Abstract
Methane is involved in a number of chemical and physical processes in the Earth's atmosphere, including global warming1. Atmospheric methane originates mainly from biogenic sources, such as rice paddies and natural wetlands; the former account for at least 30% of the global annual emission of methane to the atmosphere2. As an increase of rice production by 60% is the most appropriate way to sustain the estimated increase of the human population during the next three decades3, intensified global fertilizer application will be necessary3: but it is known that an increase of the commonly used ammonium-based fertilizers can enhance methane emission from rice agriculture. Approximately 10–30% of the methane produced by methanogens in rice paddies is consumed by methane-oxidizing bacteria associated with the roots of rice4, 5; these bacteria are generally thought to be inhibited by ammonium-based fertilizers, as was demonstrated for soils6, 7, 8 and sediments9, 10. In contrast, we show here that the activity and growth of such bacteria in the root zone of rice plants are stimulated after fertilization. Using a combination of radioactive fingerprinting11 and molecular biology12 techniques, we identify the bacteria responsible for this effect. We expect that our results will make necessary a re-evaluation of the link between fertilizer use and methane emissions, with effects on global warming studies.
Methanotrophic bacteria utilize methane as the sole carbon and energy source; they are subdivided into types I and II, on the basis of phylogeny, physiology, morphology and biochemistry, including characteristic phospholipid ester-linked fatty acids (PLFA) in their cell membranes13. These bacteria cannot be active in flooded rice soils without oxygen derived from the rice plant. This is because of their obligatory aerobic metabolism. These obligatory aerobic bacteria live in close association with the roots of wetland plants (for example, rice) which provide them with necessary oxygen14, 15 to oxidize methane which diffuses from the anoxic bulk soil to the rhizosphere. Here we examined the effect of fertilization on methane oxidation using 'microcosms' planted with rice; in each 'microcosm' there was a root-soil and a bulk-soil compartment (see Methods). The application of urea or (NH 4)2HPO4 (200 or 400 kg N ha-1) stimulated rather than inhibited methane-oxidizing activity in the root zone of rice (Fig. 1a). All samples from planted, fertilized microcosms displayed initial methane-oxidizing activity, whereas no initial activity (0–12 h) could be detected in soil samples from the rhizosphere of the unfertilized and the unplanted microcosms (Fig. 1a). The induced activities were also significantly increased by fertilization (Kruskall–Wallis test, P<0.001) and the presence of rice plants (Kruskall–Wallis test, P<0.05) (Fig. 1a). With the unfertilized samples it took 50 h before any methane depletion could be detected (Fig. 1b, filled triangles). However, ammonium addition to parallel soil slurries from unfertilized microcosms led to an immediate activation of the methane-oxidizing bacteria ( Fig. 1b, open triangles), demonstrating the essential role of ammonium in the consumption of methane in rice soil.
Figure 1: Methane-oxidizing activities in rice soil.
Shown are the effects of ammonium-based fertilization and rice plants on
methane oxidation rates of unplanted and rhizospheric soil from microcosms
that were either unfertilized or supplemented with urea (200 or 400 kg N ha
-1) or (NH4)2HPO4 (200 or 400 kg
N ha-1). The bars in a represent the arithmetic means
of the initial (black bars) and induced (hatched bars) oxidation rates of
four replicate microcosms (
1 s.d.). ND, assays without a detectable
initial oxidation rate. b, The effect of ammonium addition on methane
oxidation of rhizospheric soil slurries from unfertilized microcosms. Filled
triangles, the non-supplemented slurries; open triangles, the methane depletion
in the slurries supplemented with ammonium (end concentration, 2 mM). All
values represent the arithmetic means of four replicate microcosms (
1 s.d.).
The stimulating effect of fertilization was also demonstrated in experiments
with larger (1.5 litres of soil), non-compartmented microcosms cultivated
in a greenhouse. Methane oxidation increased substantially after urea application,
as detected both by in situ rate measurements and an increased
13C/12C ratio (less negative 
13C
value) of the emitted CH4 (M. Krüger, P. Claus & P.F.,
manuscript in preparation).
The bacteria responsible for the observations of accelerated methane oxidation were identified using radioactive fingerprinting and molecular biology techniques. In all treatments, radiolabelled methane was incorporated into PLFA fractions that contained fatty acids typical of type I and II methane-oxidizing bacteria (Fig. 2). Type II methane oxidizers dominated methane metabolism in unplanted, unfertilized soil, as indicated by recovery of 14C-PLFAs mainly in fraction 11 (Fig. 2a). In contrast, long term-fertilization of planted soil with either urea or (NH 4)2HPO4 resulted in the activity of both type I and II methane oxidizers in the rhizosphere (Fig. 2b–f ). The apparent activation of type I methane-oxidizing bacteria by fertilization was confirmed by the 14C-PLFA fingerprints obtained after addition of ammonium to soil slurries retrieved from the rhizosphere (Fig. 2b). More radioactivity was recovered in PLFA fraction 5 after addition of ammonium to such samples from unfertilized microcosms (Fig. 2b).
Figure 2: 14C-labelled phospholipid ester-linked fatty acid ( 14C-PLFA) fingerprints of methane-oxidizing bacteria from soil samples incubated with 14CH4.
The fingerprints were obtained from unplanted microcosms (a), and
the rhizosphere compartments from planted microcosms that were supplemented
with urea at 200 kg N ha-1 (c) or 400 kg N ha
-1 (d), or with (NH4)2HPO4
at 200 kg N ha-1 (e) or 400 kg N ha-1
(f). b, The fingerprint from soil originating from the rhizosphere
compartment of unfertilized microcosms that were supplemented with ammonium
(compare Fig. 1b) in post-harvest slurry incubations.
The graphs show the percentage of radioactivity detected in each PLFA fraction.
14CH4 incorporation into PLFAs that eluted mainly in fraction
5 (for example, 16:1
8, 16:17, 16:15 and 16:0) indicate
methane oxidation by type I methane oxidizers, whereas radioactive PLFAs detected
mainly in fraction 11 (for example,18:19, 18:18, 18:17,
and 18:0) indicate oxidation dominated by type II methane oxidizers. Bars
indicate the mean ( 1 s.d.) of two subsamples taken from 1
out of 4 replicate microcosms per treatment. The fingerprint in b corresponds
to the ammonium-supplemented slurry shown in Fig. 1.
The metabolism of 14CH4 was defined as dominated
by type I methane-oxidizing bacteria when the ratio of fraction 5: fraction
11 was greater than 1.1. If this ratio was lower than 1.1, type II methanotrophs
were assumed to dominate. This differentiation is based on laboratory experiments
with pure cultures of type I and II methane-oxidizing bacteria27.
The total abundance of the PLFAs specific for methanotrophic bacteria (16:18c
for type I and 18:18c for type II) can be regarded as a measure of
methanotrophic biomass16. (Fatty acids are designated by the
total number of carbon atoms. The degree of unsaturation is indicated by a
number separated from the chain length by a colon and is followed by
xc, where x indicates the position of the double bond nearest to
the aliphatic end (). The c indicates a cis-position of the
double bond on the molecule.) Our data revealed that fertilizer addition to
planted microcosms resulted in a 2–3-fold increase in the abundance
of PLFAs specific for methanotrophic bacteria (Fig. 3).
This was supported by using a most probable number approach where higher (Kruskall–Wallis
test, P<0.001) numbers of culturable methanotrophs were found in
fertilized microcosms17. Fertilization resulted mainly in growth
of type I methane oxidizers, as indicated by a ninefold increase in the type-I-specific
PLFA biomarker in microcosms fertilized with 400 kg N ha-1
of urea or (NH4)2HPO4 (Fig.
3). The type-II-specific PLFA 18:18c increased only 2–3-fold
after fertilization, but this PLFA remained the dominant methanotrophic biomarker
in fertilized samples. The dominance was most pronounced in unplanted and
unfertilized samples where the type-II-specific biomarker was 7 times more
abundant than type I biomarkers (Fig. 3).
Figure 3: Effect of fertilization on PLFA abundance in rice soil.
Shown are the total abundances of the type-I-specific PLFA 16:18c
(hatched bars) and the type-II-specific PLFA 18:1 8c (black bars) in
samples retrieved from unplanted soil and from the rhizosphere compartments
of planted microcosms that were either unfertilized or supplemented with urea
(200 or 400 kg N ha-1) or (NH4)2HPO
4 (200 or 400 kg N ha-1).
By combining the PLFA profiles and denaturing gradient gel electrophoresis (DGGE) separation of 16S rDNA amplified by the polymerase chain reaction (PCR), a higher phylogenetic resolution of the methane-oxidizing community in rice soil was obtained. The DGGE banding patterns (Fig. 4) and the respective sequences (Fig. 5) demonstrate that type and dose of fertilizer had little effect on the species composition of the methane-oxidizing community. Type I bands (RRI 1–6, RBI 1) were most closely related to Methylobacter species (Fig. 5), whereas type II bands (RRII 3–7) were related to Methylosinus and Methylocystis species. All the main bands, methanotrophic as well as non-methanotrophic, were highly related to sequences and clones from rice-field soil incubations12 (Fig. 5). DGGE bands related to type II methane oxidizers were equally intense in all samples, whereas type-I-related DGGE bands were distinctly more intense in the rhizosphere samples (Fig. 4). These intensity differences are the consequence of differences in the original number of target sequences, and hence of the numbers of the respective methane oxidizers in the soil12. Thus, as well as ammonium, the presence of the rice plant is an essential factor for type I methane-oxidizing bacteria to proliferate. This finding is supported by the PLFA data.
Figure 4: Effect of the presence of rice plants and fertilization on the composition of the methane-oxidizing community in rice soil microcosms.
Displayed are DGGE (denaturing gradient gel electrophoresis) banding patterns
obtained after amplifying DNA extracted from soil and enrichment cultures
from the highest positive MPN dilutions with an MB10
16S rDNA primer
set targeting type I methanotrophs (top panel) and an MB9
16S rDNA primer
set targeting type II methanotrophs (bottom panel). The samples were retrieved
from unplanted soil and from the rhizosphere and bulk-soil compartments of
planted microcosms that were either unfertilized or supplemented with urea
(200 or 400 kg N ha-1) or (NH4)2HPO
4 (200 or 400 kg N ha-1). The DGGE patterns of the rhizosphere
and unplanted samples correspond to the respective PLFA patterns in Fig. 2. Bands designated with acronyms were excised from the
gels, re-amplified and sequenced as described elsewhere12. The
phylogenetic positions of the derived sequences are shown in
Fig. 5.
Figure 5: Phylogenetic tree.
Shown is the relationship of partial 16S rDNA sequences retrieved from
DGGE gels (see Fig. 4) with the most closely related
members of the -, 
- and -proteobacteria. The scale bar
indicates the estimated number of base changes per nucleotide sequence position.
'c' and 'ox' are clones from rice field soil incubated
without and with methane, respectively12. MOB, methane oxidizing
bacteria.
Enrichments from the highest positive dilutions of most probable number (MPN) counts contained only type II methanotrophs (Fig. 4 , lower panel). These bands (ERRII 1–3, ERBII 1–3) matched the dominant bands obtained from soil samples, confirming the numerical dominance of type II methanotrophs in rice soil that was indicated by the PLFA analysis (Fig. 3).
Our results show that ammonium-based fertilization does not necessarily inhibit the consumption of methane in soils and sediments as has been repeatedly reported6, 7, 8, 9, 10. The absence of inhibition may be explained by the high methane availability11 in rice paddies, which counterbalances possible competitive inhibition by ammonium18. Furthermore, fertilizer-associated nitrite toxicity7 or salt effects19, 20 on methane oxidation may be eliminated by plant uptake of these ions. Such a sink for toxic substances is missing in upland soils where methane concentrations are low, explaining the reported inhibition of methane oxidation by ammonium in soils and soil incubations6, 7, 8, 9, 10.
Both type I and II conventional methanotrophs need ammonium to be active and to grow in the rice rhizosphere. However, the number of type I methane oxidizers was increased to a greater extent relative to the number of type II methane-oxidizing bacteria upon fertilization. Nitrogen availability has been suggested to select for type I methane oxidizers13,but experimental evidence has been missing until now.
Our data suggest that fertilizer application may lead to a reduction, rather than to an increase, of methane emission from wetland rice fields. Methane emission from the microcosms used in this study was reduced up to 57% by fertilizer application17. Moreover, a significant negative correlation has been found between porewater ammonium concentrations and methane emission, indicating increased methane oxidation to be the reason for this reduction17. However, we17 and other investigators21, 22 have shown that fertilization can also enhance methane production due to higher plant biomass and subsequent carbon availability for methanogens. The net emission will be the balance between enhanced consumption and production. We expect that the results we report here will therefore lead to a re-evaluation of fertilizer practice and studies on methane emission from soils and sediments.
Methods
Microcosms
Experimental system and design have been
described17, 23. The microcosms had a physical separation between
rhizosphere and bulk soil, and were incubated in growth cabinets for a period
of 90 d at 70% relative humidity and in a light/dark cycle of 12/12 h
at a photosynthetically active radiation of 450 
E m–2
s-1 (99 W m–2) and a temperature
regime of night/day of 20 °C/25 °C. The surface of the
soil was always covered with 2 cm of demineralized water. Soil and seeds (
Oryza sativa var. Roma, type japonica) originated from a Italian rice
paddy (Vercelli, northern Italy).The microcosms were either unplanted, planted
and unfertilized, or planted and supplemented with urea (200 or 400 kg N ha
-1) or (NH4)2HPO4 (200 or 400 kg
N ha-1). The fertilizer was added twice per week by injection
into the soil with a syringe. After 90 d of incubation, the soil from the
rhizosphere and bulk soil compartments was collected and soil slurries were
prepared24 for performing the analyses described below.
Methane oxidation rates
These (Fig. 1a) were determined by transferring subsamples of the above slurries to flasks that were supplemented with CH4 (10,000 p.p.m.v.). CH 4 depletion was monitored by gas chromatographic (GC) analysis of the headspace gas as described earlier24. From the sigmoidal CH 4 depletion curves (Fig. 1b, open triangles) initial (0–12 h) and induced (12–24 h) oxidation rates were calculated by linear regression. To demonstrate the nitrogen-limitation of methane oxidation in unfertilized microcosms, slurries from unfertilized microcosms were supplemented with (NH4)2SO4 (final concentration 2 mM N).
14C-PLFA profiles
Directly
after collecting the soil and preparation of soil slurries, 20 ml of slurry
was transferred to a polyethersulphon membrane (pore size 0.2 m, diameter
4.7 mm) by vacuum filtration. The membranes (with 
4 g of fresh soil)
were transferred to 150-ml flasks, supplemented with 0.15 MBq 14CH
4 and incubated at 25 °C for 3 d. PLFAs were extracted as
described elsewhere11. Radiolabelled PLFAs were separated into
15 fractions by capillary GC and collected as 14CO2
after combustion in the flame ionization detector. The radioactivity in each
fraction was determined by liquid scintillation counting.
PCR and DGGE of soil DNA
DNA was extracted from soil
and amplified by PCR as described12 . The primer sets MB10
and MB9 amplify 16S rDNA of methylotrophic bacteria belonging to the -
and - proteobacteria12, respectively. PCR products were
separated by DGGE using a 35–70% and 45–70% denaturant gradient
for MB10 and MB PCR products, respectively. For pure cultures
and enrichments, 10 l of PCR product was loaded onto the gel, whereas
for soil samples 45 l was applied. Bands were excised from the gels,
reamplified and sequenced as described12. The phylogenetic position
of the derived sequences are shown in Fig. 4.
Phylogeny
16S rDNA sequences were aligned and placed
phylogenetically with the ARB software package25. Evolutionary
distances between pairs of sequences were calculated by using the Jukes–Cantor
and Felsenstein algorithms in the ARB software. The phylogenetic trees were
constructed using the neighbour-joining algorithm supplied with the ARB package,
based on complete 16S rDNA sequences of -, - and -proteobacteria.
The partial 16S rDNA sequences retrieved from DGGE gels were added to the
tree by keeping its topology constant26. Sequences of partial
16S rRNA gene fragments of excised DGGE bands have been deposited in the GenBank
database under accession nos AF179599–AF179610.
