A major goal of microbial ecology is the identification and characterization of those microorganisms which govern transformations in natural ecosystems. This review summarizes our present knowledge of microbial interactions in the natural sulfur cycle. Central to the discussion is the recent progress made in understanding the co-occurrence in natural ecosystems of sulfur bacteria with contrasting nutritional requirements and of the spatially very close associations of bacteria, the so-called phototrophic consortia (e.g. ‘Chlorochromatium aggregatum’ or ‘Pelochromatium roseum’). In a similar way, microbial interactions may also be significant during microbial transformations other than the sulfur cycle in natural ecosystems, and could also explain the low culturability of bacteria from natural samples.
Evidence has accumulated that major biogeochemical processes in soil, sediments, and pelagic environments are often mediated by microorganisms which exhibit unknown physiological capacities. These microorganisms could either be novel, not-yet-cultured types or, alternatively, could behave in an unknown manner with a different physiology to that shown in pure laboratory cultures. Examples comprise oxidation of atmospheric methane by organisms with high affinity enzyme systems [1,2], anaerobic ammonium oxidation , and anaerobic methane oxidation . In addition, biogeochemical cycles may proceed in a different manner to that assumed previously. This has been demonstrated by the discovery of anoxygenic phototrophic bacteria which are capable of using reduced inorganic iron as electron donor during photosynthesis [5,6]. Other examples include the oxidation of sulfide by the cyanobacterium Oscillatoria limnetica[7,8], the disproportionation of thiosulfate by sulfate-reducing bacteria , and the oxidation of sulfide with nitrate as electron acceptor mediated by the giant filamentous bacterium Thioploca, which stores nitrate in its large central vacuoles .
For such detailed physiological studies, the isolation of ecologically relevant bacteria is usually a prerequisite. However, ‘nonculturable’ bacteria represent one of the most pressing problems in current microbial ecology. Since up to 50%– in some instances even 90%– of the bacterial cells in natural samples appear to be metabolically active [11–14], most of the bacteria occurring in natural samples should in principle be culturable. By comparison, the fraction of culturable cells frequently is much lower and usually less than 1%. A problem inherent in most cultivation attempts is that, due to the specific growth conditions used, only very few metabolic types of bacteria are able to grow. Accordingly, the summation of the Most Probable Number of purple sulfur bacteria, colorless sulfur bacteria and sulfate-reducing bacteria in a microbial mat accounted for as much as 5–40% of the acridine orange direct count .
However, the estimated number of species in just 100 g of forest soil is significantly higher (13 000 [17,18], or even 500 000 ) than the total number of validly described bacterial species (4272 as of July 2000; ). Clone libraries from environmental samples routinely contain many novel 16S rRNA gene sequences [21–24]. The 16S rRNA gene libraries of natural bacterial communities often do not match sequences of strains isolated from the same or similar samples [25–27]. In fact, a large cultivation campaign during which 659 bacterial isolates were obtained from grassland soil did not yield any of the 16S rRNA sequences dominating in the natural community . Only in a few instances could numerically important bacterial species be isolated [29,30]. These observations indicate that many, if not most, bacteria represent new species and still await their isolation and characterization. Obviously, the failure to cultivate ecologically relevant bacteria has to be attributed to the fact that conventional cultivation methods are still inadequate.
One considerable problem in current culture techniques is that microbial interactions cannot be reproduced adequately. In pelagic environments, total bacterial cell numbers usually are in the range of 106 ml−1. In sediments, cell densities of 109 cm−3 are reached. Provided that the bacteria are homogeneously distributed, an average distance between the cells of 122 and 12 μm can be calculated from these cell numbers in the pelagial and sediments, respectively. Based on these distances, molecular diffusion of small molecules (diffusion coefficients in the range of 1.5×10−5 cm2 s−1) in pelagic waters should take an average of 5 s while diffusive transport in sediments would take 0.05 s. Many bacteria in natural habitats are not uniformly distributed but form microcolonies, aggregates, or biofilms on solid phases [32–34]. Even free-living aquatic bacteria may occur in 10–100-μm scale patches . Here, the exchange of molecules must proceed at a higher velocity than indicated above. The smaller the scale the larger the relative contribution of organisms to the geochemical transformations: Jørgensen  concluded that the biological contribution to sulfide oxidation in the Black Sea was negligible, amounted to 30–50% in Solar Lake, and was 100% in a Beggiatoa dominated microbial mat. Therefore, essentially all transformations within microbial consortia are expected to be mediated biologically. Interestingly, the high affinity oxidation of methane observed in soil could be reproduced in a coculture of a non-novel methanotroph with a Variovorax strain . This recent finding indicates, that the transformations of bacteria in their natural environment can only be completely appreciated if the various interactions between different types of bacteria are considered as well.
Ecologically significant compounds exchanged between bacterial cells comprise (1) signalling compounds as in the case of quorum sensing, (2) growth factors, and (3) compounds directly involved in energy metabolism: in the latter category the exchange of electron donors/acceptors is especially well understood for bacteria participating in the sulfur cycle. It is this latter type of interaction which the present communication will focus on. In particular, we will discuss how such close metabolic interactions involving bacteria of the sulfur cycle will (1) change apparent biogeochemical transformations and (2) are of significance for the successful cultivation of these bacteria.
2Associations of purple sulfur bacteria and colorless sulfur bacteria
In microbial mats both purple sulfur bacteria and colorless sulfur bacteria are found in high population densities, the interesting fact is that both functional groups can be found at the same depth layers [16,37]. Purple sulfur bacteria primarily are anaerobic organisms performing an anoxygenic photosynthesis in which sulfide, or another reduced form of sulfur, is used as the electron donor. Colorless sulfur bacteria, on the contrary, use reduced forms of sulfide not only for the synthesis of cell material but also as a source of energy by their oxidation with oxygen; the in situ use of nitrate as electron acceptor is not well documented. Although oxygen is not lethal to Thiocapsa roseopersicina, the ubiquitously dominant phototrophic bacterium in microbial mats, the synthesis of photopigments in this organism is hampered by an oxygen concentration well below saturation values , thus depriving the organism from its main source of energy. Secondly, in the competition for sulfide, the phototrophic Thiocapsa would be outcompeted by the chemotrophic Thiobacillus. The fact that pigmented purple sulfur bacteria and colorless sulfur bacteria thrive in the same habitat must be related to the effect of this bacterial association on the in situ concentrations of oxygen and the transformations of sulfur compounds. This type of association may also be of relevance to the existence of consortia.
Microbial mats typically show steep gradients of oxygen and sulfide [40,41], and their simultaneous presence is restricted to a narrow layer where both concentrations are low, and either nutrient could be limiting the metabolic activities for colorless sulfur bacteria. When sulfide is limiting the end product will be sulfate. The impact of a low supply of oxygen on the metabolic activities of Thiobacillus has been mimicked experimentally by using the technique of continuous cultivation. The outcome of these studies was that not only did oxygen become undetectable, but also that the concentration of free sulfide in the culture vessel was virtually zero. Moreover, a variety of incompletely oxidized inorganic sulfur intermediates had accumulated . The ability of Thiobacillus to form incompletely oxidized sulfur intermediates when oxygen is in short supply and also to oxidize all sulfide appeared to have severe ecological implications. First of all, the resulting extremely low concentration of oxygen is unlikely to inhibit photopigment synthesis in Thiocapsa, and, more importantly, the low sulfide affinity of the phototroph compared to that of the chemotroph  becomes irrelevant since Thiocapsa can use the reduced sulfur intermediates for growth. Consequently, the aerobic Thiobacillus and the anaerobic Thiocapsa could be grown in coculture provided that oxygen was made the limiting factor . These observations can explain the co-occurrence of these nutritionally different organisms in microbial mats. Nevertheless, it is clear that these organisms do not form a consortium. In microbial mats, the immobile Thiocapsa grows in small colonies on sand grains of approximately 0.2 mm in diameter , whereas Thiobacillus is free-living and in fact shows a diurnal vertical migration coinciding with the shifting gradients of oxygen and sulfide (P. Schouten and B. Meijerink, unpublished data). However, taking advantage of the chemotaxis of the flagellated chemotroph both organisms could profit from the sulfide available provided they were united in a consortium.
3Associations of colorless sulfide-oxidizing bacteria and sulfate-reducing bacteria
The sheaths of large marine Thioploca filaments are densely covered by filamentous sulfate reducing bacteria of the genus Desulfonema. Thioploca, migrating in its own sheath, takes up nitrate in the surface layers and sulfide in the bottom layers. In doing so the organism is able to oxidize sulfide with nitrate as electron acceptor. At first sight the presence of Desulfonema appears to be superfluous. However, little sulfide can be taken up in the deeper layers where concentrations are low and thus the association with the sulfide-producing organism is of advantage to Thioploca. This close association of a sulfide-oxidizing and sulfate-reducing species indicates that a rapid recycling of sulfur compounds occurs. The organisms have not (yet) been cultivated separately. In addition to a possible supply of organic carbon by Thioploca, it may well be that the advantage for Desulfonema is found in the availability of reduced sulfur intermediates, thus avoiding the necessity to use sulfate in the same way as described below for Desulfovibrio in association with Thiobacillus.
In microbial mats not only purple sulfur bacteria and colorless sulfur bacteria co-exist, also sulfate-reducing bacteria have been reported to be most abundant in the surface layers . Sediments are difficult to examine at depth intervals less than 1 mm, thus no conclusions can be drawn with respect to a depth distribution on a 0.1 mm scale. Nevertheless, it was of interest to study the possible co-existence of obligately aerobic sulfide-oxidizing microbes and obligately anaerobic sulfide-producing organisms. Chemostat experiments with mixed cultures of Thiobacillus thioparus and Desulfovibrio desulfuricans were performed using growth media supplemented with lactate as carbon and energy source and sulfate as electron acceptor, whereas the flow of air was controlled to assure oxygen limitation . Increasing the flow of air (O2 still limiting) resulted in increased total biomass (protein) concentrations and decreased sulfide concentrations. However, when the air flow surpassed the oxygen demand needed for the complete oxidation of sulfide to sulfate by Thiobacillus both organisms were washed out: Desulfovibrio because it was faced with the presence of oxygen and Thiobacillus because sulfide was no longer available. Electronic cell count and cell sizing revealed that the numbers of Thiobacillus increased with increasing rates of air flow, as was to be expected, but the increased biomass to a large extent was due to increased numbers of the sulfate-reducing bacterium. This phenomenon was attributed to the use of reduced sulfur intermediates, instead of sulfate, as terminal electron acceptor by Desulfovibrio. These compounds were produced by Thiobacillus from sulfide since oxygen was short in supply. The metabolic interactions are depicted in Fig. 1. Realizing that 1 mol of sulfate is needed in the conversion of 2 mol of lactate to acetate, and that 50% of the resulting ATP is needed for the activation of sulfate, the use of other electron acceptors enables a doubling of the Desulfovibrio biomass. In situ concentrations of soluble sulfur compounds in microbial mats are in the picomolar range for thiosulfate, tetrathionate and pentathionate, and in the nanomolar range for polysulfide . However, in these densely populated environments, and certainly in consortia, the rapid turnover of sulfur compounds would likely result in increased populations of sulfate-reducing bacteria.
4Associations of phototrophic sulfur bacteria and sulfur- or sulfate-reducing bacteria
Green sulfur bacteria can be cocultured with sulfur- or sulfate-reducing bacteria [45,46]. In these cultures the activity of the green sulfur bacteria is controlled by the activity of the sulfide-producing organisms. Whether or not such cultures are syntrophic depends on the limitation experienced by the sulfide-producing organism: when the pool of inorganic sulfur compounds is large (as in marine ecosystems) the sulfur- or sulfate-reducing organisms may depend on the activity of the phototroph with respect to availability of degradable carbon sources, but not for the availability of electron acceptors. However, in fresh water ecosystems, usually containing low total pools of inorganic sulfur compounds, the sulfide-producing organisms often rely on the activity of the sulfide-oxidizing phototrophs for the availability of an electron acceptor. In this way, a close sulfur cycle is established through which each sulfur atom cycles many times . It remains unclear if sulfate-reducing bacteria are able to increase their cell density by using inorganic sulfur species other than sulfate, it appears that this depends on the formation of such compounds by the phototrophic partner. It is known that in the presence of both extracellular sulfur and sulfide polysulfides are formed , but their presence seems unlikely in environments with low total inorganic sulfur concentrations.
Consortia are symbiotic associations of two or more organisms in a structural organization . Phototrophic consortia were first discovered by Lauterborn at the beginning of the last century . They represent associations between a colorless central rod-shaped bacterium and several cells of green- or brown-colored epibionts (Fig. 2). This highly ordered structure makes phototrophic consortia unique in the microbial world. Overall, seven different morphotypes of motile phototrophic consortia have been repeatedly observed, and are distinguished according to the color and shape of the epibionts and the presence of intracellular gas vesicles  (Fig. 3). In ‘Chlorochromatium aggregatum’, the colorless bacterium is surrounded by green-colored rod-shaped bacteria while brown epibionts are found in ‘Pelochromatium roseum’. Phototrophic consortia of the type ‘Chlorochromatium glebulum’ are bent and contain gas-vacuolated green epibionts. A fourth type of motile consortia, ‘Pelochromatium roseo-viride’, carries an inner layer of brown-colored cells and an outer one consisting of green bacteria. Two additional types of motile phototrophic consortia have been described in which epibionts are halfmoon-shaped and either green ‘Chlorochromatium lunatum’ or brown ‘Pelochromatium selenoides’. Finally, two non-flagellated types of consortia with a different arrangement are known, namely ‘Chloroplana vacuolata’ and ‘Cylindrogloea bacterifera’. ‘C. vacuolata’ has the appearance of flat sheaths which consist of parallel rows of alternating green and colorless bacteria with gas vacuoles. ‘C. bacterifera’ is an association described only once in which a colorless filamentous bacterium is completely covered with a slime layer containing green sulfur bacteria . It has to be kept in mind, that all binary names of phototrophic consortia are without standing in nomenclature since the consortia consist of two different bacteria. Consequently, all names of phototrophic consortia are given in quotation marks.
Until recently, all attempts to enrich phototrophic consortia have been unsuccessful and none of the consortium members are in culture. The isolation of a green sulfur bacterium from ‘C. aggregatum’ was reported by Mechsner , but unfortunately the strain was lost before detailed physiological studies could be made. ‘C. aggregatum’ from a eutrophic freshwater lake exhibited a pronounced chemotaxis toward 2-oxoglutarate, and using mineral media supplemented with this carbon substrate, a stable enrichment culture could be established for the first time since the discovery of phototrophic consortia . Growth of intact consortia was observed only in the simultaneous presence of 2-oxoglutarate and light.
Based on the presence of chlorosomes in the epibiont cells of ‘C. aggregatum’, it was concluded that the latter belong to the group of green sulfur bacteria [47,55]. Available 16S rRNA data indicate that all green sulfur bacteria are phylogenetically closely related (similarity values >90%) and form an isolated branch within the eubacterial radiation . Their phylogenetic position make green sulfur bacteria a promising target for molecular ecology studies. Indeed, the molecular tools for the detection of 16S rRNA gene sequences of green sulfur bacteria in complex bacterial communities could be rapidly developed [57,58]. Applying a highly specific oligonucleotide probe for fluorescent in situ hybridization (GSB-532), the phylogenetic affiliation of the epibionts with the green sulfur bacteria could be verified . In addition, a set of PCR primers was developed which permit the specific amplification of 16S rRNA gene fragments of green sulfur bacteria and their separation by denaturing gradient gel electrophoresis (DGGE) . When this method was employed to analyze the composition of a chemocline bacterial community in which the phototrophic consortium ‘P. roseum’ dominated [58,59], none of the sequences of green sulfur bacteria which were recovered matched those available in the databases . Obviously, the epibionts of ‘P. roseum’ represent a new and unknown phylotype.
Phototrophic consortia have been found in numerous lakes worldwide (compiled in ). Despite an identical morphology of phototrophic consortia found in lakes thousands of kilometers apart, epibionts and central bacteria of consortia inhabiting different lakes may be phylogenetically distinct. In order to investigate the genetic differences and the biogeography of phototrophic consortia, partial 16S rRNA gene sequences of epibionts in phototrophic consortia from various habitats were compared . Water samples were collected from lakes in northeast Germany, Spain (Lake Banyoles area, Girona) and the USA (Washington). Single intact phototrophic consortia were mechanically separated from other bacteria by micromanipulation . It was observed that differences in the 16S rRNA gene sequences of the respective epibionts clearly correlated with geographical distances between the habitats. Thus, phototrophic consortia thriving in neighboring lakes contained epibionts with identical partial 16S rRNA gene sequences. In contrast, epibionts of morphologically similar consortia but originating from either German, Spanish or American lakes contained distinctly different sequences.
The type of interaction in phototrophic consortia is still unknown. All strains of green sulfur bacteria isolated to date are obligate anoxygenic photolithotrophs which use sulfide, elemental sulfur, and – in some instances – thiosulfate as electron donor. In the presence of sulfide and CO2, a limited number of organic carbon compounds (acetate, propionate, pyruvate) may be photoassimilated . In the stable cocultures of green sulfur bacteria and sulfate- or sulfur-reducing bacteria a close sulfur cycle is established through which each sulfur atom cycles many times. It has been proposed that a syntrophic sulfur cycle exists also in consortia and represents the physiological basis of the symbiosis . Since ambient sulfide concentrations in the natural habitat of phototrophic consortia are usually very low , a recycling of reduced sulfur would be advantageous and ensure a continuous supply of the anoxygenic photosynthetic epibionts with a suitable electron-donating substrate.
Electron transport in the epibiont cells is stimulated by exogenous sulfide . This indicates that sulfide indeed serves as photosynthetic electron donor of the epibionts similar to free-living strains of green sulfur bacteria. Several lines of evidence indicate however, that the central rod of phototrophic consortia is not a classical sulfur- or sulfate reducing bacterium. Firstly, the central rod was recently shown to be a member of the β-subgroup of Proteobacteria, whereas sulfur- and sulfate-reducing bacteria are found in the δ-subgroup . Secondly, the recently established enrichment culture of the brown-colored ‘P. roseum’ depends on millimolar concentrations of exogenous sulfide (J. Glaeser and J. Overmann, unpublished) which is inconsistent with the presence of an internal sulfur cycle within the consortia.
Several observations indicate a close interaction between the epibionts and the central rod. The number of epibionts per consortium exhibit a non-random frequency distribution with a distinct maximum, indicating a coordinated cell division of all epibionts. Furthermore, in many specimen of intact phototrophic consortia, the central rod has been observed to be in a later phase of cell division . Therefore, the chemotrophic bacterium must be able to grow and multiply while in association with the phototrophic epibiont. These observations indicate that under environmental conditions, the cell division of all epibionts and the central rod proceed synchronously. Intact consortia exhibit sudden changes in the direction of movement when suddenly illuminated with light of high intensity (‘Schreckbewegung’). Also they reverse the direction of movement when entering the dark, this so-called scotophobic response was demonstrated for ‘C. aggregatum’.
Of the cells which constitute the motile phototrophic consortia, only the central colorless bacterium is monopolarly flagellated . The photoreceptor of the scotophobic response of the consortium ‘C. aggregatum’ exhibits an action spectrum which corresponds to the absorption spectrum of the bacteriochlorophyll present in the green sulfur bacterial epibiont . Thus, a rapid interspecies signal transfer must occur between the nonmotile, light-sensing epibiont and the motile colorless central bacterium.
6Implications for the microbial ecology of sulfureta
Anoxygenic phototrophic bacteria are of significance for the biogeochemical cycling of carbon and sulfur in stratified ecosystems like meromictic lakes and microbial mats [40,41,63–66]. In the chemocline of some lakes, phototrophic consortia contribute significantly (i.e. up to 19%) to total numbers of bacteria ; they may even represent as much as two thirds of total bacterial biomass . Thus, phototrophic consortia must be significant for the carbon and sulfur turnover in these freshwater ecosystems.
While the bacterial communities in the chemocline of some stratified lakes are composed mainly of phototrophic consortia, others are dominated by gas-vacuolated forms of green sulfur bacteria . With the exception of the gliding filamentous Chloroherpeton thalassium, all green sulfur bacteria are nonmotile. Based on the experimental evidence described in the previous paragraphs it can be speculated that one major selective advantage of the stable association of green sulfur bacteria in phototrophic consortia is their motility, coupled to the chemotaxis and scotophobic response. Intact consortia sampled from the chemocline of a freshwater lake had a buoyant density much higher than that of the ambient water (1047 versus 996 kg m−3). If the consortia were immotile, such a density difference would result in sedimentation toward a zone of complete and continuous darkness. In addition, the scotophobic response and chemotaxis toward sulfide may allow the motile consortia to actively accumulate in depth horizons where conditions are favorable for their growth.
Although devoid of flagella, some free-living green sulfur bacteria like Pelodictyon phaeoclathratiforme are capable of regulating their buoyancy and hence can alter their vertical position in the water column. In these strains, gas vesicle formation is induced at very low light intensities . Movements mediated by gas vesicle formation differ considerably from flagellar motility with respect to direction and response time. In green sulfur bacteria regulation of cell density in this way takes one to several days  and therefore is not a likely mechanism to control diel migrations in the water column. In addition, cells regulating their gas vesicle content are only able to change their vertical position, but cannot respond to lateral inhomogeneities in their pelagic habitat. Based on the above considerations, one would expect that short-term fluctuations in environmental parameters and lateral heterogeneity is more pronounced in those lakes where phototrophic consortia predominate the chemocline bacterioplankton assemblage.
7General implications for the ecology and evolution of bacterial communities
Phototrophic consortia are not only of ecological significance in lake ecosystems, but also excellent model systems for the evolution of the bacterial interactions. The picture which begins to emerge from the limited dataset available is that these specialized associations evolved concomitantly with the evolutionary radiation of green sulfur bacteria, and that consortia with phylogenetically different epibionts occupy geographically confined areas. This latter observation is in contrast to the evidence gathered for some α-Proteobacteria: identical partial 16S rRNA gene sequences belonging to the so-called SAR11 phylotype have been detected both in the Atlantic and in the Pacific Ocean . Similarly, rhizobia strains which occur as symbionts in stem nodules of different Aeschynomene legume species show sequence divergences of their 16S rRNA genes of <1%. On the subspecies level, however, it has been shown that 3-chlorobenzoate-degrading soil bacteria are endemic to distinct geographic regions . The epibionts of phototrophic consortia, however, to our knowledge are the first example for endemism of bacteria on the species level. The sequence divergence of partial 16S rRNA gene sequences of different epibionts of phototrophic consortia is high (up to 4%), especially when considering the close relatedness of green sulfur bacteria as a whole . Possibly, the chemocline habitat of some stratified freshwater lakes is biogeographically more isolated than the habitats of bacteria in soil. Also, the phylogenetic breadth and age of specific interactions like those in phototrophic consortia may be considerable.
Based on the many cases which are known of a close interaction between various bacterial members of the sulfur cycle, it was not anticipated that a closed sulfur cycle may not operate in the highly specialized associations of phototrophic consortia. With the assumption of a general validity for consortia, it may well be that the selective pressure driving the formation of the highly specialized consortia during evolution of the green sulfur bacteria is not the only adaptation to low concentrations of sulfur compounds. A multidisciplinary study of the sulfur and carbon cycles in a meromictic lake ecosystem showed that the metabolism in situ of phototrophic sulfur bacteria on the one hand and sulfur- and sulfate-reducing bacteria on the other hand are only loosely coupled [65,66]. Other than a more rapid locomotion to hot spots of substrates (sulfide) or upwards toward light by chemo- or phototaxis, no other selective advantage for these associations are evident to date. These recent findings taken together, the general implication would be (1) that bacterial interactions which occur in natural bacterial communities may differ significantly from the paradigm of hydrogen or sulfur syntrophy, and may be based on a different physiology than hitherto assumed, (2) that isolation of biogeochemically significant, but so far ‘unculturable’ bacteria possibly may be substantially facilitated by exploiting such interactions between different types of bacteria.