 + to NH 2 OH, which is subsequently oxidized to nitrite and nitrate, was estimated by quantifying the amoA gene carried by bacteria (AOB) and archaea (AOA) communities ( Fig. 7A,B). In soils containing native plants (T1), we detected a low abundance of bacterial (AOB) amoA gene communities (Fig. 7A), which was increased in the presence of P. mediterranea, either in T2 (p = 0.0018) or T3 (p < 0.001). In contrast, in the initial stages of the invasion process (T3), the AOB communities were not modified by M. minutiflora in comparison to T2 (p = 0.14), but they were significantly stimulated with the dominance of the invasive plant in T4 in relation to T1 (p < 0.001), T2 (p < 0.001) and T3 (p = 0.002) (Fig. 7A). Regarding the AOA nitrifying communities ( Fig. 7B), there was an inhibition of these microbial communities in the treatment with P. mediterranea (T2) in comparison to the native species (T1) (p = 0.001), but they were progressively stimulated by the invasive species (T3), with a significant difference to T2 (p = 0.001), but not to T1 (p = 1.0). The highest abundance of AOA was registered in T4 as compared toT1 (p < 0.001), T2 (p < 0.001) and T3 (p < 0.001).

Discussion
The occupation data show a displacement of native species, particularly those belonging to the Poaceae family, which were replaced by M. minutiflora, also a Poaceae species. Such replacement of native plants by the invader might be due to competition for nutrients such as nitrate, as revealed by our results and confirmed by the literature, which demonstrates that M. minutiflora displays a high demand for nitrogen 3,4.5 . In contrast, P. mediterranea suffered a less significant displacement than the other plants in T3, possibly due to its independence for soil nitrogen uptake associated with its nitrogen fixing ability.
Our results demonstrate that the main N source in the studied site is represented by biological nitrogen fixation via P. mediterranea, which showed the highest total leaf N concentration, the lowest leaf C:N ratio and, most importantly, a depletion in leaf isotopic 15 N relative to other reference species. Such depletion is attributed to the isotopic atmospheric discrimination promoted by N-fixing bacteria associated with leguminous plants 23 . In this study, P. mediterranea was found nodulated by three species of the symbiotic nitrogen-fixing genus Bradyrhizobium, some of them clustering with species isolated from nodules of other Brazilian legumes, such as Bradyrhizobium ingae (Inga laurina), Bradyrhizobium tropiciagri (Neonotonia wightii), Bradyrhizobium macuxiense (Centrolobium paraense) and Bradyrhizobium brasilense (Vigna unguiculata). Besides, a novel bacterial species was found inhabiting the nodules of P. mediterranea and belonging to the genus Paenibacillus (P. www.nature.com/scientificreports www.nature.com/scientificreports/ periandrae) 24 . This bacterial species might also play a role on nitrogen fixation, since the genus Paenibacillus contains several nitrogen-fixing species 25 .
The N fixation occurring in P.mediterranea nodules explains the elevated N content (2.0 g/kg) in plots with this species, which is similar to the values (2.0 g/Kg) found in tropical soils in Brazil, such as in the Atlantic 26 and the Amazon 27 . The C:N ratio is an important tool to assess soil fertility in both agricultural and natural ecosystems 28 . Although total soil N and C did not differ among treatments, the C:N ratio was lowest in T2. Therefore, the presence of P. mediterranea seems to ensure N soil fertility in study area, which may be considered the main N source. Since soil N becomes available via mineralisation, we infer that not only native species use it, but that it can be opportunistically shared with invasive species, supporting the "Fluctuating Resource Hypothesis" for plant invasion 29 . Although total soil N was not altered by plant invasion, P. mediterranea stimulated the proteolytic activity in the soil, especially in T2 (P.mediterranea + native species) in relation to T1 (native species). Similar to T1, the proteolytic activity in T3 (P. mediterranea + native species + invasion < 50%) was limited in relation to T2, suggesting the establishment of a competitive relationship between the leguminous and the invasive species to resist the invasion.
Although the molecular dialogue among co-existing plants involving root-root interactions is not fully understood, root exudates may affect all the components of the rhizobiome 8,9 , allowing the defense of native species by establishing a competition against invaders 11 . Thus, we may infer that the competition established by P.mediterranea with the invader may likely be processed by root-root interactions via molecular signalling, able to detect the presence of the invader and illicit a defense response.
However, such competition seems to has been overcome by the invasive species by restoring proteolytic activity when it reached the maximum invasion in T4, possibly also via root exudation, stimulating the proteolitic activity. Since soil protease may be derived either from microbial or plant exudation 30 , the increase in this enzyme in both P. mediterranea and in fully invaded M. minutiflora plots (T2 and T4) may also be explained by the capacity of these species to act directly via protease exudation or indirectly through the stimulation of decomposing microbial populations, thereby increasing N release 13 . In previous studies, soil protease activity was increased under both leguminous species 31 and invasive grass species 32 . Similar to our results, these latter authors found a significant increase in protease activity in the rhizosphere of the invasive grass Pennisetum setaceum in relation to native species. On the other hand, despite the presence of leguminous species, soil protease was inhibited in T3, which can be explained by the root exudation of protease inhibitors such as flavonoids, plant hormones and siderophores by P. mediterreanea or microorganisms 30 , constituting a possible allelopathic defence mechanism at the beginning of invasion.
Proteolysis is a slow and limiting process of the N cycling and represents the natural supply of soil low-molecular-weight monomers (e.g., amino-acids), facilitating mineralisation 33 and making the inorganic N forms (NH 4 + and NO 3 − ) available to plants. Such availability depends on abiotic factors such as soil pH and on biotic factors such as plant species traits. Ammonium is the dominant soil inorganic N species in most mature undisturbed forests and acidic soils [34][35][36] , while nitrate is more available in agricultural soils with higher pH levels 34,35 . As under most native vegetation, the dominant chemical N form is NH 4 + , it is likely to be the target of competition between plants and microorganisms 36 . Furthermore, plant uptake of inorganic N also depends on the expression of different protein transporters, which are modulated by the soil ammonium/nitrate ratios 34 determining the N chemical preference by plants in their native habitats 35 .
Considering that the campo rupestre biome is a preserved site with acidic soils and elevated NH 4 + content in relation to NO 3 − , as found in plots with native species (T1, T2 and T3), the former appears to be the preferred N chemical species naturally used by this vegetation, as also observed by Ribeiro et al. 3 . Similar to the protease activity, NH 4 + formation was inhibited in plots weakly invaded by M. minutiflora (T3), reinforcing the idea of a competitive interaction between native and invasive species for nutritional resources. Such reduction occurred even in the presence of the N-fixing species P. mediterranea. High NH 4 + levels were also found in massively invaded plots (T4), suggesting that both M. minutiflora and native species may use NH 4 + as N source. Indeed, Incerti et al. 13 observed a greater N release from litter of invasive species as compared to natives, confirming the high dependence of invasive species on N. Thus, the competitive equilibrium for NH 4 + observed in T3 was disrupted when the invader stimulated the oxidation of NH 4 + to NO 3 − (T4), which was found elevated only in the fully invaded plots, suggesting that it was the main N form used by the invasive plants; this has also been observed by Ribeiro et al. 3 . However, in the presence of the native leguminous species M. pogocephala 3 , nitrification was stimulated in the first stage of the invasion process. In contrast, in the present study, we found low levels of NO 3 at the beginning of the invasion of P. mediterranea plots (T3), confirming the competitive state for N established between native and invasive plants. Such constraint in soil N metabolism elicited by P. mediterranea in T3 suggests a possible mechanism of biotic resistance to control exotic species invasiveness, as shown in Fig. 1E,F.
Nitrate formation in the soil is mediated by aerobic ammonia-oxidising bacteria (AOB) 37 or by ammonia-oxidising archaea (AOA) carrying amoA genes 38 . In the campo rupestre, an increase in both bacterial and archaeal populations (AOB and AOA) was found, as expressed by the amoA gene abundance in T4, contrasting with lower levels in plots with native species (T1, T2 and T3). These results showed a high reproducibility in relation to those obtained by Ribeiro et al. 3 , who, using the same experimental design in a site separated from the present study site, both spatially (almost 1 Km) and temporally (2 years), also demonstrated that the increase in soil nitrification is the main competitive strategy of M. minutiflora to invade the campo rupestre biome. Literature findings also confirm a positive feedback between different invasive plants and NO 3 − availability, mediated by both AOA and AOB nitrifying microbial populations 3,[15][16][17] . In addition, Rodríguez-Caballero et al. 32 demonstrated that the nitrifier genus Ohtaekwangia was also stimulated in the rhizosphere of the invasive species Pennisetum setaceum and Bardon et al. 39 reported a significant increase of nitrate and nitrification enzyme activity in a site invaded by Pteridium aquilinum. www.nature.com/scientificreports www.nature.com/scientificreports/ However, at the initial steps of invasion (T3), the AOB population was not stimulated and remained similar to that of non-invaded plots (T2), suggesting again a competition between P. mediterranea and the invading species involved in the control of the AOB population. This idea is reinforced by the results found by Ribeiro et al. 3 who reported that M. minutifora also strongly stimulated the AOB population, although this occurred at the very beginning of the invasion, when the dominant leguminous species was Mimosa pogocephala.
In contrast to the AOB population, the AOA population was smaller in the P. mediterranea plots (T2) in comparison to those plots without this species (T1) as well as at the beginning of the invasion (T3). Such a high AOA abundance in T1, T3 and T4 in relation to T2 may be attributed to the effect of grassy species, since AOA is particularly favoured in grasslands 40 . Therefore, the increase in AOA abundance at the first steps of this grass invasion process seems to be a key factor favouring M. minutiflora to overcome the competition against P. mediterranea and successfully invade the site. The inhibition of AOA population in presence of leguminous species has been also reported by Paungfoo-Lonhienne et al. 41 who found that AOA population was suppressed by legumes in a sugarcane cropping soil.
At the beginning of the invasion of campo rupestre by M. minutiflora, when this plant still co-existed with P. mediterranea (T3), soil nitrogen metabolism was limited, as observed for proteolytic activity, ammonium, and nitrate production as well as nitrifying AOB, but not in terms of AOA populations. Such inhibition in N metabolism was observed in the presence of P. mediterranea, but not when M. pogocephala was the dominant species 3 , suggesting that the former species elicited such an "N metabolic silence" as a strategy to constrain the expansion of the invader. Such strategy may be related to the independence of P. mediterranea in relation to soil nitrogen due to its N-fixing ability as well as the control effect over the AOB population and the inhibition of soil protease activity. However, this strategy seems to have been weakened by the stimulation of the AOA community by the grassy invasive plant, enabling it to oxidise ammonium to nitrate, which is the N species found only beneath M. minutiflora. Such a positive feedback between invader and soil nitrate allowed this species to successfully invade the campo rupestre, as observed in T4.
Therefore, the inhibition of some steps of the N cycle, elicited by P. mediterranea, seems to be a strategy of biotic resistance. Traditionally, biotic resistance is considered as the ability of resident species to constrain the invasive success of exotic species 42 through different mechanisms such as an increase in competition, predation, disease and parasites as well as a decrease in nutrient availability 43 , including N 44 . Considering that the competition for nutritional resources is a major invasive strategy used by exotic species, as proposed by the "Fluctuating Resource Hypothesis" 29 , the unavailability of some soil N chemical species as nitrate, elicited by native plants, might be one of the strategies to retard the invasion, as shown in our study. Thus, at the beginning of the invasion process, P. mediterranea limited the availability of all soil N species, but particularly constrained nitrate formation, which possibly constitutes one of the engines of the biotic resistance against M. minutiflora invasion in the campo rupestre biome. The mechanism driving the biotic resistance has been attributed to the production of plant secondary metabolites as methyl 3-(4 hydrophenyl) propionato or cyclic diterpene able to inhibit specific nitrifies groups 45 by declining the abundance of both AOA and AOB 46 . Such decrease in ammonia oxidizers population promoted by plants is a conservative strategy to preserve the NH4 +47 a phenomenon so-called as biological nitrification inhibition (BNI) as suggested by Subbarao et al. 46 .
In conclusion, this study highlighted the ability of M. minutiflora to displace the native species in the campo rupestre, particularly those from the Poaceae family, while the leguminous P. mediterranea showed some resistance. P. mediterranea resistance was attributed not only to its N-soil independence due to biological nitrogen fixation in association with Bradyrhizobium species, but also to its ability to compete with M. minutiflora at the beginning of the invasion specially via biological nitrification inhibition (BNI). At this stage, there was a constraint of proteolytic activity, ammonium and nitrate formation as well as AOB abundance. However, the invasive species was able to overcome the competition against the native species by stimulating the AOA nitrifying population. Therefore, selection of species that display biotic resistance via nitrification inhibition (BNI) might be a useful strategy to restore M. minutiflora invaded lands. Regarding specifically the invaded campo rupestre sites by M.minutiflora, we highly recommend planting P.mediterranea in future rehabilitation efforts.

Methods
Study area. The study was developed at the PESRM (3924 ha), which is located in the central-southern region of the state Minas Gerais, Brazil (20°3′20″S,44°1′11″W). The park is inserted in the Espinhaço mountain region, which has a giant hematite-rich iron deposit, largely exploited by iron ore mining companies. The dominant campo rupestre vegetation is composed of grassy, herbaceous and shrub species of the families the Poaceae, Asteraceae, Fabaceae, Myrtaceae, Melastomataceae and Orchidaceae 2 . Leguminous species occur frequently, particularly of the genera Periandra, Chamaecrista, and Mimosa 2 . This vegetation grows on shallow soils formed on ironstone outcrops (canga) originated from rock dissolutions, which are also rich in soil organic matter and nitrogen 3 .
Experimental design. The experimental site was established in an area of 600 m 2 or 0.06 ha of a grassy field (20°03′22″S,44°01′09″W). The experimental design consisted of three blocks with four treatments (plots), with three replicates/treatment/block (3 × 4 × 3 = 36), resulting in a total of 36 plots of 4 m² each. Each studied plot (4 m 2 ) was circumscribed by a 1-m strip with the same treatment, defining a large plot (3 × 3 m) to control the edge effects. Each plot (4 m 2 ) represented one treatment. Three soil sub-samples/ treatment/ block were collected, resulting in a total of nine samples/treatment.
In the study site (Fig. 1), four treatments were established: T1-Native vegetation without the legume species (P. mediterranea) (Fig. 1B) and no invasion; T2 -Native vegetation plus P. mediterranea and no invasion. (Fig. 1C,D); T3 -Native vegetation plus P. mediterranea and invasion by M. minutiflora ≤50% (Fig. 1E,F); T4 -Native vegetation plus P. mediterranea and invasion by M. minutiflora >50% (Fig. 1G,H) www.nature.com/scientificreports www.nature.com/scientificreports/ species occupation and occurrence were measured using the method proposed by Toledo and Schultze-Kraft 48 . This method uses a quadrant of 1 m², composed of 100 squares of 0.01 m² each. The quadrant was placed inside the plots and all plant species inside each square were identified.
Soil chemical analysis and soil protease activity. Each soil sample consisted of three mixed subsamples per treatment/block or nine mixed samples per treatment. Total inorganic nitrogen was determined via semi-micro-Kjeldahl digestion 49 , and N ammonia and nitrate contents were determined 50 . Soil protease activity was determined by the method described by Alef and Nannipieri 51 and expressed as mg tyrosine g −1 h −1 .

Plant leaf analysis.
Leaf samples were taken in the field, using three replicates from three individuals of each family or species (Poaceae families; native species belonging to Malpiguiaceae, Asteraceae, Melastomataceae, Euphorbiaceae and Cactaceae families; P. mediterranea and M. minutiflora) per treatment/block and dried at 70 °C for at least 48 h. After reaching constant weight, the samples were ground to a fine powder in a ball mill (Glen Creston Ltd.) and subsequently analysed for total carbon, nitrogen, and δ 15 N, using an isotope ratio mass spectrometer (Finnigan MAT Delta E, Thermo Electron, Bremen, Germany) coupled to an EA 1110 elemental analyser (Thermo Electron, Milan, Italy). The abundance of 15 N was expressed in "delta" notation (δ), which is the deviation per thousand (‰) of 15 N abundance of the sample in relation to the international standard (i.e. atmospheric N 2 , which has an atom% of 15 N = 0.3663 52 : sample atom% N 0 3663 0 366 1000 15 15 Identification of rhizobial endosymbionts of P. mediterranea. The isolation of rhizobia was carried out using P. mediterranea effective nodules (internal pink colour) on YMA medium according to Vincent 53 . Identification was performed through the analysis of the 16S rRNA genes, which were amplified 54 . Nearly complete sequences (more than 1,400 bp) were obtained by the Macrogen Corporation (Amsterdam, The Netherlands). The sequences obtained were compared to those stored in the EzBiocloud server 55 . For the phylogenetic analysis, the sequences were aligned by using the Clustal W software. Distances were calculated according to Kimura's two-parameter model 56 and used to infer phylogenetic trees with the neighbour-joining method and the MEGA5 software 57 . Confidence values for nodes in the trees were generated by bootstrap analysis using 1,000 permutations of the data sets.
AOA and AOB abundances (qPCR). The abundances of AOB and AOA communities were determined using real-time PCR (qPCR) to estimate the soil bacteria expressing the ammonia monooxygenase gene (amoA). The DNA was extracted from three mixed soil samples/treatment (0.5 g), using the MoBio Power Soil extraction kit (Carlsbad, CA) and following the recommended protocol. The DNA concentration and purity measurements were performed using a Nanodrop spectrophotometer (Nano Drop Technologies, Wilmington, DE, USA), and the PCR reactions were carried out in an Applied Biosystems Quant Studio Real-Time PCR system (Applied Biosystems, Foster City, CA) with iQSYBR-Green Supermix from Bio-Rad (Hercules, CA). Bacterial amoA gene quantification was performed using the primers amoA-1F (5′-GGGGTTTCTACTGGTGGT-3′) and amoA-2R (5′-CCTCGGGAAAGCCTTCTTC-3′) primers 58 . The thermocycling conditions were as follows: 30 s at 95 °C, followed by 40 cycles of 30 s at 95 °C and 34 s at 60 °C, 60 s at72 °C and Crenarchaeal amoA CrenamoA23fd: (5′-ATGGTCTGGCTWAGACG-3′) and CrenamoA616rd: (5′-GCCATC CATCTGTATGTCCA-3′) 59 using 95 °C for 5 min; followed by 10 cycles of 94 °C for 30 s, 55 °C for 30 s, 72 °C for 1 min; followed by 25 cycles of 92 °C for 30 s, 55 °C for 30 s, 72 °C or 1 min; followed by 72 °C at 10 min. The calibration curve was obtained by using a 10-fold serial dilution of a known concentration of positive control DNA (ranging from 10 2 to 10 8 ). The C T values obtained from each sample were then compared with the standard curve to determine the original sample DNA concentration.
Data analysis. All data were submitted to analysis of variance (ANOVA) for normally distributed variables and Kruskal-Wallis test for non-normally distributed variables (at the 5% probability level). Multiple comparisons among treatments of normally distributed variables were performed using the Tukey test, while for not normally distributed variables, we used the Nemenyi test. All analyses were conducted using the statistics software "R" (v.3.2.2).