Our collaborator from the field of crop sciences in the Legume ..... It cannot form the polymers that carbon is capable of to provide ... called H. Joel Conn that observed in 1918 that only 1.5 to 10 % of soil bacteria were detectable ...... Sequencer technology has developed in a way that has improved the resolution attainable in ...
Bacterial community responses to rhizosphere conditions in an agricultural soil, and their relationship to nitrous oxide emissions
Daniel Aleksanteri Milligan Master’s thesis University of Helsinki Department of Food and Environmental Sciences MENVI Microbiology May 2014
2
3 HELSINGIN YLIOPISTO — HELSINGFORS UNIVERSITET — UNIVERSITY OF HELSINKI Tiedekunta/Osasto — Fakultet/Sektion — Faculty
Laitos — Institution — Department
Faculty of Agriculture and Forestry
Department of Food and Environmental Sciences
Tekijä—Författare — Author
Daniel Aleksanteri Milligan Työn nimi — Arbetets titel — Title
Quantification of soil bacterial community responses to rhizosphere conditions, and their relationship to nitrous oxide emissions Oppiaine — Läroämne — Subject
Microbiology Työn laji — Arbetets art — Level
Aika — Datum — Month and year
Sivumäärä— Sidoantal — Number of pages
M.Sc. thesis
April 2014
114
Tiivistelmä—Referat — Abstract
The Legume Futures experimental field in Viikki was planted with plots of two herbaceous perennial crops. A legume Galega orientalis and the grass Bromus inermis were grown individually, and as an intercrop. Separate stands of B. inermis monocrop were fertilized with urea. Bare fallow plots acted as a control. Length-heterogeneity PCR (LH-PCR) was used to characterize soil bacteria based on the lengths of the V1– V3 regions of the 16S rRNA genes present in the community. When separated by capillary electrophoresis, fragments produced electropherogram peaks that indicated the relative abundances of various bacterial taxa. Soil pH, organic matter content, and potential denitrification were measured. Soil moisture, mineral nitrogen, and N2O emissions data were provided by a collaborator. The results confirmed previous findings that LH-PCR is highly reproducible, and indicated that it is amenable to analysis by multivariate statistic methods, so that small differences in community composition could be resolved. Seasonal variation was high, but treatment effects could also be seen. The largest differences were between fallow plots and plots sown with grass. The pure stands of legume were also distinguishable, but the three grass treatments could not be resolved from each other by any given peak. Peak correlations to a number of soil edaphic factors existed, particularly pH, which is an established determinant of bacterial communities. Measured edaphic factors are known to correlate with denitrification activity, so direct relationships are not unambiguous where correlations existed between peak data and N2O emissions. Soil nitrate and moisture in particular were also found to regulate numbers of some groups. Nevertheless, several components of the community could be identified that were better predictors of emissions than measured soil parameters. Improved bioinformatics tools are needed to link fragment lengths to sequence data and obtain the full benefit of this method. Avainsanat — Nyckelord — Keywords
denitrification, microbial ecology, LH-PCR, rhizosphere, ecosystem services, nitrous oxide, legume, Proteobacteria, Actinobacteria, soil bacteria Säilytyspaikka — Förvaringställe — Where deposited
Muita tietoja — Övriga uppgifter — Additional information
Supervisor – Professor Kristina Lindström
4
5
Preface and acknowledgements For the work recorded in this thesis I owe a lot to the colleagues and teachers that have provided their valuable assistance and input. First and foremost, I am grateful for the support of my supervisor prof. Kristina Lindström, who was kind enough to conceive of this project so that I could research the subject of my interest, the soil microbial community. I will not soon forget the opportunities she has given me to broaden my experience during my master’s studies. In the project I was fortunate to have had the cooperation of Miiro Jääskeläinen, whom it was always a pleasure to toil away in field and lab with. I received instruction in the nuances of gas sampling and chromatography from Dr. Asko Simojoki. Our collaborator from the field of crop sciences in the Legume Futures project was Dr. Fred Stoddard, who pointed me in the right direction for the statistical analysis. I am indebted to Dr. Antonella Scalise for selecting primers and conditions for nested PCR and the nitrogen group was a positive environment to work in. It was a joy to have such a diverse and experienced bunch following my numerous progress updates and providing much-appreciated feedback. In particular I benefited from the experience of Lijuan Yan, Anu Mikkonen, and Minna Santalahti in using LHPCR, and Janina Österman informed me about the rhizobial species growing with our plants. Thanks to Leena Räsänen for demonstrating Bionumerics for me, and also to Riitta Saastamoinen and Mika Kalsi, who always had time to help me find obscure lab supplies. Finally, I would like to thank Marina Kliuchko for her companionship, encouragement and support.
6
List of abbreviations ANOVA
analysis of variance
bp
base pair (length unit of DNA and RNA)
BNF
biological nitrogen fixation
CI
confidence interval
DA and DF
discriminant analysis and its result, a discriminant function
DEA
denitrification enzyme activity
DNA
deoxyribonucleic acid – encodes genetic information
DNRA
dissimilatory nitrate reduction to ammonium
% d.w.
percentage of soil dry weight
G+C content
guanine-cytosine content of DNA
LH-PCR
amplicon length heterogeneity polymerase chain reaction
N; N2; and Nr
the element nitrogen; its inert molecular form dinitrogen; and reactive nitrogen present in fertilizers, organic matter, and biomass
NapA
large subunit of periplasmic nitrate reductase
NarG
large subunit of membrane-bound nitrate reductase
NH3 and NH+4
ammonia and ammonium
NirK
copper-containing nitrite reductase
NirS
cytochrome cd1 nitrite reductase
NorB
cytochrome b unit of NO reductase
NosZ
nitrous oxide reductase
NOx (NO and NO2)
nitrogen oxide (nitric oxide and nitrogen dioxide)
NO−2 and NO3−
nitrite and nitrate
N 2O
nitrous oxide
OTU
operational taxonomic unit
cPCR and qPCR
competitive PCR and quantitative PCR
rRNA
ribosomal ribonucleic acid, involved in protein synthesis
SOM
soil organic matter
Tg
teragram, i.e. 1012 grams or a million tonnes
yr
year
7
Table of contents PREFACE AND ACKNOWLEDGEMENTS
5
LIST OF ABBREVIATIONS
6
TABLE OF CONTENTS
7
INTRODUCTION
9
LITERATURE REVIEW OF TERRESTRIAL NITROGEN CYCLING AND INVESTIGATIVE METHODS
14
1. Theoretical background
14
1.1 Nitrogen in agriculture
14
1.1.1 Sources of nitrogen and its use
14
1.1.2 Global Nr fluxes
15
1.1.3 Nitrous oxide losses
17
1.2 The nitrogen cycle
17
1.2.1 Nitrification 1.2.2 The prevalence of denitrification and other 1.2.3 Products of denitrification 1.3 Soil bacterial communities
19 NO3−
transformations in soils
20 21 23
1.3.1 Phylogeny of soil bacteria
23
1.3.2 Bacterial communities in agricultural soil
25
1.3.3 The rhizosphere microbiome
26
1.4 Nitrifying bacteria
27
1.4.1 Nitrifier community structure
28
1.4.2 Drivers of N2O emissions from nitrifiers
29
1.5 The denitrifying bacteria
30
1.5.1 Beyond nitrate reduction
31
1.5.2 Nitrous oxide reductase
32
1.5.3 The denitrification product ratio
34
1.6 Rhizobia and legumes
34
1.7 Soil characteristics
36
1.7.1 Soil pH
37
1.7.2 Soil moisture and oxygen
38
1.7.3 Fertilization
39
2 Review of methods
41
8 2.1 Denitrifying enzyme activity
42
2.2 Molecular characterization of the microbial community
43
2.2.1 Clone libraries
44
2.2.2 PCR-based approaches
44
2.3 Ecological indices and community quantification
45
2.4 LH-PCR
47
2.4.1 Bioinformatics
48
2.4.2 Execution and sources of bias
48
2.5 Targeting functional genes for fingerprinting
49
2.5.1 NirK and nirS
50
2.5.2 Horizontal gene transfer and nosZ
51
EXPERIMENTAL WORK
52
3 Research objectives
52
4 Materials and methods
54
4.1 Experimental design
54
4.2 Analysis of soil
54
4.2.1 Soil pH and organic matter
54
4.2.2 Potential denitrification
54
4.3 LH-PCR
55
4.3.1 PCR conditions
55
4.3.2 Fingerprint visualization
56
4.3.3 In silico analysis
56
4.4 Statistical analysis
57
4.4.1 Ecological indices
57
4.4.2 Analysis of variance (ANOVA)
58
4.4.3 Repeated measures ANOVA
58
4.4.4 Two-way factorial ANOVA and multivariate ANOVA
59
4.4.5 Analysis of covariance
60
4.4.6 Correlations of peak data with soil measurements
61
5 Results 5.1 Soil properties
62 62
5.1.1 Soil pH and organic matter
62
5.1.2 Denitrification enzyme activity
63
5.2 Universal LH-PCR
64
5.2.1 Peak assignment
65
5.2.2 Treatment effects on non-selective LH-PCR profiles
66
5.2.3 Peak covariance with soil pH
70
5.2.4 Peak covariance with soil nitrate
70
5.2.5 Sampling date effect on non-selective LH-PCR community profiles
71
9 5.2.6 Interactions between treatment and sampling date
75
5.2.7 Soil moisture and N2O emissions
76
5.2.8 Peak heights and soil
NO3−
5.2.9 Peak heights and DEA 5.3 Gammaproteobacteria-specific LH-PCR
78 78 79
5.3.1 Peak assignment
79
5.3.2 Treatment effect on Gammaproteobacterial LH-PCR peaks
82
5.3.3 Variation in gammaproteobacterial LH-PCR fingerprints over sampling dates
86
5.3.4 Treatment by sampling date interaction
89
5.3.5 Emissions of N2O and γ-peak 5
90
5.3.6 Soil moisture and organic matter
91
5.3.7 Correlations between non-selective and gammaproteobacterial fingerprint peaks
92
5.4 Alpha- and Betaproteobacteria -specific LH-PCR
92
5.5 Phylogenetic structure of 16S rRNA gene length heterogeneity
93
6 Discussion 6.1 Fragment lengths from accession data and LH-PCR results
96 96
6.1.1 The short-fragment range
96
6.1.2 Mid-range fragments
97
6.1.3 The long fragments
97
6.1.4 Adequacy of bioinformatic tools
99
6.1.5 Gammaproetobacterial-specific LH-PCR
99
6.1.6 Relationships between different peaks 6.2 Cultivated vegetation, fertilization and the rhizosphere community
100 102
6.2.1 Rhizosphere and bulk soil
102
6.2.2 Galega orientalis
103
6.2.3 Brome grass
104
6.3 Denitrification activity and soil organic matter
105
6.4 Fingerprints and N2O emissions
107
6.5 Temporal changes
108
CONCLUSIONS
110
REFERENCES
112
9
Introduction Nitrogen (N) is one of the required elements for all forms of life. Individual cells require 2–20% as much N as they do carbon to incorporate into nucleic acids, proteins, and other essential components (Canfield et al., 2010). The 7th element of the periodic table originates from the conversion of carbon and oxygen as hydrogen burns in stars several times larger than the sun (Henry et al., 2000), and it is plentiful on our planet. The total N of the biosphere outweighs the four other essential elements combined – carbon, phosphorus, oxygen and sulphur. Despite its ubiquity, it mainly exists in the inert molecular form of dinitrogen (N2), making N the most poorly available and sought after essential element for higher organisms (Galloway et al., 2003). Plants and animals can consume N only in reduced and oxidized forms (i.e. as reactive N, Nr). In nature most Nr is produced from N2 by bacteria in soils and water bodies (Galloway et al., 2004). However, because of the importance of the element for supporting growth, people have intervened to maximize the amount of Nr reaching crops for as long as they have been working the soil. In the earliest days of the Palaeolithic revolution this was done by cultivating virgin lands, and later by amending soil with N-rich biomass and growing crops in rotation to avoid nutrient depletion. In the 20th century synthetic fertilizers changed fundamentally our relationship to nitrogen, and the way we produce food. As a result, the quantity of Nr in terrestrial and aquatic environments has been gradually rising, particularly since the Second World War. Nr is currently released into the biosphere at twice the rate during prehistoric times (Galloway et al., 2008). The use of N fertilizers in modern agriculture has brought food security to many, but not without leading to problems stemming from the dynamic nature of the element. A lot of the Nr produced at substantial cost is lost due to the opportunism of soil microbes, despite technological efforts to minimize the relevance of these poorly understood organisms. Nutrients do not follow a sealed pipeline from enrichment to the dinner plate, but are transported in cycles by geological and meteorological processes, and by organisms that have evolved to function in their niches with staggering effectiveness. The release of nutrients into living systems has a sustaining impact on their ecology. Most of the Nr applied to crops in mineral and organic form is not utilized, which has
10 consequences on ecological systems and public health (Oenema et al., 2009). When it leaches into groundwater, levels increase in drinking water and cause cancer and heart disease. This is relevant even to areas of the world with strong environmental policies – 20% of Europeans live in areas where surface water Nr exceeds the recommended level. In waterways it causes eutrophication, which can result in algal blooms and cause fishless dead zones (Cameron et al., 2013). The impacts cost the European Union somewhere in the range of €70–320 billion per year, which is more than double what N fertilizers add to farm incomes in the region (Sutton, et al., 2011). In order to restore the nitrogen cycle to functioning in a way that supports biodiversity instead of undermining it, the amount of Nr produced needs to be reduced while at the same time increasing food production. By way of setting targets, Rockström et al. (2009) have suggested a planetary boundary, i.e. a threshold within which ecosystems have the ability to absorb Nr manufactured by us. They propose this to be 35 Tg N yr−1, which is an ambitious standard, as the actual quantity released currently stands at 121 Tg N yr−1. Due to a growing global population and pressures from biofuel production, we are on course for a two- to threefold increase in N fertilizer use by midcentury using conventional farming methods. Many suggestions that could make this objective feasible have been made. Diets can be altered to include less N-demanding meat (Reay et al., 2012), especially if the industrialised world is to set an example to rising economies that want to enjoy the same standard of living as we do. Simultaneously, while in some parts of the world fields are awash with Nr, yields elsewhere are severely limited and threatened further by a changing climate. Vulnerable farmers need to hear about sustainable practices, but also urgently require access to inputs to achieve food security. Self-sufficiency can be improved in both industrialized and developing nations by growing crops with higher rates of biological nitrogen fixation (BNF), where soil bacteria convert N2 into assimilable nutrients. Sustainable use of Nr also means improvements to the N-use efficiency of food production, where increases of 50% are thought to be attainable through management changes, without compromising yields (Erisman et al., 2008). Central to achieving this is an understanding of N fluxes in agroecosystems. Soil is the workhorse of the nitrogen cycle, as it harbours bacteria and plants that work together to convert the most abundant gas in the atmosphere to cell material via BNF. In its reactive forms, Nr is found in topsoils at 2,000–12,000 kg N ha−1, which
11 equates to 0.1–0.6% of soil mass (Cameron et al., 2013). Of the essential elements for life, nitrogen is the most chemically diverse. It cannot form the polymers that carbon is capable of to provide the scaffolding of living tissues, but N exists in soil as nine different forms in seven oxidation states (Robertson & Groffman, 2007). In mineral forms, it undergoes rapid turnover in the soil, and the functionality of soil processes is one measure of a healthy soil. Fertilizing soil with Nr results in a transient spike in dissolved nutrient availability, but in the longer term the Nr that can be assimilated by plants is determined by microbial processes of N release from organic matter (i.e. mineralization) and uptake (i.e. immobilization). The resilience of these processes is thought to be due to a high level of microbial diversity that is more stable when subject to disturbance (Hartmann & Widmer, 2006, Entry et al., 2008). Essential functions such as decomposition and nutrient cycling can be vulnerable when a lower richness of taxa is present (Braker et al., 2012). Natural and anthropogenic stresses can reduce the competitiveness of some taxa but an ecosystem that is resistant will be able to absorb these losses (Entry et al., 2008). By definition,a healthy soil, while providing protection to the crop against physical and biotic stresses, is also constantly returning to a state of equilibrium, with most of the nutrients bound in biomass. Modern agriculture resists this equilibrium, by tilling the soil and adding nutrients. It is a typical feature of fertilizers that their effect is shortlived, and that they are only momentarily available to the plant. The optimal timing and amount of fertilization required to minimize losses depends on the type of crop (Hofstra & Bouwman, 2005). In the case of nitrogen, additives are included to delay this process, but Nr is ultimately attractive to microbes as a source of energy. To generalize, the soil bacterial community is self-sufficient for nitrogen, whereas plants are self-sufficient for carbon. This forms the basis of highly evolved mutualistic relationships between plants and soil bacteria. However, in conditions where more ideal energy sources are scarce, microbes exploit Nr for respiration, rather than for building cellular components. As a result of microbial respiration, 0–25% of applied Nr is expended and released into the atmosphere as N2 or nitrous oxide (N2O). The IPCC estimates that 1.25% of applied N fertilizer is emitted as N2O, but this appears to be an underestimate (Philippot et al., 2007). N2O persists in the atmosphere for 114 years and is the third most important greenhouse gas (GHG) in terms of total radiative forcing after CO2 and CH4. It is the main cause of global warming originating directly from agricultural production. When it ultimately is degraded, it is then deposited elsewhere in rainwater, leading to
12 further eutrophication. Thus, the scope of the N cycle is global, and disturbances can be diffuse, not limited to a given point source. In order to benefit from the range of ecosystems services provided by healthy soil, we need to understand the interactions between the microbial community, vegetation, and the abiotic features of the soil that vary greatly by region and climate. The confluence of these players is found at the rhizosphere. This is the volume of soil surrounding plant roots that is influenced by the roots physically, chemically and biologically (Richardson et al., 2009a). It has often been described as “one of the most complex ecosystems on Earth” (Mendes et al., 2013). In agricultural systems it is apparent, from studies of soil under conventional systems, that plants exert an influence on the soil microbial community that even varies between cultivars of the same crop species. The link between above- and belowground diversity is so strong, that in natural systems the soil microbial community can directly affect plant community composition. The bacterial diversity of cultivated soils is generally lower than in natural systems, due to the prevalence of growing individual crops in isolation (monocropping) and the disturbing effects of tillage, fertilization and agrochemicals. Nevertheless, a gram of arable soil still contains several hundreds of species, while the number in a gram of forest soil is typically an order of magnitude higher (Hartmann & Widmer, 2006). To characterize such astonishing biodiversity, found in any given garden patch, one needs to start at the top. The complexity of soil environment means most bacteria do not grow in the laboratory, at least not without some effort. It was a lesser-known microbiologist called H. Joel Conn that observed in 1918 that only 1.5 to 10 % of soil bacteria were detectable on artificial media (Janssen, 2006). Since then, research has not only been unsuccessful at raising this estimate, but it has in fact decreased. While our knowledge of bacterial diversity has improved through molecular techniques, traditional cultivation has yet to succeed in purifying 1% of the calculated number of species. There are two broad approaches to studying the genetic diversity of soil bacteria. Total DNA is extracted, and the 16S rRNA gene is targeted for either cloning and sequencing or one of the various fingerprinting techniques that have been developed. Sequencing can be followed with the identification of the bacteria to a degree of precision, but is only practical for a small random sample of the total community. To obtain a picture of the community composition the gene is amplified and manipulated in to form a visual representation that can then be compared in similarity to other bacterial
13 communities. Some of these methods are more easily quantified than others, and some lend themselves better to corroboration with sequence data. Ultimately the techniques chosen will be based on the expected diversity of the sample material, how much we already know about the environment being studied, and how we want to apply the knowledge gained. This is not to say that taxonomy is losing its relevance to microbial ecology. When a new bacterial strain is selected for further study, several factors are considered. We are interested in characterizing an organism not purely for its novel genome, but for an understanding of how it functions in its environment (i.e. its phenotype). Closely related strains may be assigned to different genera if they are ecologically very distinct. It is also conceivable that entire bacterial families of loosely related bacteria exist, where the majority of strains display a high level of phenotypic similarity despite genetic dissimilarities. This is possible because random genetic changes occur independent of selective pressure, a process known as genetic drift. When observing community fingerprints it is important to remember therefore that genetic diversity does necessarily presume phenotypic diversity. In relating microbial diversity to function, the taxonomy of characterized species serves as a stepping-stone.
14
Literature review of terrestrial nitrogen cycling and investigative methods In this literature review I recount the history of agriculture’s impact on the nitrogen cycle, and account for the bacteria behind the relevant soil processes. I then describe the principles of the methods used in this research, and how they can be extended to test further hypotheses.
1. Theoretical background 1.1 Nitrogen in agriculture The human impact on the N balance began at least 6500 years ago with the cultivation of legume crops. Members of this family of plants show levels of BNF of around an order of magnitude greater than other crops through mutual relationships with soil bacteria called rhizobia. The contribution of legume crops to increases in BNF is designated as cultivation-induced BNF (C-BNF). Rice and sugar cane also achieve greater than average rates of BNF, but the diversity and high protein content of legumes have made them important crops for all climates and conditions. The N content of soil would traditionally be maintained by rotating legumes with other crops such as grains (Galloway et al., 2004). By improving the availability of Nr in N-limited soil, the effects of plant-microbe competition for the nutrient (Fischer et al., 2013) would be diminished, leading to better yields. Legumes increase the amount of new Nr generated, but conversion of land to farmland also increases the rate of Nr cycling due to nutrient mining. In Illinois, USA prairies would release 30 kg N ha-1 yr-1 when first ploughed for growing grain. For many decades even before intensive cultivation methods, 85% of yields were fed to livestock, allowing the N to be depleted from the soil (Davidson, 2009). Histosols are particularly rich in soil organic matter (SOM), and can release considerable amounts of Nr when initially cultivated. Crop residues and livestock waste can have the same effect when deposited directly or spread onto fields (US EPA, 2011). 1.1.1 Sources of nitrogen and its use Until the 20th century, most new Nr generated came from BNF, and the Nr available in the biosphere was balanced by another microorganismal process that returns Nr to atmospheric N2, called denitrification (Galloway et al., 2003). During the industrial era,
15 the balance was largely maintained, and the Nr that went into agriculture and manufacturing processes originated from finite geological deposits that supplied only 0.2 Tg N yr−1 (Galloway et al., 2004). This changed in 1908 with the advent of the Haber-Bosch process, an economical way to produce ammonia (NH3) directly from N2 (Erisman et al., 2008). Atmospheric N2 has a highly stable triple bond, so its conversion through BNF into NH3 is energy demanding. However, the process has a variety of uses for fertilizers, fine chemicals and explosives, and since 1960 its use has increased dramatically (Galloway et al., 2003). Of the NH3 produced this way today, 80% is used to produce fertilizers (Figure 1), and this huge increase in Nr availability is responsible for 30–50% of the crop-yield increases seen over the last century. As a result, the number of people supported by a hectare of arable land has more than doubled over the last century (Erisman et al., 2008). From total global energy consumption, 1–2 % goes to N fertilizer production and this input can represent over half of total energy expenditure for non-legume crops (Jensen et al., 2012). Almost half of the world’s population depends on N fertilizers for food. With the benefit of hindsight one can see how the proliferation of subsidized Nr has led to a dependence on its excessive use to maintain production levels. This is particularly evident in the N-use efficiencies of staple crops. Coinciding with a 7-fold increase in global fertilizer use, the N-use efficiency of cereals went from ∼80% in 1960 to ∼30% in 2000 (Philippot et al., 2007). In 2005, 100 Tg N from the Haber-Bosch process was applied to the world’s fields, but only 17 Tg N reached the end consumer as food (Figure 1; Erisman et al., 2008). This is no longer a first world problem – whereas in 1970 N-use efficiencies were greater in developing countries than in industrialized countries, the situation has reversed due to increased fertilizer use in rising economies such as China (Bouwman et al., 2009). Crop uptakes are intrinsically higher in warmer climates with longer growing seasons, so the potential for a large soil N surplus and its environmental implications is greater than in the global north. Whereas N fertilizer use has decreased in high-income countries since the 1980’s, the majority of N fertilizers are now used in low-income countries (Crews & Peoples, 2004). 1.1.2 Global Nr fluxes Our dependency on C-BNF has decreased since 1860, when it provided 15 Tg N yr−1. The rate of C-BNF rose in absolute terms to 33 Tg N in 2000, but this is now a relatively minor input, comparable to the 25 Tg N in the form of NO and NO2 (NOx)
16 coming from fossil fuel combustion alone in the same year (Galloway et al., 2003). Oxides of nitrogen are produced during combustion of fossil fuels, either from oxidation of atmospheric N2 or the release of fossil Nr (Galloway et al., 2004). This rise hardly reflects the expansion of agricultural production that has occurred in response to an increasing human population. While C-BNF has more than doubled, total natural nitrogen fixation decreased from 120 Tg N yr−1 in 1860 to 107 Tg N yr−1 (Galloway et al., 2004). This could be due to decreased nitrogen demand from plants that can regulate BNF in the rhizosphere.
Deposition[a]: 35 Tg N
Manure[d,a]: 36 – 92 Tg N
Global Nr inputs Lightning[c]: 5 Tg Nox-N
Natural terrestrial BNF[c]: 58 Tg N Marine BNF[c]: 140 Tg N
Haber-Bosch[c]: 120 Tg N Of which applied fertilizers[a,c]: 83–100 Tg N
Global crop and livestock production Nr inputs[a]: 248 Tg N
Denitrification[a]
Volatilization[a,e]: 24–33 Tg NH3-N
Fossil fuels[c]: 30 Tg Nox-N
48 Tg N2-N
7 Tg N2O-N
aBouwman"et"al,"201
2 Tg NO-N
b"Erisman"2008"
cFowler"et"al,"2013"
Withdrawal:
110[a]
Tg N
dStehfest"&"Bouwma
eBouwman"et"al."201
Cultivation-related BNF[a,c]: 39–60 Tg N
Leaching/runoff[a]: 57 Tg N
Consumed food[b]: 17 Tg N
Figure 1 Annual Nr generation and fluxes in and out of agricultural systems at the global level. N2O emissions include indirect sources originating from agricultural soil Nr. References: [a] Bouwman et al., 2011, [b] Erisman et al., 2008, [c] Fowler et al., 2013, [d] Stehfest & Bouwman, 2006, [e] Bouwman et al., 2013.
The movement of Nr from agricultural sources to the natural environment is not limited to hydrological downstream transfer, but also occurs through emissions. In the early 1990’s over half the Nr created by humans was released as NH3 and NOx (Galloway et al., 2004). These compounds can be transferred to other areas dissolved in rainfall (wet deposition), or as gases and attached to particulate matter (dry deposition; Butterbach-Bahl et al., 2011), which takes place within a matter of hours or days (Denman et al., 2007). Transport and industry produce most anthropogenic NOx (Figure 1) and atmospheric N deposition into terrestrial ecosystems is occurring at 20-fold the
17 natural rate (Erisman et al., 2008). In 2000 the global average rate of N deposition onto agricultural land was 35 Tg yr−1 (Figure 1). In Europe rates range 3–30 kg N ha−1 yr−1 in areas of short vegetation such as agricultural land. Dry deposition mainly affects closed areas, effectively doubling to tripling deposition inputs for environments such as forests (Butterbach-Bahl et al., 2011). 1.1.3 Nitrous oxide losses Atmospheric N2O has been increasing for 250 years (Sanford et al., 2012) and by the 1990s emissions of N2O reached 40–50% greater than pre-industrial levels (Denman et al., 2007). Sources include industry and biomass burning, but soils produce most of the N2O entering the atmosphere. The localization of anthropogenic N2O emission is very different from natural sources, which occur mainly between 25° S and 25° N. Human activity, on the other hand, has led to steep increases in emissions between 20° N and 40° N, measured over the last 30 years (Zaehle et al., 2011). Agricultural soils directly produce 4.2 Tg N2O-N yr−1, which comprises 70% of anthropogenic emissions of the gas. This is exceeded by the 6 Tg N2O-N yr−1 emitted from natural soils and rivers, though some of this is derived from Nr leached and deposited from agricultural soils (Bouwman et al., 2011, Thomson et al., 2012, Hartmann et al., 2013). Since the acceptance of the Kyoto Protocol in 1997, N2O emissions from non-biological sources have decreased (Thomson et al., 2012). The agricultural contribution has however remained unaddressed. There is a potential for soils to contribute to uncontrollable climate change by a positive feedback mechanism, as both elevated CO2 levels and elevated temperatures have been shown to increase N2O emissions (Brown et al., 2012). The effects of flooding due to increased incidence of extreme weather are also of a particular concern, as this could also lead to increased N2O emissions (Hansen et al., 2014). It has been noted that measures that reduce N2O produced from agricultural activities are likely to lead to reductions in GHG emissions associated with fertilizer production (Bates et al., 2008). A study using the
15
N tracer method found that 1.8–
2.8% NO3− added to land is released as N2O, and figures of up to 2.5% are supported by other research, which is considerably higher than IPCC estimates (Čuhel et al., 2010). 1.2 The nitrogen cycle The nitrogen cycle has been extensively studied due to its ecological importance, and has even been called a “model system in ecology” by Philippot et al. (2009). Figure 2 illustrates the major types of N transformations that occur in agricultural soil. Plants can
18 take up N as simple organic compounds such as amino acids. These are released when bacteria break down organic detritus, a process broadly known as mineralization (Robertson & Groffman, 2007). In intensive agricultural systems, however, the most widely applied form of Nr is urea, which is hydrolysed to NH3 in soil (Jensen et al., 2012). Plants assimilate this in the protonated state as ammonium (NH+4 ), which forms more readily at low pH. Cameron et al. (2013) provide a comprehensive review of the fate of fertilizer NH+4. In the soil, much of it remains bound strongly to clay and SOM, which have a net negative charge. Conversely, unprotonated NH3 is prone to volatilisation, anywhere in the range of 0–65% of it is volatilised as NH3. Globally, NH3 volatilization is the third most important mechanism by which Nr is lost from agricultural soils (Figure 1 and Figure 2), after leaching and denitrification.
NO3−"
N2 NH3"
N2O NO N2
N2
NH4+"
Volatilization Release of intermediates Nitrous oxide reduction nosZ
N2O norB Nitric oxide reduction
NO Nitrite nirK/ reduction nirS
Fertilization / N-deposition NO3−"
NH3
Soil
Fertilization / N-deposition
Nitrogen nif fixation Anammox
Atmosphere
NH4+
Immobilization Nitrification amo Biomass as proteins, NH2OH Mineralization nucleic hao narG/ acids, etc. − napA NO2" Uptake Nitrate nxr Soil organic N reduction NO3−" Immobilization Leaching Organic N Leaching Ground-water
Figure 2 The main processes of the bacterially mediated terrestrial nitrogen cycle. Dashed lines indicate physical transfer of N between phases and steps involved in denitrification are shown with a bold border. Genes and operons encoding some of the enzymes or enzymatic subunits involved in the cycle are encircled, and include nitrogenase (nif), ammonium monooxygenase (amo), hydroxylamine oxidoreductase (hao), nitrite oxidoreductase (nxr), nitrate reductase (narG/napA), nitrite reductase (nirK/nirS), nitric oxide reductase (norB), and nitrous oxide reductase (nosZ). The contributions of lightning, geological processes, and pollution from combustion are not depicted. Adapted from Robertson & Groffman (2007), Canfield et al. (2010), and Pilegaard (2013).
Agriculture causes 50% of total NH3 volatilisation, most of which is deposited elsewhere, leading to remote acidification and eutrophication. A small proportion is
19 oxidized in the atmosphere to yield N2O (Denman et al., 2007). Volatilisation can also occur at a low pH, due to the dissociation stoichiometry of urea. The immediate hydrolysis product is ammonium carbonate ((NH4)2CO3), which dissociates to yield NH3, NH+4 , CO2, and hydroxide (OH− ) in equal amounts. The OH− can cause a shortterm increase in pH, which promotes further NH3 formation. Soils with a high cation exchange capacity (CEC) can have lower rates of NH3 volatilisation because pH is buffered against the effect of urea hydrolysis, and a higher rate of NH+4 removal from soil solution means less is available for volatilization. Bioavailable NH3 that persists in the soil but is not assimilated can be returned to the form of N2 by the microbial processes of nitrification and denitrification (Cameron et al., 2013). 1.2.1 Nitrification The first process that can lead to the loss of mineral nitrogen from the agroecosystem is nitrification, which is mostly due to the activity of two groups of closely-related chemoautotrophs that oxidize NH3 to nitrate (NO3− ) for cell growth. A smaller proportion of nitrification is due to heterotrophs, where the reactions are not coupled to ATP production. Bacterial heterotrophic nitrifiers are similar to their autotrophic counterparts, but mainly degrade non-polar organic compounds, while their ability to oxidize NH3 appears to be incidental. Fungal nitrifiers, however, oxidize amines and amides from compounds in animal manure, crop residues, and SOM (Robertson & Groffman, 2007). In natural ecosystems nitrification rates are low and nitrogen sources are heterogeneous, but agriculture fertilizer inputs enable higher rates (Robinson et al., 2011). Nitrification is also promoted by other practices such as tillage, burning, and clear-cutting – essentially anything that increases soil NH+4 above what is required for plant growth or improves O2 availability. Nitrifiers are slow growing and poor competitors for NH+4 (Robertson & Groffman, 2007), but are active at a range of soil pH values ≥ 4 (Mørkved et al., 2007). Most plants require some nitrification for optimal growth, but there are several reasons why NO3− formation is avoided in agricultural production. Firstly, a sidereaction of NH+4 oxidation leads to the production of N2O (Figure 2). However, the main concern is the loss of Nr due to eventual denitrification and leaching. Organic nitrogen can also be leached, but this occurs at much lower rates than mineral losses (Pilegaard, 2013). The favoured source of N for most crops is NO3− , and the reason for this may be related to its mobility in soil, which is 5 to 10 times greater than that of
20 other N sources. This property also means it is susceptible to leaching, i.e. loss from the agroecosystem through groundwater (Robinson et al., 2011). Prevention of nitrification in cropping systems essentially amounts to the use of chemical inhibitors when fertilizing (Robertson & Groffman, 2007) and some plants, including a few crop species, have the same strategy, depositing nitrification inhibitors in the rhizosphere and achieving higher soil NH+4 retention (Richardson et al., 2009b). At least one synthetic nitrification inhibitor, dicyandiamide, can itself be leached into surface waters. This may raise the amount of NH+4 downstream with potential toxic effects that can impact biodiversity (Smith & Schallenberg, 2013). NO3− is also assimilated into microbial biomass, and can additionally act as an electron sink for denitrifiers. 1.2.2 The prevalence of denitrification and other NO3− transformations in soils Denitrification is responsible for more N-removal from soil than any other process, leading to the loss of 10–40% of applied fertilizer N in the form of N2 alone (Galloway et al., 2004). Conservative estimates are that in 1970–2000, 58–79 Tg N yr−1 was lost from terrestrial systems by denitrification (Bouwman et al., 2013), and 57 Tg N from agricultural soils in 2000 (Bouwman et al., 2011). An earlier study estimated that denitrification removed 87 Tg N in 1995 from non-leguminous agricultural areas alone (Hofstra & Bouwman, 2005). Because denitrification counters eutrophication, it is a boon to biodiversity and a desirable process in riparian zones and wastewater treatment. Conversely, in agroecosystems denitrification represents an economic loss, as it is a major component of fertilizer-use inefficiency, causing emissions on the field scale of 15–200 kg N ha−1 yr−1 (Hofstra & Bouwman, 2005). In intensively cultivated soils high rates of denitrification are a direct consequence of nitrification in the fertilizeraccelerated nitrogen cycle. Denitrification differs fundamentally from NO3− reduction in plants, where it is assimilatory, and NH+4 -N is immobilized for cellular components. Microorganisms link denitrification to the generation of a proton-motive force. It is a respiratory process whereby NO−3 and its reduction products are used as electron acceptors. NO−3 has a high reduction potential, so many bacteria readily denitrify where organic substrates are available and O2 pressure is low. Furthermore, anaerobic microsites are present even in well-aerated soils, so the process is ubiquitous and the reaction cascade continues as the product of each reaction is reduced through lower oxidation states to N2 (Figure 2; Galloway et al., 2003). The reactions are not all closely coupled, so the products of
21 incomplete denitrification can be released into the atmosphere. Critically, the gaseous intermediates of denitrification are not readily returned to biologically useful mineral forms, and they have environmental consequences. The microbial reduction of NO−3 is not limited to denitrification, and produces NO−2 . This is the actual substrate for denitrification sensu stricto, but a proportion of soil NO−2
may be re-oxidized by
nitrifiers. In addition to denitrification and assimilation of NO3− , there exists also the dissimilatory reduction of NO3− to NH+4 , known as ammonification. This reaction consumes a lot of electrons, and is believed to be limited to anaerobic environments rich in organic matter such as wetlands and marine sediments, where reductants are more abundant than oxidants (Mohan & Cole, 2007). In agroecosystems it would be a desirable process in the sense that it leads to Nr retention in the soil (Pilegaard, 2013), rather than its inevitable loss once the denitrification pathway is initiated. Some N2O and N2 emitted from soils result from microbial codenitrification, where NO−2 is reduced using other nitrogen compounds, yielding the same intermediates as denitrification. In grassland soils up to 92% of N2 emitted comes from fungal codenitrification. The process whereby N2 is the sole product of coupled NO−2 reduction and NH+4 oxidation is distinct from codenitrification, and is usually referred to as anaerobic ammonium oxidation (anammox). The significance of anammox is thought to be restricted to sediments, but it cannot be ruled out from the list of major soil processes, as it is hard to distinguish from codenitrification when using 15N isotope-pairing techniques. One study of agricultural soil detected at least 200 times more copies of denitrification gene nosZ than copies of the anammox gene hzo (Long et al., 2013). 1.2.3 Products of denitrification The first intermediate of denitrification is NO, which is not a persistent pollutant but reacts in the troposphere to form the greenhouse gas ozone (O3), and NO2 (Pilegaard, 2013). Plants cannot use NO for building cell material, but it is a signal molecule for root growth and even the formation of nodules required for rhizobial BNF in legumes. Some crops have been tested for the effect of bacterial NO and it has been suggested that denitrification can affect plant development in this way, as well as by stimulating seed germination and controlling pathogens (Richardson et al., 2009b). More attention is generally paid to the second intermediate of denitrification. The atmospheric concentration of N2O was fairly constant since at least 1000 AD until
22 industrialisation. This suggests that without anthropogenic disturbance, denitrification and sedimentation would balance BNF to maintain a constant level of Nr in the biosphere (Galloway et al., 2003). Of the N2O produced, 25% accumulates in the troposphere, where it acts as a greenhouse gas. The remainder reaches the stratosphere (Galloway et al., 2004), reacting with O2 to yield NO, which degrades the ozone layer at this altitude (Čuhel et al., 2010). Once emitted, N2O persists in the atmosphere for about 114 years (Thomson et al., 2012) and has increased from 270 parts per billion (ppb) in pre-industrial times to 320 ppb (Cameron et al., 2013). This is three degrees of magnitude less than atmospheric CO2 abundance, but is significant because the global warming potential of N2O is around 310 times greater (US EPA, 2011). In terms of CO2 equivalents, this denitrification product accounts for 10% of GHG emissions (Thomson et al., 2012). Most Nr that enters the denitrification pathway is reduced completely, and global terrestrial denitrification emissions of N2 increased from 36 Tg in 1900 to 68 Tg in 2000 (Bouwman et al., 2013). In the early 1990s anthropogenic NOx was emitted in the order of ∼36.2 Tg N yr−1, a third of which came from food production (Galloway et al., 2004). Of this, 1.8–12 Tg NO-N yr−1 was emitted directly from soils (Pilegaard, 2013). In comparison, terrestrial emissions of N2O increased from 8.8 Tg N in 1900 to 11.3 Tg N in 2000, and may grow to 14.2 Tg N in 2050 (Bouwman et al., 2013). Higher Nr status often leads to higher levels of photosynthesis, particularly in temperate forests, and the resulting carbon sequestration is estimated to have a climate cooling effect of 96 mW m−2. This benefit is offset by the effect of anthropogenic N2O, which has a radiative forcing of 125 mW m−2 (Zaehle et al., 2011). Affected forests are located in industrialized areas with high population pressures. It therefore remains to be seen whether net carbon sequestration can be maintained as bioenergy is increasingly exploited to replace waning fossil fuel reserves. A more long-sighted way to reduce radiative forcing would be to increase the N2O-sink activity of soils – that is to say manage soils in a way that supports N2O consumption, rather than production. Carbon sequestration also has a role to play, but it leads to increased reservoirs of organic matter that can be exploited for energy production, negating any accrued benefit to the climate. Sequestration of N2O is more final because the end product is N2 (Desloover et al., 2014), which needs to be reduced again at great energetic cost in order to be useful. The net consumption of N2O by soil microbes is frequently recorded, even in
23 agricultural soils over short periods, but the phenomenon has not been well described (Chapuis-Lardy et al., 2007). 1.3 Soil bacterial communities Most soil bacterial diversity has only been characterized at a shallow phylogenetic level. One meta-study of mainly forest and pasture soils (Janssen, 2006) found that only 10– 21% of bacterial 16S rRNA gene sequences amplified from soil could be identified by genus. Because genera often represent phenotypic units, the author concluded that our capacity to link overall genetic diversity to functional diversity is still poor. Conversely, Fierer et al. (2012) reported that when looking at very different soil communities, taxonomic and phylogenetic structure is a good predictor of functional diversity, even if this does not necessarily hold for functional genes considered individually. Microbial functions can be classified as either broad processes, carried out by very diverse groups, and narrow processes, only performed by a few specialists. 1.3.1 Phylogeny of soil bacteria Cultivation biases have allowed us to develop our understanding of a few bacterial taxa, while others have evaded discovery. The methods used to isolate bacteria lead to an over-representation of aerobic heterotrophs (Janssen, 2006), whereas the physiology of uncultured bacteria is presumably much broader. Several phyla lack pure-cultured representatives altogether, but even a phylum as well studied as the Proteobacteria includes a great deal of undescribed diversity. Furthermore, members of some novel proteobacterial families and orders have been described but not yet named, posing a challenge when interpreting sequence libraries. The phylum Bacteroidetes was found to include the most culturable soil bacteria, and 34–62% of sequences assigned to this phylum could be identified at the generic level. Figure 3 shows the distribution of taxonomic assignments of 16S rRNA gene sequences typically isolated from soil (Janssen, 2006). The relative contributions of major phyla to the bacterial community would come as a surprise to anybody who is only familiar with soil bacteria that have been studied by culture methods. In addition to the presence of novel phyla, most striking is perhaps the low proportion of sequences originating from Firmicutes (1.8%). Colonies of this phylum grow abundantly when applying the most probable number (MPN) method to enumerate soil bacteria. Nevertheless, Yin et al. (2010) found that agricultural soil Firmicutes 16S rDNA V3
24 sequences showed less similarity to known sequences than those of the other major soil phyla. Commonly used DNA-extraction methods may be inefficient at sampling the thick-walled spores characteristic of this phylum (Janssen, 2006, Yin et al., 2010). A study of 88 non-agricultural soils (Lauber et al., 2009) found the same five dominant phyla to comprise > 90% of the sequences. All phyla showed the same level diversity within a given sample, but the Acidobacteria and Actinobacteria had a greater number of phylotypes than Proteobacteria and Bacteroidetes over all the sampled soils. That is to say the beta-diversity was greater, which could be due to different lifestrategies. The Acidobacteria are generally oligotrophic, so may not thrive in an amended environment rich with plant-derived carbon, whereas there are more Bacteroidetes in copiotrophic conditions (Fierer et al., 2007). Actinobacterial subclass Rubrobacteridae are common in a range of soils, particularly those contaminated with heavy metals, and in one study (Mummey et al., 2006) were the most common group found within soil microaggregates. Twenty-two strains have been isolated from soil (Janssen, 2006), all of them aerobic heterotrophs. A study of 16 diverse soils that included more extreme environments found that the five main phyla also dominate these natural systems, though Cyanobacteria are prominent in some desert soils (Fierer et al., 2012).
25
100 Other phyla
Unassigned (2.4%) Others (5.2%)
Chloroflexi (3.2%)
Bacteroidetes (5%) 80
Subdivision 4 (0.2%) Subdivision 3 (0.5%) Verrucomicrobiae (0.03%)
Sphingobacteria (4.6%)
Spartobacteria (6.3%)
Gammaproteobacteria (8.1%)
60
Betaproteobacteria (10.0%)
Gemmatimonadetes (2.0%) Planctomycetes (2.0%) Firmicutes (1.8%) Flavobacteria (0.4%)
Verrucomicrobia (7%) Epsilonproteobacteria (0.04%) Deltaproteobacteria (2.3%)
Proteobacteria (39%) Ÿ 10-77% of soil 16S
rRNA libraries
Ÿ 528 named and
40
Acidimicrobidae (2.4%)
Ÿ At most, only 19-36%
Rubrobacteridae (5.6%)
Actinobacteria (13%)
Actinobacteridae (4.7%)
20 Acidobacteria (20%)
Subdivision 6 (4.5%)
Ÿ 5-40% of soil 16S
Subdivision 4 (7.7%)
rRNA libraries Ÿ 3 named and described genera
0
described genera
Alphaproteobacteria (18.8%)
of soil sequences can be assigned to a genus.
Ÿ 0-34% of soil 16S rRNA libraries Ÿ 158 genera in Actinobacteridae Ÿ 2 genera in Rubrobacteridae
Subdivision 7 (1.5%) Subdivision 5 (0.4%)
Subdivision 1 (3.3%)
Subdivision 3 (1.8%) Subdivision 2 (0.5%)
Figure 3 The composition of a typical soil microbial community, derived from a meta-study of 21 libraries of a total 2,920 sequences (Janssen, 2006). Average contributions of phyla and subgroups to the community are shown as percentages.
1.3.2 Bacterial communities in agricultural soil Cultivated soil will differ from the scenario in Figure 3 as it is generally richer in nutrients, and subject to mechanical disturbances, although these have a greater impact on soil fungi than bacteria. A study of arable soil in Switzerland (Hartmann & Widmer, 2006) found that the same four phyla were dominant as indicated in Figure 3, but that
26 they comprised a greater proportion of the diversity at 83% of cloned sequences. Furthermore, Actinobacteria were most abundant (35–39%), and there was less Verrucomicrobia, though differences were detectable between treatments (3–6%). It seems evident that, in comparison to some phylogenetically rich soils, the lower diversity of agricultural soil bacteria is visible at the phylum level. Agricultural soil is often left bare between harvest and the following growth season, which may impact the bacterial community in the long term. Verrucomicrobia and Acidobacteria are associated with bulk soils rather than rhizosphere soils, particularly Acidobacteria where C mineralization is low (Fierer et al., 2007). Nunes de Rocha et al. (2013) found that Acidobacteria subdivisions 1 and 3 were generally more abundant in bulk soil and leek rhizosphere than in grass and potato rhizosphere soils. The numbers of subdivision 6 strains were also substantial, but the rhizosphere effect was diminished after June. An increase in Alpha-, Beta-, and Gammaproteobacteria was also observed in rhizosphere soil relative to bulk soil. In conclusion, to detect differences in agricultural soil bacterial communities one could do worse than to focus on the quantitative changes in Acidobacteria, Proteobacteria, Actinobacteria, and Verrucomicrobia. The 5th most abundant phylum (Bacteroidetes) numbered only 2–3% in the study by Hartmann & Widmer (2006), so changes in the lesser phyla would be difficult to detect with typical fingerprinting methods, and would be best studied using approaches that specifically target these groups. 1.3.3 The rhizosphere microbiome Most bacteria in the rhizosphere, including denitrifiers, are organotrophic, and much of the SOM they use for energy is plant-derived. Plants exude large quantities of carbon from their roots, presumably in order to exert a selective pressure on the rhizomicrobial community. During the vegetation period inputs from rhizodeposition and root turnover can comprise up to 20% of photosynthate carbon (Philippot et al., 2007), or up to 40% of the carbon received by the soil (Richardson et al., 2009b). However, some researchers maintain that carbon exudation could be accidental, as this has not been experimentally ruled out. The picture is complicated by the fact that some plant pathogens, such as the oomycete Phytophthora sojae have also evolved to respond to phytogenic compounds. Nevertheless, bacterial growth in the soil is limited by carbon availability (Mendes et al., 2013), to which nitrogen is secondary. Nr is not typically
27 supplemented by plants except in leaf litter, but when soil is fertilized with N, a major effect on the microbial community is that bacteria can grow more efficiently than fungi, given their higher demand for N (Robertson & Groffman, 2007). High bacterial respiration in the rhizosphere can deplete O2 and several papers have reported significantly higher rates of denitrification in proximity to plant roots (Veresoglou et al., 2012). Bacteria in the rhizosphere support plant growth by providing resistance against biotic stress from soilborne pathogens, and abiotic stress from extreme conditions. The soils of wild and cultivated plants are also a reservoir for opportunistic human pathogens that we are only beginning to understand. One study of sugar beet seedlings showed that soil that supported the growth of plant pathogen Rhizoctonia solani also harboured more OTUs associated with potential human pathogens (Mendes et al., 2013). The body of evidence justifies the development for crop management practices that support mechanisms that can complement crop improvement to improve yields, as well as safeguard public health.
1.4 Nitrifying bacteria The soil-phase of the nitrogen cycle begins with NH3. This is the product of BNF and SOM mineralization, and is rapidly formed from the initial breakdown of most N fertilizers. With a high reduction potential, NH3 provides energy for two groups of autotrophic microbes, the ammonia oxidizing archaea (AOA) and the ammonia oxidizing bacteria (AOB). Among the bacterial members, the first and rate-limiting step for nitrification is carried out by nitrosobacteria. This group consumes NH+4 and O2 to yield NO−2 by using an enzyme called ammonium monooxygenase (AMO; Fischer et al., 2013). The second reaction is the oxidation of NO−2 to NO3−, which is carried out by nitrobacteria, using nitrite oxidoreductase (NXR). In addition to providing the substrate for denitrification, this reaction, though not well understood, also yields NO and N2O as side products. The nitrosobacteria have general names beginning with “Nitroso-” and are mostly Betaproteobacteria in the family Nitrosomonadaceae. General names for nitrobacteria begin with “Nitro-”, and are found mostly in Alphaproteobacteria and the phylum Nitrospira, with some gammaproteobacterial and possibly deltaproteobacterial members
28 (Bock & Wagner, 2013). Some heterotrophic bacteria and fungi can also nitrify using alternative pathways that are believed to be significant in forest soils. In the organic pathway, nitrification is coupled with mineralization, so that the amine (R-NH2) moiety is reduced to a nitro group (R-NO2), which is cleaved and oxidized to NO3− (Robertson & Groffman, 2007). 1.4.1 Nitrifier community structure The gene encoding subunit A of ammonia monooxygenase (amoA) exists as separate bacterial and archaeal orthologs. Sequences encoding archaeal amoA have been detected 300 times more than bacterial amoA in an agricultural soil (Fischer et al., 2013), and in general AOA are inferred to be more abundant than AOB, carrying out most nitrification activity. However, this may be only partly true in some cases, as AOB have a 10-fold higher specific activity than AOA (Prosser & Nicol, 2012). Exceptions include one study where in soil microcosms incubated at temperatures between −4 and 5°C, each form of the gene was enumerated at slightly over 106 copies g−1 dry soil. In both cases there was a positive correlation of abundance with temperature (Wertz et al., 2013). Nevertheless, the drivers for ammonium oxidizer (AO) community structure at the domain level are not understood, apart from nitrification potentially being exclusively due to AOA activity at soil pH < 5.5. Physiological studies have been hampered by difficulties in cultivating these groups. Three soil AO genomes have been sequenced, but it is not known how representative the studied strains are. Known AOA belong to Thaumarchaeota. Studies of the kinetics have shown bacterial members to have higher saturation constants and tolerance for NH3. This, and the finding that root exudates diminish AOA number seem to support the hypothesis that AOB are better adapted to the rhizosphere and fertilized soil. Nitrification inhibitors may obscure our understanding of this, as they are widely used and can affect the AOA:AOB ratio. On the other hand, conflicting results, particularly in studies of AO community pH response, indicate difficulties in drawing generalizations could stem from a broad undescribed diversity in this group (Prosser & Nicol, 2012). NH+4 is toxic at high concentrations, so it has been hypothesized that niche determinants for nitrifiers could include affinity, and tolerance for the compound and its various source complexes. Interaction with N mineralisers is important for at least some acid peat AOA, which appear to be stimulated by addition of urea and organic N but not by NH+4 . Another
29 study found soil AOB were mostly found in the top 20 cm of soil, and the AOA:AOB ratio increased at 40–50 cm (Prosser & Nicol, 2012). 1.4.2 Drivers of N2O emissions from nitrifiers Nitrosification contributes to N2O emissions in the form of NO−2 produced from NH+4 (Sanford et al., 2012). Nitrobacter spp. are inhibited by high levels of NH3/NH+4 (Cameron et al., 2013), so high NH+4 fertilizer application could result in more NO−2 in the soil solution, which would be available as an electron acceptor. Most studied AOB can also denitrify by a process called nitrifier denitrification (ND), which may be universal for betaproteobacterial AOB. In anaerobic conditions these nitrifiers can reduce NO−2 with NH+4 as the electron source. Some strains of Nitrosomonas are capable of complete reduction of NO−2 to N2, which can be coupled to organotrophic growth (Bock & Wagner, 2013). Estimates of the ND contribution to soil N2O emissions range ~0–30% (Wrage et al., 2001). An explanation that has been given for the existence of ND is NO−2 toxicity at higher concentrations (Kool et al., 2011), which appears to be particularly acute in pure cultures of AOB at pH 4. Co-cultures of Nitrobacter strains with AOB have led to growth at this pH (De Boer et al., 1995) and reduced N2O production (Kester et al., 1997), which can be attributed to the removal of NO−2 . If toxin removal explains the evolutionary origin of ND, the existence of AOB that can also reduce N2O nevertheless suggests that the process may have adapted towards meeting the energetic needs of these organisms. The hypothesis that NO−2 toxicity plays a role in ND is supported by findings that Nitrosomonas europaea expresses NO−2 reductase in the presence of NO−2 even in aerobic conditions (Kool et al., 2011). Some heterotrophic nitrifiers can also reduce NO−2
in aerobic conditions by a pathway that may be distinct from ND as
hitherto described (Pilegaard, 2013). A number of studies suggest that where soil is well aerated, most of the N2O emitted is produced by AO activity, whereas in waterlogged soils the predominant source of N2O is due to denitrifiers (Brown et al., 2012). Also, nitrification is generally the main source of N2O when NH+4 concentrations are high and water filled pore space (WFPS) is 60% or less. Applying urease and nitrification inhibitors to soil in conjunction with urea fertilization was shown to reduce N2O emissions by 37% (Zaman & Blennerhassett, 2010). The elements N and O are provided by different sources in denitrification pathways, a stable isotope incubation experiment could be conducted to determine
30 which conditions are conducive for each process (Kool et al., 2011). A poor sandy soil of pH 5.4 was supplemented with equal amounts of NH+4 and NO3− , and the nitrifiers were found to produce most of the N2O evolved over a 28 hour incubation period at WFPS 50% and 70%. Denitrifiers were only important at 90% WFPS, when they were responsible for 92% of the N2O produced. About 20 times more N2O was evolved at the highest moisture treatment and the amount of denitrifier N2O that derived from NH+4 was marginal in each case. Although the levels of N2O reported from soil ND are low, they should not be neglected, as any form of aerobic denitrification may be associated with higher N2O/(N2 + N2O) ratios due to the O2 sensitivity of N2O reduction in many studied species (Kraft et al., 2011). 1.5 The denitrifying bacteria Denitrification is carried out as a facultative form of anaerobic respiration by a broad diversity of bacteria, archaea and fungi belonging to > 60 described genera. Up to 5% of bacteria (Philippot et al., 2007) and 20% of microbial biomass (Robertson & Groffman, 2007) in the soil are estimated to belong to the group. Denitrifiers comprise a functional guild that does not remotely resemble a monophyletic group, even within the bacterial domain (Braker et al., 2012). A survey of 1500 diverse soil isolates in 1977 found almost 10% could carry out complete denitrification, and since then we have been obtaining a more complete picture of the diversity that mediates this central pathway in the nitrogen cycle. However, there remain many questions including how population shifts in the denitrifying community are linked to denitrifying activity (Hallin et al., 2007). The denitrification pathway employed by denitrifiers is fundamentally distinct from that of autotrophic NH3-oxidizers. Denitrifiers use NO3− as the substrate, while in ND, the intermediate nitrification product NO−2 is used directly. To clarify, denitrifiers utilize NO−2 that derives from the reduction of NO3− , while nitrifiers reduce NO−2 that comes from the oxidation of NH+4 (Pilegaard, 2013). The classic form of denitrification (or “denitrifier denitrification”) is initiated by one of the two described NO3− reductases, which can both be encoded for in the same strain (Philippot et al., 2007). However, membrane-bound Nar is more commonly involved in denitrification, whereas periplasmic Nap is usually active in aerobic conditions (Delgado et al., 2007). Denitrifiers are a small sub-group of the nitrate reducers, which are more widespread, as one study found 10% of soil bacteria could conserve energy using NO3− as the terminal
31 electron acceptor (Philippot et al., 2007). This group is distinct from the denitrifiers, as the majority of nitrate reducers cannot reduce other forms of Nr. The same genes are responsible for denitrification in nitrifiers and non-nitrifiers, but it remains unclear how much their regulation varies in their response to conditions favouring denitrification. 1.5.1 Beyond nitrate reduction Most denitrifiers reduce NO3− , but in contrast to most NO3− reducing organisms, they are able to further reduce NO−2 to NO (Bothe et al., 2007, p. xiii). Almost all NO is reduced further to N2O, and many denitrifiers can continue the reduction of N2O into N2, which is performed by nitrous oxide reductase (NosZ). Genes for NO−2 reductase (nirS and nirK) are frequently used as functional molecular markers of denitrification. They perform a rate-limiting step, and are found in virtually all denitrifiers, including those with a truncated reduction pathway. Only a vanishing proportion of microbial diversity can be isolated in pure culture, and denitrifiers are no exception. One study detected in marine samples up to a thousand times more nirS copies by competitive PCR (cPCR) than by culture-dependent MPN. Denitrifiers are probably no more or less amenable to culture techniques than soil bacteria in general. Soil nirK copy number has been measured with quantitative real-time PCR (qPCR) to be 100–1000 times lower than the 16S rDNA copy number (Hallin et al., 2007). Nevertheless, differences in the ability to carry out denitrification and the way related gene expression is regulated can be found between closely related strains. These differences could lead to ecosystemlevel differences, and it appears that denitrifier activity may be more dependent on the composition of the denitrifier community than shear numbers of genes present in the soil (Braker et al., 2012). Denitrifier diversity and activity may therefore be partially determined by the plants growing in the soil and their root exudates, which exert a strong selective pressure on the soil (Richardson et al., 2009b). Although differences in the regulation of denitrification at the strain level are poorly characterized, they are suggested by differences in the number of relevant enzymes encoded. For example, nitrite reduction yields the highly cytotoxic compound NO, so transcription of nir is often co-regulated with expression of the nitric oxide reductase gene (nor) in response to the presence of NO (Felgate et al., 2012). However, even this process has its exceptions, and NO can reach inhibitory levels in Rhizobium sullae (Delgado et al., 2007). Conversely, perhaps because N2O is relatively benign, closely related strains can differ in their ability to reduce N2O (Ishii et al., 2011). Copper (Cu)
32 is the required co-factor for the enzyme products of both nosZ and nirK, but interestingly the nosZ is more common in bacteria that possess nirS, which encodes the cytochrome cd1 form of NO2− reductase. The two nir genes are functionally interchangeable and the chromosomal loci of nosZ and nirS are not close to each other, so there could be an ecological explanation for this linkage (Jones et al., 2008). The ratio of nirS to nirK genes in soil varies spatially (Philippot et al., 2009) and Fischer et al. (2013) found that the abundance of denitrifiers was greater in the top 30 cm of agricultural soil than in lower horizons. The proportion of nirK genes to nirS genes was also greater in topsoil but there was no correlation between depth and the ratio of N2O reducers to NO3− reducers. This was unexpected, as the subsoil generally has lower NO3− concentrations, which would lead one to expect the prevalence of complete denitrification. The depth distribution of denitrifiers was probably more due to the lower DOC content of subsoil, as it was not affected by the higher NO3− found in subsoil planted with a legume crop. Enzymatic forms of denitrification are distinct from chemodenitrification, where NO−2 oxidizes other nitrogen compounds, or is reduced spontaneously at low pH (Pilegard, 2013). Small amounts of N2O are also produced from NO reduction carried out in order to reduce nitrosative stress (Philippot et al., 2007). This does not contribute to cell growth and is not evidence of denitrification. As diverse as they are, the contributions of these processes are minor. The denitrifiers remain the main producers of N2O in the soil, followed by the nitrifiers (Kool et al., 2011). 1.5.2 Nitrous oxide reductase A study of environmental isolates found 59 out of 71 denitrifiers could use N2O as a terminal electron acceptor (Okereke, 1993). Similarly, a survey of 68 sequenced denitrifier genomes found that 37% lack the encoding gene, nosZ (Jones et al., 2008) and environmental samples can have as many as 10 times more nir genes than nosZ genes (Jones et al., 2013). The typical range for nosZ is 105–107 copies in a gram of soil (Long et al., 2013). Among proteobacterial denitrifiers, fully sequenced Agrobacterium tumefaciens C58 does not encode nosZ, and Azospirillum lipoferum controls expression by phase variation (Philippot et al., 2009). It has been recently found that some nondenitrifying environmental bacteria can also reduce N2O, as two isolates of soils and sediments Desulfitobacterium hafniense and Anaeromyxobacter spp. encode nosZ but have neither nir gene (Jones et al., 2013). Abundances of nosZ in environmental
33 samples have been studied for some time but it is becoming evident that only a portion of the diversity has been sampled. There are two clades of the gene, the “typical” NosZ being the better studied of the two and more common among sequenced Proteobacteria. Its main feature is a Tat signal peptide, distinguishing it from clade II, which is typically translocated by the Sec system. The existence, for a single enzyme, of two systems with very different energy requirements is unusual, and may reflect ecological roles. Gene copies of both clades were detected in similar quantities in arable soils in Sweden and France (Jones et al., 2013). Most Cu-dependent enzymes have functional equivalents that do not require the cofactor, but for NosZ none are known. Felgate et al., (2012) found that under Culimited conditions a chemostat culture of Paracoccus denitrificans produced > 1000 times more N2O than when Cu was available. Inability to use N2O as an electron acceptor had little impact on growth and was compensated for by a ~ 20% increase in NO3−
consumption. A truncated form of NosZ accumulated under Cu-limited
conditions, which may have partial activity. However, it appears that the competitive advantage of having this gene is diminished in NO3− -rich environments. The authors note the environmental implications of their findings, as ~ 20% of arable lands in Europe are Cu deficient with concentrations < 1 mg/kg. It has previously been assumed that because not all denitrifiers have the nosZ gene encoding nitrous oxide reductase, it must not confer a significant competitive advantage for facultative anaerobic growth. N2O reduction is the only known entirely Cudependent enzymatic process carried out by bacteria, and a high demand for this cofactor comes from the multi-copper sulfide (Cu4S2) centre (CuZ) and di-nuclear Cu centre (CuA) (Felgate et al., 2012). Studied forms of NosZ are periplasmic, so it was believed to be limited to Gram-negative bacteria, mostly Proteobacteria and Bacteroidetes. Recently however, the gene has been annotated in genomes from Archaea, Firmicutes, and Chloroflexi (Jones et al., 2013), raising questions regarding its localization in these strains. One Bacillus isolate was reported to encode two nosZ copies (Jones et al., 2011), which interestingly produced a lower level of denitrification intermediates than isolates with only one copy. The role of denitrifier community composition in affecting the N2/(N2 + N2O) product ratio is not clear, but the modular nature of the pathway has led to a diversity of possible configurations.
34 1.5.3 The denitrification product ratio In the field, denitrification yields NO, N2O and N2 in variable quantities. Although field measurements of N2 fluxes have not yet been achieved, one study has measured the ratios in a variety of conditions applied to soils in the laboratory to determine the effects of labile C and NO3− availability on the N2O/(N2O + N2) product ratio (Senbayram et al., 2012). It was found that at high but agriculturally relevant NO3− concentrations the ratio is high because it appears NO3− and its more immediate reduction products are preferred over N2O as electron acceptors. When NO3− is more limited (at e.g. 0.2 mM), the product ratio decreases. The final step of denitrification is highly energetically favourable, but producing a diversity of N reductase enzymes presents no advantage when NO3− is plentiful. In this situation, Nar and Nir can meet the cell’s energy needs and are less sensitive to fluctuating O2 availability (Jones et al., 2008; Felgate et al., 2012). Indeed, after fertilizer application 0.5 % of applied N is likely to be emitted as NO (Pilegaard, 2013), despite its toxicity and instability. The probable explanation is that the denitrification cascade is overwhelmed at the bottom end as a result of rapid ammonia oxidation, and the products of each successive step can accumulate before they are reduced. The form in which nutrients are present in the soil affects the product ratio depending on their bioavailability and effect on soil chemistry. One study found that vineyard soils under an organic management system, where synthetic Nr is not used, had greater numbers of denitrifier genes than those under conventional management, particularly nirS (Tatti et al., 2013).
1.6 Rhizobia and legumes The ability to fix nitrogen is believed to have evolved once before spreading from this first diazotroph to various phyla of archaea and bacteria (Canfield et al., 2010). A large proportion of BNF occurs in the oceans (Figure 1), where autotrophic phytoplankton has a largely fixed C/N ratio. Photosynthesis can only proceed for as long as Nr is available, which means that denitrification limits primary production. In terrestrial environments, however, plants have various C/N ratios, which are generally higher than that of phytoplankton. This allows growth and carbon fixation in a range of environmental constraints (Gruber & Galloway, 2008). Legumes have a particularly high demand for N, and have evolved the ability to promote BNF, which benefits the plant. Without these plants, N is fixed at a rate of 1–5 kg ha−1 yr−1 by free-living
35 diazotrophs. Legumes have been measured to fix up to 200 kg N ha−1 yr−1, although this figure is lower for economically important legumes. These rates can be compared to the amount of fertilizer N (organic or synthetic) applied to agricultural soils in Europe, where national averages range 42–243 kg N ha−1 yr−1 (Erisman et al., 2011). An exceptionally diverse family, many legumes are economically important species. In 2009 legume crops were grown on 193 million ha, considerably less than the almost 700 million ha given to cereal crops, and in decline except in the case of soybean. By introducing legumes into rotational cropping systems farmers can reduce nitrogenfertilizer and agrochemical inputs, thereby reducing energy consumption by 12–34% (Jensen et al., 2012). There have been concerns that growing legumes to reduce reliance on fossil energy may not effectively reduce N2O emissions. Fertilizer use is associated with peak emissions during spring, whereas legumes produce a higher base-line level of emissions than fertilized non-legume crops. High emissions are caused in the spring following autumn ploughing in fields of the perennial forage legume alfalfa (Crews & Peoples, 2004). Legume-derived emissions appear to originate from their residues, which have a low C/N ratio, rather than as a direct consequence of BNF. The only situations where legumes have been recorded to cause higher emissions than N fertilizers are when legume pastures are terminated or when legumes are used as green manure. There is no evidence that harvesting legume crops produces more N2O than harvesting non-legume crops, which is probably because much of the fixed N is withheld from the soil in plant biomass (Jensen et al., 2012). The high rates of BNF in legume soils are due to rhizobia, nitrogen-fixing Alpha- and Betaproteobacteria, many of which have a mutualistic relationship with legume plants as differentiated bacteroid cells within specialized root structures called nodules. The enzyme that reduces atmospheric N2 to NH3 is nitrogenase, which is highly sensitive to O2. The plant supplies the organic compounds required to meet the energy demands of the reaction while a protein called leghaemoglobin regulates O2 availability within the nodule (Ott et al., 2005). Enzymes encoding enzymes for all four of the reactions in the denitrification pathway have been found in rhizobia, which have been noted to be active in both free-living and bacteroid forms. It has been suggested that although denitrification rates are low for rhizobia, their contribution to anthropogenic N2O emissions may nevertheless be significant due to the widespread cultivation of legumes (Delgado et al., 2007).
36 The expression of nirK usually occurs in response to NO−2 , but in R. sullae HCNT1 a low O2 concentration is sufficient (reviewed by Delgado et al., 2007). Isolated Bradyrhizobium japonicum strain 505 bacteroids have even been demonstrated to fix N2 using NO3− as an electron acceptor. Both members of this species and Sinorhizobium meliloti encode genes for each step of the denitrification pathway enzymes, but interestingly in the case of the S. meliloti the first two steps cannot be performed to support respiration. The amount of N2O produced by rhizobia in vivo varies greatly and consequences for net nitrogen fixing efficiency could be an issue (O’Hara & Daniel, 1985). 1.7 Soil characteristics Soil is highly heterogeneous and as a result microbial activity occurs at “hot spots” (Hofstra & Bouwman, 2005). In the case of denitrification, these are determined by the factors of pH, O2 pressure, temperature, and the availability of N-oxides and suitable electron donors. These have been described as the “proximal” controls of denitrification, which control its rate at a particular point in time (Wallenstein et al., 2006). Denitrification is not mediated directly by the confluence of these conditions, but also depends on “distal control”. Predation by protozoal grazing and viruses, environmental stress, and other factors shape the denitrifier community, which effectively determines the potential for denitrification in a soil. Some factors exert both proximal and distal control – in fact all the proximal controls of denitrification also have a distal effect, with the exception of soil NO3− availability, which has not been shown to affect denitrifier community composition (Wallenstein et al., 2006). Both distal and proximal controls can vary hugely on the micro scale, particularly in the rhizosphere. Some features of the soil are under direct influence of plant roots, particularly as a result of rhizodeposition. This affects the C/N ratio of labile SOM, which determines whether bacteria are immobilizing Nr or mineralizing it, making it available for plants and nitrifiers. The two processes can be occurring simultaneously separated by just a few tens of micrometres (Robertson & Groffman, 2007). Roots reduce O2 partial pressure through respiration and NO3− content by assimilation, while improving carbon availability by rhizodeposition (Hamonts et al., 2013). Other characteristics depend on the soil parent material and climatic effects.
37 1.7.1 Soil pH Plant and animal diversity is largely dictated by the temperature and latitude of habitats, but for microbial life in soil, pH is the major determinant of both the bacterial phyla present (Lauber et al., 2009) and phylotypes richness (Fierer et al., 2012), irrespective of habitat type. Particularly Actinobacteria, Bacteroidetes, and Acidobacteria respond to pH gradients, and diversity is greatest at neutral pH. Acidobacteria can dominate the entire soil bacterial community at pH < 5, whereas numbers of the other two phyla start to become significant at pH > 5 (Lauber et al., 2009). In many parts of the world the pH of croplands is raised to improve productivity, but to say this also raises microbial diversity is a simplification. When acidic forest soils are converted to agriculture, a rise in local bacterial diversity may be observed as a result of liming. This diversity, also called “alpha diversity” can be measured as the average species richness or the balanced representation of various taxa (i.e. how much the community is not dominated by only a few taxa). Beta diversity is a measure of how much the community varies over time or space. Rodrigues et al. (2013) found that Amazon rainforest converted to pasture had a higher bacterial taxonomic richness and phylogenetic diversity than rainforest, but that the beta diversity was lower. The community in rainforest soil would be different at different sampling locations, whereas the pasture community was more homogeneous. Proteobacteria made up over half of the bacteria in both soils but Acidobacteria were diminished in pasture soil (from 21.1% to 13.4%) and Firmicutes became more prominent following conversion (from 2.2% to 12.9%). The kinetics of most reactions of the N cycle are affected by pH, and as such are manipulable by agronomic practices (Mørkved et al., 2007). Some of these effects are enough to allow reactions to progress without the catalytic influence of microbes. When NO−2 is produced from ammonium applied to soil with a pH value below 5.5, this can also be chemically reduced. A significant proportion exists as nitrous acid (HNO2), which decomposes to NO, HNO3 and H2O, providing substrates for denitrifiers (Mørkved et al., 2007). This is a form of chemodenitrification, which can also be carried out indirectly by ferric iron-reducing bacteria. Their respiration produces ferrous iron (Fe2+), which can provide electrons for the chemo-denitrification of NO−2 (Sanford et al., 2012). In wastewater treatment it has been observed that HNO2 inhibits NosZ (Richardson et al., 2009b), though pH also affects this enzyme directly.
38 In laboratory and field experiments the overall rate of denitrification has been found to increase with increasing pH. Denitrification reaction chemistry is directly affected by pH but whether the relationship is at all attributable to the impact of pH on microbial community composition is unknown. The optimum pH for potential denitrification appears to be correlated to soil native pH, which perhaps reflects the diversity of microorganisms and pathways involved. In contrast, there is a strong trend for greater N2O emissions in low pH soils that is independent of microbial community optimum pH (Liu et al., 2010). In some studies the trend only holds for waterlogged soils, but not in drained soils (Mørkved et al., 2007). Čuhel et al. (2010) manipulated soil pH over 10 months and found that N2O emissions were the same, though overall denitrification was lower in the low pH treatment. Liming is recommended to reduce emissions, though the immediate effect can be greater respiration and nitrogen mineralization. If O2 is depleted as a result, this leads to denitrification, which can be mistaken in experiments as the direct effect of a higher soil pH rather than a possibly transient burst of activity (Mørkved et al., 2007). In a soil incubation experiment, Liu et al. (2010) observed a large increase in copy numbers of both nir genes and the nosZ gene when pH was raised from 4.0 to 6.1. The increase in nir gene copies per 16S rRNA gene copy was several times greater than the increase in nosZ copies, which indicates that strains with a truncated denitrification pathway do not necessarily show faster growth in low pH conditions. Furthermore, the ratio of nosZ transcripts to nirS transcripts was higher at pH 6.1 than at pH 8.0, leading the researchers to infer that the reason for lower N2O reduction in acidic conditions must be post-transcriptional. Also, low pH delayed the onset of N2O reduction activity with respect to Nar/Nap, Nir and Nor activity. Philippot et al. (2009) found nosZ was lower where the N2O/(N2 + N22) ratio was high, concluding that this may be due to the adaptive loss of a gene with a functionally impaired enzymatic product. Liming could be recommended for acidic agricultural soils to simultaneously maintain healthy baseline denitrification while minimizing N2O emissions. 1.7.2 Soil moisture and oxygen Denitrification is an alternative form of respiration, so its prime use is to provide bacteria with a way to conserve energy in anoxic conditions. This places it spatially apart from the mostly aerobic process of nitrification. The role of water in moderating these processes is that it diminishes the passage of N2O from – and of O2 to – the soil
39 (Pilegaard, 2013). Waterlogging has a negative impact on crop health, which is partially due to a 10,000-fold decrease in the diffusion of soil O2, which raises competition for NO3− with rhizosphere microorganisms (Hamonts et al., 2013). In many bacteria both NO3− reduction and the final step of denitrification are sensitive to O2. Of the nitrate reductase enzymes, membrane-bound NarG receives its substrate from the cytoplasm, which depends on the activity of an O2-sensitive enzyme to transport NO3− through the membrane. Periplasmic NapA does not require NO3− transport and can carry out the reaction in aerobic conditions. This was first discovered in alphaproteobacterial sulphur bacterium Paracoccus pantotrophus, which encodes both enzymes (Bell et al., 1990). It has been noted that N2O emissions can rise even when oxygen is introduced. One explanation could be that when denitrification continues to function in oxic conditions, NosZ activity decreases due to its sensitivity to O2. The reason may be that in environments where O2 availability constantly fluctuates, bacteria that do not have to change their mode of respiration can have a competitive advantage (Kraft et al., 2011). Also, when denitrification is at its maximum in waterlogged soils, the end product is mostly N2 because the demand for N2O increases. In moderately wet soils N2O can more easily diffuse to the atmosphere without being consumed (Pilegaard, 2013). Tillage introduces air into the soil and reduces denitrifier abundance by up to 60% (Long et al., 2013). Climate change in Western Europe has been associated with an increased frequency of heat waves and an increased frequency of drought. Soil moisture cycles can be affected, and along with them, the cycling of nitrogen (Hartmann et al., 2013). 1.7.3 Fertilization Innately, most agricultural soils have low potential denitrification because of the low availability of organic matter and overall microbial activity compared to grasslands. Only < 7% of the global agricultural area can be described as organic soils (Hofstra & Bouwman, 2005). Intensive agriculture, however, is characterized by highly refined fertilizer addition to oxygen-starved compacted soils, leading to higher rates of denitrification than on non-cultivated soils (Philippot et al., 2007). Along with crop type, the rate and form of N-fertilization are the most significant management-related variables affecting denitrification. Applications of mineral N in excess of 225 kg ha−1 yr−1 are associated with marked increases in denitrification rates (Hofstra & Bouwman, 2005). In Europe mean national inputs range from 42 kg ha−1 yr−1 in Portugal to 243 kg
40 ha−1 yr−1 in the Netherlands (Pilegaard, 2013). Historically, from 1860 to 2005, more emissions have originated from manure use, from which 2.0% Nr is emitted as N2O, during livestock management and soil application. Synthesized fertilizers have become more important in many regions, and from these sources 2.5% of the Nr is emitted as N2O, including off-site emissions deriving from deposition and leaching (Davidson, 2009). The effects of soil fertilization on denitrifier communities have been a major subject of molecular research in this field (Hallin et al., 2007). SOM is the main reserve of nutrients in the soil, and the amount of bioavailable NO3− in agricultural soils vary due to sources and sinks of the nutrient in a largely site-specific manner. There are three situations when NO3−
accumulates in the soil. These arise after application of N
fertilizers or manure, after harvesting, and in late winter before the growing season. When combined with anoxic conditions and a reductant such as organic carbon, denitrification can be higher at these times (Munch & Velthof, 2007). One study found that in microcosm soil to which no carbon had been added there was no expression of denitrification genes and so denitrification was most likely due mainly to the proteins already synthesized. The denitrification rate per gram of soil organic carbon was constant irrespective of pH, in all but the most heavily organic soil (Liu et al., 2010). Fields with high fertilizer inputs with high denitrification rates have little NO3− leaching. Riparian zones have an important role, as they are moist and rich in organic matter, factors that can reduce the risk of eutrophication in downstream catchments (Munch & Velthof, 2007). The quality of fertilizers used matters because synthetic and organic fertilizers have very different constituents. In the form of crop residues, the effect of nitrogen amendments depends also on whether these are incorporated as they are or burned first. Calcium ammonium nitrate can promote denitrification through a local increase in pH and when manure is added simultaneously with synthetic N, the supply of both substrates leads to higher rates (Hofstra & Bouwman, 2005). Compost provides nitrogen in a more stable form that must be mineralized before it is bioavailable, but the carbon it supplies increases overall respiration, which can create anaerobic conditions that promote denitrification (Tatti et al., 2013). In the case of a soil additive with a very high C/N ratio (>25:1), immobilization of N exceeds mineralization and less NO3− will be available for respiration. The C/N ratio of highly decomposed humus and compost is
41 low, but because this form of SOM is recalcitrant, N is not readily available (Robertson & Groffman, 2007). Notwithstanding the considerable environmental concerns of farmland fertilization, the availability of inputs should still be improved to economically vulnerable smallholders. The continuing conversion of land to agricultural use has a greater impact on emissions than would be incurred by improvements made on yields from existing farmland (US EPA, 2011). Agroecosystem management strategies should allow soil processes to function fully so that crops can benefit from SOM supplementation to provide the required nutrients, rather than depending solely on mineral fertilizers. Unlike synthetic fertilizers, SOM is heterogeneous with a wide variety of source materials, and depends on decomposition by soil microbes that display equal heterogeneity. Trials of new approaches to soil amendment produce conflicting results because it is not well understood how long-term SOM additions impact the bacterial community by releasing labile organic matter of varying composition (Bowles et al., 2014). Depolymerizing the SOM is the rate-limiting step of mineralization, so the extant microbial community also determines how the substrate is utilized. It needs to be known which kind of SOM is suitable to provide the right balance of carbon and nitrogen for the crop and its rhizosphere symbionts. Community fingerprinting could provide a costeffective way of opening the “black box” of the microbial context.
2 Review of methods Denitrification is a process that is notoriously difficult to measure (Groffman et al., 2006). An early study used pure batch and chemostat cultures of soil bacteria to determine the effect of partial pressure of O2 on NO and N2O production (Anderson & Levine, 1986). For economical appraisals, the N-balance approach is easily interpreted as it involves comparing the N-inputs and N-outputs for the agroecosystem (Hofstra & Bouwman, 2005). This allows the complete growing season to be considered, something that is challenging to model at the microbiological level. However, some inputs and outputs are difficult to estimate, and for a mechanistic understanding of N losses to denitrification there is a range of laboratory approaches available.
42 2.1 Denitrifying enzyme activity Potential denitrification usually refers to the denitrification enzyme activity (DEA), which is assessed with an excess of carbon and nitrogen substrates added to a soil slurry or intact core, in anaerobic conditions to maximize the activity of existing denitrification enzymes. The resulting N2 is small against the atmospheric background, but can be measured if the headspace is purged with an inert gas. It is difficult, however to achieve gas-tight conditions, while ensuring absorbed N2 is not released from soil aggregates or the vessel material. For this reason, DEA is most commonly assessed by the acetylene inhibition technique (AIT). Emission is measured in vitro with acetylene (C2H2) in the headspace, which inhibits NosZ activity. The rate of total denitrification can be evaluated by the production of N2O because the pathway is truncated at this point. This approach is chosen for its ease of use compared to in situ or intact core techniques (Groffman et al., 2006). The advantage of DEA is that, as a measure of microbial activity, it can be affected by management practices before changes in community composition are evident (Bowles et al., 2014). It relies on the activity of intact NirK and NirS enzymes present in the soil, so includes the vast majority of soil denitrifiers, regardless of their taxonomy. Philippot et al. (2009) also found DEA to be related to the ratio of nosZ to 16S rRNA genes present in the sample. This did not depend on absolute denitrifier gene abundance as determined by quantitative PCR, which is interesting as NosZ activity is not measured by the assay. However, several limitations mean that the method should be used with caution. Firstly, although the technique can be scaled up to pot experiments (Klemedtson et al., 1987), its use in the field not practical. There are also inadvertent effects of C2H2 that should be accounted for. Other issues are that the inhibition of N2 production may be incomplete, particularly as there exists a lot of NosZ diversity that has not been characterized in pure culture. If substrates are not supplemented, it should be borne in mind that C2H2 is also known to inhibit of NO3− production from NH4+. Furthermore, as a low molecular weight organic compound, C2H2 is a possible substrate for growth (Terry & Duxbury, 1985). One study tested for de novo synthesis of NosZ during use of AIT, which would result in overestimates of DEA. This was found not to occur within a 2-hour incubation (Hartmann et al., 2013), but a bacteriostatic antimicrobial should be considered for use in longer incubations. DEA also has limitations that are similar to
43 those described by Prosser and Nicol (2012) with respect to measuring potential nitrification. Some activity will be unaccounted for, which is inevitable when attempting to create conditions that are ideal for the entire bacterial community. The slurry method is most common, but this breaks the soil’s structure and the micro-niches it creates. It may also disrupt the close interactions that some non-culturable bacteria may depend on. Nevertheless, some studies have found it to be a useful measure. Čuhel et al. (2010) found that potential denitrification was correlated to N fluxes in the field in that when the DEA assay was performed with and without C2H2, the ratio of N2O produced correlated with the N2O/(N2O + N2) ratio in the field. Senbayram et al. (2012) performed experiments using soil supplemented with various sources of organic matter. They found that as denitrification progressed and the NO3− concentration fell below 20 mg NO3− -N /kg dry soil, N2O consumption began to exceed production from NO3− . It may be that in such conditions the ratio between NO3− concentration and available C has an effect on the denitrification product ratio. A typical AIT procedure supplements the slurry with around 18 mg NO3− -N / kg dry soil, so it should be assessed whether this is sufficient to eliminate the effect differences in sample indigenous NO3− could have. The nutrient status of the soil can make a big difference. For their potential denitrification assay, Senbayram et al. (2012) used NO3− concentrations ranging 0.5–20 mM to test their effects on DEA measured from soils undergoing long-term amendments with organic matter (OM) and mineral nitrogen (MN). Control soil that did not receive either treatment and OM soil both had greater denitrification rates under greater NO3− supplementation, but denitrification in MN soil did not appear to be limited by NO3− concentration. 2.2 Molecular characterization of the microbial community Methods that are commonly used to characterize microbial communities, including in the context of denitrification, include the PCR-based approaches terminal restriction fragment length polymorphism (T-RFLP) and denaturing gradient gel electrophoresis (DGGE). Analysis by simple RFLP is called amplified ribosomal DNA restriction analysis (ARDRA) when used for community fingerprinting. This approach allows and assessment of the level of similarity between different communities. However, the bands can be difficult to quantify, and can represent several segments of the same length, so the level of community diversity cannot be inferred. For this reason T-RFLP is used, where fluorescent primer tagged fragments correspond to the same starting
44 point of the gene (the terminal fragments). The fragments are of various lengths due to polymorphisms that result in the genomes of each organism having slightly different recognition sites for the restriction enzyme used. Gel electrophoresis produces a simpler pattern, where bands correspond more closely to specific operational taxonomic units (OTUs). 2.2.1 Clone libraries Cloning and sequencing a sample of the genes in the community, or using available genome sequences, one can show how OTUs are correlated to the phylogeny of the gene (Hallin et al., 2007). Communities may be compared taxonomically according to the proportion of OTUs shared in two samples, which is carried out when calculating the Bray-Curtis distance. Alternatively they can be compared using a combined phylogenetic tree of the sequences from two communities. The phylogenetic metric UniFrac calculates distance based on the proportion of branch lengths unique to one of the communities. However, sequencing is time-consuming, especially if it is to be exhaustive. OTUs that are present in fewer numbers are often the most sensitive to environmental disturbances, and to characterize these sequences an extensive clone library must be constructed (Entry et al., 2008). To isolate some of the less dominant specimens, DNA can be fractionated by G+C content. This is convenient but somewhat arbitrary, as it does not discriminate between phylogenetic groups, except in the case of Actinobacteria, which as a rule have high G+C content (Mühling et al., 2008). Fingerprinting gives richness, diversity, and evenness data that cannot be obtained as rapidly or inexpensively by sequencing clones (Entry et al., 2008). 2.2.2 PCR-based approaches DGGE lends itself to use in conjunction with sequencing. This method has been used to characterize denitrifier communities by amplification of nirK and nosZ genes. Fragments can be excised and cloned, making it easier to sequence a representative sample of the denitrifying organisms present. The disadvantage of DGGE is that it requires lengthy optimization of the denaturant gradient in the gel (Hallin et al., 2007). However, the ability to directly identify bands of interest if similar sequences are found in genome databases makes it informative. One of the biases of PCR in gauging bacterial numbers is that many bacteria are over-represented; a genome can have 1–13
45 copies of the 16S rRNA gene (Henry et al., 2008). Bacterial cells are also lysed with inconsistent efficiency during DNA extraction. For example, Bacillus spp. and Clostridium spp. are abundant among soil culturable bacteria but constitute only 2% of 16S rRNA libraries from the soil, which may in part be due to tenacious cell walls and spores (Janssen, 2006). When using one primer pair, some members of the community are more efficiently represented than others. Primer mismatches in the last 3–4 base pairs in particular will strongly affect initiation of amplification (Tiirola, 2002). To gain a more complete picture of the community, more primer pairs may be used (Entry et al., 2008). The 16S rRNA gene is so highly conserved, however, that bias from PCR can be minimal. Fierer et al. (2012) studied soils by extensive shotgun sequencing, where sequences are generated as a representative sample of total extracted DNA without amplification. These were found to closely match taxonomic results based on amplicon sequencing targeting the V4 region of the 16S rRNA gene, suggesting that the primers efficiently targeted most soil bacteria present. 2.3 Ecological indices and community quantification Any estimate of bacterial species in a natural environment is at best a quantitative description of the genetic diversity of the community, rather than a discrete number of groups to which organisms can be assigned among. Assigning organisms to groups of related species (i.e. at the genus level) can be even more open to interpretation than delineating species (Janssen, 2006). The definition of “species” for bacteria is beyond the scope of this thesis, but a short discussion of it may serve to make its aims clearer. When plants and animals are studied, they form discrete groups that belong together and form a gene pool separate from those of other plants and animals. What distinguishes them from bacteria is that they reproduce sexually. For this to serve as an effective means of procreation, individuals of a species are restricted to being fairly similar to each other on the genetic level. There are no constraints, on the other hand, to retain similarity to other related organisms where the opportunity to mate does not arise, and this is how new species are formed. Suffice it to say that bacteria reproduce asexually by cell division, so are free to mutate and evolve without much consideration of their sister cells. For some bacteria the ability to communicate and form multicellular structures with conspecifics is an important part of their survival. Ecology nevertheless does not necessitate the same
46 degree of similarity, as does the fusion of gametes. Bacteria can transfer genetic material even between phylogenetically diverse taxa, which can give rise to whole new phenotypes. There is not a single concept of “species” that suffices to describe all bacteria, so in community analysis researchers define operational taxonomic units (OTUs) based on threshold of sequence similarity of a gene found in all bacteria. The most common sequence studied for this purpose is that for the highly conserved small subunit ribosomal RNA (16S rRNA) gene, or its rRNA product. The sequence of this gene is one of the criteria used in the process of defining a species, so OTUs based on sequence comparisons of partial readouts of this gene will continue to be used as a surrogate for species, for as long as the majority of bacterial species remain to be described. Profiling methods can quickly measure microbial community richness, diversity and evenness. These are not entirely equivalent to the terms used in other fields of ecology, as they are based on species numbers and the definition of a species in microbiology is malleable. Instead, the individual peaks or OTUs can be considered as data points and the indices give an indication of community response to disturbances and selective pressures. Evenness refers to the similarity in abundance of the various OTUs. These characteristics are indicators of soil health and microbial functioning (Entry et al., 2008). The 16S rRNA gene is the universal phylogenetic marker for bacteria, but to specifically study functional subgroups within the community, other molecular markers are used. Central to describing communities that carry out denitrification and other soil processes is investigating the hypothesis that activity is positively correlated to the numbers of organisms displaying this activity (Hallin et al., 2007). Some organisms can have several copies of the required gene, and sometimes a required protein may be encoded but inactive, but on the whole the number of gene copies is used as a proxy for the number of bacteria carrying out the process. Two methods that have been used successfully to quantify nir genes in environmental samples are cPCR and qPCR. The latter of these is more common as it does not require a time-consuming gel-migration step (Hallin et al., 2007; Kandeler et al., 2006). This gene comes in two forms, nirK and nirS, which must be amplified using separate primer pairs, so each half of the denitrifier diversity has to be amplified separately. For a long time nosZ was treated as a relatively highly conserved gene, but a diversity of previously undescribed nosZ genes has recently been described in soils (Sanford et al., 2012, Jones et al., 2013). Combined
47 with the fact that they are absent from a lot of denitrifiers, this makes nosZ of secondary interest to nir in most studies. The role of more ubiquitous genes, on the other hand, has to be regarded critically. From soils with a high selenite content, a rhizobial NirK was found to reduce the toxic anion (Jones et al., 2008). It is evident that particularly when studying bacteria from unusual environments, possible alternative roles of enzyme products should be considered. Some studies have found a link between denitrifier community structure and abundance to ecosystem functioning, but other studies that find no link suggest it may be ecosystem-dependent. Because denitrifying bacteria are polyphyletic, metabolic roles in the nitrogen cycle cannot be determined by identification based on the 16S rRNA gene (Scala & Kerkhof, 1999). One study of soils in Germany, Sweden and Finland (Braker et al., 2012) found community structure similarity to be higher among nosZcontaining communities than among nirK- and nirS-containing communities. This could lead to easier transferability of results from different regions. Because soil denitrifiers are a large group in number as well as diversity, differences in community structure as a whole may provide an indication of the groups that are significant for nitrogen cycling. This may vary between soil types and momentary conditions. 2.4 LH-PCR Length heterogeneity polymerase chain reaction (LH-PCR) has been shown to be a highly reproducible fingerprinting method (Entry et al., 2008). Of the hyper-variable regions in the bacterial 16S rRNA gene, V1 (E. coli positions 72–101), V2 (pos. 176– 221) and V3 (pos. 451–481) show length heterogeneity due to insertions and deletions (Suzuki et al., 1998). These can be amplified together (Tiirola et al., 2003), to give fragment lengths 465–565 bp. The fragments are then separated by capillary gel electrophoresis and quantified. Amplifying the variable regions separately can give additional resolution (Ritchie et al., 2000, Mills et al., 2006; Entry et al., 2008; Entry et al., 2013) but it should be borne in mind that sequences of < 300 bp have been found to be unreliable for phylogenetic assignment (Janssen, 2006). As such, the resulting fingerprints may not represent bacterial community structure in an interpretable way, but ultimately the choice of target should depend on the nature of the sample material and the hypotheses to be tested.
48 2.4.1 Bioinformatics In order to interpret LH-PCR results, they can be combined with cloning and sequencing using the same primers to identify the taxonomy of various peaks. The economy of LH-PCR allows a broader analysis of spatial and temporal variation. Clone libraries are limited to a smaller number of samples, which can be assessed for undersampling by using rarefication curves, but even this needs a deep level of sampling to be carried out. Therefore, though they can be used in conjunction, fingerprinting methods are often used as an alternative to cloning. The growth of publically available sequence data means that sequencing ones own samples can in principle be circumvented as far as its relevance to fingerprint data is concerned. It may one day be possible to sequence a limited clone library, which would be sufficient to define the type of community the sample material belongs to. That is to say, if such community types exist, as much remains to be known about bacterial diversity. In the case of LH-PCR, however, comparison to database sequences has only been carried out once, by Tiirola et al. (2003). Only sequences from cultured bacteria were used, as the majority of sequences from uncultured bacteria could were not classified at the time. For soil communities this is no longer the case, so it may be that LH-PCR has come of age. Bioinformatic tools are only needed to benefit from libraries obtained directly from soil DNA extraction. These should preferably be the result of similar sampling and PCR conditions to those used for the LH-PCR procedure in order to obtain a comparable representation of the bacterial phyla present. A conceivable advantage of using LH-PCR for studying nitrogen cycling is that it gives information on relative abundance differences of the largest bacterial phyla, which mediate the relevant processes. This is particularly useful because it is the relative abundance of denitrifiers in the community that appear to be the best predictor of potential denitrification (Philippot et al., 2009). 2.4.2 Execution and sources of bias The aims of LH-PCR means the reaction has to be carried out with consideration to certain aspects. The thermal cycles are limited to avoid template reannealing and primer exclusion. This occurs when the product concentration exceeds a threshold of 2 nM (Suzuki et al., 1998), which corresponds to a concentration of about 670 ng µl−1 for the fragments amplified by the primers used by Tiirola et al. (2003), assuming a balanced distribution of amplicon lengths. The amplicons of poorly represented bacterial 16S
49 rRNA are less likely to re-anneal and therefore disproportionately more likely to be amplified, leading to under-representation of the dominant species and diversity overestimation (Suzuki et al., 1998). Other systematic biases are harder to avoid, but when kept consistent should not hinder the interpretation of valuable microbial community data. The way DNA is extracted from environmental samples can be expected to affect LH-PCR patterns. As no method is simultaneously ideal for all groups of bacteria, compromises are made for optimal yield. Once obtained, however, the template DNA can be used at a broad range of concentrations with no effect on the community profile, though picogram amounts approach the sensitivity limit, with gradually less product yield at these concentrations (Tiirola, 2002). Cloning and sequencing a sample of the 16S rRNA genes present is often carried out for many types of community analysis. In LH-PCR this can be followed by assignment of the cloned sequences to peaks (Mills et al., 2003; Di Bonito et al., 2013). Bernhard & Field (2000) noticed a difference of 1–2 base pairs between LH-PCR products and their corresponding sequences and noted 3 possible causes: (1) differences in electrophoretic mobility of the samples and standard due to different labels used; (2) variability in the characteristic of Taq polymerase to add adenine to the end of the PCR product; (3) sequence differences may cause migration rates that differ from those of the standard fragments. If cloning is carried out with enough coverage to give an indication of abundance, the sequences can be used to partially correct systematic difference between in silico fragment lengths and those suggested by the standard. 2.5 Targeting functional genes for fingerprinting Using the 16S rRNA gene sequence alone is not sufficient to characterize the denitrification ability of bacteria, which may present a limitation for techniques used to characterize microbial communities based on this technique. LH-PCR has been applied to functional genes as well as housekeeping genes to characterize bacterial communities performing a certain function in the environment. Two examples of this are the methanogens and diazotrophs. In the case of nitrogen fixing bacteria, the Fe subunit of dinitrogen reductase (nifH) has been used as the molecular marker for non-culture based characterization. The principles and points that need to be observed are transferable to studying any functional gene. NifH is present in all nitrogen fixers, which include many loosely related heterotrophs (Zehr et al., 2003). The phylogeny of the gene also closely
50 matches the bacterial phylogeny determined from 16S rRNA gene sequence alignment (Young, 1992). In determining the gene tree, early gene duplications were omitted, as they have been lost from many lineages and have evolved rapidly. Identifying duplications was possible by observing the chromosomal map position of the gene, which is well conserved. In Azotobacter vinelandii there is another copy in the third nitrogenase cluster (anfH) that clusters with Clostridium pasteurianum and Methanococcus lithotrophicus copies of the gene, which suggests the duplication is ancient and precedes the divergence of Eubacteria and Archaea. Transfers were ruled out by the broad range of G+C contents among orthologs (19–94% in the third base position), which were typical of the genomes in which they were found. Because of the proliferation of silent mutations, protein sequences were aligned rather than amino acid sequences. The final point to be made is that regulation of the targeted gene needs to be understood well to know what conditions are likely to produce relevant results. Like many genes, the presence of a nifH sequence does not necessarily imply the activity of its protein product in the community characterized, as this is regulated pre- and posttranslationally (Ueda et al., 1995). 2.5.1 NirK and nirS Nitrite reduction in bacteria is performed by the products of gene nirK and nirS. These genes are used as molecular markers of denitrification activity in soil, though their phylogenies show high levels of incongruence with the phylogeny of 16S rRNA. Studies of nirS distributions in soils have found that abundance correlates positively with soil nitrogen, moisture and pH. The nirS/nirK ratio responds more strongly than either gene individually, increasing with NO3− , moisture and pH (Philippot et al., 2009). This may be due to a subset of related denitrifiers that responded uniformly to these features that were targeted with the nirS primers. It has been mentioned that NirK is known to have at least one alternative function. If this is widespread for nir genes, the selective pressure that produced these other activities may result in reduced congruence with the accepted phylogeny. As a facultative pathway, the genes involved may be under different levels of selective pressure depending on how necessary they are in a given environment. These may explain why phylotype diversity can be either higher or lower than species diversity (Jones et al., 2008).
51 2.5.2 Horizontal gene transfer and nosZ Many functional genes readily undergo horizontal gene transfer (HGT), but they vary in how much they historically have done so. This will affect how much the targeted gene phylogeny reflects the 16S rRNA phylogeny. Genome localization and G+C content are important
indicators.
The
program
Colombo
(http://www.tcs.informatik.uni-
goettingen.de/colombo-sigihmm) has been used to assess whether a gene has been obtained by HGT (Palmer et al., 2009). For a gene such as nosZ, which evolved in the common ancestor of bacteria and archaea, horizontal transfers can be hard to detect (Jones et al., 2008). What can be said is that nosZ does not appear to have been transferred between distantly related bacteria in recent times (i.e. in the last 50 million years). It has been found that if two isolates have a nosZ genetic sequence similarity of greater than 80% or an amino acid sequence similarity of greater than 86%, they are very likely to belong to the same species. This is a higher threshold of similarity than the nitrate reductase gene narG, indicating that it is more highly conserved. The nosZ sequences isolated from various genera generally cluster together (Palmer et al., 2009) and one study of partial sequences found a greater congruence with 16S rRNA-based classifications than nirS, nirK, and norB, despite the existence of plasmid-borne nosZ in some strains. The Alphaproteobacteria S. meliloti and Brucella spp., on the other hand, have genes for all 4 reductase enzymes in a “denitrification island” -like cluster (Jones et al., 2008). This arrangement increases the likelihood of successful HGT, as the recipient acquires all the machinery required to carry out the pathway effectively, however this is not necessarily the case for modular pathways such as denitrification.
52
Experimental work 3 Research objectives An experimental field was set up in July 2009 to test the use of LH-PCR in monitoring the bioremediation of oil contamination, while studying the involvement of BNF and plant-growth promoting bacteria in accelerating the process. In this study conducted in 2012, I used LH-PCR as a fingerprinting method to characterize the total bacterial communities present in the soil of a legume and a grass planted for this field trial. This research was a part of the Legume Futures study of bioremediation, run under the EU Framework Programme 7. It was conducted in the research group of Professor Kristina Lindström, where the interactions between legumes and rhizobia are studied. My work was complemented by measurements made by a student from environmental soil science studying more proximal controls of emissions. Lindström et al. (2003) reviewed the potential of this symbiosis to improve the degradation of hydrocarbon contaminants in legume-mediated phytoremediation. Rhizobia raise the nitrogen status of soil, increasing demand for carbon substrates in the bacterial community, and some strains are capable of metabolizing pollutants themselves. In the field study, the bioremediation potential of fodder galega (Galega orientalis) was assessed. This is a perennial herbaceous legume that has been studied as a temperate forage plant (Fairey et al., 2000). The preferred symbiont of G. orientalis is Neorhizobium galegae (Radeva et al., 2001) and plots were inoculated with these bacteria during the establishment of the trial. The strain that was used (HAMBI540) encodes nirK and norB but not nosZ (Österman et al., 2014). As a result, it could potentially contribute to field emissions, particularly as native soil rhizobia are unable to induce nodulation in G. orientalis (Tas et al., 1996). As a negative control for rhizobial symbiotic activity in the soil, brome grass (Bromus inermis, Poaceae) was also grown. To study whether rhizobial BNF can benefit other plants, an intercrop of B. inermis and G. orientalis was planted. This grass was selected as it has been reported to grow well with Galega spp. (Fred Stoddard, University of Helsinki, email message to author, 7 Apr 2014). In the original study half of the plots were contaminated with oil, but those treatments are not dealt with here. Instead, I focus on the nature of the relationships of bacterial community structure with
53 nitrogen fluxes, and the effects of plant growth, intercropping and fertilization. The method I used (LH-PCR) was introduced to the group by Dr. Anu Mikkonen (Wallenius et al. 2010) and has been already shown during this field trial to be highly informative for monitoring the bioremediation process (Yan, 2012). The way bacterial communities are altered by hydrocarbon contamination can be visualized, showing a decrease in overall diversity. As the process progresses, the recovery of the community can be followed as it returns to a state that resembles a non-contaminated control soil community. The involvement of bacteria represented by an individual peak could be quantitatively assessed by its correlation with hydrocarbon dissipation (Mikkonen et al., 2011). Sequencer technology has developed in a way that has improved the resolution attainable in LH-PCR. The research question I attempt to meet is whether the sensitivity of LH-PCR fingerprints is sufficient to detect contributions of bacterial groups to various processes by relatively small differences in their abundance. With the statistical power offered by the economy of the method, I investigate differences in community composition that could reflect the phylogeny of active denitrifiers, one of the largest functional groups present in the soil to still defy molecular characterization. To do this, I took a reductive approach to the LH-PCR curve data, isolating each peak as an entity to itself. In doing so, I resolved how anthropogenic effects on the bacterial community must be viewed against the backdrop of spatial and temporal variation irrespective of experimental manipulation. In order to improve coverage of the bacterial community, Proteobacteria were characterised using primers that target Alpha-, Beta-, and Gammaproteobacteria individually. By targeting a phylum that includes many characterized denitrifiers we can improve the odds of detecting changes in their abundance. Ultimately, the aim was to elucidate the distal controls of vegetation and soil Nr availability and inputs on N2O emissions. That is to say, how agronomic practices bring about changes in the bacterial community structure that reflect activities that contribute to the production and consumption of N2O in the soil. As such, I hope my results can contribute to the development of sustainable agronomic management practices for maintaining soil fertility and health.
54
4 Materials and methods 4.1 Experimental design The field experiment had a randomized complete block design in four blocks. In this study, plots comprising 5 treatments were included over the four block repeats. Each plot measured 1.5 x 2.5 m. Three of the treatments were either sewn with Galega orientalis (25 kg/ha) (treatment Gao), Bromus inermis (35kg/ha) (treatment Bri), or an intercrop of the two species (6 kg/ha G. orientalis and 26 kg/ha B. inermis) (treatment Int). There were plots identical to the Bri treatment that were fertilized with 60 kg ha−1 urea dissolved in water on 8 May (treatment BriN). Other plots received water with no fertilizer. A bare fallow treatment (Fal) was included as a control. This had been treated with glyphosate twice in 2009 and 2011, but was otherwise weeded mechanically. On 11 June, the vegetated plots were harvested. Composite samples were taken 12 times, every 1–2 weeks over the growing season with an auger from 4 random points in each plot to a depth of 25 cm. The samples were passed through a 5 mm sieve on the same day before freezing until analyses were conducted. 4.2 Analysis of soil 4.2.1 Soil pH and organic matter Soil was prepared for pH measurement by mixing 20 g field-moist soil with 50 ml 10 mM CaCl2 solution on a rotary mixer (Certomat R® and Certomat H®, B. Braun, Germany) at 200 rpm for 1 hr. After the soil had settled, pH of the solution was measured (420A+, Thermo-Orion, USA). The organic matter content was estimated for comparative purposes within this experiment. Samples of ~10 g fresh soil were dried at 105°C overnight, and burnt at 550°C for four hours. The difference in masses of the dried and burnt soil is recorded as the organic matter content and calculated as a percentage of dry soil weight (% d.w.). 4.2.2 Potential denitrification The acetylene inhibition assay was carried out based on the method described by Tiedje et al. (1989). Soil slurries of 25 g field-moist soil and 25 ml 1mM glucose, and 1mM KNO3 were mixed in 125 ml bottles in an anaerobic cabinet set to 25°C and the headspace was flushed 2 times with a 25 ml syringe before sealing with a septum lid. The bottles were then pressurised with 10 ml of cabinet air and removed from the
55 cabinet. A volume of 10 ml acetylene was added to each bottle and they were placed on an end-to-end shaker. This was set to agitate the bottles at 125 rpm. After 30 min, 60 min and 90 min the headspace was homogenized by pumping with a syringe and 7 ml was transferred to helium vacuumed Exetainer 3 ml vials. Samples were stored at 4°C until analysis. Analysis of headspace composition was carried out by gas chromatography (Agilent 6890N, Agilent Technologies, USA). The flame ionisation detector (FID) measured carbon dioxide (CO2), O2, N2, C2H2 and methane while an electron capture detector (ECD) measured N2O. The standard gas mixture used for calibration contained no C2H2 but did include C2H4. These were assumed to produce the same FID signal for the purposes of calculating the volumes of the other gases in the samples. Measurements were normalized so that the sum of gas concentrations in a sample was 1 million ppm. The volume of each headspace was measured by filling the bottle and inside of the septum with water and weighing the additional mass. Because the headspace was pressurized when sampled, the atmospheric pressure-equivalent volume of the samples taken and the headspace at each stage was calculated to determine the N2O produced. 4.3 LH-PCR Total soil DNA was extracted from 0.5 g soil samples taken on five sampling dates (4 May, 1 Jun, 14 Jun, 28 June, 22 Aug). The UltraClean™ Soil DNA kit (MoBio, USA) was used according to the long version of the alternative protocol for maximum yields. A MoBio vortex adapter was used for cell disruption. DNA size, yields and purity were approximated by agarose gel electrophoresis and by spectrophotometry (NanoDrop ND1000, Thermo Fisher Scientific, USA). 4.3.1 PCR conditions The primers for amplification of the 16S rRNA gene were fD1 (5′–AGA GTT TGA TCC TGG CTC AG–3′) and 5′FAM-labelled PRUN518r (5′–ATT ACC GCG GCT GCT GG–3′) as used by Tiirola et al. (2003) and modified by Mikkonen et al. (2011). The 25 µl reaction mixture included 0.5 µl template DNA, 0.2 mM each dNTP (Finnzymes, Finland), 0.3 mM each primer, 0.05% bovine serum albumin, and 1.25 U Taq DNA polymerase (recombinant, Thermo Scientific) with its proprietary buffer (10 mM Tris-HCl, 50 mM KCl, 1 mM MgCl2, 0.08% Nonidet P40). Thermal cycling was carried out with a PTC-200 thermal cycler (MJ Research) using the calculated control
56 setting. The reaction was initially held at 94˚C for 5 min followed by 30 cycles of denaturation for 45s at 94˚C, annealing for 1 min at 55˚C, and extension for 1 min at 72˚C. The final round of polymerization was extended by 5 min, after which the reaction was brought to 4˚C. Group-specific PCR reactions were performed in 25 µl reaction mixtures with the same conditions with some changes to reaction mixture and thermal cycling. The buffer included 75 mM Tris-HCl, 20 mM (NH4)2SO4, 1 mM MgCl2, 0.01% Tween 20. The primer concentrations were lower at 0.2 mM of each primer. The same forward primer was used and the reverse primers were, according to target proteobacterial class: Alf684r (TAC GAA TTT YAC CTC TAC A; Mühling et al., 2008), Beta682r (ACG CAT TTC ACT GCT ACA CG; Ashelford et al., 2002), and Gamma871r (ACT CCC CAG GCG GTC DAC TTA; Mühling et al., 2008). The amplification of the alphaproteobacterial gene was done by touch-down PCR with the annealing temperature 55˚C in the first round, and decreasing by 1˚C for the next 4 rounds, followed by 20 cycles with annealing at 50˚C. The final extension step lasted 10 min. The betaproteobacterial gene was amplified over 25 cycles with an annealing temperature of 63˚C. The gammaproteobacterial amplification was carried out with an initial annealing temperature of 60˚C, which was reduced by 1˚C in each cycle until it reached 55˚C, which it was maintained at for 20 cycles. For nested PCR, group-specific PCR products were diluted 1000-fold and 0.5 µl was used as a template for a regular LH-PCR reaction described in the last paragraph. 4.3.2 Fingerprint visualization After each reaction, LH-PCR products were visualized by agarose gel electrophoresis to confirm they conformed to expected product size. Capillary gel electrophoresis was carried out by the Helsinki University Biotechnology Institute using the GeneScan™ 500 ROX™ size standard. The approximate active regions of the resulting curves were normalized using the Bionumerics software package (version 6, Applied Maths, Belgium) and checked for irregularities. Samples that were under- or over-loaded, or poorly aligned were subject to a second round of capillary electrophoresis. 4.3.3 In silico analysis The expected amplicon lengths for the primers used was explored for different phylogenetic groups of soil bacteria. The taxonomic assignments for submitted
57 sequences were referred to from the Ribosomal Database Project (RDP; Wang et al., 2007). These assignments tend to be stringent, which suits the purposes of this study, as to include phenotypically uncharacterized lineages in taxonomic definitions does not facilitate the interpretation of results. Predicted amplicon lengths were calculated by aligning ribosomal sequences with the primer sequences using the BLAST algorithm on the NCBI website. In silico LH-PCR curves were constructed based on accessions derived from one study, where Hartmann and Widmer (2006) sampled the top 20 cm of an agricultural soil in Switzerland under various fertilization and agrochemical regimes. Almost identical primers were used as for this study, and the soil pH was slightly higher than 6.0 on average (Maeder et al., 2002). In addition to the clone libraries compiled for that study, the predicted amplicon length of Neorhizobium galegae was also ascertained. 4.4 Statistical analysis Accepted curve data was exported to Microsoft® Excel and curve amplitudes were normalized to a standard total sum of data points across the range where a greater than background level of curve density was found. Where multiple data sets of equivalent quality were available for a given soil sample, these were averaged. Average curves for treatments on each block and at each date were calculated and peak locations identified from these. Peaks heights were then retrieved from each fingerprint, and imported into IBM SPSS Statistics (version 21, IBM, New York, USA). The peak height was calculated as the average of the maximum and second-highest readings in the peak range. The lengths of the fragments producing the peaks was estimated in relation to the standard fragments, with appropriate rounding and acknowledging that apparent < 1 bp differences between fragment lengths were possible due to differences in fragment purine content (Kaplan & Kitts, 2003). After this stage raw curve data was only used to calculate Pearson distance between peaks averaged for each treatment at each sampling date. Neighbour-joining trees were built in MEGA (version 5.2.2, Pennsylvania State University USA). 4.4.1 Ecological indices Beta diversity was measured for each treatment as the variation in the peak heights alternately over the 4 blocks and 5 sampling dates. It was calculated as the average Pearson distance of peak heights between the sampling dates for each block, then between blocks for each sampling date. The Shannon-Weiner index was used to estimate
community
diversity
from
each
fingerprint.
The
formula
is
58 𝐻 = – (𝑃! )(log ! (𝑝! )), where 𝑝! is the proportion of the ith phylotypes (Dunbar et al., 1999). Each peak of the normalized curves represented a phylotype, and was measured by its height as opposed to area under the peak. 4.4.2 Analysis of variance (ANOVA) Differences in the ecological indices between treatments were tested by two-way ANOVA with treatment and either block or sampling date as the main effects. Other forms of ANOVA were used to analyse how peak heights varied over the sampling dates and how they were affected by treatments. Confidence interval (CI) values are given, derived from the ANOVA model. The accepted type I error rate (α) for this study was P < 0.05. In addition to these main effects, variance was also attributed to the random block effect. Significant interactions were subject to simple effects analysis where appropriate. 4.4.3 Repeated measures ANOVA To detect differences in the results between sampling times, the repeated measures generalized linear model of SPSS was used with time as the within-subject factor, and with treatment and block as the between-subject variables. The model specified included sampling date, treatment and block as main effects. The results of Mauchly’s test was used to test the assumption of sphericity. Where P < 0.05, the GreenhouseGeisser correction (ε) was applied to the degrees of freedom when testing the significance of F-values. Repeated measures ANOVA (RM ANOVA) does not tolerate missing values, so the model was fitted twice, omitting either the case or the date with missing values. Testing of multiple peaks for effects of sampling time on the microbial community could lead to an inflation of type I error. To control this, the false discover rate principle was used, employing the Benjamini-Hochberg (BH) procedure to determine the familylevel critical P-value. In a family of m tests, unadjusted P values were ranked in !
increasing order. The largest rank k was found, for which 𝑃(!) ≤ ! 𝛼. The H0 was rejected for tests of ranks 1, 2, …, k. For these tests, the P-value was reported in the !
adjusted form as 𝑃× ! . Where a significant effect of sampling date was found, the Bonferroni-adjusted pairwise comparisons were examined to see if all significant comparisons were included
59 in the model. Only if all dates and cases needed to be included in the model in order to record significant differences were missing values replaced with imputed ones. The multiple imputation module in SPSS was used for this. To avoid type I error, the effects of imputation and data omission on group averages and standard deviations were confirmed to be conservative with respect to rejection of the null hypothesis. The effect size of sampling date was calculated by the following formula (Field, 2013; p. 566):
! 𝜔!"#$%&'( !"#$
𝑘−1 MS! − MS! 𝑛𝑘 = , MS! − MS! 𝑘−1 MS! + + 𝑛𝑘 MS! − MS! 𝑘
where k represents the number of dates modelled, n represents the number of plots, MSM is the model mean square of the sampling date, MSR is the mean square of the within-subject residual, and MSB is the between-subject mean square. 4.4.4 Two-way factorial ANOVA and multivariate ANOVA Experimental results were analysed as a 2-way factorial analysis of variance (ANOVA) with five treatments and four block repeats. The five sampling times were considered as a split-plot factor. The univariate general linear model in SPSS was used to find main effects and interactions. The data were modelled by the following formula: 𝑦!"# = 𝑎𝑣𝑒𝑟𝑎𝑔𝑒 + 𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡! + 𝑠𝑎𝑚𝑝𝑙𝑖𝑛𝑔 𝑡𝑖𝑚𝑒! + 𝑏𝑙𝑜𝑐𝑘! + 𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 ∙ 𝑡𝑖𝑚𝑒!" + 𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 ∙ 𝑏𝑙𝑜𝑐𝑘!" + 𝑡𝑖𝑚𝑒 ∙ 𝑏𝑙𝑜𝑐𝑘!" + 𝑟𝑒𝑠𝑖𝑑𝑢𝑎𝑙!!" , where yijk are observations. The peak variables are expected to inherently show a high degree of covariance because adjacent peaks can affect each other and an increase in any peak will always correspond to a proportional decrease in curve density elsewhere, and vice versa. Some peaks of these peaks may even represent ecologically similar groups, so viewing them separately may be equivalent to having two measures of the same guilds. When effects are spread over several different measures there are issues with both multiple testing (increased Type I error rate) and a reduction of statistical power. To improve power, the peak data was subject to multivariate analysis of variance (MANOVA). This uses the model displayed above but also takes into account
60 correlations between the dependent variables. For this purpose, the 4 May BriN-plots were omitted, as these samples were taken before application of the fertilizer. This was followed by discriminant function analysis (DA) to identify the most important peaks for discriminating between samples of different treatments and sampling dates. For this analysis, the 4 May BriN data was included as additional Bri data. 4.4.5 Analysis of covariance In some cases, models based on ANOVA were improved by correcting for confounding factors. Where peak data correlated with measured soil characteristics, these could be used as covariates. The size of the sample taken in this study is not sufficient for more than one covariate to be tested at a time, as the method is based on testing effects on estimates of adjusted means. In order for these estimates to be stable, the number of covariates modelled (C) must be limited such that
!!(!!!) !
< 0.10, where J is the
number of groups and N is the sample size (Stevens, 2002, pp. 366–7). As such, a 2-way factorial study of 20 groups (experimental plots) requires an N of 200 for stable covariates-adjusted means, and observation of ANCOVA results here should be cautious and purely exploratory. Furthermore, the covariate of the repeated measures ANCOVA design is specified for each subject and cannot be specified at the subject by sampling date level. In order for such a covariate to be informative, it must have more variation between subjects that between repeated measurements so that a subject-level average can be calculated for this purpose. Multivariate analysis of covariance (MANCOVA) was used to screen soil properties that correlated with a greater number of peaks, against treatment effects on peak heights. Effect sizes were estimated as 𝜔2 values according to the following: ! 𝜔!""!#$
=
𝜎 !!""!#$ 𝜎 !!"!#$
,
where 𝜔 2 is effect size, 𝜎 !!""!#$ is effect variance, and 𝜎 !!"!#$ is total variance. Variances were estimated from ANOVA means squared values according to Howell (2012, pp. 437–9). For the different forms of ANOVA carried out, the fit of the model was calculated according to Quinn and Keough (2002; p. 139): adjusted 𝑅! = 1 −
SS! 𝑛 − 𝑝 + 1 , SS! 𝑛 − 1
61 where SSR is the residual sum of squares, SST is the total sum of squares, n is number of measurements, and p is the number of parameters included in the model. 4.4.6 Correlations of peak data with soil measurements Peaks were tested for Pearson correlations with seven soil characteristics – pH, moisture, SOM, DEA, NO3− content, NH+4 content, and N2O emissions. Measurements of NO3− and N2O emissions were tested as log-transformations to limit the influence of outliers. These were calculated as log10(x mg NO3− -N kg−1 dry soil) and log10(y g N2ON day−1 ha−1 + 10) respectively. To determine whether there was a lag-time between community structure and N2O emissions produced, correlations were also calculated between peak data and emissions from subsequent sampling dates. For the purpose of limiting type I error rates, correlations of data with peaks of each LH-PCR fingerprint type were considered a separate family of tests and the BH procedure was employed to determine to cut-off P-value.
62
5 Results 5.1 Soil properties 5.1.1 Soil pH and organic matter The soil pH of the experimental plots was generally around 5.9. It varied between sampling dates (P < 0.001), decreasing over four sampling dates from a maximum of 5.94 on 10 May by 0.11 (± 0.096, 95% CI) pH units to a minimum of 5.83 on 14 June. This difference was considered inconsequential, so a treatment by block interaction was tested for in a two-way ANOVA. This interaction was strong (F (12, 95) = 20.2), and block differences in pH were greatest between Fal plots (P < 0.001), ranging from two plots of pH 5.6 to one plot of pH 6.3. Blocks 2–4 were quite homogeneous and the Fal plots were the only ones with anomalous pH. The exception was block 1, where the BriN plot stood out with the highest pH, 6.1 (P < 0.001). Smaller differences were found between blocks for each treatment except for the Bri plots, which had a consistent average pH of 5.8–6.0 (95% CI). SOM increased slightly by 0.3 ± 0.2 (95% CI) % from 4.8 % on 1 June to 5.2% on 8 June (F (2, 18) = 4.9, ε = 0.58, P < 0.05). There was no significant effect of treatment. Two-way factorial ANOVA indicated a treatment by block interaction (F (12,37) = 2.31, P < 0.05) and affected plots are shown in Figure 4. Block:
7%
Soil organic matter
6% 5%
1
2
3
4
c
bc
bc
ab
ab
a
a
4% 3% 2% 1% 0% Bri
BriN
Gao
Int
Fal
Figure 4 Soil organic matter of experimental plots. Error bars indicate 95% CIs and lower-case letters indicate Bonferroni-adjusted significant differences between plot means from a given block or treatment. (I.e. plots of differing block and treatment are not directly comparable.)
63 5.1.2 Denitrification enzyme activity Over all five sampling dates there was no significant treatment effect on DEA but the repeated measures ANOVA model produced an effect of sampling date (F (4, 48) = 31.9, P < 0.001). DEA was measured to be highest on the 10 May and 28 June. Of individual factorial two-way ANOVA analyses for each date, the only significant effect was found on 28 June (F (4, 12) = 7.3, P < 0.01; Figure 5), when the DEA of the Fal treatment was lower than that of the BriN treatment. On this date DEA was significantly correlated with log-transformed soil NO3− content (r = −0.505, P = 0.023) but there was no significance covariance for the effects. Over all dates DEA correlated weakly with soil NH+4 (r = 0.26, P < 0.05), but screening for this as a covariate did not bring out a significant treatment effect.
Figure 5 Average DEA measurements of soil sampled on (a) all 5 sampling dates, and (b) 28 June. Error bars indicate 95% CIs and lower-case letters indicate Bonferroni-adjusted significant differences.
64 5.2 Universal LH-PCR Pearson correlation of non-selective LH-PCR curves indicted clustering of the Fal bacterial communities at 3 disparate sampling dates (Figure 6). At the bottom of the tree are Bri plots at 2 sampling dates clustered with 2 other samplings, which together form an out-group with a low level of similarity to each other. Four of the treatments showed a high level of similarity on 1 June, clustering with the Int treatment sampled on 14 June. This branch of the neighbour-joining tree was also close to 3 treatments sampled on 28 June. There was no significant difference in Shannon diversity index or beta diversity of soil communities for the treatments over the experimental field or sampling dates. BriN (4 May) 0.002
Gao (4 May) Gao (14 June) Fal (4 May) Fal (14 June) Fal (22 August) Gao (22 August) Int (22 August) Bri (14 June) BriN (14 June) Fal (28 June) Bri (4 May) Int (4 May) Bri (28 June) BriN (28 June) Int (28 June) BriN (1 June) Int (1 June) Gao (1 June) Fal (1 June) Int (14 June) BriN (22 August) Gao (28 June)
Bri (1 June) Bri (22 August)
Figure 6 Pearson correlation-based neighbour-joining dendrogram based on average non-selective LH-PCR curves.
65 5.2.1 Peak assignment The average of the 95 LH-PCR curves obtained is displayed in Figure 7. The active area was defined as the 465–545 bp range for the purposes of normalization. Some clusters of peaks with small base-pair differences were difficult to distinguish and displayed similar patterns in most fingerprints. These (e.g. curve density comprising peak 6) were designated as a single peak. In other groups the relative prominences of the peaks varied (e.g. peaks 2 and 3), and they were designated as separate peaks. The measurements showed a high level of positive correlation between some adjacent peaks, and comparatively high correlations between peaks of fragments with high base-pair differences. Some of the more significant correlations between distant peaks are shown in Table 1. Peak correlations with soil characteristics and N2O emissions are shown in Table 2. 16
1100
11
Fluorescence signal amplitude
1000
19 20
900 800 700 600
12 13
500
1
400
2
10
3
300 200
4
5
6
7
8
22
17 18 14
15
21
9
100 0 465 470 475 480 485 490 495 500 505 510 515 520 525 530 535 540 545
Fragment length / base pairs Figure 7 The active area of the average normalized LH-PCR curve. The peak regions are numbered and indicated with glowing bands. Table 1 The correlations between several LH-PCR peak heights. Adjusted P < 10−8. Peaks 3 20 3 10 4 17 6 16 11 16
Pearson’s correlation/ r (93) 0.58 0.68 −0.68 −0.75 −0.71
66 Table 2 Pearson’s correlations of universal LH-PCR peak heights with soil characteristics and N2O emissions. Asterisks indicate adjusted P-values *
