Flooded rice paddies are one of the major biogenic sources of atmospheric methane. Apart from this contribution to the ‘greenhouse’ effect, rice paddy soil represents a suitable model system to study fundamental aspects of microbial ecology, such as diversity, structure, and dynamics of microbial communities as well as structure–function relationships between microbial groups. Flooded rice paddy soil can be considered as a system with three compartments (oxic surface soil, anoxic bulk soil, and rhizosphere) characterized by different physio-chemical conditions. After flooding, oxygen is rapidly depleted in the bulk soil. Anaerobic microorganisms, such as fermentative bacteria and methanogenic archaea, predominate within the microbial community, and thus methane is the final product of anaerobic degradation of organic matter. In the surface soil and the rhizosphere well-defined microscale chemical gradients can be measured. The oxygen profile seems to govern gradients of other electron acceptors (e.g., nitrate, iron(III), and sulfate) and reduced compounds (e.g., ammonium, iron(II), and sulfide). These gradients provide information about the activity and spatial distribution of functional groups of microorganisms. This review presents the current knowledge about the highly complex microbiology of flooded rice paddies. In Section 2 we describe the predominant microbial groups and their function with particular regard to bacterial populations utilizing polysaccharides and simple sugars, and to the methanogenic archaea. Section 3 describes the spatial and temporal development of microscale chemical gradients measured in experimentally defined model systems, including gradients of oxygen and dissolved and solid-phase iron(III) and iron(II). In Section 4, the results of measurements of microscale gradients of oxygen, pH, nitrate–nitrite, and methane in natural rice fields and natural rice soil cores taken to the laboratory will be presented. Finally, perspectives of future research are discussed (Section 5).
Rice is the world's most important agronomic plant, with 143 million ha under cultivation globally. In about 75% of this land rice grows under flooded conditions , and such anoxic soil conditions lead to the emission of CH4. This compound has a high potential for absorbing infrared light and therefore is one of the major greenhouse gases [2,3]. Up to 70–80% of atmospheric CH4 is biogenic . Estimations of the annual emission rate from flooded rice fields range between 60 Tg  and 110 Tg [6,7]. The latter value for the emission rate accounts for approximately 25% of the total annual CH4 emission into the atmosphere [6,7], indicating that flooded rice fields are an important source of biogenic CH4.
Rice field soil is studied in order to understand the processes leading to the production of methane, but also to gain general knowledge about both structure–function relationships between microbial groups and interactions of microorganisms with rice plants. Flooded rice paddy soil can be considered as a compartmentalized system, with three compartments characterized by different physio-chemical conditions: oxic surface soil, anoxic bulk soil, and rhizosphere soil plus rhizoplane (Fig. 1). Between these compartments, well-defined microscale chemical gradients can be measured and give information about the activity and spatial distribution of functional groups of microorganisms. Functional groups of microorganisms consist of metabolically related organisms, e.g., oxygen-respiring bacteria, nitrate reducers, iron reducers, sulfate reducers, but also fermenting bacteria and CH4-producing archaea. The concerted action of all functional groups of microorganisms drives the carbon, nitrogen, sulfur, and iron cycle in rice paddies, but also in other soils and sediments.
Besides the spatial distribution of functional groups of microorganisms there is also a temporal development during the growth season of rice. After flooding of the rice fields, oxygen is consumed by aerobic bacteria and chemical oxidation reactions, and oxygen is depleted fast in most regions of the soil. In the anoxic zone alternative electron acceptors are used. Nitrate is the first electron acceptor reduced after oxygen depletion followed by Mn(IV), Fe(III), SO42−, and CO2. This sequential reduction  can be explained by the thermodynamic theory which predicts a reduction according to the redox potential of the electron acceptor  (Fig. 2). Along oxic–anoxic interfaces as they may occur directly beneath the floodwater–soil interface and in the rhizosphere soil (plus rhizoplane), alternative electron acceptors will be regenerated because of the reoxidation of reduced compounds (Fig. 1).
The microbial food-web is driven by an input of organic matter into the rice soil. After the rice harvest the remaining plant material is ploughed under the soil. In principle, the pathway of anaerobic polysaccharide degradation is well known and has also been elucidated experimentally for anoxic rice soil . It mostly starts with the activity of fermenting bacteria which excrete enzymes that hydrolyze the polysaccharides. The same group of bacteria converts the resulting sugar monomers to alcohols, fatty acids, and H2. In the presence of alternative electron acceptors, these substrates are degraded completely to CO2. Under limitation of external electron acceptors, syntrophic bacteria further degrade the alcohols and fatty acids to acetate, H2-CO2 (or formate) and CO2. Acetate and H2-CO2 (or formate) finally serve as substrates for methanogenic archaea. An alternative route is the direct conversion of monomers (e.g., sugars) to acetate by homoacetogenic bacteria. Acetate then serves as substrate for acetotrophic methanogens converting it to CH4 and CO2.
In this review we will discuss the following three topics: (1) Characterization of the predominant microbial groups and their function; (2) spatial and temporal development of microscale chemical gradients measured in experimentally defined model systems of flooded rice paddy soil, including gradients of oxygen and iron; and (3) measurement of microscale gradients of oxygen, pH, nitrate–nitrite, and methane in natural rice fields and natural rice soil cores taken to the laboratory. Most of our current knowledge on the first two topics has been gained by using flooded rice microcosms incubated in the greenhouse (Fig. 1), soil slurries, or unplanted soil cores as experimental model systems. The use of such model systems has the advantage that both cultivation/incubation and sampling is possible under defined and reproducible conditions. The paddy soil used for most of the investigations to which we will refer in Sections 2 and 3 was originally taken from rice fields of the Italian Rice Research Institute in Vercelli (Italy). Since the characterization of the structural and functional composition of microbial communities in heterogeneous systems, such as rice paddy soil, strongly relies on the methods applied, the research data presented in Section 2 will be discussed in relation to the methodological approaches used to identify and enumerate predominant microbial populations. Sections 3 and 4 complement one another in the kinds of microscale chemical gradients measured. The experiments in natural rice fields (Section 4) were carried out in the Philippines.
2Rice soil microbial community
The meaningfulness of studies about diversity and structural composition of microbial communities relies on the methodological tools used. Traditionally, characterization of microbial communities depended on cultivation of microorganisms. However, cultivation permits the identification of only a very small portion of complex microbial communities, which is often below 1% of the microscopically detectable cells [11–13]. Over the last decade, important advances in molecular biology led to the development of cultivation-independent, PCR-based approaches. Such molecular approaches are based mainly on the retrieval and comparative sequence analysis of environmental small-subunit (SSU) rRNA genes (rDNA) [14–18]. Although they allow a more comprehensive analysis of microbial diversity, conclusions about the relative abundance of the microorganisms detected by such a PCR-based SSU rDNA approach have to be done with care [15,19–21]. This is because of the possibility of method-inherent bias introduced during the extraction of total community DNA, PCR-based amplification of the environmental SSU rDNA pool, and cloning. In addition, neither a defined phenotype nor an ecological role can be inferred with certainty on the basis of the phylogenetic position of an environmental SSU rDNA sequence.
Consequently, a dual approach consisting of cultivation and direct recovery of SSU rDNA was used to obtain a more objective view of the structural and functional composition of the rice paddy soil microbial community than can be obtained by either cultivation or direct recovery of SSU rDNA sequences alone. The research focused on the characterization of two functionally important groups. The first group encompassed bacteria capable of utilizing polysaccharides and simple sugars. Since anaerobic degradation of organic matter in rice paddy soil mostly starts with the hydrolysis of polysaccharides and fermentation of sugar monomers, bacteria utilizing such compounds were expected to be the most abundant in this habitat. Methanogenic archaea, the second group, represent the functionally most relevant microorganisms with respect to the global importance of flooded rice fields as a major source of atmospheric methane. To enrich for representative phenotypes and genotypes of these two functional groups of microorganisms, cultivation was performed using serial liquid dilution series in the format of anaerobic most-probable-number (MPN) counts . The isolation of microorganisms in pure culture was carried out mainly from the terminal positive dilution tubes. Using the same soil sample as the starting material, two SSU rDNA data sets were generated, the first one based on the MPN isolates and the second one based on the cultivation-independent recovery of SSU rDNA.
2.2Predominant bacterial populations
In general, MPN counts with the polysaccharides xylan and pectin as growth substrates resulted in cell numbers up to two orders of magnitude higher than those with simpler sugars, glucose and cellobiose (4.9×107–2.5×108 versus 5.9×105–5.9×106 cells per g of dry soil) . In contrast, MPN counts on three different types of cellulose, i.e., amorphous, microcrystalline, and filter paper cellulose, resulted in clearly lower cell numbers (3.1×105–5.5×106 cells per g of dry soil) than those determined on xylan and pectin [10,23]. This may be due to cultivation bias, e.g., poor culturability of this group of bacteria, or to truly low numbers .
All strains isolated in pure culture from the terminal positive dilution tubes on xylan and pectin were shown to grow also with glucose and cellobiose [15,23]. Thus, theoretically the portion of the rice paddy soil microbial community culturable with sugar dimers and monomers should have been similar to that determined with polysaccharides. The greater suitability of XB and PB for determining cell numbers in rice paddy soil might be due to the phenomenon of ‘substrate-accelerated-death’, which leads to failure of microorganisms to grow upon transfer from low-nutrient environments to media with relatively high substrate concentrations [24,25]. Polymers, on the other hand, have first to be hydrolyzed, and thus the initial concentration of growth substrates is close to zero.
The phylogenetic placement of 19 (poly)saccharolytic MPN isolates in relation to 57 SSU rDNA clone sequences directly recovered from the rice paddy soil resulted in close matches between the two data sets only for isolates from those culturable populations determined by the MPN counts to be large, i.e., those counted with xylan and pectin as substrate . These included matches with SSU rDNA similarity values greater than 98% within distinct lines of descent of the division Verrucomicrobia  (strains PB90-1 and ACB90) and the Cytophaga-Flavobacterium-Bacteroides (CFB) group (strains XB45 and PB90-2), as well as matches with similarity values greater than 95% within distinct lines of descent of clostridial cluster XIVa  (strain XB90) and the family Bacillaceae (strain SB45) (Fig. 3). The Verrucomicrobia strains PB90-1 and ACB90 are closely related to the (poly)saccharolytic strains VeGlc2, VeCb1, and VeSm13. These strains were also isolated from the anoxic bulk soil of flooded rice microcosms and were described as novel anaerobic strictly fermentative ultramicrobacteria . The small size of about 0.5 μm in length and 0.35 μm in diameter corresponding to an average cell volume of 0.03–0.04 μm3 was a stable characteristic of these strains  and not due to a miniaturization of cells, which may be observed under starvation conditions [30,31].
The close correlations between SSU rDNA data sets created by two independent approaches suggested an actual abundance of those populations defined by the organismal isolates and the affiliated environmental SSU rDNA sequences . This view was further supported by the total number of cells detected by DAPI (4′,6-diamidino-2-phenylindole) staining (4.8×108 cells per g of dry soil) and fluorescent in situ hybridization (FISH) performed with a SSU rRNA-targeted oligonucleotide probe which was considered universal for most members of the domain Bacteria (2.8×108 cells per g of dry soil). Based on the good agreement between the total number of cells detected by DAPI staining and FISH and the population sizes calculated for the polysaccharolytic strains PB90-1, XB45, PB90-2, and XB90 in the MPN counting experiments (approximately 2.5×108 cells per g of dry soil), it could be concluded that these MPN isolates were representative members of dominant bacterial groups which accounted for 5–52% of the total number of cells in the anoxic rice paddy soil [23,26]. A more qualitative support for this conclusion was the observation that bacteria with identical SSU rDNA sequences, i.e., the strains PB90-1 and ACB90, or at least highly similar sequences (>98%), i.e., the strains XB45 and PB90-2, were isolated from the terminal positive dilution tubes of MPN counting experiments carried out with different substrates (ACB=amorphous cellulose; PB=pectin; XB=xylan) (Fig. 3).
In addition to (poly)saccharolytic bacteria, homoacetogenic bacteria affiliated to the genus Sporomusa were detected in MPN counting experiments on lactate and ethanol with cell numbers between 2.3×106 and 7.5×108 cells per g of dry soil; but only if cultivated in coculture with Methanospirillum hungatei. The Sporomusa-like strain DR1/8 was isolated from the terminal positive tube of the MPN counting experiment on lactate, which gave an estimate of 7.5×108 cells per g of dry soil. The Sporomusa-like strains DR5 and DR6 were obtained in MPN counting experiments on 3,4,5-trimethoxybenzoate and ethylene glycol, respectively (Fig. 3). These MPN counts, conducted without the addition of a methanogen, suggested population sizes of only 103–104Sporomusa-like cells per g of dry soil, which may indicate that these compounds might not be useful to assess the real size of homoacetogen populations .
Why the addition of M. hungatei to the counting experiments resulted in the isolation of Sporomusa spp., while the same substrates without the addition of the methanogen led to the isolation of Desulfovibrio spp., is still not known. However, a syntrophic cooperation between the Sporomusa-like strain DR1/8 and M. hungatei obviously changed the pattern of fermentation products towards hydrogen which was recovered as methane. Thus, one explanation given by the authors  was that the addition of M. hungatei maintained the H2 partial pressure at levels similar to those found in rice paddy soil, which may have allowed homoacetogenic Sporomusa-like bacteria to dominate these cultures on lactate and ethanol. The consideration of Sporomusa spp. as predominant members of the bacterial community in the anoxic bulk soil of flooded rice microcosms is supported by the close phylogenetic correspondence between clone BSV16 that was retrieved directly from the anoxic bulk soil of a flooded rice microcosm, and the strains DR5, DR6, and DR1/8 (Fig. 3).
The growth conditions used in direct enrichment cultures often fail to reflect accurately the natural resources and conditions of the environment examined, and thus select for fast-growing bacteria which may outgrow numerically superior populations. Consequently, the microorganisms obtained in pure culture may not be of numerical or ecological significance for the biogeochemical processes in this environment. Thus, in addition to the use of polysaccharides as substrates for growth (see above), the dilution of the inoculum to extinction in the format of MPN counts and the isolation of only those bacteria in pure culture enriched in the terminal positive dilution tubes were decisive factors for the isolation of representative phenotypes of dominant rice paddy soil bacterial populations. Similar conclusions were drawn from results obtained in studies on the biodiversity of hot spring microbial mat communities, i.e., close matches between SSU rDNA data sets generated from bacterial isolates and environmental SSU rDNA were observed only when near extinction inoculum dilutions were used for enrichment of bacteria . This finding confirms the theory of the liquid dilution procedure, which predicts that the terminal tubes with growth should contain an inoculum consisting of the numerically most significant microorganisms present in the sample [22,38]. Another important factor causing close correspondence between cultivation and molecular recovery of SSU rDNA may be the use of liquid enrichment cultures instead of solidified agar media because many bacteria capable of growth in liquid media might not develop on solidified media. This view is supported by the results of two studies which also compared the microbial diversity detectable by cultivation with the diversity detectable by direct recovery of SSU rDNA [35,39]. Although in the course of both studies several hundred aerobic heterotrophic bacteria were isolated via inoculation of agar media with aliquots of a soil suspension dilution series, no close matches were detected for the untreated samples, or matches at the genus level were found only after aerobic treatments of the sediment cores for 1–21 weeks .
Although the close matches observed between some of the MPN isolates and environmental SSU rDNA clones provided strong evidence that numerically significant bacteria were cultivated, the analysis of the SSU rDNA clone library also clearly revealed the failure of culture methods to provide a comprehensive view of the total diversity in anoxic rice paddy soil. This failure of culture methods can be explained by the limited number of different growth media and enrichment conditions that can be tested, and by the fact that many bacteria might be in the viable but non-culturable state [40,41]. Nonetheless, based on the close correlations observed between the MPN isolates (PB90-1, XB45, PB90-2, and XB90) and environmental SSU rDNA, it can be assumed that the diversity represented by the clone library largely reflected the composition of the predominant bacterial groups in the rice paddy soil. Apart from the Verrucomicrobia and the CFB group, these were mainly members of the clostridial subgroups I, III, IX, XI, and XIVa , as well as clostridial-like lineages represented by the genera Oxobacter and Caloramator, and the strain SX-1 . Further evidence that various clostridial-like subgroups are typical inhabitants of the anoxic bulk soil was provided by the finding that these microorganisms apparently did not colonize the rice roots . Both cultivation and direct recovery of SSU rDNA identified members of the α and β subclasses of the class Proteobacteria as major root-associated bacterial groups , while in the anoxic bulk soil these bacterial groups appeared to be present only as a minor component of the bacterial community [23,26]. An assessment of spatial changes in the bacterial community structure along a vertical oxygen gradient in flooded paddy soil cores clearly revealed a correspondence between the abundance of α and β proteobacteria and the measured oxygen concentration . The oxic zone, i.e., the upper 2 mm directly beneath the floodwater–soil boundary layer, was colonized only by members of these two subclasses of Proteobacteria, while clostridia-like populations seemed to predominate the anoxic zone. The community diversity patterns of both the oxic and anoxic zone were clearly different from that of the air-dried paddy soil which had been used for the preparation of flooded soil cores. The latter observation indicated that the bacterial populations detected along the vertical oxygen gradient were actively growing, which supports the view of α and β proteobacteria as typical inhabitants of (micro-)oxic locations and clostridia as those of the anoxic zone in the rice paddy soil. The rhizosphere soil can be considered a transient zone between root system and the anoxic bulk soil, which is characterized by steep oxic–anoxic gradients. This transient character might also be reflected at the bacterial community level because the cultivation-independent recovery of SSU rDNA sequences suggested the abundance of α and β proteobacteria but also of various clostridia-like subgroups in this habitat .
The characterization of clostridia-like subgroups  plus Sporomusa spp.  as predominant inhabitants of the anoxic bulk soil, with many of them capable of forming spores [10,36], seems to correlate well with the major environmental factors which determine the development of defined microbial groups in this habitat . These include alternating periods of flooding and dryness, anoxic conditions during the flooded stage, and the fact that the main portion of the substrates is present in the form of polymers, such as xylan, pectin, and cellulose. Considering the periodical alternation between flooding and drainage, the capability of forming spores or other kinds of resting stages should be expected to be an important adaptation characteristic of indigenous microbial populations in rice field soil. This view is in good agreement with the high survival rate which has been determined for the anaerobic bacteria in air-dried rice field soil even after storage for 2 years; i.e., 34% of the anaerobic bacteria present in the original moist soil used as the starting material survived this treatment . However, spore formation could not be shown for all strains characterized to be representative members of dominant polysaccharolytic populations (i.e., for the strains PB90-1, PB90-2, XB45, and even the clostridia-like strain XB90), suggesting that a major portion of the anaerobic microbial community survives oxic and dry periods despite their inability to form spores. This hypothesis is further supported by the fact that methanogenic archaea are also able to survive during the dry and oxic periods at population sizes very close to those of the flooded and anoxic periods [47–49]. This observation raises the question how those anaerobic microorganisms, which are incapable of forming spores, are able to survive the dry and oxic periods of rice paddy soil. A preliminary answer might be given by the finding that the Verrucomicrobia-like strains VeGlc2, VeCb1, and VeSm13, and the methanogenic archaea isolated from rice paddy soil, are oxygen tolerant [29,50]. A significant portion (0.5–10%) of the isolates Methanosarcina strain MVF4 and Methanobacterium strain HVF5 also survived the additive effect of desiccation and exposure to oxygen . The survival rate was even higher in the presence of pyrite (FeS2), which is an important constituent of rice paddy soil . Consequently, it was hypothesized that the highly uneven, ‘raspberry’-like surface of the spheroidal pyrite crystals  may be a protective habitat for methanogens .
The functionally most important archaea in rice paddy soil are the methanogens. Radiotracer experiments had shown that acetate and H2-CO2 (or formate) are the predominant substrates for CH4 production [53,54]. The maximum amount of H2 relative to acetate that can be produced is 4 mol H2 plus 2 mol acetate per mol glucose (i.e., H2:acetate ratio of 2:1). Four H2, but only one acetate, are required for the production of one CH4. Thus, theoretically the contribution of acetate and H2-CO2 to the total CH4 produced during the anaerobic degradation of sugars should be 66 and 33%, respectively . This agrees reasonably with experimentally determined values for the contribution of acetate and H2-CO2 to CH4 production in anoxic rice paddy soil. For instance, the fraction of CH4 produced from acetate in slurries of Italian rice field soil was between 51 and 67% after 36 days of incubation at 30°C , while the percentage contribution of H2-CO2 to CH4 production in soil cores taken from flooded Italian rice fields ranged from 17 to 31%. The lower values measured for H2-dependent methanogenesis in contrast to the maximally possible contribution of 33% to total CH4 formed were probably due to the activity of homoacetogenic bacteria, which shifted the H2:acetate ratio towards acetate [57,58] (see also below).
The enumeration of H2-CO2-utilizing methanogens in the anoxic bulk soil of flooded rice microcosms detected culturable population sizes between 2.0×106 and 2.3×107 cells per g of dry soil . Strain VeH52 was obtained from the terminal positive dilution tube of one of these MPN counts. Its phylogenetic position and close correspondence to SSU rDNA directly recovered from the same sample of rice paddy soil indicated a predominance of hydrogenotrophic methanogens affiliated to Methanobacterium bryantii and Methanobacterium formicicum (Fig. 4). Counts of methanogens able to utilize acetate revealed culturable population sizes between 5.1×105 and 1.3×106 cells per g of dry soil. The cell counts on acetate nicely demonstrated the effect of the use of inocula diluted nearly to extinction. The cultures originating from the less-dilute inocula, corresponding to population sizes of <103 cells per g of dry soil, were dominated by pseudosarcinae. The affiliation of the pseudosarcinae to Methanosarcina was confirmed by strain VeA23, which was obtained from one of the lower dilution steps. Strain VeA23 was affiliated to Methanosarcina mazei (Fig. 4). In contrast, microscopic observations of the most dilute positive cultures indicated the presence of consortia consisting of two different morphotypes with nonfluorescent Methanosaeta-like multicellular filaments as the predominant type, while short fluorescent rods seemed to be only a minor component of the consortia. The recovery of SSU rDNA from these cultures confirmed the presence of methanogens affiliated to Methanosaeta concilii (culture A5.1-A, Fig. 4). However, on the intragenus level this methanogenic type formed with SSU rDNA directly retrieved from the rice paddy soil a clade which was clearly distinct from M. concilii. The second type of the consortia could be assigned to M. bryantii (culture A5.1-B, Fig. 4). The presence of Methanobacterium-like cells in the methanogenic cultures on acetate could be expected because the number of culturable H2-utilizing methanogens was higher than that of acetate-utilizing methanogens. In addition, Methanosaeta-like microorganisms might produce low levels of H2, which in this case might have enabled the Methanobacterium spp. to survive or even to grow. Taken together, the dual approach of cultivation and direct retrieval of SSU rDNA identified members of the genera Methanosaeta and Methanobacterium as predominant methanogenic populations in the anoxic bulk soil of flooded rice microcosms, whereas Methanosarcina spp. seemed to represent only a numerically minor population .
There are only a few studies on methanogenic diversity in natural rice fields to which the data obtained for the ‘flooded rice microcosm’ model system can be compared [61–67]. For instance, the cultivation of dominant hydrogenotrophic methanogens from samples of 13 soils representative of major types of rice paddy soil in three countries (France, the Philippines, and USA) resulted almost always in the isolation of Methanobacterium spp. . The cultivation approach on acetate only characterized Methanosarcina-like archaea as the dominant acetotrophic methanogens. Cultivation was carried out via MPN counting experiments. The MPN counts (cells per g of dry soil) for the 13 soils ranged from 102 to 106 on H2-CO2, from 50 to 106 on formate, and from <10 to 104 on acetate. In most of these soils, the culturable populations on formate were 5 to 400 times less abundant than those enumerated on H2-CO2. The cell counts of acetotrophic methanogens were probably subject to method-inherent bias because Methanosarcina spp. usually occur as dense aggregates which will be recorded by the MPN counts instead of single cells .
Nucleotide signature sequences affiliated to Methanosarcina, Methanosaeta, Methanogenium, or Methanobacterium were identified by direct retrieval of SSU rDNA in samples taken from nine Japanese rice field soils . Taking all reports into consideration, the following preliminary conclusions about the ubiquitous presence of defined methanogenic groups in rice paddy soils can be drawn:
1The predominant acetate-utilizing methanogens belong to either the Methanosarcinaceae [61,64–66] or the Methanosaetaceae [59,66]. M. barkeri or M. mazei are the indicator organisms for Methanosarcina-like populations, while Methanosaeta-like populations are affiliated to M. concilii.
2Members of the Methanobacteriaceae seem to be the predominant group of hydrogenotrophic methanogens in most rice paddy soils, with M. bryantii and M. formicicum being the indicator organisms for nonformatotrophic and formatotrophic populations, respectively [59,61,64–67]. In addition, members of the genus Methanobrevibacter have been identified [62,64,67].
3In addition to Methanobacteriaceae, H2-CO2-utilizing members of the Methanomicrobiaceae with affiliation to the genera Methanoculleus (also utilizes alcohols for growth)  and Methanogenium seem to be present; the latter methanogenic type has been detected especially in Japanese rice field soils .
Our current knowledge on the methanogenic diversity patterns in anoxic rice field soils raises the question of which environmental factors regulate the relative abundance of individual methanogenic groups in rice paddy soil. For instance, the acetate concentrations measured in porewater taken from the ‘flooded rice microcosm’ model system and from Italian rice field soil were between 10 and 100 μM. This range of acetate concentrations can be utilized for growth by the specialist Methanosaeta, but is clearly below the threshold values which permit growth of the currently known Methanosarcina species (typically between 200 and 1200 μM) [59,68]. Thus, the characterization of Methanosaetaceae as the dominant acetotrophic methanogens in the anoxic bulk soil of flooded rice microcosms corresponded well to the acetate concentrations measured. This suggests that the concentration of microbially available acetate represents an important regulator of the relative abundance of Methanosarcinaceae versus Methanosaetaceae. This threshold concept also explains acetate concentrations <100 μM measured during steady state CH4 production in various rice paddy soils from China, the Philippines, and Italy [69,70].
However, the explanation of the relative population sizes of Methanosarcinaceae versus Methanosaetaceae in rice paddy soils based only on their different thresholds for the utilization of acetate seems to be too simplified a concept, as deduced from temperature shift experiments with slurries of Italian rice field soil . A shift in the temperature from 30 to 15°C, the typical temperature range in Italian rice field soils, resulted in a decrease of the CH4 production. Such a shift to a lower temperature is accompanied by a change of the degradation pathway of organic matter, resulting in a decrease of the steady state H2 partial pressure and a transient accumulation of acetate, propionate, caproate, lactate, and isopropanol . As a consequence, the fraction of CH4 produced from H2-CO2 decreases and that produced from acetate increases, so that CH4 is then mainly produced from acetate [53,73,74]. Most likely, this is correlated with homoacetogenesis becoming the main fermentation reaction under this condition . Consequently, major changes in the structural composition of the methanogenic community can be expected. Prolonged incubation of the soil slurries at 30°C resulted in a dominance of Methanosarcinaceae, while Methanosaetaceae seemed not to be present. In contrast, incubation experiments at 15°C showed a much more diverse archaeal community with a relative decrease of the Methanosarcinaceae and increase of the Methanosaetaceae . Similar observations were made in temperature shift experiments in which serial transfers of methanogenic cellulose-degrading enrichment cultures resulted in the dominance of Methanosarcinaceae and Methanosaetaceae at 30 and 15°C, respectively . However, in both studies the increase of Methanosaetaceae at 15°C in relation to Methanosarcinaceae could not be explained by the threshold concept, since the acetate concentrations were at 15°C well above 1 mM and thus above the threshold of Methanosarcina species. An experimentally supported explanation for the relative increase of the Methanosaetaceae at 15°C has still not been reported. However, a currently favored hypothesis speculates that the enriched Methanosaeta species might be better adapted to low temperatures than Methanosarcina species with a lower temperature optimum for growth (μ and/or Y) rather than for enzyme activity (Vmax) . Nevertheless, taking into consideration the relative increase of H2-dependent methanogenesis at higher temperature, the dominance of Methanosarcinaceae at 30°C appears reasonable because Methanosarcina species are the only currently known acetotrophic methanogens utilizing both acetate and H2-CO2 (and also methanol). This higher metabolic versatility might also explain why members of the Methanosarcinaceae, but not Methanosaetaceae, were detected on rice roots, since excised rice roots were found to produce CH4 predominantly from H2-CO2 rather than from acetate .
Compared to the tremendous changes of the methanogenic community structure observed in response to temperature shifts from 30 to 15°C, the population dynamics of the archaeal community seemed to be low during the first 17 days after flooding of Italian rice field soil and constant incubation of the soil slurries at 25°C . As concluded from SSU rDNA-based T-RFLP analyses, only the relative SSU rRNA gene frequency of the Methanosarcinaceae increased distinctly within the first 11 days (from 15 to 29% of total archaeal gene frequency), while that of the Methanosaetaceae did not change. The relative increase of the Methanosarcinaceae was positively correlated with changes in acetate and formate concentrations, providing further evidence of the selective advantage of Methanosarcina species being able to switch between different electron donors.
Likewise, the soil aggregate size seems to have only a minor impact on the archaeal community structure, while it has a major effect on the rate of production of CH4, as determined in slurries of Italian rice field soil . The rates of CH4 production were lowest with small aggregates (<50 μm and 50–100 μm), were highest with aggregates of 200–2000 μm size, and were intermediate with aggregates of 2000–15 000 μm size. The different rates of CH4 production were positively correlated with the concentrations of acetate, propionate, and caproate that transiently accumulated in slurries of the different aggregate sizes, and also with the organic carbon content. However, these differences in the CH4 production rates did not correspond to observable changes in the archaeal community structure.
Besides the identification of methanogens with well-known phenotypic traits and ecological roles, several novel archaeal lineages have been detected in the anoxic bulk soil and on rice roots of flooded rice microcosms (rice clusters I–V) , and in slurries of Italian rice paddy soil (rice cluster VI) . To date, no representative pure cultures are available for any of these six lineages. The rice clusters I and II formed two distinct clades within the phylogenetic radiation of the orders ‘Methanosarcinales’ and Methanomicrobiales (Fig. 4) , which suggested that the corresponding microorganisms can be considered methanogenic. However, considering the phylogenetic distances to taxonomically described members of the ‘Methanosarcinales’ and Methanomicrobiales, cultured archaea belonging to rice clusters I and II would have to be given the taxonomic status of a family or even an order . Further evidence to support the hypothesis that these novel groups are methanogens comes from their molecular detection in methanogenic consortia grown in cultures that had been inoculated with the terminal positive dilution steps of MPN counts of methanogens on washed rice roots. SSU rDNA sequences recovered from a H2-CO2-using consortium either belonged to rice cluster I or were affiliated to M. bryantii. In contrast, SSU rDNA sequences retrieved from an ethanol-using consortium either grouped within rice cluster II or belonged to the genus Methanosarcina (Fig. 4).
Based on the phylogenetic branching pattern, a methanogenic phenotype appears to be unlikely for members of the rice clusters III, IV, and VI, while members of rice cluster V may or may not be methanogens. Thus, these four archaeal lineages will only be treated briefly. The rice clusters III and V belong to the Euryarchaeota. Rice cluster III grouped on the same branch as Thermoplasma acidophilum and marine group II but was only distantly related to these taxa. Rice cluster V formed a novel main line of descent without any distinct affiliation to one of the other euryarchaeal groups  (Fig. 4). The rice clusters IV and VI are deep-branching crenarchaeal assemblages with rice cluster IV closely related to environmental SSU rDNA recovered from a temperate marsh environment , whereas rice cluster VI was affiliated to environmental SSU rDNA previously recovered from agricultural soil [71,82].
3Localization of various microbial processes in flooded rice soil
Theoretically, the reduction processes in depth profiles should be ordered according to their redox potential , and under limitation of electron donors the distinction of the various processes is expected to be sharper. However, in sediments rich in organic matter, several processes can coexist, e.g., sulfate reduction and methanogenesis [83,84]. Profile studies provide information about the spatial distribution of various electron acceptors. In addition, active zones can be recognized and sometimes modeled from the profiles.
The development of microsensors for oxygen enabled the analysis of microscale gradients and the calculation of oxygen respiration rates in profiles of thin microbial mats and various sediments [85–88]. In the sediments analyzed, a significant amount of oxygen was consumed for the oxidation of reduced compounds such as sulfide, ammonium, and ferrous iron . In rice paddies not only does the soil surface contain oxygen but so does the rhizosphere, due to diffusive transport of oxygen through the aerenchyma of rice roots. The release of oxygen from the roots of rice or other aquatic plants causes radial redox gradients around the roots [90–96]. The presence of oxygen enables various chemical and microbial oxidation processes in the rhizosphere such as the oxidation of methane [97–101], ammonium [102,103], sulfide , and ferrous iron [105–108]. In rhizosphere soil the turnover rates of electron acceptors are higher than in bulk soil, and in addition also the turnover rates of organic substrates are enhanced by the production and release of root exudates.
In fact, oxygen at the soil surface and along the roots is a key factor controlling the gradients of other electron acceptors, such as nitrate, ferric iron, and sulfate. In the presence of oxygen these electron acceptors are regenerated by an oxygen-dependent oxidation of the reduced products (ammonium, iron(II), and sulfide). The regeneration of nitrate and ammonium leads to a tight coupling of nitrification and denitrification especially in the rhizosphere soil, which was shown by various experimental setups using different methodological approaches. Significant emission of 15N2 was observed after addition of 15NH4Cl to planted rice soil . Nitrifying and denitrifying activities calculated from microelectrode measurements of oxygen and nitrate showed an overlapping zone of both processes in the depth profile of model sediments [86,87]. Nitrogen emission from rice microcosms taken as an indicator for denitrification was detected only after N-fertilization and when nitrification was possible due to the oxygen supply via the aerenchyma .
The pool size of total sulfur in paddy soil is rather low and therefore sulfate reduction was also expected to occur preferentially in the oxic–anoxic interfaces. Sulfate reduction rates determined using the 35S-radiotracer techniques  were highest directly at the rice root (500 nmol cm−3 d−1; within 0–1.5 mm distance) compared with rhizosphere soil (310 nmol cm−3 d−1) and unplanted rice soil at the oxic/anoxic interface (100 nmol cm−3 d−1, ). The porewater sulfate profile showed highest concentrations at the surface of unplanted rice soil (150 μM), which decreased to 10 μM at 8 mm depth . Although there was a distinct profile of sulfate-reducing activity and sulfate concentrations, the distribution of sulfate-reducing bacteria quantified by MPN enumeration showed similar cell numbers along the vertical profile . This result raises the question of whether the MPN technique is suitable for the detection of small changes in population size. The profile of sulfate-reducing activity might not reflect an enrichment of sulfate-reducing bacteria along the sulfate gradient due to different substrate availability. In contrast, it could be the result of changes only in sulfate-reducing activity at locations where sulfate is available. However, slightly higher cell numbers of sulfate-reducing bacteria have been observed in planted and root-influenced paddy soil compared to bulk soil , suggesting that the higher sulfur turnover rate in the rhizosphere allows microbial growth. The characterization of sulfate-reducing bacteria isolated from rhizosphere soil and bulk soil suggested the presence of two different physiological types of sulfate reducers: The rhizosphere soil, ‘rich’ in substrate, was dominated by fast-growing incomplete oxidizers like Desulfovibrio species, whereas the substrate-impoverished bulk soil mainly contained complete oxidizing, spore-forming and slow-growing Desulfotomaculum species .
3.2Redox processes of iron
Similar to the nitrate and sulfate profiles, iron(III) and (II) profiles are affected by oxygen. Iron(II) is very oxygen sensitive at neutral pH and is oxidized in a fast chemical and microbial reaction to iron(III) in zones where oxygen is present. In the anoxic zone iron(II) is mobile as Fe2+ cation by diffusive transport. The greatest fraction of iron(II) in rice soil, however, is present in precipitated forms (mainly as siderite). Even lower than iron(II) is the diffusive mobility of iron(III) which occurs almost exclusively as iron(III) oxide and hydroxide precipitate. The soluble (porewater) iron(III) and iron(II) concentrations of flooded rice soil were in fact 4500 times smaller than the concentrations determined after acidic extraction (Fig. 5), which results in dissolution of most of the biologically relevant iron precipitates. Furthermore, the porewater concentrations of iron(III) and iron(II) showed parallel depth profiles which gave no information on the localization of iron-reducing or iron-oxidizing activity. In contrast, the iron profiles determined after acidic extractions of iron(II) and iron(III) from thin soil slices (100 μm) showed a spatial and temporal development within several weeks (Fig. 6) after flooding. In these iron(II) and iron(III) profiles, zones of active iron(II) oxidation and iron(III) reduction could be recognized . In the upper 3 mm oxygen is present (Fig. 7), and the predominant process of iron metabolism is oxygen-dependent iron(II) oxidation. The contribution of microbial and chemical iron(II) oxidation cannot be differentiated at present due to methodological limitations, e.g., there is no specific inhibitor known for microbial iron oxidation and it is still not known whether microbial iron oxidation shows an isotopic discrimination favoring the lighter isotope 54Fe over the heavier isotopes 56Fe, 57Fe, and 58Fe similar to the isotopic discrimination of the lighter carbon and sulfur isotopes. In the uppermost mm of the soil, both iron(II) and iron(III) concentrations are very low (20 μmol cm−3), indicating a diffusive loss of iron into the floodwater.
Below 3 mm soil depth, where oxygen is depleted, a second zone of iron oxidation is indicated by high iron(III) concentrations. In this zone iron oxidation can occur only by oxygen-independent processes. The electron acceptor used for anaerobic microbial iron(II) oxidation could be nitrate. This is supported by the availability of nitrate in this soil depth and high cell numbers of nitrate-dependent iron oxidizers . This process is also supported by the observation of iron(II) oxidation in anoxic soil slurries after nitrate addition . In the depth zone where the data suggest a nitrate-dependent iron oxidation, the reduction of iron(III) may occur simultaneously. Below 4–5 mm soil depth, nitrate becomes limiting and thus microbial iron reduction starts to be the predominant process. However, below 8 mm iron-reducing activity decreases because of limited availability of reducible iron(III) 6 weeks after flooding. Measurements of methane production showed increasing potential activities of methanogens in soil depths of decreasing iron(III) concentrations [112,115].
3.3Effect of rice roots and soil parameters on iron profiles
Rice plants growing in flooded rice paddy soil influence the iron profiles because of the diffusive release of oxygen from the roots into the rhizosphere soil. Oxygen enables the oxidation of ferrous iron at locations with a high root density leading to high concentrations of ferric iron. The soil surface is characterized by a dense root mat  and iron(III) concentrations in this soil layer are almost twice as high as in unplanted paddy soil . The iron coating on rice roots is thought to reduce the toxicity of ferrous iron for the plant [116,117]. In addition, the iron coating of rice roots can act as a scavenger for metal cations like Zn2+, Cd2+, Cu2+, and Pb2+ and thus may affect the nutrient uptake of rice plants. There are reports that iron plaques can reduce the nutrient supply possibly leading to nutrient deficiency (e.g., ). Other reports suggest that iron plaques can act as a nutrient reservoir . The presence of iron-oxidizing bacteria in the ferric iron precipitates on roots of other aquatic plants  suggests that microbial iron oxidation may also contribute substantially to iron precipitation in rhizosphere soil.
The profiles of iron(II) and iron(III) in the soil are affected by the iron content of the soil and the quality of the iron oxide. Rice soils are known to contain ferrihydrite, lepidocrocite, goethite, and hematite [121–123]. Microbial reduction of these iron oxides seems to be influenced by their crystallinity [124,125] and surface area . In paddy soil a preferential reduction of ferrihydrite and lepidocrocite is indicated by the results obtained using differential extraction procedures (Ratering, unpublished results). Ferrihydrite and lepidocrocite have both a low crystallinity and metastability, compared to goethite and hematite . Both oxides (ferrihydrite and lepidocrocite) can be extracted with hydrochloric acid in the presence of hydroxylamine, or with oxalic acid. The fraction of iron(III) extractable with these solvents was found to be in good correlation with the amounts of microbially reducible oxides [128,129].
The organic content of the soil also affects the iron profiles. In environments of low organic content microbial respiration is lower, resulting in oxygen penetration deeper into the soil . Competition for substrates between various physiological groups is more pronounced. Under these conditions methane production is suppressed and the contribution of iron(III) reduction to total organic carbon oxidation is enhanced . In soil with low organic carbon content, the CO2 consumption in the methanogenic zone is lower, which leads to a higher accumulation of siderite. This was confirmed by the detection of higher concentrations of iron(II) in the soil fraction below 4 mm depth .
Based on the iron profiles, different zones of iron redox reactions have been localized. However, the dynamics of iron reduction and iron oxidation cannot be deduced from these profiles. Unfortunately, experiments measuring the iron-oxidizing and iron-reducing activity in situ cannot be performed at present because methods for determination of either iron reduction or iron oxidation have not been developed. In addition to the shortcomings of measuring the simultaneous action of both oxidation and reduction, it is not yet possible to differentiate between microbial and chemical iron oxidation due to methodological limitations.
4Microscale chemical gradients measured in natural rice paddies
The best resolution of chemical gradients in stratified microbial communities is offered by the use of various types of microsensors that can be inserted directly into the otherwise undisturbed community. Several relevant chemical species can be measured in rice paddy soil this way (pH, NH4+, NO3−, NO2−, N2O, CO2, O2, CH4, H2). Unfortunately, no ion-specific microsensors are available for Fe2+ and Mn2+, but it is possible to analyze these species by simple microelectrodes combined with a voltametric technique . At present, the microdistribution of pH, O2, NO3−+NO2−, and CH4, has been investigated in intact rice paddy soil by specific sensors, and H2 and CH4 have been determined with gas sampling probes [133,134]. The following discussion will focus on the analyses with specific sensors as these offer the highest spatial resolution. Only in situ data and data measured in freshly collected rice paddy cores will be described. The data illustrated here were obtained at the International Rice Research Institute (IRRI) in the Philippines and on paddy soil that was kept flooded during the whole growth period. All experiments were conducted with soils that had been flooded for more than 3 weeks following transplanting. The soil at IRRI is rich in clay and contains about 1.8% (dry weight) extractable Fe. The Fe content is sufficient to prevent any build-up of free hydrogen sulfide during the growth season.
Oxygen profiles have been measured both in the field and in the laboratory using oxygen microsensors . Oxygen never penetrated more than 3–6 mm from the floodwater–soil boundary layer, although cyanobacterial photosynthesis could push the oxic–anoxic interface down to a few mm depth during the day (Fig. 8). During darkness the oxygen penetration was about 1 mm. However, when the oxygen microsensor was introduced close to a plant, about one third of the profiles exhibited secondary oxygen peaks (Fig. 8) with a width of 1–2 mm. Four profiles were measured down to a depth of 90 mm, and secondary oxygen peaks were observed at depths ranging from 35 to 75 mm in two of these profiles (one profile contained two secondary peaks). The data in Fig. 8 were measured in the field, thus it was not possible to observe whether the secondary peaks were due to oxygen within roots penetrated by the sensor, or if they were due to oxygen leaking from roots to the surrounding soil. Experiments were therefore conducted with rice plants that had been cultivated in 0.4 l glass beakers for about 1 week before the experiments. Oxygen around and within roots growing along the glass wall of these beakers could be analyzed in the laboratory, while the relative position of the microsensor and the roots could be observed through the glass wall. The measurements (Fig. 9) showed that the aerenchyma of the 0.7 mm thick roots contained 32% of atmospheric tension of oxygen, and that the oxic zone extended about 0.3 mm from the root into the otherwise anoxic soil. The aerenchyma was surrounded by 0.05–0.1 mm of root tissue, where part of the decrease in oxygen concentration occurred. The measurements also showed that the brown zone around the roots reflected not only oxidized but also oxic conditions. The extension of the oxidized zone was exactly the same as that of the oxic zone. The absence of a brown zone around some, mainly older, roots thus indicated that there was no oxygen export from root to soil. Also shown in Fig. 9 is the distribution of pH around the root. It should be mentioned that the pH sensitive cone of the electrode used was about 0.25 mm long, so that the data-points represent some kind of mean pH over a distance of 0.25 mm. Inside the root the pH was around 7.15, whereas the soil pH was 6.25. Oxidation of reduced Fe, Mn, and sulfide in the oxic rhizosphere apparently did not cause a drop in pH as suggested by other authors . Measurements around roots of deep-water rice and lowland rice growing in acidic soil did not show any local pH minimum either (N.P. Revsbech and W. Reichardt, unpublished).
Oxygen sensors for field use must be made with a very heavy glass wall , so that they do not break when pushed against medium-sized soil particles. Such sensors can be made with a robust and almost solid glass tips having diameters of more than 1 mm but otherwise the same characteristics as thin oxygen microsensors (i.e., almost the same signal in stirred and stagnant media, a rapid response). The disadvantage of robust sensors, however, is the physical disturbance caused through introduction into the substrate. When analyzing finer details such as illustrated in Fig. 9, it is necessary to use microsensors with tip diameters very much less than 0.1 mm. Structures to be analyzed must be relatively close to the soil surface to minimize the risk of breakage during introduction. Usually such detailed analysis is of little value if not followed by visual inspection, as described above.
Most lowland rice farming is based on extensive use of nitrogen fertilization. Usually this nitrogen is supplied as urea , since a supply as ammonium may lead to high losses by evaporation, and a supply as nitrate leads to a high loss by denitrification. The ammonium made available by hydrolysis of urea can, however, undergo nitrification in the oxic parts of the rice paddy, i.e., in the water, at the floodwater–soil interface, or in the oxic rhizosphere. Much of the nitrate–nitrite formed by nitrification will subsequently be denitrified, as much of the NOx− will diffuse into anoxic soil instead of being taken up by the roots. Some reports indicate high rates of coupled nitrification–denitrification in the rhizosphere , but even higher rates of nitrification should be expected at the soil surface of recently fertilized soil. Such nitrifying activity can be investigated by 15N-isotope work, and the formation of nitrite and nitrate can be investigated by Liquid-Ion-eXchanger (LIX)-based microsensors [138,139] or by NOx− biosensor [140,141]. Nitrate and nitrite may, however, be difficult to analyze by LIX sensors, as the soil surface is characterized by steep gradients of bicarbonate which may interfere with detection of low nitrate and nitrite concentrations. The NOx− biosensor shows only little interference with bicarbonate and measurements may be conducted with a resolution of <0.1 μM NOx− when operated with electrophoretic transport of ions to the tip . New types of LIX sensors are, however, constantly being developed , and LIX sensors made with tips <1 μm offer the possibility of very high spatial resolution. Current microscale biosensors have tip diameters of 20–80 μm, but such tips are sufficiently small for analysis of the floodwater–soil interface.
An example of biosensor-measured NOx− profiles in a rice paddy soil is shown in Fig. 10. A soil column was taken into the laboratory and incubated at 32°C in water containing 100 μM NH4+ and about 1 μM NOx−. The data illustrated are typical of many NOx− profiles. In the dark, when the oxygen penetration (not shown) into the soil was only 1 mm, NOx− was also depleted within 1 mm. There was no NOx− peak indicating nitrifying activity. During illumination (300 μmol photons m−2 s−1) the oxygen penetration was 2–3 mm (not shown) due to microbenthic photosynthesis, and NOx− was building up to a concentration of 3.6 μM at a depth of 1 mm. Depletion of NOx− occurred at a depth of 2.5 mm, again corresponding to the depth of oxygen depletion. Apparently the increase in oxygen penetration caused by photosynthetic oxygen production increased nitrification to levels where a NOx− peak could be detected, but the nitrifying activity was still at a very low level. Many NOx− profiles were measured, and most were almost identical to those shown in Fig. 10. One single patch did, however, exhibit nitrifying activity also in the dark, even without added ammonium. The microprofile measurements were backed by 15N isotope work, and the isotope experiments verified the finding of overall low nitrification and denitrification rates.
The very low nitrification rates illustrated in Fig. 10 must be due to a low number of nitrifying bacteria, as NH4+ was present in excess. All fields that we investigated had similar low rates, but conditions probably exist where much larger nitrifying populations can develop. Conditions promoting nitrifying activity would include regular and excessive nitrogen fertilization of the fields, a stable soil surface enabling the slow growing nitrifiers to establish themselves, and a relatively low load of organic material enabling a relatively deep oxygen penetration into the soil.
It would be possible to analyze NOx− in the rhizosphere using a similar experimental design to that illustrated in Fig. 9 for oxygen. However, even lower concentrations than found at the floodwater–soil interface can be expected here, as the sinks, i.e., roots and anoxic soil, are within 0.2 mm of the potential production zone in the oxic rhizosphere.
Much work has been done on methane emission from rice paddies, and it is thus relevant to study the microenvironments where methane oxidation limits emission to the atmosphere. An oxygen-insensitive methane biosensor has been developed [143,144], which can be used for this purpose. This methane biosensor does suffer from some H2S interference (25% of the signal for an equivalent amount of methane) and often also from H2 interference. Free sulfide was, however, not present in the investigated paddy soil due to the high iron content, and H2 is usually only present in nanomolar concentrations and does not interfere.
Examples of oxygen and methane profiles in a rice paddy soil exposed to darkness (right panel) and to illumination (left panel) are shown in Fig. 11. During darkness the oxidation of methane occurs at the very surface of the soil. During illumination the oxidation zone is pushed down to a depth of 4–6 mm due to the extension of the oxic zone by photosynthetic oxygen production.
Although methane oxidation associated with the floodwater–soil interface may seem very efficient, a significant amount of methane passes to the atmosphere across this interface due to its escape in the form of bubbles. The shunting of methane through the plant aerenchyma is, however, the main mechanism of emission [54,100,145].
Our knowledge about the microbial processes in rice paddy soil has steadily increased over the last few years. However, we are now only starting to discover the diversity and relative abundance of indigenous microbial populations and their function. Many important aspects of community structure and microbial processes remain unresolved and will need to be addressed in future research. Some of these are as follows:
1Rice fields are exposed to alternating periods of flooding and drainage with desiccation of the soil with every vegetation period. Except for a few preliminary studies [47–50], no detailed research has been conducted to examine the survival of distinct groups of (strictly) anaerobic microorganisms in air-dried paddy soil, including their adaptive mechanisms and physiological responses to oxidative and water activity stress. Understanding the effects of alternating conditions in paddy soil on the microbial community would increase our knowledge about adaptation mechanisms of microorganisms developed in response to environmental stress.
2We have now a reasonable understanding of the sequential reduction processes occurring upon flooding of rice field soil, as well as of the spatial-temporal development of microscale chemical gradients at the floodwater–soil interface and in the rhizosphere soil. However, although the activity and dynamics of the nitrate reducers, iron(III) reducers, and sulfate reducers, can be measured, only a little is known about the identities of the major microbial players and whether the microbial community structure changes during the subsequent reduction processes. Is it the kind and/or population size of the functional groups that change with time or is it the same population that just switches its metabolic activity? This knowledge is important for a better understanding of structure–function relations between microbial groups in rice paddy soil. Fundamental principles arising from this research could also be of relevance for other soil and sediment systems.
3Despite the detailed characterization of the predominant (poly)saccharolytic bacteria and methanogenic archaea, our current knowledge about the functional groups of microorganisms involved in the methanogenic degradation of organic matter in anoxic rice paddy soil is still limited. This is especially true with respect to the diversity, abundance, and dynamics of homoacetogenic populations and with respect to syntrophic associations formed during the methanogenic degradation of organic matter and the microbial community members involved.
4A major challenge will be the isolation of members of the archaeal rice clusters I–VI into pure culture to obtain detailed information about their phenotypic traits and putative ecological roles. This seems to be of special importance for the (probably) methanogenic members of rice clusters I and II, which have repeatedly been detected in slurries of Italian rice field soil [71,77,78] and also in total community DNA extracted from roots of rice plants grown in the Philippines . Isolation of bacterial and archaeal microorganisms as representatives of SSU rDNA sequences that have been retrieved directly from environments is an important task to enlarge our understanding of microbial diversity. Currently, the number of environmental SSU rDNA sequences is increasing to a much greater extent than the number of isolates with new phylogenetic and/or physiological characteristics.
5Most of our knowledge about the microbiology of rice paddy soils has been gained from greenhouse and laboratory experiments. The data obtained from profile measurements in flooded rice soil microcosms and in natural rice fields seem to indicate that the major principles of process localization can be recognized in microcosm studies. The oxygen profiles were very similar in both systems. Since the oxygen profile seems to be the regulating factor for several other compounds (nitrate, sulfate, iron) one could speculate that the profiles of alternative electron acceptors measured in flooded rice cores are in good agreement with those detectable under natural field conditions. However, the spatial resolution of the profiles and time scales of profile development are greatly affected by soil parameters, e.g., soil organic carbon, content of alternative electron acceptors, and presence of rice roots, and therefore may vary in different rice soils. Thus, to what extent our knowledge obtained from profile measurements in flooded rice soil microcosms is applicable to the situation in natural rice fields must now be assessed in additional, carefully designed field experiments. These should include not only the localization of functional processes in rice paddy soil by microsensor studies, but also the confirmation of our current knowledge on the bacterial and archaeal community structure as well as the respective population dynamics during the vegetation period of rice.
6The increasing demand for rice as a major nutriment for the growing human population may lead to increased methane emission from rice fields and, as a consequence, may result in an even higher impact of rice cultivation on future global warming than today. Thus, future research on rice cultivation techniques will have to focus on a better understanding of methane production and in situ methane oxidation as well as on the development of possible mitigation strategies for suppression of methane emission. Our knowledge about the soil microbial community associated with the currently used rice cultivars will be essential for the success of new mitigation strategies and new rice breeds.
This list of research topics might be selective in that it reflects the interests of the authors of this review. Nonetheless, experimental data on these topics would provide important pieces of information which should lead to an increased understanding of this complex and puzzling microbial ecosystem.
The rice field soil used in most of the studies discussed in this review was kindly provided by Dr. Salvatore Russo (Vercelli, Italy). We are indebted to Ralf Conrad for stimulating discussions. We thank Stefan Ratering, Ralf Oltmanns, and Dagmar Reinhardt for technical help with manuscript preparation and the figures. Financial support through the Deutsche Forschungsgemeinschaft and DANIDA (RUF) is gratefully acknowledged.