Chapter three of Thomas Brock's classic microbial ecology text (Brock, 1966) is prefaced by a quote attributed as a graduate student motto. The motto simply states, ‘microbial ecology is microbial physiology under the worst possible conditions’. After more than a quarter century, the statement still captures the essence of a fundamental obstacle to progress in microbial ecology – attributing a microbially catalysed process, measured in a complex environmental milieu, with a specific organism or organisms. This remains a challenge to contemporary microbial ecologists. In specific cases, such as microbial mat communities, which are highly structured and dominated by a relatively small number of morphologically recognizable taxa, it has been possible to measure rates of photosynthesis or sulphide oxidation and relate these specifically to the organisms present (Revsbech and Jørgensen, 1986). These studies are the exception, and there is a need to develop approaches that allow explicit links to be made between the presence of specific organisms and the processes they catalyse. This has become more pressing given the tantalizing and growing insight into the complexity and dynamics of natural bacterial and archaeal communities offered by culture-independent analyses (Amann et al., 1995; Head et al., 1998; Hugenholtz et al., 1998). There are now many major groups of Bacteria and Archaea known only from molecular sequences and, until these organisms can be cultivated, the only means of understanding their role in the environment is through culture-independent characterization linked to determination of in situ metabolic activity. Even when an organism can be cultivated, properties determined in the laboratory may not necessarily reflect the activities and physiology of their counterparts in the environment (Brock, 1987), where factors such as resource competition, environmental heterogeneity, predation and other interactions are prevalent.
This review aims to highlight selected recent developments that are revealing the metabolic capabilities and ecological role of the uncultured majority
Inferring an organism's role from its response to the environment
Characterization of an organism's habitat and establishing its distribution within the habitat has long been a tool available to microbial ecologists to infer ecological function (Brock, 1966). This approach has been useful because the presence of natural gradients of light, temperature, salinity, electron acceptors and donors, etc. is a feature of many natural environments. Determining bacterial community structure in relation to such gradients may allow the organisms present to be associated with specific ecological properties and may identify potential factors responsible for the emergence of diversity in specific groups of organisms. This is exemplified by recent studies of prochlorophytes.
Several related (≈ 2–3% 16S rRNA sequence divergence) photosynthetic picoplankton from the genus Prochlorococcus have been isolated from the same location in the North Atlantic. Physiological studies of the isolated prochlorophytes revealed that they comprised evolutionarily distinct clades each adapted to different optimal illumination conditions. These findings led to the proposal that the congruent physiological and evolutionary diversity observed was driven by adaptation to different light intensities within the water column of oligotrophic marine environments. Although this conclusion was derived from the study of pure cultures of Prochlorococcus (Moore et al., 1998), the application of culture-independent methods to determine the composition of prochlorophyte communities and the relative abundance of different genotypes in relation to gradients of light has placed the laboratory studies in an environmental context. West and Scanlan (1999) determined the relative abundance of high (HLI) and low light-adapted (LL) Prochlorococcus genotypes in samples taken from different depths in the water column. Relative quantification of the different genotypes was determined by probing of polymerase chain reaction (PCR)-amplified 16S rRNA gene fragments with oligonucleotides characteristic of the HLI and LL genotypes. Denaturing gradient gel electrophoresis (DGGE) was used to determine qualitative changes in the composition of the prochlorophyte community with depth. This indicated a clear depth- and light intensity-related, differential distribution of the HLI (distributed between 0 and 50 m) and the LL (distributed between 30 and 90 m) genotypes in water column samples from the North Atlantic (Fig. 1). The authors concluded that niche partitioning as a result of adaptation to environments experiencing different light intensity played a fundamental role in driving the diversity and distribution of the different Prochlorococcus species. However, it was also clear from the field studies that environmental factors such as nutrient status, turbulence or temperature played a role in defining the finer scale fluctuations in genotype depth distribution that was apparent in different water column profiles.
Freshwater sulphur bacteria
The co-existence of closely related sulphur bacteria from the genus Achromatium has been explained in part by culture-independent analysis of the distribution of morphologically similar, distinct genotypes in relation to redox gradients in a freshwater sediment (Gray et al., 1999a). 16S rRNA sequence analysis and fluorescence in situ hybridization (FISH) clearly demonstrated that Achromatium communities comprised several co-existing, closely related Achromatium species (Head et al., 1996; Glöckner et al., 1999; Gray et al., 1999a). The distribution of different Achromatium genotypes in relation to sediment redox profiles indicated that each was adapted to different redox conditions (Fig. 2). Although all genotypes were identified at all depths in the sediment, the relative proportions of two of the genotypes (RY 7/8 and RY 5) changed with the transition between the oxidized and the reduced zones in the sediment. The RY 7/8 cells comprised 67.8% (n = 1137) of cells in the oxidized zone and represented a significantly (P < 0.001) smaller proportion of the population (50.1%, n = 373) in the reduced zone. Conversely, the RY 5 cell type, which represented 14.8% (n = 1089) of the Achromatium population in the oxidized zone, represented a significantly (P < 0.001) larger proportion of the population (34.6%, n = 399) under more reducing conditions. This suggested that divergent co-existing Achromatium genotypes may occupy different ecological niches within the sediments they inhabit. However, in contrast to studies of populations of Prochlorococcus spp., the mechanisms that lead to the diversification observed in Achromatium communities remain to be determined.
Sulphate-reducing bacteria in microbial mats
Niche partitioning engendered by differences in physicochemical conditions can clearly occur over very different spatial scales – metres in the case of water column communities of Prochlorococcus and millimetres in sediment environment Achromatium spp.
Perhaps one of the most exciting developments in recent years has been the application of microsensor measurements in culture-independent studies of bacterial community structure to provide analysis of chemical and physical gradients on a submillimetre scale combined with information on the distribution of specific microorganisms. Studies of a hypersaline cyanobacterial mat community from Solar Lake, Sinai (Egypt), have revealed unexpected properties of sulphate-reducing bacteria (SRB) that inhabit the mats (Minz et al., 1999a,b). First, the SRB populations were greatest within the oxycline of the bacterial mat (Minz et al., 1999a), supporting a growing body of evidence that SRB do not require anoxic conditions to function (Canfield and Des Marais, 1991; Krekeler et al., 1997). Secondly, the distribution of SRB, based on analysis of genes encoding dissimilatory sulphite reductase (DSR), was related to the level of oxygen (Minz et al., 1999b). In general, SRB diversity was higher in samples from the oxycline compared with anoxic samples from greater depth in the mat. In particular, specific clades of SRB showed clear preferences for particular redox conditions, with bacteria most similar to Desulfonema and Desulfococcus identified largely in regions where oxygen was present and fluctuated between 160% air saturation during the day to below 10% at night. In contrast, DSR sequences recovered from permanently anoxic depth intervals were not closely related to currently characterized taxa of SRB. This contrasts with laboratory studies of SRB indicating that several Desulfovibrio spp. have a high affinity for oxygen although they are not effective oxygen respirers (Krekeler et al., 1997; 1998); no Desulfovibrio spp. DSR sequences were recovered from the Solar Lake mat samples (Minz et al., 1999b). The significance of oxygen-respiring or aerotolerant Desulfonema spp. has thus been inferred from studies combining culture-independent characterization of specific bacterial populations with respect to their occurrence under defined environmental conditions, suggesting fruitful new avenues for research.
Nitrifying bacteria in a nitrifying reactor
Microsensor measurement of chemical profiles in relation to the distribution of different bacterial taxa may be used not only to identify environmental conditions favoured by particular bacteria, but can also be used to quantify specific bacterially mediated processes. The net flux of chemical species in a given environment can be calculated from measured chemical profiles and diffusive properties (Jørgensen and Revsbech, 1985). When combined with FISH to provide quantitative estimates of the abundance of specific cells, specific transformation rates and kinetic parameters can be calculated at the level of populations or cells. An elegant application of this approach has been the determination of the effect of changing environmental conditions on the activity and distribution of ammonia-oxidizing and nitrite-oxidizing bacteria (AOB and NOB) in a nitrifying reactor (Schramm et al., 1999). Ammonia and nitrite oxidation rates were measured in aggregates from a fluidized bed reactor. Nitrification was localized to a 100–150 µM zone within the flocs, and FISH was used to locate and enumerate the dominant nitrifying bacteria (Nitrosospira and Nitrospira spp.) within the zone of nitrification. Profiles of ammonium, nitrite, nitrate and oxygen within the aggregate were used to calculate volumetric ammonia and nitrite oxidation rates, and enumeration of the AOB and NOB allowed cell-specific rates of ammonia and nitrite oxidation to be calculated for different points within the reactor. Under conditions in which ammonia was limiting, there was a clear change in the population structure of nitrifying bacteria in the reactor that correlated with declining ammonia concentration. Numbers of AOB and ammonia oxidation rates were significantly higher at the bottom of the reactor and, although AOB were still present in low numbers at the top of the reactor, ammonia oxidation activity could not be detected. In contrast NOB and nitrite, oxidation rates increased from the bottom to the top of the reactor. Calculation of cell-specific oxidation of electron donors indicated that the activities of the AOB were considerably higher than those of the NOB, but the NOB had higher cell numbers. The cell-specific reaction rate was calculated from the ammonia or nitrite concentrations at different points in depth profiles. These values were plotted on Lineweaver–Burk plots to determine half-saturation constants (Km) for ammonia and nitrite oxidation by the uncultured AOB and NOB populations, assuming that they were physiologically homogeneous. The Km values determined in this fashion were orders of magnitude lower than those reported for cultured Nitrosospira spp. and Nitrobacter spp., and it was speculated that the Nitrosospira and Nitrospira spp. present in this reactor were typical K strategists with high substrate affinity and low maximum activity or growth rate, compared with typical r strategists, which have low affinity for substrates but high activity and growth rates.
The value of this kind of study goes beyond understanding factors that control the activity of interacting bacterial communities and has practical consequences for the design of wastewater treatment systems. Half-saturation constants used in mathematical models that inform the design of full-scale wastewater treatment plants for example are typically based on values for pure cultures of bacteria. If these are orders of magnitude greater than for nitrifying bacteria that typically occur in wastewater treatment systems, the predicted performance may deviate significantly from that which actually occurs, with real environmental and economic consequences.
Measurement of metabolic potential in situ
The preceding discussion has focused on attempts to infer bacterial ecology by determining an organism's relationship to its chemical environment. This has worked well in relatively simple, structured environments where, in its broadest sense, the function of the key organisms under investigation is known or where they are predominant members of the bacterial communities present. In the case of the nitrifying bacteria, which fall into phylogenetically and phenotypically coherent groups, it has even been possible, to determine in situ substrate utilization kinetics. However, as studies of Achromatium spp. have illustrated, there is a limit to what can be achieved by simply measuring chemical profiles and fluxes in relation to the abundance of specific bacterial populations. When the resources used by an uncultured organism are unknown, its habitat is complex and heterogeneous, and it co-exists with a multitude of other organisms whose roles are equally unknown, it is not a trivial task to infer an organism's ecological role or metabolic capabilities.
In recent years, a number of techniques have been perfected, which allow microbial ecologists to correlate directly a specific metabolic activity with phylogenetically identifiable units (single cells or DNA sequences) in natural environments. These techniques not only allow microbial ecologists to identify specific metabolic activities of defined phylogenetic entities within a complex community but, by determining the response of specific metabolic properties to natural or experimentally imposed perturbations, it is possible to determine environmental factors that regulate the activity.
Combined microautoradiography and FISH
Microautoradiography is a powerful tool, with which the uptake of specific radiochemicals by individual cells can be determined. This method has been applied in many ecological studies (e.g. Brock and Brock, 1968; Maier and Gallardo, 1984; Karner and Fuhrman, 1997; Gray et al., 1999b); however, its major limitation is the inability to link the uptake of a substrate by a cell with its phylogenetic identity. In contrast, FISH allows the phylogenetic identity of uncultured cells to be determined, but because phylogeny and phenotype are rarely congruent, little can be concluded regarding a cell's metabolic capabilities (Amann et al., 1996; Amann and Ludwig, 2000). Recently, microautradiography and FISH have been combined to exploit the complementary strengths of the two approaches. A number of acronyms have been devised for the slightly different approaches used, e.g. MAR-FISH (microautoradiography-fluorescence insituhybridization; Lee et al., 1999), STAR-FISH (substrate tracking autoradiography-fluorescence insituhybridization; Ouverney and Fuhrman, 1999) and MICRO-FISH (microautoradiography-fluorescence insituhybridization; Cottrell and Kirchman, 2000). However, all the methods differ only in detail, and we shall refer to them collectively by the more generic term combined microautoradiography and FISH (Gray et al., 2000). To date, combined microautoradiography and FISH have been applied in a small number of studies of natural and engineered environments. The method was first used to determine organic and inorganic substrate use by ammonia-oxidizing bacteria, inorganic phosphate-accumulating bacteria and members of the Proteobacteria present in activated sludge (Lee et al., 1999). This demonstrated that environmental perturbations such as induced anoxia had dramatic effects on the uptake of labelled substrates and that members of the beta-Proteobacteria were potentially important in enhanced phosphorus removal. Its application to marine bacterioplankton has further demonstrated that alpha-Proteobacteria and members of the Cytophaga–Flavobacterium group, which together accounted for ≈ 55% of the bacterial cells present, assimilated tritiated amino acids (Ouverney and Fuhrman, 1999). One recent study has demonstrated that the newly discovered low-temperature marine Archaea, now known to be abundant in marine environments, are also capable of assimilating dissolved amino acids at nanomolar concentrations (Ouverney and Fuhrman, 2000). However, in this study, a mix of four Archaea-specific probes – one general archaeal probe, two probes specific for group I Crenarchaeota and a fourth probe specific for the marine group II Euryarchaeota– was used to increase the sensitivity of FISH detection, and it was not possible to attribute uptake of the amino acids to a specific group of marine Archaea. However, from the autoradiographic and FISH data, and measurement of a population maximum at 200 m depth, it was concluded that at least some of the Archaea were heterotrophic. Combined microautoradiography and FISH have also been used to determine the relative contribution made to the utilization of marine dissolved organic matter (DOM) by different prokaryotic groups (Cottrell and Kirchman, 2000). In a study of North American coastal and estuarine waters (Cottrell and Kirchman, 2000), it was determined that a number of the predominant bacterial groups present (members of the α, β and γ subdivisions of the Proteobacteria and members of the Cytophaga–Flavobacterium group) showed very distinct patterns of DOM uptake. When aliquots of sea water were incubated with tritiated chitin, N-acetyl glucosamine (NAG), protein and amino acids, it was shown that no particular group dominated the consumption of all DOM fractions. For instance, a high proportion of the Cytophaga–Flavobacterium group were consumers of chitin, NAG and protein even when they constituted a small fraction of the total population, but only a small proportion of this group was involved in the assimilation of dissolved amino acids. In contrast, α-Proteobacteria were the predominant bacteria consuming amino acids. The complex patterns of DOM fraction utilization, which did not correlate with the relative abundance of the bacterial groups, prompted the authors to suggest that, in models of organic matter cycling in marine environments, it would be important to take account of the composition of bacterial communities and, for instance, consumption of chitin might be dominated by members of the Cytophaga–Flavobacterium group even if they represented only a small proportion of the bacterioplankton.
The association of rather broad phylogenetic groupings with uptake of particular radiolabelled substrates has provided some insight into the role of different bacteria in natural and engineered ecosystems. However, combined microautoradiography and FISH can be equally applied at a much finer phylogenetic resolution and has been used to investigate the physiology of Thiothrix spp. involved in filamentous sludge bulking (Nielsen et al., 2000). Excessive growth of filamentous bacteria in activated sludge is responsible for impairing sludge settling, resulting in reduced performance – a phenomenon known as bulking. Many different filamentous bacteria can cause bulking of activated sludge. These can be difficult to distinguish morphologically and can have different metabolic properties. Accordingly, detailed information on the metabolic properties of different filamentous bacteria involved in bulking is considered to be of value for the formulation of strategies to control their growth. Filamentous bacteria from activated sludge are notoriously difficult to maintain in laboratory cultures and, consequently, our understanding of their metabolic properties is limited. For this reason, combined microautoradiography and FISH have been used to determine whether members of the genus Thiothrix identified in situ in bulking activated sludge are physiologically homogeneous and if the provision of inorganic electron donors promotes mixotrophy. Culture-based studies have shown that some filamentous bacteria from activated sludge (e.g. Thiothrix spp.) include representatives that are facultative chemolithoautotrophs, mixotrophic or heterotrophic (Williams and Unz, 1989). The occurrence of mixotrophic Thiothrix spp. has been considered a significant factor giving Thiothrix spp. a competitive advantage over heterotrophs and leading to bulking. The extent of uptake of [14C]-acetate and [14C]-bicarbonate by filaments that hybridized to a fluorescently labelled oligonucleotide probe specific for Thiothrix nivea and some other members of the genus Thiothrix was determined directly in activated sludge samples subjected to a range of environmental conditions (Nielsen et al., 2000). Uptake of acetate by Thiothrix filaments was noted, especially in the presence of added thiosulphate, as was the assimilation of labelled bicarbonate. However, bicarbonate uptake only occurred in the presence of thiosulphate, thiosulphate and acetate or in cells that had deposited intracellular elemental sulphur. Although these results clearly suggested that the Thiothrix cells were either heterotrophic or mixotrophic, the authors also concluded that Thiothrix filaments were capable of chemolithoautotrophic growth. This conclusion was based on an estimated incorporation of 2.8 g of carbon from bicarbonate mol−1 thiosulphate consumed, a yield comparable with cultured Thiothrix spp. growing chemoautotrophically in a chemostat (Nielsen et al., 2000). In this particular study, all the Thiothrix filaments identified behaved in a similar fashion, suggesting that they were physiologically homogeneous. Extending these studies to other filamentous bacteria promises to aid in the elucidation of environmental factors that lead to activated sludge bulking and, ultimately, the development of diagnostic tools to identify rapidly specific types of bulking and appropriate remedial strategies targeted at the specific physiological properties of the culprit organisms.
So far, the only application of combined microautoradiography and FISH to delineate the activities of different species within the same genus has been a study of mixed natural populations of sulphur bacteria from the genus Achromatium (Gray et al., 2000). Solely microautoradiographic studies of a mixed population of these bacteria from Rydal Water, Cumbria, UK, indicated that not all cells assimilated [14C]-bicarbonate and [14C]-acetate. This suggested that the Achromatium community exhibited physiological as well as phylogenetic diversity (Gray et al., 1999b). It was tentatively proposed that, within the Rydal Water community, there were Achromatium species that were chemoorganoheterotrophs or chemolithoheterotrophs together with Achromatium species that are obligate chemolithoautotrophs, facultative chemolithoautotrophs or true mixotrophs. This putative physiological diversity was invoked to explain the consistent presence of a large proportion of Achromatium cells within the community that were inactive towards one or other of the radiolabelled substrates tested. This hypothesis was consistent with findings from a similar microautoradiographic study of Achromatium spp. at a different location (Hell Kettles, County Durham, UK), which indicated that the predominant Achromatium spp. was heterotrophic (Gray et al., 1999b). By comparison of the substrate uptake patterns of Achromatium at the two sites, it was shown that, in contrast to the Rydal Water Achromatium community, the Hell Kettles assemblage did not display any autotrophic potential. However, the application of combined microautoradiography and FISH to define the physiology of individual components of the Rydal Water community indicated considerable overlap in the known carbon metabolism of co-existing Achromatium populations (Fig. 3). All Achromatium spp. at this site were capable of assimilating 14C-labelled bicarbonate and acetate, indicating that all were either facultative chemolithoautotrophs or mixotrophs. It was therefore suggested that the co-existing Achromatium spp. that were previously shown to be ecologically distinct (Gray et al., 1999a) (Fig. 2) used similar carbon sources, but did so optimally under different conditions. Based on the depth/redox-related distribution of the Achromatium subpopulations within the sediment, it was speculated that the nature of niche differentiation in these organisms might be linked to resource availability.
An unresolved aspect of this study was that uptake of radiolabelled substrates was only detected in a subset of each of the three subpopulations targeted by the oligonucleotide probes used. It was tentatively suggested that this was evidence of facultative metabolism in Achromatium subpopulations where, because of the heterogeneous nature of the sediment, there may be cells expressing different metabolic pathways; however, this conclusion is equivocal and demonstrates some of the pitfalls that must be considered when interpreting data from combined microautoradiography and FISH. For instance, in the study of Achromatium spp., the inconsistency in substrate uptake could be explained simply by functional diversity that was manifest at a higher degree of phylogenetic resolution than the 16S rRNA probes were designed to identify. In a more general sense, interpretation of combined microautoradiography and FISH data should be interpreted with caution. For example, complex metabolism cannot simply be deduced from the uptake of a single radiolabelled substrate. Although uptake of radiolabelled bicarbonate by Achromatium cells suggests that they have autotrophic potential, it is also known that heterotrophic bacteria can incorporate CO2 via alternative enzymatic pathways and, in such cases, inorganic carbon contributes little to cell growth (Gray et al., 2000). Likewise, the uptake of dissolved amino acids by marine Archaea does not preclude their use of other carbon sources, as nothing is yet known about these organisms' use of other organic and, perhaps more importantly, inorganic carbon sources (Ouverney and Fuhrman, 2000). These Archaea may yet prove to be metabolically more versatile and involved in the cycling of organic and inorganic carbon (Ouverney and Fuhrman, 2000). In addition, it should be borne in mind that, with any microautoradiographic technique, substrates may be degraded rapidly by other bacteria, so the uptake of radiolabel by an organism may not necessarily have resulted from the assimilation of the added substrate, but could also have resulted from uptake of metabolites produced from it. This can be avoided to some extent by the use of short incubation times, but this approach is by no means infallible.
In some contexts, a limitation of FISH is that it targets rRNA. The genes that code for this macromolecule are highly conserved and cannot be used to discriminate to the same phylogenetic resolution as functional genes or other more variable coding regions of the bacterial genome such as intergenic spacer regions. Potentially, future studies using microautoradiography to discriminate physiology at the species or subspecies levels could be coupled with in situ PCR methods (Hodson et al., 1995; Chen et al., 1999). This permits the in situ visualization of cells based on more variable regions of the genome and can link uptake of a specific substrate with the presence of a gene or genes that are known to be involved in metabolism of the substrate. Another logical extension of combined microautoradiography and FISH is to apply quantitative methods such as track or grain density autoradiography (Davenport and Maguire, 1984; Carney and Fahnenstiel, 1987) to evaluate relative or actual substrate uptake rates within mixed communities. Absolute measurements require the incorporation of appropriate radioactive standards. In natural communities, these methods have been confined to estimating the primary productivity of individual morphologically conspicuous phytoplankton species (Davenport and Maguire, 1984). However, with the advent of in situ cell identification by FISH, these quantification methods can now be applied to a much wider range of functional activities and physiological groups. It may even prove possible to derive kinetic parameters such as half-saturation constants and Vmax values for individual uncultured organisms through artificial manipulation of substrate concentrations in microcosms.
Stable isotope probing of nucleic acids
In the last few years, the application of stable isotope tracers to determine functionally active components of microbial communities by analysis of lipid biomarkers has been used increasingly in the study of microbial ecology (e.g. Boschker et al., 1998; Bull et al., 2000). This methodological approach is based on the stable carbon isotope labelling of individual organism-specific or group-specific lipid biomarkers (e.g. phospholipid fatty acids, PLFA) by incubation of samples with 13C-enriched substrates and analysis of extracted biomarkers by isotope ratio mass spectroscopy (Boschker et al., 1998). The wider availability of isotope ratio monitoring–gas chromatography–mass spectrometry (irm-GC-MS) for analysis of the natural isotopic signatures of lipid biomarkers has also allowed the environmental role of uncultured organisms to be inferred (Hinrichs et al., 1999). Even though diagnostic signature lipids have been identified for a wide range of taxa, this approach lacks the universal applicability that analysis of nucleic acids offers.
Recently, a method has been developed that exploits stable isotope labelling of nucleic acids to determine the metabolic capabilities of active components of complex natural communities (Radajewski et al., 2000). Stable isotope probing (SIP) of nucleic acids has been used to identify organisms potentially involved in C-1 metabolism in soils. Soil microcosms were treated with [13C]- or [12C]-methanol and incubated under aerobic conditions. Genomic DNA was extracted from soil, and 13C-enriched DNA was purified by equilibrium centrifugation in CsCl–ethidium bromide density gradients (Fig. 4). PCR was used to amplify 16S rRNA genes from the 13C-enriched DNA, and these were cloned and sequenced. A small number of eubacterial sequences from the α subclass of the Proteobacteria and group 1 of the Acidobacterium division were identified. Interestingly, these sequences were most closely related to genera not normally associated with methanol utilization but typically growing at acidic pH. It was speculated that the acidity (pH 3.5) of the soil used may have selected for atypical methylotrophs. A significant strength of SIP is that the 13C-enriched DNA will contain the entire genome of each functionally active component of the community. Cloning large fragments of the labelled DNA using bacterial artificial chromosome (BAC) vectors (e.g. Béjàet al., 2000a) opens the door for more comprehensive genome-level analysis of uncultivated bacteria that can be associated with a specific metabolic function.
The authors of the first paper reporting the use of SIP highlighted a number of limitations of the technique. First, SIP analysis of nucleic acids is less sensitive than analogous lipid biomarker analysis. Furthermore, as with all tracer-based approaches, the dilution of added 13C-labelled substrate by indigenous unlabelled substrates necessarily reduces the proportion of label that is incorporated into DNA. Thus, it is necessary to both add a large excess of labelled substrate and use long incubation times to maximize 13C uptake. Both these experimental strategies are likely to bias the results obtained. For instance, the addition of high concentrations of substrate may promote the growth of copiotrophic organisms that are atypical of those active under natural conditions. This is akin to the biases imposed by traditional culture-based methods. Extended incubation times are also likely to promote the formation of 13C-labelled metabolites that can be assimilated by secondary users of the labelled carbon. Furthermore, where complete degradation of a labelled compound involves a consortium of organisms, it may not always be possible to dissect the role of each without the use of labelled intermediates in the degradation pathway. Even then, degradation initiated by co-metabolic attack may lead to particular difficulties. However, it may be possible to follow the flux of 13C through different members of a consortium by sampling over a time course after the addition of the labelled substrate. Comparison of DGGE profiles of 16S rRNA gene fragments from the unlabelled and labelled fraction over time can, in principle, permit the identification of organisms that sequentially incorporate carbon from the primary substrate, whereas those that primarily incorporate carbon dioxide or methane as terminal products of degradation can be identified by the addition of 13C-labelled bicarbonate or methane. This principle has been elegantly demonstrated with a 4-chlorosalicylate-degrading mixed culture by immunocapture and analysis of the lipids of individual components of the consortium using irm-GC-MS, and an elaborate network of metabolite spillage and carbon sharing was unravelled (Pelz et al., 1999). SIP promises to extend this approach to more complex natural communities.
Although organisms responsible for co-metabolic initiation of catabolism may not incorporate much labelled carbon under natural conditions, in enrichment cultures, they are unlikely to survive unless they use metabolic end-products or intermediates, and this may provide clues to the identity of such organisms with subsequent, complementary analyses being used to strengthen any inferences made.
Despite its limitations SIP offers a powerful tool for linking the occurrence of particular organisms with specific processes in natural communities. At present, the technique is in its infancy, and a number of obvious improvements are likely to have a significant affect on sensitivity. For instance, it may be possible to target RNA, a molecule that is manufactured rapidly and in large quantities in active cells. This may reduce the need for long incubation periods. Although RNA pellets at the bottom of the gradient in CsCl gradients, the use of density separation media with a greater density than saturated CsCl solution (e.g. caesium trifluoroacetate; CsTFA) allows RNA to be recovered as a band in gradients (Zarlenga and Gamble, 1987). In addition, the use of 14C-labelled substrates or, in some cases, multiple labelled compounds (13C, 15N, 2H, 3H) may lower the quantities of labelled substrate required to permit the successful recovery of labelled nucleic acids. However, the use of heavy radioisotopes has associated safety considerations, and the convenience of stable isotope labels is likely to restrain the wide application of radioisotopes.
The potential applications of SIP are immense, and it offers the opportunity to elucidate the functional properties of the plethora of uncultivated organisms that are now known to dominate most natural environments. Obvious targets include uncultured low-temperature Crenarchaeota and members of the SAR11 cluster. In addition, there are a number of microbially catalysed processes that can be measured in situ but for which the organisms responsible have remained tantalizingly elusive. The organisms or consortia responsible for anaerobic methane oxidation and those involved in anaerobic benzene oxidation, for example, are prime candidates for characterization by the application of SIP.
Bromodeoxyuridine incorporation to determine metabolic activity and function
DNA containing bromodeoxyuridine (BrdU) can be purified from DNA that does not contain this thymidine analogue using immunocapture techniques (Urbach et al., 1999). This property has been used to infer the metabolic activity and function of organisms that have increased their metabolic activity in response to specified stimuli, by virtue of their incorporation of BrdU into their DNA (Borneman, 1999; Yin et al., 2000). The bromodeoxyuridine-enriched DNA from the metabolically active members of a microbial community can then be visualized either by immunofluorescence using antibromodeoxyuridine monoclonal antibodies and fluorescently labelled secondary antibodies or isolated by immunochemical capture using antibody-coated paramagnetic beads (Urbach et al., 1999). The second of these approaches has recently been used to determine the extent of functional redundancy along a soil reclamation gradient in a highly contaminated and denuded Brazilian mine spoil (Yin et al., 2000). In this study, soils sampled from four different locations along a reclamation gradient were amended with bromodeoxyuridine and a single carbon substrate (l-serine, l-threonine, sodium citrate or α-lactose hydrate). Soil samples were incubated for 24 h, after which DNA was extracted, and BrdU-labelled DNA was isolated from unlabelled DNA by immunocapture. Community structure analysis of immunocaptured and uncaptured DNA was achieved by PCR amplification of 16–23S intergenic spacers and their subsequent resolution using electrophoresis on agarose gels. This revealed that, although the four different carbon amendments had little effect on the composition of bacterial communities, as represented by uncaptured DNA, significant differences were apparent in the populations that had incorporated BrdU in their DNA. These results suggested that the experimental approach used was able to identify populations that have different metabolic functions in terms of carbon source use. The discrimination of metabolic function achieved was used to determine the range of taxa capable of responding to each substrate in the Brazilian soils. The greater the number of taxa incorporating BrdU in their DNA in response to the addition of a specific substrate was taken as an indication of higher levels of functional redundancy within the microbial community. Interestingly, it was found that the range of taxa incorporating BrdU in response to the addition of each substrate was closely correlated with the ability of the different sites to support plant regrowth, with the highest diversity of active organisms identified in soils with the highest levels of plant growth. These findings indicated that functional redundancy may be a useful indicator of the recovery of these reclaimed mine spoils and of soil quality generally; however, whether high functional redundancy is a necessary part of this improved soil fertility or simply a side-effect of it remains to be determined.
Like bromodeoxyuridine, iodonitrotetrazolium violet (INT) can also used to identify active members of microbial communities that have responded to experimentally imposed stimuli (Whiteley et al., 2000). However, rather than labelling DNA, this method exploits the reduction of tetrazolium salts to form insoluble formazan crystals intracellularly in response to respiratory activity. The intracellular deposition of formazan alters the density of the active cells, which allows their separation by density gradient centrifugation for subsequent community analysis. This method has been applied recently to characterize the response of bacteria to the addition of oxidizable substrates to sea-water samples (Whiteley et al., 2000).
As with SIP, both BrdU and INT approaches have limitations. Some of these are shared with SIP, e.g. selective stimulation of bacteria not actually active before substrate amendment. However, in addition, both these techniques have more specific restrictions. For instance, the application of the INT method is likely to be limited to natural ecosystems such as marine and freshwater environments where planktonic bacterial cells are easily purified. Furthermore, as INT reduction is linked to specific sites of reduction in cellular electron transport chains, some bacteria may either not reduce INT at all or may do so in insufficient quantities to be of use in cell purification. Consequently, data generated by the application of this method to unknown bacterial populations that encompass a wide metabolic diversity should be interpreted with caution (Smith and McFeters, 1997). Likewise, an important limitation of the BrdU approach is that not all microbial populations assimilate the nucleotide even when they are metabolically active (Urbach et al., 1999). At present, the extent to which this is the case is not well understood, and data from studies of natural ecosystems must therefore be interpreted with caution as the absence of sequences from immunocaptured DNA is not evidence that the organisms represented by them are inactive. Nevertheless, the BrdU approach to isolating DNA from a subset of the metabolically active organisms present and the INT method for isolating active cells offer some potential advantages over SIP. First, these techniques do not rely on the use of expensive labelled substrates and so, at least for BrdU, may prove to be more widely applicable (Urbach et al., 1999). Secondly, although SIP is restricted to the investigation of communities that incorporate isotopically heavy elements from labelled substrates into biomass, the BrdU and INT techniques can potentially be used to investigate the response of organisms to a wider range of environmental perturbations such as temperature variation or the detrimental effects of toxic pollutants.
Discovery of novel metabolic activities
Culture-independent analyses of microbial communities on the whole permit the identification of novel organisms by analysis of 16S rRNA sequences. In contrast, studies of operational rather than informational genes are largely restricted to the discovery of variations on a known theme (e.g. ammonia and methane monooxygenase genes or catabolic genes involved in particular transformations). It is thus difficult to discover completely novel metabolic processes without resort to cultivation of the responsible organisms. This limitation has been overcome recently by an elegant approach based on the recovery and identification of novel operational genes linked to rRNA genes recovered in bacterial artificial chromosome (BAC) libraries generated from environmental DNA (Béjàet al., 2000a,b). Although clone libraries containing large inserts from environmental DNA have been used to study uncultured bacterial communities for several years, this has mainly focused on comparative analysis of genes co-recovered with identifiable rRNA genes (e.g. Stein et al., 1996). However, more recently, a DNA polymerase gene has been recovered from the uncultured archaeon Cenarchaeum symbiosum and expressed in Escherichia coli (Schleper et al., 1997). This allowed the biochemical properties of the protein from the archaeon to be determined without access to pure cultures of the organism. Another exciting extension of this approach has been the expression and functional analysis of a novel bacterial rhodopsin gene recovered from a BAC clone that originated from the SAR86 cluster of the gamma-Proteobacteria. Uncultured bacteria from the SAR86 cluster have long been recognized in marine 16S rRNA gene clone libraries (Mullins et al., 1995), and they appear to be widespread in the marine environment. However, nothing was known regarding their physiological properties or ecological role. A SAR86-derived BAC clone was recently found to carry a homologue of the archaeal light-driven proton pump, bacteriorhodopsin (Béjàet al., 2000b). This was termed proteorhodopsin. The proteorhodopsin gene was PCR amplified from the BAC clone, and the PCR product was cloned in an outer membrane protease-deficient E. coli strain. Heterologous expression of the protein and functional analysis in the presence of retinal revealed that proteorhodopsin was a light-driven proton pump and suggested that SAR86 bacteria may exhibit a previously unknown mode of photoheterotrophy or photoautotrophy. This finding introduces the possibility that SAR86-like organisms have an important and previously unrecognized role in the marine carbon cycle.
Microbial ecologists now have at their disposal a raft of complementary culture-independent methods that can be used to investigate microbial function in situ.
Although describing the functionally active and inactive components of complex communities in a single snap-shot analysis may be informative in its own right and can lead to the generation of ecological hypotheses, in order to test the hypotheses rigorously, we need information on how organisms respond to different environmental perturbations (Begon et al., 1996). This may be achieved by applying a temporal or spatial component to analyses or eliciting different physiological responses by applying artificially imposed perturbations. These strategies have been a feature of many of the studies cited in this paper and, as such, provide useful models for future investigations using in situ measures of microbial function.
New and more sophisticated technologies are continually emerging and, as we enter the era of post-genomics, microarray technology promises to revolutionize the ways in which we analyse gene expression in both cultured bacteria and natural environments. However, much in the way of fundamental expression analysis and comparative genomics is required before we are in a position to interpret gene expression data from environmental samples in a comprehensive manner. Analysis of gene expression in complex microbial communities at the level of proteins has already been tested using traditional biochemical approaches (Duncan et al., 2000; Bott et al., 2001), but the application of the tools of proteomics (e.g. two-dimensional gel electrophoresis coupled with matrix-assisted laser desorption ionization–time of flight MS and other mass spectrometric techniques) is set to increase the quantity and quality of information on gene expression that can be obtained from complex microbial communities.
It is easy to be seduced by the elegance of the technology now available to microbial ecologists, but ingenious technology alone does not increase understanding. Only by applying these immensely powerful tools to address real ecological questions will we harness their full potential. The relatively small but growing number of genome sequences from environmentally relevant organisms dictates that the initial application of post-genomic technologies will of necessity involve addressing some rather simple questions in the first instance