The osmoadaptation of most micro-organisms involves the accumulation of K+ ions and one or more of a restricted range of low molecular mass organic solutes, collectively termed ‘compatible solutes’. These solutes are accumulated to high intracellular concentrations, in order to balance the osmotic pressure of the growth medium and maintain cell turgor pressure, which provides the driving force for cell extension growth. In this review, I discuss the alternative roles which compatible solutes may also play as intracellular reserves of carbon, energy and nitrogen, and as more general stress metabolites involved in protection of cells against other environmental stresses including heat, desiccation and freezing. Thus, the evolutionary selection for the accumulation of a specific compatible solute may not depend solely upon its function during osmoadaptation, but also upon the secondary benefits its accumulation provides, such as increased tolerance of other environmental stresses prevalent in the organism's niche or even anti-herbivory or dispersal functions in the case of dimethylsulfoniopropionate (DMSP). In the second part of the review, I discuss the ecological consequences of the release of compatible solutes to the environment, where they can provide sources of compatible solutes, carbon, nitrogen and energy for other members of the micro-flora. Finally, at the global scale the metabolism of specific compatible solutes (betaines and DMSP) in brackish water, marine and hypersaline environments may influence global climate, due to the production of the trace gases, methane and dimethylsulfide (DMS) and in the case of DMS, also couple the marine and terrestrial sulfur cycles.
In micro-organisms, no plausible mechanisms exist which allow the maintenance of an intracellular osmotic pressure lower than that of the bathing medium . All microbial cells must maintain an intracellular osmotic pressure somewhat greater than that of the growth medium in order to generate cell turgor pressure, which is considered to be the driving force for cell extension, growth and division [1–3]. Therefore, the ability to adapt to fluctuations in the osmolarity of the growth medium is of fundamental importance for growth and survival, and micro-organisms have developed a number of osmoadaptive strategies to cope with such fluctuations.
In the extremely halophilic Archaea, inorganic salts (mainly KCl) are accumulated intracellularly to molal concentrations, in order to balance the osmotic pressure of the growth medium [1,4,5] and the intracellular enzymes and organelles have evolved to function at high ionic strength . However, this KCl strategy is relatively inflexible and halobacteria generally display a relatively narrow range of adaptation for a specific high osmolarity environment . Indeed the cell envelopes of most halobacteria are structurally unstable outside this limited range of concentrations and dissolve at NaCl concentrations below approximately 1 M [1,6,7].
In contrast, in Eubacteria, actinomycetes, algae, fungi and yeasts, osmoadaptation is dependent upon the accumulation of K+ ions and one or more of a restricted range of low molecular mass organic solutes, collectively termed ‘compatible solutes’ due to their compatibility with biological function . These solutes are synthesised or accumulated from the growth medium to high intracellular concentrations, in order to balance the osmotic pressure of the growth medium and maintain cell turgor. In addition to their strictly osmotic function, compatible solutes may also protect intracellular enzymes and organelles against the potentially inhibitory effects of changes in intracellular ionic strength and water availability [4,5,8–12]. Compatible solutes share a number of common properties  and can be divided into several sub-classes, including simple sugars, heterosides, amino and imino acids, amino acid derivatives, such as betaines (fully N-methylated quaternary amino acids), ectoines and tertiary sulfonium compounds [1,2,4,5,13]. An extensive, but far from exhaustive list of the compatible solutes commonly synthesised or accumulated by active uptake from environmental sources by micro-organisms is provided in Table 1. However, it is not within the scope of this paper to review the regulation of compatible solute synthesis and accumulation, microbial osmoregulation or the detailed nature of solute compatibility at the molecular level, and these aspects have been more than adequately reviewed elsewhere [1,2,5,13,20,32–42]. In this review, I would like to discuss the potential ecological roles of compatible solutes, other than for the osmoadaptation of the individual organisms and the consequences of the release of compatible solutes to the environment on the microbial community, ecosystem function and the regulation of the global climate.
Table 1. Predominant compatible solutes accumulated by eukaryotic micro-algae, yeasts, bacteria, actinomycetes and Archaebacteria
In addition to their role in providing cells with protection against the deleterious effects of elevated environmental osmotic pressures, there is an increasing body of literature demonstrating that compatible solutes can also fulfil a range of other functions within the cell. Thus, the evolutionary pressures selecting for or against the accumulation of a specific compatible solute by an individual organism may not be based solely on the osmotic function of the solute or solutes accumulated and the range of osmotic variation to which the organism is typically subjected, but also on these secondary functions and their relation to the organism's niche.
2.1Compatible solutes as intracellular reserves
Compatible solutes are accumulated to high, often molal concentrations in the cytoplasm and therefore represent substantial intracellular stocks of carbon and/or nitrogen. Consequently, compatible solutes could fulfil a role as intracellular reserves of nutrients and energy. This may be especially true for organisms which are able to accumulate mixtures of both carbohydrate and nitrogen containing solutes. For example, the haloalkophilic phototrophic bacterium Ectothiorhodospira halochloris synthesises glycine betaine as its major compatible solute, but can also accumulate both trehalose and ectoine as secondary compatible solutes [43–45]. In this organism, ectoine accumulates and intracellular trehalose pools are small or completely absent when fixed nitrogen is abundant. Conversely, during nitrogen starvation, trehalose is accumulated and completely replaces ectoine (2 N per molecule) and partially replaces glycine betaine (1 N), thus liberating nitrogen for cell growth . If nitrogen again becomes available, the trehalose pool is metabolised and the intracellular pools of ectoine and glycine betaine increase . Moreover, the activity of the trehalase responsible for the breakdown of trehalose is stimulated by high glycine betaine concentrations  and thus is indirectly regulated by the availability of nitrogen for glycine betaine synthesis.
Similarly dimethylsulfoniopropionate (DMSP) synthesising micro-algae may accumulate nitrogenous compatible solutes such as proline, during early bloom formation when fixed nitrogen is abundant [16,47]. These are subsequently replaced by DMSP or carbohydrate compatible solutes (e.g. sucrose) during growth as exogenous nitrogen sources become increasingly limiting . Thus, the synthesis of N-containing compatible solutes allows a luxury uptake of nitrogen by the micro-algae, above their immediate growth requirements, and these solutes may subsequently serve as endogenous nitrogen reserves. This duality of function is also shown by the compatible solutes glutamate and trehalose in Escherichia coli. In this bacterium, osmoadaptation under N-limited or balanced growth conditions is distinctly biphasic and consists of a rapid accumulation of K+ ions from the bathing medium and the synthesis of the amino acid glutamate, which are then slowly replaced by trehalose during the slower second phase of osmoadaptation [28,48]. In contrast, during growth in glucose-limited continuous cultures, there was no decline of the glutamate pool and trehalose was accumulated to a lower concentration, thereby conserving carbon for cell growth and respiration . Interestingly, in E. coli transcription of the ots BA operon which encodes the enzymes for trehalose synthesis, is under the control of rpoS, a putative sigma factor for the expression of stationary phase genes [49,50]. Whilst this effectively leads to trehalose synthesis during osmotic stress, when osmoprotectants are not present in the growth medium [2,34,37], due to the cessation of cell growth , it would also tend to lead to trehalose accumulation by stationary phase cultures or those subjected to other stresses which reduce growth rates. Similarly in Saccharomyces cerevisiae and other yeasts and fungi, trehalose naturally accumulates, often to very high concentrations in stationary phase cells, including cells grown at low osmolarities . This phenomenon led to the proposal that trehalose functioned as a carbon reserve in yeasts, fungi and their spores [51,52]. However, this view has been questioned, since trehalose is accumulated even under conditions of carbon starvation and its synthesis is often at the expense of the known carbon storage compound glycogen and therefore does not result in an overall increase in cell carbohydrate content [51–54]. These observations, and the fact that trehalose is accumulated under stress conditions other than starvation, have prompted the proposal that trehalose is a stress metabolite, rather than a carbon reserve in yeast (see  for review and below), although these two roles are not mutually exclusive.
2.2Protection against stresses other than osmotic stress
The chemical properties of specific compatible solutes and their modes of interaction with biological macromolecules may also increase the tolerance of micro-organisms able to synthesise or accumulate these solutes to other environmental stresses. Such secondary benefits of compatible solutes may have played an important role in the evolution of the osmoadaptive strategies of micro-organisms in environments where these other environmental stresses frequently or sporadically occur.
Accumulation of the disaccharide trehalose has been shown to be associated with increased thermotolerance of bacterial, yeast, fungal and slime mould cells and their spores. In baker's yeast, S. cerevisiae, mild heat shock induces the rapid biosynthesis of trehalose to high cytoplasmic concentrations and the increased thermotolerance of the cells to subsequent heat treatments is directly related to their trehalose content . In yeast strains impaired in trehalose synthesis, thermotolerance also correlates well with trehalose content  and a temperature sensitive yeast strain has been shown to be defective in trehalose biosynthesis . In addition in yeast, agents which induce heat shock responses (e.g. CuSO4, H2O2 and ethanol) have been demonstrated to also cause trehalose accumulation . Similarly in E. coli stationary phase induced thermotolerance has been shown to be dependent upon the rpoS regulated expression of the ots A and B genes for trehalose synthesis, although the levels of trehalose synthesised on entry into stationary phase are very much lower than those associated with osmotic induction .
In general, survival, dormant or dispersal phases, such as the various spore types produced by actinomycetes, yeasts, fungi and slime molds are intrinsically more resistant to environmental stresses than their respective vegetative counterparts and these structures are known to accumulate very high trehalose concentrations [55,60–62]. In experiments with the actinomycete Streptomyces griseus, where the growth medium glucose concentration was experimentally manipulated to yield spores containing between 1 and 25% trehalose by dry weight, a strong correlation was found between the trehalose content of the spores and their degree of thermotolerance . Similarly for spores of the slime mould Dictyostelium discoideum, which were manipulated to have variable intracellular contents of trehalose and glycogen. The degree of thermotolerance correlated with increasing intracellular trehalose content, but not with glycogen content, and trehalose added exogenously to spore suspensions prior to heat treatments also provided a protective effect . Moreover, the stationary growth phase associated accumulation of trehalose by fungi, yeasts and E. coli[18,49–51,55] and the concurrent elevated thermotolerance of these cells may be associated with the more general changes in cell physiology, which occur upon the cessation of active growth and the entry of the cells into a resting, dispersal or survival adapted stage [64,65].
At the molecular level, all compatible solutes would be expected to increase the thermotolerance of enzymes to some extent, since they are preferentially excluded from the immediate hydration sphere of proteins [35,42,66,67]. This situation creates a thermodynamic disequilibrium due to the non-homogeneous distribution of the solute within the cell water. The degree of this disequilibrium is dependent, at least in part, upon the volume of water from which the solute is excluded and therefore is reduced by the protein occupying a small volume. Thus, excluded solutes tend to promote protein sub-unit assembly and stabilise secondarily folded, tertiary protein structures as these inter- and intra-protein interactions decrease the total surface area of the protein, and hence the volume of cell water from which the solute is excluded, thereby reducing the extent of the thermodynamic disequilibrium [35,42,66,67]. Therefore, the presence of high concentrations of compatible solutes would tend to inhibit the thermal denaturation of proteins, as unfolding of tertiary protein structures increases the protein's surface area and consequently the volume of the protein hydration sphere, and the water volume from which the solute is excluded (Fig. 1) [35,42,66,67]. In vitro studies support this hypothesis, as compatible solutes, including sugars, polyols and glycine betaine, have been demonstrated to increase the thermotolerance of individual enzymes and to inhibit the thermal denaturation of bovine serum albumin and other proteins in solution [35,67–70]. The greater efficacy of trehalose compared to other compatible solutes in vivo may be due to its interaction with membrane lipids, which are thought to be a critical target for heat treatments [71,72]. Trehalose can directly interact with the polar headgroups of membrane phospholipids via hydrogen bonding between the hydroxyl groups of the sugar and the phosphate headgroups of the lipid [73,74]. Most studies of trehalose/phospholipid interactions have mainly focused on the influence of these interactions on membrane stability during freezing, drying and freeze drying (see following sections). However, Rudolph et al.  have demonstrated that hydrogen bonding also occurs between trehalose and hydrated artificial membranes and thus trehalose may also play a role in promoting resistance of natural membranes to heat damage.
Compatible solutes, especially sugars and polyols, have long been known to increase the resistance of microbial cells to freezing . For example, the survival ratio of S. cerevisiae cells after freezing is dependent upon their initial trehalose content  and the freeze tolerance of Phycomyces spores decreases during germination as intracellular trehalose is metabolised and its concentration declines . Conversely, heat treatments, which stimulate active trehalose biosynthesis, concurrently increase the freeze tolerance of germinating conidiospores of Neurospora crassa. Indeed, for many years, the compatible solutes sucrose and glycerol have routinely been used as additives for microbial samples to be stored frozen or to be freeze dried [80,81] and exogenously added trehalose at a concentration of 0.3 M has been shown to increase the viability and preserve the acidification properties of cultures of Lactobacillus bulgaricus used in the dairy industry, to repeated freeze-thaw cycles . Similarly in vitro, a wide range of compatible solutes, including glycerol, glycine betaine, trehalose, sucrose, proline and DMSP, have been demonstrated to inhibit the denaturation of isolated cold sensitive enzymes such as phosphofructokinase during freeze-thaw cycles [83–86]. Moreover, trehalose has been shown to protect tobacco, carrot and human pancreatic cells, and even murine embryos, during freezing [87–89] and is known to be accumulated to high levels by many frost resistant insects [89–91].
At the molecular level freezing damage may be due to either dehydration or the high intracellular solute and ionic concentrations, which result from the partial freezing of the cell water , effects similar to those which occur during osmotic dehydration, although during freezing these effects may be more extreme due to the greater effective reduction in water availability. However, it has been estimated that at least 5% of cell water is unfreezable . Due to the similarity with osmotic stress, it would be expected that compatible solutes would inherently protect cytoplasmic proteins during freezing due to their exclusion from the hydration sphere of the protein, which effectively retains a layer of free water around the enzyme [35,42,66,67,84,86]. The same exclusion mechanism can also protect enzymes from the denaturing effects of the high cytoplasmic ionic concentrations which result from the partial freezing of the cell water, since, although ions may bind to exposed binding sites on the protein surface, this would not result in increased solubility of the protein and unfolding of the tertiary structure to expose further ion binding sites , due to the energetic considerations of the solvent exclusion theory, that the protein occupies a small volume to minimise the degree of thermodynamic disequilibrium (Fig. 1) [35,42,66,67]. Since unfolding of the protein's tertiary, folded structure increases the surface area of the protein, increasing the effective volume of cell water from which the solute is excluded, is thermodynamically unfavourable . Protection of individual enzymes by compatible solutes from the inhibitory effects of high ionic strength has been demonstrated in vitro [4,8,10,11]. For example, glucose 6-phosphate dehydrogenase from Spirulina subsala was 50% inhibited in the presence of 1.25 M NaCl, but retained full activity at up to 1.5 M NaCl, if a similar concentration of glycine betaine was also present, despite the overall increase in total solute concentration . Similarly 1 M glycine betaine reduced NaCl inhibition of glutamine synthetase from Synechocystis DUN 52 by 10–30% over an NaCl concentration range of 0.8–2.0 M  and a similar protection of some enzymes from increased ionic strength has also been reported for glycerol [1,4]. Thus, the interactions of compatible solutes with biological macro-molecules can explain the increased tolerance these solutes provide against freezing stresses.
Many organisms, ranging from bacteria to plants and invertebrates, are able to survive long periods of drought in a desiccated state, where greater than 99% of the cell water has been lost [92–94]. Upon subsequent re-wetting, these organisms rapidly swell and resume active metabolism, often within periods of only a few minutes . Amongst these anhydrobiotic organisms, the best known examples are resurrection plants, such as Selaginella lepidophylla and Myrothanus flabellifolia, adults and larvae of soil nematodes like Ditylenchus dipsaci, tardigrades such as Adoribiotus coronifer, the brine shrimp Artemia salina, some parasitic protozoa and the macrocysts of the slime mould Dictyostelium[73,92–94]. Whilst the underlying mechanisms of desiccation resistance are not fully understood, a common feature linking all the above organisms is the presence of very high concentrations of the disaccharide trehalose in their dried tissues. Indeed, trehalose can account for as much as 20% of total tissue dry weight and desiccation tolerance correlates directly with trehalose content [73,93].
Similarly, in micro-organisms desiccation tolerance has been shown to correlate with the accumulation of the non-reducing disaccharides, trehalose and to a lesser extent sucrose , although there is also a single report that the tetrahydropyrimidine compatible solutes ectoine and hydroxyectoine promote increased desiccation tolerance in at least some strains of E. coli. For example, S. cerevisiae exponential phase cells which contained little trehalose exhibited negligible desiccation tolerance, whereas stationary phase cells accumulated trehalose and this correlated with increased desiccation tolerance . This correlation between trehalose content and desiccation tolerance appears to be independent of growth phase, since exponential phase yeast cultures subjected to heat shocks rapidly synthesised trehalose and simultaneously acquired increased resistance to desiccation . Indeed, even additions of trehalose as an exogenous excipient to yeast cultures prior to drying increased the desiccation tolerance of both exponential and stationary phase cells . Equally, in E. coli, induction of trehalose synthesis by osmotic shock, or addition of trehalose to culture suspensions prior to drying, significantly increased their desiccation resistance and stationary phase cells were found to be inherently more resistant than exponential phase cells [95,97]. Similar protection of cells by exogenously added trehalose or sucrose was also observed during studies of E. coli and Bacillus thuringiensis during freeze drying . This effect was proposed to be dependent upon membrane phase transitions during the drying process, which allowed the sugars to diffuse into the cells  and similar phase transitions would allow intracellular trehalose to diffuse from the cells to protect the outer face of the cell membrane, as for high levels of protection the presence of the sugar on both sides of the membrane appears to be required , although in yeast a specific trehalose carrier may be involved in transferring cytoplasmic trehalose to the outer face of the membrane during drying .
In contrast to trehalose, the compatible solute glycine betaine, whether added as an excipient or accumulated intracellularly by E. coli, had no influence on the tolerance of the cells to a subsequent desiccation challenge . On the other hand, it has been reported that intracellular accumulation of glycine betaine by several strains of lactobacilli did increase their tolerance to drying [101,102]. However, it should be noted that in these experiments considerable quantities of cell water were still present following drying (water activity=0.12) and the cells were only maintained under desiccating conditions for 72 h [101,102]. Thus, the increased survival may have been due to protection of the cells by glycine betaine functioning as a compatible solute during the drying process, rather than a true protection of the cells in the dried state. In contrast, in our manipulations , the E. coli cells treated with exogenous trehalose or osmotically induced to accumulate trehalose intracellularly retained 2–6% of the initial culture viability following 50 days storage in an evacuated desiccator, whereas for cells treated with glycine betaine or induced to accumulate glycine betaine intracellularly, and untreated controls, no viable cells were present after 7–20 days storage under the same conditions (Fig. 2).
In vitro studies have also clearly demonstrated that trehalose, sucrose and other sugars are able to stabilise isolated natural or artificial membrane vesicles and a wide range of proteins and other bio-molecules in the dried state, with trehalose generally being the most effective additive [73,74,92,103–108]. This stabilisation has been most intensively studied for membrane lipids [73,74,103,108]. During drying in the absence of sugars, membrane vesicles undergo fusion and leak their contents due to phase transitions of the lipid bilayers between the liquid-crystal and gel phases during both the drying and rehydration of the membranes . Thus, vesicles of isolated lobster endoplasmic reticulum dried without sugars suffer fusion and complete loss of Ca2+ transport activity upon rehydration, whereas vesicles dried with disaccharides retained their contents and transport activity when reconstituted . Thus, in vivo trehalose would be expected to prevent inter-membrane fusions of the cell membrane and intra-membrane fusions between the cell and the other intracellular membrane systems of micro-organisms, thereby maintaining the integrity of these membrane systems and their barrier function.
The biological function and specificity of function of isolated enzymes dried with disaccharides is also retained upon subsequent rehydration [73,104,106,107]. For example, Colaço et al.  demonstrated that a wide range of DNA restriction endonucleases and other DNA modifying enzymes commonly used in molecular biology could be air-dried without significant loss of activity in solutions of trehalose, but not those of other sugars or polyols. A remarkable and unique property of these trehalose-dried enzyme preparations was that no loss in biological activity or the specificity of this activity was observed following prolonged storage (1–3 months) at temperatures of 55 and 70°C, temperatures which would completely denature the various native enzymes, which normally need to be transported and stored at sub-zero temperatures . Similarly, blood typing antibodies dried in trehalose solutions can be stored for periods of years at 37°C, without any significant losses in biological activity or the specificity of this activity [105,107]. Indeed, the unique ability of trehalose to preserve the function of dried biological macromolecules and even sensitive mammalian cells, such as human erythrocytes, has attracted increasing interest from biotechnology companies world-wide and is the subject of several international patents [107–110].
All compatible solutes would be expected to provide some level of protection to cells and intracellular macromolecules during at least part of the drying period involved in desiccation, due to the principles of the excluded solute theory, which would offer protection against both the reduced intracellular water potential and the resultant increases of intracellular ionic strength caused by the water loss [4,8,10,11,35,42,66,67,86] (Fig. 1). However, in the truly desiccated state, insufficient cell water remains for this mechanism to operate. Since, it is estimated that 0.3–0.4 g H2O g protein−1 is required to maintain even a single mono-layer of bound water around proteins. Whereas water contents of below 0.1 g H2O g dry weight−1 are typical of desiccated cells and for dried colonies of the cyanobacterium Nostoc commune, Artemia cysts and some plant seeds values ranging between 0.02 and 0.06 g H2O g dry weight−1 have been recorded (see  and references therein). Thus, mechanisms other than solute exclusion are required to account for the protection provided by some solutes, principally disaccharide sugars, during desiccation.
Sugars and in particular the disaccharides, trehalose and sucrose have been proposed to stabilise dried biological membranes through direct hydrogen bonding between the sugar and the phosphate of the phospholipid head group. The phospholipids of biological lipid bilayers are normally hydrated, with 10–12 water molecules hydrogen bonded to each phosphate group in the case of phosphatidylcholine [73,92] and this hydration shell is thought to provide the physical barrier which inhibits the fusion of adjacent lipid bilayers . Removal of this hydrogen bonded water during drying in the absence of sugars has two major effects on membranes. Firstly, it removes the barrier which prevents the fusion of adjacent lipid bilayers, resulting in massive fusion of the dehydrated membranes [103,108,111]. Secondly, the removal of hydrogen bonded water decreases the spacing between the phospholipid headgroups, resulting in an increased packing density of the lipids within the membrane [73,92,94,108]. The greater proximity of the lipid hydrocarbon side chains enhances Van der Waals interactions between adjacent lipids, causing an increase in the phase transition temperature (Tm) of the membrane and a transition of the membrane from the liquid crystalline to the gel phase at environmentally relevant temperatures. Additionally, there may be a lateral separation of the constituent lipids present in the membrane lipid mixture due to their varying individual Tms, which causes defects in the lipid bilayer [73,74,92,112]. These membrane defects and phase transitions between the liquid crystal and gel phases disrupt the membrane barrier function, resulting in the leakage of solutes. In contrast, membrane vesicles and liposomes dried with disaccharides do not undergo fusion and retain their contents when rehydrated [73,103,108,110,111]. There is now a large body of evidence that disaccharides preserve the structural integrity of membranes in the dried state by substituting for the structural water hydrogen bonded to the phospholipid headgroups. This maintains the normal headgroup spacing, the loosely packed structure and low Tm of the hydrated membrane and thereby prevents the occurrence of membrane defects and phase transitions during drying and rehydration (Fig. 3). Nuclear magnetic resonance and X-ray diffraction studies have demonstrated a close interaction between trehalose and the hydrophilic regions of the lipid headgroups, including the phosphate moieties, and the glycerol and carbonyls of the interfacial region . Similarly infrared spectroscopy studies indicate the presence of hydrogen bonds between the phosphate of the headgroup and the hydroxyl groups of the sugar molecules  and molecular modelling has shown that trehalose could be accommodated between the phosphates of adjacent lipids . In addition, direct measurements of membrane phase transition temperatures of dried isolated, artificial and natural membranes, and the membranes of whole bacteria, yeasts and pollen grains have demonstrated that sugars reduce their Tms relative to those of membranes dried without sugars [75,98,111,113–115]. For example, for intact cells of E. coli and B. thuringiensis dried in the absence of sugars, the Tm as measured by Fourier infrared spectroscopy increased by 30–40°C. Whereas the Tm for cells dried with either sucrose or trehalose remained close to that of the hydrated cells of 5–10°C . Thus, at room temperature the membranes of cells dried without sugars would be in gel phase and would undergo a phase transition upon rehydration, whilst the membranes of cells dried with sucrose or trehalose would be in the liquid crystal phase and not suffer a damaging phase transition when rehydrated. In the same study, 70 and 57% of trehalose and 56 and 44% of sucrose dried E. coli and B. thuringiensis cells respectively, survived drying, compared to only 8 and 14% of cells dried without sugars .
A similar water replacement hypothesis has also been proposed for the stabilisation of dried proteins by sugars, with the sugar forming multiple external hydrogen bonds with the protein, replacing essential structural water molecules [86,98,116,117]. The major experimental evidence supporting this hypothesis has come from infrared spectroscopy studies of isolated dried proteins and intact cells, which show that the shift of the amide II peak to a lower wave number when proteins are dried does not occur in the presence of sugars [98,117]. However, whilst such hydrogen bonding may occur between dried proteins and sugars, this cannot account for the extremely high stability of trehalose dried proteins, as other sugars and polyols would be equally capable of forming multiple hydrogen bonds with the protein, but in practice are considerably inferior to trehalose in preserving proteins in the desiccated state . For example, if the spatial distribution of the hydroxyl groups of the sugar was the crucial factor, then glucose should be equally as effective as trehalose (di-glucose), whereas if molecular mimicry of water was the major criterion, then scyllo-inositol should be the most effective protectant, since all its hydroxyl groups are axial. However, in vitro experiments have clearly demonstrated that both glucose and scyllo-inositol are very much less effective at stabilising dried proteins than trehalose .
A second hypothesis, the ‘glassy state theory’, proposes that it is the propensity of sugar solutions to form amorphous glasses when dried, which is the critical factor, as freedom of motion of biomolecules entrapped within these sugar glasses would be extremely limited and the kinetics of degradative molecular reactions therefore insignificant [118,119]. Thus, the biological molecules embedded within the glass would be stable at all temperatures below the glass-liquid transition temperature (Tg) of the glass. However, as pointed out by Colaço and colleagues [106,107], the glassy state theory again provides only a partial explanation for the extraordinary stability of trehalose dried proteins. Since it does not take into consideration the chemical nature of the glass former itself and cannot account for the apparent stability of some trehalose dried enzymes over long storage periods at temperatures considerably greater than their experimentally determined Tg. These authors attribute the much greater efficacy of trehalose compared to other glass forming sugars and polyols in preserving the structural integrity and biological activity of dried proteins to the chemical stability of this non-reducing disaccharide and in particular the inability of trehalose to participate in the Maillard reaction. The Maillard reaction consists of a cascade of chemical reactions which are responsible for the typical ‘browning’ and spoiling of dried food products [120–122]. These reactions are driven by water loss as the initial spontaneous condensation reaction between the reactive carbonyl group of the sugar and the amino group of the protein, to form a Schiff's base is reversible in the presence of water. However, at low water activities, the reaction equilibrium favours the formation of the Schiff's base and irreversible Amadori or Heyns rearrangements of this base initiates a complex cascade of further reactions, which ultimately result in fragmentation and cross linking of the proteins and the production of the melanoid pigments which are responsible for the characteristic browning of the dried product [120,121]. This hypothesis is supported by direct experimental evidence from studies of isolated proteins and peptides dried in various sugars and polyols, and stored over a range of temperatures. These studies demonstrated a rapid formation of melanoid pigments and a complex array of cross linked proteins with corresponding losses of biological activity in all the samples with the exception of those dried with trehalose . Thus, at least for isolated proteins and especially at higher but nonetheless environmentally relevant temperatures, the relative reactivity of the glass former itself would appear to have a significant influence on the stability of dry macromolecules. However, it must be noted that the cytoplasmic glasses, which would form upon the drying of whole cells or organisms, would be chemically very much more complex than the single protein glasses which have generally been used in laboratory stability studies and thus the chemistry within these glasses would be expected to be correspondingly more complicated than in these single component systems.
In conclusion, the ability of disaccharides to preserve dried micro- and multi-cellular organisms appears to be linked to their ability to directly replace water hydrogen bonded to cellular macromolecules and their inherent glass forming properties, whereas the greater efficacy of trehalose is related to its anomalously high glass transition temperature , and its chemically inert nature, which prevents deleterious chemical reactions between the dried biomolecules and the glass matrix. This unique combination of physical and chemical properties makes trehalose nature's choice as a desiccation protectant and is at least to a large degree responsible for longevity of anhydrobiotic organisms even under harsh environmental conditions.
2.2.4Trehalose a panacea for environmental stress
In addition to its osmotic role, trehalose provides cells with protection against heat, freezing, desiccation and other environmental stresses (Table 2). Induction of trehalose synthesis in response to one stress concomitantly increases the tolerance of cells to other stresses [55,56,79,97], indicating that trehalose accumulation is a crucial factor in these stress responses. Similarly the critical role of trehalose synthesis in stress tolerance is confirmed by in vitro studies, which have shown that trehalose can protect isolated enzymes, antibodies, natural and artificial membranes against desiccation, high or low temperatures, and normally inhibitory concentrations of some chemical species [73,75,84,86,92,103–111,123]. As a consequence of these diverse functions, trehalose has previously been proposed to be a general stress metabolite in yeast cells .
Table 2. Some conditions known to induce trehalose synthesis in bacteria, yeasts, fungi, slime moulds and their spores
The regulation of trehalose synthesis in fungi, yeast and bacteria also indicates a more general role of trehalose as a stress protectant, as its synthesis is induced upon the cessation of growth [18,49,50,55,96]. Hence trehalose accumulation is associated with the general changes in cell physiology which occur during the transition from an active growth phase to a survival or dispersal phase [64,65]. Indeed, in E. coli the expression of many genes associated with tolerance to adverse environmental conditions, including those for trehalose biosynthesis, are regulated by the rpoS encoded stationary phase sigma factor [49,50,124,125]. The accumulation of trehalose by the true survival and dispersal phases represented by the diverse spore types, produced by yeasts, fungi, actinomycetes and slime moulds is also consistent with a general function in stress tolerance and the resistance of these spores to several environmental stresses is directly correlated with their trehalose content [55,60,61,63]. Indeed, in at least some spores, trehalose appears to play a role in the maintenance of dormancy, as the lag time for germination is directly related to trehalose content  and there is a rapid breakdown of trehalose during germination, which is often not coupled to metabolism, as the breakdown products are excreted into the growth medium [55,126,127].
2.3DMSP an anti-grazing and bacteriocidal compound?
DMSP is accumulated as a compatible solute by many marine micro-algae, especially the abundant bloom forming planktonic groups, Haptophyta and Dinophyta [17,132]. In addition to its osmoregulatory function, it has been proposed that DMSP may also play a role as a grazing deterrent in these algae . Ingestion of the micro-algae by zooplankton or protozoa leads to the mixing of DMSP with DMSP-lyases, produced by the algae [133,134], resulting in the production of dimethylsulfide (DMS) and acrylate, which is proposed to be the active component [133,135]. Several studies have demonstrated increased fluxes of DMS in grazed compared to ungrazed algal cultures or in natural populations during periods of high grazing pressure [136–139], which would result in concomitant increases in acrylate production rates. Moreover, in a recent study of protozoan grazing of micro-algae , some protozoa were found to be unable to subsist on a diet of Emiliana huxleyi strains having high DMSP-lyase activity and even those species able to grow on high activity strains avoided them when offered a prey choice. Additionally, in at least some cases, the protozoa showed a positive discrimination against strains with high DMSP-lyase activity  and thus production of DMSP and DMSP-lyase may represent an adaptive anti-grazing strategy in these micro-algae .
Similarly due to the antibiotic effects of acrylate , DMSP has been proposed to also provide protection for micro-algae against bacterial attack . However, the inhibitory concentrations of acrylate required, 0.1–500 mM dependent upon the bacterial strain and the pH utilised [135,142], are very much greater than those generally found in the water column during micro-algal blooms, which range from nM to low μM concentrations [143,144]. Although, in a recent study, Noordkamp et al.  have demonstrated that acrylate can accumulate at inhibitory (millimolar) concentrations within the mucus layer which surrounds some colonies of Phaeocystis sp. Several authors have previously noted that young colonies of Phaeocystis sp. are remarkably bacteria-free and suggested that this may be associated with the production of acrylate [141,146]. Laboratory studies indicate that Phaeocystis sp. have inherently high DMS/DMSP exudation rates compared to other studied micro-algae  and possess a DMSP-lyase located on the outer face of the cell membrane, which would efficiently convert exudated DMSP to DMS and acrylate . Bacterial attack has also previously been shown to stimulate DMS fluxes and thus rates of acrylate formation in natural micro-algal blooms . However, it is difficult to conclude whether this is a specific reaction to bacterial attack or a simple side effect, due to the bacterial lysis of the cells, as similar effects have also been observed during viral infection . Additionally, the antibiotic role of acrylate is remains somewhat questionable, as isolated DMSP-lyase producing bacterial strains actually use the proposed bacteriocidal principle, acrylate as a growth substrate, but appear not to further metabolise the DMS moiety generated by cleavage of DMSP by DMSP-lyase [149,150].
2.4DMSP, DMS emissions and cloud formation as a dispersal strategy
In a recent review, Hamilton and Lenton  proposed that DMS fluxes emanating from oceanic phytoplankton blooms of DMSP accumulating micro-algae may serve a dispersal function and a similar mechanism may also exist for the zoospores and gametes of DMSP synthesising macro-algae . This trait may again be adaptive  as dispersal is considered to be an organism's third priority following survival and reproduction, even when conditions at the point of arrival are never better than those at the point of departure . The dispersal theory  hypothesises that oxidation of DMS to sulfate aerosols, the formation of cloud condensation nuclei and clouds themselves in the atmosphere above the blooms (see Section 3.4 for details of the mechanism), induces local winds due to the release of latent heat of condensation, thereby generating rising parcels of warm air (thermals), which draw air in at their bases. At wind speeds above a threshold of approximately 20 km h−1, wave crests begin to break up and the bubbles formed within these ‘white tops’ can both concentrate small hydrophobic cells and eject them into the atmosphere when they burst at the ocean surface . Once in the air, the rising thermals further energised by incipient cloud formation could lift these cells into the atmosphere leading to their dispersal by air currents, within the protective blanket of the clouds generated by their own DMS emissions .
Whilst this hypothesis must be considered as speculative due to the absence of direct experimental evidence, there is some supporting circumstantial evidence. It is well known that bubble surfaces do concentrate microbial cells, often by more than two orders of magnitude compared to the bulk water [154–156]. A proportion of these cells are ejected into the air within jet droplets, when the bubbles burst at the water surface  and could be transported to higher levels in the atmosphere. Micro-plankton cells of a wide range of algal species have been found in the air spora and their concentrations can be quantitatively significant [156–158]. Indeed, some cases of severe skin allergies and respiratory problems in humans, and livestock deaths have been associated with the presence of toxic dinoflagellates in marine aerosols blowing inland [159–161]. Further, circumstantial evidence is provided by the facts that DMS emissions increase when nutrients are locally exhausted and growth ceases, during grazing of the micro-algal blooms and following viral infection or bacterial attack [47,133,136–139,147,148,162–164], all phenomena which can be considered signals of unfavourable environmental conditions, which may indicate periods ripe for dispersal. Similar environmental cues of adverse conditions are known to initiate the production of dispersive, defensive or resting forms in a broad range of plants and animals (see  and references therein). Similarly small ultra-planktonic species, which potentially would benefit most from the proposed dispersal mechanism, are generally higher DMS emitters than larger planktonic species  and at least for dinophytes there appears to be a relationship between cell size and intracellular DMSP concentrations, with smaller species accumulating higher levels of DMSP [151,165]. Additionally, DMS fluxes in warm, oligotrophic, tropical seas are often high, although the micro-algal biomasses are generally low, indicating an increased investment per cell in DMS production  and E. huxleyi strains isolated from these waters are inherently higher DMS emitters than those isolated from temperate zones. Such sunny, windless areas may represent zones where both the seeding of convective winds would be most effective and the potential benefits of induced winds, cloud formation and dispersal to planktonic micro-algae may be greatest .
3Ecological significance of compatible solute release to the environment
Since compatible solutes are accumulated to very high, often molal intracellular concentrations, they can represent a significant proportion of total cell carbon and/or nitrogen. Therefore, particularly in hypersaline environments due to the high concentrations of solutes accumulated and in benthic microbial mat ecosystems, as a result of the large microbial biomasses present, these solutes may represent significant stocks of carbon and nitrogen within the ecosystem. Compatible solutes will be released to the environment due to passive diffusion across the cell membrane, during ‘sloppy’ grazing, upon cell death, or during cell lysis by bacteriophages, viruses or parasitic bacteria. Additionally, hypo-osmotic shock can cause an extremely rapid release of compatible solutes from cells via compatible solute specific efflux systems which exhibit properties that mimic those of ‘mechanosensitive channels’ whose opening probability is regulated by membrane tension [40,41,167–169]. However, in most environments, e.g. the oceans, seas or saline lakes, changes in osmolarity, if they occur, will be gradual and occur over periods of days, weeks or months rather than over seconds or minutes, and intracellular compatible solute concentrations would be expected to be regulated principally by changes in synthesis, degradation and transport rates, and/or dilution due to cell growth. Thus, the significance of rapid mechanosensitive efflux systems may be limited to particular niches which are subject to very rapid dilution.
Once released to the environment, compatible solutes can serve as osmoprotectants or carbon, energy and nitrogen sources for other members of the community. In addition, due to specific microbial pathways of degradation, the metabolism of DMSP and glycine betaine may act as sources of the climate regulating trace gases DMS and methane.
3.1Released compatible solutes as osmoregulators for other organisms
Many micro-organisms possess transport systems for compatible solutes whose transcription and/or activity is directly regulated by osmotic pressure [2,33,34,37–40], whilst in cells synthesising the specific compatible solute, these transport systems may function primarily to recycle compatible solutes passively diffusing from the cell [170–172]. Such transport systems could also scavenge compatible solutes released to the environment by other members of the microbial community and many bacteria possess transport systems for compatible solutes which they are unable to synthesis [2,23–25,171–174]. For example, many heterotrophic bacteria although unable to synthesise glycine betaine possess glycine betaine transport systems and in these bacteria the uptake of glycine betaine from the growth medium suppresses the synthesis of endogenous compatible solutes [2,33,34,37,40]. In most cases, these bacteria are equally incapable of utilising glycine betaine as a growth substrate and the expression and activity of the transport systems are directly regulated by osmotic stress [2,33,34,37,40]. Thus, the sole function of these transport systems would appear to be the exploitation of environmental sources of glycine betaine as a compatible solute. Additionally, in at least some bacteria, glycine betaine transport systems may also function in the accumulation of other compatible solutes. For example, in E. coli the uptake of ectoine during osmotic stress has been shown to be dependent on the proP and proU encoded glycine betaine transport systems . Thus, in this bacterium the same transport systems serve for the accumulation of glycine betaine, other betaines, proline and ectoine [2,33,34,37,40,173], and are probably also involved in the transport of the glycine betaine structural analogue DMSP , as in other bacteria glycine betaine has been shown to act as a competitive inhibitor of DMSP uptake and vice versa [174–176]. This duality of function may be a common feature of glycine betaine/DMSP transport systems as a recent study of natural seawater samples clearly demonstrated that DMSP and glycine betaine competed for the same transport systems and accumulated DMSP or glycine betaine could be rapidly chased from cells by addition of the alternative solute . Similarly the glucosylglycerol accumulating cyanobacterium Synechocystis possesses a single transport system which recognises glucosylglycerol, trehalose and sucrose as substrates, although this cyanobacterium is unable to either synthesise trehalose or utilise it as a growth substrate and sucrose plays only a minor role in osmoadaptation . However, the presence of trehalose in the growth medium suppresses glucosylglycerol synthesis and the cytoplasmic glucosylglycerol pool is metabolised and replaced by trehalose. Additionally in mutant strains unable to synthesise glucosylglycerol, additions of glucosylglycerol, trehalose and to a lesser extent sucrose to the bathing medium relieved growth inhibition at supra-optimal osmolarities . Thus, the compatible solute transport systems of bacteria exhibit a high degree of functional overlap, often recognising several diverse solutes as substrates [2,33,34,172–177] and in at least some bacteria, these transport systems can also function in the uptake of compatible solutes as growth substrates [172,175,177]. In some glycine betaine and DMSP catabolising bacterial strains, osmotic stress suppresses the metabolism of glycine betaine or DMSP, and these solutes are accumulated in the cytoplasm as compatible solutes [178,179]. Physiological studies of Rhizobium meliloti have demonstrated that this effect is due to relative changes in the activities of the enzymes involved in glycine betaine synthesis and degradation with growth medium osmolarity . During growth at low osmolarity the enzymes of the catabolic pathway dominate and glycine betaine serves as a carbon nitrogen and energy source, whereas at high osmolarity the activities of the catabolic enzymes are inhibited, whilst those of the enzymes of the synthetic pathway are stimulated or remain unchanged resulting in the accumulation of glycine betaine as a compatible solute in the cytoplasm .
In natural ecosystems, the possession of transport systems for compatible solutes may provide significant ecological advantages, due to the relatively low energy cost of compatible solute uptake, compared to de novo biosynthesis. In some cases, accumulation by transport may also provide secondary advantages, if the degree of compatibility of the accumulated solute is greater than that of the solute(s) normally synthesised. For example, Desulfovibrio halophilus, a moderately halophilic sulfate reducing bacterium isolated from Solar Lake, Sinai, where the salinity can vary between 8 and 14%, synthesises trehalose as a compatible solute, but can also accumulate glycine betaine via a weak transport activity, when it is present in the growth medium . Whilst growth rates at salinities below 9% w/v NaCl are similar in the presence or absence of glycine betaine, at salinities between 9 and 15% NaCl, growth rates in glycine betaine supplemented growth media are significantly greater than those in the absence of glycine betaine and the degree of this stimulation increases with increasing NaCl concentration, with a maximum stimulation of 250% recorded at 15% NaCl . Thus, over most of the natural salinity range of the environment from which it was isolated, D. halophilus cells exploiting environmental sources of glycine betaine would possess an inherent growth advantage over those synthesising trehalose. Indeed due to the energetic benefits of scavenging environmental compatible solutes compared to de novo synthesis, it is interesting to speculate that some organism's osmoadaptive strategies may be uniquely based on solute transport systems rather than synthetic pathways.
3.2Compatible solutes as carbon, nitrogen and energy sources in the environment
The compatible solutes accumulated by micro-organisms may represent significant stocks of carbon and nitrogen, which could be utilised by other members of the micro-flora, when they are released to the environment. This may be particularly true for marine and hypersaline laminated microbial mat ecosystems, due to the high bacterial biomasses present and high concentrations of particularly sugars (trehalose and sucrose), glucosylglycerol and glycine betaine which have been recorded in microbial mat systems [181–183]. Whilst these studies did not distinguish between intracellular and extracellular pools of compatible solutes, it is inevitable that a portion of these solutes will be released to the environment, where they may serve as carbon, energy and/or nitrogen sources to other members of the microbial community.
Carbohydrate compatible solutes, such as glycerol, glucosylglycerol, trehalose and sucrose can be exploited as carbon and energy sources by an extremely wide range of both aerobic and anaerobic heterotrophic bacteria and it has previously been proposed that these solutes may represent significant carbon sources for heterotrophs in microbial mat ecosystems [183–185]. More surprising is the fact that these compatible solutes can also serve as carbon sources for photosynthetic bacteria during photoheterotrophic growth. For example, in a recent survey of purple sulfur and nonsulfur bacteria isolated from coastal marine sediments, the utilisation of trehalose and sucrose and their constituent monosaccharides glucose and fructose was found to be a common characteristic, especially amongst purple nonsulfur bacteria, with 50 and 70% of the tested strains utilising trehalose and sucrose respectively and almost all strains metabolising glucose and fructose . Similarly the cyanobacterium Synechocystis PCC6803 has been shown to utilise sucrose present in the growth medium as a carbon source . In these organisms, the use of compatible solutes as carbon sources may represent a significant ecological advantage in their natural environments, due to the much greater efficiency of photoheterotrophic compared to photoautotrophic growth.
The tertiary sulfonium compound DMSP, which is synthesised as a compatible solute by micro-algae, macro-algae and some aquatic angiosperms, is also metabolised as a carbon source by a metabolically diverse range of heterotrophic bacteria, in both the water column and sediments [149,150,175,176]. In addition DMS an intermediate in the catabolism of DMSP can be utilised as an electron donor for photosynthesis by some anaerobic phototrophic bacteria [187,188] and the dimethyl sulfoxide (DMSO) generated during this photo-oxidation can subsequently be reduced back to DMS, by facultatively and strictly anaerobic bacteria, including sulfate reducing bacteria [189–191]. Consequently, from an ecological point of view, in benthic microbial systems, the DMS-DMSO couple could fulfil a role similar to that played by quantitatively important the sulfide-sulfate couple . Indeed, due to the diversity of aerobic and anaerobic pathways involved in DMSP turnover, the site and relative balance between the competing metabolic pathways involved in DMSP catabolism may have significant implications for the fluxes of the climatically active trace gases, DMS and methane to the atmosphere (see Sections 3.3.2 and 3.4.2).
The amino acids, proline, glutamate and alanine which are common compatible solutes in a wide variety of Gram-positive and Gram-negative bacteria, actinomycetes and micro-algae ([2,15,26,28,29] and Table 1) can be utilised as carbon, nitrogen and energy sources by a vast array of heterotrophic and phototrophic micro-organisms. In contrast, catabolism of glycine betaine, a compatible solute synthesised by many cyanobacteria and anaerobic phototrophic bacteria [22–24,43,170,171,184], is very much rarer. Under aerobic conditions, glycine betaine can be progressively demethylated to glycine with dimethylglycine and sarcosine (methylglycine) as intermediates [179,193], whereas under anaerobic conditions it is metabolised by several fermentative or reductive pathways, which yield trimethylamine as sole or dominant end-product (see  and Section 3.3.1 for details). The efficient recycling of glycine betaine may be of particular importance in cyanobacterial microbial mats, especially in hypersaline environments where glycine betaine can represent 20% of the total nitrogen in the surficial sediments , since primary productivity in these systems is normally considered to be limited by nitrogen availability, as evidenced by the high rates of nitrogen fixation associated with cyanobacterial mats . However, as discussed below, due to the specific metabolic pathways involved, glycine betaine catabolism may represent a significant source of the greenhouse gas methane in marine and hypersaline environments.
3.3Metabolism of methylated compatible solutes as a biogenic source of methane
3.3.1The role of betaines and hypersaline ecosystems
Betaines, particularly glycine betaine, are commonly synthesised as compatible solutes by cyanobacteria and phototrophic bacteria [22–24,170,171,185], indeed glycine betaine accumulation is almost a defining characteristic of halotolerant cyanobacteria and phototrophic bacteria [22–24,184,195]. Thus, it would be expected that glycine betaine would be present in significant quantities in saline environments and this may be especially true of microbial mat communities which are dominated by high biomasses of cyanobacteria and phototrophic bacteria. In a hypersaline Spirulina mat system in Westend pond, Virgin Islands, glycine betaine was present at concentrations of up to 100 μmol g dry weight sed−1 and large quantities have also been reported for the sediments of other hypersaline mats . Whilst most of this glycine betaine would be intracellular, it represents a potentially large pool for heterotrophic bacteria and glycine betaine would inevitably be released to the environment by the mechanisms discussed previously.
Glycine betaine is metabolised by a variety of metabolic pathways, most of which lead to the formation of trimethylamine as the sole or dominant end-product, with dimethylglycine and methylamine as subsidiary products (Fig. 4). Under aerobic conditions, trimethylamine can be completely oxidised to ammonia and CO2[179,193]. However, in sediments oxygen penetration is generally limited to the upper few millimetres and most degradation of organic matter proceeds via anaerobic pathways  and in hypersaline environments, aerobic mineralisation rates may be further reduced due to the lower solubility of oxygen in warm, highly saline waters. Under anaerobic conditions trimethylamine and other methylated amines are almost exclusively degraded by methanogenic bacteria [184,193,197–200], yielding CO2, ammonia and the potent greenhouse gas methane as end-products (Fig. 4). Thus, in the sediments of marine and hypersaline environments betaine catabolism may constitute a significant source of methane. Generally, in marine sediments due to the high concentration of sulfate in seawater, methanogens are out-competed for shared substrates by sulfate reducing bacteria, due to the greater energy yield of sulfate reduction compared to methanogenesis. However, the presence of large quantities of non-competitive substrates such as methylated amines (substrates which are not utilised by sulfate reducing bacteria) [198–200] may allow methanogens to maintain significant populations, which could compete with sulfate reducers for competitive substrates such as H2 and acetate, leading to further increases in methane production rates.
In marine sediments, porewater trimethylamine concentrations are highly variable, ranging between 0 and 1000 μM, with extractable and bound trimethylamine contents of between 0.5 and 400 nmol g dry weight sed−1 ( and references therein). However, it is difficult to conclude how much of this trimethylamine is directly derived from glycine betaine as several other environmental sources exist (see  for details), including trimethylamine N-oxide which serves as an osmotic solute in a wide variety of marine vertebrates and invertebrates [32,200], although the fact that the highest concentrations of bound and extractable trimethylamine have been recorded in sediments having high glycine betaine contents  indicates that this may be the major source of trimethylamine and thus of methane in anoxic marine sediments. Even less information exists concerning concentrations of glycine betaine and trimethylamine in the sediments of hypersaline environments, where their concentrations and turnover would be expected to be most significant, due to the ubiquity of glycine betaine as a compatible solute in halophilic and halotolerant eubacteria [22–24,184,195]. In a cyanobacterial mat in Westend pond (9–13% total salts), King  reported trimethylamine concentrations of 1–6 μM in the sediment porewater and in depth profiles, trimethylamine concentrations decreased exponentially with depth and paralleled those of glycine betaine. Similarly, Oremland and King  estimated the trimethylamine concentrations to be around 1 μM in the surficial sediments of Mono Lake, California (8.8% total salts).
Most studies on the contribution of trimethylamine to methanogenesis in marine environments have focussed on salt marsh sediments, due to the accumulation of glycine betaine as an osmoregulatory solute by many salt marsh grasses [9,202]. These studies have demonstrated a rapid conversion of trimethylamine to methane and CO2 and concluded that trimethylamine could account for as much as 90% of total salt marsh methanogenesis . Similarly studies of methanogenesis in intertidal sediments have also documented the importance of methylated amines. For example, in sediments from Lowes Cove, Maine, rates of methanogenesis were markedly stimulated by additions of glycine betaine, trimethylamine and choline, but not by non-methylated substrates. In these sediments, trimethylamine was estimated to be the source of approximately 60% of the total methane production [197,203], indicating that methylated amines including glycine betaine could contribute substantially to methane production rates in the marine environment.
Methane production in marine sediments, stimulated by glycine betaine and its fermentation products, may result in significant methane fluxes to the atmosphere, as rates of methane oxidation can be low [204,205]. This limitation of methane oxidation may be due to the inability of methylotrophic bacteria to adapt to high salt concentrations [183,206,207], a hypothesis which is supported by reports that methane oxidation activity is absent in at least some hypersaline environments [208,209] and addition of salt to rice paddy fields was found to cause a relatively greater inhibition of methane oxidation rates compared to methane production rates .
Relatively few studies have investigated the contribution of methylated amines to methanogenesis in hypersaline ecosystems. In hypersaline cyanobacterial mats, methane emissions were described as being ‘as high as from the classical methane generating ecosystem-rice paddies’ and the isotopically heavy signature of the methane produced (δ13C −35 to −42‰) suggested that photosynthetically derived methylamines and/or methylsulfides were the most probable source of this methane . However, the measured porewater methane concentrations were much lower than those generally found in freshwater systems and studies using specific inhibitors of methane oxidation demonstrated that the high methane fluxes were due to the almost complete absence of methane oxidation, rather than to intrinsically high rates of methanogenesis . The absence of methane oxidation was attributed to the sensitivity of methylotrophic bacteria to the elevated salinity and the relatively low affinity of methylotrophs for methane, which saturates at much lower concentrations in hypersaline compared to freshwater environments. Giani et al.  reported that rates of methanogenesis in the sediments of Solar Lake were stimulated by additions of methylamines and it has been estimated that trimethylamine accounted for approximately 33% of total methane production in Mono Lake, California . Similarly, King  reported a strong correlation between depth profiles of glycine betaine, trimethylamine and methane concentrations, and rates of methanogenesis in a hypersaline cyanobacterial mat, with maximal production rates of 67 nmol CH4 cm−3 h−1 recorded in the upper 0.5 cm sediment layer. In these sediments, additions of glycine betaine, trimethylamine or trimethylamine N-oxide significantly stimulated rates of methanogenesis and there was a high conversion efficiency of the added substrates to methane . Additionally, the ubiquity of trimethylamine utilisation amongst pure cultures of methanogens isolated from hypersaline environments  supports the hypothesis that it is an important in situ substrate for methanogens . Thus, hypersaline ecosystems may represent large, point sources of betaine derived methane fluxes to the atmosphere, particularly if the reported absence of methane oxidation activity in the limited number of studied hypersaline sites [208,209] is a more widespread characteristic of these environments.
In conclusion, whilst much more research over a greater variety of saline environments is required to assess the role of betaine derived trimethylamine to overall methane fluxes to the atmosphere, the ubiquity of betaines (mainly glycine betaine) as compatible solutes in prokaryotes, and sea and salt marsh grasses, and the vast areas occupied by marine and hypersaline waters, indicate that this contribution could be large. Additionally, the presence of non-competitive substrates such as trimethylamine in these environments may allow the development of large populations of methanogens in the presence of high sulfate concentrations, where for energetic reasons, sulfate reduction would normally dominate [184,193,197–200], thereby potentially allowing methanogenic bacteria to compete with sulfate reducers for competitive substrates such as H2 and acetate, which could further enhance rates of methane formation. The evidence that increasing salinity has a greater inhibitory effect on methane oxidation rates than on methane production rates [205,208,210] would tend to increase the proportion of the produced methane which fluxes to the atmosphere, especially in hypersaline ecosystems, where methane oxidation activity appears to be extremely low [208,209]. Thus, betaine derived trimethylamine may significantly contribute both directly and indirectly to methane emissions from marine and hypersaline environments and consequently to the total global budget of this potent greenhouse gas.
3.3.2The role of DMSP
DMSP is synthesised as a compatible solute by many micro- and macro-algae, and aquatic angiosperms [17,132,212–214]. DMSP released to the environment can be rapidly converted to DMS by DMSP-lyases produced by the plants themselves or the associated bacterial flora [134,149,150,215]. This DMS may either flux to the atmosphere, be further metabolised by bacteria [190,216] or be biologically or photochemically oxidised to dimethylsulfoxide (DMSO) [187,188,217]. Under aerobic conditions, DMS can be oxidised by chemolithotrophic bacteria such as Thiobacillus sp. to CO2 and sulfate [216,218,219] or to CO2 and thiosulfate by methylotrophs such as Methylophaga sulfidovorans. Additionally, both DMSP and DMS can be metabolised via a demethylation/demethiolation pathway [216,221–224]. This pathway was initially proposed for anoxic sediments, but has since been demonstrated in aerobic environments  and both aerobic and anaerobic demethylating bacteria have been isolated in axenic culture [175,178,226–228]. Whilst currently no methanogens have been isolated which can directly metabolise DMSP as a growth substrate, strains have been isolated which utilise DMS and certain demethylation intermediates, such as methylmercaptopropionate and methanethiol [175,226–227], and thus DMSP catabolism could contribute to methane production.
In marine sediments, a number of inhibitor studies have indicated that methanogens play a substantial role in DMSP turnover and additions of DMSP, DMS or demethylation intermediates have been shown to stimulate methane production rates [222,223,229–231]. However, it should be noted that in sediments DMSP is not the sole source of DMS, as metabolism of sulfur containing amino acids can also generate DMS [227,229]. In contrast to trimethylamine, an intermediate of glycine betaine metabolism DMS, methylmercaptopropionate and methanethiol are competitive substrates for methanogens and sulfate reducers [222,223,228]. Thus, for energetic reasons sulfate reducing bacteria would be expected to play a dominant role in their turnover in anaerobic, sulfate rich marine sediments. Although, in a study of anoxic salt marsh sediments  up to 28% of methane production was estimated to be derived from DMS, indicating that DMSP metabolism may significantly contribute to methane fluxes in at least some sediments. However, as most DMSP production is associated with oceanic blooms of micro-algae [232–234] on a global scale, the vast majority of DMSP degradation occurs in aerobic surface waters, rather than anaerobic sediments.
Studies of DMSP turnover in marine surface waters have also indicated that the demethylation/demethiolation pathway plays a significant role and estimates of 60–100% of total DMSP consumption via this pathway are not uncommon [230,235–239]. However, none of these studies have quantified the extent of methane generation associated with this process, although Scranton and Brewer  have reported the production of trace quantities of methane by aerobic phytoplankton cultures. Profiles of methane concentrations in oceanic waters generally show a sub-surface peak at 50–150 m depth, implying an in situ production of methane [242–245]. However, little information exists for marine surface waters on either rates of methane production or rates of methane oxidation, which would also be a major determinant of methane concentrations and hence fluxes to the atmosphere. In a single study in the Carico Bassin, Caribbean Sea, integrated methane production rates for the 0–150 m depth horizon were 90 nmol m−2 day−1 and methane oxidation rates were estimated to be equivalent to approximately 10% of the diffusive methane flux to surface waters from the sub-surface concentration peak at 110 m depth .
Whilst methane profiles show that biogenic methane production does occur within the photic zone in oceanic waters [240–245], it is not possible to evaluate from the existing literature the contribution of DMSP to this process. Although DMSP can account for 1–10% of total phytoplankton carbon, the resultant relatively high concentrations of methylated substrates such as DMSP, DMS and intermediate demethylation products in the environment and the substantial role of the demethylation/demethiolation pathway in the catabolism of these substrates [230,235–239,246] indicate that this contribution is potentially large. Similarly the presence of methane in these aerobic waters does not necessarily imply aerobic methanogenesis, as significant methanogenic activity can occur in anaerobic micro-niches associated with faecal and other organic particles, or within the gut systems of zooplankton [240,243,245,247–249]. Grazing experiments support the hypothesis that zooplankton mediate methane production in oxygenated seawater  and Karl and Tilbrook  proposed that degassing of pre-formed methane from faecal pellets was a major source of methane in marine waters. Marty  has demonstrated the presence of methanogens in various suspended matter particles collected from oxygenated waters of the Mediterranean and the coexistence of bacterial methanogenesis and methane oxidation in oxic seawater has been demonstrated in laboratory experiments . Similarly recent studies using PCR techniques have identified archaeal clones and a number of 16S rDNA sequences closely related to Methanococcoides methylutens in the digestive tract and faeces of fish species, and in suspended particulate matter in the North Sea [253,254]. Sinking algae, faecal and other particles (marine snow) would naturally tend to concentrate at the pycnocline and thus may contribute to the methane peaks observed in this zone [240–245].
Methane concentrations in oceanic surface waters are generally only slightly oversaturated with respect to the present day anthropogenically enriched atmosphere [240–243,245,247,249] and therefore sea-air methane fluxes are extremely low on an areal basis . However, due to the vast extent of the oceans, overall fluxes are estimated to be in the region of 1012 g year−1 or approximately 10% of the natural, global methane flux to the atmosphere [255–258]. Thus, despite the low areal fluxes, the oceans make a significant contribution to the global methane budget and as discussed above DMSP catabolism provides a potential source for much of this methane.
3.4DMSP and DMS, coupling of the global sulfur cycle and climate regulation
3.4.1Sources of DMS
DMS is an important sulfur containing trace gas in the atmosphere, whose production has global consequences [259,260]. As discussed in Section 3.3.2, DMS can be generated during the catabolism of sulfur containing amino acids [229,231,246,261]. However, by far the major environmental source of atmospheric DMS is DMSP which is accumulated as a compatible solute by micro-algae, macro-algae and some aquatic angiosperms [17,165,212–214]. Whilst salt marshes, coastal macro-algal beds and microbial mats can represent large point sources of DMS due to the high tissue DMSP contents and locally dense biomasses [212–214,219,262], 80–90% of the global DMS flux is emitted from the world's oceans [232–234,263]. Although DMSP accumulating micro-algae are the undoubted source of these ocean to atmosphere fluxes, the correlation between DMS production rates and algal biomasses (chlorophyll concentration) is generally poor . These poor correlations are probably due to at least three factors. Firstly, differences in species composition within the algal blooms, as DMSP accumulation is confined principally to the Dinophyceae (dinoflagellates), Prymnesiophyceae (including coccolithophorids) and some members of the Chrysophyceae and Bacillariophyceae (diatoms) [17,132,165], and intracellular DMSP concentrations vary widely, both between species and within species, with growth phase and environmental variables such as temperature and nutrient availability [16,47,239,264,265]. Similarly the activity of the DMSP-lyases produced can vary considerably between algal strains , and thus influence the balance between DMSP conversion to DMS and microbial consumption of DMSP. Secondly, the release of DMSP and/or DMS to the environment may be intermittent and thus highly variable during bloom initiation, growth and collapse, since DMSP/DMS exudation rates have been shown to be species dependent and to increase during cell senescence and lysis on the cessation of growth [147,162,266], during grazing by zooplankton or protozoa [133,136–139,164] or during viral or bacterial attack [137,138,148,163] and these emissions may be linked to anti-grazing and/or aerial dispersal strategies as discussed in Sections 2.3. and 2.4. Thirdly, as discussed in Section 3.4.2, the relative balance between the various microbial sinks for DMSP and DMS can significantly influence the quantity of released DMSP, which ultimately fluxes to the atmosphere as DMS. Thus, the density and species composition of the bloom associated microbial community and whether or not the micro-algae themselves produce DMSP-lyases will significantly influence DMS fluxes to the atmosphere.
3.4.2DMS fluxes to the atmosphere and the global sulfur cycle
DMS fluxes to the atmosphere arise due to the lysis of the compatible solute DMSP in the water column by DMSP-lysases produced by the micro-algae and/or the bloom associated bacterial community [134,149,150,215,266,267]. However, dissolved DMSP is also metabolised directly by bacteria via a demethylation/demethiolation pathway (see Fig. 5 and Section 3.3.2), and this may be the dominant process in many water bodies, accounting for 60–100% of total DMSP turnover [232,235–239]. The produced DMS can also be completely oxidised by chemolithotrophic bacteria of the genus Thiobacillus or methylotrophs, yielding CO2 and sulfate or thiosulfate respectively as end-products [216,218–220] and DMS consumption rates by bacteria have been estimated to be approximately 10-fold greater than the fluxes between the surface waters and the atmosphere [236,268]. Alternatively, dissolved DMSP can be biologically or photochemically oxidised to DMSO [187,188,217], although in aerobic waters photo-oxidation is thought to be the dominant process in DMSO formation . Concentrations of DMSO in the water column can be significantly greater than those of DMS and calculations of turnover times indicate that rates of photo-oxidation of DMS to DMSO are comparable with those for sea-air DMS fluxes [236,269,270]. In anaerobic zones, particularly in the sediment, the reverse reaction can occur, with DMSO being reduced to DMS by a diverse range of anaerobic and facultatively anaerobic bacteria [189–191]. However, rates of reduction of DMSO to DMS in aerobic waters are probably extremely limited and a single study has reported rates of DMSO turnover over periods of days in natural seawater samples and then only a small fraction of the consumed DMSO was recovered as DMS .
Despite the presence of these large sinks for both DMSP and DMS in marine surface waters, typical water column concentrations of DMS reported for oceanic surface waters range between 0.1 and 5 nM, with a global mean concentration of 3 nM [232,271], although higher concentrations ranging between 30 and 400 nM have been reported associated with specific micro-algal blooms [47,138,272–275]. The transfer of this DMS to the atmosphere, depends primarily upon the degree of over-saturation of the surface water with respect to the atmospheric concentration, and the transfer velocity across air-water interface (Kw) which is related directly to wind speed, the diffusivity of DMS and the water temperature [255,276].
Although sea to air fluxes are estimated to account for less than 1% of total micro-algal DMSP production [236,263], fluxes of DMS to atmosphere ranging between 1 and 30 μmol S m−2 day−1 have been reported [47,267,268,276,277] with a calculated global mean flux of approximately 10 μmol S m−2 day−1. Globally, total annual DMS fluxes to the atmosphere have been estimated to range between 0.5 and 1.2×1012 mol S year−1 or an estimated 50–60% of the total natural sulfur flux to the atmosphere, including volcanic degassing [232–234,271,277–280]. These quantities of DMS are sufficient to complete the global sulfur cycle, with DMS providing the principal transfer vehicle for the exchange of sulfur between the oceans and continents  (Fig. 6), a role which had previously been attributed to hydrogen sulfide, even though all attempts to find sufficient concentrations of H2S to complete the cycle had systematically failed . However, since the industrial revolution, the significance of oceanic DMS emissions to terrestrial sulfur budgets has decreased considerably and they may now be of only marginal significance in large areas of the industrialised world. Present-day global anthropogenic emissions of sulfur dioxide are currently estimated to be approximately 3-fold greater than total natural sulfur emission rates [233,234,281]. For example, Turner et al.  calculated that on an areal basis DMS emissions from the North Sea were trivial compared to the European land mass average for anthropogenic SO2 emissions in winter and equivalent to only 16% of the summer emissions.
3.4.3Oxidation of atmospheric DMS and cloud formation
Ocean-atmosphere fluxes of DMS have been proposed to influence climate on a global scale  via the formation of sulfate aerosols, cloud condensation nuclei and clouds, which augment the planetary albedo, resulting in a greater proportion of the incident solar radiation being reflected back into space, resulting in a cooling effect (Fig. 6). This process forms one of the climate regulating feedback loops of the Gaia hypothesis and is proposed to be involved in planetary homeostasis [140,260,283]. In the atmosphere, DMS is oxidised principally by hydroxyl and nitrate radicals, mainly to sulfur dioxide which is subsequently oxidised to sulfate [263,284], although a secondary pathway leads to the formation methane sulfonic acid (MSA), which has been used as a proxy for historical atmospheric DMS concentrations in the analysis of long-term ice core records [285,286]. Hydroxyl radicals are formed through the photo-dissociation of ozone by solar UV radiation and the subsequent reaction of the excited oxygen species with water . Due to this mechanism the concentration of hydroxyl radicals in the atmosphere is directly related to light intensity and atmospheric DMS concentrations can fluctuate considerably, with higher concentrations observed on cloudy compared to sunny days [287,288]. Over the remote oceans which are not significantly affected by anthropogenic NOx emissions, hydroxyl radicals are the dominant oxidant for DMS and atmospheric DMS concentrations can also undergo significant diurnal variations with typically early morning maxima [232,287].
Particle nucleation or deposition onto the surfaces of sea-salt aerosols of gaseous sulfuric acid generates sulfate aerosols. The nucleation of particles from gaseous sulfuric acid is a non-linear process, which is influenced by atmospheric concentrations of water and ammonia, and competition with deposition of sulfuric acid on pre-existing particle surfaces [263,289]. Growth of sulfate aerosols due to further deposition results in the formation of condensation nuclei (CN) or larger cloud condensation nuclei (CCN), which on advection into water saturated air cause cloud condensation . Within clouds, the products of DMS oxidation can also dissolve in pre-existing water droplets, affecting both their size and density, and thereby the overall cloud albedo . The oxidation of DMS in the atmosphere is thought to require more or less 1 day  and the whole process over 3 days , although rates may be significantly faster over tropical oceans, where high concentrations of water vapour and solar UV radiation combine to produce high hydroxyl radical concentrations, making these areas the most photochemically reactive regions of the earth . Additionally, faster routes for DMS oxidation have been proposed  and observed rates of oxidation during diurnal cycles indicate that daytime oxidation rates are approximately double the modelled values , indicating that more rapid mechanisms for DMS oxidation do exist.
3.4.4Feedbacks between cloud formation and nutrients
The acidic nature of cloud water, pH 4.5–5.6, which is partially due to the oxidation products of DMS , results in the acidic cloud droplets scavenging alkaline ammonia from the atmosphere and this process may be important for particle nucleation and growth [263,289], as well as enriching any subsequent rainfall with nitrogen which could provide a positive feedback mechanism, by stimulating the growth of DMS emitting phytoplankton . Similarly Aeolian dust, the major source of iron inputs to the remote ocean environment, may be trapped within acidic sulfate cloud droplets . Within the acidic environment of cloud droplets, iron will be predominantly present as the ferrous hydroxide (Fe(OH)2+) which can undergo photo-reduction to the relatively soluble ferrous ion (Fe2+), thereby increasing the bioavailability of iron, a micro-nutrient commonly considered to limit phytoplankton productivity in the remote oceans [296–298]. Again, this process may promote particle nucleation and the growth of sulfate aerosols, as the photo-reduction of Fe(OH)2+ is thought to produce hydroxyl radicals and thus would further stimulate rates of DMS oxidation . A feedback loop has been proposed between the atmospheric and oceanic iron and sulfur cycles , based on DMS emissions promoting via acid CNN formation, iron dissolution and thereby inputs of bio-available iron to the oceans, which in turn stimulate phytoplankton productivity and hence emissions of DMS, to complete the loop (Fig. 6). There is some historical evidence to support this hypothesis, as analysis of the DMS proxy MSA and non-sea salt sulfate in Antarctic ice cores has demonstrated increased atmospheric DMS concentrations during the last glaciation , and this is suggested to coincide with increased levels of primary productivity and iron inputs to the Southern Ocean .
3.4.5DMS fluxes and global climate regulation
Ocean-atmosphere fluxes of DMS and the subsequent oxidation of DMS to sulfate aerosols, CN and CNN has been proposed to have an influence global climate due to increased planetary albedo . Sub-micrometre sulfate aerosols scatter light and promote cloud formation in areas where concentrations of CNN are a limiting factor, and dissolution of DMS oxidation products in pre-existing cloud droplets affects both their density and size. The net consequence of these three processes is an increase in planetary albedo and thus an increase in the proportion of incident radiation reflected back into space, resulting in a cooling effect. Charlson et al.  proposed that the DMS emitting micro-algae could benefit directly from the resulting climate regulation, through increased cloud formation providing a filter against damaging UV radiation and increased nutrient inputs to the oceans from rainfall (Fig. 6). Thus, a feedback loop would exist between cloud formation, nutrient inputs, phytoplankton primary productivity and DMS emissions to the atmosphere. Lovelock and Kump [300,301] extended this hypothesis by proposing that the formation of a thermocline in oceanic surface waters was the dominant factor in controlling micro-algal productivity, due to the limited exchange of nutrients across the thermocline to the surface waters from the nutrient rich deep waters. Therefore, cooling of the surface waters under DMS induced clouds could directly benefit micro-algae by causing the breakdown of the thermocline and increased fluxes of nutrients to the photic zone.
Links between oceanic blooms of DMSP synthesising micro-algae, fluxes of DMS to the atmosphere and cloud formation are well established. Direct measurements made over relatively unpolluted oceanic zones have shown a strong correlation between DMS fluxes and the concentration of cloud condensation nuclei in the atmosphere [302–305]. However, in the Northern hemisphere, present-day anthropogenic emissions of sulfur dioxide dominate DMS emissions as a source of sulfate aerosols [233,280,281]. Indeed, climate models indicate that the cooling effect resulting from anthropogenic sulfur dioxide emissions may have reduced or completely negated climate forcing due to increased emissions of greenhouse gases in industrialised parts of the globe [281,306,307]. Inclusion of the effects of sulfate aerosols of anthropogenic origin significantly improves the agreement between model simulations of greenhouse warming and observed global mean and large-scale temperature patterns . Thus, these models indirectly support the proposal that DMS emissions could influence climate on a regional and global scale via changes in planetary albedo, resulting from sulfate aerosols and increased cloud formation.
Although there is now strong evidence that oceanic DMS emissions do influence global climate via changes in albedo, the hypothesis that this process is involved in Gaian planetary homeostasis remains controversial [140,151,308]. A major criticism of the hypothesised climate regulation by DMS emitting micro-algae is that no plausible Darwinian selection mechanisms exist which could explain how altruistic climate regulation could evolve by natural selection, since the benefits to the altruistic DMS emitters are estimated to be negligible compared to the costs of DMSP synthesis and thus blooms of altruistic DMS emitting micro-algae would be susceptible to invasion by non-altruists (i.e. non DMS emitting ‘cheats’), which not bearing the costs of DMS emissions would out-compete the altruists . However, the recent hypothesis of Hamilton and Lenton  that DMS emissions and cloud formation represent a dispersal mechanism for DMSP synthesising micro-algae (see Section 2.4 for details) could provide a direct benefit at the individual or clonal group level and may be adaptive due to the universal imperative of all organisms to disperse to new habitats [151,153]. Similarly recent observations that at least some planktonic grazers avoid micro-algal species which are high DMS emitters  may provide a mechanism for the maintenance of blooms with high populations of DMS emitters by limiting the growth of ‘cheats’, which may be subject to higher grazing pressures than their altruistic neighbours. Thus, the true evolutionary pressure for or against DMSP synthesis and the associated DMS emissions may be very different from those calculated by Caldeira .
A second criticism of climate regulation involving DMS emitting micro-algae is based on Antarctic ice core data for the DMS proxy MSA during the last ice age . These data show the opposite trend to that predicted by the climate regulation model of Charlson et al. , with higher MSA concentrations, indicative of higher DMS concentrations, occurring during the height of the last glaciation . In contrast, ice core data for MSA and non-sea salt sulfate concentrations in ice cores from Greenland are in accord with the climate regulation model, with decreasing concentrations of MSA observed during the glacial advance and a strong correlation observed between MSA and non-sea salt sulfate ratios with temperature, with the highest ratios corresponding to the warmer periods . However, as pointed out by the authors, several other possible mechanisms, including changes in atmospheric chemistry, could also explain the decreased concentrations of MSA during the glacial period. Similarly it has been proposed that increased inputs of iron from Aeolian dust particles transported from Patagonian deserts may have stimulated primary production and hence DMS emissions in the iron limited waters of the Southern Ocean during the last ice age [299,309]. This hypothesis is supported by sediment core data from the Southern Ocean, which show that increased fluxes of detritus to the sediment coincided with increased iron inputs during the last glaciation, indicating that levels of primary production in the surface waters were greater during this period . Such a stimulation of primary production by iron supply may have obscured any effects of temperature changes on DMS fluxes leading to the observed peaks of MSA concentrations in Antarctic ice cores during the glacial maximum . Thus, overall, no clear relationship between historical temperature records and DMS fluxes exists which is consistent with all the existing ice core data  and further data on MSA concentrations and the processes influencing its formation are required before MSA can confidently be used as a quantitative proxy for historical atmospheric DMS concentrations.
In conclusion, whilst there is now strong evidence that oceanic DMS fluxes associated with blooms of DMSP synthesising micro-algae do influence planetary albedo and cloud formation, and thereby affect climate through increased cloud albedo, further study is required to elucidate whether these fluxes do and have in the past been involved in the regulation of global climate as proposed by Charlson et al.  and how such a feedback between marine micro-algae and global climate could have arisen within the framework of natural selection.
As outlined in the preceding sections, the compatible solutes commonly accumulated as osmolytes by micro-organisms can fulfil a number of other ecological roles. These alternative functions include increased tolerance to other environmental stresses, such as heat or cold stresses and desiccation, roles as intracellular reserves of carbon, energy and nitrogen and in the case of DMSP possible roles in grazing defence, dispersal and even the regulation of the Earth's climate. Thus, the evolution of osmolyte systems in micro-organisms may have been influenced by factors other than the osmotic pressure and the degree of variability of osmotic pressure in their immediate environment, their mode of life, the relative availability of nutrients for compatible solute synthesis and the degree of compatibility of the solute over the relevant intracellular concentration range, but also by other selection pressures related to the secondary roles which can be fulfilled by specific compatible solutes and the relevance of these functions to the organism's niche. Thus, the specific osmoadaptive strategies adopted by micro-organisms may reflect an evolutionary compromise between the universal imperative to regulate their intracellular osmotic pressure with respect to that of the growth medium and the secondary benefits which can be provided by the accumulation of specific osmoregulatory solutes.
The author gratefully acknowledges financial support from the European Union and MURST (Ministry of University, Science and Technology, Italy) 40% 1997/98 (Prof. Ireneo Ferrari). Pierluigi Viaroli is thanked for comments on previous versions of the manuscript, Lucia ‘scruffy’ Maffi for her global influence and two anonymous referees for their constructive comments. This work represents a contribution to the ELOISE programme (ELOISE No. 125), carried out within the framework of the project ROBUST (Grant ENV4-CT96-0218).