The Abundance of Microbial Functional Genes in Grassy Woodlands Is Influenced More by Soil Nutrient Enrichment than by Recent Weed Invasion or Livestock Exclusion

Abstract

A diverse soil microbial community is involved in nitrogen cycling, and these microbes can be affected by land management practices and weed invasion. We surveyed 20 woodlands with a history of livestock grazing, with livestock recently excluded from 10 sites. We investigated whether soil nutrients were lower when grazing was excluded and higher when exotic grasses dominated the understory. Second, using quantitative real-time PCR, we investigated whether microbial nitrogen functional gene (NFG) abundance was altered with soil nutrient enrichment, livestock exclusion, and exotic grass invasion. The target genes were chiA (decomposition-ammonification), nifH (nitrogen fixation), nirK and narG (denitrification), and bacterial amoA (nitrification). Woodland soils were enriched in phosphorus and nitrogen compared to reference condition sites, but soil nutrients were not lower following livestock exclusion. Total nitrogen and nifH were negatively correlated in grazed woodlands, suggesting that aboveground herbivory reduces the capacity for belowground nitrogen fixation. Woodlands dominated by exotic grasses had higher levels of nitrate, narG, and nirK than those dominated by native grasses. We hypothesize that the increase in potential for denitrification was due to increases in soil nitrate, rather than changes in plant composition. Overall, soil physicochemistry explained more variation in NFG abundance than livestock presence or plant invasion, particularly for chiA and bacterial amoA, with significant relationships between the abundance of all five NFGs and total nitrogen or nitrate. All woodlands investigated had a history of anthropogenic disturbance and nutrification, and soil nutrient levels and the abundance of NFGs are likely to be related to long-term land management practices.

Microbes, together with plants, are major drivers of the soil nitrogen cycle. In soils, the rates of organic matter decomposition, nitrogen fixation, nitrification, and denitrification are largely mediated by the microbiota. Many microbially driven processes in soils can be impacted by land management practices and changes in the plant community (12, 20, 50), such that alterations in microbial community composition or in the abundance or activity of specific groups can alter nitrogen availability to plants or nitrogen loss from the ecosystem (60).

Grassy eucalypt woodlands were once common in southeastern Australia. However, very little of the original vegetation cover remains due to clearing land for pasture and crops (38). As a result, most remaining woodlands exist as patches within highly modified agricultural landscapes. Woodlands are structurally and functionally distinct from the surrounding agricultural land; however, their close proximity means they are vulnerable to the effects of agricultural management. Furthermore, woodland remnants are often managed in a manner similar to that of the land they border. As such, most have been regularly grazed by livestock over the past 100 to 150 years (2).

Livestock grazing in grassy woodlands could potentially alter the nitrogen cycle. Livestock redistribute and concentrate nutrients and increase the rate of nutrient exchange between the soil, plants, and animals (57), and grazing has been found to enhance the activity of soil nitrifying and denitrifying bacteria (33). The exclusion of livestock grazing from woodlands in agricultural landscapes has become a common management tool for conservation and restoration of remnant native vegetation, with benefits for vegetation communities (45), invertebrate abundance and diversity, and leaf litter decomposition rates (24). Therefore, it is possible that grazing exclusion could lead to a reduction in soil nutrients and a change in the soil microbial composition and abundance.

Grassy eucalypt woodlands were originally located on low-nutrient soils (37); however, nitrogen and phosphorus can be transported into woodlands from fertilized pasture and crop lands by livestock, wind, and runoff events (8, 39). Soil nitrogen and phosphorous concentrations can reach levels well above those required by the native vegetation (25, 40), and soil enrichment combined with soil disturbance from livestock can promote invasion by cool-season exotic annual grasses (e.g., Lolium and Bromus spp.) and herbs (e.g., Echium spp.) (25, 39).

Plant invasions can alter ecosystem processes and function by changing aboveground-belowground interactions (9, 11, 26). This is more likely to occur if the invasive plant has different ecophysical properties than the native vegetation it replaces (56, 60), as is the case with exotic grasses that invade eucalypt woodlands. Exotic grasses are annual, not perennial, and have lower C and N content and a higher specific leaf area (28) than the native grasses. In addition, various plants have been shown to exert specific selection effects on the rhizosphere microbial community (44). Therefore, there is considerable potential for alterations in the nitrogen cycle through impacts on the soil microbiota following grass invasion of woodland remnants.

We examined the effects of weed invasion and grazing management on the microbial ecology of the nitrogen cycle through the quantification of five metabolic genes known to code for enzymes involved in bacterial nitrogen cycling in soil (Table 1 and Fig. 1). These enzymes are involved in four major groups of processes of the nitrogen cycle: decomposition-ammonification, nitrification, denitrification, and nitrogen fixation (Table 1).

TABLE 1.

Summary of the genes investigated, the enzymes they encode, and their function in the nitrogen cycle

Gene	Enzyme	Process(es)	Details
chiA	Chitinase	Decomposition-ammonification	Hydrolysis of N-acetylglucosamine (conversion of chitin to chitobiose and liberation of carbon and nitrogen)
nifH	Dinitrogen reductase	Nitrogen fixation	Reduction of N2 to NH4+
amoA	α subunit of ammonia monooxygenase	Nitrification	Oxidation of NH4+ to NH2OH (ultimately leads to production of NO3− by nitrifiers)
nirK	Nitrite reductase	Denitrification	Reduction of NO2− to NO
narG	Nitrate reductase	Denitrification	Reduction of NO3− to NO2−

FIG. 1.

Simplified diagram of part of the nitrogen cycle and the nitrogen functional genes involved. The soil factors (carbon, nitrogen, nitrate, phosphorus, and pH) that were associated with an increase (↑) or decrease (↓) in the abundance of each gene are noted above the large white arrows.

The nitrogen functional gene (NFG) approach has been used successfully to investigate the effects of land (6, 48, 54) and water management (49) on parts of the nitrogen cycle. The NFG approach is process specific and can indicate the biological capacity for these functions to occur, with NFG abundance related to process rates and substrate availability and microbial population density in some environments (6, 32, 50, 55).

The major aims of this study were to determine the relationships between nitrogen functional gene abundance and soil nutrients with respect to livestock grazing exclusion and invasion by exotic annual grasses in woodland remnants. We hypothesized that NFG abundance is affected by the nutrient status of soils which, in turn, is affected by the extent of exotic plant invasion and management practices, such as livestock grazing.

MATERIALS AND METHODS

Field sites.

Twenty grassy woodland sites were chosen in the southwestern slopes and southern tablelands of New South Wales, Australia. The sites were located on privately owned land near three towns in Australia, with six sites within 40 km of Boorowa (−34.433°N, 148.733°E), seven sites within 20 km of Murrumbateman (−34.966°N, 149.033°E), and seven sites within 40 km of Bungendore (−35.250N, 149.433°E). The woodland overstory was dominated by one to four species of Eucalyptus, the most common being Eucalyptus melliodora, E. blakelyi, E. goniocalyx, E. nortonii, E. pauciflora, and E. albens. The understory was dominated by native perennial grasses and/or exotic annual grasses.

We established a 50- by 50-m plot in each site, situated so that at least two sides of the plot were on the most productive field-woodland boundary. The fields were predominately managed as pastures consisting of native grasses, weeds and/or improved pasture (deliberately sown with palatable exotic grass species). For more details of the study sites, see Lindsay and Cunningham (24).

All sites had a long history of cattle and sheep grazing (>100 years). Ten of the sites were still grazed by cattle and/or sheep, with no fence separating the remnant woodland from adjacent pasture. The other 10 sites had livestock grazing excluded for an average of 7 years (6 to 25 years) at the commencement of the study. All sites still had low levels of grazing by rabbits (Oryctolagus cuniculus) and kangaroos (Macropus giganteus).

Plant surveys.

Vegetation surveys were conducted in spring (October and November) using a point transect approach. Surveys were performed along four 50-m-long transects spaced 10 m apart, parallel to the soil sample transects. The plant specimen that intersected each transect at 1-m intervals was identified. Tree canopy cover was determined every 5 m. Plant samples were identified to species where possible, with nomenclature following Harden (13, 13a, 13b, 13c). A posthoc grouping of sites was performed based on the dominant grass type in the woodland understory. The three groups were annual exotic-dominated sites, native perennial-dominated sites, or sites with an even mix of native and exotic plants. Annual exotic-dominated sites had >66% annual exotic grass cover in the understory, mixed sites had 33 to 66% annual exotic grass cover, and native-dominated sites had <33% annual exotic grass cover. There were three mixed grass and perennial grass sites in each of the grazed and ungrazed categories and four annual grass sites in each of the grazed and ungrazed categories. The most common annual exotic grasses were barley grass (Hordeum leporinum), soft brome (Bromus molliformis), rye grass (Lolium rigidum), giant brome (Bromus diandrus), and rat's tail fescue (Vulpia bromoides).

Soil collection.

Soil samples were collected in May 2006 (autumn) along three transects, 12.5 m apart, running down the slope (from high to low) 5, 15, 25, 35 and 45 m from the woodland edge. Soil samples were collected with a “push in” metal soil collector (diameter, 4.7 cm; depth, 0 to 5 cm). Samples from the same position on each of the three transects were combined together to give five samples per site. Soil was transported in an insulated container. Bulk density was determined with a metal sampling device (4.5 cm wide by 10 cm deep). Five samples were taken per site across all three transects. This was done in a zigzag method (i.e., position 1 transect 1, position 2 transect 2, etc.).

Quantification of nitrogen functional genes.

We quantified a suite of microbial genes associated with N cycling in soil samples from 15 of the 20 woodland sites (15 × 5 samples). Genes were selected for quantification based on providing broad coverage of the major N transformation steps (Table 1). When multiple functional genes are known for a single N transformation (e.g., narG and napA are both functional markers for nitrate reduction), a single indicator gene was selected to represent the process (Table 1). In these cases, gene selection was based on previous results where the genes have provided informative responses of soil N cycling communities to aboveground management practices (6, 50). For decomposition-ammonification, where a multitude of pathways exists, the bacterial chitinase gene chiA was chosen, as chitin is a major source of carbon and nitrogen for bacteria and other organisms in soil ecosystems (21). These group A bacterial (including actinobacteria) chitinase genes were targeted, as they possess well-conserved regions (47) allowing for specific targeting across broad phylogenetic range and can account for 70 to 80% of chitin degradation in soils (58).

DNA was extracted from 0.2-g subsamples of soil using the Powersoil DNA kit (MoBio Laboratories Inc., Carlsbad, CA). The 0.2-g sample was taken from the same combined and homogenized soil samples as used for the soil nutrient analysis. A mechanical disruptor (speed setting of 4.5; 30 s) (FastPrep Bio 101; MP Biomedicals, Solon, OH) was used to enhance extraction efficiency. DNA was quantified fluorometrically (FLUOstar; BMG Labtechnologies Inc., Durham, NC) using Picogreen (Molecular Probes Inc., Eugene, OR), and final extracts were eluted into 50 μl of sterile distilled deionized water and stored at −80°C.

All reactions were conducted on an ABI Prism 700 sequence detection system (Applied Biosystems Australia, Mulgrave, Victoria, Australia) in triplicate. PCR of narG was based on the methods of Philippot et al. (35), PCR of nifH was based on the methods of Rösch et al. (42), and PCR of betaproteobacterium amoA was based on the methods of Stephen et al. (46). PCR conditions and PCR chemistry were as described by Wakelin et al. (49). The quantification of the chiA gene was based on the GA1F and GA1R primer set (58). Primers were added to give 0.5 μM in the master mix. Following hot-start activation of Taq polymerase (95°C for 15 min), 40 cycles, with 1 cycle consisting of 95°C for 20 s, 57°C for 30 s, and 72°C for 30 s, were used. Quantification of nirK was based on the nirK-1F and nirK-5R primer set (4), using 0.5 μM in the master mix. Thermocycle conditions were also based on touchdown PCR; in the first 10 cycles, the annealing temperature was reduced from 50°C to 45°C, after which it was maintained at 45°C for a further 30 cycles. Denaturation was achieved at 95°C for 20 s, primer annealing was achieved at 50°C or 45°C for 30 s, and extension was achieved at 72°C for 30 s. Gene copy numbers were derived from standard curves generated by serially diluting genomic DNA extracted from pure bacterial reference cultures from 10⁷ to 10³ copies per reaction mixture. Calculations for standard curves required estimates of genome size and gene copy number per genome. Genome size estimates (in megabases) for bacterial reference taxa were as follows: 4.6 for Serratia marcescens (chiA standard), 6.8 for Rhizobium leguminosarum (nifH) and Rhizobium trifolii (nirK) (17), 2.8 for Nitrosomonas europaea (amoA), and 4.2 for Escherichia coli (narG) (3). Two copies of amoA are present in N. europaea (19), and calculations for the other taxa were based on one copy of each gene of interest. A melting curve analysis was performed at the end of each quantitative real-time PCR to ensure that a single product was generated. Reactions were also run on a 1.3% agarose gel to confirm specificity. The mean number of copies (n = 3) per mg soil was calculated for each sample, and a dry weight correction was used, as DNA quantification was performed on moist soil.

Soil physicochemistry.

Bulk density was determined by the method of Lindsay and French (26) and calculated based on dry weight in grams per cubic centimeter. Soil for nutrient analysis was dried at 40°C for 48 to 72 h after collection and crushed to pass though a 2-mm sieve. Green plant material and rocks were removed during sieving. Ammonium and nitrate plus nitrite levels were determined by method 7C2, plant-available phosphorous level was determined by method 9B2, and soil pH was determined by method 4B1 of Rayment and Higginson (41). Total soil carbon and total nitrogen analysis was performed by the Environmental Analysis Laboratory, Lismore, Australia.

REML.

We used residual maximum likelihood analysis (REML) in GenStat version 11.3 (34) to investigate whether soil nutrient levels varied with grazing and grass type and whether there was an interaction between grazing and grass type. We choose REML over analysis of variance, as the design was unbalanced. We treated grazing as a fixed factor with two levels (grazed or ungrazed), grass type as a fixed factor with three levels (annual, mixed, or perennial) and site as a random factor. For posthoc contrasts, we compared the predicted means with the least significant difference (LSD) at α = 0.05. REML analysis, as described above, was also performed with the gene abundance data. All gene abundance data were log(x +1) or square root transformed, and all soil nutrient data were log(x+1) transformed. We also used REML to determine whether exotic grass cover and exotic grass species richness were lower in sites from which grazing had been excluded.

Multiple linear regressions.

We investigated the relationships between gene abundance and soil nutrients and whether these relationships changed with grazing exclusion (grazing factor) and weed invasion (grass type factor) by general linear models (GLMs). We initially explored possible GLMs using the GenStat regression procedure of all subsets (34). Model fit comparison was done with the Akaike information criterion and the adjusted R². Explanatory variables explored were grazing factor (grazed or ungrazed), grass type factor (annual, mixed, or perennial), soil carbon, nitrogen, plant-available phosphorus, nitrate, ammonia, pH, and bulk density. We investigated the relationships between the abundance of each gene by calculation of correlation coefficients.

We also determined multiple regression models to explain exotic annual grass cover. In addition to the variables used in the models for gene abundance, we also explored tree canopy, shrub, and ground cover. Native grass cover was strongly negatively correlated with exotic grass cover and was not included in the model.

RESULTS

Soil properties.

There was a large spread in the soil properties between sites with total carbon ranging from 15.7 to 288 g/kg (63.7 ± 48.2) (mean ± 1 standard deviation [SD]), total nitrogen from 0.481 to 19.0 g/kg (3.75 ± 3.1), and nitrate from undetectable to 140 mg/kg (8.91 ± 18.4). All sites had acidic soils, with a pH range of 3.53 to 5.45 (4.29 ± 0.34).

Soil nutrient levels were not lower in sites with livestock grazing excluded (Table 2), except for a trend for lower plant-available phosphorus (P = 0.095) (Table 2). Nitrate was the only nutrient that was significantly higher in the sites dominated by exotic annual grass (18.6 ± 4.1 mg/kg) compared to the sites with mixed native and exotic grasses (4.25 ± 0.83 mg/kg) or sites dominated by perennial native grasses (0.66 ± 0.20 mg/kg) (Table 2).

TABLE 2.

Summary of the residual maximum likelihood (REML) analysis for soil properties with dominant understory grass type and grazing treatment in 20 woodland remnants^a

Source	df	REML analysis for soil propertyb
		Ammonia level		Bulk density		Carbon level		Nitrate level		Nitrogen level		pH		Phosphorus level
		Wald	P	Wald	P	Wald	P	Wald	P	Wald	P	Wald	P	Wald	P
Grass type	2, 14	0.32	0.852	4.28	0.156	2.14	0.369	10.09	0.024	3.95	0.176	0.41	0.816	0.38	0.831
Grazing	1, 14	0.95	0.346	0.01	0.922	0.98	0.34	0.02	0.894	0.37	0.553	2.6	0.129	3.23	0.095
Grazing × grass type	2, 14	2.02	0.39	2.03	0.387	1.22	0.557	0.76	0.69	0.03	0.987	3.56	0.205	3.05	0.254

^aSummary of the residual maximum likelihood analysis for soil properties with dominant understory grass type (exotic annual, mixed exotic and native, or native perennial grasses) and grazing treatment (livestock grazed or grazing excluded).

^bThe Wald statistic and P values are shown. The P value of ≤0.05 is shown in boldface type.

Exotic grass cover (F_1,18 = 0.18 and P = 0.675) and exotic grass richness (F_1,18 = 0.28 and P = 0.601) were not lower when livestock grazing was excluded. Exotic annual grass cover was best explained by soil nitrate (P = 0.002) (positive relationship) and ammonium (P = 0.013) (negative relationship) (overall model R² = 54.1 and P = 0.001).

Nitrogen functional genes.

All five genes were detected at all sites and treatments, but not in all samples. The gene with the highest abundance was narG (1.1 × 10⁵ to 2.4 × 10⁷ copies/g soil) (minimum to maximum) followed by chiA (3.1 × 10⁴ to 2.4 × 10⁶) and nirK (2.2 × 10⁴ to 1.6 × 10⁶), with nifH having the lowest abundance (5.0 × 10² to 2.1 × 10⁴ copies/g soil). The only gene we did not detect in all samples was bacterial amoA (0 to 5.1 × 10⁵ copies/g soil), being below the limit of detection in 10 samples. These samples were all in native grass-dominated sites with grazing excluded. Regardless, there was no significant difference in the abundance of amoA or of any other gene between grazing treatments (Table 3).

TABLE 3.

Summary of the REML analysis for the abundance of functional nitrogen genes with grazing treatment and dominant understory grass type^a

Source	df	REML analysis for the abundance of the following functional nitrogen geneb:
		chiA		narG		nirK		nifH		amoA
		Wald	P	Wald	P	Wald	P	Wald	P	Wald	P
Grass type	2, 53	3.84	0.157	6.65	0.043	9.06	0.017	6.44	0.047	0.04	0.819
Grazing	1, 11	2.08	0.176	0.22	0.65	0.12	0.739	0.87	0.369	0.07	0.801
Grazing × grass type	2, 44	0.13	0.939	0.17	0.918	5.14	0.093	1.4	0.503	0.03	0.984

^aSummary of the REML analysis for the abundance of functional nitrogen genes with grazing treatment (livestock grazed or grazing excluded) and dominant understory grass type (exotic annual, mixed exotic and native, or native perennial grasses).

^bThe Wald statistic and P values are shown. P values of ≤0.05 are shown in boldface type.

The abundance of chiA was positively correlated with the abundance of both amoA (r = 0.472 and P = <0.001) and narG (r = 0.532 and P = <0.001). There was also a positive correlation between nirK and nifH (r = 0.313 and P = 0.005).

Overall, total soil carbon, total nitrogen, and soil pH best explained the variation in the abundance of the five genes (Fig. 1 and Table 4). The abundance of chiA and amoA were best explained by soil carbon and soil pH, with a positive correlation with both (Fig. 2 and Table 4). There were also positive correlations between chiA and total nitrogen (R² = 33.9, F_1,73 = 38.0, and P = 0.001), amoA and nitrogen (R² = 5.4, F_1,73 = 5.16, and P = 0.026), and amoA and plant-available phosphorus (R² = 11.1, F_1,73 = 9.91, and P = 0.002).

TABLE 4.

Summary of the multiple linear regressions between the abundance of functional nitrogen genes and soil properties

Response variable	Explanatory variable	R2 (adjusted)	Pa	F ratio or t statisticb
amoA	Overall	29.5	<0.001	16.5
	C		0.001	4.66
	pH		0.001	4.58
chiA	Overall	46.0	<0.001	32.5
	C		<0.001	7.45
	pH		<0.001	5.13
nifH	Overall	28.6	<0.001	10.9
	N		0.001	−5.28
	Graze (slope)		0.003	−2.67
	Nitrogen × graze		0.002	3.35
narG	Overall model	55.6	<0.001	19.3
	N		0.049	2.01
	Annual vs. perennial (intercept)		0.003	3.03
	Annual vs. perennial (slope)		0.005	−2.91
	Annual vs. mixed (slope)		0.004	−3.0
	Mixed vs. perennial (slope)		0.34	0.96
	N × annual grass vs. perennial		0.005	−2.91
	N × mixed grass vs. perennial		0.354	−0.93
	N × annual grass vs. mixed		0.004	−3.0
nirK	Overall model	31.1	<0.001	6.78
	N		<0.001	3.94
	pH		0.001	3.36
	Graze (slope)		0.42	0.68
	N × graze		0.001	−3.33
	pH × graze		0.827	0.22

^aP values of ≤0.05 are shown in boldface type.

^bThe F ratio is presented for overall model fit, and the t statistic is shown for fit of explanatory variables and group comparisons.

FIG. 2.

Fitted regression lines between the relative abundance of chiA and pH values (a) and soil carbon (b). The level of carbon in soil is shown in grams per kilogram. The 95% confidence intervals are also shown. Very similar patterns were observed between amoA, carbon, and pH (data not shown).

The abundance of nifH was significantly lower in sites dominated by native grasses than in the sites with mixed native and exotic grasses and sites dominated by exotic grasses (Fig. 3 a and Table 3). There was also a trend for higher abundance of nifH in sites where grazing was excluded (Fig. 3a). The abundance of nifH was negatively correlated with soil nitrogen, and this relationship was significantly different between the two grazing treatments (Table 4). The relationship was strongly negative for the grazed sites, but there was no significant relationship for the sites with grazing excluded (Fig. 4 b).

FIG. 3.

Average relative abundance of nifH (a) and narG (b) in relation to the dominant grass type within woodlands grazed by livestock (black) and with grazing excluded (white). Each value represents the mean ± 1 standard error (error bar).

FIG. 4.

Fitted regression lines for relationships of the nirK (a) and nifH (b) genes with soil nitrogen and how these vary between woodlands with livestock grazing present (open triangles, broken line) or grazing excluded (closed triangle, solid line).

There was a greater abundance of narG in the sites dominated by exotic grasses than in the sites dominated by native grasses and the sites with mixed native and exotic grasses (Fig. 3b and Table 3). The abundance of narG was positively correlated with soil nitrogen, and this relationship also varied with grass type (Table 4). The regression line for the annual grass-dominated sites had a significantly steeper slope and different intercept than that for the perennial grass-dominated sites (Fig. 3). The abundance of narG was also positively correlated with nitrate (R² = 38.8, F_1,73 = 47.3, and P = 0.001) and phosphorus (R² = 30.1, F_1,73 = 32.5, and P = 0.001).

The abundance of nirK was lower in the sites dominated by native grasses than in the sites dominated by exotic grasses or the sites with mixed native and exotic grasses (Table 4). The interaction between grass type and grazing type was also close to significant (P = 0.093). Posthoc analysis of this interaction indicated that nirK was more abundant in the native grass sites that were still grazed than in the sites with grazing excluded. The abundance of nirK was best explained by soil pH and nitrogen (Table 4), with a positive relationship with both. The interaction between nirK and nitrogen varied with grazing treatment, with a significantly steeper regression line for the sites with grazing excluded (Table 4 and Fig. 4a).

DISCUSSION

The abundance of the functional nitrogen cycling genes was strongly related to the soil properties, especially total carbon, total nitrogen, and pH. Soil physicochemical characteristics have been shown to have a greater influence on microbial community structure and function than the present land management practices (51). This was also the case in this study, whereby the abundance of amoA and chiA did not vary with ground-layer vegetation type or current grazing type. However, long-term agricultural management practices had influenced the underlying woodland soil properties, with average soil carbon and nitrogen levels twice those of reference condition woodlands with little disturbance (37) and the average nitrate and ammonia levels five times higher than those of reference condition woodlands. The maximum soil nutrient levels were 10- to 70-fold higher than the maxima recorded in reference grassy woodlands.

Impact of livestock grazing on soil properties and NFG abundance.

We did not detect any difference in soil properties between woodlands with and without livestock grazing. Not surprisingly, correlations between grazing and abundances of NFGs were weak. In the current study, livestock had been excluded for an average of only 7 years. However, we have detected improvements in the vegetation condition, invertebrate abundance and diversity, and leaf litter decomposition rates in these woodlands following livestock exclusion (24). Soil physical characteristics could take longer to respond to management changes than other aspects of the ecosystem. Nutrients added to remnant vegetation through agricultural activities are often very persistent and may remain elevated for hundreds of years after agriculture (29). Many forms of phosphorus are stable, water insoluble, or immobile (15), and ammonium can be tightly held onto clay and organic colloids, thus inhibiting conversion to other nitrogen species and leaching from the soil.

The relationships between nifH and nirK and soil nutrients differed with grazing treatments, but there was no difference in the abundance of any of these genes with grazing treatment per se. The relationship between nirK and soil nitrogen was weaker in the woodlands still grazed by livestock than those where grazing had been excluded. There would have been more frequent, abundant, and recent input of labile nitrogen from dung and urine in grazed sites, such that nitrogen may have been less limiting to the denitrifying bacteria.

The negative relationship between nifH gene abundance and nitrogen was significant only for the sites still grazed by livestock, suggesting that frequent livestock grazing could inhibit dinitrogenase reductase and lead to a reduction in the biological capacity for nitrogen fixation. The sites still grazed had more recent nitrogen inputs from livestock, redistributing nutrients from adjacent fertilized pasture, and a large proportion of the nitrogen in dung and urine is freely available to plants and microbes (43). High-nitrogen inputs, in various forms, can reduce the ecosystem dependence on free-living nitrogen-fixing organisms (36). If combined nitrogen (e.g., nitrate, urea) is present, microbes may use this instead of fixing nitrogen gas, as this takes less energy to assimilate. However, certain bacterial taxa may contain the nifH gene but are not actively involved in soil nitrogen fixation. As such, the relationship between nifH abundance (or relative enrichment in the DNA pool) and environmental properties will be strongest only in the systems that are highly nitrogen constrained for durations consistent with species selection via community turnover. This is also true for bacteria containing the narG, nirK, and chiA genes, but not for bacteria containing amoA.

Relationship between NFG abundance and soil properties.

Soil pH appeared to be an important factor influencing the biological capacity for nitrification, decomposition-ammonification, and denitrification, with gene abundance increasing as pH increased. Soil pH is an important factor controlling soil microbial composition and function (47, 50-52). The abundance of narG (18) and bacterial amoA (50), nitrification rate (10), and chitinase activity (21) have all been shown to increase with increasing pH in other ecosystems. Nitrification can become inhibited at low pH due to decreased ammonium availability (61), and pH can regulate both instantaneous denitrification rates and the denitrifier community in the long term (53).

There was a positive relationship between the soil phosphorous level and amoA and narG abundance. Both amoA and narG gene abundance have been shown to increase following the addition of phosphate fertilizer, and community composition of soil fungi and bacteria can change after the addition of phosphorus (50). Increases in the level of phosphorus in soil typically lead to a reduction in native plant diversity in grassy woodlands (7, 25), and this change in the plant community, independent of exotic grass invasion, could also alter the abundance and composition of nitrifying and denitrifying bacteria in the soil (14, 60). We also found significant positive relationships between chiA abundance and total soil carbon and nitrogen levels. This was not surprising, as decomposition of chitin contributes to both the labile carbon and nitrogen pools in the soil.

The abundance of both narG and nirK were positively correlated with soil nitrogen, and narG abundance was also positively correlated with nitrate. The addition of nitrogen, in particular nitrate, can increase denitrification rates both directly via increasing substrate concentrations and indirectly via distal controls, such as increases in litter decomposition rates (53, 54). Wakelin et al. (50) found that narG abundance increased with agricultural intensification, which could relate to higher soil fertility.

We could not detect amoA in 10 samples from woodlands with native grasses which had grazing excluded. This gene may have been present but below the limit of detection. These samples all had nitrate levels below the average (<1 mg/kg in seven samples and undetectable in three samples). This suggests that nitrification may have been occurring at a low rate in these areas. These sites were not grazed by livestock or fertilized, and as such, they would have received little external input of nitrate. It is also possible that ammonia-oxidizing bacteria (AOB) could have a more patchy distribution in nutrient-poor areas with low productivity, such as eucalypt woodlands which are no longer disturbed by agricultural activities.

Our choice of primers targeted betaproteobacterial ammonia oxidizers, and thus did not capture ammonia-oxidizing archaea (AOA). AOA have been found in higher abundance than ammonia-oxidizing bacteria (AOB) in acidic soils from several ecosystems (31, 32). All the soils in this study were acidic, but the samples in which amoA was not detected were no more acidic than the average. It remains unclear whether AOA play a key functional role in soil nitrification and, if so, under what range of environmental conditions (22, 31, 32). Some studies have suggested that AOA make a highly significant contribution to nitrification (5), while others have concluded they are far less functionally important than AOB (16). Recent work by Moin et al. (31) showed that the abundance of bacterial and archaeal amoA genes was strongly correlated in salt marsh sediments and that archaeal amoA genes have similar responses to some biogeochemical variables.

Impact of grass invasion on NFG abundance.

Woodland remnants dominated by exotic annual grasses had higher soil nitrate levels and a greater biological capacity for denitrification and nitrogen fixation than those dominated by native perennial grasses. There were numerous exotic clovers (Trifolium spp.) present in the woodlands dominated by exotic grasses and the woodlands with mixed exotic and native grasses, but these were absent from the woodlands dominated by native grasses. Therefore, it is possible that the higher abundance of nifH in the sites that had been invaded by exotic plants was due to the legume-rhizobium associations of clover, whereas the low levels of nifH at the sites dominated by native grasses were from free-living nitrogen-fixing bacteria. The nifH present would have to be characterized in more detail in order to confirm this.

The relationship between narG and soil nitrogen was strongest in the woodlands dominated by exotic grasses. The bacteria in the invaded sites could be more nutrient limited than those in the native-grass-dominated woodlands. There could also be more competition between the denitrifying bacteria and the annual grasses for nitrate in these sites.

Plant species composition is a major driver of bacterial community structure and function, and changes in microbial activity or composition can occur following weed invasion (56). The microbial denitrifying community may have increased in abundance or changed in composition as a result of the annual grass invasion or increased soil nitrate. The most common change detected following weed invasion is an increase in nitrifier abundance or nitrification rates (10, 14, 20, 23, 27).

Nitrogen is present in many forms in the soil, and some plant species can take up more than one type, depending on the soil chemistry. Early successional species generally have a preference for nitrate and can have high rates of nitrate uptake, especially when soil nitrate levels are elevated (30). Higher rates of nitrification would benefit the annual grasses which can access nitrate, such as Bromus hordeaceus (1, 59). However, we did not find an increase in the potential for decomposition-ammonification or nitrification with weed invasion. The increase in the biological potential for denitrification could be a response to increases in the nitrate and nitrite reductase enzyme substrates, nitrite and nitrate, rather than to changes in the plant community.

It is likely that the high soil nitrate levels assisted annual invasion, but the elevated nitrate was not a direct result of invasion. Soil nitrogen and phosphorus were in excess to what most native species require to survive at many sites, and the annual grasses could have been exploiting these unutilized resources.

Changes to the microbial community as a result of historical livestock grazing and nutrient addition could outweigh some of the subsequent changes following the more recent invasion of annual grasses. We do not know how long the exotic grasses had been present in the woodlands, but some land managers reported anecdotally their woodlands had been invaded for at least 20 years. In mixed-species systems, changes to gross nitrogen cycling rates mediated by the presence of specific exotic grasses may be patchy (14). Wakelin et al. (50) also found little difference in the microbial community and function between pastures of exotic annual grasses and those with native perennial grasses. Exotic grasses may have to completely dominate the understory vegetation (i.e., 100% cover) before we can detect changes in the abundance for some of the nitrogen functional genes. Alternatively, we may need to examine woodlands with no livestock grazing history before and after annual grass invasion, though such sites are rare. It is also possible that grass invasion may alter the diversity, rather than the abundance, of some of these functional genes; however, this was beyond the scope of the present study.

The methodology we used does not provide a direct measure of the rates of nitrogen transformations, and the presence of a functional gene does not always mean that the function is operating. However, changes in nitrogen functional gene abundance have been related to changes in bacterial population size, and we consider them useful ecological indicators of the major functional components of the soil nitrogen cycle. Abundance of the amoA gene has been positively correlated with ammonia oxidation rates (6, 32), and nifH gene abundance has been positively correlated with nitrogen fixation rates (50). While we do not have complete coverage of all of the nitrogen cycle genes, this data set does cover the major processes and gives insight into factors that relate to alterations in the nitrogen cycle in woodlands under different land management practices.

Conclusions.

Soil nitrogen appears to have played a greater role in influencing the functional nitrogen gene abundance than the more recent weed invasion and removal of livestock grazing. There were significant correlations between all five nitrogen functional genes with soil nitrogen or nitrate, with four of them having positive correlations. In these grassy woodlands, the relationship between nirK and narG with nitrogen was much stronger than between amoA and nitrogen, suggesting that nitrogen may influence the biological potential for denitrification more than the biological potential for nitrification.

Soil nutrient pools that reflect agricultural land use history can persist for decades (12). The elevations and variations in woodland soil nutrient levels are likely to be related to long-term land management practices, and the abundance of functional nitrogen genes could also be related to long-term land management practices. This lasting legacy of past land practices on soil biota and physicochemistry will make restoration of grassy woodland remnants more challenging than we may have considered hitherto. However, if the higher levels of biological potential for denitrification in the most disturbed sites are persistent, this could assist in reducing the soil nitrogen pool.