Assessment of the importance of dissimilatory nitrate reduction to ammonium for the terrestrial nitrogen cycle

Introduction Conclusions References


Introduction
Our understanding of the nitrogen (N) cycle has increased in recent years due to newly discovered processes and the finding that various groups of microorganisms are involved in N transformations, e.g.archaeal ammonia (NH 3 ) oxidation and codenitrification (Hayatsu et al., 2008;Francis et al., 2007).Among the many processes that constitute the N cycle are two processes of dissimilatory nitrate (NO − 3 ) reduction that occur under similar conditions of low oxygen concentrations (Tiedje et al., 1982): denitrification, which is the reduction of NO − 3 to gaseous N compounds (NO, N 2 O and N 2 ), and dissimilatory NO − 3 reduction to ammonium (NH + 4 ) (DNRA), which is also termed fermentative NO − 3 reduction, NO − 3 ammonification or fermentative ammonification.In both processes nitrite (NO − 2 ) is an intermediate product (Philippot and Højberg, 1999).Hence, our discussion on DNRA is in most points equally valid for dissimilatory NO − 3 and NO − reduction between denitrification and DNRA (Tiedje et al., 1982).DNRA is favoured under higher C/NO − 3 ratios when the electron acceptor (NO − 3 ) becomes limiting (Tiedje et al., 1982).
While the importance of DNRA in marine ecosystems (Burgin and Hamilton, 2007), the responsible enzymes and bioenergetics (Philippot and Højberg, 1999;Simon, 2002;Takaya, 2002;Kraft et al., 2011) as well as molecular techniques to track DNRA microorganisms (Philippot, 2005) were recently reviewed no such current review is available for DNRA in soils.In recent years N cycling studies have increasingly investigated DNRA in various ecosystems.Thus we think it is timely to revisit this often forgotten process, summarise the current knowledge of DNRA in terrestrial ecosystems and to explore its importance for soil N cycling.We will discuss how various environmental factors influence DNRA in soil and approaches to investigate the importance of DNRA in soil.Two approaches have been used: first, mi-crobiological techniques have been applied to identify soil microorganisms capable of performing DNRA and to assess their abundance, in particular in comparison with denitrifying microbes; second, 15 N has been used as a tracer to qualitatively investigate NH + 4 production from added 15 NO − 3 in order to elucidate if DNRA occurs in soil.Moreover, 15 N tracing techniques also allow the quantification of gross rates for DNRA, which will be highlighted in this review.

Environmental conditions for DNRA
The soil oxidation state is a principal factor that influences the importance of DNRA compared to denitrification (Matheson et al., 2002;Brunel et al., 1992) with DNRA by bacteria and fungi occurring under more reducing (anoxic) conditions (Takaya, 2002;Yin et al., 2002;Page et al., 2003).On the other hand, other studies showed that DNRA is less sensitive to variable redox conditions (Pett-Ridge et al., 2006) and less sensitive to O 2 than denitrification (Fazzolari et al., 1998).In the later study soil aggregates were incubated under various O 2 levels with the same NO − 3 concentrations combined with different levels of glucose C additions.The authors showed that the effect of variable O 2 on DNRA was dependent on the C/NO − 3 ratio and concluded that C rather than O 2 was the key factor regulating NO − 3 partitioning between denitrification and DNRA (Fazzolari et al., 1998).This study as well as the one by Smith (1982) confirmed the importance of the C/NO − 3 ratio on partitioning NO − 3 reduction between DNRA and denitrification as hypothesised by Tiedje et al. (1982).Yin et al. (1998) showed that significant DNRA occurred only at a C/NO − 3 ratio above 12.However, Matheson et al. (2002) argued that the effect of C/NO − 3 ratio on DNRA may be an artefact of experimental approaches.Experiments investigating DNRA under different C/NO − 3 ratios usually artificially alter either the organic C or NO − 3 content that can result in enhanced O 2 consumption due to stimulation of microbial activity or in altered soil redox potential due to that NO − 3 is an oxidising agent itself (Matheson et al., 2002).Therefore, Matheson et al. (2002) concluded that experimental evidence of the effect of C/NO − 3 ratio are most likely due to altered soil oxidation state, which is hence the key partitioning factor.However, in a tropical forest the natural difference in the C/NO − 3 ratio explained 44 % of the variability of gross DNRA rates determined by an in-situ 15 N tracing experiment (Silver et al., 2005).In contrast, in a laboratory incubation with intact soil cores from another tropical forest no correlation between DNRA and C/NO − 3 was detected (Sotta et al., 2008).These contrasting results could be due to the occurrence of DNRA in anaerobic micro-sites and that the bulk soil C/NO − 3 ratio may not be a representative indicator for the ratio at the site of activity.Tiedje et al. (1988) pointed out that it is the availability of a suitable organic C source, supporting respiration or fermentation, that regulates the population of DNRA Biogeosciences, 8, 1779Biogeosciences, 8, -1791Biogeosciences, 8, , 2011 www.biogeosciences.net/8/1779/2011/bacteria.Several studies showed that the addition of glucose, a carbohydrate that supports respiration as well as fermentation, stimulated DNRA (Buresh and Patrick, 1978;Smith and Zimmerman, 1981;Yin et al., 2002;Caskey and Tiedje, 1979;Fazzolari Correa and Germon, 1991;Yin et al., 1998).
In a 15 N labelling study with soil slurries Morley and Baggs (2010) reported that DNRA appeared to be stimulated more by carbohydrates (glucose and mannitol) than amino acids and butyrate, but Yin et al. (1998) reported that the carbohydrates glycerol and succinate do not support DNRA.Moreover, in two anaerobic soils, addition of glucose did not influence DNRA (Chen et al., 1995).Chen et al. (1995) identified several possible explanations, including the high native soil C content compared to the amount of added glucose, unfavourable redox conditions for DNRA or effects of soil rewetting.DeCatanzaro et al. (1987) also did not find an effect of glucose addition on DNRA, but revealed that DNRA was stimulated by alfalfa addition.This was apparently an effect of sulphur in alfalfa, which was released as -SH group during sulphur mineralisation of organic matter and served as reducing agent.Under anaerobic conditions, sulphide stimulates DNRA, by serving as an electron donor, and depresses denitrification, by repressing NO and N 2 O reductase (Myers, 1972;Brunet and Garcia-Gill, 1996).In this line De-Catanzaro et al. (1987) found in the above mentioned study a stimulation of DNRA when simultaneously adding glucose and sulphide, which contrasted the finding from only glucose addition.
Other C sources, like straw, glycerol, methanol and succinate did not promote DNRA (Buresh and Patrick, 1978;deCatanzaro et al., 1987;Yin et al., 1998).The reasons for this are not fully understood and deserve further investigations.Buresh and Patrick (1978) as well as Yin et al. (1998) attributed this finding to the fact that the mentioned C sources are poor substrates for fermentation.As DNRA was thought to be a solely fermentative process (Cole and Brown, 1980) these substrates hence also did not promote DNRA.However, as two distinct pathways of DNRA exist, one fermentative and one respiratory (Moreno-Vivián and Ferguson, 1998;Simon, 2002;Mohan et al., 2004), this can not be the sole explanation.For the respiratory DNRA Simon (2002) listed formate, H 2 and sulphide as substrates.Possibly the above mentioned C sources do neither favour the fermentative nor the respiratory DNRA pathway.This may also explain why in some cases addition of glucose does not support DNRA and some of the above mentioned contradictions in the response to C/NO − 3 , as this may depend on whether respiratory or fermentative DNRA bacteria are present.This is supported by DeCatanzaro et al. (1987) in respect to sulphide.
Several studies investigated the effect of pH on DNRA, though findings were partly contradictory.Higher DNRA was associated with alkaline conditions (Nõmmik, 1956;Stevens et al., 1998;Fazzolari Correa and Germon, 1991;Gamble et al., 1977) and Woods (1938) reported a pH optimum of 6.5 for NO − 2 reduction and of 7.5 for NO − 3 re-duction.In contrast other studies found a negative relationship between DNRA and soil pH (Davidson and Ståhl, 2000;Waring and Gilliam, 1983).For denitrification many studies found an effect of pH, however it appeared that this effect may be indirect due to changes in the availability of organic C ( Šimek and Cooper, 2002).Under acidic conditions the slow breakdown of organic matter decreases the availability of organic C for microorganisms and, hence, denitrification.
It is unknown if this is also the case for DNRA, but Waring and Gilliam (1983) reported that DNRA increased at lower pH (<4) in poorly drained soils, which was linked to the soluble C content.Therefore, contrasting findings of the pH effect on DNRA may partly be related to soil C availability and, hence, be of indirect nature.
Taken together, the oxidation status and the C/NO − 3 ratio appear to be the most important factors regulating the importance of DNRA in soil, while the effect of pH was not consistent.Other investigations found a correlation between DNRA and SOM, moisture or soil N (Gamble et al., 1977;Davidson and Ståhl, 2000).However, there are not enough data available in the literature to make a comprehensive analysis on the importance of the various factors.Hence, studies are needed to systematically investigate the main controlling factors of DNRA in soil.

Effects of plants on DNRA in soils
The presence of roots alters the activity and abundance of dissimilatory NO − 3 reducers in soils, as a consequence of altered substrate and oxygen availability (Philippot et al., 2009).It is well established that denitrification is generally stimulated by the presence of roots, due to exudates and oxygen consumption (Klemedtsson et al., 1987;Woldendorp, 1963).However, no study has investigated the direct effect of plants on DNRA in upland soils, but some information is available for wetland/freshwater plants, though findings are not conclusive.In the presence of reed sweetgrass (Glyceria maxima) DNRA bacteria (53 %) dominated the NO − 3 reducer community in a pot experiment (Nijburg and Laanbroek, 1997a), while in unplanted soil denitrifiers dominated (71 %).In contrast, the presence of reed (Thypha angustifolia) had little effect on the functional groups of NO − 3 reducers, with DNRA bacteria accounting for 12 and 19 % in the bulk and rhizospheric sediment from a freshwater lake, respectively (Brunel et al., 1992).A higher contribution of DNRA to the recovery of added 15 NO − 3 was found in soil cores containing reed roots compared to root free cores (Nijburg and Laanbroek, 1997b).In contrast a 15 NO − 3 labelling microcosm study found that DNRA accounted for 49 % of NO − 3 consumption in unplanted soil, while in the presence of plants DNRA accounted for less than 1 % (Matheson et al., 2002).At the same time denitrification was higher in the planted soil.A riparian zone study by Dhondt et al. (2003) also showed that during the growing season denitrification www.biogeosciences.net/8/1779/2011/Biogeosciences, 8, 1779-1791, 2011 was dominant while DNRA predominated when plant activity was low.
Effective soil N retention is achieved by a tight coupling of DNRA with plant (and microbial) NH + 4 uptake as was observed in tropical upland soils (Templer et al., 2008).This highlights the need to better understand plant interactions with DNRA and N retention in upland soils by conducting parallel studies with planted and unplanted soils.Apart from assimilation, N retention may also occur due to adsorption of NH + 4 , produced via DNRA, on clay minerals or organic matter, and is governed by the cation exchange capacity.

Production of N 2 O during DNRA
DNRA is generally seen as a process that conserves N in the ecosystem.However, many microorganisms conducting DNRA also produce N 2 O (Cole, 1988).Kaspar (1982) suggested that N 2 O production by DNRA microorganisms is a detoxification mechanism, in order to avoid high concentrations of NO − 2 .In a batch culture, a soil Citrobacter sp.produced N 2 O and NH + 4 by enzymatically reducing NO − 2 (Smith, 1982).The use of 13 NO − 3 labelling proved that several microorganisms were able to simultaneously produce NH + 4 and N 2 O via dissimilatory pathways, whereby NH + 4 accounted typically for >90 % of the total product (Bleakley and Tiedje, 1982).This finding is in line with an anaerobic batch incubation study where all DNRA isolates from three different soils showed N 2 O production, which accounted for 5-10 % of added NO − 3 (Smith and Zimmerman, 1981).However, as stated by Cole (1988), the N 2 O production rate by DNRA microorganisms is typically in the range of 1 % of the NO − 2 or NO − 3 reduction.Based on a 15 NO − 3 labelling study, Stevens et al. (1998) concluded that DNRA became a more important process for N 2 O production with increasing pH, possibly as a mechanism to reduce harmful NO − 2 that tended to accumulate under high pH.However, as both DNRA and denitrification use the same substrates (NO − 3 and NO − 2 ) the contribution of these two processes to total N 2 O production can not be investigated based on 15 NO − 3 labelling alone.Thus the identification of the responsible microorganisms is required (Stevens et al., 1998).To achieve this, quantifying enzyme activity rather than investigating microbial species or functional genes is needed, as discussed for linking denitrifier density to functioning by Philippot and Hallin (2005).

Soil microorganisms involved in DNRA
The aim of this section is to summarise studies that compared the abundance of soil DNRA microorganisms to denitrifiers.The capability for NO − 3 respiration and for DNRA is widely spread among bacteria (Philippot, 2005;Simon, 2002).Tiedje (1988) listed several genera of soil DNRA bac-teria, which are either obligate anaerobes (Clostridium), facultative anaerobes (Citrobacter, Enterobacter, Erwinia, Escherichia, Klebsiella) or aerobes (Bacillus, Pseudomonas), and recently DNRA been shown for several rhizobial species (Polcyn and Podeszwa, 2009).In addition a soil Arthrobacter strain, an abundant soil genus worldwide which is regarded as an obligate aerobe, showed DNRA when incubated anaerobically (Eschbach et al., 2003).This was also shown for eight Nitrobacter strains that were regarded as obligate aerobe NO − 2 oxidiser (Freitag et al., 1987).Moreover, the capability for DNRA is widely distributed among common soil fungi, mostly belonging to the ascomycota (Zhou et al., 2002).Many bacteria capable of DNRA are found in the Enterobacteriacea, which is the only group of procaryotes with no known denitrifier (Zumft, 1997).Mohan and Cole (2007) pointed out that there is no known bacterium capable of both, denitrification and DNRA.Recently, however, Behrendt et al. (2010) provided evidence by growth tests that two newly described Paenibacillus species, including one fen soil isolate, showed a versatile metabolism and were capable of heterotrophic nitrification, DNRA and denitrification.Moreover, Zhou et al. (2002) showed that denitrification and DNRA are alternatively expressed in a common soil fungus (Fusarium oxysporum) depending on oxygen status and available C source (Zhou et al., 2002).These authors classified the metabolism of this fungus depending on O 2 status as: DNRA under anoxic conditions, denitrification when O 2 supply was limited and aerobic respiration under sufficient O 2 supply.
Microorganisms which reduce NO − 3 via a dissimilatory pathway can be classified either as (a) denitrifiers, producing gaseous N compounds, (b) NO − 2 accumulators which reduce NO − 3 only to NO − 2 , or (c) DNRA microorganisms that reduce NO − 3 or NO − 2 to NH + 4 .Several authors compared the abundance of DNRA and denitrifying bacteria in soils, which, however, does not provide information on the activity of these bacterial groups in soil.This would require alternative approaches, e.g. 15 N labelling as discussed below (Sect.6).Evidence for a direct competition between DNRA bacteria and denitrifiers for NO − 3 comes from a soil inoculation study using 15 NO − 3 as a tracer (see Sect. 6) (Fazzolari et al., 1990).When a non-sterile soil was inoculated with the DNRA bacterium Enterobacter amnigenus, the production of 15 NH + 4 from 15 NO − 3 increased compared to non-inoculated control soil while at the same time N 2 O production decreased.In contrast, inoculation with the denitrifier Agrobacterium radiobacter resulted in faster N 2 O production while no 15 NH + 4 was produced.Simultaneous inoculation with both bacteria resulted in intermediate results (Fazzolari et al., 1990).
Studies, investigating microorganisms with a dissimilatory NO − 3 reduction pathway, often found that non-denitrifying NO − 3 reducers were most abundant.In all 19 soils investigated by Gamble et al. (1977) the number of NO − 2 accumulators outnumbered denitrifiers, with an average ratio of 4:1.
Biogeosciences, 8, 1779-1791, 2011 www.biogeosciences.net/8/1779/2011/This is consistent with the ratio of 4:1-3:1 for DNRA bacteria to denitrifiers in soil reported by Tiedje et al. (1982) and agrees with Bengtsson and Bergwall (2000) who reported a higher number of DNRA bacteria than denitrifiers for a spruce forest soil.Moreover, Smith and Zimmerman (1981) found that non-denitrifying bacteria dominated NO − 3 reducers, but most were NO − 2 accumulators.However, the majority of NO − 2 accumulators were capable of DNRA when NO − 3 was limited (Smith and Zimmerman, 1981), which was also found in a Klebsiella sp.(Dunn et al., 1979).In contrast, Brunel et al. (1992) found that, after addition of glycerol, only few strains of NO − 2 accumulators were capable of DNRA but more were able to produce N 2 O.In this context it is interesting to notice that the growth of the DNRA bacterium Enterobacter amnigenus was only related to NO − 3 reduction to NO − 2 but not to the reduction of NO − 2 to NH + 4 (Fazzolari et al., 1990).These authors concluded that NO − 2 reduction may serve as an electron sink but not for energy generation.In three paddy soils, the number of DNRA bacteria was only 19-35 % of the number of denitrifiers (Yin et al., 1998).However, in two other paddy soils Yin et al. (2002) found that the number of DNRA was higher than denitrifiers when the soil was pre-incubated or when C was added, which again points to the importance of C as discussed in Sect. 2. However, all discussed results are based on culturable microorganisms.Moreover, the activity of a DNRA bacterium differed between pure culture and soil inoculation (Fazzolari et al., 1990).Investigating the abundance of bacteria by functional genes, Kandeler et al. (2009) reported that denitrifiers accounted for less than half of the total NO − 3 reducer community in a forest soil.Molecular approaches for studying the microbial community of NO − 3 reducers in-situ were recently reviewed by Philippot (2005) and are therefore not repeated here.The review by Philippot (2005) highlighted the importance of functional genes, of culture independent approaches and of quantitative information when investigating denitrifier and NO − 3 reducer communities.Moreover, to link microbial diversity with functional activity Philippot and Hallin (2005) pointed out the need for investigating enzyme activities rather than functional genes (DNA as well as mRNA).This could provide a strong tool for investigating and comparing the DNRA and denitrification activity in soil.
6 Investigating DNRA by 15 N tracing techniques 15 N tracing techniques are commonly used to investigate the fate of N in terrestrial ecosystems (Hart and Myrold, 1996).These techniques are also used to quantify gross transformation rates (see Sect. 6.2).To confirm the occurrence of DNRA in soil, various researchers applied 15 NO − 3 and measured the 15 N enrichment of NH + 4 after incubation.Commonly, these studies applied also NH + 4 to inhibit immobilisation of NO − 3 due to high NH + 4 concentrations.If NO − 3 immobilisation occurred at significant rates, 15 N enrichment of NH + 4 may also be the result of immobilisation (i.e.assimilatory NO − 3 reduction) and subsequent remineralisation.However, there seems to be no study that has systematically investigated if the assumption of negligible NO − 3 immobilisation holds true.

15 N labelling to confirm DNRA occurrence
More than 50 years ago Nõmmik (1956) showed that a small amount of added 15 NO − 3 was converted to NH + 4 , but only under strictly anaerobic conditions.Therefore, it was concluded that DNRA is "extremely insignificant" under the prevailing conditions in arable soil (Nõmmik, 1956).This was also true for six tropical soils from the Philippines, where only a small fraction (<2 %) of added 15 NO − 3 was recovered as NH + 4 (MacRae et al., 1968).After incubating ten agricultural soils with varying texture and properties, Fazzolari Correa and Germon (1991) showed that 10-38 % of added 15 NO − 3 was reduced to NH + 4 via DNRA when a labile C source was added.This is in the same range as found by Stanford et al. (1975) and Wan et al. (2009).In these studies only very small amounts of 15 NO − 3 were recovered as 15 NH + 4 without addition of labile C.However, Fazzolari Correa and Germon (1991) pointed out that all the conditions required for DNRA can be present in agricultural soils.Furthermore, several studies using anaerobic soil incubations attributed 15 NO − 3 recovery as 15 NH + 4 to DNRA (Buresh and Patrick, 1978;Ambus et al., 1992;Chen et al., 1995;Dhondt et al., 2003;Yin et al., 1998).
In soil cores from a riparian fen, DNRA was only detected below a depth of 5 cm (Ambus et al., 1992).However, when the same soil was incubated as slurry, DNRA did not differ between three soil layers (0-5, 5-10 and 10-20 cm), but the ratio of DNRA to total NO − 3 reduction increased with depth (Ambus et al., 1992).This finding points to the effect that experimental conditions (e.g.soil slurry vs. core incubation) can have on experimental results, which impedes the comparison of results from different studies.Another riparian buffer zone study using slurry incubations showed that DNRA was only significant during the dormant season.This was attributed to low redox potentials and high inputs of labile C during that time (Dhondt et al., 2003).The studies by Ambus et al. (1992) and Dhondt et al. (2003) both point to the occurrence of DNRA under more reduced conditions compared to denitrification.However, some studies indicated that DNRA and denitrification can occur simultaneously in soil (Paul and Beauchamp, 1989;Stevens and Laughlin, 1998;Morley and Baggs, 2010), possibly in different micro-sites that differ in redox state.Furthermore, in contrast to Ambus et al. (1992), slurry incubations with soil from a riparian zone by Davis et al. (2008) resulted in higher DNRA rates in the surface soil (0-15 cm) compared to sub-soil (130-155 cm).A similar result was observed in an adjacent cropping system.The different results between the two studies may be related to

Gross DNRA transformation rates
Higher numbers of DNRA microorganisms compared to denitrifiers may not necessarily reflect a higher DNRA activity in soil.Thus, investigations of gross transformation rates are needed to evaluate the importance of DNRA.This can be achieved by 15 N labelling techniques in conjunction with data analysis via analytical or numerical models.Most 15 N labelling studies to date did not consider DNRA, because it was assumed to be a negligible process.Here, we will summarise gross DNRA rates reported from soils and explore if DNRA is indeed negligible or must be considered as an important N pathway in soil.
The first reported gross DNRA rate measurement in soils was presented by Ambus et al. (1992) for a riparian fen.Unfortunately, no equation for the calculation of the gross rates was provided.An analytical solution to calculate gross rates for DNRA, based on the increase of the 15 N enrichment of the NH + 4 pool after addition of 15 NO − 3 as a tracer, was developed by Silver et al. (2001) (the actual equations are presented in: Huygens et al., 2008).The derived analytical equations were applied to investigate DNRA in various ecosystems, mostly tropical forest soils (Table 1).Several studies showed that DNRA was a significant and sometimes dominant fate of NO − 3 in terrestrial ecosystems (Table 1).Some studies compared DNRA and denitrification rates.In a tropical forest soil DNRA was threefold higher than denitrification (Silver et al., 2001) and in a spruce forest the gross rate of DNRA was three orders of magnitude higher than gross denitrification (Bengtsson and Bergwall, 2000).This was also found for freshly sampled soil from another tropical forest (Pett-Ridge et al., 2006).However, when this soil was pre-incubated for 3-6 weeks under different redox regimes denitrification exceeded DNRA in all cases.Furthermore, Pett-Ridge et al. (2006) found that gross DNRA was unexpectedly higher in aerobic soils than in anoxic soils or in soils with fluctuating redox conditions.They explained this observation by higher NO − 3 concentrations in aerobic soil and the possibility of the occurrence of DNRA in anaerobic soil aggregates.Previously, Silver et al. (2001) showed that the rate of DNRA can be limited by the availability of NO − 3 that is caused by a small pool size in conjunction with high turnover.The DNRA rate constant, calculated as gross rate divided by NO − 3 concentration, was, however, highest in anoxic soils (Pett-Ridge et al., 2006).Comparing the rate constant of denitrification and DNRA revealed that DNRA bacteria are more competitive for NO − 3 under fluctuating redox conditions (Pett-Ridge et al., 2006).Thus, it appears that under certain environmental conditions DNRA bacteria are able to compete successfully with denitrifying bacteria for NO − 3 , which supports the theoretical advantage of DNRA under low NO − 3 concentrations (Tiedje et al., 1982; see Sect. 2).
Unfortunately analytical solutions for quantifying gross transformation rates, and particularly DNRA, introduce inconsistencies.These inconsistencies can occur when the assumption that no 15 N is recycled into the labelled pool does not apply, when inappropriate kinetic settings for N transformations are used (Rütting and Müller, 2007) or if NO − 3 consumption and DNRA are calculated separately.Using an analytical model, Templer et al. (2008) found a higher gross rate for DNRA compared to total NO − 3 consumption in one out of three tropical forest soils, which could have been due to inconsistencies.To overcome the problems associated with analytical solutions we recommend using numerical data analysis via so called 15 N tracing models (Rütting and Müller, 2007;Rütting et al., 2011), which enables a simultaneous analysis of all NO − 3 consumption pathways in a coherent model framework.This recommendation is in line with Silver at al. ( 2001) who stated that "numerical modeling may provide an alternative approach to explore the role of DNRA under a variety of scenarios".The only 15 N tracing model that included DNRA was presented by Müller at al. (2004;2007).An alternative approach to numerical tracing models was presented by Tietema and van Dam (1996), who combined 15 N experiments with a simulation model.In this model DNRA was simulated as a function of microbial biomass, but was independent of substrate concentrations.However, to our knowledge this methodology has subsequently been applied in only one other study (Verburg et al., 1999), in which, moreover, DNRA was solely simulated as function of soil C content, as no tracing of 15 NO − 3 into 15 NH + 4 was determined.
Application of the 15 N tracing model developed by Müller et al. (2004Müller et al. ( , 2007) ) showed that DNRA is likely to occur in numerous ecosystems and was sometimes the dominant NO − 3 consumption process, as can be seen from the results of the studies summarised in Table 1.However, other investigations did not find evidence for DNRA (Cookson et al., 2006;Laughlin et al., 2008), confirming that DNRA may be important only in some, but not all ecosystems (Stanford et al., 1975).Such observations, however, may also be related to the experimental conditions.One additional advantage by using numerical 15 N tracing models is that correlations between N transformations can be investigated (Müller et al., 2007).This enabled the detection of functional linkage between DNRA and the organic pathway of heterotrophic nitrification (oxidation of organic N) in a Nothofagus forest on an Andisol in southern Chile (Rütting et al., 2008).The authors considered this link to be an adaptation of the microbial community with the result that N losses could be minimised.Recently two new bacterial species were described that performed simultaneously DNRA and heterotrophic nitrification (Behrendt et al., 2010)  producing process, when coupled to the organic pathway of heterotrophic nitrification, to mineralisation of soil organic nitrogen (SON) as described by Schimel and Bennett (2004); (b) alternative pathways transferring NO − 3 to NH + 4 in soil (Burger and Jackson, 2004) that can be evaluated by 15 N tracing studies in combination with numerical data analysis if all the shown N pools are measured.direct mineralisation pathway (Fig. 1a).In the Nothofagus forest DNRA accounted for more than 90 % of total NO − 3 consumption (Rütting et al., 2008;Huygens et al., 2007).However, the transfer of 15 N from NO − 3 to NH + 4 could in fact be due to three different pathways (Fig. 1b): (1) DNRA, (2) plant N efflux and (3) remineralisation by microorganisms (Burger and Jackson, 2004).Using data from a microcosm 15 N study and simulation models Burger and Jackson (2004) provided evidence that each of the three pathways was on its own able to explain the observed 15 N dynamics.Numerical 15 N tracing models (e.g.Müller et al., 2007) have the potential to investigate the most likely pathway of NO − 3 reduction to NH + 4 when the 15 N enrichment of roots, soil organic N and microbial biomass are measured in addition to the mineral N pools (Fig. 1b).The alternative pathways should be tested to identify via a likelihood analysis whether DNRA or alternative pathways occurred.In the above mentioned Nothofagus study these alternative pathways (plant N efflux and remineralisation) could be ruled out as no roots were present in the laboratory incubation and the 15 N enrichment in five organic N fractions was too low to explain the 15 NH + 4 enrichment by remineralisation (Rütting et al., 2008;Huygens et al., 2007).More detailed studies are needed to investigate the importance of the alternative pathways proposed by Burger and Jackson (2004), i.e.DNRA, plant N efflux and remineralisation by microorganisms, by combining 15 N labelling studies with numerical data analysis.
Along a climate gradient in China, ranging from tropical to temperate zone, Zhang et al. (2011b) investigated gross N dynamics in 13 forest soils.In all but one soil gross DNRA activity was observed and DNRA was in half of the investigated soils the sole NO − 3 consumption pathway.Highest gross DNRA was found for the tropical soil (0.155 µg N g −1 soil day −1 ), whereas rates were similar for sub-tropical and temperate forest soils (0.01-0.11 µg N g −1 soil day −1 ; Table 1).The soil not exhibiting DNRA dynamic was collected from a sub-tropical, evergreen broad-leaf forest, while all other soils were either from deciduous broad-leaf or conif-erous forests.This may point to an effect of the quality and quantity of plant litter on the DNRA activity, even though other studies did observe high gross DNRA rates in soil from tropical evergreen forests (e.g.Templer et al., 2008).Moreover, some studies compared the gross DNRA in soils from temperate and sub-tropical broad-leaf and coniferous forests (Zhang et al., 2011a;Zhang et al., 2011b;Staelens et al., 2011).In three out of four sites higher DNRA rates were observed in soil underneath broad-leaf species, which may be related to the fact that broad-leaves usually contain a higher amount of labile C compared to coniferous needles, which may stimulate DNRA (see Sect. 2).However, in the fourth site the DNRA was about twice as high under coniferous forest than broad leaf-forest, but the reason for this is not clear.
It was also shown that the soil type had a significant effect on DNRA, with higher rates in clay compared to sandy soil (Sotta et al., 2008).No explanation was provided for the differences, which are however likely due to the higher soil C and N content and higher anaerobicity in the clay soil.This is in line with findings from a Swedish catchment area, where higher gross DNRA was observed for organic compared to mineral soils, but the differences could be explained by differences in soil organic matter content and the gravimetric water content (Rütting et al, unpublished results).

Effect of global change on DNRA
The functional importance of DNRA in soil is its capacity to increase N retention, as NO − 3 is transformed to NH + 4 .Ammonium is available for plant and microbial uptake, but is less prone to losses via leaching or as gaseous compounds (Buresh and Patrick, 1978;Tiedje, 1988;Silver et al., 2001;Huygens et al., 2007).Current climate change scenarios suggest that many ecosystems may become increasingly N limited in the future.This is mainly due to increased atmospheric CO 2 concentration, which can lead to a higher plant N demand (Hungate et al., 2003;Luo et al., 2004)  deposition (Johnson, 2006).However, Hungate et al. (2003) showed that the expected increase in N deposition will not cover the additional N demand under elevated CO 2 , indicating that N retention processes such as DNRA may become more important for ecosystem productivity.Tietema and van Dam (1996) investigated the effect of increased N deposition on the N cycle in two coniferous forest soils.At one site they found higher DNRA under preindustrial (1-2 kg N ha −1 yr −1 ) compared to increased deposition (31-37 kg N ha −1 yr −1 ), while no effect was found for the second forest.In a forest N fertilisation experiment in Sweden no clear pattern of the relationship between the amount of fertiliser applied (0-180 kg N ha −1 yr −1 ) and 15 NO − 3 reduction to 15 NH + 4 was found (Bengtsson and Bergwall, 2000).Kandeler et al. (2009), however, found lower total nitrate reductase activity in soil where N deposition was decreased, although the total number of nitrate reduction genes was not affected.The effect of elevated CO 2 on N cycling rates, including DNRA, was recently investigated in soils from two long-term free air CO 2 enrichment studies on temperate grassland (Müller et al., 2009;Rütting et al., 2010).In both of these studies DNRA was stimulated under elevated CO 2 by 140 and 44 %, respectively, most likely due to an increased C input into the soil that stimulated microbial activity and possibly increased anaerobicity.

Importance and regulation of DNRA
Summarising the findings of 15 N labelling studies indicates that the occurrence of DNRA is more widely spread in soils than previously thought (Table 1).Notably, only one study has reported gross DNRA rates in an arable soil (Inselsbacher et al., 2010) and none for boreal ecosystems, indi-cating that further studies in these ecosystems are needed.In the biomes where data are available (temperate and (sub-)tropical forest as well as temperate grassland) the importance of DNRA (measured as contribution to total NO − 3 consumption, %C NO 3 ) range from negligible to dominant (Table 1).This poses the question which environmental factors do regulate the rates and the importance of DNRA in soil.The best way to approach this question would be to conduct a meta-analysis.However, this is hindered by two facts.First, there is a lack of data on environmental factors in the literature.Notably, almost none of the studies on gross DNRA have reported the soil organic matter content, which has been identified as the best predictor for DNRA in a Swedish catchment (Rütting et al., unpublished results).Second, no general agreement has been reached on the most suitable method to quantify gross DNRA rates.As mentioned above two approaches have been used, applying either analytical or numerical data analysis.Moreover, some of the studies have used the classical pool dilution technique which is not suitable to resolve the simultaneously occurring NO − 3 reduction pathways (e.g.Silver et al., 2001;Sotta et al., 2008;Inselsbacher et al., 2010).The quantification of gross DNRA requires a "mirror labeling" approach (Barraclough, 1997) in order to assure the same microbial activity in both 15 N labeling treatments.
We conducted some statistical analyses with the available data set (see Supplement Table SI-1) using SigmaPlot (Version 11, Systat Software, Inc.).Based on Pearson's correlation (Table 2a) there were significant positive correlations of gross DNRA with mean annual precipitation (MAP) and total soil C (TC) and gross DNRA tended to be poritively correlated to the C/N ratio.As a next step we conducted multiple linear regressions (stepwise forward approach) to find combinations of variables that best predict gross DNRA in soils.
Gross DNRA rates were best explained by a combination of mean annual temperature (MAT) and MAP (R 2 = 0.38) according to: DNRA= − 0.0714 + 0.0102 • MAT + 6.13 × 10 −5 • MAP (1) These findings provide an indication that gross DNRA rates are influenced by the climatic conditions as well as edaphic factors.
High gross rates must not necessarily mean a high importance in an ecosystem.Therefore we calculated the portion of DNRA to total gross NO − 3 consumption (%C NO 3 ) and conducted statistical analyses in order to identify the importance of DNRA in the studied ecosystems.There is a significant correlation of %C NO 3 with the concentration of both mineral N forms (NH + 4 and NO − 3 ), MAT as well as the water filled pore space (WFPS; Table 2b).Notably, the correlation to WFPS and MAT are negative.The results from the multiple linear regression showed that %C NO 3 can be predicted from a linear combination of total soil N (TN) and WFPS (R 2 = 0.76): %C NO 3 =90.279+ 101.185 • TN − 1.338 • WFPS (2) Taken together, these results indicate that highest gross DNRA can be expected in soils with a high soil C (and possibly organic matter) content in humid regions.Indeed, highest gross DNRA has been reported for tropical and temperate rainforests (Table 1).However, the relative importance of DNRA (measured as % C NO 3 ) seems to be higher in temperate climates (Table 2a) in soil with lower soil moisture.This later finding seems to contradict with the anoxic nature of DNRA (see Sect. 2), but may be explained by the higher NO − 3 availability in less moist soils or by the higher tolerance of DNRA to changing redox conditions and O 2 compared to denitrification (Pett-Ridge et al., 2006;Fazzolari et al., 1998).

Conclusions
More than thirty years ago Cole and Brown (1980) concluded that the significance of DNRA in anaerobic soil was unknown.Now, with the use of 15 N labelling techniques and the quantification of gross DNRA rates, the hypothesis that DNRA "may be much more important than presently realized" (Stevens et al., 1998) seems to be confirmed.Gross DNRA rates can be quantified via 15 N tracing studies in combination with numerical data analysis and 15 N tracing models that consider DNRA as well as all N transformations that interact with each other (Rütting et al., 2011).A particularly powerful tool for future investigations can be the combination of 15 N tracing and molecular approaches (Wallenstein and Vilgalys, 2005;Philippot and Hallin, 2005).Summarising the findings of several studies, we conclude that DNRA is a significant, or even dominant, NO − 3 consumption process in some ecosystems (Table 1).The importance of DNRA may even increase under current climate change scenarios.Previously, it was concluded that the potential for significant DNRA exists in most soils, but that it is only expressed under anoxic conditions when C is readily available, possibly in anaerobic micro-sites (Smith and Zimmerman, 1981;Caskey and Tiedje, 1979).Yin et al. (1998) showed that a soil C/NO − 3 ratio above 12 seems to be a threshold for significant DNRA activity, but more studies are needed to ascertain if this threshold is a general feature or variable depending on soil properties.As Burgin and Hamilton (2007) concluded for aquatic systems, more work is also needed to understand the importance of DNRA in various terrestrial ecosystems.Therefore, future investigations on the soil N cycle in different terrestrial ecosystems (forest, agricultural land, grassland, wetland) should focus not only on 'classical' N cycling processes such as nitrification, denitrification and mineralisation, but also should include processes such as DNRA, because the occurrence of this process is often an indicator for ecosystem N retention.Finally, the N mineralisation paradigm of Schimel and Bennett (2004) should be adapted and include DNRA as an alternative NH + 4 producing process, in particular in conjunction with the postulated link to the organic pathway of heterotrophic nitrification (Fig. 1).An improved understanding of the conditions that govern whether NO − 3 is reduced to gaseous N or NH + 4 could also provide possible mitigation scenarios for N 2 O. Supplement related to this article is available online at: http://www.biogeosciences.net/8/1779/2011/bg-8-1779-2011-supplement.pdf.

aFig. 1 .
Fig. 1.Schematic overview of the importance of DNRA in soil: (a) as an alternative NH +4 producing process, when coupled to the organic pathway of heterotrophic nitrification, to mineralisation of soil organic nitrogen (SON) as described bySchimel and Bennett (2004); (b) alternative pathways transferring NO − 3 to NH + 4 in soil(Burger and Jackson, 2004) that can be evaluated by 15 N tracing studies in combination with numerical data analysis if all the shown N pools are measured.
Rütting et al.:Assessment of the importance of dissimilatory nitrate reduction to ammonium different soil depths investigated, due to changing substrate availability and redox conditions with depth.

Table 1 .
, supporting the above proposed functional linkages.Such a functional link, if proved to be a general pattern in soil, could provide an alternative pathway of NH + Summary of gross DNRA rates [µg N g −1 soil day −1 ] in terrestrial ecosystems (mean ± standard deviation, if available, or range of values) calculated by analytical (A) or numerical (N) 15 N tracing models as well as the portion of DNRA to total gross NO − 3 consumption (% C NO 3 ).

Table 2 .
Results from Pearson Product Moment Correlation of (a) gross DNRA rates as well as (b) portion of gross DNRA to total gross NO − 3 consumption (%C NO 3 ) to measured environmental factors: soil pH (measured in water), total soil carbon and nitrogen (TC and TN), concentration of ammonium and nitrate ([NH + 4 ] and [NO − 3 ]), C/N ratio, mean annual temperature and precipitation (MAT and MAP), incubation temperature (T inc ), gravimetric water content (GWC) and water filled pore space (WFPS).Variables used for multiple linear regression of gross DNRA and %C NO 3 as dependent variable, respectively.Not all variables could be considered due to the degree of freedom. *